v8.2.0

v8.0.1

Changelog.md

@@ -1,5 +1,111 @@
 # Versions
 
+
+
+
+
+##### 7.1.2 MQTT-Update - (2021-06-17)
+
+* NEW: 7.1.2: bug fix setting hostname, Flash-LED not off during rebootNEW: 7.1.1: bug fix wlan password with "="  (again)
+
+* MQTT error message: changes "no error", send retain flag
+
+* Update wlan handling to esp-idf 4.1
+
+* Upgrade digital CNN to v8.7.0  (added new images)
+
+* Bug fix: MQTT, WLAN, LED-Controll, GPIO usage, fixed IP, calculation flow rate
+
+  
+
+##### 7.0.1 MQTT-Update - (2021-05-13)
+
+* NEW: 7.0.1: bug fix wlan password with "=" 
+
+* Upgrade digital CNN to v8.5.0  (added new images)
+
+* New MQTT topics: flow rate (units/minute), time stamp (last correct read readout)
+
+* Update MQTT/Error topic to " " in case no error (instead of empty string)
+
+* Portrait or landscape image orientation in rotated image (avoid cropping)
+
+##### 6.7.2 Image Processing in Memory - (2021-05-01)
+
+* NEW 6.7.2: Updated html for setup modus - remove reboot on edit configuration)
+
+* NEW 6.7.1: Improved stability of camera (back to v6.6.1) - remove black strips and areas
+
+* Upgrade digital CNN to v8.3.0  (added new type of digits)
+
+* Internal update: TFlite (v2.5), esp32cam, startup sequence
+
+* Rollback to espressif v2.1.0, as v3.2.0 shows unstable reboot
+
+* Bugfix: WLan-passwords, reset of hostname
+
+
+##### 6.6.1 Image Processing in Memory - (2021-04-05)
+
+* NEW 6.6.1: failed SD card initialization indicated by fast blinking LED at startup
+* Improved SD-card handling (increase compatibility with more type of cards)
+
+##### 6.5.0 Image Processing in Memory - (2021-03-25)
+
+* Upgrade digital CNN to v8.2.0  (added new type of digits)
+* Supporting alignment structures in ROI definition
+* Bug fixing: definition of  hostname in `config.ini`
+
+##### 6.4.0 Image Processing in Memory - (2021-03-20)
+
+* Additional alignment marks for settings the ROIs (analog and digit)
+* Upgrade analog CNN to v7.0.0 (added new type of pointer)
+
+##### 6.3.1 Image Processing in Memory - (2021-03-16)
+
+* NEW: 6.3.1: bug fixing in initial edit reference image and `config.ini` (Spelling error in `InitialRotate`)
+* Initial setup mode: bug fixing, error correction
+* Bug-fixing
+
+##### 6.2.2 Image Processing in Memory - (2021-03-10)
+
+* NEW 6.2.2: bug fixing
+* NEW 6.2.1: Changed brightness and contrast to default if not enabled (resolves to bright images)
+* Determination of fixed illumination settings during startup - speed up of 5s in each run
+* Update digital CNN to v8.1.1 (additional digital images trained)
+* Extended error message in MQTT error message
+
+
+* Image brightness is now adjustable 
+
+
+* Bug fixing: minor topics 
+
+
+##### 6.1.0 Image Processing in Memory - (2021-01-20)
+
+* Disabling of analog / digital counters in configuration 
+* Improved Alignment Algorithm (`AlignmentAlgo`  = `Default`,  `Accurate` , `Fast`)
+* Analog counters: `ExtendedResolution` (last digit is extended by sub comma value of CNN)
+* `config.ini`: additional parameter `hostname`  (additional to wlan.ini)
+* Switching of GPIO12/13 via http-interface: `/GPIO?GPIO=12&Status=high/low`
+* Bug fixing: html configuration page, wlan password ("=" now possible)
+
+##### 6.0.0 Image Processing in Memory - (2021-01-02)
+
+* **Major change**: image processing fully in memory - no need of SD card buffer anymore
+
+  * Need to limit camera resolution to VGA (due to memory limits)
+* MQTT: Last Will Testament (LWT) implemented: "connection lost" in case of connection lost to `TopicError`
+* Disabled `CheckDigitIncreaseConsistency` in default configuration - must now be explicit enabled if needed
+* Update digital CNN to v7.2.1 (additional digital images trained) 
+* Setting of arbitrary time server in `config.ini`
+* Option for fixed IP-, DNS-Settings in `wlan.ini`
+* Increased stability (internal image and camera handling)
+* Bug fixing: edit digits, handling PreValue, html-bugs
+
+
+
 ##### 5.0.0 Setup Modus - (2020-12-06)
 
 * Implementation of initial setup modus for fresh installation
@@ -127,4 +233,4 @@
 
 ##### 0.1.0 (2020-08-07)
 
-* Initial Version
+* Initial Version

FeatureRequest.md

@@ -1,6 +1,6 @@
 ## Feature Requests
 
-**There are a lot of ideas for further improvements, but only limited capacity on side of the developer. **Therefore I have created this page as a collection of ideas. 
+**There are a lot of ideas for further improvements, but only limited capacity on side of the developer.** Therefore I have created this page as a collection of ideas. 
 
 1. Who ever has a new idea can put it here, so it that it is not forgotten. 
 
@@ -11,20 +11,81 @@
 
 ____
 
-#### #3 Allow grouping of digits to multiple reading values
+#### #10 Improve and bug fix logging of images
 
-* https://github.com/jomjol/AI-on-the-edge-device/issues/123
+* https://github.com/jomjol/AI-on-the-edge-device/issues/307
 
-Implementation of two different independent readouts in one setup
+  
+
+#### #9 Basic auth for the UI
+
+* https://github.com/jomjol/AI-on-the-edge-device/issues/283
+
+* Implementation of an authentication mechanism.
+
+#### #8 MQTT configurable readout intervall
+
+Make the readout intervall configurable via MQTT.
+
+* Change the mqtt part to receive and process input and not only sending
+
+#### #7 Extended Error Handling
+
+Check different types of error (e.g. tflite not availabe) and generate an error on the html page.
 
 To do:
 
-* Extend the configuration, setting and processing flow for two independend readouts
+* Make a list of "important" errors
+* Implement a checking algo
+* Extend the firmware and html page for the error handling
+
+#### ~~#6 Check for double ROI names~~ - implemented v8.0.0
+
+~~Check during configuration, that ROI names are unique.~~
+
+~~To do:~~
+
+* ~~Implementation of ROI name checking in html code before saving analog or digital ROIs~~
+
+  
+
+#### #5 Configurable decimal separator (point or comma) 
+
+Decimal separator configurable for different systems
+
+To do:
+
+* Implementation of decimal point into postprocessing module
+* Extension of configuration
+* Adaption of the html configuration to implement shifting
 
 
-#### 
 
-https://github.com/jomjol/AI-on-the-edge-device/issues/123
+#### ~~#4 Initial Shifting and Rotation~~ - implemented v7.0.0
+
+* ~~https://github.com/jomjol/AI-on-the-edge-device/issues/123~~
+
+~~Implementation of a shifting additional to the initial rotation of the raw camera input~~
+
+~~To do:~~
+
+* ~~Implementation of shifting~~
+* ~~Extension of configuration~~
+* ~~Adaption of the html configuration to implement shifting~~
+
+
+
+#### ~~#3 Allow grouping of digits to multiple reading values~~ - implemented v8.0.0
+
+* ~~https://github.com/jomjol/AI-on-the-edge-device/issues/123~~
+
+~~Implementation of two different independent readouts in one setup~~
+
+~~To do:~~
+
+* ~~Extend the configuration, setting and processing flow for two independend readouts~~
+
+
 
 
 
@@ -47,15 +108,16 @@ To do:
 
 ____
 
-#### #1 Optional GPIO for external flash/lighting
+#### ~~#1 Optional GPIO for external flash/lighting~~ - implemented (v8.0.0)
 
-* https://github.com/jomjol/AI-on-the-edge-device/issues/133
+* ~~https://github.com/jomjol/AI-on-the-edge-device/issues/133~~
 
-Implementation of an an extrnal flash / lightning through GPIOs.
-* available GPIOs: 12 & 13 (currently in use for html switching)
+~~Implementation of an an extrnal flash / lightning through GPIOs.~~
 
-To do:
+* ~~available GPIOs: 12 & 13 (currently in use for html switching)~~
 
-* Implementation of a software module for external light source (e.g. WS8132 LED controller, ...)
-* Update of the camera module to use the external light instead of the internal flash light
-* Adopt the configuration algorithm with a configurable light source
+~~To do:~~
+
+* ~~Implementation of a software module for external light source (e.g. WS8132 LED controller, ...)~~
+* ~~Update of the camera module to use the external light instead of the internal flash light~~
+* ~~Adopt the configuration algorithm with a configurable light source~~

README.md

@@ -4,12 +4,16 @@ This is an example of Artificial Intelligence (AI) calculations on a very cheap
 
 ### Details on **function**, **installation** and **configuration** can be found on the **[Wiki Page](https://github.com/jomjol/AI-on-the-edge-device/wiki)**
 
-A 3d-printable housing can be found here: https://www.thingiverse.com/thing:4571627
+A 3d-printable housing can be found here: https://www.thingiverse.com/thing:4573481
+
+respectively ESP32-Cam housing only: https://www.thingiverse.com/thing:4571627
 
 <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/watermeter_all.jpg" width="200"><img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/main.jpg" width="200"><img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/size.png" width="200"> 
 
 <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/watermeter.jpg" width="600"> 
 
+<img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/powermeter.jpg" width="600"> 
+
 
 
 
@@ -24,7 +28,9 @@ If you would like to support the developer with a cup of coffee you can do that
 <input type="image" src="https://www.paypalobjects.com/en_US/DK/i/btn/btn_donateCC_LG.gif" border="0" name="submit" title="PayPal - The safer, easier way to pay online!" alt="Donate with PayPal button" />
 <img alt="" border="0" src="https://www.paypal.com/en_DE/i/scr/pixel.gif" width="1" height="1" />
 </form>
+If you have any technical topics, you can file a issue in this repository. 
 
+In other cases you can contact the developer via email: <img src="https://raw.githubusercontent.com/jomjol/AI-on-the-edge-device/master/images/mail.jpg" height="25"> 
 
 ## Change log
 
@@ -41,55 +47,34 @@ If you would like to support the developer with a cup of coffee you can do that
 
 
 
-##### 6.4.0 Image Processing in Memory - (2021-03-20)
+##### 8.2.0 - Multi Meter Support (2021-08-24)
 
-* Additional alignment marks for settings the ROIs (analog and digit)
-* Upgrade analog CNN to v7.0.0 (added new type of pointer)
+* Improve server responsiveness
+* Flow status and prevalue status in overview
+* Improved prevalue handling 
 
-##### 6.3.1 Image Processing in Memory - (2021-03-16)
+##### 8.1.0 - Multi Meter Support (2021-08-12)
 
-* NEW: 6.3.1: bug fixing in initial edit reference image and `config.ini` (Spelling error in `InitialRotate`)
-* Initial setup mode: bug fixing, error correction
-* Bug-fixing
+* GPIO: using the general mqtt main topic for GPIO
 
-##### 6.2.2 Image Processing in Memory - (2021-03-10)
+* Upgrade digital CNN to v12.0.0  (added new images)
+* Update tfmicro to new master (2021-08-07)
+* Bug fix: remove text in mqtt value, remove connect limit in wlan reconnet
 
-* NEW 6.2.2: bug fixing
-* NEW 6.2.1: Changed brightness and contrast to default if not enabled (resolves to bright images)
-* Determination of fixed illumination settings during startup - speed up of 5s in each run
-* Update digital CNN to v8.1.1 (additional digital images trained)
-* Extended error message in MQTT error message
-
-
-* Image brightness is now adjustable 
-
-
-* Bug fixing: minor topics 
-
-
-##### 6.1.0 Image Processing in Memory - (2021-01-20)
-
-* Disabling of analog / digital counters in configuration 
-* Improved Alignment Algorithm (`AlignmentAlgo`  = `Default`,  `Accurate` , `Fast`)
-* Analog counters: `ExtendedResolution` (last digit is extended by sub comma value of CNN)
-* `config.ini`: additional parameter `hostname`  (additional to wlan.ini)
-* Switching of GPIO12/13 via http-interface: `/GPIO?GPIO=12&Status=high/low`
-* Bug fixing: html configuration page, wlan password ("=" now possible)
-
-##### 6.0.0 Image Processing in Memory - (2021-01-02)
-
-* **Major change**: image processing fully in memory - no need of SD card buffer anymore
-  
-  * Need to limit camera resolution to VGA (due to memory limits)
-* MQTT: Last Will Testament (LWT) implemented: "connection lost" in case of connection lost to `TopicError`
-* Disabled `CheckDigitIncreaseConsistency` in default configuration - must now be explicit enabled if needed
-* Update digital CNN to v7.2.1 (additional digital images trained) 
-* Setting of arbitrary time server in `config.ini`
-* Option for fixed IP-, DNS-Settings in `wlan.ini`
-* Increased stability (internal image and camera handling)
-* Bug fixing: edit digits, handling PreValue, html-bugs
+##### 8.0.5 - Multi Meter Support (2021-08-01)
 
+* NEW 8.0.5: bug fix: saving prevalue
+* NEW 8.0.4: bug fix: load config.ini after upgrade
+* NEW 8.0.3: bug fix: reboot during `config.ini` handling, html error
+* NEW 8.0.2: saving roundes prevalue, bug fix html server
+* NEW 8.0.1: bug fix: html handling of parameter `FixedExposure` and `ImageSize`
+* Dual / multi meter support (more than 1 number to be recognized)
+  This is implemented with the feature "number" on the ROI definition as well as selected options
+* MQTT: standardization of the naming - including new topics (`json`,  `freeMem `, `uptime`)c
+* Preparation for extended GPIO support (thanks to Zwerk2k) - not tested and fully functional yet
+* Bug fixing: html server, memory leak, MQTT connect, hostname, turn of flash LED
 
+<span style="color: red;">**ATTENTION: the configuration and prevalue files are modified automatically and will not be backward compatible!**</span> 
 
 
 ## Additional ideas
@@ -102,6 +87,12 @@ There are some ideas and feature request, which are not followed currently - mai
 
 ## History
 
+##### 7.1.2 MQTT-Update - (2021-06-17)
+
+**7.0.1 MQTT-Update - (2021-05-13)**
+
+##### 6.7.2 Image Processing in Memory - (2021-05-01)
+
 ##### 5.0.0 Setup Modus - (2020-12-06)
 
 ##### 4.1.1 Configuration editor - (2020-12-02)
@@ -122,8 +113,3 @@ There are some ideas and feature request, which are not followed currently - mai
 
 #### [Full Changelog](Changelog.md)
 
-
-
-## Solved topics
-
-* n.a.

code/.helper/copy.bat

@@ -1,3 +1,3 @@
-copy "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\code\.pio\build\esp32cam\firmware.bin" "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\firmware\firmware.bin"
-copy "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\code\.pio\build\esp32cam\bootloader.bin" "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\firmware\bootloader.bin"
-copy "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\code\.pio\build\esp32cam\partitions.bin" "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\firmware\partitions.bin"
+copy "..\..\code\.pio\build\esp32cam\firmware.bin" "..\..\firmware\firmware.bin"
+copy "..\..\code\.pio\build\esp32cam\bootloader.bin" "..\..\firmware\bootloader.bin"
+copy "..\..\code\.pio\build\esp32cam\partitions.bin" "..\..\firmware\partitions.bin"

code/.helper/makezip.bat

@@ -1 +1 @@
-powershell Compress-Archive "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\sd-card\html\*.*" "C:\Users\Muell\Documents\Programmieren\GitHub\AI-on-the-edge-device\firmware\html.zip"
+powershell Compress-Archive "..\..\sd-card\html\*.*" "..\..\firmware\html.zip"

code/components/connect_wlan/connect_wlan._cpp_

@@ -1,539 +0,0 @@
-#include "connect_wlan.h"
-
-#include <string.h>
-#include "freertos/FreeRTOS.h"
-#include "freertos/task.h"
-#include "freertos/event_groups.h"
-#include "esp_wifi.h"
-#include "esp_log.h"
-
-#include <fstream>
-#include <vector>
-#include <sstream>
-
-#include "Helper.h"
-
-static const char *TAG = "connect_wlan";
-
-std::string ssid = "";
-std::string passphrase = "";
-std::string hostname = "";
-std::string ipaddress = "";
-std::string gw = "";
-std::string netmask = "";
-std::string dns = "";
-
-std::string std_hostname = "watermeter";
-
-#define BLINK_GPIO GPIO_NUM_33
-
-static EventGroupHandle_t s_wifi_event_group;
-
-#define WIFI_CONNECTED_BIT BIT0
-#define WIFI_FAIL_BIT      BIT1
-static int s_retry_num = 0;
-
-
-std::vector<string> ZerlegeZeile(std::string input, std::string _delimiter = "")
-{
-	std::vector<string> Output;
-	std::string delimiter = " =,";
-    if (_delimiter.length() > 0){
-        delimiter = _delimiter;
-    }
-
-	input = trim(input, delimiter);
-	size_t pos = findDelimiterPos(input, delimiter);
-	std::string token;
-	while (pos != std::string::npos) {
-		token = input.substr(0, pos);
-		token = trim(token, delimiter);
-		Output.push_back(token);
-		input.erase(0, pos + 1);
-		input = trim(input, delimiter);
-		pos = findDelimiterPos(input, delimiter);
-	}
-	Output.push_back(input);
-
-	return Output;
-}
-
-
-void blinkstatus(int dauer, int _anzahl)
-{
-    gpio_reset_pin(BLINK_GPIO);
-    gpio_set_direction(BLINK_GPIO, GPIO_MODE_OUTPUT);
-    for (int i = 0; i < _anzahl; ++i)
-    {
-        gpio_set_level(BLINK_GPIO, 0);
-        vTaskDelay(dauer / portTICK_PERIOD_MS);
-        gpio_set_level(BLINK_GPIO, 1);
-        vTaskDelay(dauer / portTICK_PERIOD_MS);          
-    }
-}
-
-
-
-
-void strinttoip4(std::string ip, int &a, int &b, int &c, int &d) {
-    std::stringstream s(ip);
-    char ch; //to temporarily store the '.'
-    s >> a >> ch >> b >> ch >> c >> ch >> d;
-}
-
-
-
-
-static void event_handler_neu(void* arg, esp_event_base_t event_base,
-                                int32_t event_id, void* event_data)
-{
-    if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_START) {
-        blinkstatus(200, 1);
-        esp_wifi_connect();
-    } else if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_DISCONNECTED) {
-        blinkstatus(200, 5);
-        esp_wifi_connect();
-        s_retry_num++;
-        ESP_LOGI(TAG, "retry to connect to the AP");
-    } else if (event_base == IP_EVENT && event_id == IP_EVENT_STA_GOT_IP) {
-        blinkstatus(1000, 3);
-        ip_event_got_ip_t* event = (ip_event_got_ip_t*) event_data;
-        ESP_LOGI(TAG, "got ip:" IPSTR, IP2STR(&event->ip_info.ip));
-        s_retry_num = 0;
-        xEventGroupSetBits(s_wifi_event_group, WIFI_CONNECTED_BIT);
-    }
-}
-
-
-void initialise_wifi()
-{
-    s_wifi_event_group = xEventGroupCreate();
-    ESP_ERROR_CHECK(esp_netif_init());
-    ESP_ERROR_CHECK(esp_event_loop_create_default());
-    esp_netif_create_default_wifi_sta();
-
-    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
-    ESP_ERROR_CHECK(esp_wifi_init(&cfg));
-
-    esp_event_handler_instance_t instance_any_id;
-    esp_event_handler_instance_t instance_got_ip;
-    ESP_ERROR_CHECK(esp_event_handler_instance_register(WIFI_EVENT,
-                                                        ESP_EVENT_ANY_ID,
-                                                        &event_handler_neu,
-                                                        NULL,
-                                                        &instance_any_id));
-    ESP_ERROR_CHECK(esp_event_handler_instance_register(IP_EVENT,
-                                                        IP_EVENT_STA_GOT_IP,
-                                                        &event_handler_neu,
-                                                        NULL,
-                                                        &instance_got_ip));
-
-
-    wifi_config_t wifi_config = { };
-    strcpy((char*)wifi_config.sta.ssid, (const char*)ssid.c_str());
-    strcpy((char*)wifi_config.sta.password, (const char*)passphrase.c_str());
-
-    ESP_ERROR_CHECK(esp_wifi_set_mode(WIFI_MODE_STA) );
-    ESP_ERROR_CHECK(esp_wifi_set_config(ESP_IF_WIFI_STA, &wifi_config) );
-    ESP_ERROR_CHECK(esp_wifi_start() );
-
-    ESP_LOGI(TAG, "wifi_init_sta finished.");
-
-    // Waiting until either the connection is established (WIFI_CONNECTED_BIT) or connection failed for the maximum
-    // number of re-tries (WIFI_FAIL_BIT). The bits are set by event_handler() (see above) 
-    EventBits_t bits = xEventGroupWaitBits(s_wifi_event_group,
-            WIFI_CONNECTED_BIT | WIFI_FAIL_BIT,
-            pdFALSE,
-            pdFALSE,
-            portMAX_DELAY);
-
-    // xEventGroupWaitBits() returns the bits before the call returned, hence we can test which event actually
-    // happened. 
-    if (bits & WIFI_CONNECTED_BIT) {
-        ESP_LOGI(TAG, "connected to ap SSID:%s password:%s",
-                 ssid.c_str(), passphrase.c_str());
-    } else if (bits & WIFI_FAIL_BIT) {
-        ESP_LOGI(TAG, "Failed to connect to SSID:%s, password:%s",
-                 ssid.c_str(), passphrase.c_str());
-    } else {
-        ESP_LOGE(TAG, "UNEXPECTED EVENT");
-    }
-
-    // The event will not be processed after unregister 
-    ESP_ERROR_CHECK(esp_event_handler_instance_unregister(IP_EVENT, IP_EVENT_STA_GOT_IP, instance_got_ip));
-    ESP_ERROR_CHECK(esp_event_handler_instance_unregister(WIFI_EVENT, ESP_EVENT_ANY_ID, instance_any_id));
-    vEventGroupDelete(s_wifi_event_group);
-
-    tcpip_adapter_ip_info_t ip_info;
-    ESP_ERROR_CHECK(tcpip_adapter_get_ip_info(TCPIP_ADAPTER_IF_STA, &ip_info));
-    ipaddress = std::string(ip4addr_ntoa(&ip_info.ip));
-    netmask = std::string(ip4addr_ntoa(&ip_info.netmask));
-    gw = std::string(ip4addr_ntoa(&ip_info.gw));
-    printf("IPv4 :  %s\n", ip4addr_ntoa(&ip_info.ip));
-    printf("HostName :  %s\n", hostname.c_str());
-}
-
-
-void initialise_wifi_fixed_ip2()
-{
-    s_wifi_event_group = xEventGroupCreate();
-    ESP_ERROR_CHECK(esp_netif_init());
-    ESP_ERROR_CHECK(esp_event_loop_create_default());
-    esp_netif_t *my_sta = esp_netif_create_default_wifi_sta();
-
-    esp_netif_dhcpc_stop(my_sta);
-
-    esp_netif_ip_info_t ip_info;
-
-    int a, b, c, d;
-
-    strinttoip4(ipaddress, a, b, c, d);
-    IP4_ADDR(&ip_info.ip, a, b, c, d);
-
-    strinttoip4(gw, a, b, c, d);
-   	IP4_ADDR(&ip_info.gw, a, b, c, d);
-
-    strinttoip4(netmask, a, b, c, d);
-   	IP4_ADDR(&ip_info.netmask, a, b, c, d);
-
-    esp_netif_set_ip_info(my_sta, &ip_info);
-
-
-    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
-    ESP_ERROR_CHECK(esp_wifi_init(&cfg));
-
-    if (dns.length() > 0) {
-        esp_netif_dns_info_t dns_info;
-        ip4_addr_t ip;
-        ip.addr = esp_ip4addr_aton(dns.c_str());
-        ip_addr_set_ip4_u32(&dns_info.ip, ip.addr);
-        ESP_ERROR_CHECK(esp_netif_set_dns_info(my_sta, ESP_NETIF_DNS_MAIN, &dns_info));
-    }
-
-
-    esp_event_handler_instance_t instance_any_id;
-    esp_event_handler_instance_t instance_got_ip;
-    ESP_ERROR_CHECK(esp_event_handler_instance_register(WIFI_EVENT,
-                                                        ESP_EVENT_ANY_ID,
-                                                        &event_handler_neu,
-                                                        NULL,
-                                                        &instance_any_id));
-    ESP_ERROR_CHECK(esp_event_handler_instance_register(IP_EVENT,
-                                                        IP_EVENT_STA_GOT_IP,
-                                                        &event_handler_neu,
-                                                        NULL,
-                                                        &instance_got_ip));
-
-
-    wifi_config_t wifi_config = { };
-    strcpy((char*)wifi_config.sta.ssid, (const char*)ssid.c_str());
-    strcpy((char*)wifi_config.sta.password, (const char*)passphrase.c_str());
-
-    ESP_ERROR_CHECK(esp_wifi_set_mode(WIFI_MODE_STA) );
-    ESP_ERROR_CHECK(esp_wifi_set_config(ESP_IF_WIFI_STA, &wifi_config) );
-    ESP_ERROR_CHECK(esp_wifi_start() );
-
-    ESP_LOGI(TAG, "wifi_init_sta finished.");
-
-    // Waiting until either the connection is established (WIFI_CONNECTED_BIT) or connection failed for the maximum
-    // number of re-tries (WIFI_FAIL_BIT). The bits are set by event_handler() (see above) 
-    EventBits_t bits = xEventGroupWaitBits(s_wifi_event_group,
-            WIFI_CONNECTED_BIT | WIFI_FAIL_BIT,
-            pdFALSE,
-            pdFALSE,
-            portMAX_DELAY);
-
-    // xEventGroupWaitBits() returns the bits before the call returned, hence we can test which event actually
-    // happened. 
-    if (bits & WIFI_CONNECTED_BIT) {
-        ESP_LOGI(TAG, "connected to ap SSID:%s password:%s",
-                 ssid.c_str(), passphrase.c_str());
-    } else if (bits & WIFI_FAIL_BIT) {
-        ESP_LOGI(TAG, "Failed to connect to SSID:%s, password:%s",
-                 ssid.c_str(), passphrase.c_str());
-    } else {
-        ESP_LOGE(TAG, "UNEXPECTED EVENT");
-    }
-
-    // The event will not be processed after unregister 
-    ESP_ERROR_CHECK(esp_event_handler_instance_unregister(IP_EVENT, IP_EVENT_STA_GOT_IP, instance_got_ip));
-    ESP_ERROR_CHECK(esp_event_handler_instance_unregister(WIFI_EVENT, ESP_EVENT_ANY_ID, instance_any_id));
-    vEventGroupDelete(s_wifi_event_group);
-
-    tcpip_adapter_ip_info_t ip_info2;
-    ESP_ERROR_CHECK(tcpip_adapter_get_ip_info(TCPIP_ADAPTER_IF_STA, &ip_info2));
-    ipaddress = std::string(ip4addr_ntoa(&ip_info2.ip));
-    netmask = std::string(ip4addr_ntoa(&ip_info2.netmask));
-    gw = std::string(ip4addr_ntoa(&ip_info2.gw));
-}
-
-void ConnectToWLAN()
-{
-    if (ipaddress.length() == 0 || gw.length() == 0 || netmask.length() == 0)
-    {
-        printf("Connect to WLAN with dyn. IP\n");
-        initialise_wifi();    
-    }
-    else
-    {
-        printf("Connect to WLAN with fixed IP\n");
-        initialise_wifi_fixed_ip2();
-    }
-}
-
-
-bool ChangeHostName(std::string fn, std::string _newhostname)
-{
-    if (_newhostname == hostname)
-        return false;
-
-    string line = "";
-    std::vector<string> zerlegt;
-
-    bool found = false;
-
-    std::vector<string> neuesfile;
-
-    FILE* pFile;
-    fn = FormatFileName(fn);
-    pFile = OpenFileAndWait(fn.c_str(), "r");
-
-    printf("file loaded\n");
-
-    if (pFile == NULL)
-        return false;
-
-    char zw[1024];
-    fgets(zw, 1024, pFile);
-    line = std::string(zw);
-
-    while ((line.size() > 0) || !(feof(pFile)))
-    {
-        printf("%s", line.c_str());
-        zerlegt = ZerlegeZeile(line, "=");
-        zerlegt[0] = trim(zerlegt[0], " ");
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "HOSTNAME")){
-            line = "hostname = \"" + _newhostname + "\"\n";
-            found = true;
-        }
-
-        neuesfile.push_back(line);
-
-        if (fgets(zw, 1024, pFile) == NULL)
-        {
-            line = "";
-        }
-        else
-        {
-            line = std::string(zw);
-        }
-    }
-
-    if (!found)
-    {
-        line = "hostname = \"" + _newhostname + "\"\n";
-        neuesfile.push_back(line);        
-    }
-
-    fclose(pFile);
-
-    pFile = OpenFileAndWait(fn.c_str(), "w+");
-
-    for (int i = 0; i < neuesfile.size(); ++i)
-    {
-        fputs(neuesfile[i].c_str(), pFile);
-    }
-
-    fclose(pFile);
-
-    return true;
-}
-
-
-void LoadWlanFromFile(std::string fn)
-{
-    string line = "";
-    std::vector<string> zerlegt;
-    hostname = std_hostname;
-
-    FILE* pFile;
-    fn = FormatFileName(fn);
-
-    pFile = OpenFileAndWait(fn.c_str(), "r");
-    printf("file loaded\n");
-
-    if (pFile == NULL)
-        return;
-
-    char zw[1024];
-    fgets(zw, 1024, pFile);
-    line = std::string(zw);
-
-    while ((line.size() > 0) || !(feof(pFile)))
-    {
-        printf("%s", line.c_str());
-        zerlegt = ZerlegeZeile(line, "=");
-        zerlegt[0] = trim(zerlegt[0], " ");
-        for (int i = 2; i < zerlegt.size(); ++i)
-            zerlegt[i] = zerlegt[i-1] + zerlegt[i];
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "HOSTNAME")){
-            hostname = trim(zerlegt[1]);
-            if ((hostname[0] == '"') && (hostname[hostname.length()-1] == '"')){
-                hostname = hostname.substr(1, hostname.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "SSID")){
-            ssid = trim(zerlegt[1]);
-            if ((ssid[0] == '"') && (ssid[ssid.length()-1] == '"')){
-                ssid = ssid.substr(1, ssid.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "PASSWORD")){
-            passphrase = zerlegt[1];
-            if ((passphrase[0] == '"') && (passphrase[passphrase.length()-1] == '"')){
-                passphrase = passphrase.substr(1, passphrase.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "IP")){
-            ipaddress = zerlegt[1];
-            if ((ipaddress[0] == '"') && (ipaddress[ipaddress.length()-1] == '"')){
-                ipaddress = ipaddress.substr(1, ipaddress.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "GATEWAY")){
-            gw = zerlegt[1];
-            if ((gw[0] == '"') && (gw[gw.length()-1] == '"')){
-                gw = gw.substr(1, gw.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "NETMASK")){
-            netmask = zerlegt[1];
-            if ((netmask[0] == '"') && (netmask[netmask.length()-1] == '"')){
-                netmask = netmask.substr(1, netmask.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "DNS")){
-            dns = zerlegt[1];
-            if ((dns[0] == '"') && (dns[dns.length()-1] == '"')){
-                dns = dns.substr(1, dns.length()-2);
-            }
-        }
-
-
-        if (fgets(zw, 1024, pFile) == NULL)
-        {
-            line = "";
-        }
-        else
-        {
-            line = std::string(zw);
-        }
-    }
-
-    fclose(pFile);
-
-    // Check if Hostname was empty in .ini if yes set to std_hostname
-    if(hostname.length() <= 0){
-        hostname = std_hostname;
-    }
-
-    printf("\nWLan: %s, %s\n", ssid.c_str(), passphrase.c_str());
-    printf("Hostename: %s\n", hostname.c_str());
-    printf("Fixed IP: %s, Gateway %s, Netmask %s, DNS %s\n", ipaddress.c_str(), gw.c_str(), netmask.c_str(), dns.c_str());
-
-}
-
-void LoadNetConfigFromFile(std::string _fn, std::string &_ip, std::string &_gw, std::string &_netmask, std::string &_dns)
-{
-    string line = "";
-    std::vector<string> zerlegt;
-
-
-
-    FILE* pFile;
-    _fn = FormatFileName(_fn);
-    pFile = OpenFileAndWait(_fn.c_str(), "r");
-
-    if (pFile == NULL)
-        return;
-
-    char zw[1024];
-    fgets(zw, 1024, pFile);
-    line = std::string(zw);
-
-    while ((line.size() > 0) || !(feof(pFile)))
-    {
-        printf("%s", line.c_str());
-        zerlegt = ZerlegeZeile(line, "=");
-        zerlegt[0] = trim(zerlegt[0], " ");
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "IP")){
-            _ip = zerlegt[1];
-            if ((_ip[0] == '"') && (_ip[_ip.length()-1] == '"')){
-                _ip = _ip.substr(1, _ip.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "GATEWAY")){
-            _gw = zerlegt[1];
-            if ((_gw[0] == '"') && (_gw[_gw.length()-1] == '"')){
-                _gw = _gw.substr(1, _gw.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "NETMASK")){
-            _netmask = zerlegt[1];
-            if ((_netmask[0] == '"') && (_netmask[_netmask.length()-1] == '"')){
-                _netmask = _netmask.substr(1, _netmask.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "DNS")){
-            _dns = zerlegt[1];
-            if ((_dns[0] == '"') && (_dns[_dns.length()-1] == '"')){
-                _dns = _dns.substr(1, _dns.length()-2);
-            }
-        }
-
-        if (fgets(zw, 1024, pFile) == NULL)
-        {
-            line = "";
-        }
-        else
-        {
-            line = std::string(zw);
-        }
-    }
-
-    fclose(pFile);
-}
-
-
-std::string getHostname(){
-    return hostname;
-}
-
-std::string getIPAddress(){
-    return ipaddress;
-}
-
-std::string getSSID(){
-    return ssid;
-}
-
-std::string getNetMask(){
-    return netmask;
-}
-
-std::string getGW(){
-    return gw;
-}
-

code/components/connect_wlan/connect_wlan._h_

@@ -1,21 +0,0 @@
-//#ifndef CONNECT_WLAN_H
-//#define CONNECT_WLAN_H
-
-#include <string>
-#include "driver/gpio.h"
-
-const int CONNECTED_BIT = BIT0;
-
-void ConnectToWLAN();
-
-void LoadWlanFromFile(std::string fn);
-
-bool ChangeHostName(std::string fn, std::string _newhostname);
-
-std::string getHostname();
-std::string getIPAddress();
-std::string getSSID();
-std::string getNetMask();
-std::string getGW();
-
-//#endif

code/components/connect_wlan/connect_wlan.cpp

@@ -1,489 +0,0 @@
-#include "connect_wlan.h"
-
-
-
-#include <string.h>
-#include "esp_wifi.h"
-#include "esp_event_loop.h"
-#include "freertos/event_groups.h"
-#include "esp_log.h"
-
-#include <fstream>
-#include <string>
-#include <vector>
-#include <sstream>
-#include <iostream>
-
-#include <arpa/inet.h>
-
-#include "Helper.h"
-
-
-
-
-static const char *MAIN_TAG = "connect_wlan";
-
-std::string ssid = "";
-std::string passphrase = "";
-std::string hostname = "";
-std::string ipaddress = "";
-std::string gw = "";
-std::string netmask = "";
-std::string dns = "";
-std::string std_hostname = "watermeter";
-
-static EventGroupHandle_t wifi_event_group;
-
-
-#define BLINK_GPIO GPIO_NUM_33
-
-
-std::vector<string> ZerlegeZeile(std::string input, std::string _delimiter = "")
-{
-	std::vector<string> Output;
-	std::string delimiter = " =,";
-    if (_delimiter.length() > 0){
-        delimiter = _delimiter;
-    }
-
-	input = trim(input, delimiter);
-	size_t pos = findDelimiterPos(input, delimiter);
-	std::string token;
-	while (pos != std::string::npos) {
-		token = input.substr(0, pos);
-		token = trim(token, delimiter);
-		Output.push_back(token);
-		input.erase(0, pos + 1);
-		input = trim(input, delimiter);
-		pos = findDelimiterPos(input, delimiter);
-	}
-	Output.push_back(input);
-
-	return Output;
-}
-
-
-
-
-void wifi_connect(){
-    wifi_config_t cfg = { };
-    strcpy((char*)cfg.sta.ssid, (const char*)ssid.c_str());
-    strcpy((char*)cfg.sta.password, (const char*)passphrase.c_str());
-    
-    ESP_ERROR_CHECK( esp_wifi_disconnect() );
-    ESP_ERROR_CHECK( esp_wifi_set_config(ESP_IF_WIFI_STA, &cfg) );
-    ESP_ERROR_CHECK( esp_wifi_connect() );
-}
-
-
-
-void blinkstatus(int dauer, int _anzahl)
-{
-    gpio_reset_pin(BLINK_GPIO);
-    gpio_set_direction(BLINK_GPIO, GPIO_MODE_OUTPUT);
-    for (int i = 0; i < _anzahl; ++i)
-    {
-        gpio_set_level(BLINK_GPIO, 0);
-        vTaskDelay(dauer / portTICK_PERIOD_MS);
-        gpio_set_level(BLINK_GPIO, 1);
-        vTaskDelay(dauer / portTICK_PERIOD_MS);          
-    }
-}
-
-static esp_err_t event_handler(void *ctx, system_event_t *event)
-{
-    switch(event->event_id) {
-    case SYSTEM_EVENT_STA_START:
-        blinkstatus(200, 1);
-        wifi_connect();
-        break;
-    case SYSTEM_EVENT_STA_GOT_IP:
-        xEventGroupSetBits(wifi_event_group, CONNECTED_BIT);
-        blinkstatus(1000, 3);
-        break;
-    case SYSTEM_EVENT_STA_DISCONNECTED:
-        blinkstatus(200, 5);
-        esp_wifi_connect();
-        xEventGroupClearBits(wifi_event_group, CONNECTED_BIT);
-        break;
-    default:
-        break;
-    }
-    return ESP_OK;
-}
-
-void initialise_wifi()
-{
-    ESP_ERROR_CHECK(esp_event_loop_init(event_handler, NULL) );
-    wifi_event_group = xEventGroupCreate();  
-
-    esp_log_level_set("wifi", ESP_LOG_NONE); // disable wifi driver logging
-    tcpip_adapter_init();
-    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
-    ESP_ERROR_CHECK( esp_wifi_init(&cfg) );
-    ESP_ERROR_CHECK( esp_wifi_set_mode(WIFI_MODE_STA) );
-    ESP_ERROR_CHECK( esp_wifi_start() );
-    esp_err_t ret = tcpip_adapter_set_hostname(TCPIP_ADAPTER_IF_STA , hostname.c_str());
-    if(ret != ESP_OK ){
-      ESP_LOGE(MAIN_TAG,"failed to set hostname:%d",ret);  
-    }
-    xEventGroupWaitBits(wifi_event_group,CONNECTED_BIT,true,true,portMAX_DELAY);
-    tcpip_adapter_ip_info_t ip_info;
-    ESP_ERROR_CHECK(tcpip_adapter_get_ip_info(TCPIP_ADAPTER_IF_STA, &ip_info));
-    ipaddress = std::string(ip4addr_ntoa(&ip_info.ip));
-    netmask = std::string(ip4addr_ntoa(&ip_info.netmask));
-    gw = std::string(ip4addr_ntoa(&ip_info.gw));
-    printf("IPv4 :  %s\n", ip4addr_ntoa(&ip_info.ip));
-    printf("HostName :  %s\n", hostname.c_str());
-}
-
-///////////////////////////////////////////////////////////
-///////////////////////////////////////////////////////////
-
-void strinttoip4(std::string ip, int &a, int &b, int &c, int &d) {
-    std::stringstream s(ip);
-    char ch; //to temporarily store the '.'
-    s >> a >> ch >> b >> ch >> c >> ch >> d;
-}
-
-void initialise_wifi_fixed_ip()
-{
-
-    wifi_event_group = xEventGroupCreate();
-
-    ESP_ERROR_CHECK(esp_netif_init());
-    ESP_ERROR_CHECK(esp_event_loop_create_default());
-
-    esp_netif_t *my_sta = esp_netif_create_default_wifi_sta();
-
-    esp_netif_dhcpc_stop(my_sta);
-
-    esp_netif_ip_info_t ip_info;
-
-    int a, b, c, d;
-
-    strinttoip4(ipaddress, a, b, c, d);
-    IP4_ADDR(&ip_info.ip, a, b, c, d);
-
-    strinttoip4(gw, a, b, c, d);
-   	IP4_ADDR(&ip_info.gw, a, b, c, d);
-
-    strinttoip4(netmask, a, b, c, d);
-   	IP4_ADDR(&ip_info.netmask, a, b, c, d);
-
-    esp_netif_set_ip_info(my_sta, &ip_info);
-
-    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
-    ESP_ERROR_CHECK(esp_wifi_init(&cfg));
-
-    if (dns.length() > 0) {
-        esp_netif_dns_info_t dns_info;
-        ip4_addr_t ip;
-        ip.addr = esp_ip4addr_aton(dns.c_str());
-        ip_addr_set_ip4_u32(&dns_info.ip, ip.addr);
-        ESP_ERROR_CHECK(esp_netif_set_dns_info(my_sta, ESP_NETIF_DNS_MAIN, &dns_info));
-    }
-
-    ESP_ERROR_CHECK(esp_event_loop_init(event_handler, NULL) );
-
-    wifi_config_t wifi_config = { };
-    strcpy((char*)wifi_config.sta.ssid, (const char*)ssid.c_str());
-    strcpy((char*)wifi_config.sta.password, (const char*)passphrase.c_str());
-
-    ESP_ERROR_CHECK(esp_wifi_set_mode(WIFI_MODE_STA) );
-    ESP_ERROR_CHECK(esp_wifi_set_config(ESP_IF_WIFI_STA, &wifi_config) );
-    ESP_ERROR_CHECK(esp_wifi_start() );
-
-    ESP_LOGI(MAIN_TAG, "wifi_init_sta finished.");
-
-    EventBits_t bits = xEventGroupWaitBits(wifi_event_group,CONNECTED_BIT,true,true,portMAX_DELAY);
-
-    if (bits & CONNECTED_BIT) {
-        ESP_LOGI(MAIN_TAG, "connected to ap SSID:%s password:%s",
-                 ssid.c_str(), passphrase.c_str());
-    } else  {
-        ESP_LOGI(MAIN_TAG, "Failed to connect to SSID:%s, password:%s",
-                 ssid.c_str(), passphrase.c_str());
-    }
-    tcpip_adapter_ip_info_t ip_info2;
-    ESP_ERROR_CHECK(tcpip_adapter_get_ip_info(TCPIP_ADAPTER_IF_STA, &ip_info2));
-    ipaddress = std::string(ip4addr_ntoa(&ip_info2.ip));
-    netmask = std::string(ip4addr_ntoa(&ip_info2.netmask));
-    gw = std::string(ip4addr_ntoa(&ip_info2.gw));
-
-//    vEventGroupDelete(wifi_event_group);
-}
-
-
-
-void ConnectToWLAN()
-{
-    if (ipaddress.length() == 0 || gw.length() == 0 || netmask.length() == 0)
-    {
-        printf("Connect to WLAN with dyn. IP\n");
-        initialise_wifi();    
-    }
-    else
-    {
-        printf("Connect to WLAN with fixed IP\n");
-        initialise_wifi_fixed_ip();
-    }
-}
-
-
-
-bool ChangeHostName(std::string fn, std::string _newhostname)
-{
-    if (_newhostname == hostname)
-        return false;
-
-    string line = "";
-    std::vector<string> zerlegt;
-
-    bool found = false;
-
-    std::vector<string> neuesfile;
-
-    FILE* pFile;
-    fn = FormatFileName(fn);
-    pFile = OpenFileAndWait(fn.c_str(), "r");
-
-    printf("file loaded\n");
-
-    if (pFile == NULL)
-        return false;
-
-    char zw[1024];
-    fgets(zw, 1024, pFile);
-    line = std::string(zw);
-
-    while ((line.size() > 0) || !(feof(pFile)))
-    {
-        printf("%s", line.c_str());
-        zerlegt = ZerlegeZeile(line, "=");
-        zerlegt[0] = trim(zerlegt[0], " ");
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "HOSTNAME")){
-            line = "hostname = \"" + _newhostname + "\"\n";
-            found = true;
-        }
-
-        neuesfile.push_back(line);
-
-        if (fgets(zw, 1024, pFile) == NULL)
-        {
-            line = "";
-        }
-        else
-        {
-            line = std::string(zw);
-        }
-    }
-
-    if (!found)
-    {
-        line = "hostname = \"" + _newhostname + "\"\n";
-        neuesfile.push_back(line);        
-    }
-
-    fclose(pFile);
-
-    pFile = OpenFileAndWait(fn.c_str(), "w+");
-
-    for (int i = 0; i < neuesfile.size(); ++i)
-    {
-        fputs(neuesfile[i].c_str(), pFile);
-    }
-
-    fclose(pFile);
-
-    return true;
-}
-
-
-void LoadWlanFromFile(std::string fn)
-{
-    string line = "";
-    std::vector<string> zerlegt;
-    hostname = std_hostname;
-
-    FILE* pFile;
-    fn = FormatFileName(fn);
-
-    pFile = OpenFileAndWait(fn.c_str(), "r");
-    printf("file loaded\n");
-
-    if (pFile == NULL)
-        return;
-
-    char zw[1024];
-    fgets(zw, 1024, pFile);
-    line = std::string(zw);
-
-    while ((line.size() > 0) || !(feof(pFile)))
-    {
-        printf("%s", line.c_str());
-        zerlegt = ZerlegeZeile(line, "=");
-        zerlegt[0] = trim(zerlegt[0], " ");
-        for (int i = 2; i < zerlegt.size(); ++i)
-            zerlegt[i] = zerlegt[i-1] + zerlegt[i];
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "HOSTNAME")){
-            hostname = trim(zerlegt[1]);
-            if ((hostname[0] == '"') && (hostname[hostname.length()-1] == '"')){
-                hostname = hostname.substr(1, hostname.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "SSID")){
-            ssid = trim(zerlegt[1]);
-            if ((ssid[0] == '"') && (ssid[ssid.length()-1] == '"')){
-                ssid = ssid.substr(1, ssid.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "PASSWORD")){
-            passphrase = zerlegt[1];
-            if ((passphrase[0] == '"') && (passphrase[passphrase.length()-1] == '"')){
-                passphrase = passphrase.substr(1, passphrase.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "IP")){
-            ipaddress = zerlegt[1];
-            if ((ipaddress[0] == '"') && (ipaddress[ipaddress.length()-1] == '"')){
-                ipaddress = ipaddress.substr(1, ipaddress.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "GATEWAY")){
-            gw = zerlegt[1];
-            if ((gw[0] == '"') && (gw[gw.length()-1] == '"')){
-                gw = gw.substr(1, gw.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "NETMASK")){
-            netmask = zerlegt[1];
-            if ((netmask[0] == '"') && (netmask[netmask.length()-1] == '"')){
-                netmask = netmask.substr(1, netmask.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "DNS")){
-            dns = zerlegt[1];
-            if ((dns[0] == '"') && (dns[dns.length()-1] == '"')){
-                dns = dns.substr(1, dns.length()-2);
-            }
-        }
-
-
-        if (fgets(zw, 1024, pFile) == NULL)
-        {
-            line = "";
-        }
-        else
-        {
-            line = std::string(zw);
-        }
-    }
-
-    fclose(pFile);
-
-    // Check if Hostname was empty in .ini if yes set to std_hostname
-    if(hostname.length() <= 0){
-        hostname = std_hostname;
-    }
-
-    printf("\nWLan: %s, %s\n", ssid.c_str(), passphrase.c_str());
-    printf("Hostename: %s\n", hostname.c_str());
-    printf("Fixed IP: %s, Gateway %s, Netmask %s, DNS %s\n", ipaddress.c_str(), gw.c_str(), netmask.c_str(), dns.c_str());
-
-}
-
-void LoadNetConfigFromFile(std::string fn, std::string &_ip, std::string &_gw, std::string &_netmask, std::string &_dns)
-{
-    string line = "";
-    std::vector<string> zerlegt;
-
-    FILE* pFile;
-    fn = FormatFileName(fn);
-    pFile = OpenFileAndWait(fn.c_str(), "r");
-
-    printf("file loaded\n");
-
-    if (pFile == NULL)
-        return;
-
-    char zw[1024];
-    fgets(zw, 1024, pFile);
-    line = std::string(zw);
-
-    while ((line.size() > 0) || !(feof(pFile)))
-    {
-        printf("%s", line.c_str());
-        zerlegt = ZerlegeZeile(line, "=");
-        zerlegt[0] = trim(zerlegt[0], " ");
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "IP")){
-            _ip = zerlegt[1];
-            if ((_ip[0] == '"') && (_ip[_ip.length()-1] == '"')){
-                _ip = _ip.substr(1, _ip.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "GATEWAY")){
-            _gw = zerlegt[1];
-            if ((_gw[0] == '"') && (_gw[_gw.length()-1] == '"')){
-                _gw = _gw.substr(1, _gw.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "NETMASK")){
-            _netmask = zerlegt[1];
-            if ((_netmask[0] == '"') && (_netmask[_netmask.length()-1] == '"')){
-                _netmask = _netmask.substr(1, _netmask.length()-2);
-            }
-        }
-
-        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "DNS")){
-            _dns = zerlegt[1];
-            if ((_dns[0] == '"') && (_dns[_dns.length()-1] == '"')){
-                _dns = _dns.substr(1, _dns.length()-2);
-            }
-        }
-
-        if (fgets(zw, 1024, pFile) == NULL)
-        {
-            line = "";
-        }
-        else
-        {
-            line = std::string(zw);
-        }
-    }
-
-    fclose(pFile);
-}
-
-
-std::string getHostname(){
-    return hostname;
-}
-
-std::string getIPAddress(){
-    return ipaddress;
-}
-
-std::string getSSID(){
-    return ssid;
-}
-
-std::string getNetMask(){
-    return netmask;
-}
-
-std::string getGW(){
-    return gw;
-}
-

code/components/connect_wlan/connect_wlan.h

@@ -1,20 +0,0 @@
-#ifndef CONNECT_WLAN_H
-#define CONNECT_WLAN_H
-
-#include <string>
-#include "driver/gpio.h"
-
-const int CONNECTED_BIT = BIT0;
-void ConnectToWLAN();
-
-void LoadWlanFromFile(std::string fn);
-
-bool ChangeHostName(std::string fn, std::string _newhostname);
-
-std::string getHostname();
-std::string getIPAddress();
-std::string getSSID();
-std::string getNetMask();
-std::string getGW();
-
-#endif

code/components/esp32-camera-master/conversions/include/img_converters.h

@@ -62,7 +62,8 @@ bool frame2jpg_cb(camera_fb_t * fb, uint8_t quality, jpg_out_cb cb, void * arg);
  * @param height    Height in pixels of the source image
  * @param format    Format of the source image
  * @param quality   JPEG quality of the resulting image
- * @param out       Pointer to be populated with the address of the resulting buffer
+ * @param out       Pointer to be populated with the address of the resulting buffer.
+ *                  You MUST free the pointer once you are done with it.
  * @param out_len   Pointer to be populated with the length of the output buffer
  *
  * @return true on success

code/components/esp32-camera-master/conversions/to_bmp.c

@@ -317,7 +317,7 @@ bool fmt2bmp(uint8_t *src, size_t src_len, uint16_t width, uint16_t height, pixf
     }
     *out = out_buf;
     *out_len = out_size;
-    return true;
+    return true; 
 }
 
 bool frame2bmp(camera_fb_t * fb, uint8_t ** out, size_t * out_len)

code/components/esp32-camera-master/driver/camera.c

@@ -1321,7 +1321,7 @@ esp_err_t camera_init(const camera_config_t* config)
     }
 
     vsync_intr_disable();
-    err = gpio_install_isr_service(ESP_INTR_FLAG_LEVEL1 | ESP_INTR_FLAG_IRAM);
+    err = gpio_install_isr_service(ESP_INTR_FLAG_LOWMED | ESP_INTR_FLAG_IRAM);
     if (err != ESP_OK) {
     	if (err != ESP_ERR_INVALID_STATE) {
     		ESP_LOGE(TAG, "gpio_install_isr_service failed (%x)", err);

code/components/esp32-camera-master/examples/take_picture.c

@@ -29,7 +29,7 @@
 
 // ================================ CODE ======================================
 
-#include <esp_event_loop.h>
+#include <esp_event.h>
 #include <esp_log.h>
 #include <esp_system.h>
 #include <nvs_flash.h>

code/components/esp32-camera-master/idf_component.yml

@@ -1,5 +1,3 @@
-name: "esp32-camera"
-
 version: "1.0.0"
-
 description: This package hosts ESP32 compatible driver for OV2640 image sensors. Additionally it provides a few tools, which allow converting the captured frame data to the more common BMP and JPEG formats.
+url: https://github.com/espressif/esp32-camera

code/components/jomjol_configfile/CMakeLists.txt

@@ -2,6 +2,6 @@ FILE(GLOB_RECURSE app_sources ${CMAKE_CURRENT_SOURCE_DIR}/*.*)
 
 idf_component_register(SRCS ${app_sources}
                     INCLUDE_DIRS "."
-                    REQUIRES jomjol_helper)
+                    REQUIRES jomjol_logfile)
 
 

code/components/jomjol_configfile/configFile.cpp

@@ -0,0 +1,105 @@
+#include <string.h>
+#include "freertos/FreeRTOS.h"
+#include "freertos/task.h"
+
+#include "Helper.h"
+#include "configFile.h"
+
+//static const char *TAGCONFIGFILE = "configFile";
+
+ConfigFile::ConfigFile(std::string filePath)
+{
+    std::string config = FormatFileName(filePath);
+    pFile = OpenFileAndWait(config.c_str(), "r");
+}
+
+ConfigFile::~ConfigFile()
+{
+    fclose(pFile);
+}
+
+bool ConfigFile::isNewParagraph(std::string input)
+{
+	if ((input[0] == '[') || ((input[0] == ';') && (input[1] == '[')))
+	{
+		return true;
+	}
+	return false;
+}
+
+bool ConfigFile::GetNextParagraph(std::string& aktparamgraph, bool &disabled, bool &eof)
+{
+	while (getNextLine(&aktparamgraph, disabled, eof) && !isNewParagraph(aktparamgraph));
+
+	if (isNewParagraph(aktparamgraph))
+		return true;
+	return false;
+}
+
+bool ConfigFile::getNextLine(std::string *rt, bool &disabled, bool &eof)
+{
+    eof = false;
+	char zw[1024] = "";
+	if (pFile == NULL)
+	{
+		*rt = "";
+		return false;
+	}
+
+	if (fgets(zw, 1024, pFile))
+	{
+		printf("%s", zw);
+		if ((strlen(zw) == 0) && feof(pFile))
+		{
+			*rt = "";
+			eof = true;
+			return false;
+		}
+	}
+	else
+	{
+		*rt = "";
+		eof = true;
+		return false;
+	}
+	*rt = zw;
+	*rt = trim(*rt);
+	while ((zw[0] == ';' || zw[0] == '#' || (rt->size() == 0)) && !(zw[1] == '['))			// Kommentarzeilen (; oder #) und Leerzeilen überspringen, es sei denn es ist ein neuer auskommentierter Paragraph
+	{
+		fgets(zw, 1024, pFile);
+		printf("%s", zw);		
+		if (feof(pFile))
+		{
+			*rt = "";
+            eof = true;
+			return false;
+		}
+		*rt = zw;
+		*rt = trim(*rt);
+	}
+
+    disabled = ((*rt)[0] == ';');
+	return true;
+}
+
+std::vector<string> ConfigFile::ZerlegeZeile(std::string input, std::string delimiter)
+{
+	std::vector<string> Output;
+//	std::string delimiter = " =,";
+
+	input = trim(input, delimiter);
+	size_t pos = findDelimiterPos(input, delimiter);
+	std::string token;
+	while (pos != std::string::npos) {
+		token = input.substr(0, pos);
+		token = trim(token, delimiter);
+		Output.push_back(token);
+		input.erase(0, pos + 1);
+		input = trim(input, delimiter);
+		pos = findDelimiterPos(input, delimiter);
+	}
+	Output.push_back(input);
+
+	return Output;
+
+}

code/components/jomjol_configfile/configFile.h

@@ -0,0 +1,16 @@
+#include <string>
+#include <vector>
+
+class ConfigFile {
+public:
+    ConfigFile(std::string filePath);
+    ~ConfigFile();
+
+    bool isNewParagraph(std::string input);
+    bool GetNextParagraph(std::string& aktparamgraph, bool &disabled, bool &eof);
+	bool getNextLine(std::string* rt, bool &disabled, bool &eof);
+    std::vector<std::string> ZerlegeZeile(std::string input, std::string delimiter = " =, \t");
+    
+private:
+    FILE* pFile;
+};

code/components/jomjol_controlGPIO/CMakeLists.txt

@@ -3,7 +3,7 @@ FILE(GLOB_RECURSE app_sources ${CMAKE_CURRENT_SOURCE_DIR}/*.*)
 list(APPEND EXTRA_COMPONENT_DIRS $ENV{IDF_PATH}/examples/common_components/protocol_examples_common)
 
 idf_component_register(SRCS ${app_sources}
-                    INCLUDE_DIRS "."
-                    REQUIRES esp_http_server jomjol_logfile)
+                    INCLUDE_DIRS "." "../../include"
+                    REQUIRES esp_http_server jomjol_logfile jomjol_configfile jomjol_mqtt jomjol_flowcontroll)
 
 

code/components/jomjol_controlGPIO/server_GPIO.cpp

@@ -1,4 +1,5 @@
 #include <string>
+#include <functional>
 #include "string.h"
 
 #include <string.h>
@@ -6,22 +7,395 @@
 #include "freertos/task.h"
 #include "esp_system.h"
 #include "esp_event.h"
+
+#include "server_tflite.h"
+
+//#define LOG_LOCAL_LEVEL ESP_LOG_DEBUG
 #include "esp_log.h"
-#include "driver/gpio.h"
 //#include "errno.h"
 
 #include <sys/stat.h>
+#include <vector>
+//#include <regex>
+
+#include "defines.h"
 
 #include "server_GPIO.h"
 
 #include "ClassLogFile.h"
-
+#include "configFile.h"
 #include "Helper.h"
+#include "interface_mqtt.h"
 
-// #define DEBUG_DETAIL_ON 
+static const char *TAG_SERVERGPIO = "server_GPIO";
+QueueHandle_t gpio_queue_handle = NULL;
 
-esp_err_t handler_switch_GPIO(httpd_req_t *req)
+#define DEBUG_DETAIL_ON 
+
+GpioPin::GpioPin(gpio_num_t gpio, const char* name, gpio_pin_mode_t mode, gpio_int_type_t interruptType, uint8_t dutyResolution, std::string mqttTopic, bool httpEnable) 
 {
+    _gpio = gpio;
+    _name = name; 
+    _mode = mode;
+    _interruptType = interruptType;    
+    _mqttTopic = mqttTopic;
+}
+
+GpioPin::~GpioPin()
+{
+    ESP_LOGD(TAG_SERVERGPIO,"reset GPIO pin %d", _gpio);
+    if (_interruptType != GPIO_INTR_DISABLE) {
+        //hook isr handler for specific gpio pin
+        gpio_isr_handler_remove(_gpio);
+    }
+    gpio_reset_pin(_gpio);
+}
+
+static void IRAM_ATTR gpio_isr_handler(void* arg)
+{
+    GpioResult gpioResult;
+    gpioResult.gpio = *(gpio_num_t*) arg;
+    gpioResult.value = gpio_get_level(gpioResult.gpio);
+    BaseType_t ContextSwitchRequest = pdFALSE;
+ 
+    xQueueSendToBackFromISR(gpio_queue_handle,(void*)&gpioResult,&ContextSwitchRequest);
+   
+    if(ContextSwitchRequest){
+        taskYIELD();
+    }
+}
+
+static void gpioHandlerTask(void *arg) {
+    ESP_LOGD(TAG_SERVERGPIO,"start interrupt task");
+    while(1){
+        if(uxQueueMessagesWaiting(gpio_queue_handle)){
+            while(uxQueueMessagesWaiting(gpio_queue_handle)){
+                GpioResult gpioResult;
+                xQueueReceive(gpio_queue_handle,(void*)&gpioResult,10);
+                ESP_LOGD(TAG_SERVERGPIO,"gpio: %d state: %d", gpioResult.gpio, gpioResult.value);
+                ((GpioHandler*)arg)->gpioInterrupt(&gpioResult);
+            }  
+        }
+
+        ((GpioHandler*)arg)->taskHandler();
+        vTaskDelay(pdMS_TO_TICKS(1000));
+    }
+}
+
+void GpioPin::gpioInterrupt(int value) {
+    if (_mqttTopic != "") {
+        ESP_LOGD(TAG_SERVERGPIO, "gpioInterrupt %s %d", _mqttTopic.c_str(), value);
+
+        MQTTPublish(_mqttTopic, value ? "true" : "false");
+        currentState = value;
+    }
+}
+
+void GpioPin::init()
+{
+    gpio_config_t io_conf;
+    //set interrupt
+    io_conf.intr_type = _interruptType;
+    //set as output mode
+    io_conf.mode = (_mode == GPIO_PIN_MODE_OUTPUT) || (_mode == GPIO_PIN_MODE_BUILT_IN_FLASH_LED) ? gpio_mode_t::GPIO_MODE_OUTPUT : gpio_mode_t::GPIO_MODE_INPUT;
+    //bit mask of the pins that you want to set,e.g.GPIO18/19
+    io_conf.pin_bit_mask = (1ULL << _gpio);
+    //set pull-down mode
+    io_conf.pull_down_en = _mode == GPIO_PIN_MODE_INPUT_PULLDOWN ? gpio_pulldown_t::GPIO_PULLDOWN_ENABLE : gpio_pulldown_t::GPIO_PULLDOWN_DISABLE;
+    //set pull-up mode
+    io_conf.pull_up_en = _mode == GPIO_PIN_MODE_INPUT_PULLDOWN ? gpio_pullup_t::GPIO_PULLUP_ENABLE : gpio_pullup_t::GPIO_PULLUP_DISABLE;
+    //configure GPIO with the given settings
+    gpio_config(&io_conf);
+
+    if (_interruptType != GPIO_INTR_DISABLE) {
+        //hook isr handler for specific gpio pin
+        ESP_LOGD(TAG_SERVERGPIO, "GpioPin::init add isr handler for GPIO %d\r\n", _gpio);
+        gpio_isr_handler_add(_gpio, gpio_isr_handler, (void*)&_gpio);
+    }
+
+    if ((_mqttTopic != "") && ((_mode == GPIO_PIN_MODE_OUTPUT) || (_mode == GPIO_PIN_MODE_OUTPUT_PWM) || (_mode == GPIO_PIN_MODE_BUILT_IN_FLASH_LED))) {
+        std::function<bool(std::string, char*, int)> f = std::bind(&GpioPin::handleMQTT, this, std::placeholders::_1, std::placeholders::_2, std::placeholders::_3);
+        MQTTregisterSubscribeFunction(_mqttTopic, f);
+    }
+}
+
+bool GpioPin::getValue(std::string* errorText)
+{   
+    if ((_mode != GPIO_PIN_MODE_INPUT) && (_mode != GPIO_PIN_MODE_INPUT_PULLUP) && (_mode != GPIO_PIN_MODE_INPUT_PULLDOWN)) {
+        (*errorText) = "GPIO is not in input mode";
+    }
+
+    return gpio_get_level(_gpio) == 1;
+}
+
+void GpioPin::setValue(bool value, gpio_set_source setSource, std::string* errorText)
+{
+    ESP_LOGD(TAG_SERVERGPIO, "GpioPin::setValue %d\r\n", value);
+
+    if ((_mode != GPIO_PIN_MODE_OUTPUT) && (_mode != GPIO_PIN_MODE_OUTPUT_PWM) && (_mode != GPIO_PIN_MODE_BUILT_IN_FLASH_LED)) {
+        (*errorText) = "GPIO is not in output mode";
+    } else {
+        gpio_set_level(_gpio, value);
+
+        if ((_mqttTopic != "") && (setSource != GPIO_SET_SOURCE_MQTT)) {
+            MQTTPublish(_mqttTopic, value ? "true" : "false");
+        }
+    }
+}
+
+void GpioPin::publishState() {
+    int newState = gpio_get_level(_gpio);
+    if (newState != currentState) {
+        ESP_LOGD(TAG_SERVERGPIO,"publish state of GPIO %d new state %d", _gpio, newState);
+        MQTTPublish(_mqttTopic, newState ? "true" : "false");
+        currentState = newState;
+    }
+}
+
+bool GpioPin::handleMQTT(std::string, char* data, int data_len) {
+    ESP_LOGD(TAG_SERVERGPIO, "GpioPin::handleMQTT data %.*s\r\n", data_len, data);
+
+    std::string dataStr(data, data_len);
+    dataStr = toLower(dataStr);
+    std::string errorText = "";
+    if ((dataStr == "true") || (dataStr == "1")) {
+        setValue(true, GPIO_SET_SOURCE_MQTT, &errorText);
+    } else if ((dataStr == "false") || (dataStr == "0")) {
+        setValue(false, GPIO_SET_SOURCE_MQTT, &errorText);    
+    } else {
+        errorText = "wrong value ";
+        errorText.append(data, data_len);
+    }
+
+    if (errorText != "") {
+        ESP_LOGE(TAG_SERVERGPIO, "%s", errorText.c_str());
+    }
+
+    return (errorText == "");
+}
+
+
+esp_err_t callHandleHttpRequest(httpd_req_t *req)
+{
+    ESP_LOGD(TAG_SERVERGPIO,"callHandleHttpRequest");
+
+    GpioHandler *gpioHandler = (GpioHandler*)req->user_ctx;
+    return gpioHandler->handleHttpRequest(req);
+}
+
+void taskGpioHandler(void *pvParameter)
+{
+    ESP_LOGD(TAG_SERVERGPIO,"taskGpioHandler");
+    ((GpioHandler*)pvParameter)->init();
+}
+
+GpioHandler::GpioHandler(std::string configFile, httpd_handle_t httpServer) 
+{
+    ESP_LOGI(TAG_SERVERGPIO,"start GpioHandler");
+    _configFile = configFile;
+    _httpServer = httpServer;
+
+    ESP_LOGI(TAG_SERVERGPIO, "register GPIO Uri");
+    registerGpioUri();
+}
+
+GpioHandler::~GpioHandler()  {
+    if (gpioMap != NULL) {
+        clear();
+        delete gpioMap;
+    }
+}
+
+void GpioHandler::init()
+{
+    // TickType_t xDelay = 60000 / portTICK_PERIOD_MS;
+    // printf("wait before start %ldms\r\n", (long) xDelay);
+    // vTaskDelay( xDelay );
+
+    if (gpioMap == NULL) {
+        gpioMap = new std::map<gpio_num_t, GpioPin*>();
+    } else {
+        clear();
+    }
+    
+    
+    ESP_LOGI(TAG_SERVERGPIO, "read GPIO config and init GPIO");
+    if (!readConfig()) {
+        clear();
+        delete gpioMap;
+        gpioMap = NULL;
+        ESP_LOGI(TAG_SERVERGPIO, "GPIO init comleted, handler is disabled");
+        return;
+    }
+
+
+    for(std::map<gpio_num_t, GpioPin*>::iterator it = gpioMap->begin(); it != gpioMap->end(); ++it) {
+        it->second->init();
+    }
+
+    std::function<void()> f = std::bind(&GpioHandler::handleMQTTconnect, this);
+    MQTTregisterConnectFunction("gpio-handler", f);
+
+    if (xHandleTaskGpio == NULL) {
+        gpio_queue_handle = xQueueCreate(10,sizeof(GpioResult));
+        BaseType_t  xReturned = xTaskCreate(&gpioHandlerTask, "gpio_int", configMINIMAL_STACK_SIZE * 8, (void *)this, tskIDLE_PRIORITY + 2, &xHandleTaskGpio);
+        if(xReturned == pdPASS ) {
+            ESP_LOGD(TAG_SERVERGPIO, "xHandletaskGpioHandler started");
+        } else {
+            ESP_LOGD(TAG_SERVERGPIO, "xHandletaskGpioHandler not started %d ", (int)xHandleTaskGpio);
+        }
+    }
+
+    ESP_LOGI(TAG_SERVERGPIO, "GPIO init comleted, is enabled");
+}
+
+void GpioHandler::taskHandler() {
+    if (gpioMap != NULL) {
+        for(std::map<gpio_num_t, GpioPin*>::iterator it = gpioMap->begin(); it != gpioMap->end(); ++it) {
+            if ((it->second->getInterruptType() == GPIO_INTR_DISABLE))
+                it->second->publishState();
+        }
+    }
+}
+
+
+void GpioHandler::handleMQTTconnect()
+{
+    if (gpioMap != NULL) {
+        for(std::map<gpio_num_t, GpioPin*>::iterator it = gpioMap->begin(); it != gpioMap->end(); ++it) {
+            if ((it->second->getMode() == GPIO_PIN_MODE_INPUT) || (it->second->getMode() == GPIO_PIN_MODE_INPUT_PULLDOWN) || (it->second->getMode() == GPIO_PIN_MODE_INPUT_PULLUP))
+                it->second->publishState();
+        }
+    }
+}
+
+void GpioHandler::deinit() {
+    MQTTunregisterConnectFunction("gpio-handler");
+    clear();
+    if (xHandleTaskGpio != NULL) {
+        vTaskDelete(xHandleTaskGpio);
+        xHandleTaskGpio = NULL;
+    }
+}
+
+void GpioHandler::gpioInterrupt(GpioResult* gpioResult) {
+    if ((gpioMap != NULL) && (gpioMap->find(gpioResult->gpio) != gpioMap->end())) {
+        (*gpioMap)[gpioResult->gpio]->gpioInterrupt(gpioResult->value);
+    }
+}
+
+bool GpioHandler::readConfig() 
+{
+    if (!gpioMap->empty())
+        clear();
+
+    ConfigFile configFile = ConfigFile(_configFile); 
+
+    std::vector<std::string> zerlegt;
+    std::string line = "";
+    bool disabledLine = false;
+    bool eof = false;
+        
+    while ((!configFile.GetNextParagraph(line, disabledLine, eof) || (line.compare("[GPIO]") != 0)) && !disabledLine && !eof) {}
+    if (eof)
+        return false;
+    
+    _isEnabled = !disabledLine;
+
+    if (!_isEnabled)
+        return false;
+
+//    std::string mainTopicMQTT = "";
+    std::string mainTopicMQTT = GetMQTTMainTopic();
+    if (mainTopicMQTT.length() > 0)
+    {
+        mainTopicMQTT = mainTopicMQTT + "/GPIO";
+        ESP_LOGD(TAG_SERVERGPIO, "MAINTOPICMQTT found\r\n");
+    }
+
+    bool registerISR = false;
+    while (configFile.getNextLine(&line, disabledLine, eof) && !configFile.isNewParagraph(line))
+    {
+        zerlegt = configFile.ZerlegeZeile(line);
+        // const std::regex pieces_regex("IO([0-9]{1,2})");
+        // std::smatch pieces_match;
+        // if (std::regex_match(zerlegt[0], pieces_match, pieces_regex) && (pieces_match.size() == 2))
+        // {
+        //     std::string gpioStr = pieces_match[1];
+        ESP_LOGD(TAG_SERVERGPIO, "conf param %s\r\n", toUpper(zerlegt[0]).c_str());
+        if (toUpper(zerlegt[0]) == "MAINTOPICMQTT") {
+//            ESP_LOGD(TAG_SERVERGPIO, "MAINTOPICMQTT found\r\n");
+//            mainTopicMQTT = zerlegt[1];
+        } else if ((zerlegt[0].rfind("IO", 0) == 0) && (zerlegt.size() >= 6))
+        {
+            ESP_LOGI(TAG_SERVERGPIO,"Enable GP%s in %s mode", zerlegt[0].c_str(), zerlegt[1].c_str());
+            std::string gpioStr = zerlegt[0].substr(2, 2);
+            gpio_num_t gpioNr = (gpio_num_t)atoi(gpioStr.c_str());
+            gpio_pin_mode_t pinMode = resolvePinMode(toLower(zerlegt[1]));
+            gpio_int_type_t intType = resolveIntType(toLower(zerlegt[2]));
+            uint16_t dutyResolution = (uint8_t)atoi(zerlegt[3].c_str());
+            bool mqttEnabled = toLower(zerlegt[4]) == "true";
+            bool httpEnabled = toLower(zerlegt[5]) == "true";
+            char gpioName[100];
+            if (zerlegt.size() >= 7) {
+                strcpy(gpioName, trim(zerlegt[6]).c_str());
+            } else {
+                sprintf(gpioName, "GPIO%d", gpioNr);
+            }
+            std::string mqttTopic = mqttEnabled ? (mainTopicMQTT + "/" + gpioName) : "";
+            GpioPin* gpioPin = new GpioPin(gpioNr, gpioName, pinMode, intType,dutyResolution, mqttTopic, httpEnabled);
+            (*gpioMap)[gpioNr] = gpioPin;
+
+            if (intType != GPIO_INTR_DISABLE) {
+                registerISR = true;
+            }
+        }
+    }
+
+    if (registerISR) {
+        //install gpio isr service
+        gpio_install_isr_service(ESP_INTR_FLAG_LOWMED | ESP_INTR_FLAG_IRAM);
+    }
+
+    return true;
+}
+
+void GpioHandler::clear() 
+{
+    ESP_LOGD(TAG_SERVERGPIO, "GpioHandler::clear\r\n");
+
+    if (gpioMap != NULL) {
+        for(std::map<gpio_num_t, GpioPin*>::iterator it = gpioMap->begin(); it != gpioMap->end(); ++it) {
+            delete it->second;
+        }
+        gpioMap->clear();
+    }
+
+    // gpio_uninstall_isr_service(); can't uninstall, isr service is used by camera
+}
+ 
+void GpioHandler::registerGpioUri() 
+{
+    ESP_LOGI(TAG_SERVERGPIO, "server_GPIO - Registering URI handlers");
+    
+    httpd_uri_t camuri = { };
+    camuri.method    = HTTP_GET;
+    camuri.uri       = "/GPIO";
+    camuri.handler   = callHandleHttpRequest;
+    camuri.user_ctx  = (void*)this;    
+    httpd_register_uri_handler(_httpServer, &camuri);
+}
+
+esp_err_t GpioHandler::handleHttpRequest(httpd_req_t *req)
+{
+    ESP_LOGD(TAG_SERVERGPIO, "handleHttpRequest");
+
+    if (gpioMap == NULL) {
+        std::string resp_str = "GPIO handler not initialized";
+        httpd_resp_send(req, resp_str.c_str(), resp_str.length());    
+        return ESP_OK;
+    }
+
 #ifdef DEBUG_DETAIL_ON 
     LogFile.WriteHeapInfo("handler_switch_GPIO - Start");    
 #endif
@@ -30,95 +404,183 @@ esp_err_t handler_switch_GPIO(httpd_req_t *req)
     char _query[200];
     char _valueGPIO[30];    
     char _valueStatus[30];    
-    std::string gpio, status, zw;
-    int gpionum = 0;
-    gpio_num_t gpio_num;
+    std::string gpio, status;
 
-    if (httpd_req_get_url_query_str(req, _query, 200) == ESP_OK)
-    {
-        printf("Query: "); printf(_query); printf("\n");
+    if (httpd_req_get_url_query_str(req, _query, 200) == ESP_OK) {
+        ESP_LOGD(TAG_SERVERGPIO, "Query: %s", _query);
         
         if (httpd_query_key_value(_query, "GPIO", _valueGPIO, 30) == ESP_OK)
         {
-            printf("GPIO is found"); printf(_valueGPIO); printf("\n"); 
+            ESP_LOGD(TAG_SERVERGPIO, "GPIO is found %s", _valueGPIO); 
             gpio = std::string(_valueGPIO);
+        } else {
+            std::string resp_str = "GPIO No is not defined";
+            httpd_resp_send(req, resp_str.c_str(), resp_str.length());    
+            return ESP_OK;
         }
         if (httpd_query_key_value(_query, "Status", _valueStatus, 30) == ESP_OK)
         {
-            printf("Status is found"); printf(_valueStatus); printf("\n"); 
+            ESP_LOGD(TAG_SERVERGPIO, "Status is found %s", _valueStatus); 
             status = std::string(_valueStatus);
         }
-    };
+    } else {
+        const char* resp_str = "Error in call. Use /GPIO?GPIO=12&Status=high";
+        httpd_resp_send(req, resp_str, strlen(resp_str));    
+        return ESP_OK;
+    }
 
     status = toUpper(status);
-    if (!(status == "HIGH") && !(status == "LOW"))
+    if ((status != "HIGH") && (status != "LOW") && (status != "TRUE") && (status != "FALSE") && (status != "0") && (status != "1") && (status != ""))
     {
-        zw = "Status not valid: " + status;;
+        std::string zw = "Status not valid: " + status;
         httpd_resp_sendstr_chunk(req, zw.c_str());
         httpd_resp_sendstr_chunk(req, NULL);          
         return ESP_OK;    
     }
 
-    gpionum = stoi(gpio);
-    
-    // frei: 16; 12-15; 2; 4  // nur 12 und 13 funktionieren 2: reboot, 4: BlitzLED, 14/15: DMA für SDKarte ???
+    int gpionum = stoi(gpio);
 
-    switch (gpionum) {
-        case 12:
-            gpio_num = GPIO_NUM_12;
-            break;
-        case 13:
-            gpio_num = GPIO_NUM_13;
-            break;
-        default:
-            zw = "GPIO" + std::to_string(gpionum) + " not support - only 12 & 13 free";
+    // frei: 16; 12-15; 2; 4  // nur 12 und 13 funktionieren 2: reboot, 4: BlitzLED, 15: PSRAM, 14/15: DMA für SDKarte ???
+    gpio_num_t gpio_num = resolvePinNr(gpionum);
+    if (gpio_num == GPIO_NUM_NC)
+    {
+        std::string zw = "GPIO" + std::to_string(gpionum) + " not support - only 12 & 13 free";
             httpd_resp_sendstr_chunk(req, zw.c_str());
             httpd_resp_sendstr_chunk(req, NULL);          
-            return ESP_OK;    
+            return ESP_OK;
     }
 
-    if (status == "HIGH")  
-        gpio_set_level(gpio_num, 1);
+    if (gpioMap->count(gpio_num) == 0) {
+        char resp_str [30];
+        sprintf(resp_str, "GPIO%d is not registred", gpio_num);
+        httpd_resp_send(req, resp_str, strlen(resp_str));  
+        return ESP_OK;     
+    }
+    
+    if (status == "") 
+    {
+        std::string resp_str = "";
+        status = (*gpioMap)[gpio_num]->getValue(&resp_str) ? "HIGH" : "LOW";
+        if (resp_str == "") {
+            resp_str = status;
+        }
+        httpd_resp_sendstr_chunk(req, resp_str.c_str());
+        httpd_resp_sendstr_chunk(req, NULL);
+    }
     else
-        gpio_set_level(gpio_num, 0); 
-
-
-    zw = "GPIO" + std::to_string(gpionum) + " switched to " + status;
-    httpd_resp_sendstr_chunk(req, zw.c_str());
-    httpd_resp_sendstr_chunk(req, NULL);          
+    {
+        std::string resp_str = "";
+        (*gpioMap)[gpio_num]->setValue((status == "HIGH") || (status == "TRUE") || (status == "1"), GPIO_SET_SOURCE_HTTP, &resp_str);
+        if (resp_str == "") {
+            resp_str = "GPIO" + std::to_string(gpionum) + " switched to " + status;
+        }
+        httpd_resp_sendstr_chunk(req, resp_str.c_str());
+        httpd_resp_sendstr_chunk(req, NULL);
+    }
+          
     return ESP_OK;    
 };
 
-void initGPIO()
+void GpioHandler::flashLightEnable(bool value) 
 {
-    gpio_config_t io_conf;
-    //disable interrupt
-    io_conf.intr_type = GPIO_INTR_DISABLE;
-    //set as output mode
-    io_conf.mode = GPIO_MODE_OUTPUT;
-    //bit mask of the pins that you want to set,e.g.GPIO18/19
-//    io_conf.pin_bit_mask = ((1ULL<<GPIO_OUTPUT_IO_0) | (1ULL<<GPIO_OUTPUT_IO_1));
-//    io_conf.pin_bit_mask = ((1ULL << GPIO_NUM_12) | (1ULL << GPIO_NUM_2) | (1ULL << GPIO_NUM_4) | (1ULL << GPIO_NUM_12) | (1ULL << GPIO_NUM_13) | (1ULL << GPIO_NUM_14) | (1ULL << GPIO_NUM_15));
-    io_conf.pin_bit_mask = ((1ULL << GPIO_NUM_12) | (1ULL << GPIO_NUM_13));
-    //disable pull-down mode
-    io_conf.pull_down_en = (gpio_pulldown_t) 0;
-    //disable pull-up mode
-    io_conf.pull_up_en = (gpio_pullup_t) 0;
-    //configure GPIO with the given settings
-    gpio_config(&io_conf);
+    ESP_LOGD(TAG_SERVERGPIO, "GpioHandler::flashLightEnable %s\r\n", value ? "true" : "false");
+
+    if (gpioMap != NULL) {
+        for(std::map<gpio_num_t, GpioPin*>::iterator it = gpioMap->begin(); it != gpioMap->end(); ++it) 
+        {
+            if (it->second->getMode() == GPIO_PIN_MODE_BUILT_IN_FLASH_LED) //|| (it->second->getMode() == GPIO_PIN_MODE_EXTERNAL_FLASH_PWM) || (it->second->getMode() == GPIO_PIN_MODE_EXTERNAL_FLASH_WS281X))
+            {
+                std::string resp_str = "";
+                it->second->setValue(value, GPIO_SET_SOURCE_INTERNAL, &resp_str);
+
+                if (resp_str == "") {
+                    ESP_LOGD(TAG_SERVERGPIO, "Flash light pin GPIO %d switched to %s\r\n", (int)it->first, (value ? "on" : "off"));
+                } else {
+                    ESP_LOGE(TAG_SERVERGPIO, "Can't set flash light pin GPIO %d.  Error: %s\r\n", (int)it->first, resp_str.c_str());
+                }
+            }
+        }
+    }
 }
 
-
-void register_server_GPIO_uri(httpd_handle_t server)
+gpio_num_t GpioHandler::resolvePinNr(uint8_t pinNr) 
 {
-    ESP_LOGI(TAGPARTGPIO, "server_GPIO - Registering URI handlers");
-    
-    httpd_uri_t camuri = { };
-    camuri.method    = HTTP_GET;
-    camuri.uri       = "/GPIO";
-    camuri.handler   = handler_switch_GPIO;
-    camuri.user_ctx  = (void*) "switch GPIO";    
-    httpd_register_uri_handler(server, &camuri);
-
-    initGPIO();
+    switch(pinNr)  {
+        case 0:
+            return GPIO_NUM_0;
+        case 1:
+            return GPIO_NUM_1;
+        case 3:
+            return GPIO_NUM_3;
+        case 4:
+            return GPIO_NUM_4;
+        case 12:
+            return GPIO_NUM_12;
+        case 13:
+            return GPIO_NUM_13;
+        default: 
+            return GPIO_NUM_NC;   
+    }
+}
+
+gpio_pin_mode_t GpioHandler::resolvePinMode(std::string input) 
+{
+    if( input == "disabled" ) return GPIO_PIN_MODE_DISABLED;
+    if( input == "input" ) return GPIO_PIN_MODE_INPUT;
+    if( input == "input-pullup" ) return GPIO_PIN_MODE_INPUT_PULLUP;
+    if( input == "input-pulldown" ) return GPIO_PIN_MODE_INPUT_PULLDOWN;
+    if( input == "output" ) return GPIO_PIN_MODE_OUTPUT;
+    if( input == "built-in-led" ) return GPIO_PIN_MODE_BUILT_IN_FLASH_LED;
+    if( input == "output-pwm" ) return GPIO_PIN_MODE_OUTPUT_PWM;
+    if( input == "external-flash-pwm" ) return GPIO_PIN_MODE_EXTERNAL_FLASH_PWM;
+    if( input == "external-flash-ws281x" ) return GPIO_PIN_MODE_EXTERNAL_FLASH_WS281X;
+
+    return GPIO_PIN_MODE_DISABLED;
+}
+
+gpio_int_type_t GpioHandler::resolveIntType(std::string input) 
+{
+    if( input == "disabled" ) return GPIO_INTR_DISABLE;
+    if( input == "rising-edge" ) return GPIO_INTR_POSEDGE;
+    if( input == "falling-edge" ) return GPIO_INTR_NEGEDGE;
+    if( input == "rising-and-falling" ) return GPIO_INTR_ANYEDGE ;
+    if( input == "low-level-trigger" ) return GPIO_INTR_LOW_LEVEL;
+    if( input == "high-level-trigger" ) return GPIO_INTR_HIGH_LEVEL;
+
+
+    return GPIO_INTR_DISABLE;
+}
+
+static GpioHandler *gpioHandler = NULL;
+
+void gpio_handler_create(httpd_handle_t server) 
+{
+    if (gpioHandler == NULL)
+        gpioHandler = new GpioHandler(CONFIG_FILE, server);
+}
+
+void gpio_handler_init() 
+{
+    if (gpioHandler != NULL) {
+        gpioHandler->init();
+    }
+}
+
+void gpio_handler_deinit() {
+    if (gpioHandler != NULL) {
+        gpioHandler->deinit();
+   }
+}
+
+void gpio_handler_destroy()
+{
+    if (gpioHandler != NULL) {
+        delete gpioHandler;
+        gpioHandler = NULL;
+    }
+}
+
+GpioHandler* gpio_handler_get()
+{
+    return gpioHandler;
 }

code/components/jomjol_controlGPIO/server_GPIO.h

@@ -1,10 +1,99 @@
+#ifndef SERVER_GPIO_H
+#define SERVER_GPIO_H
+
 #include <esp_log.h>
 
 #include <esp_http_server.h>
+#include <map>
+#include "driver/gpio.h"
 
 //#include "ClassControllCamera.h"
 
-static const char *TAGPARTGPIO = "server_GPIO";
+typedef enum {
+    GPIO_PIN_MODE_DISABLED              = 0x0,
+    GPIO_PIN_MODE_INPUT                 = 0x1,
+    GPIO_PIN_MODE_INPUT_PULLUP          = 0x2,
+    GPIO_PIN_MODE_INPUT_PULLDOWN        = 0x3,
+    GPIO_PIN_MODE_OUTPUT                = 0x4,
+    GPIO_PIN_MODE_BUILT_IN_FLASH_LED    = 0x5,
+    GPIO_PIN_MODE_OUTPUT_PWM            = 0x6,
+    GPIO_PIN_MODE_EXTERNAL_FLASH_PWM    = 0x7,
+    GPIO_PIN_MODE_EXTERNAL_FLASH_WS281X = 0x8,
+} gpio_pin_mode_t;
 
-void register_server_GPIO_uri(httpd_handle_t server);
+struct GpioResult {
+    gpio_num_t gpio;
+    int value;
+};
 
+typedef enum {
+    GPIO_SET_SOURCE_INTERNAL  = 0,
+    GPIO_SET_SOURCE_MQTT  = 1,
+    GPIO_SET_SOURCE_HTTP  = 2,
+} gpio_set_source;
+
+class GpioPin {
+public:
+    GpioPin(gpio_num_t gpio, const char* name, gpio_pin_mode_t mode, gpio_int_type_t interruptType, uint8_t dutyResolution, std::string mqttTopic, bool httpEnable);
+    ~GpioPin();
+
+    void init();
+    bool getValue(std::string* errorText);
+    void setValue(bool value, gpio_set_source setSource, std::string* errorText);
+    bool handleMQTT(std::string, char* data, int data_len);
+    void publishState();
+    void gpioInterrupt(int value);
+    gpio_int_type_t getInterruptType() { return _interruptType; }
+    gpio_pin_mode_t getMode() { return _mode; }
+
+private:
+    gpio_num_t _gpio;
+    const char* _name;
+    gpio_pin_mode_t _mode;
+    gpio_int_type_t _interruptType;
+    std::string _mqttTopic;
+    int currentState = -1;
+};
+
+esp_err_t callHandleHttpRequest(httpd_req_t *req);
+void taskGpioHandler(void *pvParameter);
+
+class GpioHandler {
+public:
+    GpioHandler(std::string configFile, httpd_handle_t httpServer);
+    ~GpioHandler();
+    
+    void init();
+    void deinit();
+    void registerGpioUri();
+    esp_err_t handleHttpRequest(httpd_req_t *req);
+    void taskHandler();
+    void gpioInterrupt(GpioResult* gpioResult);  
+    void flashLightEnable(bool value);
+    bool isEnabled() { return _isEnabled; }
+    void handleMQTTconnect();
+
+private:
+    std::string _configFile;
+    httpd_handle_t _httpServer;
+    std::map<gpio_num_t, GpioPin*> *gpioMap = NULL;
+    TaskHandle_t xHandleTaskGpio = NULL;
+    bool _isEnabled = false;
+
+    bool readConfig();
+    void clear();
+    
+    gpio_num_t resolvePinNr(uint8_t pinNr);
+    gpio_pin_mode_t resolvePinMode(std::string input);
+    gpio_int_type_t resolveIntType(std::string input);
+};
+
+void gpio_handler_create(httpd_handle_t server);
+void gpio_handler_init();
+void gpio_handler_deinit();
+void gpio_handler_destroy();
+GpioHandler* gpio_handler_get();
+
+
+
+#endif //SERVER_GPIO_H

code/components/jomjol_controlcamera/CMakeLists.txt

@@ -4,6 +4,6 @@ list(APPEND EXTRA_COMPONENT_DIRS $ENV{IDF_PATH}/examples/common_components/proto
 
 idf_component_register(SRCS ${app_sources}
                     INCLUDE_DIRS "."
-                    REQUIRES esp32-camera-master esp_http_server jomjol_logfile jomjol_image_proc nvs_flash)
+                    REQUIRES esp32-camera-master esp_http_server jomjol_logfile jomjol_image_proc nvs_flash jomjol_fileserver_ota jomjol_controlGPIO)
 
 

code/components/jomjol_controlcamera/ClassControllCamera.cpp

@@ -9,11 +9,14 @@
 #include "Helper.h"
 #include "CImageBasis.h"
 
+#include "server_ota.h"
+#include "server_GPIO.h"
+
 
 #define BOARD_ESP32CAM_AITHINKER
 
 
-#include <esp_event_loop.h>
+#include <esp_event.h>
 #include <esp_log.h>
 #include <esp_system.h>
 #include <nvs_flash.h>
@@ -48,7 +51,7 @@
 #define CAM_PIN_HREF 23
 #define CAM_PIN_PCLK 22
 
-static const char *TAG = "example:take_picture";
+static const char *TAGCAMERACLASS = "server_part_camera"; 
 
 static camera_config_t camera_config = {
     .pin_pwdn = CAM_PIN_PWDN,
@@ -220,13 +223,17 @@ void CCamera::SetQualitySize(int qual, framesize_t resol)
 void CCamera::EnableAutoExposure(int flashdauer)
 {
     LEDOnOff(true);
-    LightOnOff(true);
+    if (flashdauer > 0)
+        LightOnOff(true);
     const TickType_t xDelay = flashdauer / portTICK_PERIOD_MS;
     vTaskDelay( xDelay );
 
     camera_fb_t * fb = esp_camera_fb_get();
     if (!fb) {
         ESP_LOGE(TAGCAMERACLASS, "Camera Capture Failed");
+        LEDOnOff(false);
+        LightOnOff(false);
+        doReboot();
     }
     esp_camera_fb_return(fb);        
 
@@ -269,8 +276,11 @@ esp_err_t CCamera::CaptureToBasisImage(CImageBasis *_Image, int delay)
 
     camera_fb_t * fb = esp_camera_fb_get();
     if (!fb) {
-        ESP_LOGE(TAGCAMERACLASS, "Camera Capture Failed");
+        ESP_LOGE(TAGCAMERACLASS, "CaptureToBasisImage: Camera Capture Failed");
         LEDOnOff(false);
+        LightOnOff(false);
+        doReboot();
+
         return ESP_FAIL;
     }
 
@@ -353,8 +363,11 @@ esp_err_t CCamera::CaptureToFile(std::string nm, int delay)
 
     camera_fb_t * fb = esp_camera_fb_get();
     if (!fb) {
-        ESP_LOGE(TAGCAMERACLASS, "Camera Capture Failed");
+        ESP_LOGE(TAGCAMERACLASS, "CaptureToFile: Camera Capture Failed");
         LEDOnOff(false);
+        LightOnOff(false);
+        doReboot();
+
         return ESP_FAIL;
     }
     LEDOnOff(false);    
@@ -443,7 +456,11 @@ esp_err_t CCamera::CaptureToHTTP(httpd_req_t *req, int delay)
     fb = esp_camera_fb_get();
     if (!fb) {
         ESP_LOGE(TAGCAMERACLASS, "Camera capture failed");
+        LEDOnOff(false);
+        LightOnOff(false);
         httpd_resp_send_500(req);
+//        doReboot();
+
         return ESP_FAIL;
     }
 
@@ -480,15 +497,20 @@ esp_err_t CCamera::CaptureToHTTP(httpd_req_t *req, int delay)
 
 void CCamera::LightOnOff(bool status)
 {
-	// Init the GPIO
-    gpio_pad_select_gpio(FLASH_GPIO);
-    /* Set the GPIO as a push/pull output */
-    gpio_set_direction(FLASH_GPIO, GPIO_MODE_OUTPUT);  
+    GpioHandler* gpioHandler = gpio_handler_get();
+    if ((gpioHandler != NULL) && (gpioHandler->isEnabled())) {
+        gpioHandler->flashLightEnable(status);
+    }  else {
+        // Init the GPIO
+        gpio_pad_select_gpio(FLASH_GPIO);
+        /* Set the GPIO as a push/pull output */
+        gpio_set_direction(FLASH_GPIO, GPIO_MODE_OUTPUT);  
 
-    if (status)  
-        gpio_set_level(FLASH_GPIO, 1);
-    else
-        gpio_set_level(FLASH_GPIO, 0);      
+        if (status)  
+            gpio_set_level(FLASH_GPIO, 1);
+        else
+            gpio_set_level(FLASH_GPIO, 0);
+    }
 }
 
 void CCamera::LEDOnOff(bool status)

code/components/jomjol_controlcamera/ClassControllCamera.h

@@ -16,9 +16,6 @@
 #define CAMERA_MODEL_AI_THINKER
 
 
-static const char *TAGCAMERACLASS = "server_part_camera"; 
-
-
 class CCamera {
     protected:
         int ActualQuality;

code/components/jomjol_controlcamera/server_camera.cpp

@@ -12,14 +12,17 @@
 char scratch2[SCRATCH_BUFSIZE2];
 
 //#define DEBUG_DETAIL_ON   
-
+static const char *TAGPARTCAMERA = "server_camera";
 
 
 void PowerResetCamera(){
         ESP_LOGD(TAGPARTCAMERA, "Resetting camera by power down line");
-        gpio_config_t conf = { 0 };
+        gpio_config_t conf;
+        conf.intr_type = GPIO_INTR_DISABLE;
         conf.pin_bit_mask = 1LL << GPIO_NUM_32;
         conf.mode = GPIO_MODE_OUTPUT;
+        conf.pull_down_en = GPIO_PULLDOWN_DISABLE;
+        conf.pull_up_en = GPIO_PULLUP_DISABLE;
         gpio_config(&conf);
 
         // carefull, logic is inverted compared to reset pin

code/components/jomjol_controlcamera/server_camera.h

@@ -7,8 +7,6 @@
 
 //#include "ClassControllCamera.h"
 
-static const char *TAGPARTCAMERA = "server_camera";
-
 void register_server_camera_uri(httpd_handle_t server);
 
 void PowerResetCamera();

code/components/jomjol_fileserver_ota/CMakeLists.txt

@@ -1,7 +1,7 @@
 FILE(GLOB_RECURSE app_sources ${CMAKE_CURRENT_SOURCE_DIR}/*.*)
 
 idf_component_register(SRCS ${app_sources}
-                    INCLUDE_DIRS "."
-                    REQUIRES tfmicro esp_http_server app_update esp_http_client nvs_flash jomjol_tfliteclass jomjol_flowcontroll spiffs jomjol_helper)
+                    INCLUDE_DIRS "." "../../include"
+                    REQUIRES tfmicro esp_http_server app_update esp_http_client nvs_flash jomjol_tfliteclass jomjol_flowcontroll spiffs jomjol_helper jomjol_controlGPIO)
 
 

code/components/jomjol_fileserver_ota/server_file.cpp

@@ -26,9 +26,12 @@
 #include <esp_spiffs.h>
 #include "esp_http_server.h"
 
+#include "defines.h"
 #include "ClassLogFile.h"
 
 #include "server_help.h"
+#include "interface_mqtt.h"
+#include "server_GPIO.h"
 
 #include "Helper.h"
 #include "miniz.h"
@@ -53,17 +56,17 @@ struct file_server_data {
     char scratch[SCRATCH_BUFSIZE];
 };
 
-static const char *TAG = "file_server";
+static const char *TAG_FILESERVER = "file_server";
 
 /* Handler to redirect incoming GET request for /index.html to /
  * This can be overridden by uploading file with same name */
-static esp_err_t index_html_get_handler(httpd_req_t *req)
-{
-    httpd_resp_set_status(req, "307 Temporary Redirect");
-    httpd_resp_set_hdr(req, "Location", "/");
-    httpd_resp_send(req, NULL, 0);  // Response body can be empty
-    return ESP_OK;
-}
+// static esp_err_t index_html_get_handler(httpd_req_t *req)
+// {
+//     httpd_resp_set_status(req, "307 Temporary Redirect");
+//     httpd_resp_set_hdr(req, "Location", "/");
+//     httpd_resp_send(req, NULL, 0);  // Response body can be empty
+//     return ESP_OK;
+// }
 
 /* Send HTTP response with a run-time generated html consisting of
  * a list of all files and folders under the requested path.
@@ -95,7 +98,7 @@ static esp_err_t http_resp_dir_html(httpd_req_t *req, const char *dirpath, const
     printf("entrypath: <%s>\n", entrypath);    
 
     if (!dir) {
-        ESP_LOGE(TAG, "Failed to stat dir : %s", dirpath);
+        ESP_LOGE(TAG_FILESERVER, "Failed to stat dir : %s", dirpath);
         /* Respond with 404 Not Found */
         httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, "Directory does not exist");
         return ESP_FAIL;
@@ -115,7 +118,7 @@ static esp_err_t http_resp_dir_html(httpd_req_t *req, const char *dirpath, const
             if (chunksize > 0){
                 if (httpd_resp_send_chunk(req, chunk, chunksize) != ESP_OK) {
                 fclose(fd);
-                ESP_LOGE(TAG, "File sending failed!");
+                ESP_LOGE(TAG_FILESERVER, "File sending failed!");
                 return ESP_FAIL;
                 }
             }
@@ -154,11 +157,11 @@ static esp_err_t http_resp_dir_html(httpd_req_t *req, const char *dirpath, const
             strlcpy(entrypath + dirpath_len, entry->d_name, sizeof(entrypath) - dirpath_len);
             printf("Entrypath: %s\n", entrypath);
             if (stat(entrypath, &entry_stat) == -1) {
-                ESP_LOGE(TAG, "Failed to stat %s : %s", entrytype, entry->d_name);
+                ESP_LOGE(TAG_FILESERVER, "Failed to stat %s : %s", entrytype, entry->d_name);
                 continue;
             }
             sprintf(entrysize, "%ld", entry_stat.st_size);
-            ESP_LOGI(TAG, "Found %s : %s (%s bytes)", entrytype, entry->d_name, entrysize);
+            ESP_LOGI(TAG_FILESERVER, "Found %s : %s (%s bytes)", entrytype, entry->d_name, entrysize);
 
             /* Send chunk of HTML file containing table entries with file name and size */
             httpd_resp_sendstr_chunk(req, "<tr><td><a href=\"");
@@ -206,19 +209,19 @@ static esp_err_t logfileact_get_handler(httpd_req_t *req)
     LogFile.WriteToFile("logfileact_get_handler");
     char filepath[FILE_PATH_MAX];
     FILE *fd = NULL;
-    struct stat file_stat;
+    //struct stat file_stat;
     printf("uri: %s\n", req->uri);
 
-    const char filename = 'log_current.txt'; 
+    const char* filename = "log_current.txt"; 
 
-    printf("uri: %s, filename: %s, filepath: %s\n", req->uri, &filename, filepath);
+    printf("uri: %s, filename: %s, filepath: %s\n", req->uri, filename, filepath);
 
     std::string currentfilename = LogFile.GetCurrentFileName();
 
 
     fd = OpenFileAndWait(currentfilename.c_str(), "r");
     if (!fd) {
-        ESP_LOGE(TAG, "Failed to read existing file : %s", filepath);
+        ESP_LOGE(TAG_FILESERVER, "Failed to read existing file : %s", filepath);
         /* Respond with 500 Internal Server Error */
         httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to read existing file");
         return ESP_FAIL;
@@ -226,8 +229,8 @@ static esp_err_t logfileact_get_handler(httpd_req_t *req)
 
     httpd_resp_set_hdr(req, "Access-Control-Allow-Origin", "*");
 
-//    ESP_LOGI(TAG, "Sending file : %s (%ld bytes)...", &filename, file_stat.st_size);
-    set_content_type_from_file(req, &filename);
+//    ESP_LOGI(TAG_FILESERVER, "Sending file : %s (%ld bytes)...", &filename, file_stat.st_size);
+    set_content_type_from_file(req, filename);
 
     /* Retrieve the pointer to scratch buffer for temporary storage */
     char *chunk = ((struct file_server_data *)req->user_ctx)->scratch;
@@ -239,7 +242,7 @@ static esp_err_t logfileact_get_handler(httpd_req_t *req)
         /* Send the buffer contents as HTTP response chunk */
         if (httpd_resp_send_chunk(req, chunk, chunksize) != ESP_OK) {
             fclose(fd);
-            ESP_LOGE(TAG, "File sending failed!");
+            ESP_LOGE(TAG_FILESERVER, "File sending failed!");
             /* Abort sending file */
             httpd_resp_sendstr_chunk(req, NULL);
             /* Respond with 500 Internal Server Error */
@@ -252,7 +255,7 @@ static esp_err_t logfileact_get_handler(httpd_req_t *req)
 
     /* Close file after sending complete */
     fclose(fd);
-    ESP_LOGI(TAG, "File sending complete");
+    ESP_LOGI(TAG_FILESERVER, "File sending complete");
 
     /* Respond with an empty chunk to signal HTTP response completion */
     httpd_resp_send_chunk(req, NULL, 0);
@@ -284,7 +287,7 @@ static esp_err_t download_get_handler(httpd_req_t *req)
 
 
     if (!filename) {
-        ESP_LOGE(TAG, "Filename is too long");
+        ESP_LOGE(TAG_FILESERVER, "Filename is too long");
         /* Respond with 500 Internal Server Error */
         httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Filename too long");
         return ESP_FAIL;
@@ -297,11 +300,11 @@ static esp_err_t download_get_handler(httpd_req_t *req)
         if (buf_len > 1) {
             char buf[buf_len];
             if (httpd_req_get_url_query_str(req, buf, buf_len) == ESP_OK) {
-                ESP_LOGI(TAG, "Found URL query => %s", buf);
+                ESP_LOGI(TAG_FILESERVER, "Found URL query => %s", buf);
                 char param[32];
                 /* Get value of expected key from query string */
                 if (httpd_query_key_value(buf, "readonly", param, sizeof(param)) == ESP_OK) {
-                    ESP_LOGI(TAG, "Found URL query parameter => readonly=%s", param);
+                    ESP_LOGI(TAG_FILESERVER, "Found URL query parameter => readonly=%s", param);
                     readonly = param && strcmp(param,"true")==0;
                 }
             }
@@ -316,7 +319,7 @@ static esp_err_t download_get_handler(httpd_req_t *req)
 
         /* If file not present on SPIFFS check if URI
          * corresponds to one of the hardcoded paths */
-        ESP_LOGE(TAG, "Failed to stat file : %s", filepath);
+        ESP_LOGE(TAG_FILESERVER, "Failed to stat file : %s", filepath);
         /* Respond with 404 Not Found */
         httpd_resp_send_err(req, HTTPD_404_NOT_FOUND, "File does not exist");
         return ESP_FAIL;
@@ -324,7 +327,7 @@ static esp_err_t download_get_handler(httpd_req_t *req)
 
     fd = OpenFileAndWait(filepath, "r");
     if (!fd) {
-        ESP_LOGE(TAG, "Failed to read existing file : %s", filepath);
+        ESP_LOGE(TAG_FILESERVER, "Failed to read existing file : %s", filepath);
         /* Respond with 500 Internal Server Error */
         httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to read existing file");
         return ESP_FAIL;
@@ -332,7 +335,7 @@ static esp_err_t download_get_handler(httpd_req_t *req)
 
     httpd_resp_set_hdr(req, "Access-Control-Allow-Origin", "*");
 
-    ESP_LOGI(TAG, "Sending file : %s (%ld bytes)...", filename, file_stat.st_size);
+    ESP_LOGI(TAG_FILESERVER, "Sending file : %s (%ld bytes)...", filename, file_stat.st_size);
     set_content_type_from_file(req, filename);
 
     /* Retrieve the pointer to scratch buffer for temporary storage */
@@ -345,7 +348,7 @@ static esp_err_t download_get_handler(httpd_req_t *req)
         /* Send the buffer contents as HTTP response chunk */
         if (httpd_resp_send_chunk(req, chunk, chunksize) != ESP_OK) {
             fclose(fd);
-            ESP_LOGE(TAG, "File sending failed!");
+            ESP_LOGE(TAG_FILESERVER, "File sending failed!");
             /* Abort sending file */
             httpd_resp_sendstr_chunk(req, NULL);
             /* Respond with 500 Internal Server Error */
@@ -358,7 +361,7 @@ static esp_err_t download_get_handler(httpd_req_t *req)
 
     /* Close file after sending complete */
     fclose(fd);
-    ESP_LOGI(TAG, "File sending complete");
+    ESP_LOGI(TAG_FILESERVER, "File sending complete");
 
     /* Respond with an empty chunk to signal HTTP response completion */
     httpd_resp_send_chunk(req, NULL, 0);
@@ -385,13 +388,13 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
 
     /* Filename cannot have a trailing '/' */
     if (filename[strlen(filename) - 1] == '/') {
-        ESP_LOGE(TAG, "Invalid filename : %s", filename);
+        ESP_LOGE(TAG_FILESERVER, "Invalid filename : %s", filename);
         httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Invalid filename");
         return ESP_FAIL;
     }
 
     if (stat(filepath, &file_stat) == 0) {
-        ESP_LOGE(TAG, "File already exists : %s", filepath);
+        ESP_LOGE(TAG_FILESERVER, "File already exists : %s", filepath);
         /* Respond with 400 Bad Request */
         httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST, "File already exists");
         return ESP_FAIL;
@@ -399,7 +402,7 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
 
     /* File cannot be larger than a limit */
     if (req->content_len > MAX_FILE_SIZE) {
-        ESP_LOGE(TAG, "File too large : %d bytes", req->content_len);
+        ESP_LOGE(TAG_FILESERVER, "File too large : %d bytes", req->content_len);
         /* Respond with 400 Bad Request */
         httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST,
                             "File size must be less than "
@@ -411,13 +414,13 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
 
     fd = OpenFileAndWait(filepath, "w");
     if (!fd) {
-        ESP_LOGE(TAG, "Failed to create file : %s", filepath);
+        ESP_LOGE(TAG_FILESERVER, "Failed to create file : %s", filepath);
         /* Respond with 500 Internal Server Error */
         httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to create file");
         return ESP_FAIL;
     }
 
-    ESP_LOGI(TAG, "Receiving file : %s...", filename);
+    ESP_LOGI(TAG_FILESERVER, "Receiving file : %s...", filename);
 
     /* Retrieve the pointer to scratch buffer for temporary storage */
     char *buf = ((struct file_server_data *)req->user_ctx)->scratch;
@@ -429,7 +432,7 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
 
     while (remaining > 0) {
 
-        ESP_LOGI(TAG, "Remaining size : %d", remaining);
+        ESP_LOGI(TAG_FILESERVER, "Remaining size : %d", remaining);
         /* Receive the file part by part into a buffer */
         if ((received = httpd_req_recv(req, buf, MIN(remaining, SCRATCH_BUFSIZE))) <= 0) {
             if (received == HTTPD_SOCK_ERR_TIMEOUT) {
@@ -442,7 +445,7 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
             fclose(fd);
             unlink(filepath);
 
-            ESP_LOGE(TAG, "File reception failed!");
+            ESP_LOGE(TAG_FILESERVER, "File reception failed!");
             /* Respond with 500 Internal Server Error */
             httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to receive file");
             return ESP_FAIL;
@@ -455,7 +458,7 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
             fclose(fd);
             unlink(filepath);
 
-            ESP_LOGE(TAG, "File write failed!");
+            ESP_LOGE(TAG_FILESERVER, "File write failed!");
             /* Respond with 500 Internal Server Error */
             httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Failed to write file to storage");
             return ESP_FAIL;
@@ -468,7 +471,7 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
 
     /* Close file upon upload completion */
     fclose(fd);
-    ESP_LOGI(TAG, "File reception complete");
+    ESP_LOGI(TAG_FILESERVER, "File reception complete");
 
     std::string directory = std::string(filepath);
 	size_t zw = directory.find("/");
@@ -483,10 +486,10 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
     int start_fn = strlen(((struct file_server_data *)req->user_ctx)->base_path);
     printf("Directory: %s, start_fn: %d, found: %d\n", directory.c_str(), start_fn, found);
 	directory = directory.substr(start_fn, found - start_fn + 1);
-    printf("Directory danach: %s\n", directory.c_str());    
+    printf("Directory danach 1: %s\n", directory.c_str());    
 
     directory = "/fileserver" + directory;
-    printf("Directory danach: %s\n", directory.c_str());   
+    printf("Directory danach 2: %s\n", directory.c_str());   
 
     /* Redirect onto root to see the updated file list */
     httpd_resp_set_status(req, "303 See Other");
@@ -496,6 +499,15 @@ static esp_err_t upload_post_handler(httpd_req_t *req)
     httpd_resp_set_status(req, "303 See Other");
     httpd_resp_set_hdr(req, "Location", directory.c_str());
     httpd_resp_sendstr(req, "File uploaded successfully");
+
+/*
+    if (strcmp(filepath, CONFIG_FILE) == 0) {
+        printf("New config found. Reload handler.");
+        gpio_handler_deinit();
+        MQTTdestroy();
+    }
+*/
+
     return ESP_OK;
 }
 
@@ -567,19 +579,19 @@ static esp_err_t delete_post_handler(httpd_req_t *req)
 
         /* Filename cannot have a trailing '/' */
         if (filename[strlen(filename) - 1] == '/') {
-            ESP_LOGE(TAG, "Invalid filename : %s", filename);
+            ESP_LOGE(TAG_FILESERVER, "Invalid filename : %s", filename);
             httpd_resp_send_err(req, HTTPD_500_INTERNAL_SERVER_ERROR, "Invalid filename");
             return ESP_FAIL;
         }
 
         if (stat(filepath, &file_stat) == -1) {
-            ESP_LOGE(TAG, "File does not exist : %s", filename);
+            ESP_LOGE(TAG_FILESERVER, "File does not exist : %s", filename);
             /* Respond with 400 Bad Request */
             httpd_resp_send_err(req, HTTPD_400_BAD_REQUEST, "File does not exist");
             return ESP_FAIL;
         }
 
-        ESP_LOGI(TAG, "Deleting file : %s", filename);
+        ESP_LOGI(TAG_FILESERVER, "Deleting file : %s", filename);
         /* Delete file */
         unlink(filepath);
 
@@ -596,10 +608,10 @@ static esp_err_t delete_post_handler(httpd_req_t *req)
         int start_fn = strlen(((struct file_server_data *)req->user_ctx)->base_path);
         printf("Directory: %s, start_fn: %d, found: %d\n", directory.c_str(), start_fn, found);
         directory = directory.substr(start_fn, found - start_fn + 1);
-        printf("Directory danach: %s\n", directory.c_str());    
+        printf("Directory danach 3: %s\n", directory.c_str());    
 
         directory = "/fileserver" + directory;
-        printf("Directory danach: %s\n", directory.c_str());   
+        printf("Directory danach 4: %s\n", directory.c_str());   
     }
     
 
@@ -623,7 +635,7 @@ void delete_all_in_directory(std::string _directory)
     std::string filename;
 
     if (!dir) {
-        ESP_LOGE(TAG, "Failed to stat dir : %s", _directory.c_str());
+        ESP_LOGE(TAG_FILESERVER, "Failed to stat dir : %s", _directory.c_str());
         return;
     }
 
@@ -632,7 +644,7 @@ void delete_all_in_directory(std::string _directory)
         if (!(entry->d_type == DT_DIR)){
             if (strcmp("wlan.ini", entry->d_name) != 0){                    // auf wlan.ini soll nicht zugegriffen werden !!!
                 filename = _directory + "/" + std::string(entry->d_name);
-                ESP_LOGI(TAG, "Deleting file : %s", filename.c_str());
+                ESP_LOGI(TAG_FILESERVER, "Deleting file : %s", filename.c_str());
                 /* Delete file */
                 unlink(filename.c_str());    
             }
@@ -722,19 +734,19 @@ void register_server_file_uri(httpd_handle_t server, const char *base_path)
     /* Validate file storage base path */
     if (!base_path) {
 //    if (!base_path || strcmp(base_path, "/spiffs") != 0) {
-        ESP_LOGE(TAG, "File server base_path not set");
+        ESP_LOGE(TAG_FILESERVER, "File server base_path not set");
 //        return ESP_ERR_INVALID_ARG;
     }
 
     if (server_data) {
-        ESP_LOGE(TAG, "File server already started");
+        ESP_LOGE(TAG_FILESERVER, "File server already started");
 //        return ESP_ERR_INVALID_STATE;
     }
 
     /* Allocate memory for server data */
     server_data = (file_server_data *) calloc(1, sizeof(struct file_server_data));
     if (!server_data) {
-        ESP_LOGE(TAG, "Failed to allocate memory for server data");
+        ESP_LOGE(TAG_FILESERVER, "Failed to allocate memory for server data");
 //        return ESP_ERR_NO_MEM;
     }
     strlcpy(server_data->base_path, base_path,

code/components/jomjol_fileserver_ota/server_help.cpp

@@ -107,6 +107,8 @@ esp_err_t set_content_type_from_file(httpd_req_t *req, const char *filename)
         return httpd_resp_set_type(req, "text/html");
     } else if (IS_FILE_EXT(filename, ".jpeg")) {
         return httpd_resp_set_type(req, "image/jpeg");
+    } else if (IS_FILE_EXT(filename, ".jpg")) {
+        return httpd_resp_set_type(req, "image/jpeg");
     } else if (IS_FILE_EXT(filename, ".ico")) {
         return httpd_resp_set_type(req, "image/x-icon");
     }

code/components/jomjol_fileserver_ota/server_ota.cpp

@@ -12,7 +12,7 @@
 #include "freertos/task.h"
 #include "esp_system.h"
 #include "esp_event.h"
-#include "esp_event_loop.h"
+#include "esp_event.h"
 #include "esp_log.h"
 #include <esp_ota_ops.h>
 #include "esp_http_client.h"
@@ -28,6 +28,7 @@
 
 #include "server_tflite.h"
 #include "server_file.h"
+#include "server_GPIO.h"
 
 #include "ClassLogFile.h"
 
@@ -46,6 +47,7 @@ static char ota_write_data[BUFFSIZE + 1] = { 0 };
 
 
 #define OTA_URL_SIZE 256
+static const char *TAGPARTOTA = "server_ota";
 
 
 static void infinite_loop(void)
@@ -60,14 +62,14 @@ static void infinite_loop(void)
 
 
 
-static bool ota_example_task(std::string fn)
+static bool ota_update_task(std::string fn)
 {
     esp_err_t err;
     /* update handle : set by esp_ota_begin(), must be freed via esp_ota_end() */
     esp_ota_handle_t update_handle = 0 ;
     const esp_partition_t *update_partition = NULL;
 
-    ESP_LOGI(TAGPARTOTA, "Starting OTA example");
+    ESP_LOGI(TAGPARTOTA, "Starting OTA update");
 
     const esp_partition_t *configured = esp_ota_get_boot_partition();
     const esp_partition_t *running = esp_ota_get_running_partition();
@@ -374,7 +376,9 @@ esp_err_t handler_ota_update(httpd_req_t *req)
 
     const char* resp_str;    
 
-    if (ota_example_task(fn))
+    KillTFliteTasks();
+    gpio_handler_deinit();
+    if (ota_update_task(fn))
     {
         resp_str = "Firmware Update Successfull!<br><br>You can restart now.";
     }
@@ -400,8 +404,6 @@ void hard_restart() {
 
 void task_reboot(void *pvParameter)
 {
-
-
     while(1)
     {
         vTaskDelay(5000 / portTICK_PERIOD_MS);
@@ -413,12 +415,14 @@ void task_reboot(void *pvParameter)
 }
 
 void doReboot(){
-    LogFile.WriteToFile("Reboot - now");
-    KillTFliteTasks();
+    ESP_LOGI(TAGPARTOTA, "Reboot in 5sec");
+    LogFile.WriteToFile("Reboot in 5sec");
     xTaskCreate(&task_reboot, "reboot", configMINIMAL_STACK_SIZE * 64, NULL, 10, NULL);
+    // KillTFliteTasks(); // kills itself 
+    gpio_handler_destroy();
     vTaskDelay(5000 / portTICK_PERIOD_MS);
     esp_restart();
-    hard_restart();    
+    hard_restart();
 }
 
 

code/components/jomjol_fileserver_ota/server_ota.h

@@ -4,9 +4,8 @@
 
 //#include "ClassControllCamera.h"
 
-static const char *TAGPARTOTA = "server_ota";
-
 void register_server_ota_sdcard_uri(httpd_handle_t server);
 void CheckOTAUpdate();
 void doReboot();
+void hard_restart();
 

code/components/jomjol_flowcontroll/CMakeLists.txt

@@ -2,6 +2,6 @@ FILE(GLOB_RECURSE app_sources ${CMAKE_CURRENT_SOURCE_DIR}/*.*)
 
 idf_component_register(SRCS ${app_sources}
                     INCLUDE_DIRS "."
-                    REQUIRES jomjol_tfliteclass jomjol_helper jomjol_controlcamera jomjol_mqtt jomjol_fileserver_ota jomjol_image_proc connect_wlan)
+                    REQUIRES jomjol_tfliteclass jomjol_helper jomjol_controlcamera jomjol_mqtt jomjol_fileserver_ota jomjol_image_proc jomjol_wlan)
 
 

code/components/jomjol_flowcontroll/ClassFlow.cpp

@@ -94,6 +94,23 @@ string ClassFlow::getReadout()
 	return string();
 }
 
+std::string ClassFlow::GetParameterName(std::string _input)
+{
+    string _param;
+    int _pospunkt = _input.find_first_of(".");
+    if (_pospunkt > -1)
+    {
+        _param = _input.substr(_pospunkt+1, _input.length() - _pospunkt - 1);
+    }
+    else
+    {
+        _param = _input;
+    }
+//    printf("Parameter: %s, Pospunkt: %d\n", _param.c_str(), _pospunkt);
+	return _param;
+}
+
+
 bool ClassFlow::getNextLine(FILE* pfile, string *rt)
 {
 	char zw[1024];
@@ -102,24 +119,21 @@ bool ClassFlow::getNextLine(FILE* pfile, string *rt)
 		*rt = "";
 		return false;
 	}
-	fgets(zw, 1024, pfile);
-	printf("%s", zw);
-	if ((strlen(zw) == 0) && feof(pfile))
+	if (!fgets(zw, 1024, pfile))
 	{
 		*rt = "";
+		printf("END OF FILE\n");
 		return false;
 	}
+	printf("%s", zw);
 	*rt = zw;
 	*rt = trim(*rt);
 	while ((zw[0] == ';' || zw[0] == '#' || (rt->size() == 0)) && !(zw[1] == '['))			// Kommentarzeilen (; oder #) und Leerzeilen überspringen, es sei denn es ist ein neuer auskommentierter Paragraph
 	{
-		fgets(zw, 1024, pfile);
-		printf("%s", zw);		
-		if (feof(pfile))
-		{
-			*rt = "";
+		*rt = "";
+		if (!fgets(zw, 1024, pfile))
 			return false;
-		}
+		printf("%s", zw);		
 		*rt = zw;
 		*rt = trim(*rt);
 	}

code/components/jomjol_flowcontroll/ClassFlow.h

@@ -37,6 +37,8 @@ protected:
 
 	virtual void SetInitialParameter(void);
 
+	std::string GetParameterName(std::string _input);
+
 	bool disabled;
 
 public:

code/components/jomjol_flowcontroll/ClassFlowAlignment.cpp

@@ -19,6 +19,7 @@ void ClassFlowAlignment::SetInitialParameter(void)
     initalrotate = 0;
     anz_ref = 0;
     initialmirror = false;
+    initialflip = false;
     SaveAllFiles = false;
     namerawimage =  "/sdcard/img_tmp/raw.jpg";
     FileStoreRefAlignment = "/sdcard/config/align.txt";
@@ -72,6 +73,11 @@ bool ClassFlowAlignment::ReadParameter(FILE* pfile, string& aktparamgraph)
     while (this->getNextLine(pfile, &aktparamgraph) && !this->isNewParagraph(aktparamgraph))
     {
         zerlegt = ZerlegeZeile(aktparamgraph);
+        if ((toUpper(zerlegt[0]) == "FLIPIMAGESIZE") && (zerlegt.size() > 1))
+        {
+            if (toUpper(zerlegt[1]) == "TRUE")
+                initialflip = true;
+        }
         if ((toUpper(zerlegt[0]) == "INITIALMIRROR") && (zerlegt.size() > 1))
         {
             if (toUpper(zerlegt[1]) == "TRUE")
@@ -153,7 +159,13 @@ bool ClassFlowAlignment::doFlow(string time)
         delete AlignAndCutImage;
     AlignAndCutImage = new CAlignAndCutImage(ImageBasis, ImageTMP);   
 
-    CRotateImage rt(AlignAndCutImage, ImageTMP);
+    CRotateImage rt(AlignAndCutImage, ImageTMP, initialflip);
+    if (initialflip)
+    {
+        int _zw = ImageBasis->height;
+        ImageBasis->height = ImageBasis->width;
+        ImageBasis->width = _zw;
+    }
 
     if (initialmirror){
         printf("do mirror\n");
@@ -161,7 +173,7 @@ bool ClassFlowAlignment::doFlow(string time)
         if (SaveAllFiles) AlignAndCutImage->SaveToFile(FormatFileName("/sdcard/img_tmp/mirror.jpg"));
     }
  
-    if (initalrotate != 0)
+    if ((initalrotate != 0) || initialflip)
     {
         rt.Rotate(initalrotate);
         if (SaveAllFiles) AlignAndCutImage->SaveToFile(FormatFileName("/sdcard/img_tmp/rot.jpg"));
@@ -176,6 +188,12 @@ bool ClassFlowAlignment::doFlow(string time)
 
     if (SaveAllFiles)
     {
+        if (initialflip)
+        {
+            int _zw = ImageTMP->width;
+            ImageTMP->width = ImageTMP->height;
+            ImageTMP->height = _zw;
+        }
         DrawRef(ImageTMP);
         ImageTMP->SaveToFile(FormatFileName("/sdcard/img_tmp/alg_roi.jpg"));
     }
@@ -209,7 +227,7 @@ void ClassFlowAlignment::SaveReferenceAlignmentValues()
         time(&rawtime);
         timeinfo = localtime(&rawtime);
 
-        strftime(buffer, 80, "%Y-%m-%d_%H-%M-%S", timeinfo);
+        strftime(buffer, 80, "%Y-%m-%dT%H:%M:%S", timeinfo);
         zwtime = std::string(buffer);
     }
 

code/components/jomjol_flowcontroll/ClassFlowAlignment.h

@@ -15,6 +15,7 @@ class ClassFlowAlignment :
 protected:
     float initalrotate;
     bool initialmirror;
+    bool initialflip;
     RefInfo References[2];
     int anz_ref;
     string namerawimage;

code/components/jomjol_flowcontroll/ClassFlowAnalog.cpp

@@ -1,8 +1,10 @@
 #include "ClassFlowAnalog.h"
 
 #include <math.h>
-#include <iomanip>
+#include <iomanip> 
 #include <sys/types.h>
+#include <sstream>      // std::stringstream
+
   
 // #define OHNETFLITE
 
@@ -44,9 +46,9 @@ ClassFlowAnalog::ClassFlowAnalog(std::vector<ClassFlow*>* lfc) : ClassFlowImage(
 }
 
 
-int ClassFlowAnalog::AnzahlROIs()
+int ClassFlowAnalog::AnzahlROIs(int _analog = 0)
 {
-    int zw = ROI.size();
+    int zw = ANALOG[_analog]->ROI.size();
     if (extendedResolution)
         zw++;
     
@@ -54,27 +56,27 @@ int ClassFlowAnalog::AnzahlROIs()
 } 
 
 
-string ClassFlowAnalog::getReadout()
+string ClassFlowAnalog::getReadout(int _analog = 0)
 {
     string result = "";    
-    if (ROI.size() == 0)
+    if (ANALOG[_analog]->ROI.size() == 0)
         return result;
 
 
-    float zahl = ROI[ROI.size() - 1]->result;
+    float zahl = ANALOG[_analog]->ROI[ANALOG[_analog]->ROI.size() - 1]->result;
     int ergebnis_nachkomma = ((int) floor(zahl * 10)) % 10;
 
     int prev = -1;
 
-    prev = ZeigerEval(ROI[ROI.size() - 1]->result, prev);
+    prev = ZeigerEval(ANALOG[_analog]->ROI[ANALOG[_analog]->ROI.size() - 1]->result, prev);
     result = std::to_string(prev);
 
     if (extendedResolution)
         result = result + std::to_string(ergebnis_nachkomma);
 
-    for (int i = ROI.size() - 2; i >= 0; --i)
+    for (int i = ANALOG[_analog]->ROI.size() - 2; i >= 0; --i)
     {
-        prev = ZeigerEval(ROI[i]->result, prev);
+        prev = ZeigerEval(ANALOG[_analog]->ROI[i]->result, prev);
         result = std::to_string(prev) + result;
     }
 
@@ -151,8 +153,8 @@ bool ClassFlowAnalog::ReadParameter(FILE* pfile, string& aktparamgraph)
         }
         if (zerlegt.size() >= 5)
         {
-            roianalog* neuroi = new roianalog;
-            neuroi->name = zerlegt[0];
+            analog* _analog = GetANALOG(zerlegt[0], true);
+            roianalog* neuroi = _analog->ROI[_analog->ROI.size()-1];
             neuroi->posx = std::stoi(zerlegt[1]);
             neuroi->posy = std::stoi(zerlegt[2]);
             neuroi->deltax = std::stoi(zerlegt[3]);
@@ -160,7 +162,7 @@ bool ClassFlowAnalog::ReadParameter(FILE* pfile, string& aktparamgraph)
             neuroi->result = -1;
             neuroi->image = NULL;
             neuroi->image_org = NULL;
-            ROI.push_back(neuroi);
+//            ROI.push_back(neuroi);
         }
 
         if ((toUpper(zerlegt[0]) == "SAVEALLFILES") && (zerlegt.size() > 1))
@@ -176,15 +178,75 @@ bool ClassFlowAnalog::ReadParameter(FILE* pfile, string& aktparamgraph)
         }
     }
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        ROI[i]->image = new CImageBasis(modelxsize, modelysize, 3);
-        ROI[i]->image_org = new CImageBasis(ROI[i]->deltax, ROI[i]->deltay, 3);
-    }
+   for (int _ana = 0; _ana < ANALOG.size(); ++_ana)
+        for (int i = 0; i < ANALOG[_ana]->ROI.size(); ++i)
+        {
+            ANALOG[_ana]->ROI[i]->image = new CImageBasis(modelxsize, modelysize, 3);
+            ANALOG[_ana]->ROI[i]->image_org = new CImageBasis(ANALOG[_ana]->ROI[i]->deltax, ANALOG[_ana]->ROI[i]->deltay, 3);
+        }
 
     return true;
 }
 
+analog* ClassFlowAnalog::FindANALOG(string _name_number)
+{
+
+    for (int i = 0; i < ANALOG.size(); ++i)
+    {
+        if (ANALOG[i]->name == _name_number)
+            return ANALOG[i];
+    }
+
+    return NULL;
+}
+
+
+
+analog* ClassFlowAnalog::GetANALOG(string _name, bool _create = true)
+{
+    string _analog, _roi;
+    int _pospunkt = _name.find_first_of(".");
+//    printf("Name: %s, Pospunkt: %d\n", _name.c_str(), _pospunkt);
+    if (_pospunkt > -1)
+    {
+        _analog = _name.substr(0, _pospunkt);
+        _roi = _name.substr(_pospunkt+1, _name.length() - _pospunkt - 1);
+    }
+    else
+    {
+        _analog = "default";
+        _roi = _name;
+    }
+
+    analog *_ret = NULL;
+
+    for (int i = 0; i < ANALOG.size(); ++i)
+    {
+        if (ANALOG[i]->name == _analog)
+            _ret = ANALOG[i];
+    }
+
+    if (!_create)         // nicht gefunden und soll auch nicht erzeugt werden
+        return _ret;
+
+
+    if (_ret == NULL)
+    {
+        _ret = new analog;
+        _ret->name = _analog;
+        ANALOG.push_back(_ret);
+    }
+
+    roianalog* neuroi = new roianalog;
+    neuroi->name = _roi;
+    _ret->ROI.push_back(neuroi);
+
+    printf("GetANALOG - ANALOG %s - roi %s\n", _analog.c_str(), _roi.c_str());
+
+    return _ret;
+}
+
+
 
 string ClassFlowAnalog::getHTMLSingleStep(string host)
 {
@@ -236,16 +298,29 @@ bool ClassFlowAnalog::doAlignAndCut(string time)
 
     CAlignAndCutImage *caic = flowpostalignment->GetAlignAndCutImage();    
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        printf("Analog %d - Align&Cut\n", i);
-        
-        caic->CutAndSave(ROI[i]->posx, ROI[i]->posy, ROI[i]->deltax, ROI[i]->deltay, ROI[i]->image_org);
-        if (SaveAllFiles) ROI[i]->image_org->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ROI[i]->name + ".jpg"));
+    for (int _ana = 0; _ana < ANALOG.size(); ++_ana)
+        for (int i = 0; i < ANALOG[_ana]->ROI.size(); ++i)
+        {
+            printf("Analog %d - Align&Cut\n", i);
+            
+            caic->CutAndSave(ANALOG[_ana]->ROI[i]->posx, ANALOG[_ana]->ROI[i]->posy, ANALOG[_ana]->ROI[i]->deltax, ANALOG[_ana]->ROI[i]->deltay, ANALOG[_ana]->ROI[i]->image_org);
+            if (SaveAllFiles)
+            {
+                if (ANALOG[_ana]->name == "default")
+                    ANALOG[_ana]->ROI[i]->image_org->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ANALOG[_ana]->ROI[i]->name + ".jpg"));
+                else
+                    ANALOG[_ana]->ROI[i]->image_org->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ANALOG[_ana]->name + "_" + ANALOG[_ana]->ROI[i]->name + ".jpg"));
+            } 
 
-        ROI[i]->image_org->Resize(modelxsize, modelysize, ROI[i]->image);
-        if (SaveAllFiles) ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ROI[i]->name + ".bmp"));
-    }
+            ANALOG[_ana]->ROI[i]->image_org->Resize(modelxsize, modelysize, ANALOG[_ana]->ROI[i]->image);
+            if (SaveAllFiles)
+            {
+                if (ANALOG[_ana]->name == "default")
+                    ANALOG[_ana]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ANALOG[_ana]->ROI[i]->name + ".bmp"));
+                else
+                    ANALOG[_ana]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ANALOG[_ana]->name + "_" + ANALOG[_ana]->ROI[i]->name + ".bmp"));
+            } 
+        }
 
     return true;
 } 
@@ -256,13 +331,14 @@ void ClassFlowAnalog::DrawROI(CImageBasis *_zw)
     int g = 255;
     int b = 0;
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        _zw->drawRect(ROI[i]->posx, ROI[i]->posy, ROI[i]->deltax, ROI[i]->deltay, r, g, b, 1);
-        _zw->drawCircle((int) (ROI[i]->posx + ROI[i]->deltax/2), (int)  (ROI[i]->posy + ROI[i]->deltay/2), (int) (ROI[i]->deltax/2), r, g, b, 2);
-        _zw->drawLine((int) (ROI[i]->posx + ROI[i]->deltax/2), (int) ROI[i]->posy, (int) (ROI[i]->posx + ROI[i]->deltax/2), (int) (ROI[i]->posy + ROI[i]->deltay), r, g, b, 2);
-        _zw->drawLine((int) ROI[i]->posx, (int) (ROI[i]->posy + ROI[i]->deltay/2), (int) ROI[i]->posx + ROI[i]->deltax, (int) (ROI[i]->posy + ROI[i]->deltay/2), r, g, b, 2);
-    }
+    for (int _ana = 0; _ana < ANALOG.size(); ++_ana)
+        for (int i = 0; i < ANALOG[_ana]->ROI.size(); ++i)
+        {
+            _zw->drawRect(ANALOG[_ana]->ROI[i]->posx, ANALOG[_ana]->ROI[i]->posy, ANALOG[_ana]->ROI[i]->deltax, ANALOG[_ana]->ROI[i]->deltay, r, g, b, 1);
+            _zw->drawCircle((int) (ANALOG[_ana]->ROI[i]->posx + ANALOG[_ana]->ROI[i]->deltax/2), (int)  (ANALOG[_ana]->ROI[i]->posy + ANALOG[_ana]->ROI[i]->deltay/2), (int) (ANALOG[_ana]->ROI[i]->deltax/2), r, g, b, 2);
+            _zw->drawLine((int) (ANALOG[_ana]->ROI[i]->posx + ANALOG[_ana]->ROI[i]->deltax/2), (int) ANALOG[_ana]->ROI[i]->posy, (int) (ANALOG[_ana]->ROI[i]->posx + ANALOG[_ana]->ROI[i]->deltax/2), (int) (ANALOG[_ana]->ROI[i]->posy + ANALOG[_ana]->ROI[i]->deltay), r, g, b, 2);
+            _zw->drawLine((int) ANALOG[_ana]->ROI[i]->posx, (int) (ANALOG[_ana]->ROI[i]->posy + ANALOG[_ana]->ROI[i]->deltay/2), (int) ANALOG[_ana]->ROI[i]->posx + ANALOG[_ana]->ROI[i]->deltax, (int) (ANALOG[_ana]->ROI[i]->posy + ANALOG[_ana]->ROI[i]->deltay/2), r, g, b, 2);
+        }
 } 
 
 bool ClassFlowAnalog::doNeuralNetwork(string time)
@@ -282,43 +358,46 @@ bool ClassFlowAnalog::doNeuralNetwork(string time)
     string zwcnn = "/sdcard" + cnnmodelfile;
     zwcnn = FormatFileName(zwcnn);
     printf(zwcnn.c_str());printf("\n");
-    tflite->LoadModel(zwcnn); 
+    if (!tflite->LoadModel(zwcnn)) {
+        printf("Can't read model file /sdcard%s\n", cnnmodelfile.c_str());
+        delete tflite;
+        return false;
+    } 
     tflite->MakeAllocate();
 #endif
 
-    for (int i = 0; i < ROI.size(); ++i)
+    for (int _ana = 0; _ana < ANALOG.size(); ++_ana)
     {
-        printf("Analog %d - TfLite\n", i);
-        ioresize = "/sdcard/img_tmp/ra" + std::to_string(i) + ".bmp";
-        ioresize = FormatFileName(ioresize);
-
-
-        float f1, f2;
-        f1 = 0; f2 = 0;
-
-#ifndef OHNETFLITE
-//        LogFile.WriteToFile("ClassFlowAnalog::doNeuralNetwork vor CNN tflite->LoadInputImage(ioresize)");
-//        tflite->LoadInputImage(ioresize);
-        tflite->LoadInputImageBasis(ROI[i]->image);        
-        tflite->Invoke();
-        if (debugdetailanalog) LogFile.WriteToFile("Nach Invoke");
-
-
-        f1 = tflite->GetOutputValue(0);
-        f2 = tflite->GetOutputValue(1);
-#endif
-
-        float result = fmod(atan2(f1, f2) / (M_PI * 2) + 2, 1);
-//        printf("Result sin, cos, ziffer: %f, %f, %f\n", f1, f2, result);  
-        ROI[i]->result = result * 10;
-
-        printf("Result Analog%i: %f\n", i, ROI[i]->result); 
-
-        if (isLogImage)
+        for (int i = 0; i < ANALOG[_ana]->ROI.size(); ++i)
         {
-            LogImage(logPath, ROI[i]->name, &ROI[i]->result, NULL, time, ROI[i]->image_org);
+            printf("Analog %d - TfLite\n", i);
+
+            float f1, f2;
+            f1 = 0; f2 = 0;
+
+    #ifndef OHNETFLITE
+            tflite->LoadInputImageBasis(ANALOG[_ana]->ROI[i]->image);        
+            tflite->Invoke();
+            if (debugdetailanalog) LogFile.WriteToFile("Nach Invoke");
+
+
+            f1 = tflite->GetOutputValue(0);
+            f2 = tflite->GetOutputValue(1);
+    #endif
+
+            float result = fmod(atan2(f1, f2) / (M_PI * 2) + 2, 1);
+    //        printf("Result sin, cos, ziffer: %f, %f, %f\n", f1, f2, result);  
+            ANALOG[_ana]->ROI[i]->result = result * 10;
+
+            printf("Result Analog%i: %f\n", i, ANALOG[_ana]->ROI[i]->result); 
+
+            if (isLogImage)
+            {
+                LogImage(logPath, ANALOG[_ana]->ROI[i]->name, &ANALOG[_ana]->ROI[i]->result, NULL, time, ANALOG[_ana]->ROI[i]->image_org);
+            }
         }
     }
+
 #ifndef OHNETFLITE
         delete tflite;
 #endif    
@@ -331,18 +410,78 @@ std::vector<HTMLInfo*> ClassFlowAnalog::GetHTMLInfo()
 {
     std::vector<HTMLInfo*> result;
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        HTMLInfo *zw = new HTMLInfo;
-        zw->filename = ROI[i]->name + ".bmp";
-        zw->filename_org = ROI[i]->name + ".jpg";
-        zw->val = ROI[i]->result;
-        zw->image = ROI[i]->image;
-        zw->image_org = ROI[i]->image_org;
-        result.push_back(zw);
-    }
+    for (int _ana = 0; _ana < ANALOG.size(); ++_ana)
+        for (int i = 0; i < ANALOG[_ana]->ROI.size(); ++i)
+        {
+                if (ANALOG[_ana]->name == "default")
+                    ANALOG[_ana]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ANALOG[_ana]->ROI[i]->name + ".bmp"));
+                else
+                    ANALOG[_ana]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ANALOG[_ana]->name + "_" + ANALOG[_ana]->ROI[i]->name + ".bmp"));
+
+
+            HTMLInfo *zw = new HTMLInfo;
+            if (ANALOG[_ana]->name == "default")
+            {
+                zw->filename = ANALOG[_ana]->ROI[i]->name + ".bmp";
+                zw->filename_org = ANALOG[_ana]->ROI[i]->name + ".jpg";
+            }
+            else
+            {
+                zw->filename = ANALOG[_ana]->name + "_" + ANALOG[_ana]->ROI[i]->name + ".bmp";
+                zw->filename_org = ANALOG[_ana]->name + "_" + ANALOG[_ana]->ROI[i]->name + ".jpg";
+            }
+
+            zw->val = ANALOG[_ana]->ROI[i]->result;
+            zw->image = ANALOG[_ana]->ROI[i]->image;
+            zw->image_org = ANALOG[_ana]->ROI[i]->image_org;
+
+            result.push_back(zw);
+        }
 
     return result;
 }
 
 
+
+int ClassFlowAnalog::getAnzahlANALOG()
+{
+    return ANALOG.size();
+}
+
+string ClassFlowAnalog::getNameANALOG(int _analog)
+{
+    if (_analog < ANALOG.size())
+        return ANALOG[_analog]->name;
+
+    return "ANALOG DOES NOT EXIST";
+}
+
+analog* ClassFlowAnalog::GetANALOG(int _analog)
+{
+    if (_analog < ANALOG.size())
+        return ANALOG[_analog];
+
+    return NULL;
+}
+
+
+
+void ClassFlowAnalog::UpdateNameNumbers(std::vector<std::string> *_name_numbers)
+{
+    for (int _dig = 0; _dig < ANALOG.size(); _dig++)
+    {
+        std::string _name = ANALOG[_dig]->name;
+        bool found = false;
+        for (int i = 0; i < (*_name_numbers).size(); ++i)
+        {
+            if ((*_name_numbers)[i] == _name)
+                found = true;
+        }
+        if (!found)
+            (*_name_numbers).push_back(_name);
+    }
+}
+
+
+
+

code/components/jomjol_flowcontroll/ClassFlowAnalog.h

@@ -10,12 +10,19 @@ struct roianalog {
     string name;
 };
 
+struct analog {
+    string name;
+    std::vector<roianalog*> ROI;
+};
+
 
 class ClassFlowAnalog :
     public ClassFlowImage
 {
 protected:
-    std::vector<roianalog*> ROI;
+//    std::vector<roianalog*> ROI;
+    std::vector<analog*> ANALOG;
+
     string cnnmodelfile;
     int modelxsize, modelysize;
     int ZeigerEval(float zahl, int ziffer_vorgaenger);
@@ -24,7 +31,8 @@ protected:
 
     ClassFlowAlignment* flowpostalignment;
 
-	void SetInitialParameter(void);        
+	void SetInitialParameter(void);   
+
 
 public:
     bool extendedResolution;
@@ -34,14 +42,23 @@ public:
     bool ReadParameter(FILE* pfile, string& aktparamgraph);
     bool doFlow(string time);
     string getHTMLSingleStep(string host);
-    string getReadout();   
+    string getReadout(int _analog);   
 
     void DrawROI(CImageBasis *_zw); 
 
     bool doNeuralNetwork(string time); 
     bool doAlignAndCut(string time);
    	std::vector<HTMLInfo*> GetHTMLInfo();   
-    int AnzahlROIs(); 
+    int AnzahlROIs(int _analog);
+
+    int getAnzahlANALOG();
+    analog* GetANALOG(int _analog);
+    analog* GetANALOG(string _name, bool _create);
+    analog* FindANALOG(string _name_number);    
+    string getNameANALOG(int _analog);     
+
+    void UpdateNameNumbers(std::vector<std::string> *_name_numbers);
+
 
     string name(){return "ClassFlowAnalog";}; 
 };

code/components/jomjol_flowcontroll/ClassFlowControll.cpp

@@ -1,6 +1,7 @@
 #include "ClassFlowControll.h"
 
 #include "connect_wlan.h"
+#include "read_wlanini.h"
 
 #include "freertos/task.h"
 
@@ -11,6 +12,9 @@
 #include "Helper.h"
 #include "server_ota.h"
 
+
+//#include "CImg.h"
+
 #include "server_help.h"
 
 //#define DEBUG_DETAIL_ON  
@@ -21,13 +25,16 @@ static const char* TAG = "flow_controll";
 std::string ClassFlowControll::doSingleStep(std::string _stepname, std::string _host){
     std::string _classname = "";
     std::string result = "";
+//    printf("_stepname: %s\n", _stepname.c_str());
     if ((_stepname.compare("[MakeImage]") == 0) || (_stepname.compare(";[MakeImage]") == 0)){
         _classname = "ClassFlowMakeImage";
     }
     if ((_stepname.compare("[Alignment]") == 0) || (_stepname.compare(";[Alignment]") == 0)){
         _classname = "ClassFlowAlignment";
     }
-    if ((_stepname.compare("[Digits]") == 0) || (_stepname.compare(";[Digits]") == 0)){
+    if ((_stepname.compare(0, 7, "[Digits") == 0) || (_stepname.compare(0, 8, ";[Digits") == 0)) {
+//    if ((_stepname.compare("[Digits]") == 0) || (_stepname.compare(";[Digits]") == 0)){
+//        printf("Digits!!!\n");
         _classname = "ClassFlowDigit";
     }
     if ((_stepname.compare("[Analog]") == 0) || (_stepname.compare(";[Analog]") == 0)){
@@ -47,6 +54,25 @@ std::string ClassFlowControll::doSingleStep(std::string _stepname, std::string _
     return result;
 }
 
+std::string ClassFlowControll::TranslateAktstatus(std::string _input)
+{
+    if (_input.compare("ClassFlowMakeImage") == 0)
+        return ("Take Image");
+    if (_input.compare("ClassFlowAlignment") == 0)
+        return ("Aligning");
+    if (_input.compare("ClassFlowAnalog") == 0)
+        return ("Analog ROIs");
+    if (_input.compare("ClassFlowDigit") == 0)
+        return ("Digital ROIs");
+    if (_input.compare("ClassFlowMQTT") == 0)
+        return ("Sending MQTT");
+    if (_input.compare("ClassFlowPostProcessing") == 0)
+        return ("Processing");
+
+    return "Unkown Status";
+}
+
+
 std::vector<HTMLInfo*> ClassFlowControll::GetAllDigital()
 {
     for (int i = 0; i < FlowControll.size(); ++i)
@@ -68,6 +94,19 @@ std::vector<HTMLInfo*> ClassFlowControll::GetAllAnalog()
 }
 
 
+
+string ClassFlowControll::GetMQTTMainTopic()
+{
+    for (int i = 0; i < FlowControll.size(); ++i)
+        if (FlowControll[i]->name().compare("ClassFlowMQTT") == 0)
+            return ((ClassFlowMQTT*) (FlowControll[i]))->GetMQTTMainTopic();
+
+
+    return "";
+}
+
+
+
 void ClassFlowControll::SetInitialParameter(void)
 {
     AutoStart = false;
@@ -78,7 +117,7 @@ void ClassFlowControll::SetInitialParameter(void)
     flowpostprocessing = NULL;
     disabled = false;
     aktRunNr = 0;
-    aktstatus = "Startup";
+    aktstatus = "Booting ...";
 
 }
 
@@ -109,13 +148,14 @@ ClassFlow* ClassFlowControll::CreateClassFlow(std::string _type)
         cfc = new ClassFlowAnalog(&FlowControll);
         flowanalog = (ClassFlowAnalog*) cfc;
     }
-    if (toUpper(_type).compare("[DIGITS]") == 0)
+    if (toUpper(_type).compare(0, 7, "[DIGITS") == 0)
     {
         cfc = new ClassFlowDigit(&FlowControll);
         flowdigit = (ClassFlowDigit*) cfc;
     }
     if (toUpper(_type).compare("[MQTT]") == 0)
         cfc = new ClassFlowMQTT(&FlowControll);
+        
     if (toUpper(_type).compare("[POSTPROCESSING]") == 0)
     {
         cfc = new ClassFlowPostProcessing(&FlowControll); 
@@ -168,9 +208,12 @@ void ClassFlowControll::InitFlow(std::string config)
         }
         else
         {
-            fgets(zw, 1024, pFile);
-            printf("%s", zw);
-            line = std::string(zw);
+            line = "";
+            if (fgets(zw, 1024, pFile) && !feof(pFile))
+                {
+                    printf("Read: %s", zw);
+                    line = std::string(zw);
+                }
         }
     }
 
@@ -188,9 +231,9 @@ void ClassFlowControll::doFlowMakeImageOnly(string time){
     for (int i = 0; i < FlowControll.size(); ++i)
     {
         if (FlowControll[i]->name() == "ClassFlowMakeImage") {
-            zw_time = gettimestring("%Y%m%d-%H%M%S");
-            aktstatus = zw_time + ": " + FlowControll[i]->name();
-            string zw = "FlowControll.doFlowMakeImageOnly - " + FlowControll[i]->name();
+//            zw_time = gettimestring("%Y%m%d-%H%M%S");
+            zw_time = gettimestring("%H:%M:%S");
+            aktstatus = TranslateAktstatus(FlowControll[i]->name()) + " (" + zw_time + ")";
             FlowControll[i]->doFlow(time);
         }
     }
@@ -205,13 +248,16 @@ bool ClassFlowControll::doFlow(string time)
     int repeat = 0;
 
 #ifdef DEBUG_DETAIL_ON 
-    LogFile.WriteHeapInfo("ClassFlowAnalog::doFlow - Start");
+    LogFile.WriteHeapInfo("ClassFlowControll::doFlow - Start");
 #endif
 
     for (int i = 0; i < FlowControll.size(); ++i)
     {
-        zw_time = gettimestring("%Y%m%d-%H%M%S");
-        aktstatus = zw_time + ": " + FlowControll[i]->name();
+        zw_time = gettimestring("%H:%M:%S");
+        aktstatus = TranslateAktstatus(FlowControll[i]->name()) + "(" + zw_time + ")";
+
+//        zw_time = gettimestring("%Y%m%d-%H%M%S");
+//        aktstatus = zw_time + ": " + FlowControll[i]->name();
         
        
         string zw = "FlowControll.doFlow - " + FlowControll[i]->name();
@@ -234,28 +280,54 @@ bool ClassFlowControll::doFlow(string time)
         }
         
 #ifdef DEBUG_DETAIL_ON  
-        LogFile.WriteHeapInfo("ClassFlowAnalog::doFlow");
+        LogFile.WriteHeapInfo("ClassFlowControll::doFlow");
 #endif
 
     }
-    zw_time = gettimestring("%Y%m%d-%H%M%S");    
-    aktstatus = zw_time + ": Flow is done";
+    zw_time = gettimestring("%H:%M:%S");
+    aktstatus = "Flow finished (" + zw_time + ")";
     return result;
 }
 
-void ClassFlowControll::UpdateAktStatus(std::string _flow)
+
+string ClassFlowControll::getReadoutAll(int _type)
 {
-    aktstatus = gettimestring("%Y%m%d-%H%M%S");
-    aktstatus = aktstatus + "\t" + std::to_string(aktRunNr) + "\t";
-    
-    if (_flow == "ClassFlowMakeImage")
-        aktstatus = aktstatus + "Taking Raw Image";
-    else
-        if (_flow == "ClassFlowAlignment")
-            aktstatus = aktstatus + "Aligning Image";
+    std::vector<NumberPost*> numbers = flowpostprocessing->GetNumbers();
+    std::string out = "";
 
+    for (int i = 0; i < numbers.size(); ++i)
+    {
+        out = out + numbers[i]->name + "\t";
+        switch (_type) {
+            case READOUT_TYPE_VALUE:
+                out = out + numbers[i]->ReturnValueNoError;
+                break;
+            case READOUT_TYPE_PREVALUE:
+                if (flowpostprocessing->PreValueUse)
+                {
+                    if (numbers[i]->PreValueOkay)
+                        out = out + numbers[i]->ReturnPreValue;
+                    else
+                        out = out + "PreValue too old";                
+                }
+                else
+                    out = out + "PreValue deactivated";
+                break;
+            case READOUT_TYPE_RAWVALUE:
+                out = out + numbers[i]->ReturnRawValue;
+                break;
+            case READOUT_TYPE_ERROR:
+                out = out + numbers[i]->ErrorMessageText;
+                break;
+        }
+        if (i < numbers.size()-1)
+            out = out + "\r\n";
+    }
 
-}
+//    printf("OUT: %s", out.c_str());
+
+    return out;
+}	
 
 
 string ClassFlowControll::getReadout(bool _rawvalue = false, bool _noerror = false)
@@ -281,17 +353,17 @@ string ClassFlowControll::getReadout(bool _rawvalue = false, bool _noerror = fal
     return result;
 }
 
-string ClassFlowControll::GetPrevalue()	
+string ClassFlowControll::GetPrevalue(std::string _number)	
 {
     if (flowpostprocessing)
     {
-        return flowpostprocessing->GetPreValue();   
+        return flowpostprocessing->GetPreValue(_number);   
     }
 
     return std::string();    
 }
 
-std::string ClassFlowControll::UpdatePrevalue(std::string _newvalue)
+std::string ClassFlowControll::UpdatePrevalue(std::string _newvalue, std::string _numbers, bool _extern)
 {
     float zw;
     char* p;
@@ -313,7 +385,7 @@ std::string ClassFlowControll::UpdatePrevalue(std::string _newvalue)
 
     if (flowpostprocessing)
     {
-        flowpostprocessing->SavePreValue(zw);
+        flowpostprocessing->SetPreValue(zw, _numbers, _extern);
         return _newvalue;    
     }
 
@@ -382,6 +454,9 @@ bool ClassFlowControll::ReadParameter(FILE* pfile, string& aktparamgraph)
             {
                 // reboot notwendig damit die neue wlan.ini auch benutzt wird !!!
                 fclose(pfile);
+                printf("do reboot\n");
+                esp_restart();
+                hard_restart();                   
                 doReboot();
             }
         }
@@ -402,6 +477,7 @@ bool ClassFlowControll::ReadParameter(FILE* pfile, string& aktparamgraph)
     return true;
 }
 
+
 int ClassFlowControll::CleanTempFolder() {
     const char* folderPath = "/sdcard/img_tmp";
     
@@ -435,7 +511,7 @@ int ClassFlowControll::CleanTempFolder() {
 
 esp_err_t ClassFlowControll::SendRawJPG(httpd_req_t *req)
 {
-    return flowmakeimage->SendRawJPG(req);
+    return flowmakeimage != NULL ? flowmakeimage->SendRawJPG(req) : ESP_FAIL;
 }
 
 
@@ -447,6 +523,12 @@ esp_err_t ClassFlowControll::GetJPGStream(std::string _fn, httpd_req_t *req)
     esp_err_t result = ESP_FAIL;
     bool Dodelete = false;    
 
+    if (flowalignment == NULL)
+    {
+        printf("Can't continue, flowalignment is NULL\n");
+        return ESP_FAIL;
+    }
+
     if (_fn == "alg.jpg")
     {
         _send = flowalignment->ImageBasis;  
@@ -460,6 +542,18 @@ esp_err_t ClassFlowControll::GetJPGStream(std::string _fn, httpd_req_t *req)
         flowalignment->DrawRef(_imgzw);
         if (flowdigit) flowdigit->DrawROI(_imgzw);
         if (flowanalog) flowanalog->DrawROI(_imgzw);
+
+/*/////////////////////////////////////        
+        cimg_library::CImg<unsigned char> cimg(_imgzw->rgb_image, _imgzw->bpp, _imgzw->width, _imgzw->height, 1);
+    
+        //Convert cimg type
+//        cimg.permute_axes("yzcx");
+        cimg.draw_text(300, 300, "Dies ist ein Test", "black");
+
+        
+        //Convert back to stb type to save
+//        cimg.permute_axes("cxyz");
+*////////////////////////////////////
         _send = _imgzw;
         Dodelete = true;
     }
@@ -478,7 +572,9 @@ esp_err_t ClassFlowControll::GetJPGStream(std::string _fn, httpd_req_t *req)
             if (htmlinfo[i]->image_org)
                 _send = htmlinfo[i]->image_org;        
         }
+        delete htmlinfo[i];
     }
+    htmlinfo.clear();
 
     htmlinfo = GetAllAnalog();
     for (int i = 0; i < htmlinfo.size(); ++i)
@@ -493,7 +589,9 @@ esp_err_t ClassFlowControll::GetJPGStream(std::string _fn, httpd_req_t *req)
             if (htmlinfo[i]->image_org)
                 _send = htmlinfo[i]->image_org;        
         }
+        delete htmlinfo[i];
     }
+    htmlinfo.clear();
 
     if (_send)
     {

code/components/jomjol_flowcontroll/ClassFlowControll.h

@@ -11,6 +11,12 @@
 #include "ClassFlowMQTT.h"
 
 
+#define READOUT_TYPE_VALUE 0
+#define READOUT_TYPE_PREVALUE 1
+#define READOUT_TYPE_RAWVALUE 2
+#define READOUT_TYPE_ERROR 3
+
+
 class ClassFlowControll :
     public ClassFlow
 {
@@ -30,18 +36,21 @@ protected:
 	std::string aktstatus;
 	int aktRunNr;
 
-	void UpdateAktStatus(std::string _flow);
-
 public:
 	void InitFlow(std::string config);
 	bool doFlow(string time);
 	void doFlowMakeImageOnly(string time);
 	bool getStatusSetupModus(){return SetupModeActive;};
 	string getReadout(bool _rawvalue, bool _noerror);
-	string UpdatePrevalue(std::string _newvalue);
-	string GetPrevalue();	
+	string getReadoutAll(int _type);	
+	string UpdatePrevalue(std::string _newvalue, std::string _numbers, bool _extern);
+	string GetPrevalue(std::string _number = "");	
 	bool ReadParameter(FILE* pfile, string& aktparamgraph);	
 
+	string TranslateAktstatus(std::string _input);
+
+	string GetMQTTMainTopic();
+
 	esp_err_t GetJPGStream(std::string _fn, httpd_req_t *req);
 	esp_err_t SendRawJPG(httpd_req_t *req);
 

code/components/jomjol_flowcontroll/ClassFlowDigit.cpp

@@ -26,7 +26,9 @@ void ClassFlowDigit::SetInitialParameter(void)
     previousElement = NULL;    
     SaveAllFiles = false;
     disabled = false;
-
+    DecimalShift = 0;
+    DecimalShiftEnabled = false;
+    isLogImageSelect = false;
 }    
 
 ClassFlowDigit::ClassFlowDigit() : ClassFlowImage(TAG)
@@ -63,16 +65,16 @@ ClassFlowDigit::ClassFlowDigit(std::vector<ClassFlow*>* lfc, ClassFlow *_prev) :
     }    
 }
 
-string ClassFlowDigit::getReadout()
+string ClassFlowDigit::getReadout(int _digit = 0)
 {
     string rst = "";
 
-    for (int i = 0; i < ROI.size(); ++i)
+    for (int i = 0; i < DIGIT[_digit]->ROI.size(); ++i)
     {
-        if (ROI[i]->resultklasse == 10)
+        if (DIGIT[_digit]->ROI[i]->resultklasse == 10)
             rst = rst + "N";
         else
-            rst = rst + std::to_string(ROI[i]->resultklasse);
+            rst = rst + std::to_string(DIGIT[_digit]->ROI[i]->resultklasse);
     }
 
     return rst;
@@ -88,9 +90,19 @@ bool ClassFlowDigit::ReadParameter(FILE* pfile, string& aktparamgraph)
         if (!this->GetNextParagraph(pfile, aktparamgraph)) 
             return false;
 
-    if ((aktparamgraph.compare("[Digits]") != 0) && (aktparamgraph.compare(";[Digits]") != 0))       // Paragraph passt nich zu MakeImage
+    printf("aktparamgraph: %s\n", aktparamgraph.c_str());
+
+    if ((aktparamgraph.compare(0, 7, "[Digits") != 0) && (aktparamgraph.compare(0, 8, ";[Digits") != 0))       // Paragraph passt nich zu MakeImage
         return false;
 
+    int _pospkt = aktparamgraph.find_first_of(".");
+    int _posklammerzu = aktparamgraph.find_first_of("]");
+    if (_pospkt > -1)
+        NameDigit = aktparamgraph.substr(_pospkt+1, _posklammerzu - _pospkt-1);
+    else
+        NameDigit = "";
+    printf("Name Digit: %s\n", NameDigit.c_str());
+
     if (aktparamgraph[0] == ';')
     {
         disabled = true;
@@ -108,6 +120,13 @@ bool ClassFlowDigit::ReadParameter(FILE* pfile, string& aktparamgraph)
             LogImageLocation = "/sdcard" + zerlegt[1];
             isLogImage = true;            
         }
+
+        if ((zerlegt[0] == "LogImageSelect") && (zerlegt.size() > 1))
+        {
+            LogImageSelect = zerlegt[1];
+            isLogImageSelect = true;            
+        }
+
         if ((zerlegt[0] == "Model") && (zerlegt.size() > 1))
         {
             cnnmodelfile = zerlegt[1];
@@ -119,8 +138,8 @@ bool ClassFlowDigit::ReadParameter(FILE* pfile, string& aktparamgraph)
         }
         if (zerlegt.size() >= 5)
         {
-            roi* neuroi = new roi;
-            neuroi->name = zerlegt[0];
+            digit* _digit = GetDIGIT(zerlegt[0], true);
+            roi* neuroi = _digit->ROI[_digit->ROI.size()-1];
             neuroi->posx = std::stoi(zerlegt[1]);
             neuroi->posy = std::stoi(zerlegt[2]);
             neuroi->deltax = std::stoi(zerlegt[3]);
@@ -128,7 +147,6 @@ bool ClassFlowDigit::ReadParameter(FILE* pfile, string& aktparamgraph)
             neuroi->resultklasse = -1;
             neuroi->image = NULL;
             neuroi->image_org = NULL;            
-            ROI.push_back(neuroi);
         }
 
         if ((toUpper(zerlegt[0]) == "SAVEALLFILES") && (zerlegt.size() > 1))
@@ -139,15 +157,72 @@ bool ClassFlowDigit::ReadParameter(FILE* pfile, string& aktparamgraph)
 
     }
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        ROI[i]->image = new CImageBasis(modelxsize, modelysize, 3);
-        ROI[i]->image_org = new CImageBasis(ROI[i]->deltax, ROI[i]->deltay, 3);
-    }
+   for (int _dig = 0; _dig < DIGIT.size(); ++_dig)
+        for (int i = 0; i < DIGIT[_dig]->ROI.size(); ++i)
+        {
+            DIGIT[_dig]->ROI[i]->image = new CImageBasis(modelxsize, modelysize, 3);
+            DIGIT[_dig]->ROI[i]->image_org = new CImageBasis(DIGIT[_dig]->ROI[i]->deltax, DIGIT[_dig]->ROI[i]->deltay, 3);
+        }
 
     return true;
 }
 
+digit* ClassFlowDigit::FindDIGIT(string _name_number)
+{
+    for (int i = 0; i < DIGIT.size(); ++i)
+    {
+        if (DIGIT[i]->name == _name_number)
+            return DIGIT[i];
+    }
+
+    return NULL;
+}
+
+
+digit* ClassFlowDigit::GetDIGIT(string _name, bool _create = true)
+{
+    string _digit, _roi;
+    int _pospunkt = _name.find_first_of(".");
+//    printf("Name: %s, Pospunkt: %d\n", _name.c_str(), _pospunkt);
+    if (_pospunkt > -1)
+    {
+        _digit = _name.substr(0, _pospunkt);
+        _roi = _name.substr(_pospunkt+1, _name.length() - _pospunkt - 1);
+    }
+    else
+    {
+        _digit = "default";
+        _roi = _name;
+    }
+
+    digit *_ret = NULL;
+
+    for (int i = 0; i < DIGIT.size(); ++i)
+    {
+        if (DIGIT[i]->name == _digit)
+            _ret = DIGIT[i];
+    }
+
+    if (!_create)         // nicht gefunden und soll auch nicht erzeugt werden, ggf. geht eine NULL zurück
+        return _ret;
+
+    if (_ret == NULL)
+    {
+        _ret = new digit;
+        _ret->name = _digit;
+        DIGIT.push_back(_ret);
+    }
+
+    roi* neuroi = new roi;
+    neuroi->name = _roi;
+    _ret->ROI.push_back(neuroi);
+
+    printf("GetDIGIT - digit %s - roi %s\n", _digit.c_str(), _roi.c_str());
+
+    return _ret;
+}
+
+
 
 string ClassFlowDigit::getHTMLSingleStep(string host)
 {
@@ -198,17 +273,32 @@ bool ClassFlowDigit::doAlignAndCut(string time)
 
     CAlignAndCutImage *caic = flowpostalignment->GetAlignAndCutImage();
 
-    for (int i = 0; i < ROI.size(); ++i)
+    for (int _dig = 0; _dig < DIGIT.size(); ++_dig)
     {
-        printf("DigitalDigit %d - Align&Cut\n", i);
-        
-        caic->CutAndSave(ROI[i]->posx, ROI[i]->posy, ROI[i]->deltax, ROI[i]->deltay, ROI[i]->image_org);
-        if (SaveAllFiles) ROI[i]->image_org->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ROI[i]->name + ".jpg"));
+        printf("DIGIT[_dig]->ROI.size() %d\n", DIGIT[_dig]->ROI.size());
+        for (int i = 0; i < DIGIT[_dig]->ROI.size(); ++i)
+        {
+            printf("DigitalDigit %d - Align&Cut\n", i);
+            
+            caic->CutAndSave(DIGIT[_dig]->ROI[i]->posx, DIGIT[_dig]->ROI[i]->posy, DIGIT[_dig]->ROI[i]->deltax, DIGIT[_dig]->ROI[i]->deltay, DIGIT[_dig]->ROI[i]->image_org);
+            if (SaveAllFiles)
+            {
+                if (DIGIT[_dig]->name == "default")
+                    DIGIT[_dig]->ROI[i]->image_org->SaveToFile(FormatFileName("/sdcard/img_tmp/" + DIGIT[_dig]->ROI[i]->name + ".jpg"));
+                else
+                    DIGIT[_dig]->ROI[i]->image_org->SaveToFile(FormatFileName("/sdcard/img_tmp/" + DIGIT[_dig]->name + "_" + DIGIT[_dig]->ROI[i]->name + ".jpg"));
+            } 
 
-        ROI[i]->image_org->Resize(modelxsize, modelysize, ROI[i]->image);
-        if (SaveAllFiles) ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + ROI[i]->name + ".bmp"));
+            DIGIT[_dig]->ROI[i]->image_org->Resize(modelxsize, modelysize, DIGIT[_dig]->ROI[i]->image);
+            if (SaveAllFiles)
+            {
+                if (DIGIT[_dig]->name == "default")
+                    DIGIT[_dig]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + DIGIT[_dig]->ROI[i]->name + ".bmp"));
+                else
+                    DIGIT[_dig]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + DIGIT[_dig]->name + "_" + DIGIT[_dig]->ROI[i]->name + ".bmp"));
+            } 
+        }
     }
-
     return true;
 } 
 
@@ -223,26 +313,42 @@ bool ClassFlowDigit::doNeuralNetwork(string time)
     CTfLiteClass *tflite = new CTfLiteClass;  
     string zwcnn =  FormatFileName("/sdcard" + cnnmodelfile);
     printf(zwcnn.c_str());printf("\n");
-    tflite->LoadModel(zwcnn); 
+    if (!tflite->LoadModel(zwcnn)) {
+        printf("Can't read model file /sdcard%s\n", cnnmodelfile.c_str());
+        delete tflite;
+        return false;
+    } 
+
     tflite->MakeAllocate();
 #endif
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        printf("DigitalDigit %d - TfLite\n", i);
-
-        ROI[i]->resultklasse = 0;
-#ifndef OHNETFLITE
-        ROI[i]->resultklasse = tflite->GetClassFromImageBasis(ROI[i]->image);
-
-#endif
-        printf("Result Digit%i: %d\n", i, ROI[i]->resultklasse);
-
-        if (isLogImage)
+    for (int _dig = 0; _dig < DIGIT.size(); ++_dig)
+        for (int i = 0; i < DIGIT[_dig]->ROI.size(); ++i)
         {
-            LogImage(logPath, ROI[i]->name, NULL, &ROI[i]->resultklasse, time, ROI[i]->image_org);
+            printf("DigitalDigit %d - TfLite\n", i);
+
+            DIGIT[_dig]->ROI[i]->resultklasse = 0;
+    #ifndef OHNETFLITE
+            DIGIT[_dig]->ROI[i]->resultklasse = tflite->GetClassFromImageBasis(DIGIT[_dig]->ROI[i]->image);
+
+    #endif
+            printf("Result Digit%i: %d\n", i, DIGIT[_dig]->ROI[i]->resultklasse);
+
+            if (isLogImage)
+            {
+                if (isLogImageSelect)
+                {
+                    if (LogImageSelect.find(DIGIT[_dig]->ROI[i]->name) != std::string::npos)
+                    {
+                        LogImage(logPath, DIGIT[_dig]->ROI[i]->name, NULL, &DIGIT[_dig]->ROI[i]->resultklasse, time, DIGIT[_dig]->ROI[i]->image_org);
+                    }
+                }
+                else
+                {
+                    LogImage(logPath, DIGIT[_dig]->ROI[i]->name, NULL, &DIGIT[_dig]->ROI[i]->resultklasse, time, DIGIT[_dig]->ROI[i]->image_org);
+                }
+            }
         }
-    }
 #ifndef OHNETFLITE
         delete tflite;
 #endif
@@ -251,25 +357,82 @@ bool ClassFlowDigit::doNeuralNetwork(string time)
 
 void ClassFlowDigit::DrawROI(CImageBasis *_zw)
 {
-    for (int i = 0; i < ROI.size(); ++i)
-        _zw->drawRect(ROI[i]->posx, ROI[i]->posy, ROI[i]->deltax, ROI[i]->deltay, 0, 0, 255, 2);
+    for (int _dig = 0; _dig < DIGIT.size(); ++_dig)
+        for (int i = 0; i < DIGIT[_dig]->ROI.size(); ++i)
+            _zw->drawRect(DIGIT[_dig]->ROI[i]->posx, DIGIT[_dig]->ROI[i]->posy, DIGIT[_dig]->ROI[i]->deltax, DIGIT[_dig]->ROI[i]->deltay, 0, 0, (255 - _dig*100), 2);
 }     
 
 std::vector<HTMLInfo*> ClassFlowDigit::GetHTMLInfo()
 {
     std::vector<HTMLInfo*> result;
 
-    for (int i = 0; i < ROI.size(); ++i)
-    {
-        HTMLInfo *zw = new HTMLInfo;
-        zw->filename = ROI[i]->name + ".bmp";
-        zw->filename_org = ROI[i]->name + ".jpg";
-        zw->val = ROI[i]->resultklasse;
-        zw->image = ROI[i]->image;
-        zw->image_org = ROI[i]->image_org;
-        result.push_back(zw);
-    }
+    for (int _dig = 0; _dig < DIGIT.size(); ++_dig)
+        for (int i = 0; i < DIGIT[_dig]->ROI.size(); ++i)
+        {
+                if (DIGIT[_dig]->name == "default")
+                    DIGIT[_dig]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + DIGIT[_dig]->ROI[i]->name + ".bmp"));
+                else
+                    DIGIT[_dig]->ROI[i]->image->SaveToFile(FormatFileName("/sdcard/img_tmp/" + DIGIT[_dig]->name + "_" + DIGIT[_dig]->ROI[i]->name + ".bmp"));
+
+
+            HTMLInfo *zw = new HTMLInfo;
+            if (DIGIT[_dig]->name == "default")
+            {
+                zw->filename = DIGIT[_dig]->ROI[i]->name + ".bmp";
+                zw->filename_org = DIGIT[_dig]->ROI[i]->name + ".jpg";
+            }
+            else
+            {
+                zw->filename = DIGIT[_dig]->name + "_" + DIGIT[_dig]->ROI[i]->name + ".bmp";
+                zw->filename_org = DIGIT[_dig]->name + "_" + DIGIT[_dig]->ROI[i]->name + ".jpg";
+            }
+
+            zw->val = DIGIT[_dig]->ROI[i]->resultklasse;
+            zw->image = DIGIT[_dig]->ROI[i]->image;
+            zw->image_org = DIGIT[_dig]->ROI[i]->image_org;
+            result.push_back(zw);
+        }
 
     return result;
 }
 
+int ClassFlowDigit::getAnzahlDIGIT()
+{
+    return DIGIT.size();
+}
+
+string ClassFlowDigit::getNameDIGIT(int _digit)
+{
+    if (_digit < DIGIT.size())
+        return DIGIT[_digit]->name;
+
+    return "DIGIT DOES NOT EXIST";
+}
+
+digit* ClassFlowDigit::GetDIGIT(int _digit)
+{
+    if (_digit < DIGIT.size())
+        return DIGIT[_digit];
+
+    return NULL;
+}
+
+void ClassFlowDigit::UpdateNameNumbers(std::vector<std::string> *_name_numbers)
+{
+    for (int _dig = 0; _dig < DIGIT.size(); _dig++)
+    {
+        std::string _name = DIGIT[_dig]->name;
+        bool found = false;
+        for (int i = 0; i < (*_name_numbers).size(); ++i)
+        {
+            if ((*_name_numbers)[i] == _name)
+                found = true;
+        }
+        if (!found)
+            (*_name_numbers).push_back(_name);
+    }
+}
+
+
+
+

code/components/jomjol_flowcontroll/ClassFlowDigit.h

@@ -5,6 +5,8 @@
 
 #include <string>
 
+
+
 struct roi {
     int posx, posy, deltax, deltay;
     int resultklasse;
@@ -13,20 +15,34 @@ struct roi {
     roi* next;
 };
 
+struct digit {
+    string name;
+    std::vector<roi*> ROI;
+};
+
 class ClassFlowDigit :
     public ClassFlowImage
 {
 protected:
-    std::vector<roi*> ROI;
+//    std::vector<roi*> ROI;
+    std::vector<digit*> DIGIT;
     string cnnmodelfile;
     int modelxsize, modelysize;
     bool SaveAllFiles;
+    string NameDigit;
+    int DecimalShift;
+    bool DecimalShiftEnabled;
+
+    bool isLogImageSelect;
+    string LogImageSelect;
+
 
     ClassFlowAlignment* flowpostalignment;
  
     bool doNeuralNetwork(string time); 
     bool doAlignAndCut(string time); 
 
+
 	void SetInitialParameter(void);    
 
 public:
@@ -36,9 +52,18 @@ public:
     bool ReadParameter(FILE* pfile, string& aktparamgraph);
     bool doFlow(string time);
     string getHTMLSingleStep(string host); 
-    string getReadout();
+    string getReadout(int _digit);
    	std::vector<HTMLInfo*> GetHTMLInfo();
 
+    int getAnzahlDIGIT();
+    digit* GetDIGIT(int _digit);
+    digit* GetDIGIT(string _name, bool _create);
+    digit* FindDIGIT(string _name_number);
+
+    string getNameDIGIT(int _digit);
+
+    void UpdateNameNumbers(std::vector<std::string> *_name_numbers);
+
     void DrawROI(CImageBasis *_zw);        
 
     string name(){return "ClassFlowDigit";};

code/components/jomjol_flowcontroll/ClassFlowMQTT.cpp

@@ -1,6 +1,8 @@
+#include <sstream>
 #include "ClassFlowMQTT.h"
 #include "Helper.h"
 
+#include "time_sntp.h"
 #include "interface_mqtt.h"
 #include "ClassFlowPostProcessing.h"
 
@@ -11,6 +13,13 @@ void ClassFlowMQTT::SetInitialParameter(void)
     uri = "";
     topic = "";
     topicError = "";
+    topicRate = "";
+    topicTimeStamp = "";
+    maintopic = "";
+    mainerrortopic = ""; 
+
+    topicUptime = "";
+    topicFreeMem = "";
     clientname = "watermeter";
     OldValue = "";
     flowpostprocessing = NULL;  
@@ -19,6 +28,9 @@ void ClassFlowMQTT::SetInitialParameter(void)
     previousElement = NULL;
     ListFlowControll = NULL; 
     disabled = false;
+    MQTTenable = false;
+    
+    
 
 }       
 
@@ -86,40 +98,98 @@ bool ClassFlowMQTT::ReadParameter(FILE* pfile, string& aktparamgraph)
         {
             this->uri = zerlegt[1];
         }
-        if ((toUpper(zerlegt[0]) == "TOPIC") && (zerlegt.size() > 1))
-        {
-            this->topic = zerlegt[1];
-        }
-        if ((toUpper(zerlegt[0]) == "TOPICERROR") && (zerlegt.size() > 1))
-        {
-            this->topicError = zerlegt[1];
-        }
+
         if ((toUpper(zerlegt[0]) == "CLIENTID") && (zerlegt.size() > 1))
         {
             this->clientname = zerlegt[1];
         }
 
+        if (((toUpper(zerlegt[0]) == "TOPIC") || (toUpper(zerlegt[0]) == "MAINTOPIC")) && (zerlegt.size() > 1))
+        {
+            maintopic = zerlegt[1];
+        }
     }
 
-    if ((uri.length() > 0) && (topic.length() > 0)) 
+    if (!MQTTisConnected() && (uri.length() > 0) && (maintopic.length() > 0)) 
     {
-        MQTTInit(uri, clientname, user, password, topicError, 60);
+        mainerrortopic = maintopic + "/connection";
+        MQTTInit(uri, clientname, user, password, mainerrortopic, 60); 
+        MQTTPublish(mainerrortopic, "connected");
+        MQTTenable = true;
     }
    
     return true;
 }
 
 
+string ClassFlowMQTT::GetMQTTMainTopic()
+{
+    return maintopic;
+}
+
+
 bool ClassFlowMQTT::doFlow(string zwtime)
 {
+    if (!MQTTenable)
+        return true;
+
     std::string result;
     std::string resulterror = "";
+    std::string resultrate = "";
+    std::string resulttimestamp = "";
     string zw = "";
+    string namenumber = "";
+
+    MQTTPublish(mainerrortopic, "connected");
     
+    zw = maintopic + "/" + "uptime";
+    char uptimeStr[11];
+    sprintf(uptimeStr, "%ld", (long)getUpTime());
+    MQTTPublish(zw, uptimeStr);
+
+    zw = maintopic + "/" + "freeMem";
+    char freeheapmem[11];
+    sprintf(freeheapmem, "%zu", esp_get_free_heap_size());
+    MQTTPublish(zw, freeheapmem);
+
     if (flowpostprocessing)
     {
-        result =  flowpostprocessing->getReadoutParam(false, true);
-        resulterror = flowpostprocessing->getReadoutError();
+        std::vector<NumberPost*> NUMBERS = flowpostprocessing->GetNumbers();
+
+        for (int i = 0; i < NUMBERS.size(); ++i)
+        {
+            result =  NUMBERS[i]->ReturnValueNoError;
+            resulterror = NUMBERS[i]->ErrorMessageText;
+            resultrate = std::to_string(NUMBERS[i]->FlowRateAct);
+            resulttimestamp = NUMBERS[i]->timeStamp;
+
+            namenumber = NUMBERS[i]->name;
+            if (namenumber == "default")
+                namenumber = maintopic + "/";
+            else
+                namenumber = maintopic + "/" + namenumber + "/";
+
+            zw = namenumber + "value";    
+            MQTTPublish(zw, result);
+
+            zw = namenumber + "error";    
+            MQTTPublish(zw, resulterror, 1);
+
+            zw = namenumber + "rate";    
+            MQTTPublish(zw, resultrate);
+
+            zw = namenumber + "timestamp";
+            MQTTPublish(zw, resulttimestamp);
+
+
+            std::string json="{\"value\":"+result;
+            json += ",\"error\":\""+resulterror;
+            json += "\",\"rate\":"+resultrate;
+            json += ",\"timestamp\":\""+resulttimestamp+"\"}";
+
+            zw = namenumber + "json";
+            MQTTPublish(zw, json);
+        }
     }
     else
     {
@@ -134,14 +204,9 @@ bool ClassFlowMQTT::doFlow(string zwtime)
                     result = result + "\t" + zw;
             }
         }
+        MQTTPublish(topic, result);
     }
     
-    MQTTPublish(topic, result);
-
-    if (topicError.length() > 0) {
-        MQTTPublish(topicError, resulterror);
-    }
-
     OldValue = result;
     
     return true;

code/components/jomjol_flowcontroll/ClassFlowMQTT.h

@@ -9,10 +9,13 @@ class ClassFlowMQTT :
     public ClassFlow
 {
 protected:
-    std::string uri, topic, topicError, clientname;
+    std::string uri, topic, topicError, clientname, topicRate, topicTimeStamp, topicUptime, topicFreeMem;
     std::string OldValue;
 	ClassFlowPostProcessing* flowpostprocessing;  
-    std::string user, password;  
+    std::string user, password; 
+    bool MQTTenable;
+
+    std::string maintopic, mainerrortopic; 
 	void SetInitialParameter(void);        
 
 public:
@@ -20,6 +23,8 @@ public:
     ClassFlowMQTT(std::vector<ClassFlow*>* lfc);
     ClassFlowMQTT(std::vector<ClassFlow*>* lfc, ClassFlow *_prev);
 
+    string GetMQTTMainTopic();
+
     bool ReadParameter(FILE* pfile, string& aktparamgraph);
     bool doFlow(string time);
     string name(){return "ClassFlowMQTT";};

code/components/jomjol_flowcontroll/ClassFlowMakeImage.cpp

@@ -14,12 +14,23 @@ static const char* TAG = "flow_make_image";
 esp_err_t ClassFlowMakeImage::camera_capture(){
     string nm =  namerawimage;
     Camera.CaptureToFile(nm);
+    time(&TimeImageTaken);
+    localtime(&TimeImageTaken);
+
     return ESP_OK;
 }
 
 void ClassFlowMakeImage::takePictureWithFlash(int flashdauer)
 {
+    // für den Fall, dass das Bild geflippt wird, muss es hier zurück gesetzt werden ////
+    rawImage->width = image_width;          
+    rawImage->height = image_height;
+    /////////////////////////////////////////////////////////////////////////////////////
+    printf("Flashdauer: %d\n", flashdauer);
     Camera.CaptureToBasisImage(rawImage, flashdauer);
+    time(&TimeImageTaken);
+    localtime(&TimeImageTaken);
+
     if (SaveAllFiles) rawImage->SaveToFile(namerawimage);
 }
 
@@ -82,6 +93,12 @@ bool ClassFlowMakeImage::ReadParameter(FILE* pfile, string& aktparamgraph)
             if (toUpper(zerlegt[1]) == "TRUE")
                 SaveAllFiles = true;
         }
+        
+        if ((toUpper(zerlegt[0]) == "WAITBEFORETAKINGPICTURE") && (zerlegt.size() > 1))
+        {
+            waitbeforepicture = stoi(zerlegt[1]);
+        }
+
 
         if ((toUpper(zerlegt[0]) == "BRIGHTNESS") && (zerlegt.size() > 1))
         {
@@ -114,9 +131,9 @@ bool ClassFlowMakeImage::ReadParameter(FILE* pfile, string& aktparamgraph)
     rawImage->CreateEmptyImage(image_width, image_height, 3);
 
     waitbeforepicture_store = waitbeforepicture;
-    if (FixedExposure)
+    if (FixedExposure && (waitbeforepicture > 0))
     {
-        printf("Fixed Exposure enabled!\n");
+//        printf("Fixed Exposure enabled!\n");
         int flashdauer = (int) (waitbeforepicture * 1000);
         Camera.EnableAutoExposure(flashdauer);
         waitbeforepicture = 0.2;
@@ -165,6 +182,9 @@ bool ClassFlowMakeImage::doFlow(string zwtime)
 esp_err_t ClassFlowMakeImage::SendRawJPG(httpd_req_t *req)
 {
     int flashdauer = (int) (waitbeforepicture * 1000);
+    time(&TimeImageTaken);
+    localtime(&TimeImageTaken);
+
     return Camera.CaptureToHTTP(req, flashdauer);
 }
 
@@ -175,6 +195,9 @@ ImageData* ClassFlowMakeImage::SendRawImage()
     ImageData *id;
     int flashdauer = (int) (waitbeforepicture * 1000);
     Camera.CaptureToBasisImage(zw, flashdauer);
+    time(&TimeImageTaken);
+    localtime(&TimeImageTaken);
+
     id = zw->writeToMemoryAsJPG();    
     delete zw;
     return id;  

code/components/jomjol_flowcontroll/ClassFlowPostProcessing.cpp

@@ -1,8 +1,6 @@
 #include "ClassFlowPostProcessing.h"
 
 #include "Helper.h"
-#include "ClassFlowAnalog.h"
-#include "ClassFlowDigit.h"
 #include "ClassFlowMakeImage.h"
 #include "ClassLogFile.h"
 
@@ -11,127 +9,223 @@
 
 #include <time.h>
 
-string ClassFlowPostProcessing::GetPreValue()
+#include "time_sntp.h"
+
+
+#define PREVALUE_TIME_FORMAT_OUTPUT "%Y-%m-%dT%H:%M:%S"
+#define PREVALUE_TIME_FORMAT_INPUT "%d-%d-%dT%d:%d:%d"
+
+
+string ClassFlowPostProcessing::GetPreValue(std::string _number)
 {
     std::string result;
-    bool isAnalog = false;
-    bool isDigit = false;
+    int index = -1;
 
-    int AnzahlAnalog = 0;
-    result = RundeOutput(PreValue, -DecimalShift);
+    if (_number == "")
+        _number = "default";
 
-    for (int i = 0; i < ListFlowControll->size(); ++i)
-    {
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowAnalog") == 0)
-        {
-            isAnalog = true;
-            AnzahlAnalog = ((ClassFlowAnalog*)(*ListFlowControll)[i])->AnzahlROIs();
-        }
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowDigit") == 0)
-        {
-            isDigit = true;
-        }
-    }
+    for (int i = 0; i < NUMBERS.size(); ++i)
+        if (NUMBERS[i]->name == _number)
+            index = i;
 
-    if (isDigit && isAnalog)
-        result = RundeOutput(PreValue, AnzahlAnalog - DecimalShift);
+//    result = RundeOutput(NUMBERS[index]->PreValue, -NUMBERS[index]->DecimalShift);
+    result = RundeOutput(NUMBERS[index]->PreValue, NUMBERS[index]->Nachkomma);
+
+//    if (NUMBERS[index]->digit_roi && NUMBERS[index]->analog_roi)
+//        result = RundeOutput(NUMBERS[index]->PreValue, NUMBERS[index]->AnzahlAnalog - NUMBERS[index]->DecimalShift);
 
     return result;
 }
 
+void ClassFlowPostProcessing::SetPreValue(float zw, string _numbers, bool _extern)
+{
+    printf("SetPrevalue: %f, %s\n", zw, _numbers.c_str());
+    for (int j = 0; j < NUMBERS.size(); ++j)
+    {
+//        printf("Number %d, %s\n", j, NUMBERS[j]->name.c_str());
+        if (NUMBERS[j]->name == _numbers)
+        {
+            NUMBERS[j]->PreValue = zw;
+            if (_extern)
+            {
+                time(&(NUMBERS[j]->lastvalue));
+                localtime(&(NUMBERS[j]->lastvalue));
+            }
+//            printf("Found %d! - set to %f\n", j,  NUMBERS[j]->PreValue);
+        }
+    }
+    UpdatePreValueINI = true;
+    SavePreValue();
+}
+
+
 bool ClassFlowPostProcessing::LoadPreValue(void)
 {
+    std::vector<string> zerlegt;
     FILE* pFile;
     char zw[1024];
-    string zwtime, zwvalue;
+    string zwtime, zwvalue, name;
+    bool _done = false;
+
+    UpdatePreValueINI = false;       // Konvertierung ins neue Format
+
 
     pFile = fopen(FilePreValue.c_str(), "r");
     if (pFile == NULL)
         return false;
 
     fgets(zw, 1024, pFile);
-    printf("%s", zw);
+    printf("Read Zeile Prevalue.ini: %s", zw);
     zwtime = trim(std::string(zw));
-
-    fgets(zw, 1024, pFile);
-    fclose(pFile);
-    printf("%s", zw);
-    zwvalue = trim(std::string(zw));
-    PreValue = stof(zwvalue.c_str());
-
-    time_t tStart;
-    int yy, month, dd, hh, mm, ss;
-    struct tm whenStart;
-
-    sscanf(zwtime.c_str(), "%d-%d-%d_%d-%d-%d", &yy, &month, &dd, &hh, &mm, &ss);
-    whenStart.tm_year = yy - 1900;
-    whenStart.tm_mon = month - 1;
-    whenStart.tm_mday = dd;
-    whenStart.tm_hour = hh;
-    whenStart.tm_min = mm;
-    whenStart.tm_sec = ss;
-    whenStart.tm_isdst = -1;
-
-    tStart = mktime(&whenStart);
-
-    time_t now;
-    time(&now);
-    localtime(&now);
-    double difference = difftime(now, tStart);
-    difference /= 60;
-    if (difference > PreValueAgeStartup)
+    if (zwtime.length() == 0)
         return false;
 
-    Value = PreValue;
-    ReturnValue = to_string(Value);
-    ReturnValueNoError = ReturnValue; 
-
-    bool isAnalog = false;
-    bool isDigit = false;
-    int AnzahlAnalog = 0;
-
-    for (int i = 0; i < ListFlowControll->size(); ++i)
+    zerlegt = HelperZerlegeZeile(zwtime, "\t");
+    if (zerlegt.size() > 1)     // neues Format
     {
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowAnalog") == 0)
-            isAnalog = true;
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowDigit") == 0)
-            isDigit = true;
-    }
+        while ((zerlegt.size() > 1) && !_done)
+        {
+            name = trim(zerlegt[0]);
+            zwtime = trim(zerlegt[1]);
+            zwvalue = trim(zerlegt[2]);
 
-    if (isDigit || isAnalog)
+            for (int j = 0; j < NUMBERS.size(); ++j)
+            {
+                if (NUMBERS[j]->name == name)
+                {
+                    NUMBERS[j]->PreValue = stof(zwvalue.c_str());
+                    NUMBERS[j]->ReturnPreValue = RundeOutput(NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma);
+
+                    time_t tStart;
+                    int yy, month, dd, hh, mm, ss;
+                    struct tm whenStart;
+
+                    sscanf(zwtime.c_str(), PREVALUE_TIME_FORMAT_INPUT, &yy, &month, &dd, &hh, &mm, &ss);
+                    whenStart.tm_year = yy - 1900;
+                    whenStart.tm_mon = month - 1;
+                    whenStart.tm_mday = dd;
+                    whenStart.tm_hour = hh;
+                    whenStart.tm_min = mm;
+                    whenStart.tm_sec = ss;
+                    whenStart.tm_isdst = -1;
+
+                    NUMBERS[j]->lastvalue = mktime(&whenStart);
+
+                    time(&tStart);
+                    localtime(&tStart);
+                    double difference = difftime(tStart, NUMBERS[j]->lastvalue);
+                    difference /= 60;
+                    if (difference > PreValueAgeStartup)
+                    {
+                        NUMBERS[j]->PreValueOkay = false;
+                    }
+                    else
+                    {
+                        NUMBERS[j]->PreValueOkay = true;
+                        NUMBERS[j]->Value = NUMBERS[j]->PreValue;
+                        NUMBERS[j]->ReturnValue = to_string(NUMBERS[j]->Value);
+                        NUMBERS[j]->ReturnValueNoError = NUMBERS[j]->ReturnValue; 
+
+                        if (NUMBERS[j]->digit_roi || NUMBERS[j]->analog_roi)
+                        {
+                            NUMBERS[j]->ReturnValue = RundeOutput(NUMBERS[j]->Value, NUMBERS[j]->AnzahlAnalog - NUMBERS[j]->DecimalShift);
+                            NUMBERS[j]->ReturnValueNoError = NUMBERS[j]->ReturnValue;
+                        }
+                    }
+
+                }
+            }
+
+            if (!fgets(zw, 1024, pFile))
+                _done = true;
+            else
+            {
+                printf("Read Zeile Prevalue.ini: %s", zw);
+                zerlegt = HelperZerlegeZeile(trim(std::string(zw)), "\t");
+                if (zerlegt.size() > 1)
+                {
+                    name = trim(zerlegt[0]);
+                    zwtime = trim(zerlegt[1]);
+                    zwvalue = trim(zerlegt[2]);
+                }
+            }
+        }
+        fclose(pFile);
+    }   
+    else        // altes Format
     {
-        ReturnValue = RundeOutput(Value, AnzahlAnalog - DecimalShift);
-        ReturnValueNoError = ReturnValue;
-    }
-   
+        fgets(zw, 1024, pFile);
+        fclose(pFile);
+        printf("%s", zw);
+        zwvalue = trim(std::string(zw));
+        NUMBERS[0]->PreValue = stof(zwvalue.c_str());
+
+        time_t tStart;
+        int yy, month, dd, hh, mm, ss;
+        struct tm whenStart;
+
+        sscanf(zwtime.c_str(), PREVALUE_TIME_FORMAT_INPUT, &yy, &month, &dd, &hh, &mm, &ss);
+        whenStart.tm_year = yy - 1900;
+        whenStart.tm_mon = month - 1;
+        whenStart.tm_mday = dd;
+        whenStart.tm_hour = hh;
+        whenStart.tm_min = mm;
+        whenStart.tm_sec = ss;
+        whenStart.tm_isdst = -1;
+
+        printf("TIME: %d, %d, %d, %d, %d, %d\n", whenStart.tm_year, whenStart.tm_mon, whenStart.tm_wday, whenStart.tm_hour, whenStart.tm_min, whenStart.tm_sec);
+
+        NUMBERS[0]->lastvalue = mktime(&whenStart);
+
+        time(&tStart);
+        localtime(&tStart);
+        double difference = difftime(tStart, NUMBERS[0]->lastvalue);
+        difference /= 60;
+        if (difference > PreValueAgeStartup)
+            return false;
+
+        NUMBERS[0]->Value = NUMBERS[0]->PreValue;
+        NUMBERS[0]->ReturnValue = to_string(NUMBERS[0]->Value);
+        NUMBERS[0]->ReturnValueNoError = NUMBERS[0]->ReturnValue; 
+
+        if (NUMBERS[0]->digit_roi || NUMBERS[0]->analog_roi)
+        {
+            NUMBERS[0]->ReturnValue = RundeOutput(NUMBERS[0]->Value, NUMBERS[0]->AnzahlAnalog - NUMBERS[0]->DecimalShift);
+            NUMBERS[0]->ReturnValueNoError = NUMBERS[0]->ReturnValue;
+        }
+
+        UpdatePreValueINI = true;       // Konvertierung ins neue Format
+        SavePreValue();
+    } 
+
     return true;
 }
 
-void ClassFlowPostProcessing::SavePreValue(float value, string zwtime)
+void ClassFlowPostProcessing::SavePreValue()
 {
     FILE* pFile;
+    string _zw;
+
+    if (!UpdatePreValueINI)         // PreValues unverändert --> File muss nicht neu geschrieben werden
+        return;
 
     pFile = fopen(FilePreValue.c_str(), "w");
 
-    if (strlen(zwtime.c_str()) == 0)
+    for (int j = 0; j < NUMBERS.size(); ++j)
     {
-        time_t rawtime;
-        struct tm* timeinfo;
         char buffer[80];
+        struct tm* timeinfo = localtime(&NUMBERS[j]->lastvalue);
+        strftime(buffer, 80, PREVALUE_TIME_FORMAT_OUTPUT, timeinfo);
+        NUMBERS[j]->timeStamp = std::string(buffer);
+//        printf("SaverPreValue %d, Value: %f, Nachkomma %d\n", j, NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma);
 
-        time(&rawtime);
-        timeinfo = localtime(&rawtime);
+        _zw = NUMBERS[j]->name + "\t" + NUMBERS[j]->timeStamp + "\t" + RundeOutput(NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma) + "\n";
+        printf("Write PreValue Zeile: %s\n", _zw.c_str());
 
-        strftime(buffer, 80, "%Y-%m-%d_%H-%M-%S", timeinfo);
-        zwtime = std::string(buffer);
+        fputs(_zw.c_str(), pFile);
     }
 
-    PreValue = value;
-
-    fputs(zwtime.c_str(), pFile);
-    fputs("\n", pFile);
-    fputs(to_string(value).c_str(), pFile);
-    fputs("\n", pFile);
+    UpdatePreValueINI = false;
 
     fclose(pFile);
 }
@@ -139,25 +233,105 @@ void ClassFlowPostProcessing::SavePreValue(float value, string zwtime)
 
 ClassFlowPostProcessing::ClassFlowPostProcessing(std::vector<ClassFlow*>* lfc)
 {
+//    FlowRateAct = 0;
     PreValueUse = false;
     PreValueAgeStartup = 30;
-    AllowNegativeRates = false;
-    MaxRateValue = 0.1;
     ErrorMessage = false;
     ListFlowControll = NULL;
-    PreValueOkay = false;
-    useMaxRateValue = false;
-    checkDigitIncreaseConsistency = false;
-    DecimalShift = 0;    
-    ErrorMessageText = "";
+//    PreValueOkay = false;
+//    DecimalShift = 0;    
+//    ErrorMessageText = "";
+//    timeStamp = "";
     FilePreValue = FormatFileName("/sdcard/config/prevalue.ini");
     ListFlowControll = lfc;
+    flowMakeImage = NULL;
+    UpdatePreValueINI = false;
+
+    for (int i = 0; i < ListFlowControll->size(); ++i)
+    {
+        if (((*ListFlowControll)[i])->name().compare("ClassFlowMakeImage") == 0)
+        {
+            flowMakeImage = (ClassFlowMakeImage*) (*ListFlowControll)[i];
+        }
+    }
+}
+
+void ClassFlowPostProcessing::handleDecimalSeparator(string _decsep, string _value)
+{
+    string _digit, _decpos;
+    int _pospunkt = _decsep.find_first_of(".");
+//    printf("Name: %s, Pospunkt: %d\n", _decsep.c_str(), _pospunkt);
+    if (_pospunkt > -1)
+        _digit = _decsep.substr(0, _pospunkt);
+    else
+        _digit = "default";
+
+    for (int j = 0; j < NUMBERS.size(); ++j)
+    {
+        int _zwdc = 0;
+
+        try
+        {
+            _zwdc = stoi(_value);
+        }
+        catch(const std::exception& e)
+        {
+            printf("ERROR - Decimalshift is not a number: %s\n", _value.c_str());
+        }
+        
+        if (_digit == "default")                        // erstmal auf default setzen (falls sonst nichts gesetzt)
+            NUMBERS[j]->DecimalShift = _zwdc;
+
+        if (NUMBERS[j]->name == _digit)
+            NUMBERS[j]->DecimalShift = _zwdc;
+
+        NUMBERS[j]->Nachkomma = NUMBERS[j]->AnzahlAnalog - NUMBERS[j]->DecimalShift;
+    }
+}
+
+void ClassFlowPostProcessing::handleMaxRateValue(string _decsep, string _value)
+{
+    string _digit, _decpos;
+    int _pospunkt = _decsep.find_first_of(".");
+//    printf("Name: %s, Pospunkt: %d\n", _decsep.c_str(), _pospunkt);
+    if (_pospunkt > -1)
+        _digit = _decsep.substr(0, _pospunkt);
+    else
+        _digit = "default";
+
+    for (int j = 0; j < NUMBERS.size(); ++j)
+    {
+        float _zwdc = 1;
+
+        try
+        {
+            _zwdc = stof(_value);
+        }
+        catch(const std::exception& e)
+        {
+            printf("ERROR - MaxRateValue is not a number: %s\n", _value.c_str());
+        }
+
+
+        if (_digit == "default")                        // erstmal auf default setzen (falls sonst nichts gesetzt)
+        {
+            NUMBERS[j]->useMaxRateValue = true;
+            NUMBERS[j]->MaxRateValue = _zwdc;
+        }
+
+        if (NUMBERS[j]->name == _digit)
+        {
+            NUMBERS[j]->useMaxRateValue = true;
+            NUMBERS[j]->MaxRateValue = _zwdc;
+        }
+    }
 }
 
 
 bool ClassFlowPostProcessing::ReadParameter(FILE* pfile, string& aktparamgraph)
 {
     std::vector<string> zerlegt;
+    int _n;
 
     aktparamgraph = trim(aktparamgraph);
 
@@ -169,53 +343,148 @@ bool ClassFlowPostProcessing::ReadParameter(FILE* pfile, string& aktparamgraph)
     if (aktparamgraph.compare("[PostProcessing]") != 0)       // Paragraph passt nich zu MakeImage
         return false;
 
+    InitNUMBERS();
+
+
     while (this->getNextLine(pfile, &aktparamgraph) && !this->isNewParagraph(aktparamgraph))
     {
         zerlegt = this->ZerlegeZeile(aktparamgraph);
-        if ((toUpper(zerlegt[0]) == "DECIMALSHIFT") && (zerlegt.size() > 1))
+        std::string _param = GetParameterName(zerlegt[0]);
+
+        if ((toUpper(_param) == "DECIMALSHIFT") && (zerlegt.size() > 1))
         {
-            DecimalShift = stoi(zerlegt[1]);
+            handleDecimalSeparator(zerlegt[0], zerlegt[1]);
+        }
+        if ((toUpper(_param) == "MAXRATEVALUE") && (zerlegt.size() > 1))
+        {
+            handleMaxRateValue(zerlegt[0], zerlegt[1]);
         }
 
-        if ((toUpper(zerlegt[0]) == "PREVALUEUSE") && (zerlegt.size() > 1))
+        if ((toUpper(_param) == "PREVALUEUSE") && (zerlegt.size() > 1))
         {
             if (toUpper(zerlegt[1]) == "TRUE")
             {
                 PreValueUse = true;
             }
         }
-        if ((toUpper(zerlegt[0]) == "CHECKDIGITINCREASECONSISTENCY") && (zerlegt.size() > 1))
+        if ((toUpper(_param) == "CHECKDIGITINCREASECONSISTENCY") && (zerlegt.size() > 1))
         {
             if (toUpper(zerlegt[1]) == "TRUE")
-                checkDigitIncreaseConsistency = true;
+                for (_n = 0; _n < NUMBERS.size(); ++_n)
+                    NUMBERS[_n]->checkDigitIncreaseConsistency = true;
         }        
-        if ((toUpper(zerlegt[0]) == "ALLOWNEGATIVERATES") && (zerlegt.size() > 1))
+        if ((toUpper(_param) == "ALLOWNEGATIVERATES") && (zerlegt.size() > 1))
         {
             if (toUpper(zerlegt[1]) == "TRUE")
-                AllowNegativeRates = true;
+                for (_n = 0; _n < NUMBERS.size(); ++_n)
+                    NUMBERS[_n]->AllowNegativeRates = true;
         }
-        if ((toUpper(zerlegt[0]) == "ERRORMESSAGE") && (zerlegt.size() > 1))
+        if ((toUpper(_param) == "ERRORMESSAGE") && (zerlegt.size() > 1))
         {
             if (toUpper(zerlegt[1]) == "TRUE")
                 ErrorMessage = true;
         }
-        if ((toUpper(zerlegt[0]) == "PREVALUEAGESTARTUP") && (zerlegt.size() > 1))
+        if ((toUpper(_param) == "PREVALUEAGESTARTUP") && (zerlegt.size() > 1))
         {
             PreValueAgeStartup = std::stoi(zerlegt[1]);
         }
-        if ((toUpper(zerlegt[0]) == "MAXRATEVALUE") && (zerlegt.size() > 1))
-        {
-            useMaxRateValue = true;
-            MaxRateValue = std::stof(zerlegt[1]);
-        }
     }
 
     if (PreValueUse) {
-        PreValueOkay = LoadPreValue();
+        LoadPreValue();
     }
+
     return true;
 }
 
+void ClassFlowPostProcessing::InitNUMBERS()
+{
+//    ClassFlowDigit* _cdigit = NULL;
+//    ClassFlowAnalog* _canalog = NULL;
+    int anzDIGIT = 0;
+    int anzANALOG = 0;
+    std::vector<std::string> name_numbers;
+
+    flowAnalog = NULL;
+    flowDigit = NULL;
+
+    for (int i = 0; i < ListFlowControll->size(); ++i)
+    {
+        if (((*ListFlowControll)[i])->name().compare("ClassFlowDigit") == 0)
+        {
+            flowDigit = (ClassFlowDigit*) (*ListFlowControll)[i];
+            anzDIGIT = flowDigit->getAnzahlDIGIT();
+        }
+        if (((*ListFlowControll)[i])->name().compare("ClassFlowAnalog") == 0)
+        {
+            flowAnalog = (ClassFlowAnalog*)(*ListFlowControll)[i];
+            anzANALOG = flowAnalog->getAnzahlANALOG();
+        }
+    }
+
+    if (flowDigit)
+        flowDigit->UpdateNameNumbers(&name_numbers);
+    if (flowAnalog)
+        flowAnalog->UpdateNameNumbers(&name_numbers);
+
+    printf("Anzahl NUMBERS: %d - DIGITS: %d, ANALOG: %d\n", name_numbers.size(), anzDIGIT, anzANALOG);
+
+    for (int _num = 0; _num < name_numbers.size(); ++_num)
+    {
+        NumberPost *_number = new NumberPost;
+
+        _number->name = name_numbers[_num];
+        
+        _number->digit_roi = NULL;
+        if (flowDigit)
+            _number->digit_roi = flowDigit->FindDIGIT(name_numbers[_num]);
+        
+        if (_number->digit_roi)
+            _number->AnzahlDigital = _number->digit_roi->ROI.size();
+        else
+            _number->AnzahlDigital = 0;
+
+        _number->analog_roi = NULL;
+        if (flowAnalog)
+            _number->analog_roi = flowAnalog->FindANALOG(name_numbers[_num]);
+
+
+        if (_number->analog_roi)
+            _number->AnzahlAnalog = _number->analog_roi->ROI.size();
+        else
+            _number->AnzahlAnalog = 0;
+
+        _number->ReturnRawValue = "";      // Rohwert (mit N & führenden 0)    
+        _number->ReturnValue = "";         // korrigierter Rückgabewert, ggf. mit Fehlermeldung
+        _number->ReturnValueNoError = "";  // korrigierter Rückgabewert ohne Fehlermeldung
+        _number->ErrorMessageText = "";        // Fehlermeldung bei Consistency Check
+        _number->ReturnPreValue = "";
+        _number->PreValueOkay = false;
+        _number->AllowNegativeRates = false;
+        _number->MaxRateValue = 0.1;
+        _number->useMaxRateValue = false;
+        _number->checkDigitIncreaseConsistency = false;
+        _number->PreValueOkay = false;
+        _number->useMaxRateValue = false;
+        _number->DecimalShift = 0;
+
+        _number->FlowRateAct = 0;          // m3 / min
+        _number->PreValue = 0;             // letzter Wert, der gut ausgelesen wurde
+        _number->Value = 0;                // letzer ausgelesener Wert, inkl. Korrekturen
+        _number->ReturnRawValue = "";      // Rohwert (mit N & führenden 0)    
+        _number->ReturnValue = "";         // korrigierter Rückgabewert, ggf. mit Fehlermeldung
+        _number->ReturnValueNoError = "";  // korrigierter Rückgabewert ohne Fehlermeldung
+        _number->ErrorMessageText = "";        // Fehlermeldung bei Consistency Check
+
+        _number->Nachkomma = _number->AnzahlAnalog;
+
+        NUMBERS.push_back(_number);
+    }
+
+    for (int i = 0; i < NUMBERS.size(); ++i)
+        printf("Number %s, Anz DIG: %d, Anz ANA %d\n", NUMBERS[i]->name.c_str(), NUMBERS[i]->AnzahlDigital, NUMBERS[i]->AnzahlAnalog);
+}
+
 string ClassFlowPostProcessing::ShiftDecimal(string in, int _decShift){
 
     if (_decShift == 0){
@@ -264,142 +533,127 @@ bool ClassFlowPostProcessing::doFlow(string zwtime)
     string digit = "";
     string analog = "";
     string zwvalue;
-    bool isdigit = false;
-    bool isanalog = false;
-    int AnzahlAnalog = 0;
     string zw;
     time_t imagetime = 0;
     string rohwert;
 
-    ErrorMessageText = "";
-
-
-    for (int i = 0; i < ListFlowControll->size(); ++i)
-    {
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowMakeImage") == 0)
-        {
-            imagetime = ((ClassFlowMakeImage*)(*ListFlowControll)[i])->getTimeImageTaken();
-        }
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowDigit") == 0)
-        {
-            isdigit = true;
-            digit = (*ListFlowControll)[i]->getReadout();
-        }
-        if (((*ListFlowControll)[i])->name().compare("ClassFlowAnalog") == 0)
-        {
-            isanalog = true;
-            analog = (*ListFlowControll)[i]->getReadout();
-            AnzahlAnalog = ((ClassFlowAnalog*)(*ListFlowControll)[i])->AnzahlROIs();
-        }
-    }
+//    ErrorMessageText = "";
 
+    imagetime = flowMakeImage->getTimeImageTaken();
     if (imagetime == 0)
         time(&imagetime);
 
     struct tm* timeinfo;
     timeinfo = localtime(&imagetime);
-
     char strftime_buf[64];
-    strftime(strftime_buf, sizeof(strftime_buf), "%Y-%m-%d_%H-%M-%S", timeinfo);
+    strftime(strftime_buf, sizeof(strftime_buf), "%Y-%m-%dT%H:%M:%S", timeinfo);
     zwtime = std::string(strftime_buf);
 
+    printf("Anzahl NUMBERS: %d\n", NUMBERS.size());
 
-    //    // TESTING ONLY////////////////////
-    //    isdigit = true; digit = "12N";
-    //    isanalog = true; analog = "456";
-
-    ReturnRawValue = "";
-
-    if (isdigit)
-        ReturnRawValue = digit;
-    if (isdigit && isanalog)
-        ReturnRawValue = ReturnRawValue + ".";
-    if (isanalog)
-        ReturnRawValue = ReturnRawValue + analog; 
-
-
-    if (!isdigit)
+    for (int j = 0; j < NUMBERS.size(); ++j)
     {
-        AnzahlAnalog = 0;
-    }
+        NUMBERS[j]->ReturnRawValue = "";
+        NUMBERS[j]->ErrorMessageText = "";
 
-    ReturnRawValue = ShiftDecimal(ReturnRawValue, DecimalShift);   
+        if (NUMBERS[j]->digit_roi)
+            NUMBERS[j]->ReturnRawValue = flowDigit->getReadout(j);
+        if (NUMBERS[j]->digit_roi && NUMBERS[j]->analog_roi)
+            NUMBERS[j]->ReturnRawValue = NUMBERS[j]->ReturnRawValue + ".";
+        if (NUMBERS[j]->analog_roi)
+            NUMBERS[j]->ReturnRawValue = NUMBERS[j]->ReturnRawValue + flowAnalog->getReadout(j); 
 
-    rohwert = ReturnRawValue;
+        NUMBERS[j]->ReturnRawValue = ShiftDecimal(NUMBERS[j]->ReturnRawValue, NUMBERS[j]->DecimalShift);   
 
-    if (!PreValueUse || !PreValueOkay)
-    {
-        ReturnValue = ReturnRawValue;
-        ReturnValueNoError = ReturnRawValue;
+        rohwert = NUMBERS[j]->ReturnRawValue;
 
-        if ((findDelimiterPos(ReturnValue, "N") == std::string::npos) && (ReturnValue.length() > 0))
+        if (!PreValueUse || !NUMBERS[j]->PreValueOkay)
         {
-            while ((ReturnValue.length() > 1) && (ReturnValue[0] == '0'))
+            NUMBERS[j]->ReturnValue = NUMBERS[j]->ReturnRawValue;
+            NUMBERS[j]->ReturnValueNoError = NUMBERS[j]->ReturnRawValue;
+
+            if ((findDelimiterPos(NUMBERS[j]->ReturnValue, "N") == std::string::npos) && (NUMBERS[j]->ReturnValue.length() > 0))
             {
-                ReturnValue.erase(0, 1);
+                while ((NUMBERS[j]->ReturnValue.length() > 1) && (NUMBERS[j]->ReturnValue[0] == '0'))
+                {
+                    NUMBERS[j]->ReturnValue.erase(0, 1);
+                }
+                NUMBERS[j]->Value = std::stof(NUMBERS[j]->ReturnValue);
+                NUMBERS[j]->ReturnValueNoError = NUMBERS[j]->ReturnValue;
+
+                NUMBERS[j]->PreValueOkay = true;
+                NUMBERS[j]->PreValue = NUMBERS[j]->Value;
+                NUMBERS[j]->ReturnPreValue = RundeOutput(NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma);
+                NUMBERS[j]->lastvalue = flowMakeImage->getTimeImageTaken();
+                zwtime = ConvertTimeToString(NUMBERS[j]->lastvalue, PREVALUE_TIME_FORMAT_OUTPUT);
+
+                UpdatePreValueINI = true;
+                SavePreValue();
+            }
+        }
+        else
+        {
+            zw = ErsetzteN(NUMBERS[j]->ReturnRawValue, NUMBERS[j]->PreValue); 
+
+            NUMBERS[j]->Value = std::stof(zw);
+            if (NUMBERS[j]->checkDigitIncreaseConsistency)
+            {
+                NUMBERS[j]->Value = checkDigitConsistency(NUMBERS[j]->Value, NUMBERS[j]->DecimalShift, NUMBERS[j]->analog_roi != NULL, NUMBERS[j]->PreValue);
+            }
+
+            zwvalue = RundeOutput(NUMBERS[j]->Value, NUMBERS[j]->AnzahlAnalog - NUMBERS[j]->DecimalShift);
+
+            if ((!NUMBERS[j]->AllowNegativeRates) && (NUMBERS[j]->Value < NUMBERS[j]->PreValue))
+            {
+                NUMBERS[j]->ErrorMessageText = NUMBERS[j]->ErrorMessageText + "Neg. Rate - Read: " + zwvalue + " - Raw: " + NUMBERS[j]->ReturnRawValue + " - Pre: " + RundeOutput(NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma) + " "; 
+                NUMBERS[j]->Value = NUMBERS[j]->PreValue;
+                zwvalue = RundeOutput(NUMBERS[j]->Value, NUMBERS[j]->AnzahlAnalog - NUMBERS[j]->DecimalShift);
+            }
+
+            if (NUMBERS[j]->useMaxRateValue && (abs(NUMBERS[j]->Value - NUMBERS[j]->PreValue) > NUMBERS[j]->MaxRateValue))
+            {
+                NUMBERS[j]->ErrorMessageText = NUMBERS[j]->ErrorMessageText + "Rate too high - Read: " + RundeOutput(NUMBERS[j]->Value, NUMBERS[j]->Nachkomma) + " - Pre: " + RundeOutput(NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma);
+                NUMBERS[j]->Value = NUMBERS[j]->PreValue;
+                zwvalue = RundeOutput(NUMBERS[j]->Value, NUMBERS[j]->Nachkomma);
+            }
+
+            NUMBERS[j]->ReturnValueNoError = zwvalue;
+            NUMBERS[j]->ReturnValue = zwvalue;
+            if (NUMBERS[j]->ErrorMessage && (NUMBERS[j]->ErrorMessageText.length() > 0))
+                NUMBERS[j]->ReturnValue = NUMBERS[j]->ReturnValue + "\t" + NUMBERS[j]->ErrorMessageText;
+
+
+            double difference = difftime(imagetime, NUMBERS[j]->lastvalue);      // in Sekunden
+            difference /= 60;                                          // in Minuten
+            NUMBERS[j]->FlowRateAct = (NUMBERS[j]->Value - NUMBERS[j]->PreValue) / difference;
+            NUMBERS[j]->lastvalue = imagetime;
+
+            if (NUMBERS[j]->ErrorMessageText.length() == 0)
+            {
+                NUMBERS[j]->PreValue = NUMBERS[j]->Value;
+                NUMBERS[j]->ReturnPreValue = RundeOutput(NUMBERS[j]->PreValue, NUMBERS[j]->Nachkomma);
+                NUMBERS[j]->ErrorMessageText = "no error";
+                UpdatePreValueINI = true;
             }
-            Value = std::stof(ReturnValue);
-            ReturnValueNoError = ReturnValue;
-
-            PreValueOkay = true;
-            PreValue = Value;
-            
-            SavePreValue(Value, zwtime);
         }
-        return true;
     }
 
-    zw = ErsetzteN(ReturnRawValue); 
-
-    Value = std::stof(zw);
-    if (checkDigitIncreaseConsistency)
-    {
-        Value = checkDigitConsistency(Value, DecimalShift, isanalog);
-    }
-
-    zwvalue = RundeOutput(Value, AnzahlAnalog - DecimalShift);
-
-    if ((!AllowNegativeRates) && (Value < PreValue))
-    {
-        ErrorMessageText = ErrorMessageText + "Negative Rate - Returned old value - read value: " + zwvalue + " - raw value: " + ReturnRawValue + " - checked value: " + std::to_string(Value) + " "; 
-        Value = PreValue;
-        zwvalue = RundeOutput(Value, AnzahlAnalog - DecimalShift);
-    }
-
-    if (useMaxRateValue && (abs(Value - PreValue) > MaxRateValue))
-    {
-        ErrorMessageText = ErrorMessageText + "Rate too high - Returned old value - read value: " + zwvalue + " - checked value: " + RundeOutput(Value, AnzahlAnalog - DecimalShift) + " ";
-        Value = PreValue;
-        zwvalue = RundeOutput(Value, AnzahlAnalog - DecimalShift);
-    }
-
-    ReturnValueNoError = zwvalue;
-    ReturnValue = zwvalue;
-    if (ErrorMessage && (ErrorMessageText.length() > 0))
-        ReturnValue = ReturnValue + "\t" + ErrorMessageText;
-
-    if (ErrorMessageText.length() == 0)
-    {
-        PreValue = Value;
-        
-        SavePreValue(Value, zwtime);
-       
-    }
+    SavePreValue();
     return true;
 }
 
-string ClassFlowPostProcessing::getReadout()
+string ClassFlowPostProcessing::getReadout(int _number)
 {
-    return ReturnValue;
+    return NUMBERS[_number]->ReturnValue;
 }
 
-string ClassFlowPostProcessing::getReadoutParam(bool _rawValue, bool _noerror)
+string ClassFlowPostProcessing::getReadoutParam(bool _rawValue, bool _noerror, int _number)
 {
     if (_rawValue)
-        return ReturnRawValue;
+        return NUMBERS[_number]->ReturnRawValue;
     if (_noerror)
-        return ReturnValueNoError;
-    return ReturnValue;
+        return NUMBERS[_number]->ReturnValueNoError;
+    return NUMBERS[_number]->ReturnValue;
 }
 
 string ClassFlowPostProcessing::RundeOutput(float _in, int _anzNachkomma){
@@ -426,7 +680,7 @@ string ClassFlowPostProcessing::RundeOutput(float _in, int _anzNachkomma){
 }
 
 
-string ClassFlowPostProcessing::ErsetzteN(string input)
+string ClassFlowPostProcessing::ErsetzteN(string input, float _prevalue)
 {
     int posN, posPunkt;
     int pot, ziffer;
@@ -447,7 +701,7 @@ string ClassFlowPostProcessing::ErsetzteN(string input)
             pot = posPunkt - posN;
         }
 
-        zw = PreValue / pow(10, pot);
+        zw =_prevalue / pow(10, pot);
         ziffer = ((int) zw) % 10;
         input[posN] = ziffer + 48;
 
@@ -457,7 +711,7 @@ string ClassFlowPostProcessing::ErsetzteN(string input)
     return input;
 }
 
-float ClassFlowPostProcessing::checkDigitConsistency(float input, int _decilamshift, bool _isanalog){
+float ClassFlowPostProcessing::checkDigitConsistency(float input, int _decilamshift, bool _isanalog, float _preValue){
     int aktdigit, olddigit;
     int aktdigit_before, olddigit_before;
     int pot, pot_max;
@@ -475,12 +729,12 @@ float ClassFlowPostProcessing::checkDigitConsistency(float input, int _decilamsh
     {
         zw = input / pow(10, pot-1);
         aktdigit_before = ((int) zw) % 10;
-        zw = PreValue / pow(10, pot-1);
+        zw = _preValue / pow(10, pot-1);
         olddigit_before = ((int) zw) % 10;
 
         zw = input / pow(10, pot);
         aktdigit = ((int) zw) % 10;
-        zw = PreValue / pow(10, pot);
+        zw = _preValue / pow(10, pot);
         olddigit = ((int) zw) % 10;
 
         no_nulldurchgang = (olddigit_before <= aktdigit_before);
@@ -506,8 +760,18 @@ float ClassFlowPostProcessing::checkDigitConsistency(float input, int _decilamsh
     return input;
 }
 
-
-string ClassFlowPostProcessing::getReadoutError() 
+string ClassFlowPostProcessing::getReadoutRate(int _number)
 {
-    return ErrorMessageText;
+    return std::to_string(NUMBERS[_number]->FlowRateAct);
+}
+
+string ClassFlowPostProcessing::getReadoutTimeStamp(int _number)
+{
+   return NUMBERS[_number]->timeStamp; 
+}
+
+
+string ClassFlowPostProcessing::getReadoutError(int _number) 
+{
+    return NUMBERS[_number]->ErrorMessageText;
 }

code/components/jomjol_flowcontroll/ClassFlowPostProcessing.h

@@ -1,47 +1,95 @@
 #pragma once
 #include "ClassFlow.h"
+#include "ClassFlowMakeImage.h"
+#include "ClassFlowAnalog.h"
+#include "ClassFlowDigit.h"
+
 
 #include <string>
 
 
+struct NumberPost {
+//    int PreValueAgeStartup; 
+    float MaxRateValue;
+    bool useMaxRateValue;
+    bool ErrorMessage;
+    bool PreValueOkay;
+    bool AllowNegativeRates;
+    bool checkDigitIncreaseConsistency;
+    time_t lastvalue;
+    string timeStamp;
+    float FlowRateAct;          // m3 / min
+    float PreValue;             // letzter Wert, der gut ausgelesen wurde
+    float Value;                // letzer ausgelesener Wert, inkl. Korrekturen
+    string ReturnRawValue;      // Rohwert (mit N & führenden 0)    
+    string ReturnValue;         // korrigierter Rückgabewert, ggf. mit Fehlermeldung
+    string ReturnPreValue;  // korrigierter Rückgabewert ohne Fehlermeldung
+    string ReturnValueNoError;
+    string ErrorMessageText;        // Fehlermeldung bei Consistency Check
+    int AnzahlAnalog;
+    int AnzahlDigital;
+    int DecimalShift;
+    int Nachkomma;
+//    ClassFlowAnalog* ANALOG;
+//    ClassFlowDigit* DIGIT;
+
+    digit *digit_roi;
+    analog *analog_roi;
+
+
+
+    string name;
+};
+
+
+
+
 class ClassFlowPostProcessing :
     public ClassFlow
 {
 protected:
-    bool PreValueUse;
+    std::vector<NumberPost*> NUMBERS;
+    bool UpdatePreValueINI;
+
     int PreValueAgeStartup; 
-    bool AllowNegativeRates;
-    float MaxRateValue;
-    bool useMaxRateValue;
     bool ErrorMessage;
-    bool PreValueOkay;
-    bool checkDigitIncreaseConsistency;
-    int DecimalShift;
+
+
+    ClassFlowAnalog* flowAnalog;
+    ClassFlowDigit* flowDigit;    
+
 
     string FilePreValue;
-    float PreValue;             // letzter Wert, der gut ausgelesen wurde
-    float Value;                // letzer ausgelesener Wert, inkl. Korrekturen
-    string ReturnRawValue;      // Rohwert (mit N & führenden 0)    
-    string ReturnValue;         // korrigierter Rückgabewert, ggf. mit Fehlermeldung
-    string ReturnValueNoError;  // korrigierter Rückgabewert ohne Fehlermeldung
-    string ErrorMessageText;        // Fehlermeldung bei Consistency Check
+
+    ClassFlowMakeImage *flowMakeImage;
 
     bool LoadPreValue(void);
     string ShiftDecimal(string in, int _decShift);
 
-    string ErsetzteN(string);
-    float checkDigitConsistency(float input, int _decilamshift, bool _isanalog);
+    string ErsetzteN(string, float _prevalue);
+    float checkDigitConsistency(float input, int _decilamshift, bool _isanalog, float _preValue);
     string RundeOutput(float _in, int _anzNachkomma);
 
+    void InitNUMBERS();
+    void handleDecimalSeparator(string _decsep, string _value);
+    void handleMaxRateValue(string _decsep, string _value);
+
+
 public:
+    bool PreValueUse;
+
     ClassFlowPostProcessing(std::vector<ClassFlow*>* lfc);
     bool ReadParameter(FILE* pfile, string& aktparamgraph);
     bool doFlow(string time);
-    string getReadout();
-    string getReadoutParam(bool _rawValue, bool _noerror);
-    string getReadoutError();
-    void SavePreValue(float value, string time = "");
-    string GetPreValue();
+    string getReadout(int _number);
+    string getReadoutParam(bool _rawValue, bool _noerror, int _number = 0);
+    string getReadoutError(int _number = 0);
+    string getReadoutRate(int _number = 0);
+    string getReadoutTimeStamp(int _number = 0);
+    void SavePreValue();
+    string GetPreValue(std::string _number = "");
+    void SetPreValue(float zw, string _numbers, bool _extern = false);
+    std::vector<NumberPost*> GetNumbers(){return NUMBERS;};
 
     string name(){return "ClassFlowPostProcessing";};
 };

code/components/jomjol_helper/Helper.cpp

@@ -10,6 +10,7 @@
 #include <string.h>
 #include <esp_log.h>
 
+
 #include "ClassLogFile.h"
 //#include "ClassLogFile.h"
 
@@ -77,8 +78,9 @@ void memCopyGen(uint8_t* _source, uint8_t* _target, int _size)
 
 
 
-FILE* OpenFileAndWait(const char* nm, char* _mode, int _waitsec)
+FILE* OpenFileAndWait(const char* nm, const char* _mode, int _waitsec)
 {
+	printf("open config file %s in mode %s\n", nm, _mode);
 	FILE *pfile = fopen(nm, _mode);
 
 	if (pfile == NULL)
@@ -313,6 +315,14 @@ string toUpper(string in)
 	return in;
 }
 
+string toLower(string in)
+{
+	for (int i = 0; i < in.length(); ++i)
+		in[i] = tolower(in[i]);
+	
+	return in;
+}
+
 // CPU Temp
 extern "C" uint8_t temprature_sens_read();
 float temperatureRead()
@@ -358,3 +368,30 @@ int removeFolder(const char* folderPath, const char* logTag) {
 
 	return deleted;
 }
+
+
+
+std::vector<string> HelperZerlegeZeile(std::string input, std::string _delimiter = "")
+{
+	std::vector<string> Output;
+	std::string delimiter = " =,";
+    if (_delimiter.length() > 0){
+        delimiter = _delimiter;
+    }
+
+	input = trim(input, delimiter);
+	size_t pos = findDelimiterPos(input, delimiter);
+	std::string token;
+	while (pos != std::string::npos) {
+		token = input.substr(0, pos);
+		token = trim(token, delimiter);
+		Output.push_back(token);
+		input.erase(0, pos + 1);
+		input = trim(input, delimiter);
+		pos = findDelimiterPos(input, delimiter);
+	}
+	Output.push_back(input);
+
+	return Output;
+}
+

code/components/jomjol_helper/Helper.h

@@ -1,6 +1,7 @@
 #pragma once
 #include <string>
 #include <fstream>
+#include <vector>
 
 
 using namespace std;
@@ -10,7 +11,7 @@ void FindReplace(std::string& line, std::string& oldString, std::string& newStri
 
 void CopyFile(string input, string output);
 
-FILE* OpenFileAndWait(const char* nm, char* _mode, int _waitsec = 1);
+FILE* OpenFileAndWait(const char* nm, const char* _mode, int _waitsec = 1);
 
 size_t findDelimiterPos(string input, string delimiter);
 //string trim(string istring);
@@ -22,6 +23,7 @@ string getFileType(string filename);
 int mkdir_r(const char *dir, const mode_t mode);
 int removeFolder(const char* folderPath, const char* logTag);
 
+string toLower(string in);
 string toUpper(string in);
 
 float temperatureRead();
@@ -30,6 +32,8 @@ time_t addDays(time_t startTime, int days);
 
 void memCopyGen(uint8_t* _source, uint8_t* _target, int _size);
 
+std::vector<string> HelperZerlegeZeile(std::string input, std::string _delimiter);
+
 ///////////////////////////
 size_t getInternalESPHeapSize();
 size_t getESPHeapSize();

code/components/jomjol_image_proc/CImageBasis.cpp

@@ -354,12 +354,10 @@ CImageBasis::CImageBasis(CImageBasis *_copyfrom, int _anzrepeat)
     int memsize = width * height * channels;
     rgb_image = (unsigned char*)GET_MEMORY(memsize);
 
-    TickType_t xDelay;
     int anz = 1;
     while (!rgb_image && (anz < _anzrepeat))    
     {
-		    printf("Create Image from Copy - Speicher ist voll - Versuche es erneut: %d.\n", anz);
-        xDelay = 1000 / portTICK_PERIOD_MS;
+	    printf("Create Image from Copy - Speicher ist voll - Versuche es erneut: %d.\n", anz);
         rgb_image = (unsigned char*) malloc(memsize);
         anz++;
     }

code/components/jomjol_image_proc/CRotateImage.cpp

@@ -1,7 +1,7 @@
 #include "CRotateImage.h"
 
 
-CRotateImage::CRotateImage(CImageBasis *_org, CImageBasis *_temp)
+CRotateImage::CRotateImage(CImageBasis *_org, CImageBasis *_temp, bool _flip)
 {
     rgb_image = _org->rgb_image;
     channels = _org->channels;
@@ -9,8 +9,10 @@ CRotateImage::CRotateImage(CImageBasis *_org, CImageBasis *_temp)
     height = _org->height;
     bpp = _org->bpp;
     externalImage = true;   
-    ImageTMP = _temp;    
+    ImageTMP = _temp;   
+    ImageOrg = _org; 
     islocked = false;
+    doflip = _flip;
 }
 
 void CRotateImage::Mirror(){
@@ -58,12 +60,33 @@ void CRotateImage::Mirror(){
 
 void CRotateImage::Rotate(float _angle, int _centerx, int _centery)
 {
+    int org_width, org_height;
     float m[2][3];
 
     float x_center = _centerx;
     float y_center = _centery;
     _angle = _angle / 180 * M_PI;
 
+    if (doflip)
+    {
+        org_width = width;
+        org_height = height;
+        height = org_width;
+        width = org_height;
+        x_center =  x_center - (org_width/2) + (org_height/2);
+        y_center =  y_center + (org_width/2) - (org_height/2);
+        if (ImageOrg)
+        {
+            ImageOrg->height = height;
+            ImageOrg->width = width;
+        }
+    }
+    else
+    {
+        org_width = width;
+        org_height = height;
+    }
+
     m[0][0] = cos(_angle);
     m[0][1] = sin(_angle);
     m[0][2] = (1 - m[0][0]) * x_center - m[0][1] * y_center;
@@ -72,6 +95,12 @@ void CRotateImage::Rotate(float _angle, int _centerx, int _centery)
     m[1][1] = m[0][0];
     m[1][2] = m[0][1] * x_center + (1 - m[0][0]) * y_center;
 
+    if (doflip)
+    {
+        m[0][2] = m[0][2] + (org_width/2) - (org_height/2);
+        m[1][2] = m[1][2] - (org_width/2) + (org_height/2);
+    }
+
     int memsize = width * height * channels;
     uint8_t* odata;
     if (ImageTMP)
@@ -101,9 +130,9 @@ void CRotateImage::Rotate(float _angle, int _centerx, int _centery)
             x_source += int(m[0][2]);
             y_source += int(m[1][2]);
 
-            if ((x_source >= 0) && (x_source < width) && (y_source >= 0) && (y_source < height))
+            if ((x_source >= 0) && (x_source < org_width) && (y_source >= 0) && (y_source < org_height))
             {
-                p_source = rgb_image + (channels * (y_source * width + x_source));
+                p_source = rgb_image + (channels * (y_source * org_width + x_source));
                 for (int _channels = 0; _channels < channels; ++_channels)
                     p_target[_channels] = p_source[_channels];
             }

code/components/jomjol_image_proc/CRotateImage.h

@@ -4,10 +4,11 @@
 class CRotateImage: public CImageBasis
 {
     public:
-        CImageBasis *ImageTMP;
-        CRotateImage(std::string _image) : CImageBasis(_image) {ImageTMP = NULL;};
-        CRotateImage(uint8_t* _rgb_image, int _channels, int _width, int _height, int _bpp) : CImageBasis(_rgb_image, _channels, _width, _height, _bpp) {ImageTMP = NULL;};
-        CRotateImage(CImageBasis *_org, CImageBasis *_temp);
+        CImageBasis *ImageTMP, *ImageOrg;
+        bool doflip;
+        CRotateImage(std::string _image, bool _flip = false) : CImageBasis(_image) {ImageTMP = NULL; ImageOrg = NULL; doflip = _flip;};
+        CRotateImage(uint8_t* _rgb_image, int _channels, int _width, int _height, int _bpp, bool _flip = false) : CImageBasis(_rgb_image, _channels, _width, _height, _bpp) {ImageTMP = NULL;  ImageOrg = NULL; doflip = _flip;};
+        CRotateImage(CImageBasis *_org, CImageBasis *_temp, bool _flip = false);
 
         void Rotate(float _angle);
         void Rotate(float _angle, int _centerx, int _centery);

code/components/jomjol_logfile/ClassLogFile.cpp

@@ -12,7 +12,7 @@ ClassLogFile LogFile("/sdcard/log/message", "log_%Y-%m-%d.txt");
 
 void ClassLogFile::WriteHeapInfo(std::string _id)
 {
-std::string _zw = "\t" + _id;
+    std::string _zw = "\t" + _id;
     if (loglevel > 0)
         _zw =  _zw + "\t" + getESPHeapInfo();
 
@@ -77,7 +77,7 @@ void ClassLogFile::WriteToDedicatedFile(std::string _fn, std::string info, bool
             time(&rawtime);
             timeinfo = localtime(&rawtime);
 
-            strftime(buffer, 80, "%Y-%m-%d_%H-%M-%S", timeinfo);
+            strftime(buffer, 80, "%Y-%m-%dT%H:%M:%S", timeinfo);
 
             zwtime = std::string(buffer);
             info = zwtime + ": " + info;

code/components/jomjol_mqtt/interface_mqtt.cpp

@@ -1,12 +1,14 @@
 #include "interface_mqtt.h"
 
-
+//#define LOG_LOCAL_LEVEL ESP_LOG_DEBUG
 #include "esp_log.h"
 #include "mqtt_client.h"
 #include "ClassLogFile.h"
 
-static const char *TAG = "interface_mqtt";
+static const char *TAG_INTERFACEMQTT = "interface_mqtt";
 
+std::map<std::string, std::function<void()>>* connectFunktionMap = NULL;  
+std::map<std::string, std::function<bool(std::string, char*, int)>>* subscribeFunktionMap = NULL;  
 bool debugdetail = true;
 
 // #define CONFIG_BROKER_URL "mqtt://192.168.178.43:1883"
@@ -16,51 +18,74 @@ esp_mqtt_event_id_t esp_mmqtt_ID = MQTT_EVENT_ANY;
 bool mqtt_connected = false;
 esp_mqtt_client_handle_t client = NULL;
 
-void MQTTPublish(std::string _key, std::string _content){
+void MQTTPublish(std::string _key, std::string _content, int retained_flag){
     if (client && mqtt_connected) {
         int msg_id;
         std::string zw;
-        msg_id = esp_mqtt_client_publish(client, _key.c_str(), _content.c_str(), 0, 1, 0);
+        msg_id = esp_mqtt_client_publish(client, _key.c_str(), _content.c_str(), 0, 1, retained_flag);
         zw = "sent publish successful in MQTTPublish, msg_id=" + std::to_string(msg_id) + ", " + _key + ", " + _content;
         if (debugdetail) LogFile.WriteToFile(zw);
-        ESP_LOGI(TAG, "sent publish successful in MQTTPublish, msg_id=%d, %s, %s", msg_id, _key.c_str(), _content.c_str());
+        ESP_LOGD(TAG_INTERFACEMQTT, "sent publish successful in MQTTPublish, msg_id=%d, %s, %s", msg_id, _key.c_str(), _content.c_str());
     }
     else {
-        ESP_LOGI(TAG, "Problem with Publish, client=%d, mqtt_connected %d", (int) client, (int) mqtt_connected);
+        ESP_LOGW(TAG_INTERFACEMQTT, "Problem with Publish, client=%d, mqtt_connected %d", (int) client, (int) mqtt_connected);
     }
 }
 
 
 static esp_err_t mqtt_event_handler_cb(esp_mqtt_event_handle_t event)
 {
+    int msg_id;
+    std::string topic = "";
     switch (event->event_id) {
+        case MQTT_EVENT_BEFORE_CONNECT:
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_BEFORE_CONNECT");
+            break;
         case MQTT_EVENT_CONNECTED:
-            ESP_LOGI(TAG, "MQTT_EVENT_CONNECTED");
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_CONNECTED");
             mqtt_connected = true;
+            MQTTconnected();
             break;
         case MQTT_EVENT_DISCONNECTED:
-            ESP_LOGI(TAG, "MQTT_EVENT_DISCONNECTED");
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_DISCONNECTED");
+            break;
+        case MQTT_EVENT_SUBSCRIBED:
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_SUBSCRIBED, msg_id=%d", event->msg_id);
+            msg_id = esp_mqtt_client_publish(client, "/topic/qos0", "data", 0, 0, 0);
+            ESP_LOGI(TAG_INTERFACEMQTT, "sent publish successful, msg_id=%d", msg_id);
+            break;
+        case MQTT_EVENT_UNSUBSCRIBED:
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_UNSUBSCRIBED, msg_id=%d", event->msg_id);
             break;
         case MQTT_EVENT_PUBLISHED:
-            ESP_LOGI(TAG, "MQTT_EVENT_PUBLISHED, msg_id=%d", event->msg_id);
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_PUBLISHED, msg_id=%d", event->msg_id);
             break;
         case MQTT_EVENT_DATA:
-            ESP_LOGI(TAG, "MQTT_EVENT_DATA");
-            printf("TOPIC=%.*s\r\n", event->topic_len, event->topic);
-            printf("DATA=%.*s\r\n", event->data_len, event->data);
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_DATA");
+            ESP_LOGI(TAG_INTERFACEMQTT, "TOPIC=%.*s\r\n", event->topic_len, event->topic);
+            ESP_LOGI(TAG_INTERFACEMQTT, "DATA=%.*s\r\n", event->data_len, event->data);
+            topic.assign(event->topic, event->topic_len);
+            if (subscribeFunktionMap != NULL) {
+                if (subscribeFunktionMap->find(topic) != subscribeFunktionMap->end()) {
+                    ESP_LOGD(TAG_INTERFACEMQTT, "call handler function\r\n");
+                    (*subscribeFunktionMap)[topic](topic, event->data, event->data_len);
+                }
+            } else {
+                ESP_LOGW(TAG_INTERFACEMQTT, "no handler available\r\n");
+            }
             break;
         case MQTT_EVENT_ERROR:
-            ESP_LOGI(TAG, "MQTT_EVENT_ERROR");
+            ESP_LOGI(TAG_INTERFACEMQTT, "MQTT_EVENT_ERROR");
             break;
         default:
-            ESP_LOGI(TAG, "Other event id:%d", event->event_id);
+            ESP_LOGI(TAG_INTERFACEMQTT, "Other event id:%d", event->event_id);
             break;
     }
     return ESP_OK;
 }
 
 static void mqtt_event_handler(void *handler_args, esp_event_base_t base, int32_t event_id, void *event_data) {
-    ESP_LOGD(TAG, "Event dispatched from event loop base=%s, event_id=%d", base, event_id);
+    ESP_LOGD(TAG_INTERFACEMQTT, "Event dispatched from event loop base=%s, event_id=%d", base, event_id);
     mqtt_event_handler_cb((esp_mqtt_event_handle_t) event_data);
 }
 
@@ -74,6 +99,7 @@ void MQTTInit(std::string _mqttURI, std::string _clientid, std::string _user, st
         .client_id = _clientid.c_str(),
         .lwt_topic = _LWTContext.c_str(),
         .lwt_msg = _zwmessage.c_str(),
+        .lwt_retain = 1,
         .lwt_msg_len = _lzw,
         .keepalive = _keepalive
     };
@@ -81,12 +107,100 @@ void MQTTInit(std::string _mqttURI, std::string _clientid, std::string _user, st
     if (_user.length() && _password.length()){
         mqtt_cfg.username = _user.c_str();
         mqtt_cfg.password = _password.c_str();
-        printf("Connect to MQTT: %s, %s", mqtt_cfg.username, mqtt_cfg.password);
+        ESP_LOGI(TAG_INTERFACEMQTT, "Connect to MQTT: %s, %s", mqtt_cfg.username, mqtt_cfg.password);
     };
 
     client = esp_mqtt_client_init(&mqtt_cfg);
     esp_mqtt_client_register_event(client, esp_mmqtt_ID, mqtt_event_handler, client);
     esp_mqtt_client_start(client);
 
-    MQTTPublish(_LWTContext, "");
+    MQTTPublish(_LWTContext, "", 1);
 }
+
+void MQTTdestroy() {
+    if (client != NULL) {
+        esp_mqtt_client_stop(client);
+        esp_mqtt_client_destroy(client);
+    }
+}
+
+bool MQTTisConnected() {
+    return mqtt_connected;
+}
+
+void MQTTregisterConnectFunction(std::string name, std::function<void()> func){
+    ESP_LOGD(TAG_INTERFACEMQTT, "MQTTregisteronnectFunction %s\r\n", name.c_str());
+    if (connectFunktionMap == NULL) {
+        connectFunktionMap = new std::map<std::string, std::function<void()>>();
+    }
+
+    if ((*connectFunktionMap)[name] != NULL) {
+        ESP_LOGW(TAG_INTERFACEMQTT, "connect function %s already registred", name.c_str());
+        return;
+    }
+
+    (*connectFunktionMap)[name] = func;
+
+    if (mqtt_connected) {
+        func();
+    }
+}
+
+void MQTTunregisterConnectFunction(std::string name){
+    ESP_LOGD(TAG_INTERFACEMQTT, "MQTTregisteronnectFunction %s\r\n", name.c_str());
+    if ((connectFunktionMap != NULL) && (connectFunktionMap->find(name) != connectFunktionMap->end())) {
+        connectFunktionMap->erase(name);
+    }
+}
+
+void MQTTregisterSubscribeFunction(std::string topic, std::function<bool(std::string, char*, int)> func){
+    ESP_LOGD(TAG_INTERFACEMQTT, "MQTTregisterSubscribeFunction %s\r\n", topic.c_str());
+    if (subscribeFunktionMap == NULL) {
+        subscribeFunktionMap = new std::map<std::string, std::function<bool(std::string, char*, int)>>();
+    }
+
+    if ((*subscribeFunktionMap)[topic] != NULL) {
+        ESP_LOGW(TAG_INTERFACEMQTT, "topic %s already registred for subscription", topic.c_str());
+        return;
+    }
+
+    (*subscribeFunktionMap)[topic] = func;
+
+    if (mqtt_connected) {
+        int msg_id = esp_mqtt_client_subscribe(client, topic.c_str(), 0);
+        ESP_LOGD(TAG_INTERFACEMQTT, "topic %s subscribe successful, msg_id=%d", topic.c_str(), msg_id);
+    }
+}
+
+void MQTTconnected(){
+    if (mqtt_connected) {
+        if (connectFunktionMap != NULL) {
+            for(std::map<std::string, std::function<void()>>::iterator it = connectFunktionMap->begin(); it != connectFunktionMap->end(); ++it) {
+                it->second();
+                ESP_LOGD(TAG_INTERFACEMQTT, "call connect function %s", it->first.c_str());
+            }
+        }
+
+        if (subscribeFunktionMap != NULL) {
+            for(std::map<std::string, std::function<bool(std::string, char*, int)>>::iterator it = subscribeFunktionMap->begin(); it != subscribeFunktionMap->end(); ++it) {
+                int msg_id = esp_mqtt_client_subscribe(client, it->first.c_str(), 0);
+                ESP_LOGD(TAG_INTERFACEMQTT, "topic %s subscribe successful, msg_id=%d", it->first.c_str(), msg_id);
+            }
+        }
+    }
+}
+
+void MQTTdestroySubscribeFunction(){
+    if (subscribeFunktionMap != NULL) {
+        if (mqtt_connected) {
+            for(std::map<std::string, std::function<bool(std::string, char*, int)>>::iterator it = subscribeFunktionMap->begin(); it != subscribeFunktionMap->end(); ++it) {
+                int msg_id = esp_mqtt_client_unsubscribe(client, it->first.c_str());
+                ESP_LOGI(TAG_INTERFACEMQTT, "topic %s unsubscribe successful, msg_id=%d", it->first.c_str(), msg_id);
+            }
+        }
+
+        subscribeFunktionMap->clear();
+        delete subscribeFunktionMap;
+        subscribeFunktionMap = NULL;
+    }
+}

code/components/jomjol_mqtt/interface_mqtt.h

@@ -1,7 +1,23 @@
+#ifndef INTERFACE_MQTT_H
+#define INTERFACE_MQTT_H
+
 #include <string>
+#include <map>
+#include <functional>
 
 void MQTTInit(std::string _mqttURI, std::string _clientid, std::string _user, std::string _password, std::string _LWTContext, int _keepalive);
+void MQTTdestroy();
 
 //void MQTTInit(std::string _mqttURI, std::string _clientid, std::string _user = "", std::string _password = "");
 
-void MQTTPublish(std::string _key, std::string _content);
+void MQTTPublish(std::string _key, std::string _content, int retained_flag = 0);
+
+bool MQTTisConnected();
+
+void MQTTregisterConnectFunction(std::string name, std::function<void()> func);
+void MQTTunregisterConnectFunction(std::string name);
+void MQTTregisterSubscribeFunction(std::string topic, std::function<bool(std::string, char*, int)> func);
+void MQTTdestroySubscribeFunction();
+void MQTTconnected();
+
+#endif //INTERFACE_MQTT_H

code/components/jomjol_tfliteclass/CTfLiteClass.cpp

@@ -17,6 +17,7 @@ float CTfLiteClass::GetOutputValue(int nr)
     return output2->data.f[nr];
 }
 
+
 int CTfLiteClass::GetClassFromImageBasis(CImageBasis *rs)
 {
     if (!LoadInputImageBasis(rs))
@@ -98,7 +99,8 @@ void CTfLiteClass::GetOutPut()
 
 void CTfLiteClass::Invoke()
 {
-    interpreter->Invoke();
+    if (interpreter != nullptr)
+      interpreter->Invoke();
 }
 
 
@@ -155,9 +157,9 @@ void CTfLiteClass::MakeAllocate()
 }
 
 void CTfLiteClass::GetInputTensorSize(){
+#ifdef DEBUG_DETAIL_ON    
     float *zw = this->input;
     int test = sizeof(zw);
-#ifdef DEBUG_DETAIL_ON    
     printf("Input Tensor Dimension: %d\n", test);       
 #endif
 }
@@ -184,13 +186,11 @@ unsigned char* CTfLiteClass::ReadFileToCharArray(std::string _fn)
 
     unsigned char *result = (unsigned char*) malloc(size);
     int anz = 1;
-    TickType_t xDelay;
     while (!result && (anz < 6))    // maximal 5x versuchen (= 5s)
     {
 #ifdef DEBUG_DETAIL_ON      
 		    printf("Speicher ist voll - Versuche es erneut: %d.\n", anz);
 #endif
-        xDelay = 1000 / portTICK_PERIOD_MS;
         result = (unsigned char*) malloc(size);
         anz++;
     }
@@ -208,7 +208,7 @@ unsigned char* CTfLiteClass::ReadFileToCharArray(std::string _fn)
     return result;
 }
 
-void CTfLiteClass::LoadModel(std::string _fn){
+bool CTfLiteClass::LoadModel(std::string _fn){
 
 #ifdef SUPRESS_TFLITE_ERRORS
     this->error_reporter = new tflite::OwnMicroErrorReporter;
@@ -219,9 +219,14 @@ void CTfLiteClass::LoadModel(std::string _fn){
     unsigned char *rd;
     rd = ReadFileToCharArray(_fn.c_str());
 
+    if (rd == NULL) 
+      return false;
+
     this->model = tflite::GetModel(rd);
     free(rd);
     TFLITE_MINIMAL_CHECK(model != nullptr); 
+    
+    return true;
 }
 
 

code/components/jomjol_tfliteclass/CTfLiteClass.h

@@ -9,7 +9,7 @@
 #include "tensorflow/lite/micro/micro_error_reporter.h"
 #include "tensorflow/lite/micro/micro_interpreter.h"
 #include "tensorflow/lite/schema/schema_generated.h"
-#include "tensorflow/lite/version.h"
+//#include "tensorflow/lite/version.h"
 #include "tensorflow/lite/micro/kernels/micro_ops.h"
 #include "esp_err.h"
 #include "esp_log.h"
@@ -56,7 +56,7 @@ class CTfLiteClass
     public:
         CTfLiteClass();
         ~CTfLiteClass();        
-        void LoadModel(std::string _fn);
+        bool LoadModel(std::string _fn);
         void MakeAllocate();
         void GetInputTensorSize();
         bool LoadInputImageBasis(CImageBasis *rs);
@@ -64,6 +64,7 @@ class CTfLiteClass
         void GetOutPut();
         int GetOutClassification();
         int GetClassFromImageBasis(CImageBasis *rs);
+        std::string GetStatusFlow();
 
         float GetOutputValue(int nr);
         void GetInputDimension(bool silent);

code/components/jomjol_tfliteclass/server_tflite.cpp

@@ -8,6 +8,7 @@
 #include <iomanip>
 #include <sstream>
 
+#include "defines.h"
 #include "Helper.h"
 
 #include "esp_camera.h"
@@ -17,8 +18,9 @@
 #include "ClassFlowControll.h"
 
 #include "ClassLogFile.h"
+#include "server_GPIO.h"
 
-//#define DEBUG_DETAIL_ON       
+// #define DEBUG_DETAIL_ON       
 
 
 ClassFlowControll tfliteflow;
@@ -37,6 +39,9 @@ bool auto_isrunning = false;
 
 int countRounds = 0;
 
+static const char *TAGTFLITE = "server_tflite";
+
+
 int getCountFlowRounds() {
     return countRounds;
 }
@@ -64,9 +69,11 @@ void KillTFliteTasks()
 #ifdef DEBUG_DETAIL_ON          
     printf("Handle: xHandleblink_task_doFlow: %ld\n", (long) xHandleblink_task_doFlow);  
 #endif  
-    if (xHandleblink_task_doFlow)
+    if (xHandleblink_task_doFlow != NULL)
     {
-        vTaskDelete(xHandleblink_task_doFlow);
+        TaskHandle_t xHandleblink_task_doFlowTmp = xHandleblink_task_doFlow;
+        xHandleblink_task_doFlow = NULL;
+        vTaskDelete(xHandleblink_task_doFlowTmp);
 #ifdef DEBUG_DETAIL_ON      
         printf("Killed: xHandleblink_task_doFlow\n");
 #endif
@@ -75,9 +82,11 @@ void KillTFliteTasks()
 #ifdef DEBUG_DETAIL_ON      
     printf("Handle: xHandletask_autodoFlow: %ld\n", (long) xHandletask_autodoFlow);  
 #endif
-    if (xHandletask_autodoFlow)
+    if (xHandletask_autodoFlow != NULL)
     {
-        vTaskDelete(xHandletask_autodoFlow);
+        TaskHandle_t xHandletask_autodoFlowTmp = xHandletask_autodoFlow;
+        xHandletask_autodoFlow = NULL;
+        vTaskDelete(xHandletask_autodoFlowTmp);
 #ifdef DEBUG_DETAIL_ON      
         printf("Killed: xHandletask_autodoFlow\n");
 #endif
@@ -87,11 +96,10 @@ void KillTFliteTasks()
 
 void doInit(void)
 {
-    string config = "/sdcard/config/config.ini";
 #ifdef DEBUG_DETAIL_ON             
     printf("Start tfliteflow.InitFlow(config);\n");
 #endif
-    tfliteflow.InitFlow(config);
+    tfliteflow.InitFlow(CONFIG_FILE);
 #ifdef DEBUG_DETAIL_ON      
     printf("Finished tfliteflow.InitFlow(config);\n");
 #endif
@@ -136,7 +144,7 @@ esp_err_t handler_init(httpd_req_t *req)
     printf("handler_doinit uri:\n"); printf(req->uri); printf("\n");
 #endif
 
-    char* resp_str = "Init started<br>";
+    const char* resp_str = "Init started<br>";
     httpd_resp_send(req, resp_str, strlen(resp_str));     
 
     doInit();
@@ -159,8 +167,6 @@ esp_err_t handler_doflow(httpd_req_t *req)
     LogFile.WriteHeapInfo("handler_doflow - Start");       
 #endif
 
-    char* resp_str;
-
     printf("handler_doFlow uri: "); printf(req->uri); printf("\n");
 
     if (flowisrunning)
@@ -173,7 +179,7 @@ esp_err_t handler_doflow(httpd_req_t *req)
     {
         xTaskCreate(&blink_task_doFlow, "blink_doFlow", configMINIMAL_STACK_SIZE * 64, NULL, tskIDLE_PRIORITY+1, &xHandleblink_task_doFlow);
     }
-    resp_str = "doFlow gestartet - dauert ca. 60 Sekunden";
+    const char* resp_str = "doFlow gestartet - dauert ca. 60 Sekunden";
     httpd_resp_send(req, resp_str, strlen(resp_str));  
     /* Respond with an empty chunk to signal HTTP response completion */
     httpd_resp_send_chunk(req, NULL, 0);       
@@ -196,6 +202,8 @@ esp_err_t handler_wasserzaehler(httpd_req_t *req)
 
     bool _rawValue = false;
     bool _noerror = false;
+    bool _all = false;
+    std::string _type = "value";
     string zw;
 
     printf("handler_wasserzaehler uri:\n"); printf(req->uri); printf("\n");
@@ -206,6 +214,22 @@ esp_err_t handler_wasserzaehler(httpd_req_t *req)
     if (httpd_req_get_url_query_str(req, _query, 100) == ESP_OK)
     {
 //        printf("Query: "); printf(_query); printf("\n");
+        if (httpd_query_key_value(_query, "all", _size, 10) == ESP_OK)
+        {
+#ifdef DEBUG_DETAIL_ON       
+            printf("all is found"); printf(_size); printf("\n"); 
+#endif
+            _all = true;
+        }
+
+        if (httpd_query_key_value(_query, "type", _size, 10) == ESP_OK)
+        {
+#ifdef DEBUG_DETAIL_ON       
+            printf("all is found"); printf(_size); printf("\n"); 
+#endif
+            _type = std::string(_size);
+        }
+
         if (httpd_query_key_value(_query, "rawvalue", _size, 10) == ESP_OK)
         {
 #ifdef DEBUG_DETAIL_ON       
@@ -222,6 +246,29 @@ esp_err_t handler_wasserzaehler(httpd_req_t *req)
         }        
     }  
 
+    httpd_resp_set_hdr(req, "Access-Control-Allow-Origin", "*");
+
+    if (_all)
+    {
+        httpd_resp_set_type(req, "text/plain");
+        printf("TYPE: %s\n", _type.c_str());
+        int _intype = READOUT_TYPE_VALUE;
+        if (_type == "prevalue")
+            _intype = READOUT_TYPE_PREVALUE;
+        if (_type == "raw")
+            _intype = READOUT_TYPE_RAWVALUE;
+        if (_type == "error")
+            _intype = READOUT_TYPE_ERROR;
+
+
+        zw = tfliteflow.getReadoutAll(_intype);
+        printf("ZW: %s\n", zw.c_str());
+        if (zw.length() > 0)
+            httpd_resp_sendstr_chunk(req, zw.c_str()); 
+        httpd_resp_sendstr_chunk(req, NULL);   
+        return ESP_OK;
+    }
+
     zw = tfliteflow.getReadout(_rawValue, _noerror);
     if (zw.length() > 0)
         httpd_resp_sendstr_chunk(req, zw.c_str()); 
@@ -429,7 +476,7 @@ esp_err_t handler_editflow(httpd_req_t *req)
 
 //        printf("Parameter host: "); printf(_host.c_str()); printf("\n"); 
 //        string zwzw = "Do " + _task + " start\n"; printf(zwzw.c_str());
-        bool changed = Camera.SetBrightnessContrastSaturation(bri, con, sat);
+        Camera.SetBrightnessContrastSaturation(bri, con, sat);
         std::string zw = tfliteflow.doSingleStep("[MakeImage]", _host);
         httpd_resp_sendstr_chunk(req, zw.c_str()); 
     } 
@@ -483,6 +530,36 @@ esp_err_t handler_editflow(httpd_req_t *req)
 };
 
 
+esp_err_t handler_statusflow(httpd_req_t *req)
+{
+#ifdef DEBUG_DETAIL_ON       
+    LogFile.WriteHeapInfo("handler_prevalue - Start");       
+#endif
+
+    const char* resp_str;
+    string zw;
+
+#ifdef DEBUG_DETAIL_ON       
+    printf("handler_prevalue:\n"); printf(req->uri); printf("\n");
+#endif
+
+    zw = tfliteflow.getActStatus();
+    
+    resp_str = zw.c_str();
+
+    httpd_resp_set_hdr(req, "Access-Control-Allow-Origin", "*");
+    httpd_resp_send(req, resp_str, strlen(resp_str));   
+    /* Respond with an empty chunk to signal HTTP response completion */
+    httpd_resp_send_chunk(req, NULL, 0);      
+
+#ifdef DEBUG_DETAIL_ON       
+    LogFile.WriteHeapInfo("handler_prevalue - Start");       
+#endif
+
+    return ESP_OK;
+};
+
+
 esp_err_t handler_prevalue(httpd_req_t *req)
 {
 #ifdef DEBUG_DETAIL_ON       
@@ -498,6 +575,7 @@ esp_err_t handler_prevalue(httpd_req_t *req)
 
     char _query[100];
     char _size[10] = "";
+    char _numbers[50] = "default";
 
     if (httpd_req_get_url_query_str(req, _query, 100) == ESP_OK)
     {
@@ -511,15 +589,24 @@ esp_err_t handler_prevalue(httpd_req_t *req)
             printf("Value: "); printf(_size); printf("\n"); 
 #endif
         }
+
+        httpd_query_key_value(_query, "numbers", _numbers, 50);
     }      
 
     if (strlen(_size) == 0)
-        zw = tfliteflow.GetPrevalue();
+    {
+        zw = tfliteflow.GetPrevalue(std::string(_numbers));
+    }
     else
-        zw = "SetPrevalue to " + tfliteflow.UpdatePrevalue(_size);
+    {
+        zw = "SetPrevalue to " + tfliteflow.UpdatePrevalue(_size, _numbers, true);
+    }
     
     resp_str = zw.c_str();
 
+    httpd_resp_set_hdr(req, "Access-Control-Allow-Origin", "*");
+
+
     httpd_resp_send(req, resp_str, strlen(resp_str));   
     /* Respond with an empty chunk to signal HTTP response completion */
     httpd_resp_send_chunk(req, NULL, 0);      
@@ -535,17 +622,17 @@ void task_autodoFlow(void *pvParameter)
 {
     int64_t fr_start, fr_delta_ms;
 
+    printf("task_autodoFlow: start\r\n");
     doInit();
-    
-    auto_isrunning = tfliteflow.isAutoStart(auto_intervall);
+    gpio_handler_init();
 
+    auto_isrunning = tfliteflow.isAutoStart(auto_intervall);
     if (isSetupModusActive()) {
         auto_isrunning = false;
         std::string zw_time = gettimestring(LOGFILE_TIME_FORMAT);
         tfliteflow.doFlowMakeImageOnly(zw_time);
 
     }
-
     while (auto_isrunning)
     {
         std::string _zw = "task_autodoFlow - next round - Round #" + std::to_string(++countRounds);
@@ -590,6 +677,7 @@ void task_autodoFlow(void *pvParameter)
     }
     vTaskDelete(NULL); //Delete this task if it exits from the loop above
     xHandletask_autodoFlow = NULL;
+    printf("task_autodoFlow: end\r\n");
 }
 
 void TFliteDoAutoStart()
@@ -597,6 +685,11 @@ void TFliteDoAutoStart()
     xTaskCreate(&task_autodoFlow, "task_autodoFlow", configMINIMAL_STACK_SIZE * 64, NULL, tskIDLE_PRIORITY+1, &xHandletask_autodoFlow);
 }
 
+std::string GetMQTTMainTopic()
+{
+    return tfliteflow.GetMQTTMainTopic();
+}
+
 
 
 void register_server_tflite_uri(httpd_handle_t server)
@@ -621,6 +714,10 @@ void register_server_tflite_uri(httpd_handle_t server)
     camuri.user_ctx  = (void*) "Light Off"; 
     httpd_register_uri_handler(server, &camuri);  
 
+    camuri.uri       = "/statusflow.html";
+    camuri.handler   = handler_statusflow;
+    camuri.user_ctx  = (void*) "Light Off"; 
+    httpd_register_uri_handler(server, &camuri);  
     
     camuri.uri       = "/editflow.html";
     camuri.handler   = handler_editflow;

code/components/jomjol_tfliteclass/server_tflite.h

@@ -1,12 +1,11 @@
 #include <esp_log.h>
+#include <string>
 
 #include <esp_http_server.h>
 #include "CImageBasis.h"
 
 //#include "ClassControllCamera.h"
 
-static const char *TAGTFLITE = "server_tflite";
-
 void register_server_tflite_uri(httpd_handle_t server);
 
 void KillTFliteTasks();
@@ -15,6 +14,8 @@ void TFliteDoAutoStart();
 
 bool isSetupModusActive();
 
+std::string GetMQTTMainTopic();
+
 esp_err_t GetJPG(std::string _filename, httpd_req_t *req);
 
 esp_err_t GetRawJPG(httpd_req_t *req);

code/components/jomjol_time_sntp/time_sntp.cpp

@@ -18,6 +18,7 @@
 static const char *TAG = "sntp";
 
 bool setTimeAlwaysOnReboot = true;
+time_t bootTime;
 
 static void obtain_time(void);
 static void initialize_sntp(void);
@@ -27,6 +28,17 @@ void time_sync_notification_cb(struct timeval *tv)
     ESP_LOGI(TAG, "Notification of a time synchronization event");
 }
 
+std::string ConvertTimeToString(time_t _time, const char * frm)
+{
+    struct tm timeinfo;
+    char strftime_buf[64];
+    localtime_r(&_time, &timeinfo);
+    strftime(strftime_buf, sizeof(strftime_buf), frm, &timeinfo);
+
+    std::string result(strftime_buf);
+    return result;
+}
+
 std::string gettimestring(const char * frm)
 {
     time_t now;
@@ -114,4 +126,17 @@ static void initialize_sntp(void)
     sntp_setservername(0, "pool.ntp.org");
 //    sntp_set_time_sync_notification_cb(time_sync_notification_cb);
     sntp_init();
+}
+
+void setBootTime() 
+{
+    time(&bootTime);
+}
+
+time_t getUpTime() 
+{
+    time_t now;
+    time(&now);
+
+    return now - bootTime;
 }

code/components/jomjol_time_sntp/time_sntp.h

@@ -15,5 +15,10 @@
 void setup_time(void);
 
 std::string gettimestring(const char * frm);
+std::string ConvertTimeToString(time_t _time, const char * frm);
+
 void setTimeZone(std::string _tzstring);
-void reset_servername(std::string _servername);
+void reset_servername(std::string _servername);
+
+void setBootTime();
+time_t getUpTime();

code/components/jomjol_wlan/CMakeLists.txt

@@ -0,0 +1,7 @@
+FILE(GLOB_RECURSE app_sources ${CMAKE_CURRENT_SOURCE_DIR}/*.*)
+
+idf_component_register(SRCS ${app_sources}
+                    INCLUDE_DIRS "."
+                    REQUIRES nvs_flash jomjol_helper)
+
+

code/components/jomjol_wlan/connect_wlan.cpp

@@ -0,0 +1,289 @@
+#include "connect_wlan.h"
+
+#include <string.h>
+#include "freertos/FreeRTOS.h"
+#include "freertos/task.h"
+#include "freertos/event_groups.h"
+#include "driver/gpio.h"
+#include "esp_system.h"
+#include "esp_wifi.h"
+#include "esp_event.h"
+#include "esp_log.h"
+#include "nvs_flash.h"
+
+#include "lwip/err.h"
+#include "lwip/sys.h"
+
+#include <fstream>
+#include <string>
+#include <vector>
+#include <sstream>
+#include <iostream>
+
+
+
+#define EXAMPLE_ESP_MAXIMUM_RETRY  1000
+
+/* FreeRTOS event group to signal when we are connected*/
+static EventGroupHandle_t s_wifi_event_group;
+
+/* The event group allows multiple bits for each event, but we only care about two events:
+ * - we are connected to the AP with an IP
+ * - we failed to connect after the maximum amount of retries */
+#define WIFI_CONNECTED_BIT BIT0
+#define WIFI_FAIL_BIT      BIT1
+
+static const char *TAG = "wifi station";
+
+static int s_retry_num = 0;
+
+///////////////////////////////////////////////////////////
+#define BLINK_GPIO GPIO_NUM_33
+
+int BlinkDauer;
+int BlinkAnzahl;
+bool BlinkOff;
+bool BlinkIsRunning = false;
+
+std::string hostname = "";
+std::string std_hostname = "watermeter";
+std::string ipadress = "";
+std::string ssid = "";
+
+std::string getIPAddress()
+{
+    return ipadress;
+}
+
+std::string getSSID()
+{
+    return ssid;
+}
+
+
+void task_doBlink(void *pvParameter)
+{
+    ESP_LOGI("BLINK", "Blinken - start");
+    while (BlinkIsRunning)
+    {
+//        ESP_LOGI("BLINK", "Blinken - wait");
+        vTaskDelay(100 / portTICK_PERIOD_MS);
+    }
+
+    BlinkIsRunning = true;
+
+	// Init the GPIO
+    gpio_pad_select_gpio(BLINK_GPIO);
+    /* Set the GPIO as a push/pull output */
+    gpio_set_direction(BLINK_GPIO, GPIO_MODE_OUTPUT);  
+
+    for (int i = 0; i < BlinkAnzahl; ++i)
+    {
+		if (BlinkAnzahl > 1)
+		{
+	        gpio_set_level(BLINK_GPIO, 1);
+    	    vTaskDelay(BlinkDauer / portTICK_PERIOD_MS);
+		}
+        gpio_set_level(BLINK_GPIO, 0);      
+        vTaskDelay(BlinkDauer / portTICK_PERIOD_MS);
+    }
+
+    if (BlinkOff)
+        gpio_set_level(BLINK_GPIO, 1);
+
+    ESP_LOGI("BLINK", "Blinken - done");
+    BlinkIsRunning = false;
+
+    vTaskDelete(NULL); //Delete this task if it exits from the loop above
+}
+
+void LEDBlinkTask(int _dauer, int _anz, bool _off)
+{
+	BlinkDauer = _dauer;
+	BlinkAnzahl = _anz;
+	BlinkOff = _off;
+
+    xTaskCreate(&task_doBlink, "task_doBlink", configMINIMAL_STACK_SIZE * 8, NULL, tskIDLE_PRIORITY+1, NULL);
+}
+/////////////////////////////////////////////////////////
+
+
+static void event_handler(void* arg, esp_event_base_t event_base,
+                                int32_t event_id, void* event_data)
+{
+    if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_START) {
+        LEDBlinkTask(200, 1, true);
+        esp_wifi_connect();
+    } else if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_DISCONNECTED) {
+//        if (s_retry_num < EXAMPLE_ESP_MAXIMUM_RETRY){ 
+            esp_wifi_connect();
+            s_retry_num++;
+            ESP_LOGI(TAG, "retry to connect to the AP");
+        ESP_LOGI(TAG,"connect to the AP fail");
+    } else if (event_base == IP_EVENT && event_id == IP_EVENT_STA_GOT_IP) {
+        ip_event_got_ip_t* event = (ip_event_got_ip_t*) event_data;
+        ESP_LOGI(TAG, "got ip:" IPSTR, IP2STR(&event->ip_info.ip));
+        ipadress = std::string(ip4addr_ntoa((const ip4_addr*) &event->ip_info.ip));
+        s_retry_num = 0;
+        xEventGroupSetBits(s_wifi_event_group, WIFI_CONNECTED_BIT);
+        LEDBlinkTask(1000, 5, true);
+    }
+}
+
+void strinttoip4(const char *ip, int &a, int &b, int &c, int &d) {
+    std::string zw = std::string(ip);
+    std::stringstream s(zw);
+    char ch; //to temporarily store the '.'
+    s >> a >> ch >> b >> ch >> c >> ch >> d;
+}
+
+
+void wifi_init_sta(const char *_ssid, const char *_password, const char *_hostname, const char *_ipadr, const char *_gw,  const char *_netmask, const char *_dns)
+{
+    ssid = std::string(_ssid);
+
+    s_wifi_event_group = xEventGroupCreate();
+
+    ESP_ERROR_CHECK(esp_netif_init());
+    ESP_ERROR_CHECK(esp_event_loop_create_default());
+
+/////////////////////////////////////////////////////////////////
+
+
+    esp_netif_t *my_sta = esp_netif_create_default_wifi_sta();
+
+    if ((_ipadr != NULL) && (_gw != NULL) && (_netmask != NULL))
+    {
+        ESP_LOGI(TAG, "set IP %s, GW %s, Netmask %s manual", _ipadr, _gw, _netmask);
+        esp_netif_dhcpc_stop(my_sta);
+
+        esp_netif_ip_info_t ip_info;
+        int a, b, c, d;
+        strinttoip4(_ipadr, a, b, c, d);
+        IP4_ADDR(&ip_info.ip, a, b, c, d);
+        strinttoip4(_gw, a, b, c, d);
+        IP4_ADDR(&ip_info.gw, a, b, c, d);
+        strinttoip4(_netmask, a, b, c, d);
+        IP4_ADDR(&ip_info.netmask, a, b, c, d);
+
+        esp_netif_set_ip_info(my_sta, &ip_info);
+    }
+
+    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
+    ESP_ERROR_CHECK(esp_wifi_init(&cfg));
+
+    if ((_ipadr != NULL) && (_gw != NULL) && (_netmask != NULL))
+    {
+        if (_dns == NULL)
+            _dns = _gw;
+            
+        ESP_LOGI(TAG, "set DNS manual");
+        esp_netif_dns_info_t dns_info;
+        ip4_addr_t ip;
+        ip.addr = esp_ip4addr_aton(_dns);
+        ip_addr_set_ip4_u32(&dns_info.ip, ip.addr);
+        ESP_ERROR_CHECK(esp_netif_set_dns_info(my_sta, ESP_NETIF_DNS_MAIN, &dns_info));
+    }
+/////////////////////////////////////////////////////////////////
+
+
+//    esp_netif_create_default_wifi_sta();
+
+//    wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
+//    ESP_ERROR_CHECK(esp_wifi_init(&cfg));
+
+
+
+/* 
+////////////////////////////// esp-idf 4.2 //////////////////////////
+    esp_event_handler_instance_t instance_any_id;
+    esp_event_handler_instance_t instance_got_ip;
+    ESP_ERROR_CHECK(esp_event_handler_instance_register(WIFI_EVENT,
+                                                        ESP_EVENT_ANY_ID,
+                                                        &event_handler,
+                                                        NULL,
+                                                        &instance_any_id));
+    ESP_ERROR_CHECK(esp_event_handler_instance_register(IP_EVENT,
+                                                        IP_EVENT_STA_GOT_IP,
+                                                        &event_handler,
+                                                        NULL,
+                                                        &instance_got_ip));
+////////////////////////// ENDE esp-idf 4.2 ///////////////////////////
+*/
+
+    ESP_ERROR_CHECK(esp_event_handler_register(WIFI_EVENT, ESP_EVENT_ANY_ID, &event_handler, NULL));
+    ESP_ERROR_CHECK(esp_event_handler_register(IP_EVENT, IP_EVENT_STA_GOT_IP, &event_handler, NULL));
+
+    wifi_config_t wifi_config = { };
+
+    strcpy((char*)wifi_config.sta.ssid, (const char*)_ssid);
+    strcpy((char*)wifi_config.sta.password, (const char*)_password);
+
+    ESP_ERROR_CHECK(esp_wifi_set_mode(WIFI_MODE_STA) );
+    ESP_ERROR_CHECK(esp_wifi_set_config(WIFI_IF_STA, &wifi_config) );
+    ESP_ERROR_CHECK(esp_wifi_start() );
+
+    if (_hostname != NULL)
+    {
+        esp_err_t ret = tcpip_adapter_set_hostname(TCPIP_ADAPTER_IF_STA , _hostname);
+        hostname = std::string(_hostname);
+        if(ret != ESP_OK ){
+            ESP_LOGE(TAG,"failed to set hostname:%d",ret);  
+        }
+    }
+
+    ESP_LOGI(TAG, "wifi_init_sta finished.");
+
+    /* Waiting until either the connection is established (WIFI_CONNECTED_BIT) or connection failed for the maximum
+     * number of re-tries (WIFI_FAIL_BIT). The bits are set by event_handler() (see above) */
+    EventBits_t bits = xEventGroupWaitBits(s_wifi_event_group,
+            WIFI_CONNECTED_BIT | WIFI_FAIL_BIT,
+            pdFALSE,
+            pdFALSE,
+            portMAX_DELAY);
+
+    /* xEventGroupWaitBits() returns the bits before the call returned, hence we can test which event actually
+     * happened. */
+    if (bits & WIFI_CONNECTED_BIT) {
+        ESP_LOGI(TAG, "connected to ap SSID:%s password:%s",
+                 _ssid, _password);
+    } else if (bits & WIFI_FAIL_BIT) {
+        ESP_LOGI(TAG, "Failed to connect to SSID:%s, password:%s",
+                 _ssid, _password);
+    } else {
+        ESP_LOGE(TAG, "UNEXPECTED EVENT");
+    }
+
+    /* The event will not be processed after unregister */
+/* 
+////////////////////////////// esp-idf 4.2 //////////////////////////
+    ESP_ERROR_CHECK(esp_event_handler_instance_unregister(IP_EVENT, IP_EVENT_STA_GOT_IP, instance_got_ip));
+    ESP_ERROR_CHECK(esp_event_handler_instance_unregister(WIFI_EVENT, ESP_EVENT_ANY_ID, instance_any_id));
+////////////////////////// ENDE esp-idf 4.2 ///////////////////////////
+*/
+
+/* Deaktiveren, damit bei einen Verbindungsabbruch neu aufgebaut wird
+    ESP_ERROR_CHECK(esp_event_handler_unregister(IP_EVENT, IP_EVENT_STA_GOT_IP, &event_handler));
+    ESP_ERROR_CHECK(esp_event_handler_unregister(WIFI_EVENT, ESP_EVENT_ANY_ID, &event_handler));
+    vEventGroupDelete(s_wifi_event_group);
+*/ 
+
+/*
+    while (BlinkIsRunning)
+    {
+        vTaskDelay(100 / portTICK_PERIOD_MS);
+    }
+*/
+}
+
+
+void wifi_init_sta(const char *_ssid, const char *_password, const char *_hostname)
+{
+    wifi_init_sta(_ssid, _password, _hostname, NULL, NULL, NULL, NULL);
+}
+
+void wifi_init_sta(const char *_ssid, const char *_password)
+{
+    wifi_init_sta(_ssid, _password, NULL, NULL, NULL, NULL, NULL);
+}
+

code/components/jomjol_wlan/connect_wlan.h

@@ -0,0 +1,16 @@
+#ifndef CONNECT_WLAN_H
+#define CONNECT_WLAN_H
+
+#include <string>
+
+void wifi_init_sta(const char *_ssid, const char *_password, const char *_hostname, const char *_ipadr, const char *_gw,  const char *_netmask, const char *_dns);
+void wifi_init_sta(const char *_ssid, const char *_password, const char *_hostname);
+void wifi_init_sta(const char *_ssid, const char *_password);
+
+std::string getIPAddress();
+std::string getSSID();
+
+extern std::string hostname;
+extern std::string std_hostname;
+
+#endif

code/components/jomjol_wlan/read_wlanini.cpp

@@ -0,0 +1,257 @@
+#include "read_wlanini.h"
+
+#include "Helper.h"
+
+#include "connect_wlan.h"
+
+#include <fstream>
+#include <string>
+#include <vector>
+#include <sstream>
+#include <iostream>
+#include <string.h>
+
+
+std::vector<string> ZerlegeZeile(std::string input, std::string _delimiter = "")
+{
+	std::vector<string> Output;
+	std::string delimiter = " =,";
+    if (_delimiter.length() > 0){
+        delimiter = _delimiter;
+    }
+
+	input = trim(input, delimiter);
+	size_t pos = findDelimiterPos(input, delimiter);
+	std::string token;
+	while (pos != std::string::npos) {
+		token = input.substr(0, pos);
+		token = trim(token, delimiter);
+		Output.push_back(token);
+		input.erase(0, pos + 1);
+		input = trim(input, delimiter);
+		pos = findDelimiterPos(input, delimiter);
+	}
+	Output.push_back(input);
+
+	return Output;
+}
+
+
+
+void LoadWlanFromFile(std::string fn, char *&_ssid, char *&_password, char *&_hostname, char *&_ipadr, char *&_gw,  char *&_netmask, char *&_dns)
+{
+    std::string ssid = "";
+    std::string passphrase = "";
+    std::string ipaddress = "";
+    std::string gw = "";
+    std::string netmask = "";
+    std::string dns = "";
+
+    std::string line = "";
+    std::vector<string> zerlegt;
+    hostname = std_hostname;
+
+    FILE* pFile;
+    fn = FormatFileName(fn);
+
+    pFile = OpenFileAndWait(fn.c_str(), "r");
+    printf("file loaded\n");
+
+    if (pFile == NULL)
+        return;
+
+    char zw[1024];
+    fgets(zw, 1024, pFile);
+    line = std::string(zw);
+
+    while ((line.size() > 0) || !(feof(pFile)))
+    {
+//        printf("%s", line.c_str());
+        zerlegt = ZerlegeZeile(line, "=");
+        zerlegt[0] = trim(zerlegt[0], " ");
+        for (int i = 2; i < zerlegt.size(); ++i)
+            zerlegt[1] = zerlegt[1] + "=" + zerlegt[i];
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "HOSTNAME")){
+            hostname = trim(zerlegt[1]);
+            if ((hostname[0] == '"') && (hostname[hostname.length()-1] == '"')){
+                hostname = hostname.substr(1, hostname.length()-2);
+            }
+        }
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "SSID")){
+            ssid = trim(zerlegt[1]);
+            if ((ssid[0] == '"') && (ssid[ssid.length()-1] == '"')){
+                ssid = ssid.substr(1, ssid.length()-2);
+            }
+        }
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "PASSWORD")){
+            passphrase = zerlegt[1];
+            if ((passphrase[0] == '"') && (passphrase[passphrase.length()-1] == '"')){
+                passphrase = passphrase.substr(1, passphrase.length()-2);
+            }
+        }
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "IP")){
+            ipaddress = zerlegt[1];
+            if ((ipaddress[0] == '"') && (ipaddress[ipaddress.length()-1] == '"')){
+                ipaddress = ipaddress.substr(1, ipaddress.length()-2);
+            }
+        }
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "GATEWAY")){
+            gw = zerlegt[1];
+            if ((gw[0] == '"') && (gw[gw.length()-1] == '"')){
+                gw = gw.substr(1, gw.length()-2);
+            }
+        }
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "NETMASK")){
+            netmask = zerlegt[1];
+            if ((netmask[0] == '"') && (netmask[netmask.length()-1] == '"')){
+                netmask = netmask.substr(1, netmask.length()-2);
+            }
+        }
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "DNS")){
+            dns = zerlegt[1];
+            if ((dns[0] == '"') && (dns[dns.length()-1] == '"')){
+                dns = dns.substr(1, dns.length()-2);
+            }
+        }
+
+
+        if (fgets(zw, 1024, pFile) == NULL)
+        {
+            line = "";
+        }
+        else
+        {
+            line = std::string(zw);
+        }
+    }
+
+    fclose(pFile);
+
+    // Check if Hostname was empty in .ini if yes set to std_hostname
+    if(hostname.length() == 0){
+        hostname = std_hostname;
+    }
+
+    _hostname = new char[hostname.length() + 1];
+    strcpy(_hostname, hostname.c_str());
+
+    _ssid = new char[ssid.length() + 1];
+    strcpy(_ssid, ssid.c_str());
+
+    _password = new char[passphrase.length() + 1];
+    strcpy(_password, passphrase.c_str());
+
+    if (ipaddress.length() > 0)
+    {
+        _ipadr = new char[ipaddress.length() + 1];
+        strcpy(_ipadr, ipaddress.c_str());
+    }
+    else
+        _ipadr = NULL;
+
+    if (gw.length() > 0)
+    {
+        _gw = new char[gw.length() + 1];
+        strcpy(_gw, gw.c_str());
+    }
+    else
+        _gw = NULL;
+
+    if (netmask.length() > 0)
+    {
+        _netmask = new char[netmask.length() + 1];
+        strcpy(_netmask, netmask.c_str());
+    }
+    else
+        _netmask = NULL;
+
+    if (dns.length() > 0)
+    {
+        _dns = new char[dns.length() + 1];
+        strcpy(_dns, dns.c_str());
+    }
+    else
+        _dns = NULL;
+}
+
+
+
+
+bool ChangeHostName(std::string fn, std::string _newhostname)
+{
+    if (_newhostname == hostname)
+        return false;
+
+    string line = "";
+    std::vector<string> zerlegt;
+
+    bool found = false;
+
+    std::vector<string> neuesfile;
+
+    FILE* pFile;
+    fn = FormatFileName(fn);
+    pFile = OpenFileAndWait(fn.c_str(), "r");
+
+    printf("file loaded\n");
+
+    if (pFile == NULL)
+        return false;
+
+    char zw[1024];
+    fgets(zw, 1024, pFile);
+    line = std::string(zw);
+
+    while ((line.size() > 0) || !(feof(pFile)))
+    {
+        printf("%s", line.c_str());
+        zerlegt = ZerlegeZeile(line, "=");
+        zerlegt[0] = trim(zerlegt[0], " ");
+
+        if ((zerlegt.size() > 1) && (toUpper(zerlegt[0]) == "HOSTNAME")){
+            line = "hostname = \"" + _newhostname + "\"\n";
+            found = true;
+        }
+
+        neuesfile.push_back(line);
+
+        if (fgets(zw, 1024, pFile) == NULL)
+        {
+            line = "";
+        }
+        else
+        {
+            line = std::string(zw);
+        }
+    }
+
+    if (!found)
+    {
+        line = "\nhostname = \"" + _newhostname + "\"\n";
+        neuesfile.push_back(line);        
+    }
+
+    fclose(pFile);
+
+    pFile = OpenFileAndWait(fn.c_str(), "w+");
+
+    for (int i = 0; i < neuesfile.size(); ++i)
+    {
+        printf(neuesfile[i].c_str());
+        fputs(neuesfile[i].c_str(), pFile);
+    }
+
+    fclose(pFile);
+
+    printf("*** Update hostname done ***\n");
+
+    return true;
+}
+

code/components/jomjol_wlan/read_wlanini.h

@@ -0,0 +1,11 @@
+#ifndef READ_WLANINI_H
+#define READ_WLANINI_H
+
+#include <string>
+
+void LoadWlanFromFile(std::string fn, char *&_ssid, char *&_password, char *&_hostname, char *&_ipadr, char *&_gw,  char *&_netmask, char *&_dns);
+
+bool ChangeHostName(std::string fn, std::string _newhostname);
+
+
+#endif

code/components/tfmicro/CMakeLists.txt

@@ -23,7 +23,7 @@ if(NOT DEFINED ENV{IDF_PATH})
 endif()
 
 idf_component_register(
-  SRCS tensorflow/lite/micro/micro_error_reporter.cc tensorflow/lite/micro/simple_memory_allocator.cc tensorflow/lite/micro/memory_helpers.cc tensorflow/lite/micro/test_helpers.cc tensorflow/lite/micro/recording_micro_allocator.cc tensorflow/lite/micro/micro_time.cc tensorflow/lite/micro/recording_simple_memory_allocator.cc tensorflow/lite/micro/micro_string.cc tensorflow/lite/micro/micro_profiler.cc tensorflow/lite/micro/debug_log.cc tensorflow/lite/micro/all_ops_resolver.cc tensorflow/lite/micro/micro_utils.cc tensorflow/lite/micro/micro_interpreter.cc tensorflow/lite/micro/micro_allocator.cc tensorflow/lite/micro/benchmarks/keyword_scrambled_model_data.cc tensorflow/lite/micro/memory_planner/linear_memory_planner.cc tensorflow/lite/micro/memory_planner/greedy_memory_planner.cc tensorflow/lite/micro/testing/test_conv_model.cc tensorflow/lite/c/common.c tensorflow/lite/core/api/error_reporter.cc tensorflow/lite/core/api/flatbuffer_conversions.cc tensorflow/lite/core/api/op_resolver.cc tensorflow/lite/core/api/tensor_utils.cc tensorflow/lite/kernels/internal/quantization_util.cc tensorflow/lite/kernels/kernel_util.cc tensorflow/lite/schema/schema_utils.cc tensorflow/lite/micro/kernels/prelu.cc tensorflow/lite/micro/kernels/dequantize.cc tensorflow/lite/micro/kernels/pad.cc tensorflow/lite/micro/kernels/shape.cc tensorflow/lite/micro/kernels/l2norm.cc tensorflow/lite/micro/kernels/tanh.cc tensorflow/lite/micro/kernels/resize_nearest_neighbor.cc tensorflow/lite/micro/kernels/logical.cc tensorflow/lite/micro/kernels/kernel_util.cc tensorflow/lite/micro/kernels/ceil.cc tensorflow/lite/micro/kernels/arg_min_max.cc tensorflow/lite/micro/kernels/softmax.cc tensorflow/lite/micro/kernels/sub.cc tensorflow/lite/micro/kernels/add.cc tensorflow/lite/micro/kernels/floor.cc tensorflow/lite/micro/kernels/kernel_runner.cc tensorflow/lite/micro/kernels/split_v.cc tensorflow/lite/micro/kernels/hard_swish.cc tensorflow/lite/micro/kernels/pooling.cc tensorflow/lite/micro/kernels/concatenation.cc tensorflow/lite/micro/kernels/mul.cc tensorflow/lite/micro/kernels/unpack.cc tensorflow/lite/micro/kernels/round.cc tensorflow/lite/micro/kernels/quantize.cc tensorflow/lite/micro/kernels/ethosu.cc tensorflow/lite/micro/kernels/svdf.cc tensorflow/lite/micro/kernels/maximum_minimum.cc tensorflow/lite/micro/kernels/reshape.cc tensorflow/lite/micro/kernels/reduce.cc tensorflow/lite/micro/kernels/strided_slice.cc tensorflow/lite/micro/kernels/neg.cc tensorflow/lite/micro/kernels/pack.cc tensorflow/lite/micro/kernels/elementwise.cc tensorflow/lite/micro/kernels/comparisons.cc tensorflow/lite/micro/kernels/fully_connected.cc tensorflow/lite/micro/kernels/depthwise_conv.cc tensorflow/lite/micro/kernels/split.cc tensorflow/lite/micro/kernels/logistic.cc tensorflow/lite/micro/kernels/circular_buffer.cc tensorflow/lite/micro/kernels/conv.cc tensorflow/lite/micro/kernels/activations.cc 
+  SRCS tensorflow/lite/micro/simple_memory_allocator.cc tensorflow/lite/micro/debug_log.cc tensorflow/lite/micro/micro_error_reporter.cc tensorflow/lite/micro/memory_helpers.cc tensorflow/lite/micro/test_helpers.cc tensorflow/lite/micro/recording_micro_allocator.cc tensorflow/lite/micro/micro_time.cc tensorflow/lite/micro/recording_simple_memory_allocator.cc tensorflow/lite/micro/micro_string.cc tensorflow/lite/micro/micro_profiler.cc tensorflow/lite/micro/flatbuffer_utils.cc tensorflow/lite/micro/micro_graph.cc tensorflow/lite/micro/mock_micro_graph.cc tensorflow/lite/micro/all_ops_resolver.cc tensorflow/lite/micro/micro_utils.cc tensorflow/lite/micro/micro_interpreter.cc tensorflow/lite/micro/micro_allocator.cc tensorflow/lite/micro/system_setup.cc tensorflow/lite/micro/memory_planner/linear_memory_planner.cc tensorflow/lite/micro/memory_planner/greedy_memory_planner.cc tensorflow/lite/schema/schema_utils.cc tensorflow/lite/c/common.c tensorflow/lite/core/api/tensor_utils.cc tensorflow/lite/core/api/error_reporter.cc tensorflow/lite/core/api/flatbuffer_conversions.cc tensorflow/lite/core/api/op_resolver.cc tensorflow/lite/kernels/kernel_util.cc tensorflow/lite/kernels/internal/quantization_util.cc tensorflow/lite/kernels/internal/reference/portable_tensor_utils.cc  tensorflow/lite/micro/kernels/activations.cc tensorflow/lite/micro/kernels/activations_common.cc tensorflow/lite/micro/kernels/add.cc tensorflow/lite/micro/kernels/add_n.cc tensorflow/lite/micro/kernels/arg_min_max.cc tensorflow/lite/micro/kernels/batch_to_space_nd.cc tensorflow/lite/micro/kernels/cast.cc tensorflow/lite/micro/kernels/ceil.cc tensorflow/lite/micro/kernels/circular_buffer.cc tensorflow/lite/micro/kernels/comparisons.cc tensorflow/lite/micro/kernels/concatenation.cc tensorflow/lite/micro/kernels/conv.cc tensorflow/lite/micro/kernels/conv_common.cc tensorflow/lite/micro/kernels/cumsum.cc tensorflow/lite/micro/kernels/depth_to_space.cc tensorflow/lite/micro/kernels/depthwise_conv.cc tensorflow/lite/micro/kernels/depthwise_conv_common.cc tensorflow/lite/micro/kernels/dequantize.cc tensorflow/lite/micro/kernels/detection_postprocess.cc tensorflow/lite/micro/kernels/elementwise.cc tensorflow/lite/micro/kernels/elu.cc tensorflow/lite/micro/kernels/ethosu.cc tensorflow/lite/micro/kernels/exp.cc tensorflow/lite/micro/kernels/expand_dims.cc tensorflow/lite/micro/kernels/fill.cc tensorflow/lite/micro/kernels/floor.cc tensorflow/lite/micro/kernels/floor_div.cc tensorflow/lite/micro/kernels/floor_mod.cc tensorflow/lite/micro/kernels/fully_connected.cc tensorflow/lite/micro/kernels/fully_connected_common.cc tensorflow/lite/micro/kernels/gather.cc tensorflow/lite/micro/kernels/gather_nd.cc tensorflow/lite/micro/kernels/hard_swish.cc tensorflow/lite/micro/kernels/hard_swish_common.cc tensorflow/lite/micro/kernels/if.cc tensorflow/lite/micro/kernels/kernel_runner.cc tensorflow/lite/micro/kernels/kernel_util.cc tensorflow/lite/micro/kernels/l2norm.cc tensorflow/lite/micro/kernels/l2_pool_2d.cc tensorflow/lite/micro/kernels/leaky_relu.cc tensorflow/lite/micro/kernels/logical.cc tensorflow/lite/micro/kernels/logical_common.cc tensorflow/lite/micro/kernels/logistic.cc tensorflow/lite/micro/kernels/logistic_common.cc tensorflow/lite/micro/kernels/log_softmax.cc tensorflow/lite/micro/kernels/maximum_minimum.cc tensorflow/lite/micro/kernels/mul.cc tensorflow/lite/micro/kernels/neg.cc tensorflow/lite/micro/kernels/pack.cc tensorflow/lite/micro/kernels/pad.cc tensorflow/lite/micro/kernels/pooling.cc tensorflow/lite/micro/kernels/pooling_common.cc tensorflow/lite/micro/kernels/prelu.cc tensorflow/lite/micro/kernels/quantize.cc tensorflow/lite/micro/kernels/quantize_common.cc tensorflow/lite/micro/kernels/reduce.cc tensorflow/lite/micro/kernels/reshape.cc tensorflow/lite/micro/kernels/resize_bilinear.cc tensorflow/lite/micro/kernels/resize_nearest_neighbor.cc tensorflow/lite/micro/kernels/round.cc tensorflow/lite/micro/kernels/shape.cc tensorflow/lite/micro/kernels/softmax.cc tensorflow/lite/micro/kernels/softmax_common.cc tensorflow/lite/micro/kernels/space_to_batch_nd.cc tensorflow/lite/micro/kernels/space_to_depth.cc tensorflow/lite/micro/kernels/split.cc tensorflow/lite/micro/kernels/split_v.cc tensorflow/lite/micro/kernels/squeeze.cc tensorflow/lite/micro/kernels/strided_slice.cc tensorflow/lite/micro/kernels/sub.cc tensorflow/lite/micro/kernels/svdf.cc tensorflow/lite/micro/kernels/svdf_common.cc tensorflow/lite/micro/kernels/tanh.cc tensorflow/lite/micro/kernels/transpose.cc tensorflow/lite/micro/kernels/transpose_conv.cc tensorflow/lite/micro/kernels/unpack.cc tensorflow/lite/micro/kernels/zeros_like.cc
   INCLUDE_DIRS . third_party/gemmlowp third_party/flatbuffers/include third_party/ruy)
 
 # Reduce the level of paranoia to be able to compile TF sources
@@ -32,7 +32,7 @@ target_compile_options(${COMPONENT_LIB} PRIVATE
   -Wno-missing-field-initializers
   -Wno-type-limits)
 
-target_compile_options(${COMPONENT_LIB} PRIVATE -fno-unwind-tables -ffunction-sections -fdata-sections -fmessage-length=0 -DTF_LITE_STATIC_MEMORY -DTF_LITE_DISABLE_X86_NEON -O3 -Werror -Wsign-compare -Wdouble-promotion -Wshadow -Wunused-variable -Wmissing-field-initializers -Wunused-function -Wswitch -Wvla -Wall -Wextra -Wstrict-aliasing -Wno-unused-parameter)
-target_compile_options(${COMPONENT_LIB} PRIVATE $<$<COMPILE_LANGUAGE:CXX>: -std=c++11 -fno-rtti -fno-exceptions -fno-threadsafe-statics -fno-unwind-tables -ffunction-sections -fdata-sections -fmessage-length=0 -DTF_LITE_STATIC_MEMORY -DTF_LITE_DISABLE_X86_NEON -O3 -Werror -Wsign-compare -Wdouble-promotion -Wshadow -Wunused-variable -Wmissing-field-initializers -Wunused-function -Wswitch -Wvla -Wall -Wextra -Wstrict-aliasing -Wno-unused-parameter   >)
+target_compile_options(${COMPONENT_LIB} PRIVATE -Wimplicit-function-declaration -Werror -fno-unwind-tables -ffunction-sections -fdata-sections -fmessage-length=0 -DTF_LITE_STATIC_MEMORY -DTF_LITE_DISABLE_X86_NEON -Wsign-compare -Wdouble-promotion -Wshadow -Wunused-variable -Wmissing-field-initializers -Wunused-function -Wswitch -Wvla -Wall -Wextra -Wstrict-aliasing -Wno-unused-parameter -DESP)
+target_compile_options(${COMPONENT_LIB} PRIVATE $<$<COMPILE_LANGUAGE:CXX>: -std=c++11 -fno-rtti -fno-exceptions -fno-threadsafe-statics -Werror -fno-unwind-tables -ffunction-sections -fdata-sections -fmessage-length=0 -DTF_LITE_STATIC_MEMORY -DTF_LITE_DISABLE_X86_NEON -Wsign-compare -Wdouble-promotion -Wshadow -Wunused-variable -Wmissing-field-initializers -Wunused-function -Wswitch -Wvla -Wall -Wextra -Wstrict-aliasing -Wno-unused-parameter -DESP >)
 target_compile_options(${COMPONENT_LIB} INTERFACE $<$<IN_LIST:-DTF_LITE_STATIC_MEMORY,$<TARGET_PROPERTY:${COMPONENT_LIB},COMPILE_OPTIONS>>:-DTF_LITE_STATIC_MEMORY>)
 target_link_libraries(${COMPONENT_LIB} PRIVATE -lm)

code/components/tfmicro/tensorflow/core/public/version.h

@@ -1,139 +0,0 @@
-/* Copyright 2015 The TensorFlow Authors. All Rights Reserved.
-
-Licensed under the Apache License, Version 2.0 (the "License");
-you may not use this file except in compliance with the License.
-You may obtain a copy of the License at
-
-    http://www.apache.org/licenses/LICENSE-2.0
-
-Unless required by applicable law or agreed to in writing, software
-distributed under the License is distributed on an "AS IS" BASIS,
-WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-See the License for the specific language governing permissions and
-limitations under the License.
-==============================================================================*/
-
-#ifndef TENSORFLOW_CORE_PUBLIC_VERSION_H_
-#define TENSORFLOW_CORE_PUBLIC_VERSION_H_
-
-// TensorFlow uses semantic versioning, see http://semver.org/.
-
-// Also update tensorflow/tensorflow.bzl and
-// tensorflow/tools/pip_package/setup.py
-#define TF_MAJOR_VERSION 2
-#define TF_MINOR_VERSION 5
-#define TF_PATCH_VERSION 0
-
-// TF_VERSION_SUFFIX is non-empty for pre-releases (e.g. "-alpha", "-alpha.1",
-// "-beta", "-rc", "-rc.1")
-#define TF_VERSION_SUFFIX ""
-
-#define TF_STR_HELPER(x) #x
-#define TF_STR(x) TF_STR_HELPER(x)
-
-// e.g. "0.5.0" or "0.6.0-alpha".
-#define TF_VERSION_STRING                                            \
-  (TF_STR(TF_MAJOR_VERSION) "." TF_STR(TF_MINOR_VERSION) "." TF_STR( \
-      TF_PATCH_VERSION) TF_VERSION_SUFFIX)
-
-// GraphDef compatibility versions (the versions field in graph.proto).
-//
-// Each graph has producer and min_consumer versions, and each
-// consumer has its own version and a min_producer.  In addition, graphs can
-// mark specific consumer versions as bad (to prevent bugs from executing).
-// A consumer will execute a graph if the consumer's version is at least the
-// graph's min_consumer, the graph's producer version is at least the consumer's
-// min_producer, and the consumer version isn't specifically disallowed by the
-// graph.
-//
-// By default, newly created graphs have producer version TF_GRAPH_DEF_VERSION
-// min_consumer TF_GRAPH_DEF_MIN_CONSUMER, and no other bad consumer versions.
-//
-// Version history:
-//
-// 0. Graphs created before GraphDef versioning
-// 1. First real version (2dec2015)
-// 2. adjust_contrast only takes float, doesn't perform clamping (11dec2015)
-// 3. Remove TileGrad, since it was equivalent to reduce_sum (30dec2015)
-// 4. When support for this version is removed, we can safely make AttrValue
-//    parsing more strict with respect to empty list values (see
-//    111635679, 7jan2016).
-// 5. Graphs are wholly-validated during Session::Create() (7jan2016).
-// 6. TensorFlow is scalar strict within Google (27jan2016).
-// 7. Remove TopK in favor of TopKV2 (5feb2016).
-// 8. Replace RandomCrop from C++ with pure Python (5feb2016).
-// 9. Deprecate batch_norm_with_global_normalization (16feb2016).
-// 10. Deprecate conv3d_backprop_{filter,input} (10jun2016).
-// 11. Deprecate {batch}_self_adjoint_eig (3aug2016).
-// 12. Graph consumers understand the node_def field of FunctionDef (22aug2016).
-// 13. Deprecate multiple batch linear algebra ops (9sep2016).
-// 14. Deprecate batch_matrix_* ops. (10sep2016).
-// 15. Deprecate batch_fft_* ops. (14sep2016).
-// 16. Deprecate tensor_array (v1) ops in favor of v2 (10nov2016).
-// 17. Deprecate inv (11nov2016).
-// 17. Expose reverse_v2 (10nov2016)
-// 18. Add VariableV2 (30nov2016)
-// 19. Deprecated ops created by models moved out of core SkipGram, NegTrain.
-//     (08dec2016)
-// 20. Catch all version 1.0 changes to Python API generation. SplitV is now
-//     used for tf.split, ReverseV2 is now used by tf.reverse, ConcatV2 is
-//     now used by tf.concat. Graphs use flooring
-//     division and mod semantics. TensorArrayV3. (12dec2016)
-//     Also considered the version for when it is required for reduction
-//     ops' indices to be scalar or vector, and not higher rank.
-//     Some earlier graph def versions allowed this.
-// 21. Dropped FunctionDef.Node support, switched to node_def introduced
-//     in version 12. (11jan2017)
-// 22. Placeholder now can specify and enforce scalar and partial
-//     shapes, particularly when restoring a graph from GraphDef
-//     produced at version 22 or later.  (04/10/2016)
-// 23. Remove NonMaxSuppression in favor of NonMaxSuppressionV2.
-// 24. Deprecate lookup ops (v1) ops in favor of v2 (30may2017)
-// 25. Deprecate stack (v1) ops in favor of v2 (2017/6/15).
-// 25. Deprecate RandomPoisson (v1) ops in favor of v2 (2017/10/25).
-// 26. Add a bool 'stripped_default_attrs' to MetaInfoDef indicating
-//     whether default-valued attrs have been stripped from the nodes in the
-//     GraphDef. (7dec2017)
-// 27. Deprecate TensorArray ops v2 in favor of v3 and deprecated io_ops
-//     deprecated in favor of V2 ops. (2018/01/23)
-// 28. Deprecate MatrixExponential op in favor of Python implementation.
-//     (2018/08/21).
-// (2019/02/15). Added `control_ret` field to FunctionDef proto, and
-//     `control_output` field to OpDef proto.
-// 29. Deprecate StatefulStandardNormal op in favor of StatefulStandardNormalV2.
-//     (2019/03/25).
-// (2019/04/17). Added `arg_attr` field to FunctionDefProto.
-// 30. (2019/05/09) First date based GraphDef version. GraphDef
-//     versions advance by 1 each day after this point.
-
-#define TF_GRAPH_DEF_VERSION_MIN_PRODUCER 0
-#define TF_GRAPH_DEF_VERSION_MIN_CONSUMER 0
-#define TF_GRAPH_DEF_VERSION 578  // Updated: 2020/11/7
-
-// Checkpoint compatibility versions (the versions field in SavedSliceMeta).
-//
-// The checkpoint versions have the same semantics as GraphDef versions, but the
-// numbering scheme is separate.  We have no plans to ever deprecate checkpoint
-// versions, but it's good to have this in place in case we ever need to.
-//
-// Version history:
-//
-// 0. Checkpoints saved before checkpoint versioning.
-// 1. First real version (10feb2015).
-#define TF_CHECKPOINT_VERSION_MIN_PRODUCER 0
-#define TF_CHECKPOINT_VERSION_MIN_CONSUMER 0
-#define TF_CHECKPOINT_VERSION 1
-
-/// Version query functions (defined in generated version_info.cc)
-
-// Host compiler version (declared elsewhere to be __VERSION__)
-extern const char* tf_compiler_version();
-// The git commit designator when tensorflow was built
-// If no git repository, this will be "internal".
-extern const char* tf_git_version();
-// Value of the _GLIBCXX_USE_CXX11_ABI flag, or 0 if it's not set.
-extern int tf_cxx11_abi_flag();
-// Returns 1 if build is monolithic, or 0 otherwise.
-extern int tf_monolithic_build();
-
-#endif  // TENSORFLOW_CORE_PUBLIC_VERSION_H_

code/components/tfmicro/tensorflow/lite/c/builtin_op_data.h

@@ -63,13 +63,11 @@ typedef struct {
 } TfLiteMirrorPaddingParams;
 
 // Possible fused activation functions.
-// TODO(aselle): rename to TfLiteActivation
 typedef enum {
   kTfLiteActNone = 0,
   kTfLiteActRelu,
-  kTfLiteActReluN1To1,                    // min(max(-1, x), 1)
-  kTfLiteActRelu1 = kTfLiteActReluN1To1,  // kTfLiteActRelu1 will be deprecated.
-  kTfLiteActRelu6,                        // min(max(0, x), 6)
+  kTfLiteActReluN1To1,  // min(max(-1, x), 1)
+  kTfLiteActRelu6,      // min(max(0, x), 6)
   kTfLiteActTanh,
   kTfLiteActSignBit,
   kTfLiteActSigmoid,
@@ -88,6 +86,19 @@ typedef struct {
   int dilation_height_factor;
 } TfLiteConvParams;
 
+typedef struct {
+  TfLitePadding padding;
+  int stride_width;
+  int stride_height;
+  int stride_depth;
+  int dilation_width_factor;
+  int dilation_height_factor;
+  int dilation_depth_factor;
+  TfLiteFusedActivation activation;
+} TfLiteConv3DParams;
+
+typedef TfLiteConv3DParams TfLiteConv3DTransposeParams;
+
 typedef struct {
   TfLitePadding padding;
   int stride_width;
@@ -214,6 +225,10 @@ typedef struct {
 typedef struct {
   bool adj_x;
   bool adj_y;
+  // Parameters for BatchMatMul version 4 or above.
+  // If set to true and the weights are quantized, then non constant inputs
+  // are quantized at evaluation time with asymmetric quantization.
+  bool asymmetric_quantize_inputs;
 } TfLiteBatchMatMulParams;
 
 typedef struct {
@@ -314,8 +329,9 @@ typedef struct {
 } TfLitePadV2Params;
 
 typedef struct {
-  // TODO(ahentz): We can't have dynamic data in this struct, at least not yet.
-  // For now we will fix the maximum possible number of dimensions.
+  // These fields are only used in old models for backward compatibility.
+  // In the current implementation, we use the 2nd input of the op as the shape,
+  // and these fields are unused.
   int shape[TFLITE_RESHAPE_PARAMS_MAX_DIMENSION_COUNT];
   int num_dimensions;
 } TfLiteReshapeParams;
@@ -351,6 +367,7 @@ typedef struct {
 
 typedef struct {
   int axis;
+  int batch_dims;
 } TfLiteGatherParams;
 
 typedef struct {
@@ -474,6 +491,17 @@ typedef struct {
   int init_subgraph_index;
 } TfLiteCallOnceParams;
 
+typedef struct {
+  int table_id;
+  TfLiteType key_dtype;
+  TfLiteType value_dtype;
+} TfLiteHashtableParams;
+
+typedef struct {
+  const char* container;
+  const char* shared_name;
+} TfLiteVarHandleParams;
+
 #ifdef __cplusplus
 }  // extern "C"
 #endif  // __cplusplus

code/components/tfmicro/tensorflow/lite/c/c_api_types.h

@@ -0,0 +1,109 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+// This file declares types used by the pure C inference API defined in c_api.h,
+// some of which are also used in the C++ and C kernel and interpreter APIs.
+
+#ifndef TENSORFLOW_LITE_C_C_API_TYPES_H_
+#define TENSORFLOW_LITE_C_C_API_TYPES_H_
+
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Define TFL_CAPI_EXPORT macro to export a function properly with a shared
+// library.
+#ifdef SWIG
+#define TFL_CAPI_EXPORT
+#elif defined(TFL_STATIC_LIBRARY_BUILD)
+#define TFL_CAPI_EXPORT
+#else  // not definded TFL_STATIC_LIBRARY_BUILD
+#if defined(_WIN32)
+#ifdef TFL_COMPILE_LIBRARY
+#define TFL_CAPI_EXPORT __declspec(dllexport)
+#else
+#define TFL_CAPI_EXPORT __declspec(dllimport)
+#endif  // TFL_COMPILE_LIBRARY
+#else
+#define TFL_CAPI_EXPORT __attribute__((visibility("default")))
+#endif  // _WIN32
+#endif  // SWIG
+
+typedef enum TfLiteStatus {
+  kTfLiteOk = 0,
+
+  // Generally referring to an error in the runtime (i.e. interpreter)
+  kTfLiteError = 1,
+
+  // Generally referring to an error from a TfLiteDelegate itself.
+  kTfLiteDelegateError = 2,
+
+  // Generally referring to an error in applying a delegate due to
+  // incompatibility between runtime and delegate, e.g., this error is returned
+  // when trying to apply a TfLite delegate onto a model graph that's already
+  // immutable.
+  kTfLiteApplicationError = 3,
+
+  // Generally referring to serialized delegate data not being found.
+  // See tflite::delegates::Serialization.
+  kTfLiteDelegateDataNotFound = 4,
+
+  // Generally referring to data-writing issues in delegate serialization.
+  // See tflite::delegates::Serialization.
+  kTfLiteDelegateDataWriteError = 5,
+
+  // Generally referring to data-reading issues in delegate serialization.
+  // See tflite::delegates::Serialization.
+  kTfLiteDelegateDataReadError = 5,
+} TfLiteStatus;
+
+// Types supported by tensor
+typedef enum {
+  kTfLiteNoType = 0,
+  kTfLiteFloat32 = 1,
+  kTfLiteInt32 = 2,
+  kTfLiteUInt8 = 3,
+  kTfLiteInt64 = 4,
+  kTfLiteString = 5,
+  kTfLiteBool = 6,
+  kTfLiteInt16 = 7,
+  kTfLiteComplex64 = 8,
+  kTfLiteInt8 = 9,
+  kTfLiteFloat16 = 10,
+  kTfLiteFloat64 = 11,
+  kTfLiteComplex128 = 12,
+  kTfLiteUInt64 = 13,
+  kTfLiteResource = 14,
+  kTfLiteVariant = 15,
+  kTfLiteUInt32 = 16,
+} TfLiteType;
+
+// Legacy. Will be deprecated in favor of TfLiteAffineQuantization.
+// If per-layer quantization is specified this field will still be populated in
+// addition to TfLiteAffineQuantization.
+// Parameters for asymmetric quantization. Quantized values can be converted
+// back to float using:
+//     real_value = scale * (quantized_value - zero_point)
+typedef struct TfLiteQuantizationParams {
+  float scale;
+  int32_t zero_point;
+} TfLiteQuantizationParams;
+
+#ifdef __cplusplus
+}  // extern C
+#endif
+#endif  // TENSORFLOW_LITE_C_C_API_TYPES_H_

code/components/tfmicro/tensorflow/lite/c/common.c

@@ -14,6 +14,8 @@ limitations under the License.
 ==============================================================================*/
 
 #include "tensorflow/lite/c/common.h"
+#include "tensorflow/lite/c/c_api_types.h"
+
 #ifndef TF_LITE_STATIC_MEMORY
 #include <stdlib.h>
 #include <string.h>
@@ -43,8 +45,10 @@ int TfLiteIntArrayEqualsArray(const TfLiteIntArray* a, int b_size,
 #ifndef TF_LITE_STATIC_MEMORY
 
 TfLiteIntArray* TfLiteIntArrayCreate(int size) {
-  TfLiteIntArray* ret =
-      (TfLiteIntArray*)malloc(TfLiteIntArrayGetSizeInBytes(size));
+  int alloc_size = TfLiteIntArrayGetSizeInBytes(size);
+  if (alloc_size <= 0) return NULL;
+  TfLiteIntArray* ret = (TfLiteIntArray*)malloc(alloc_size);
+  if (!ret) return ret;
   ret->size = size;
   return ret;
 }
@@ -179,9 +183,9 @@ void TfLiteTensorRealloc(size_t num_bytes, TfLiteTensor* tensor) {
   }
   // TODO(b/145340303): Tensor data should be aligned.
   if (!tensor->data.raw) {
-    tensor->data.raw = malloc(num_bytes);
+    tensor->data.raw = (char*)malloc(num_bytes);
   } else if (num_bytes > tensor->bytes) {
-    tensor->data.raw = realloc(tensor->data.raw, num_bytes);
+    tensor->data.raw = (char*)realloc(tensor->data.raw, num_bytes);
   }
   tensor->bytes = num_bytes;
 }
@@ -197,12 +201,16 @@ const char* TfLiteTypeGetName(TfLiteType type) {
       return "INT16";
     case kTfLiteInt32:
       return "INT32";
+    case kTfLiteUInt32:
+      return "UINT32";
     case kTfLiteUInt8:
       return "UINT8";
     case kTfLiteInt8:
       return "INT8";
     case kTfLiteInt64:
       return "INT64";
+    case kTfLiteUInt64:
+      return "UINT64";
     case kTfLiteBool:
       return "BOOL";
     case kTfLiteComplex64:
@@ -215,11 +223,15 @@ const char* TfLiteTypeGetName(TfLiteType type) {
       return "FLOAT16";
     case kTfLiteFloat64:
       return "FLOAT64";
+    case kTfLiteResource:
+      return "RESOURCE";
+    case kTfLiteVariant:
+      return "VARIANT";
   }
   return "Unknown type";
 }
 
-TfLiteDelegate TfLiteDelegateCreate() {
+TfLiteDelegate TfLiteDelegateCreate(void) {
   TfLiteDelegate d = {
       .data_ = NULL,
       .Prepare = NULL,

code/components/tfmicro/tensorflow/lite/c/common.h

@@ -40,26 +40,12 @@ limitations under the License.
 #include <stddef.h>
 #include <stdint.h>
 
+#include "tensorflow/lite/c/c_api_types.h"  // IWYU pragma: export
+
 #ifdef __cplusplus
 extern "C" {
 #endif  // __cplusplus
 
-typedef enum TfLiteStatus {
-  kTfLiteOk = 0,
-
-  // Generally referring to an error in the runtime (i.e. interpreter)
-  kTfLiteError = 1,
-
-  // Generally referring to an error from a TfLiteDelegate itself.
-  kTfLiteDelegateError = 2,
-
-  // Generally referring to an error in applying a delegate due to
-  // incompatibility between runtime and delegate, e.g., this error is returned
-  // when trying to apply a TfLite delegate onto a model graph that's already
-  // immutable.
-  kTfLiteApplicationError = 3
-} TfLiteStatus;
-
 // The list of external context types known to TF Lite. This list exists solely
 // to avoid conflicts and to ensure ops can share the external contexts they
 // need. Access to the external contexts is controlled by one of the
@@ -80,7 +66,7 @@ struct TfLiteRegistration;
 
 // An external context is a collection of information unrelated to the TF Lite
 // framework, but useful to a subset of the ops. TF Lite knows very little
-// about about the actual contexts, but it keeps a list of them, and is able to
+// about the actual contexts, but it keeps a list of them, and is able to
 // refresh them if configurations like the number of recommended threads
 // change.
 typedef struct TfLiteExternalContext {
@@ -98,7 +84,8 @@ typedef struct TfLiteIntArray {
 // https://github.com/google/re2/commit/b94b7cd42e9f02673cd748c1ac1d16db4052514c
 #if (!defined(__clang__) && defined(__GNUC__) && __GNUC__ == 6 && \
      __GNUC_MINOR__ >= 1) ||                                      \
-    defined(HEXAGON) || (__clang_major__ == 7 && __clang_minor__ == 1)
+    defined(HEXAGON) ||                                           \
+    (defined(__clang__) && __clang_major__ == 7 && __clang_minor__ == 1)
   int data[0];
 #else
   int data[];
@@ -254,22 +241,6 @@ void TfLiteFloatArrayFree(TfLiteFloatArray* a);
     }                                      \
   } while (0)
 
-// Define TFL_CAPI_EXPORT macro to export a function properly with a shared
-// library.
-#ifdef SWIG
-#define TFL_CAPI_EXPORT
-#else
-#if defined(_WIN32)
-#ifdef TFL_COMPILE_LIBRARY
-#define TFL_CAPI_EXPORT __declspec(dllexport)
-#else
-#define TFL_CAPI_EXPORT __declspec(dllimport)
-#endif  // TFL_COMPILE_LIBRARY
-#else
-#define TFL_CAPI_EXPORT __attribute__((visibility("default")))
-#endif  // _WIN32
-#endif  // SWIG
-
 // Single-precision complex data type compatible with the C99 definition.
 typedef struct TfLiteComplex64 {
   float re, im;  // real and imaginary parts, respectively.
@@ -285,23 +256,6 @@ typedef struct TfLiteFloat16 {
   uint16_t data;
 } TfLiteFloat16;
 
-// Types supported by tensor
-typedef enum {
-  kTfLiteNoType = 0,
-  kTfLiteFloat32 = 1,
-  kTfLiteInt32 = 2,
-  kTfLiteUInt8 = 3,
-  kTfLiteInt64 = 4,
-  kTfLiteString = 5,
-  kTfLiteBool = 6,
-  kTfLiteInt16 = 7,
-  kTfLiteComplex64 = 8,
-  kTfLiteInt8 = 9,
-  kTfLiteFloat16 = 10,
-  kTfLiteFloat64 = 11,
-  kTfLiteComplex128 = 12,
-} TfLiteType;
-
 // Return the name of a given type, for error reporting purposes.
 const char* TfLiteTypeGetName(TfLiteType type);
 
@@ -318,22 +272,12 @@ typedef enum TfLiteQuantizationType {
 typedef struct TfLiteQuantization {
   // The type of quantization held by params.
   TfLiteQuantizationType type;
-  // Holds a reference to one of the quantization param structures specified
-  // below.
+  // Holds an optional reference to a quantization param structure. The actual
+  // type depends on the value of the `type` field (see the comment there for
+  // the values and corresponding types).
   void* params;
 } TfLiteQuantization;
 
-// Legacy. Will be deprecated in favor of TfLiteAffineQuantization.
-// If per-layer quantization is specified this field will still be populated in
-// addition to TfLiteAffineQuantization.
-// Parameters for asymmetric quantization. Quantized values can be converted
-// back to float using:
-//     real_value = scale * (quantized_value - zero_point)
-typedef struct TfLiteQuantizationParams {
-  float scale;
-  int32_t zero_point;
-} TfLiteQuantizationParams;
-
 // Parameters for asymmetric quantization across a dimension (i.e per output
 // channel quantization).
 // quantized_dimension specifies which dimension the scales and zero_points
@@ -353,7 +297,9 @@ typedef union TfLitePtrUnion {
    * GetTensorData<TYPE>(tensor) instead, otherwise only access .data, as other
    * members are deprecated. */
   int32_t* i32;
+  uint32_t* u32;
   int64_t* i64;
+  uint64_t* u64;
   float* f;
   TfLiteFloat16* f16;
   double* f64;
@@ -430,6 +376,17 @@ typedef struct TfLiteCustomAllocation {
   size_t bytes;
 } TfLiteCustomAllocation;
 
+// The flags used in `Interpreter::SetCustomAllocationForTensor`.
+// Note that this is a bitmask, so the values should be 1, 2, 4, 8, ...etc.
+typedef enum TfLiteCustomAllocationFlags {
+  kTfLiteCustomAllocationFlagsNone = 0,
+  // Skips checking whether allocation.data points to an aligned buffer as
+  // expected by the TFLite runtime.
+  // NOTE: Setting this flag can cause crashes when calling Invoke().
+  // Use with caution.
+  kTfLiteCustomAllocationFlagsSkipAlignCheck = 1,
+} TfLiteCustomAllocationFlags;
+
 // A tensor in the interpreter system which is a wrapper around a buffer of
 // data including a dimensionality (or NULL if not currently defined).
 #ifndef TF_LITE_STATIC_MEMORY
@@ -499,8 +456,8 @@ typedef struct TfLiteTensor {
 } TfLiteTensor;
 
 // A structure representing an instance of a node.
-// This structure only exhibits the inputs, outputs and user defined data, not
-// other features like the type.
+// This structure only exhibits the inputs, outputs, user defined data and some
+// node properties (like statefulness), not other features like the type.
 typedef struct TfLiteNode {
   // Inputs to this node expressed as indices into the simulator's tensors.
   TfLiteIntArray* inputs;
@@ -533,8 +490,11 @@ typedef struct TfLiteNode {
   // created by calling `interpreter.ModifyGraphWithDelegate`.
   // WARNING: This is an experimental interface that is subject to change.
   struct TfLiteDelegate* delegate;
+
+  // Whether this op might have side effect (e.g. stateful op).
+  bool might_have_side_effect;
 } TfLiteNode;
-#else  // defined(TF_LITE_STATIC_MEMORY)?
+#else   // defined(TF_LITE_STATIC_MEMORY)?
 // NOTE: This flag is opt-in only at compile time.
 //
 // Specific reduced TfLiteTensor struct for TF Micro runtime. This struct
@@ -683,6 +643,7 @@ typedef struct TfLiteContext {
   // TfLiteDelegates can traverse the current execution plan by iterating
   // through each member of this array and using GetNodeAndRegistration() to
   // access details about a node. i.e.
+  //
   // TfLiteIntArray* execution_plan;
   // TF_LITE_ENSURE_STATUS(context->GetExecutionPlan(context, &execution_plan));
   // for (int exec_index = 0; exec_index < execution_plan->size; exec_index++) {
@@ -691,6 +652,28 @@ typedef struct TfLiteContext {
   //    TfLiteRegistration* reg;
   //    context->GetNodeAndRegistration(context, node_index, &node, &reg);
   // }
+  // Note: the memory pointed by '`*execution_plan` is OWNED by TfLite runtime.
+  // Future calls to GetExecutionPlan invalidates earlier outputs. The following
+  // code snippet shows the issue of such an invocation pattern. After calling
+  // CheckNode, subsequent access to `plan_1st` is undefined.
+  //
+  // void CheckNode(const TfLiteNode* node) {
+  //   ...
+  //   TfLiteIntArray* plan_2nd;
+  //   TF_LITE_ENSURE_STATUS(context->GetExecutionPlan(context, &plan_2nd));
+  //   ...
+  // }
+  //
+  // TfLiteIntArray* plan_1st;
+  // TF_LITE_ENSURE_STATUS(context->GetExecutionPlan(context, &plan_1st));
+  // for (int exec_index = 0; exec_index < plan_1st->size; exec_index++) {
+  //    int node_index = plan_1st->data[exec_index];
+  //    TfLiteNode* node;
+  //    TfLiteRegistration* reg;
+  //    context->GetNodeAndRegistration(context, node_index, &node, &reg);
+  //    CheckNode(node);
+  // }
+  //
   // WARNING: This is an experimental interface that is subject to change.
   TfLiteStatus (*GetExecutionPlan)(struct TfLiteContext* context,
                                    TfLiteIntArray** execution_plan);
@@ -820,6 +803,18 @@ typedef struct TfLiteContext {
   // WARNING: This method may not be available on all platforms.
   TfLiteEvalTensor* (*GetEvalTensor)(const struct TfLiteContext* context,
                                      int tensor_idx);
+
+  // Retrieves named metadata buffer from the TFLite model.
+  // Returns kTfLiteOk if metadata is successfully obtained from the flatbuffer
+  // Model: that is, there exists a `metadata` entry with given `name` string.
+  // (see TFLite's schema.fbs).
+  // The corresponding `buffer` information is populated in `ptr` & `bytes`.
+  // The data from `ptr` is valid for the lifetime of the Interpreter.
+  //
+  // WARNING: This is an experimental interface that is subject to change.
+  TfLiteStatus (*GetModelMetadata)(const struct TfLiteContext* context,
+                                   const char* name, const char** ptr,
+                                   size_t* bytes);
 } TfLiteContext;
 
 typedef struct TfLiteRegistration {
@@ -961,7 +956,7 @@ typedef struct TfLiteDelegate {
 
 // Build a 'null' delegate, with all the fields properly set to their default
 // values.
-TfLiteDelegate TfLiteDelegateCreate();
+TfLiteDelegate TfLiteDelegateCreate(void);
 
 #ifdef __cplusplus
 }  // extern "C"

code/components/tfmicro/tensorflow/lite/core/api/flatbuffer_conversions.cc

@@ -169,6 +169,10 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseAdd(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_ADD_N: {
+      return ParseAddN(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_ARG_MAX: {
       return ParseArgMax(op, error_reporter, allocator, builtin_data);
     }
@@ -181,6 +185,14 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParsePool(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_BATCH_MATMUL: {
+      return ParseBatchMatMul(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_BATCH_TO_SPACE_ND: {
+      return ParseBatchToSpaceNd(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_CEIL: {
       return ParseCeil(op, error_reporter, allocator, builtin_data);
     }
@@ -193,6 +205,14 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseConv2D(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_CUMSUM: {
+      return ParseCumsum(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_DEPTH_TO_SPACE: {
+      return ParseDepthToSpace(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_DEPTHWISE_CONV_2D: {
       return ParseDepthwiseConv2D(op, error_reporter, allocator, builtin_data);
     }
@@ -201,14 +221,46 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseDequantize(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_DIV: {
+      return ParseDiv(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_ELU: {
+      return ParseElu(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_EXP: {
+      return ParseExp(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_EXPAND_DIMS: {
+      return ParseExpandDims(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_FILL: {
+      return ParseFill(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_FLOOR: {
       return ParseFloor(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_FLOOR_DIV: {
+      return ParseFloorDiv(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_FLOOR_MOD: {
+      return ParseFloorMod(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_FULLY_CONNECTED: {
       return ParseFullyConnected(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_GATHER_ND: {
+      return ParseGatherNd(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_GREATER: {
       return ParseGreater(op, error_reporter, allocator, builtin_data);
     }
@@ -229,6 +281,10 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParsePool(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_LEAKY_RELU: {
+      return ParseLeakyRelu(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_LESS: {
       return ParseLess(op, error_reporter, allocator, builtin_data);
     }
@@ -257,6 +313,10 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseLogistic(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_LOG_SOFTMAX: {
+      return ParseLogSoftmax(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_MAXIMUM: {
       return ParseMaximum(op, error_reporter, allocator, builtin_data);
     }
@@ -297,6 +357,10 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParsePadV2(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_POW: {
+      return ParsePow(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_PRELU: {
       return ParsePrelu(op, error_reporter, allocator, builtin_data);
     }
@@ -309,6 +373,10 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseReducer(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_REDUCE_ALL: {
+      return ParseReducer(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_REDUCE_MAX: {
       return ParseReducer(op, error_reporter, allocator, builtin_data);
     }
@@ -362,6 +430,14 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseSoftmax(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_SPACE_TO_BATCH_ND: {
+      return ParseSpaceToBatchNd(op, error_reporter, allocator, builtin_data);
+    }
+
+    case BuiltinOperator_SPACE_TO_DEPTH: {
+      return ParseSpaceToDepth(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_SPLIT: {
       return ParseSplit(op, error_reporter, allocator, builtin_data);
     }
@@ -378,6 +454,10 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseSquare(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_SQUEEZE: {
+      return ParseSqueeze(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_STRIDED_SLICE: {
       return ParseStridedSlice(op, error_reporter, allocator, builtin_data);
     }
@@ -398,23 +478,20 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return ParseTanh(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_TRANSPOSE_CONV: {
+      return ParseTransposeConv(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_UNPACK: {
       return ParseUnpack(op, error_reporter, allocator, builtin_data);
     }
 
+    case BuiltinOperator_ZEROS_LIKE: {
+      return ParseZerosLike(op, error_reporter, allocator, builtin_data);
+    }
+
     case BuiltinOperator_CAST: {
-      auto params = safe_allocator.Allocate<TfLiteCastParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* schema_params = op->builtin_options_as_CastOptions()) {
-        TF_LITE_ENSURE_STATUS(ConvertTensorType(schema_params->in_data_type(),
-                                                &params->in_data_type,
-                                                error_reporter));
-        TF_LITE_ENSURE_STATUS(ConvertTensorType(schema_params->out_data_type(),
-                                                &params->out_data_type,
-                                                error_reporter));
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
+      return ParseCast(op, error_reporter, allocator, builtin_data);
     }
     case BuiltinOperator_LSH_PROJECTION: {
       auto params = safe_allocator.Allocate<TfLiteLSHProjectionParams>();
@@ -483,16 +560,7 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
     case BuiltinOperator_HASHTABLE_LOOKUP:
       // no-op.
       return kTfLiteOk;
-    case BuiltinOperator_DIV: {
-      auto params = safe_allocator.Allocate<TfLiteDivParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* schema_params = op->builtin_options_as_DivOptions()) {
-        params->activation =
-            ConvertActivation(schema_params->fused_activation_function());
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
+
     case BuiltinOperator_LOCAL_RESPONSE_NORMALIZATION: {
       auto params = safe_allocator.Allocate<TfLiteLocalResponseNormParams>();
       TF_LITE_ENSURE(error_reporter, params != nullptr);
@@ -584,66 +652,9 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       *builtin_data = params.release();
       return kTfLiteOk;
     }
-    case BuiltinOperator_SPACE_TO_DEPTH: {
-      auto params = safe_allocator.Allocate<TfLiteSpaceToDepthParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* schema_params =
-              op->builtin_options_as_SpaceToDepthOptions()) {
-        params->block_size = schema_params->block_size();
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
-    case BuiltinOperator_DEPTH_TO_SPACE: {
-      auto params = safe_allocator.Allocate<TfLiteDepthToSpaceParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* schema_params =
-              op->builtin_options_as_DepthToSpaceOptions()) {
-        params->block_size = schema_params->block_size();
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
+
     case BuiltinOperator_GATHER: {
-      auto params = safe_allocator.Allocate<TfLiteGatherParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      params->axis = 0;
-      if (const auto* gather_params = op->builtin_options_as_GatherOptions()) {
-        params->axis = gather_params->axis();
-      }
-
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
-
-    case BuiltinOperator_SQUEEZE: {
-      auto params = safe_allocator.Allocate<TfLiteSqueezeParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* schema_params = op->builtin_options_as_SqueezeOptions()) {
-        const auto* squeeze_dims = schema_params->squeeze_dims();
-        if (squeeze_dims != nullptr) {
-          TF_LITE_ENSURE_STATUS(FlatBufferIntVectorToArray(
-              sizeof(params->squeeze_dims), squeeze_dims, params->squeeze_dims,
-              error_reporter, "squeeze"));
-          params->num_squeeze_dims = squeeze_dims->size();
-        } else {
-          params->num_squeeze_dims = 0;
-        }
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
-    case BuiltinOperator_TRANSPOSE_CONV: {
-      auto params = safe_allocator.Allocate<TfLiteTransposeConvParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* transpose_conv_params =
-              op->builtin_options_as_TransposeConvOptions()) {
-        params->padding = ConvertPadding(transpose_conv_params->padding());
-        params->stride_width = transpose_conv_params->stride_w();
-        params->stride_height = transpose_conv_params->stride_h();
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
+      return ParseGather(op, error_reporter, allocator, builtin_data);
     }
     case BuiltinOperator_SPARSE_TO_DENSE: {
       auto params = safe_allocator.Allocate<TfLiteSparseToDenseParams>();
@@ -656,7 +667,6 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       return kTfLiteOk;
     }
     case BuiltinOperator_DELEGATE: {
-      // TODO(ycling): Revisit when supporting saving delegated models.
       TF_LITE_REPORT_ERROR(error_reporter,
                            "DELEGATE op shouldn't exist in model.");
       return kTfLiteError;
@@ -683,16 +693,6 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       *builtin_data = params.release();
       return kTfLiteOk;
     }
-    case BuiltinOperator_LEAKY_RELU: {
-      auto params = safe_allocator.Allocate<TfLiteLeakyReluParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* leaky_relu_params =
-              op->builtin_options_as_LeakyReluOptions()) {
-        params->alpha = leaky_relu_params->alpha();
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
     case BuiltinOperator_MIRROR_PAD: {
       auto params = safe_allocator.Allocate<TfLiteMirrorPaddingParams>();
       TF_LITE_ENSURE(error_reporter, params != nullptr);
@@ -750,17 +750,6 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       *builtin_data = params.release();
       return kTfLiteOk;
     }
-    case BuiltinOperator_BATCH_MATMUL: {
-      auto params = safe_allocator.Allocate<TfLiteBatchMatMulParams>();
-      TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* bmm_params =
-              op->builtin_options_as_BatchMatMulOptions()) {
-        params->adj_x = bmm_params->adj_x();
-        params->adj_y = bmm_params->adj_y();
-      }
-      *builtin_data = params.release();
-      return kTfLiteOk;
-    }
     case BuiltinOperator_CALL_ONCE: {
       auto params = safe_allocator.Allocate<TfLiteCallOnceParams>();
       TF_LITE_ENSURE(error_reporter, params != nullptr);
@@ -771,50 +760,75 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
       *builtin_data = params.release();
       return kTfLiteOk;
     }
-    case BuiltinOperator_CUMSUM: {
-      auto params = safe_allocator.Allocate<TfLiteCumsumParams>();
+    case BuiltinOperator_CONV_3D:
+    case BuiltinOperator_CONV_3D_TRANSPOSE: {
+      auto params = safe_allocator.Allocate<TfLiteConv3DParams>();
       TF_LITE_ENSURE(error_reporter, params != nullptr);
-      if (const auto* cumsum_params = op->builtin_options_as_CumsumOptions()) {
-        params->exclusive = cumsum_params->exclusive();
-        params->reverse = cumsum_params->reverse();
+      if (const auto* conv3d_params = op->builtin_options_as_Conv3DOptions()) {
+        params->padding = ConvertPadding(conv3d_params->padding());
+        params->activation =
+            ConvertActivation(conv3d_params->fused_activation_function());
+        params->stride_depth = conv3d_params->stride_d();
+        params->stride_height = conv3d_params->stride_h();
+        params->stride_width = conv3d_params->stride_w();
+        params->dilation_depth_factor = conv3d_params->dilation_d_factor();
+        params->dilation_height_factor = conv3d_params->dilation_h_factor();
+        params->dilation_width_factor = conv3d_params->dilation_w_factor();
+      }
+      *builtin_data = params.release();
+      return kTfLiteOk;
+    }
+    case BuiltinOperator_HASHTABLE: {
+      auto params = safe_allocator.Allocate<TfLiteHashtableParams>();
+      TF_LITE_ENSURE(error_reporter, params != nullptr);
+      if (const auto* hashtable_params =
+              op->builtin_options_as_HashtableOptions()) {
+        params->table_id = hashtable_params->table_id();
+        TF_LITE_ENSURE_STATUS(ConvertTensorType(
+            hashtable_params->key_dtype(), &params->key_dtype, error_reporter));
+        TF_LITE_ENSURE_STATUS(ConvertTensorType(hashtable_params->value_dtype(),
+                                                &params->value_dtype,
+                                                error_reporter));
+      }
+      *builtin_data = params.release();
+      return kTfLiteOk;
+    }
+    case BuiltinOperator_VAR_HANDLE: {
+      auto params = safe_allocator.Allocate<TfLiteVarHandleParams>();
+      TF_LITE_ENSURE(error_reporter, params != nullptr);
+      params->container = nullptr;
+      params->shared_name = nullptr;
+      if (const auto* var_handle_params =
+              op->builtin_options_as_VarHandleOptions()) {
+        if (var_handle_params->container())
+          params->container = var_handle_params->container()->c_str();
+        if (var_handle_params->shared_name())
+          params->shared_name = var_handle_params->shared_name()->c_str();
       }
       *builtin_data = params.release();
       return kTfLiteOk;
     }
     // Below are the ops with no builtin_data structure.
-    case BuiltinOperator_BATCH_TO_SPACE_ND:
     // TODO(aselle): Implement call in BuiltinOptions, but nullptrs are
     // ok for now, since there is no call implementation either.
     case BuiltinOperator_CALL:
     case BuiltinOperator_CONCAT_EMBEDDINGS:
     case BuiltinOperator_COS:
     case BuiltinOperator_CUSTOM:
-    case BuiltinOperator_ELU:
     case BuiltinOperator_EMBEDDING_LOOKUP:
     case BuiltinOperator_EQUAL:
-    case BuiltinOperator_EXP:
-    case BuiltinOperator_EXPAND_DIMS:
-    case BuiltinOperator_LOG_SOFTMAX:
     case BuiltinOperator_MATRIX_DIAG:
     case BuiltinOperator_MATRIX_SET_DIAG:
     case BuiltinOperator_RELU_N1_TO_1:
     case BuiltinOperator_SELECT:
     case BuiltinOperator_SELECT_V2:
     case BuiltinOperator_SLICE:
-    case BuiltinOperator_SPACE_TO_BATCH_ND:
     case BuiltinOperator_TILE:
     case BuiltinOperator_TOPK_V2:
     case BuiltinOperator_TRANSPOSE:
-    case BuiltinOperator_POW:
-    case BuiltinOperator_FLOOR_DIV:
-    case BuiltinOperator_ZEROS_LIKE:
-    case BuiltinOperator_FILL:
-    case BuiltinOperator_FLOOR_MOD:
     case BuiltinOperator_RANGE:
     case BuiltinOperator_SQUARED_DIFFERENCE:
     case BuiltinOperator_REVERSE_V2:
-    case BuiltinOperator_ADD_N:
-    case BuiltinOperator_GATHER_ND:
     case BuiltinOperator_WHERE:
     case BuiltinOperator_RANK:
     case BuiltinOperator_NON_MAX_SUPPRESSION_V4:
@@ -823,6 +837,16 @@ TfLiteStatus ParseOpDataTfLite(const Operator* op, BuiltinOperator op_type,
     case BuiltinOperator_DENSIFY:
     case BuiltinOperator_SEGMENT_SUM:
     case BuiltinOperator_BROADCAST_TO:
+    case BuiltinOperator_RFFT2D:
+    case BuiltinOperator_IMAG:
+    case BuiltinOperator_REAL:
+    case BuiltinOperator_COMPLEX_ABS:
+    case BuiltinOperator_HASHTABLE_FIND:
+    case BuiltinOperator_HASHTABLE_IMPORT:
+    case BuiltinOperator_HASHTABLE_SIZE:
+    case BuiltinOperator_READ_VARIABLE:
+    case BuiltinOperator_ASSIGN_VARIABLE:
+    case BuiltinOperator_BROADCAST_ARGS:
       return kTfLiteOk;
     case BuiltinOperator_PLACEHOLDER_FOR_GREATER_OP_CODES:
       return kTfLiteError;
@@ -850,6 +874,9 @@ TfLiteStatus ConvertTensorType(TensorType tensor_type, TfLiteType* type,
     case TensorType_INT32:
       *type = kTfLiteInt32;
       return kTfLiteOk;
+    case TensorType_UINT32:
+      *type = kTfLiteUInt32;
+      return kTfLiteOk;
     case TensorType_UINT8:
       *type = kTfLiteUInt8;
       return kTfLiteOk;
@@ -859,6 +886,9 @@ TfLiteStatus ConvertTensorType(TensorType tensor_type, TfLiteType* type,
     case TensorType_INT64:
       *type = kTfLiteInt64;
       return kTfLiteOk;
+    case TensorType_UINT64:
+      *type = kTfLiteUInt64;
+      return kTfLiteOk;
     case TensorType_STRING:
       *type = kTfLiteString;
       return kTfLiteOk;
@@ -871,6 +901,12 @@ TfLiteStatus ConvertTensorType(TensorType tensor_type, TfLiteType* type,
     case TensorType_COMPLEX128:
       *type = kTfLiteComplex128;
       return kTfLiteOk;
+    case TensorType_RESOURCE:
+      *type = kTfLiteResource;
+      return kTfLiteOk;
+    case TensorType_VARIANT:
+      *type = kTfLiteVariant;
+      return kTfLiteOk;
     default:
       *type = kTfLiteNoType;
       TF_LITE_REPORT_ERROR(error_reporter,
@@ -912,6 +948,11 @@ TfLiteStatus ParseAdd(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+TfLiteStatus ParseAddN(const Operator* op, ErrorReporter* error_reporter,
+                       BuiltinDataAllocator* allocator, void** builtin_data) {
+  return kTfLiteOk;
+}
+
 TfLiteStatus ParseArgMax(const Operator* op, ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data) {
   CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
@@ -962,6 +1003,56 @@ TfLiteStatus ParseArgMin(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseBatchMatMul(const Operator* op, ErrorReporter* error_reporter,
+                              BuiltinDataAllocator* allocator,
+                              void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  auto params = safe_allocator.Allocate<TfLiteBatchMatMulParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  if (const auto* bmm_params = op->builtin_options_as_BatchMatMulOptions()) {
+    params->adj_x = bmm_params->adj_x();
+    params->adj_y = bmm_params->adj_y();
+    params->asymmetric_quantize_inputs =
+        bmm_params->asymmetric_quantize_inputs();
+  }
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseBatchToSpaceNd(const Operator*, ErrorReporter*,
+                                 BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseCast(const Operator* op, ErrorReporter* error_reporter,
+                       BuiltinDataAllocator* allocator, void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  auto params = safe_allocator.Allocate<TfLiteCastParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  if (const auto* schema_params = op->builtin_options_as_CastOptions()) {
+    TF_LITE_ENSURE_STATUS(ConvertTensorType(
+        schema_params->in_data_type(), &params->in_data_type, error_reporter));
+    TF_LITE_ENSURE_STATUS(ConvertTensorType(schema_params->out_data_type(),
+                                            &params->out_data_type,
+                                            error_reporter));
+  }
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1030,6 +1121,24 @@ TfLiteStatus ParseConv2D(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseCumsum(const Operator* op, ErrorReporter* error_reporter,
+                         BuiltinDataAllocator* allocator, void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  auto params = safe_allocator.Allocate<TfLiteCumsumParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  if (const auto* cumsum_params = op->builtin_options_as_CumsumOptions()) {
+    params->exclusive = cumsum_params->exclusive();
+    params->reverse = cumsum_params->reverse();
+  }
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1038,6 +1147,31 @@ TfLiteStatus ParseCos(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
   return kTfLiteOk;
 }
 
+TfLiteStatus ParseDepthToSpace(const Operator* op,
+                               ErrorReporter* error_reporter,
+                               BuiltinDataAllocator* allocator,
+                               void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  std::unique_ptr<TfLiteDepthToSpaceParams,
+                  SafeBuiltinDataAllocator::BuiltinDataDeleter>
+      params = safe_allocator.Allocate<TfLiteDepthToSpaceParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+
+  const auto* schema_params = op->builtin_options_as_DepthToSpaceOptions();
+  if (schema_params != nullptr) {
+    params->block_size = schema_params->block_size();
+  } else {
+    // TODO(b/157480169): We should either return kTfLiteError or fill in some
+    // reasonable defaults in the params struct. We are not doing so until we
+    // better undertand the ramifications of changing the legacy behavior.
+  }
+
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 TfLiteStatus ParseDepthwiseConv2D(const Operator* op,
                                   ErrorReporter* error_reporter,
                                   BuiltinDataAllocator* allocator,
@@ -1082,6 +1216,29 @@ TfLiteStatus ParseDequantize(const Operator*, ErrorReporter*,
   return kTfLiteOk;
 }
 
+TfLiteStatus ParseDiv(const Operator* op, ErrorReporter* error_reporter,
+                      BuiltinDataAllocator* allocator, void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  auto params = safe_allocator.Allocate<TfLiteDivParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  if (const auto* schema_params = op->builtin_options_as_DivOptions()) {
+    params->activation =
+        ConvertActivation(schema_params->fused_activation_function());
+  }
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseElu(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
+                      void**) {
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1090,6 +1247,30 @@ TfLiteStatus ParseEqual(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseExp(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
+                      void**) {
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseExpandDims(const Operator*, ErrorReporter*,
+                             BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseFill(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
+                       void**) {
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1098,6 +1279,22 @@ TfLiteStatus ParseFloor(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseFloorDiv(const Operator*, ErrorReporter*,
+                           BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseFloorMod(const Operator*, ErrorReporter*,
+                           BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
 TfLiteStatus ParseFullyConnected(const Operator* op,
                                  ErrorReporter* error_reporter,
                                  BuiltinDataAllocator* allocator,
@@ -1144,6 +1341,35 @@ TfLiteStatus ParseFullyConnected(const Operator* op,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseGather(const Operator* op, ErrorReporter* error_reporter,
+                         BuiltinDataAllocator* allocator, void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  auto params = safe_allocator.Allocate<TfLiteGatherParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  params->axis = 0;
+  params->batch_dims = 0;
+  if (const auto* gather_params = op->builtin_options_as_GatherOptions()) {
+    params->axis = gather_params->axis();
+    params->batch_dims = gather_params->batch_dims();
+  }
+
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseGatherNd(const Operator*, ErrorReporter*,
+                           BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1168,6 +1394,30 @@ TfLiteStatus ParseHardSwish(const Operator*, ErrorReporter*,
   return kTfLiteOk;
 }
 
+TfLiteStatus ParseIf(const Operator* op, ErrorReporter* error_reporter,
+                     BuiltinDataAllocator* allocator, void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  std::unique_ptr<TfLiteIfParams, SafeBuiltinDataAllocator::BuiltinDataDeleter>
+      params = safe_allocator.Allocate<TfLiteIfParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+
+  const IfOptions* schema_params = op->builtin_options_as_IfOptions();
+
+  if (schema_params != nullptr) {
+    params->then_subgraph_index = schema_params->then_subgraph_index();
+    params->else_subgraph_index = schema_params->else_subgraph_index();
+  } else {
+    // TODO(b/157480169): We should either return kTfLiteError or fill in some
+    // reasonable defaults in the params struct. We are not doing so until we
+    // better undertand the ramifications of changing the legacy behavior.
+  }
+
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 TfLiteStatus ParseL2Normalization(const Operator* op,
                                   ErrorReporter* error_reporter,
                                   BuiltinDataAllocator* allocator,
@@ -1195,6 +1445,22 @@ TfLiteStatus ParseL2Normalization(const Operator* op,
   return kTfLiteOk;
 }
 
+TfLiteStatus ParseLeakyRelu(const Operator* op, ErrorReporter* error_reporter,
+                            BuiltinDataAllocator* allocator,
+                            void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  auto params = safe_allocator.Allocate<TfLiteLeakyReluParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  if (const auto* leaky_relu_params =
+          op->builtin_options_as_LeakyReluOptions()) {
+    params->alpha = leaky_relu_params->alpha();
+  }
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1251,6 +1517,14 @@ TfLiteStatus ParseLogistic(const Operator*, ErrorReporter*,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseLogSoftmax(const Operator*, ErrorReporter*,
+                             BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1378,6 +1652,14 @@ TfLiteStatus ParsePool(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParsePow(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
+                      void**) {
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1599,6 +1881,39 @@ TfLiteStatus ParseSoftmax(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseSpaceToBatchNd(const Operator*, ErrorReporter*,
+                                 BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
+TfLiteStatus ParseSpaceToDepth(const Operator* op,
+                               ErrorReporter* error_reporter,
+                               BuiltinDataAllocator* allocator,
+                               void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  std::unique_ptr<TfLiteSpaceToDepthParams,
+                  SafeBuiltinDataAllocator::BuiltinDataDeleter>
+      params = safe_allocator.Allocate<TfLiteSpaceToDepthParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+
+  const auto* schema_params = op->builtin_options_as_SpaceToDepthOptions();
+  if (schema_params != nullptr) {
+    params->block_size = schema_params->block_size();
+  } else {
+    // TODO(b/157480169): We should either return kTfLiteError or fill in some
+    // reasonable defaults in the params struct. We are not doing so until we
+    // better undertand the ramifications of changing the legacy behavior.
+  }
+
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 TfLiteStatus ParseSplit(const Operator* op, ErrorReporter* error_reporter,
                         BuiltinDataAllocator* allocator, void** builtin_data) {
   CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
@@ -1647,6 +1962,39 @@ TfLiteStatus ParseSplitV(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+TfLiteStatus ParseSqueeze(const Operator* op, ErrorReporter* error_reporter,
+                          BuiltinDataAllocator* allocator,
+                          void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+
+  std::unique_ptr<TfLiteSqueezeParams,
+                  SafeBuiltinDataAllocator::BuiltinDataDeleter>
+      params = safe_allocator.Allocate<TfLiteSqueezeParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+
+  const SqueezeOptions* schema_params = op->builtin_options_as_SqueezeOptions();
+
+  if (schema_params != nullptr) {
+    const auto* squeeze_dims = schema_params->squeeze_dims();
+    if (squeeze_dims != nullptr) {
+      TF_LITE_ENSURE_STATUS(FlatBufferIntVectorToArray(
+          sizeof(params->squeeze_dims), squeeze_dims, params->squeeze_dims,
+          error_reporter, "squeeze"));
+      params->num_squeeze_dims = squeeze_dims->size();
+    } else {
+      params->num_squeeze_dims = 0;
+    }
+  } else {
+    // TODO(b/157480169): We should either return kTfLiteError or fill in some
+    // reasonable defaults in the params struct. We are not doing so until we
+    // better undertand the ramifications of changing the legacy behavior.
+  }
+
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
+
 // We have this parse function instead of directly returning kTfLiteOk from the
 // switch-case in ParseOpData because this function is used as part of the
 // selective registration for the OpResolver implementation in micro.
@@ -1753,6 +2101,40 @@ TfLiteStatus ParseTanh(const Operator*, ErrorReporter*, BuiltinDataAllocator*,
                        void**) {
   return kTfLiteOk;
 }
+//
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseTranspose(const Operator*, ErrorReporter*,
+                            BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
+TfLiteStatus ParseTransposeConv(const Operator* op,
+                                ErrorReporter* error_reporter,
+                                BuiltinDataAllocator* allocator,
+                                void** builtin_data) {
+  CheckParsePointerParams(op, error_reporter, allocator, builtin_data);
+
+  SafeBuiltinDataAllocator safe_allocator(allocator);
+  std::unique_ptr<TfLiteTransposeConvParams,
+                  SafeBuiltinDataAllocator::BuiltinDataDeleter>
+      params = safe_allocator.Allocate<TfLiteTransposeConvParams>();
+  TF_LITE_ENSURE(error_reporter, params != nullptr);
+  const TransposeConvOptions* transpose_conv_params =
+      op->builtin_options_as_TransposeConvOptions();
+  if (transpose_conv_params != nullptr) {
+    params->padding = ConvertPadding(transpose_conv_params->padding());
+    params->stride_width = transpose_conv_params->stride_w();
+    params->stride_height = transpose_conv_params->stride_h();
+  } else {
+    // TODO(b/157480169): We should either return kTfLiteError or fill in some
+    // reasonable defaults in the params struct. We are not doing so until we
+    // better undertand the ramifications of changing the legacy behavior.
+  }
+  *builtin_data = params.release();
+  return kTfLiteOk;
+}
 
 TfLiteStatus ParseUnpack(const Operator* op, ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data) {
@@ -1779,6 +2161,14 @@ TfLiteStatus ParseUnpack(const Operator* op, ErrorReporter* error_reporter,
   return kTfLiteOk;
 }
 
+// We have this parse function instead of directly returning kTfLiteOk from the
+// switch-case in ParseOpData because this function is used as part of the
+// selective registration for the OpResolver implementation in micro.
+TfLiteStatus ParseZerosLike(const Operator*, ErrorReporter*,
+                            BuiltinDataAllocator*, void**) {
+  return kTfLiteOk;
+}
+
 TfLiteStatus ParseOpData(const Operator* op, BuiltinOperator op_type,
                          ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data) {

code/components/tfmicro/tensorflow/lite/core/api/flatbuffer_conversions.h

@@ -75,15 +75,30 @@ TfLiteStatus ParseAbs(const Operator* op, ErrorReporter* error_reporter,
 TfLiteStatus ParseAdd(const Operator* op, ErrorReporter* error_reporter,
                       BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseAddN(const Operator* op, ErrorReporter* error_reporter,
+                       BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParseArgMax(const Operator* op, ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data);
 
 TfLiteStatus ParseArgMin(const Operator* op, ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseBatchMatMul(const Operator* op, ErrorReporter* error_reporter,
+                              BuiltinDataAllocator* allocator,
+                              void** builtin_data);
+
+TfLiteStatus ParseBatchToSpaceNd(const Operator* op,
+                                 ErrorReporter* error_reporter,
+                                 BuiltinDataAllocator* allocator,
+                                 void** builtin_data);
+
 TfLiteStatus ParseCeil(const Operator* op, ErrorReporter* error_reporter,
                        BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseCast(const Operator* op, ErrorReporter* error_reporter,
+                       BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParseConcatenation(const Operator* op,
                                 ErrorReporter* error_reporter,
                                 BuiltinDataAllocator* allocator,
@@ -95,6 +110,14 @@ TfLiteStatus ParseConv2D(const Operator* op, ErrorReporter* error_reporter,
 TfLiteStatus ParseCos(const Operator* op, ErrorReporter* error_reporter,
                       BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseCumsum(const Operator* op, ErrorReporter* error_reporter,
+                         BuiltinDataAllocator* allocator, void** builtin_data);
+
+TfLiteStatus ParseDepthToSpace(const Operator* op,
+                               ErrorReporter* error_reporter,
+                               BuiltinDataAllocator* allocator,
+                               void** builtin_data);
+
 TfLiteStatus ParseDepthwiseConv2D(const Operator* op,
                                   ErrorReporter* error_reporter,
                                   BuiltinDataAllocator* allocator,
@@ -104,17 +127,48 @@ TfLiteStatus ParseDequantize(const Operator* op, ErrorReporter* error_reporter,
                              BuiltinDataAllocator* allocator,
                              void** builtin_data);
 
+TfLiteStatus ParseDiv(const Operator* op, ErrorReporter* error_reporter,
+                      BuiltinDataAllocator* allocator, void** builtin_data);
+
+TfLiteStatus ParseElu(const Operator* op, ErrorReporter* error_reporter,
+                      BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParseEqual(const Operator* op, ErrorReporter* error_reporter,
                         BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseExp(const Operator* op, ErrorReporter* error_reporter,
+                      BuiltinDataAllocator* allocator, void** builtin_data);
+
+TfLiteStatus ParseExpandDims(const Operator* op, ErrorReporter* error_reporter,
+                             BuiltinDataAllocator* allocator,
+                             void** builtin_data);
+
+TfLiteStatus ParseFill(const Operator* op, ErrorReporter* error_reporter,
+                       BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParseFloor(const Operator* op, ErrorReporter* error_reporter,
                         BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseFloorDiv(const Operator* op, ErrorReporter* error_reporter,
+                           BuiltinDataAllocator* allocator,
+                           void** builtin_data);
+
+TfLiteStatus ParseFloorMod(const Operator* op, ErrorReporter* error_reporter,
+                           BuiltinDataAllocator* allocator,
+                           void** builtin_data);
+
 TfLiteStatus ParseFullyConnected(const Operator* op,
                                  ErrorReporter* error_reporter,
                                  BuiltinDataAllocator* allocator,
                                  void** builtin_data);
 
+TfLiteStatus ParseGather(const Operator* op, ErrorReporter* error_reporter,
+                         BuiltinDataAllocator* allocator, void** builtin_data);
+
+TfLiteStatus ParseGatherNd(const Operator* op, ErrorReporter* error_reporter,
+                           BuiltinDataAllocator* allocator,
+                           void** builtin_data);
+
 TfLiteStatus ParseGreater(const Operator* op, ErrorReporter* error_reporter,
                           BuiltinDataAllocator* allocator, void** builtin_data);
 
@@ -127,11 +181,18 @@ TfLiteStatus ParseHardSwish(const Operator* op, ErrorReporter* error_reporter,
                             BuiltinDataAllocator* allocator,
                             void** builtin_data);
 
+TfLiteStatus ParseIf(const Operator* op, ErrorReporter* error_reporter,
+                     BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParseL2Normalization(const Operator* op,
                                   ErrorReporter* error_reporter,
                                   BuiltinDataAllocator* allocator,
                                   void** builtin_data);
 
+TfLiteStatus ParseLeakyRelu(const Operator* op, ErrorReporter* error_reporter,
+                            BuiltinDataAllocator* allocator,
+                            void** builtin_data);
+
 TfLiteStatus ParseLess(const Operator* op, ErrorReporter* error_reporter,
                        BuiltinDataAllocator* allocator, void** builtin_data);
 
@@ -158,6 +219,10 @@ TfLiteStatus ParseLogistic(const Operator* op, ErrorReporter* error_reporter,
                            BuiltinDataAllocator* allocator,
                            void** builtin_data);
 
+TfLiteStatus ParseLogSoftmax(const Operator* op, ErrorReporter* error_reporter,
+                             BuiltinDataAllocator* allocator,
+                             void** builtin_data);
+
 TfLiteStatus ParseMaximum(const Operator* op, ErrorReporter* error_reporter,
                           BuiltinDataAllocator* allocator, void** builtin_data);
 
@@ -186,6 +251,9 @@ TfLiteStatus ParsePadV2(const Operator* op, ErrorReporter* error_reporter,
 TfLiteStatus ParsePool(const Operator* op, ErrorReporter* error_reporter,
                        BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParsePow(const Operator* op, ErrorReporter* error_reporter,
+                      BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParsePrelu(const Operator* op, ErrorReporter* error_reporter,
                         BuiltinDataAllocator* allocator, void** builtin_data);
 
@@ -230,12 +298,25 @@ TfLiteStatus ParseSin(const Operator* op, ErrorReporter* error_reporter,
 TfLiteStatus ParseSoftmax(const Operator* op, ErrorReporter* error_reporter,
                           BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseSpaceToBatchNd(const Operator* op,
+                                 ErrorReporter* error_reporter,
+                                 BuiltinDataAllocator* allocator,
+                                 void** builtin_data);
+
+TfLiteStatus ParseSpaceToDepth(const Operator* op,
+                               ErrorReporter* error_reporter,
+                               BuiltinDataAllocator* allocator,
+                               void** builtin_data);
+
 TfLiteStatus ParseSplit(const Operator* op, ErrorReporter* error_reporter,
                         BuiltinDataAllocator* allocator, void** builtin_data);
 
 TfLiteStatus ParseSplitV(const Operator* op, ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseSqueeze(const Operator* op, ErrorReporter* error_reporter,
+                          BuiltinDataAllocator* allocator, void** builtin_data);
+
 TfLiteStatus ParseSqrt(const Operator* op, ErrorReporter* error_reporter,
                        BuiltinDataAllocator* allocator, void** builtin_data);
 
@@ -256,9 +337,22 @@ TfLiteStatus ParseSvdf(const Operator* op, ErrorReporter* error_reporter,
 TfLiteStatus ParseTanh(const Operator* op, ErrorReporter* error_reporter,
                        BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseTranspose(const Operator* op, ErrorReporter* error_reporter,
+                            BuiltinDataAllocator* allocator,
+                            void** builtin_data);
+
+TfLiteStatus ParseTransposeConv(const Operator* op,
+                                ErrorReporter* error_reporter,
+                                BuiltinDataAllocator* allocator,
+                                void** builtin_data);
+
 TfLiteStatus ParseUnpack(const Operator* op, ErrorReporter* error_reporter,
                          BuiltinDataAllocator* allocator, void** builtin_data);
 
+TfLiteStatus ParseZerosLike(const Operator* op, ErrorReporter* error_reporter,
+                            BuiltinDataAllocator* allocator,
+                            void** builtin_data);
+
 }  // namespace tflite
 
 #endif  // TENSORFLOW_LITE_CORE_API_FLATBUFFER_CONVERSIONS_H_

code/components/tfmicro/tensorflow/lite/core/api/op_resolver.cc

@@ -30,8 +30,7 @@ TfLiteStatus GetRegistrationFromOpCode(
   auto builtin_code = GetBuiltinCode(opcode);
   int version = opcode->version();
 
-  if (builtin_code > BuiltinOperator_MAX ||
-      builtin_code < BuiltinOperator_MIN) {
+  if (builtin_code > BuiltinOperator_MAX) {
     TF_LITE_REPORT_ERROR(
         error_reporter,
         "Op builtin_code out of range: %d. Are you using old TFLite binary "
@@ -43,7 +42,9 @@ TfLiteStatus GetRegistrationFromOpCode(
     if (*registration == nullptr) {
       TF_LITE_REPORT_ERROR(
           error_reporter,
-          "Didn't find op for builtin opcode '%s' version '%d'\n",
+          "Didn't find op for builtin opcode '%s' version '%d'. "
+          "An older version of this builtin might be supported. "
+          "Are you using an old TFLite binary with a newer model?\n",
           EnumNameBuiltinOperator(builtin_code), version);
       status = kTfLiteError;
     }

code/components/tfmicro/tensorflow/lite/core/api/op_resolver.h

@@ -15,6 +15,7 @@ limitations under the License.
 #ifndef TENSORFLOW_LITE_CORE_API_OP_RESOLVER_H_
 #define TENSORFLOW_LITE_CORE_API_OP_RESOLVER_H_
 
+#include <memory>
 #include <vector>
 
 #include "tensorflow/lite/c/common.h"
@@ -45,6 +46,22 @@ class OpResolver {
   }
 
   virtual ~OpResolver() {}
+
+ private:
+  /// Returns true if this OpResolver may contain any "user defined" ops.
+  /// By "user defined" ops, we mean any op definitions other than those
+  /// contained in tflite::ops::builtin::BuiltinOpResolver.
+  ///
+  /// If this method returns true, it doesn't necessarily mean that the
+  /// OpResolver contains a user-defined op, just that the absence of
+  /// user-defined ops can't be guaranteed.
+  ///
+  /// Note that "user-defined" ops are not the same as "custom" ops;
+  /// BuiltinOpResolver may support certain "custom" ops, in addition to
+  /// "builtin" ops, and may not support all of the "builtin" op enum values.
+  virtual bool MayContainUserDefinedOps() const { return true; }
+
+  friend class OpResolverInternal;
 };
 
 // Handles the logic for converting between an OperatorCode structure extracted

code/components/tfmicro/tensorflow/lite/core/api/profiler.h

@@ -1,194 +0,0 @@
-/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.
-
-Licensed under the Apache License, Version 2.0 (the "License");
-you may not use this file except in compliance with the License.
-You may obtain a copy of the License at
-
-    http://www.apache.org/licenses/LICENSE-2.0
-
-Unless required by applicable law or agreed to in writing, software
-distributed under the License is distributed on an "AS IS" BASIS,
-WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-See the License for the specific language governing permissions and
-limitations under the License.
-==============================================================================*/
-#ifndef TENSORFLOW_LITE_CORE_API_PROFILER_H_
-#define TENSORFLOW_LITE_CORE_API_PROFILER_H_
-
-#include <cstdint>
-
-namespace tflite {
-
-// A simple utility for enabling profiled event tracing in TensorFlow Lite.
-class Profiler {
- public:
-  // As certain Profiler instance might be only interested in certain event
-  // types, we define each event type value to allow a Profiler to use
-  // bitmasking bitwise operations to determine whether an event should be
-  // recorded or not.
-  enum class EventType {
-    // Default event type, the metadata field has no special significance.
-    DEFAULT = 1,
-
-    // The event is an operator invocation and the event_metadata field is the
-    // index of operator node.
-    OPERATOR_INVOKE_EVENT = 2,
-
-    // The event is an invocation for an internal operator of a TFLite delegate.
-    // The event_metadata field is the index of operator node that's specific to
-    // the delegate.
-    DELEGATE_OPERATOR_INVOKE_EVENT = 4,
-
-    // The event is a recording of runtime instrumentation such as the overall
-    // TFLite runtime status, the TFLite delegate status (if a delegate
-    // is applied), and the overall model inference latency etc.
-    // Note, the delegate status and overall status are stored as separate
-    // event_metadata fields. In particular, the delegate status is encoded
-    // as DelegateStatus::full_status().
-    GENERAL_RUNTIME_INSTRUMENTATION_EVENT = 8,
-  };
-
-  virtual ~Profiler() {}
-
-  // Signals the beginning of an event and returns a handle to the profile
-  // event. The `event_metadata1` and `event_metadata2` have different
-  // interpretations based on the actual Profiler instance and the `event_type`.
-  // For example, as for the 'SubgraphAwareProfiler' defined in
-  // lite/core/subgraph.h, when the event_type is OPERATOR_INVOKE_EVENT,
-  // `event_metadata1` represents the index of a TFLite node, and
-  // `event_metadata2` represents the index of the subgraph that this event
-  // comes from.
-  virtual uint32_t BeginEvent(const char* tag, EventType event_type,
-                              int64_t event_metadata1,
-                              int64_t event_metadata2) = 0;
-  // Similar w/ the above, but `event_metadata2` defaults to 0.
-  uint32_t BeginEvent(const char* tag, EventType event_type,
-                      int64_t event_metadata) {
-    return BeginEvent(tag, event_type, event_metadata, /*event_metadata2*/ 0);
-  }
-
-  // Signals an end to the specified profile event with 'event_metadata's, This
-  // is useful when 'event_metadata's are not available when the event begins
-  // or when one wants to overwrite the 'event_metadata's set at the beginning.
-  virtual void EndEvent(uint32_t event_handle, int64_t event_metadata1,
-                        int64_t event_metadata2) {}
-  // Signals an end to the specified profile event.
-  virtual void EndEvent(uint32_t event_handle) = 0;
-
-  // Appends an event of type 'event_type' with 'tag' and 'event_metadata'
-  // which started at 'start' and ended at 'end'
-  // Note:
-  // In cases were ProfileSimmarizer and tensorflow::StatsCalculator are used
-  // they assume the value is in "usec", if in any case subclasses
-  // didn't put usec, then the values are not meaningful.
-  // TODO karimnosseir: Revisit and make the function more clear.
-  void AddEvent(const char* tag, EventType event_type, uint64_t start,
-                uint64_t end, int64_t event_metadata) {
-    AddEvent(tag, event_type, start, end, event_metadata,
-             /*event_metadata2*/ 0);
-  }
-
-  virtual void AddEvent(const char* tag, EventType event_type, uint64_t start,
-                        uint64_t end, int64_t event_metadata1,
-                        int64_t event_metadata2) {}
-
- protected:
-  friend class ScopedProfile;
-};
-
-// Adds a profile event to `profiler` that begins with the construction
-// of the object and ends when the object goes out of scope.
-// The lifetime of tag should be at least the lifetime of `profiler`.
-// `profiler` may be null, in which case nothing is profiled.
-class ScopedProfile {
- public:
-  ScopedProfile(Profiler* profiler, const char* tag,
-                Profiler::EventType event_type = Profiler::EventType::DEFAULT,
-                int64_t event_metadata = 0)
-      : profiler_(profiler), event_handle_(0) {
-    if (profiler) {
-      event_handle_ = profiler_->BeginEvent(tag, event_type, event_metadata);
-    }
-  }
-
-  ~ScopedProfile() {
-    if (profiler_) {
-      profiler_->EndEvent(event_handle_);
-    }
-  }
-
- protected:
-  Profiler* profiler_;
-  uint32_t event_handle_;
-};
-
-class ScopedOperatorProfile : public ScopedProfile {
- public:
-  ScopedOperatorProfile(Profiler* profiler, const char* tag, int node_index)
-      : ScopedProfile(profiler, tag, Profiler::EventType::OPERATOR_INVOKE_EVENT,
-                      static_cast<uint32_t>(node_index)) {}
-};
-
-class ScopedDelegateOperatorProfile : public ScopedProfile {
- public:
-  ScopedDelegateOperatorProfile(Profiler* profiler, const char* tag,
-                                int node_index)
-      : ScopedProfile(profiler, tag,
-                      Profiler::EventType::DELEGATE_OPERATOR_INVOKE_EVENT,
-                      static_cast<uint32_t>(node_index)) {}
-};
-
-class ScopedRuntimeInstrumentationProfile : public ScopedProfile {
- public:
-  ScopedRuntimeInstrumentationProfile(Profiler* profiler, const char* tag)
-      : ScopedProfile(
-            profiler, tag,
-            Profiler::EventType::GENERAL_RUNTIME_INSTRUMENTATION_EVENT, -1) {}
-
-  void set_runtime_status(int64_t delegate_status, int64_t interpreter_status) {
-    if (profiler_) {
-      delegate_status_ = delegate_status;
-      interpreter_status_ = interpreter_status;
-    }
-  }
-
-  ~ScopedRuntimeInstrumentationProfile() {
-    if (profiler_) {
-      profiler_->EndEvent(event_handle_, delegate_status_, interpreter_status_);
-    }
-  }
-
- private:
-  int64_t delegate_status_;
-  int64_t interpreter_status_;
-};
-
-}  // namespace tflite
-
-#define TFLITE_VARNAME_UNIQ_IMPL(name, ctr) name##ctr
-#define TFLITE_VARNAME_UNIQ(name, ctr) TFLITE_VARNAME_UNIQ_IMPL(name, ctr)
-
-#define TFLITE_SCOPED_TAGGED_DEFAULT_PROFILE(profiler, tag)          \
-  tflite::ScopedProfile TFLITE_VARNAME_UNIQ(_profile_, __COUNTER__)( \
-      (profiler), (tag))
-
-#define TFLITE_SCOPED_TAGGED_OPERATOR_PROFILE(profiler, tag, node_index)     \
-  tflite::ScopedOperatorProfile TFLITE_VARNAME_UNIQ(_profile_, __COUNTER__)( \
-      (profiler), (tag), (node_index))
-
-#define TFLITE_SCOPED_DELEGATE_OPERATOR_PROFILE(profiler, tag, node_index) \
-  tflite::ScopedDelegateOperatorProfile TFLITE_VARNAME_UNIQ(               \
-      _profile_, __COUNTER__)((profiler), (tag), (node_index))
-
-#define TFLITE_ADD_RUNTIME_INSTRUMENTATION_EVENT(                          \
-    profiler, tag, delegate_status, interpreter_status)                    \
-  do {                                                                     \
-    if (!profiler) {                                                       \
-      const auto handle = profiler->BeginEvent(                            \
-          tag, Profiler::EventType::GENERAL_RUNTIME_INSTRUMENTATION_EVENT, \
-          delegate_status, interpreter_status);                            \
-      profiler->EndEvent(handle);                                          \
-    }                                                                      \
-  } while (false);
-
-#endif  // TENSORFLOW_LITE_CORE_API_PROFILER_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/common.h

@@ -178,14 +178,54 @@ inline int32_t MultiplyByQuantizedMultiplier(int64_t x,
   // - input x is in the range -(1<<47) <= x < (1<<47)
   assert(quantized_multiplier >= 0);
   assert(shift >= -31 && shift < 8);
+  assert(x >= -(static_cast<int64_t>(1) << 47) &&
+         x < (static_cast<int64_t>(1) << 47));
 
-  int32_t reduced_multiplier = (quantized_multiplier + (1 << 15)) >> 16;
+  int32_t reduced_multiplier = (quantized_multiplier < 0x7FFF0000)
+                                   ? ((quantized_multiplier + (1 << 15)) >> 16)
+                                   : 0x7FFF;
   int total_shift = 15 - shift;
   x = (x * (int64_t)reduced_multiplier) + ((int64_t)1 << (total_shift - 1));
   int32_t result = x >> total_shift;
   return result;
 }
 
+#ifdef USE_NEON
+// Round uses ARM's rounding shift right.
+inline int32x4x4_t MultiplyByQuantizedMultiplier4Rows(
+    int32x4x4_t input_val, int32_t quantized_multiplier, int shift) {
+  const int left_shift = std::max(shift, 0);
+  const int right_shift = std::min(shift, 0);
+  int32x4x4_t result;
+
+  int32x4_t multiplier_dup = vdupq_n_s32(quantized_multiplier);
+  int32x4_t left_shift_dup = vdupq_n_s32(left_shift);
+  int32x4_t right_shift_dup = vdupq_n_s32(right_shift);
+
+  result.val[0] =
+      vrshlq_s32(vqrdmulhq_s32(vshlq_s32(input_val.val[0], left_shift_dup),
+                               multiplier_dup),
+                 right_shift_dup);
+
+  result.val[1] =
+      vrshlq_s32(vqrdmulhq_s32(vshlq_s32(input_val.val[1], left_shift_dup),
+                               multiplier_dup),
+                 right_shift_dup);
+
+  result.val[2] =
+      vrshlq_s32(vqrdmulhq_s32(vshlq_s32(input_val.val[2], left_shift_dup),
+                               multiplier_dup),
+                 right_shift_dup);
+
+  result.val[3] =
+      vrshlq_s32(vqrdmulhq_s32(vshlq_s32(input_val.val[3], left_shift_dup),
+                               multiplier_dup),
+                 right_shift_dup);
+
+  return result;
+}
+#endif
+
 template <typename T>
 int CountLeadingZeros(T integer_input) {
   static_assert(std::is_unsigned<T>::value,
@@ -239,79 +279,125 @@ inline Integer FloorLog2(Integer n) {
   }
 }
 
-// generate INT16 LUT for function(), e.g., table exp(x) and 1/(1+x) used in
-// softmax
-// func - the function to build the LUT for (e.g exp(x))
-// min,max - table limits
-// table - pointer to buffer
-// num - number of elements in the LUT
-inline void gen_lut(double (*func)(double), double min, double max,
-                    int16_t* table, const int num) {
-  // size of table should equal to num + 1
-  // last element only for slope calculation
-  double step = (max - min) / (num - 1);
-  double half_step = step / 2.0;
-  for (int i = 0; i < num - 1; i++) {
-    double sample_val = TfLiteRound(func(min + i * step) * 32768.0);
-    double midpoint_interp_val =
-        TfLiteRound((func(min + (i + 1) * step) * 32768.0 +
-                     TfLiteRound(func(min + i * step) * 32768.0)) /
-                    2.0);
-    double midpoint_val =
-        TfLiteRound(func(min + i * step + half_step) * 32768.0);
-    double midpoint_err = midpoint_interp_val - midpoint_val;
-    double bias = TfLiteRound(midpoint_err / 2.0);
-    table[i] = std::min(std::max(sample_val - bias, -32768.0), 32767.0);
-  }
-  table[num - 1] =
-      std::min(std::max(TfLiteRound(func(max) * 32768.0), -32768.0), 32767.0);
+// The size of the LUT depends on the type of input. For int8 inputs a simple
+// 256 entries LUT is used. For int16 inputs the high 9 bits are used for
+// indexing and the 7 remaining bits are used for interpolation. We thus use a
+// 513-entries LUT for int16 cases, 512 for the 9-bit indexing and 1 extra entry
+// to interpolate the last value.
+template <typename LutInT>
+constexpr int lut_size() {
+  static_assert(std::is_same<LutInT, int8_t>::value ||
+                    std::is_same<LutInT, int16_t>::value,
+                "Only LUTs with int8 or int16 inputs are supported.");
+  return std::is_same<LutInT, int8_t>::value ? 256 : 513;
 }
 
-// generate INT16 LUT for function(), e.g., table exp(x) and 1/(1+x) used in
-// softmax
-// func - the function to build the LUT for (e.g exp(x))
-// min,max - table limits
-// table - pointer to buffer
-// num - number of elements in the LUT
-inline void gen_lut(float (*func)(float), float min, float max, int16_t* table,
-                    const int num) {
-  // size of table should equal to num + 1
-  // last element only for slope calculation
-  float step = (max - min) / (num - 1);
-  float half_step = step / 2.0f;
-  for (int i = 0; i < num - 1; i++) {
-    float sample_val = TfLiteRound(func(min + i * step) * 32768.0f);
-    float midpoint_interp_val =
-        TfLiteRound((func(min + (i + 1) * step) * 32768.0f +
-                     TfLiteRound(func(min + i * step) * 32768.0f)) /
-                    2.0f);
-    float midpoint_val =
-        TfLiteRound(func(min + i * step + half_step) * 32768.0f);
-    float midpoint_err = midpoint_interp_val - midpoint_val;
-    float bias = TfLiteRound(midpoint_err / 2.0f);
-    table[i] = std::min(std::max(sample_val - bias, -32768.0f), 32767.0f);
+// Generate a LUT for 'func' which can be used to approximate functions like
+// exp, log, ...
+//
+// - func: the function to build the LUT for (e.g exp(x))
+// - input_min, input_max: range of the func inputs
+// - output_min, output_max: range of the func outputs
+// - lut: pointer to the LUT table to fill, the table must be of size
+// lut_size<LutInT>()
+template <typename FloatT, typename LutInT, typename LutOutT>
+inline void gen_lut(FloatT (*func)(FloatT), FloatT input_min, FloatT input_max,
+                    FloatT output_min, FloatT output_max, LutOutT* lut) {
+  static_assert(std::is_same<LutInT, int8_t>::value ||
+                    std::is_same<LutInT, int16_t>::value,
+                "Only LUTs with int8 or int16 inputs are supported.");
+  static_assert(std::is_same<LutOutT, int8_t>::value ||
+                    std::is_same<LutOutT, int16_t>::value,
+                "Only LUTs with int8 or int16 outputs are supported.");
+  static_assert(std::is_floating_point<FloatT>::value,
+                "FloatT must be a floating-point type.");
+
+  const int nb_steps = std::is_same<LutInT, int8_t>::value ? 256 : 512;
+  const FloatT step = (input_max - input_min) / nb_steps;
+  const FloatT half_step = step / 2;
+  const FloatT output_scaling_inv =
+      static_cast<FloatT>(std::numeric_limits<LutOutT>::max() -
+                          std::numeric_limits<LutOutT>::min() + 1) /
+      (output_max - output_min);
+  const FloatT table_min =
+      static_cast<FloatT>(std::numeric_limits<LutOutT>::min());
+  const FloatT table_max =
+      static_cast<FloatT>(std::numeric_limits<LutOutT>::max());
+
+  for (int i = 0; i < nb_steps; i++) {
+    const FloatT val = func(input_min + i * step);
+    const FloatT val_midpoint = func(input_min + i * step + half_step);
+    const FloatT val_next = func(input_min + (i + 1) * step);
+
+    const FloatT sample_val = TfLiteRound(val * output_scaling_inv);
+    const FloatT midpoint_interp_val =
+        TfLiteRound((val_next * output_scaling_inv +
+                     TfLiteRound(val * output_scaling_inv)) /
+                    2);
+    const FloatT midpoint_val = TfLiteRound(val_midpoint * output_scaling_inv);
+    const FloatT midpoint_err = midpoint_interp_val - midpoint_val;
+    const FloatT bias = TfLiteRound(midpoint_err / 2);
+
+    lut[i] = static_cast<LutOutT>(std::min<FloatT>(
+        std::max<FloatT>(sample_val - bias, table_min), table_max));
+  }
+
+  const bool with_extra_interpolation_value =
+      std::is_same<LutInT, int16_t>::value;
+  if (with_extra_interpolation_value) {
+    lut[nb_steps] = static_cast<LutOutT>(std::min<FloatT>(
+        std::max<FloatT>(TfLiteRound(func(input_max) * output_scaling_inv),
+                         table_min),
+        table_max));
   }
-  table[num - 1] = std::min(
-      std::max(TfLiteRound(func(max) * 32768.0f), -32768.0f), 32767.0f);
 }
 
-// int16_t func table lookup, e.g., lookup exp() and 1/(1+x) used in softmax
-inline int16_t generic_int16_table_lookup(int16_t value, const int16_t* lut) {
-  // 512 base value, lut[513] only for calculate slope
-  uint16_t index = static_cast<uint16_t>(256 + (value >> 7));
+// LUT must have 513 values
+template <typename LutOutT>
+inline LutOutT lut_lookup_with_interpolation(int16_t value,
+                                             const LutOutT* lut) {
+  static_assert(std::is_same<LutOutT, int8_t>::value ||
+                    std::is_same<LutOutT, int16_t>::value,
+                "Only LUTs with int8 or int16 outputs are supported.");
+  // 512 base values, lut[513] is only used to calculate the slope
+  const uint16_t index = static_cast<uint16_t>(256 + (value >> 7));
   assert(index < 512 && "LUT index out of range.");
-  int16_t offset = value & 0x7f;
+  const int16_t offset = value & 0x7f;
 
-  // base and slope are Q0.15
-  int16_t base = lut[index];
-  int16_t slope = lut[index + 1] - lut[index];
+  // Base and slope are Q0.x
+  const LutOutT base = lut[index];
+  const LutOutT slope = lut[index + 1] - lut[index];
 
-  // Q0.15 * Q0.7 = Q0.22
-  // Round and convert from Q0.22 to Q0.15
-  int32_t delta = (static_cast<int32_t>(slope) * offset + 64) >> 7;
+  // Q0.x * Q0.7 = Q0.(x + 7)
+  // Round and convert from Q0.(x + 7) to Q0.x
+  const int delta = (slope * offset + 64) >> 7;
 
   // Q0.15 + Q0.15
-  return base + delta;
+  return static_cast<LutOutT>(base + delta);
+}
+
+// int16_t -> int16_t table lookup with interpolation
+// LUT must have 513 values
+inline int16_t lut_lookup(int16_t value, const int16_t* lut) {
+  return lut_lookup_with_interpolation(value, lut);
+}
+
+// int16_t -> int8_t table lookup with interpolation
+// LUT must have 513 values
+inline int8_t lut_lookup(int16_t value, const int8_t* lut) {
+  return lut_lookup_with_interpolation(value, lut);
+}
+
+// int8_t -> int8_t table lookup without interpolation
+// LUT must have 256 values
+inline int8_t lut_lookup(int8_t value, const int8_t* lut) {
+  return lut[128 + value];
+}
+
+// int8_t -> int16_t table lookup without interpolation
+// LUT must have 256 values
+inline int16_t lut_lookup(int8_t value, const int16_t* lut) {
+  return lut[128 + value];
 }
 
 // Table of sigmoid(i/24) at 0.16 format - 256 elements.
@@ -533,7 +619,8 @@ log_x_for_x_greater_than_or_equal_to_1_impl(
   //                   InputIntegerBits - z_b_headroom - 0.25);
   const FixedPointAccum z_a_pow_2_adj = SaturatingAddNonGemmlowp(
       FixedPointAccum::FromRaw(SaturatingRoundingMultiplyByPOTParam(
-          InputIntegerBits - z_a_headroom_plus_1, 31 - kAccumIntegerBits)),
+          static_cast<int32_t>(InputIntegerBits - z_a_headroom_plus_1),
+          31 - kAccumIntegerBits)),
       shifted_quarter);
 
   // z_b is treated like z_a, but premultiplying by sqrt(0.5).
@@ -543,7 +630,8 @@ log_x_for_x_greater_than_or_equal_to_1_impl(
       SaturatingRoundingMultiplyByPOTParam(z_a.raw(), z_b_headroom);
   const FixedPointAccum z_b_pow_2_adj = SaturatingSub(
       FixedPointAccum::FromRaw(SaturatingRoundingMultiplyByPOTParam(
-          InputIntegerBits - z_b_headroom, 31 - kAccumIntegerBits)),
+          static_cast<int32_t>(InputIntegerBits - z_b_headroom),
+          31 - kAccumIntegerBits)),
       shifted_quarter);
 
   const FixedPoint0 r = FixedPoint0::FromRaw(std::min(r_a_raw, r_b_raw));

code/components/tfmicro/tensorflow/lite/kernels/internal/cppmath.h

@@ -19,9 +19,8 @@ limitations under the License.
 
 namespace tflite {
 
-#if defined(TF_LITE_USE_GLOBAL_CMATH_FUNCTIONS) ||                           \
-    (defined(__ANDROID__) && !defined(__NDK_MAJOR__)) || defined(ARDUINO) || \
-    defined(__ZEPHYR__)
+#if defined(TF_LITE_USE_GLOBAL_CMATH_FUNCTIONS) || \
+    (defined(__ANDROID__) && !defined(__NDK_MAJOR__)) || defined(__ZEPHYR__)
 #define TF_LITE_GLOBAL_STD_PREFIX
 #else
 #define TF_LITE_GLOBAL_STD_PREFIX std
@@ -34,6 +33,7 @@ namespace tflite {
   }
 
 DECLARE_STD_GLOBAL_SWITCH1(TfLiteRound, round);
+DECLARE_STD_GLOBAL_SWITCH1(TfLiteExpm1, expm1);
 
 }  // namespace tflite
 

code/components/tfmicro/tensorflow/lite/kernels/internal/optimized/neon_check.h

@@ -15,26 +15,6 @@ limitations under the License.
 #ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_OPTIMIZED_NEON_CHECK_H_
 #define TENSORFLOW_LITE_KERNELS_INTERNAL_OPTIMIZED_NEON_CHECK_H_
 
-#if defined(__ARM_NEON__) || defined(__ARM_NEON)
-#define USE_NEON
-#include <arm_neon.h>
-#endif
-
-#if defined __GNUC__ && defined __SSE4_1__ && !defined TF_LITE_DISABLE_X86_NEON
-#define USE_NEON
-#include "NEON_2_SSE.h"
-#endif
-
-// NEON_OR_PORTABLE(SomeFunc, args) calls NeonSomeFunc(args) if USE_NEON is
-// defined, PortableSomeFunc(args) otherwise.
-#ifdef USE_NEON
-// Always use Neon code
-#define NEON_OR_PORTABLE(funcname, ...) Neon##funcname(__VA_ARGS__)
-
-#else
-// No NEON available: Use Portable code
-#define NEON_OR_PORTABLE(funcname, ...) Portable##funcname(__VA_ARGS__)
-
-#endif  // defined(USE_NEON)
+// TFLM does not need to utilize any Neon optimizations.
 
 #endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_OPTIMIZED_NEON_CHECK_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/portable_tensor.h

@@ -15,7 +15,6 @@ limitations under the License.
 #ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_PORTABLE_TENSOR_H_
 #define TENSORFLOW_LITE_KERNELS_INTERNAL_PORTABLE_TENSOR_H_
 
-#include <complex>
 #include <vector>
 
 #include "tensorflow/lite/c/common.h"

code/components/tfmicro/tensorflow/lite/kernels/internal/quantization_util.cc

@@ -289,7 +289,7 @@ void PreprocessSoftmaxScaling(double beta, double input_scale,
     input_beta_real_multiplier = (1ll << 31) - 1.0;
   }
 #else   // TFLITE_EMULATE_FLOAT
-  const double input_beta_real_multiplier = std::min(
+  const double input_beta_real_multiplier = std::min<double>(
       beta * input_scale * (1 << (31 - input_integer_bits)), (1ll << 31) - 1.0);
 #endif  // TFLITE_EMULATE_FLOAT
 

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/add.h

@@ -15,6 +15,8 @@ limitations under the License.
 #ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ADD_H_
 #define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ADD_H_
 
+#include <type_traits>
+
 #include "fixedpoint/fixedpoint.h"
 #include "tensorflow/lite/kernels/internal/common.h"
 
@@ -27,25 +29,14 @@ inline void Add(const ArithmeticParams& params,
                 const RuntimeShape& input1_shape, const T* input1_data,
                 const RuntimeShape& input2_shape, const T* input2_data,
                 const RuntimeShape& output_shape, T* output_data) {
+  T activation_min, activation_max;
+  GetActivationParams(params, &activation_min, &activation_max);
+
   const int flat_size =
       MatchingElementsSize(input1_shape, input2_shape, output_shape);
   for (int i = 0; i < flat_size; ++i) {
     output_data[i] = ActivationFunctionWithMinMax(
-        input1_data[i] + input2_data[i], params.quantized_activation_min,
-        params.quantized_activation_max);
-  }
-}
-
-inline void Add(const ArithmeticParams& params,
-                const RuntimeShape& input1_shape, const float* input1_data,
-                const RuntimeShape& input2_shape, const float* input2_data,
-                const RuntimeShape& output_shape, float* output_data) {
-  const int flat_size =
-      MatchingElementsSize(input1_shape, input2_shape, output_shape);
-  for (int i = 0; i < flat_size; i++) {
-    auto x = input1_data[i] + input2_data[i];
-    output_data[i] = ActivationFunctionWithMinMax(
-        x, params.float_activation_min, params.float_activation_max);
+        input1_data[i] + input2_data[i], activation_min, activation_max);
   }
 }
 
@@ -202,21 +193,12 @@ inline void Add(const ArithmeticParams& params,
   }
 }
 
-// TODO(jiawen): We can implement BroadcastAdd on buffers of arbitrary
-// dimensionality if the runtime code does a single loop over one dimension
-// that handles broadcasting as the base case. The code generator would then
-// generate max(D1, D2) nested for loops.
-// TODO(benoitjacob): BroadcastAdd is intentionally duplicated from
-// reference_ops.h. Once an optimized version is implemented and NdArrayDesc<T>
-// is no longer referenced in this file, move NdArrayDesc<T> from types.h to
-// reference_ops.h.
-inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
-                               const RuntimeShape& input1_shape,
-                               const float* input1_data,
-                               const RuntimeShape& input2_shape,
-                               const float* input2_data,
-                               const RuntimeShape& output_shape,
-                               float* output_data) {
+template <typename T>
+inline typename std::enable_if<!is_small_integer<T>::value, void>::type
+BroadcastAdd4DSlow(const ArithmeticParams& params,
+                   const RuntimeShape& input1_shape, const T* input1_data,
+                   const RuntimeShape& input2_shape, const T* input2_data,
+                   const RuntimeShape& output_shape, T* output_data) {
   NdArrayDesc<4> desc1;
   NdArrayDesc<4> desc2;
   NdArrayDescsForElementwiseBroadcast(input1_shape, input2_shape, &desc1,
@@ -224,6 +206,9 @@ inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
   const RuntimeShape extended_output_shape =
       RuntimeShape::ExtendedShape(4, output_shape);
 
+  T activation_min, activation_max;
+  GetActivationParams(params, &activation_min, &activation_max);
+
   // In Tensorflow, the dimensions are canonically named (batch_number, row,
   // col, channel), with extents (batches, height, width, depth), with the
   // trailing dimension changing most rapidly (channels has the smallest stride,
@@ -240,51 +225,10 @@ inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
       for (int x = 0; x < extended_output_shape.Dims(2); ++x) {
         for (int c = 0; c < extended_output_shape.Dims(3); ++c) {
           output_data[Offset(extended_output_shape, b, y, x, c)] =
-              ActivationFunctionWithMinMax(
+              ActivationFunctionWithMinMax<T>(
                   input1_data[SubscriptToIndex(desc1, b, y, x, c)] +
                       input2_data[SubscriptToIndex(desc2, b, y, x, c)],
-                  params.float_activation_min, params.float_activation_max);
-        }
-      }
-    }
-  }
-}
-
-inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
-                               const RuntimeShape& input1_shape,
-                               const int32_t* input1_data,
-                               const RuntimeShape& input2_shape,
-                               const int32_t* input2_data,
-                               const RuntimeShape& output_shape,
-                               int32_t* output_data) {
-  NdArrayDesc<4> desc1;
-  NdArrayDesc<4> desc2;
-  NdArrayDescsForElementwiseBroadcast(input1_shape, input2_shape, &desc1,
-                                      &desc2);
-  const RuntimeShape extended_output_shape =
-      RuntimeShape::ExtendedShape(4, output_shape);
-
-  // In Tensorflow, the dimensions are canonically named (batch_number, row,
-  // col, channel), with extents (batches, height, width, depth), with the
-  // trailing dimension changing most rapidly (channels has the smallest stride,
-  // typically 1 element).
-  //
-  // In generated C code, we store arrays with the dimensions reversed. The
-  // first dimension has smallest stride.
-  //
-  // We name our variables by their Tensorflow convention, but generate C code
-  // nesting loops such that the innermost loop has the smallest stride for the
-  // best cache behavior.
-  for (int b = 0; b < extended_output_shape.Dims(0); ++b) {
-    for (int y = 0; y < extended_output_shape.Dims(1); ++y) {
-      for (int x = 0; x < extended_output_shape.Dims(2); ++x) {
-        for (int c = 0; c < extended_output_shape.Dims(3); ++c) {
-          output_data[Offset(extended_output_shape, b, y, x, c)] =
-              ActivationFunctionWithMinMax(
-                  input1_data[SubscriptToIndex(desc1, b, y, x, c)] +
-                      input2_data[SubscriptToIndex(desc2, b, y, x, c)],
-                  params.quantized_activation_min,
-                  params.quantized_activation_max);
+                  activation_min, activation_max);
         }
       }
     }
@@ -295,10 +239,11 @@ inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
 // is 32-bit for both cases. The overflow does not happen due to the
 // choice of the shift (20 or 15, accordingly - see add.cc for more comments).
 template <typename T>
-inline void BroadcastAdd4DSlow(
-    const ArithmeticParams& params, const RuntimeShape& input1_shape,
-    const T* input1_data, const RuntimeShape& input2_shape,
-    const T* input2_data, const RuntimeShape& output_shape, T* output_data) {
+inline typename std::enable_if<is_small_integer<T>::value, void>::type
+BroadcastAdd4DSlow(const ArithmeticParams& params,
+                   const RuntimeShape& input1_shape, const T* input1_data,
+                   const RuntimeShape& input2_shape, const T* input2_data,
+                   const RuntimeShape& output_shape, T* output_data) {
   NdArrayDesc<4> desc1;
   NdArrayDesc<4> desc2;
   NdArrayDescsForElementwiseBroadcast(input1_shape, input2_shape, &desc1,

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/add_n.h

@@ -0,0 +1,86 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ADD_N_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ADD_N_H_
+
+#include <algorithm>
+#include <limits>
+
+#include "tensorflow/lite/kernels/internal/common.h"
+
+namespace tflite {
+namespace reference_ops {
+
+// T is expected to be either float or int.
+template <typename T>
+inline void AddN(const RuntimeShape& input_shape, const size_t num_inputs,
+                 const T* const* input_data, T* output_data) {
+  // All inputs and output should have the same shape, this is checked during
+  // Prepare stage.
+  const size_t size = input_shape.FlatSize();
+  for (size_t i = 0; i < size; ++i) {
+    T x = 0;
+    for (size_t j = 0; j < num_inputs; ++j) {
+      x += input_data[j][i];
+    }
+    output_data[i] = x;
+  }
+}
+
+inline void AddN(const ArithmeticParams& params,
+                 const RuntimeShape& input_shape, const size_t num_inputs,
+                 const int8_t* const* input_data, int8_t* output_data) {
+  TFLITE_DCHECK_LE(params.quantized_activation_min,
+                   params.quantized_activation_max);
+  // Input offset is negative input zero point. Activation tensors are
+  // asymmetric quantized so they span the full int8 range.
+  // All inputs should have same zero-point and scale, this is checked during
+  // Prepare stage.
+  TFLITE_DCHECK_GE(-params.input1_offset, std::numeric_limits<int8_t>::min());
+  TFLITE_DCHECK_LE(-params.input1_offset, std::numeric_limits<int8_t>::max());
+
+  // All inputs and output should have the same shape, this is checked during
+  // Prepare stage.
+  const size_t size = input_shape.FlatSize();
+  for (size_t i = 0; i < size; ++i) {
+    // accumulate in scaled_x before clamping to avoid overflow
+    const int32_t x = params.input1_offset;  // x = 0
+    const int32_t shifted_x = x * (1 << params.left_shift);
+    int32_t scaled_x = MultiplyByQuantizedMultiplierSmallerThanOneExp(
+        shifted_x, params.input1_multiplier, params.input1_shift);
+
+    for (size_t j = 0; j < num_inputs; ++j) {
+      const int32_t y = params.input1_offset + input_data[j][i];
+      const int32_t shifted_y = y * (1 << params.left_shift);
+      int32_t scaled_y = MultiplyByQuantizedMultiplierSmallerThanOneExp(
+          shifted_y, params.input1_multiplier, params.input1_shift);
+      scaled_x += scaled_y;
+    }
+
+    const int32_t raw_output =
+        MultiplyByQuantizedMultiplierSmallerThanOneExp(
+            scaled_x, params.output_multiplier, params.output_shift) +
+        params.output_offset;
+    const int32_t clamped_output =
+        std::min(params.quantized_activation_max,
+                 std::max(params.quantized_activation_min, raw_output));
+    output_data[i] = static_cast<int8_t>(clamped_output);
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ADD_N_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/arg_min_max.h

@@ -15,12 +15,23 @@ limitations under the License.
 #ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ARG_MIN_MAX_H_
 #define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ARG_MIN_MAX_H_
 
+#include <functional>
+
 #include "tensorflow/lite/kernels/internal/types.h"
 
 namespace tflite {
 
 namespace reference_ops {
 
+template <typename T>
+std::function<bool(T, T)> GetComparefunction(bool is_arg_max) {
+  if (is_arg_max) {
+    return std::greater<T>();
+  } else {
+    return std::less<T>();
+  }
+}
+
 template <typename T1, typename T2, typename T3, typename Cmp>
 void ArgMinMax(const RuntimeShape& input1_shape, const T1* input1_data,
                const T3* input2_data, const RuntimeShape& output_shape,
@@ -62,6 +73,15 @@ void ArgMinMax(const RuntimeShape& input1_shape, const T1* input1_data,
     }
   }
 }
+
+template <typename T1, typename T2, typename T3>
+void ArgMinMax(const RuntimeShape& input1_shape, const T1* input1_data,
+               const T3* input2_data, const RuntimeShape& output_shape,
+               T2* output_data, const bool is_arg_max) {
+  ArgMinMax(input1_shape, input1_data, input2_data, output_shape, output_data,
+            GetComparefunction<T1>(is_arg_max));
+}
+
 }  // namespace reference_ops
 }  // namespace tflite
 

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/batch_matmul.h

@@ -0,0 +1,275 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_BATCH_MATMUL_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_BATCH_MATMUL_H_
+
+#include <algorithm>
+#include <cstdint>
+
+#include "tensorflow/lite/kernels/internal/common.h"
+#include "tensorflow/lite/kernels/internal/compatibility.h"
+#include "tensorflow/lite/kernels/internal/tensor_utils_common.h"
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+namespace reference_ops {
+namespace batch_matmul {
+
+// Determine which dimension is the broadcast dimension.
+inline int broadcast_dim(int lhs_dim, int rhs_dim) {
+  if (lhs_dim == rhs_dim) return lhs_dim;
+  if (lhs_dim == 1) return rhs_dim;
+  TFLITE_DCHECK_EQ(rhs_dim, 1);
+  return lhs_dim;
+}
+
+// Compute the "extent" for iterating on this dimension.
+// If we are broadcasting, then don't advance (i.e return 0).
+inline int extent(const RuntimeShape& shape, int x) {
+  if (shape.Dims(x) == 1) {
+    return 0;
+  }
+  int prod = 1;
+  for (int i = x + 1; i < shape.DimensionsCount(); ++i) {
+    prod *= shape.Dims(i);
+  }
+  return prod;
+}
+
+}  // namespace batch_matmul
+
+template <typename Ta, typename Tb, typename Tout>
+inline void BatchMatMul(const RuntimeShape& lhs_shape, const Ta* lhs_data,
+                        const RuntimeShape& rhs_shape, const Tb* rhs_data,
+                        const RuntimeShape& output_shape, Tout* output_data) {
+  const RuntimeShape extended_lhs_shape =
+      RuntimeShape::ExtendedShape(5, lhs_shape);
+  const RuntimeShape extended_rhs_shape =
+      RuntimeShape::ExtendedShape(5, rhs_shape);
+
+  const int batch_dim0 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(0), extended_rhs_shape.Dims(0));
+  const int batch_dim1 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(1), extended_rhs_shape.Dims(1));
+  const int batch_dim2 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(2), extended_rhs_shape.Dims(2));
+
+  const int lhs_ext0 = batch_matmul::extent(extended_lhs_shape, 0);
+  const int lhs_ext1 = batch_matmul::extent(extended_lhs_shape, 1);
+  const int lhs_ext2 = batch_matmul::extent(extended_lhs_shape, 2);
+  const int rhs_ext0 = batch_matmul::extent(extended_rhs_shape, 0);
+  const int rhs_ext1 = batch_matmul::extent(extended_rhs_shape, 1);
+  const int rhs_ext2 = batch_matmul::extent(extended_rhs_shape, 2);
+
+  // Set params for each matrix multiply.
+  const int lhs_rows = extended_lhs_shape.Dims(3);
+  const int rhs_cols = extended_rhs_shape.Dims(4);
+  const int accum_depth = extended_lhs_shape.Dims(4);
+
+  for (int b0 = 0; b0 < batch_dim0; ++b0) {
+    const Ta* lhs_ptr0 = lhs_data + (b0 * lhs_ext0);
+    const Tb* rhs_ptr0 = rhs_data + (b0 * rhs_ext0);
+    for (int b1 = 0; b1 < batch_dim1; ++b1) {
+      const Ta* lhs_ptr1 = lhs_ptr0 + b1 * lhs_ext1;
+      const Tb* rhs_ptr1 = rhs_ptr0 + b1 * rhs_ext1;
+      for (int b2 = 0; b2 < batch_dim2; ++b2) {
+        const Ta* lhs_ptr2 = lhs_ptr1 + b2 * lhs_ext2;
+        const Tb* rhs_ptr2 = rhs_ptr1 + b2 * rhs_ext2;
+        Tout* out_ptr = output_data + ((b0 * batch_dim1 * batch_dim2) +
+                                       b1 * batch_dim2 + b2) *
+                                          lhs_rows * rhs_cols;
+        for (int j = 0; j < rhs_cols; ++j) {
+          for (int i = 0; i < lhs_rows; ++i) {
+            Tout total = 0;
+            for (int k = 0; k < accum_depth; ++k) {
+              total += static_cast<Tout>(lhs_ptr2[accum_depth * i + k]) *
+                       static_cast<Tout>(rhs_ptr2[j * accum_depth + k]);
+            }
+            int idx = lhs_rows * j + i;
+            out_ptr[idx] = total;
+          }
+        }
+      }
+    }
+  }
+}
+
+inline void BatchMatMul(const RuntimeShape& lhs_shape, const int8_t* lhs_data,
+                        const RuntimeShape& rhs_shape, const int8_t* rhs_data,
+                        const float* scaling_factors,
+                        const int32_t* input_offset, int32_t* row_sums,
+                        const RuntimeShape& output_shape, float* output_data,
+                        bool* compute_row_sums) {
+  const RuntimeShape extended_lhs_shape =
+      RuntimeShape::ExtendedShape(5, lhs_shape);
+  const RuntimeShape extended_rhs_shape =
+      RuntimeShape::ExtendedShape(5, rhs_shape);
+
+  const int batch_dim0 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(0), extended_rhs_shape.Dims(0));
+  const int batch_dim1 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(1), extended_rhs_shape.Dims(1));
+  const int batch_dim2 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(2), extended_rhs_shape.Dims(2));
+
+  const int lhs_ext0 = batch_matmul::extent(extended_lhs_shape, 0);
+  const int lhs_ext1 = batch_matmul::extent(extended_lhs_shape, 1);
+  const int lhs_ext2 = batch_matmul::extent(extended_lhs_shape, 2);
+  const int rhs_ext0 = batch_matmul::extent(extended_rhs_shape, 0);
+  const int rhs_ext1 = batch_matmul::extent(extended_rhs_shape, 1);
+  const int rhs_ext2 = batch_matmul::extent(extended_rhs_shape, 2);
+
+  // Set params for each matrix multiply.
+  const int lhs_rows = extended_lhs_shape.Dims(3);
+  const int rhs_cols = extended_rhs_shape.Dims(4);
+  const int accum_depth = extended_lhs_shape.Dims(4);
+
+  const int ioff_ext0 = rhs_ext0 == 0 ? 0 : rhs_cols;
+  const int ioff_ext1 = rhs_ext1 == 0 ? 0 : rhs_cols;
+  const int ioff_ext2 = rhs_ext2 == 0 ? 0 : rhs_cols;
+  const int woff_ext0 = lhs_ext0 == 0 ? 0 : lhs_rows;
+  const int woff_ext1 = lhs_ext1 == 0 ? 0 : lhs_rows;
+  const int woff_ext2 = lhs_ext2 == 0 ? 0 : lhs_rows;
+
+  if (!compute_row_sums || *compute_row_sums) {
+    int num_weights_matrices = 1;
+    for (int i = 1; i < extended_lhs_shape.DimensionsCount() - 2; ++i) {
+      num_weights_matrices *= extended_lhs_shape.Dims(i);
+    }
+    tensor_utils::ReductionSumVector(
+        lhs_data, row_sums, num_weights_matrices * lhs_rows, accum_depth);
+    if (compute_row_sums) {
+      *compute_row_sums = false;
+    }
+  }
+
+  for (int b0 = 0; b0 < batch_dim0; ++b0) {
+    const int8_t* lhs_ptr0 = lhs_data + (b0 * lhs_ext0);
+    const int8_t* rhs_ptr0 = rhs_data + (b0 * rhs_ext0);
+    const int32_t* ioff_ptr0 = input_offset + (b0 * ioff_ext0);
+    const float* scale_ptr0 = scaling_factors + (b0 * ioff_ext0);
+    const int32_t* woff_ptr0 = row_sums + (b0 * woff_ext0);
+    for (int b1 = 0; b1 < batch_dim1; ++b1) {
+      const int8_t* lhs_ptr1 = lhs_ptr0 + b1 * lhs_ext1;
+      const int8_t* rhs_ptr1 = rhs_ptr0 + b1 * rhs_ext1;
+      const int32_t* ioff_ptr1 = ioff_ptr0 + (b1 * ioff_ext1);
+      const float* scale_ptr1 = scale_ptr0 + (b1 * ioff_ext1);
+      const int32_t* woff_ptr1 = woff_ptr0 + (b1 * woff_ext1);
+      for (int b2 = 0; b2 < batch_dim2; ++b2) {
+        const int8_t* lhs_ptr2 = lhs_ptr1 + b2 * lhs_ext2;
+        const int8_t* rhs_ptr2 = rhs_ptr1 + b2 * rhs_ext2;
+        const int32_t* ioff_ptr2 = ioff_ptr1 + (b2 * ioff_ext2);
+        const float* scale_ptr2 = scale_ptr1 + (b2 * ioff_ext2);
+        const int32_t* woff_ptr2 = woff_ptr1 + (b2 * woff_ext2);
+        float* out_ptr = output_data + ((b0 * batch_dim1 * batch_dim2) +
+                                        b1 * batch_dim2 + b2) *
+                                           lhs_rows * rhs_cols;
+        for (int j = 0; j < rhs_cols; ++j) {
+          const float batch_scaling_factor = scale_ptr2[j];
+          const float batch_offset = static_cast<float>(ioff_ptr2[j]);
+          for (int i = 0; i < lhs_rows; ++i) {
+            int32_t total = 0;
+            for (int k = 0; k < accum_depth; ++k) {
+              total +=
+                  lhs_ptr2[accum_depth * i + k] * rhs_ptr2[j * accum_depth + k];
+            }
+            int32_t row_sum = woff_ptr2[i];
+            total -= row_sum * batch_offset;
+            int idx = lhs_rows * j + i;
+            out_ptr[idx] += batch_scaling_factor * total;
+          }
+        }
+      }
+    }
+  }
+}
+
+template <typename T, typename AccumT>
+inline void BatchMatMul(const FullyConnectedParams& params,
+                        const RuntimeShape& lhs_shape, const T* lhs_data,
+                        const RuntimeShape& rhs_shape, const T* rhs_data,
+                        const RuntimeShape& output_shape, T* output_data) {
+  const RuntimeShape extended_lhs_shape =
+      RuntimeShape::ExtendedShape(5, lhs_shape);
+  const RuntimeShape extended_rhs_shape =
+      RuntimeShape::ExtendedShape(5, rhs_shape);
+
+  const int batch_dim0 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(0), extended_rhs_shape.Dims(0));
+  const int batch_dim1 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(1), extended_rhs_shape.Dims(1));
+  const int batch_dim2 = batch_matmul::broadcast_dim(
+      extended_lhs_shape.Dims(2), extended_rhs_shape.Dims(2));
+
+  const int lhs_ext0 = batch_matmul::extent(extended_lhs_shape, 0);
+  const int lhs_ext1 = batch_matmul::extent(extended_lhs_shape, 1);
+  const int lhs_ext2 = batch_matmul::extent(extended_lhs_shape, 2);
+  const int rhs_ext0 = batch_matmul::extent(extended_rhs_shape, 0);
+  const int rhs_ext1 = batch_matmul::extent(extended_rhs_shape, 1);
+  const int rhs_ext2 = batch_matmul::extent(extended_rhs_shape, 2);
+
+  // Set params for each matrix multiply.
+  const int lhs_rows = extended_lhs_shape.Dims(3);
+  const int rhs_cols = extended_rhs_shape.Dims(4);
+  const int accum_depth = extended_lhs_shape.Dims(4);
+
+  const int32_t input_offset = params.input_offset;
+  const int32_t filter_offset = params.weights_offset;
+  const int32_t output_offset = params.output_offset;
+  const int32_t output_multiplier = params.output_multiplier;
+  const int output_shift = params.output_shift;
+  const int32_t output_activation_min = params.quantized_activation_min;
+  const int32_t output_activation_max = params.quantized_activation_max;
+  TFLITE_DCHECK_LE(output_activation_min, output_activation_max);
+
+  for (int b0 = 0; b0 < batch_dim0; ++b0) {
+    const T* lhs_ptr0 = lhs_data + (b0 * lhs_ext0);
+    const T* rhs_ptr0 = rhs_data + (b0 * rhs_ext0);
+    for (int b1 = 0; b1 < batch_dim1; ++b1) {
+      const T* lhs_ptr1 = lhs_ptr0 + b1 * lhs_ext1;
+      const T* rhs_ptr1 = rhs_ptr0 + b1 * rhs_ext1;
+      for (int b2 = 0; b2 < batch_dim2; ++b2) {
+        const T* lhs_ptr2 = lhs_ptr1 + b2 * lhs_ext2;
+        const T* rhs_ptr2 = rhs_ptr1 + b2 * rhs_ext2;
+        T* out_ptr = output_data +
+                     ((b0 * batch_dim1 * batch_dim2) + b1 * batch_dim2 + b2) *
+                         lhs_rows * rhs_cols;
+
+        for (int j = 0; j < rhs_cols; ++j) {
+          for (int i = 0; i < lhs_rows; ++i) {
+            AccumT total = 0;
+            for (int k = 0; k < accum_depth; ++k) {
+              AccumT lhs_val = lhs_ptr2[accum_depth * i + k];
+              AccumT rhs_val = rhs_ptr2[accum_depth * j + k];
+              total += (lhs_val + filter_offset) * (rhs_val + input_offset);
+            }
+            int32_t total_scaled = MultiplyByQuantizedMultiplier(
+                total, output_multiplier, output_shift);
+            total_scaled += output_offset;
+            total_scaled = std::max(total_scaled, output_activation_min);
+            total_scaled = std::min(total_scaled, output_activation_max);
+            const int idx = lhs_rows * j + i;
+            out_ptr[idx] = static_cast<T>(total_scaled);
+          }
+        }
+      }
+    }
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_BATCH_MATMUL_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/batch_to_space_nd.h

@@ -0,0 +1,101 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_BATCH_TO_SPACE_ND_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_BATCH_TO_SPACE_ND_H_
+
+#include <cmath>
+
+#include "ruy/profiler/instrumentation.h"  // from @ruy
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+namespace reference_ops {
+
+// TODO(b/135760455): Move this method anonymous namespace in a cc file.
+inline RuntimeShape ExtendShapeBatchToSpace(const RuntimeShape& shape) {
+  if (shape.DimensionsCount() == 4) {
+    return shape;
+  }
+  RuntimeShape new_shape(4, 1);
+  new_shape.SetDim(0, shape.Dims(0));
+  new_shape.SetDim(1, shape.Dims(1));
+  new_shape.SetDim(3, shape.Dims(2));
+  return new_shape;
+}
+
+template <typename T>
+inline void BatchToSpaceND(const RuntimeShape& unextended_input1_shape,
+                           const T* input1_data,
+                           const RuntimeShape& unextended_input2_shape,
+                           const int32_t* block_shape_data,
+                           const RuntimeShape& unextended_input3_shape,
+                           const int32_t* crops_data,
+                           const RuntimeShape& unextended_output_shape,
+                           T* output_data) {
+  ruy::profiler::ScopeLabel label("BatchToSpaceND");
+  TFLITE_DCHECK_GE(unextended_input1_shape.DimensionsCount(), 3);
+  TFLITE_DCHECK_LE(unextended_input1_shape.DimensionsCount(), 4);
+  TFLITE_DCHECK_EQ(unextended_input1_shape.DimensionsCount(),
+                   unextended_output_shape.DimensionsCount());
+
+  const RuntimeShape input1_shape =
+      ExtendShapeBatchToSpace(unextended_input1_shape);
+  const RuntimeShape output_shape =
+      ExtendShapeBatchToSpace(unextended_output_shape);
+
+  const int output_width = output_shape.Dims(2);
+  const int output_height = output_shape.Dims(1);
+  const int output_batch_size = output_shape.Dims(0);
+
+  const int depth = input1_shape.Dims(3);
+  const int input_width = input1_shape.Dims(2);
+  const int input_height = input1_shape.Dims(1);
+  const int input_batch_size = input1_shape.Dims(0);
+
+  const int block_shape_height = block_shape_data[0];
+  const int block_shape_width =
+      unextended_input1_shape.DimensionsCount() == 4 ? block_shape_data[1] : 1;
+  const int crops_top = crops_data[0];
+  const int crops_left =
+      unextended_input1_shape.DimensionsCount() == 4 ? crops_data[2] : 0;
+  for (int in_batch = 0; in_batch < input_batch_size; ++in_batch) {
+    const int out_batch = in_batch % output_batch_size;
+    const int spatial_offset = in_batch / output_batch_size;
+    for (int in_h = 0; in_h < input_height; ++in_h) {
+      const int out_h = in_h * block_shape_height +
+                        spatial_offset / block_shape_width - crops_top;
+      if (out_h < 0 || out_h >= output_height) {
+        continue;
+      }
+      for (int in_w = 0; in_w < input_width; ++in_w) {
+        const int out_w = in_w * block_shape_width +
+                          spatial_offset % block_shape_width - crops_left;
+
+        if (out_w < 0 || out_w >= output_width) {
+          continue;
+        }
+        T* out = output_data + Offset(output_shape, out_batch, out_h, out_w, 0);
+        const T* in =
+            input1_data + Offset(input1_shape, in_batch, in_h, in_w, 0);
+        memcpy(out, in, depth * sizeof(T));
+      }
+    }
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_BATCH_TO_SPACE_ND_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/binary_function.h

@@ -23,9 +23,6 @@ namespace tflite {
 
 namespace reference_ops {
 
-// TODO(ycling): Refactoring. Remove BroadcastLogical and use the more
-// generalized and efficient BroadcastBinaryFunction.
-//
 // Also appears to duplicate MinimumMaximum.
 //
 // R: Result type. T1: Input 1 type. T2: Input 2 type.
@@ -63,7 +60,6 @@ inline void BroadcastBinaryFunction4DSlow(
 }
 
 // R: Result type. T1: Input 1 type. T2: Input 2 type.
-// TODO(renjieliu): Refactor other binary functions to use this one.
 template <typename R, typename T1, typename T2>
 inline void BinaryFunction(const RuntimeShape& input1_shape,
                            const T1* input1_data,

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/concatenation.h

@@ -68,8 +68,7 @@ inline void Concatenation(const ConcatenationParams& params,
   }
 }
 
-// TODO(prabhumk): This is the same as the optimized implementation.
-// TODO(prabhumk): The quantized implementation of concatentation isn't fully
+// TODO(b/174275780): The quantized implementation of concatentation isn't fully
 // quantized as it takes scale as a floating point value. This should be fixed
 // when optimizng this routine further.
 inline void ConcatenationWithScaling(const ConcatenationParams& params,

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/conv.h

@@ -15,16 +15,13 @@ limitations under the License.
 #ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_CONV_H_
 #define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_CONV_H_
 
-#include "tensorflow/lite/kernels/internal/types.h"
 #include "tensorflow/lite/kernels/internal/common.h"
-
-
+#include "tensorflow/lite/kernels/internal/types.h"
 
 namespace tflite {
 
 namespace reference_ops {
 
-
 inline void Conv(const ConvParams& params, const RuntimeShape& input_shape,
                  const float* input_data, const RuntimeShape& filter_shape,
                  const float* filter_data, const RuntimeShape& bias_shape,
@@ -108,8 +105,8 @@ inline void Conv(const ConvParams& params, const RuntimeShape& input_shape,
                  uint8_t* output_data, const RuntimeShape& im2col_shape,
                  uint8_t* im2col_data, void* cpu_backend_context) {
   (void)cpu_backend_context;  // only used in optimized code.
-  (void)im2col_data;   // only used in optimized code.
-  (void)im2col_shape;  // only used in optimized code.
+  (void)im2col_data;          // only used in optimized code.
+  (void)im2col_shape;         // only used in optimized code.
   const int stride_width = params.stride_width;
   const int stride_height = params.stride_height;
   const int dilation_width_factor = params.dilation_width_factor;

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/cumsum.h

@@ -0,0 +1,175 @@
+/* Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_CUMSUM_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_CUMSUM_H_
+
+#include <algorithm>
+#include <cstdint>
+#include <limits>
+
+#include "tensorflow/lite/kernels/internal/common.h"
+#include "tensorflow/lite/kernels/internal/compatibility.h"
+
+namespace tflite {
+namespace reference_ops {
+
+template <typename T>
+inline void CumSum(const T* input_data, const RuntimeShape& shape, int32_t axis,
+                   bool exclusive, bool reverse, T* output_data) {
+  const int32_t rank = shape.DimensionsCount();
+  TFLITE_DCHECK_GE(rank, 1);
+  TFLITE_DCHECK_GE(axis, 0);
+  TFLITE_DCHECK_LT(axis, rank);
+
+  size_t inner = 1;
+  size_t outer = 1;
+  size_t depth = 1;
+  for (int32_t i = 0; i < rank; i++) {
+    if (i < axis)
+      inner *= shape.Dims(i);
+    else if (i > axis)
+      outer *= shape.Dims(i);
+    else
+      depth = shape.Dims(i);
+  }
+
+  for (size_t outer_index = 0; outer_index < outer; outer_index++) {
+    size_t outer_index_adj;
+    if (reverse)
+      outer_index_adj = (outer - 1) - outer_index;
+    else
+      outer_index_adj = outer_index;
+    for (size_t inner_index = 0; inner_index < inner; inner_index++) {
+      T accumulator = 0;
+      size_t inner_index_adj;
+      if (reverse)
+        inner_index_adj = (inner - 1) - inner_index;
+      else
+        inner_index_adj = inner_index;
+      for (size_t depth_index = 0; depth_index < depth; depth_index++) {
+        size_t depth_index_adj;
+        if (reverse)
+          depth_index_adj = (depth - 1) - depth_index;
+        else
+          depth_index_adj = depth_index;
+
+        size_t index = outer_index_adj;
+        index += inner_index_adj * depth * outer;
+        index += depth_index_adj * outer;
+
+        if (exclusive) {
+          output_data[index] = accumulator;
+          accumulator += input_data[index];
+        } else {
+          accumulator += input_data[index];
+          output_data[index] = accumulator;
+        }
+      }
+    }
+  }
+}
+
+//
+// Quantized INT8 CUMSUM
+//
+inline void CumSum(const ArithmeticParams& params, const int8_t* input_data,
+                   const RuntimeShape& shape, int32_t axis, bool exclusive,
+                   bool reverse, int8_t* output_data) {
+  TFLITE_DCHECK_LE(params.quantized_activation_min,
+                   params.quantized_activation_max);
+  // Input offset is negative input zero point. Activation tensors are
+  // asymmetric quantized so they span the full int8 range.
+  // All inputs should have same zero-point and scale, this is checked during
+  // Prepare stage.
+  TFLITE_DCHECK_GE(-params.input1_offset, std::numeric_limits<int8_t>::min());
+  TFLITE_DCHECK_LE(-params.input1_offset, std::numeric_limits<int8_t>::max());
+
+  const int32_t rank = shape.DimensionsCount();
+  TFLITE_DCHECK_GE(rank, 1);
+  TFLITE_DCHECK_GE(axis, 0);
+  TFLITE_DCHECK_LT(axis, rank);
+
+  size_t inner = 1;
+  size_t outer = 1;
+  size_t depth = 1;
+  for (int32_t i = 0; i < rank; i++) {
+    if (i < axis)
+      inner *= shape.Dims(i);
+    else if (i > axis)
+      outer *= shape.Dims(i);
+    else
+      depth = shape.Dims(i);
+  }
+
+  for (size_t outer_index = 0; outer_index < outer; outer_index++) {
+    size_t outer_index_adj;
+    if (reverse)
+      outer_index_adj = (outer - 1) - outer_index;
+    else
+      outer_index_adj = outer_index;
+    for (size_t inner_index = 0; inner_index < inner; inner_index++) {
+      int32_t accumulator = params.input1_offset;  // accumulator = 0
+      accumulator *= (1 << params.left_shift);
+      accumulator = MultiplyByQuantizedMultiplierSmallerThanOneExp(
+          accumulator, params.input1_multiplier, params.input1_shift);
+
+      size_t inner_index_adj;
+      if (reverse)
+        inner_index_adj = (inner - 1) - inner_index;
+      else
+        inner_index_adj = inner_index;
+
+      for (size_t depth_index = 0; depth_index < depth; depth_index++) {
+        size_t depth_index_adj;
+        if (reverse)
+          depth_index_adj = (depth - 1) - depth_index;
+        else
+          depth_index_adj = depth_index;
+
+        size_t index = outer_index_adj;
+        index += inner_index_adj * depth * outer;
+        index += depth_index_adj * outer;
+
+        const int32_t y = params.input1_offset + input_data[index];
+        const int32_t shifted_y = y * (1 << params.left_shift);
+        const int32_t scaled_y = MultiplyByQuantizedMultiplierSmallerThanOneExp(
+            shifted_y, params.input1_multiplier, params.input1_shift);
+
+        int32_t scaled_output;
+        if (exclusive) {
+          scaled_output = accumulator;
+          accumulator += scaled_y;
+        } else {
+          accumulator += scaled_y;
+          scaled_output = accumulator;
+        }
+
+        const int32_t raw_output =
+            MultiplyByQuantizedMultiplierSmallerThanOneExp(
+                scaled_output, params.output_multiplier, params.output_shift) +
+            params.output_offset;
+        const int32_t clamped_output =
+            std::min(params.quantized_activation_max,
+                     std::max(params.quantized_activation_min, raw_output));
+        output_data[index] = static_cast<int8_t>(clamped_output);
+      }
+    }
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_CUMSUM_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/depth_to_space.h

@@ -0,0 +1,79 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_DEPTH_TO_SPACE_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_DEPTH_TO_SPACE_H_
+
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+namespace reference_ops {
+
+template <typename T>
+inline void DepthToSpace(const tflite::DepthToSpaceParams& op_params,
+                         const RuntimeShape& unextended_input_shape,
+                         const T* input_data,
+                         const RuntimeShape& unextended_output_shape,
+                         T* output_data) {
+  TFLITE_DCHECK_LE(unextended_input_shape.DimensionsCount(), 4);
+  TFLITE_DCHECK_LE(unextended_output_shape.DimensionsCount(), 4);
+  const RuntimeShape input_shape =
+      RuntimeShape::ExtendedShape(4, unextended_input_shape);
+  const RuntimeShape output_shape =
+      RuntimeShape::ExtendedShape(4, unextended_output_shape);
+
+  const int input_depth = input_shape.Dims(3);
+  const int input_width = input_shape.Dims(2);
+  const int input_height = input_shape.Dims(1);
+  const int input_batch = input_shape.Dims(0);
+
+  const int output_depth = output_shape.Dims(3);
+  const int output_width = output_shape.Dims(2);
+  const int output_height = output_shape.Dims(1);
+  const int output_batch = output_shape.Dims(0);
+
+  const int32_t block_size = op_params.block_size;
+
+  TFLITE_DCHECK_EQ(input_width * block_size, output_width);
+  TFLITE_DCHECK_EQ(input_height * block_size, output_height);
+  TFLITE_DCHECK_EQ(input_depth, output_depth * block_size * block_size);
+  TFLITE_DCHECK_EQ(input_batch, output_batch);
+
+  for (int out_b = 0; out_b < output_batch; ++out_b) {
+    for (int out_h = 0; out_h < output_height; ++out_h) {
+      for (int out_w = 0; out_w < output_width; ++out_w) {
+        for (int out_d = 0; out_d < output_depth; ++out_d) {
+          const int in_d =
+              out_d + ((out_h % block_size) * block_size + out_w % block_size) *
+                          output_depth;
+
+          const int in_w = out_w / block_size;
+          const int in_h = out_h / block_size;
+          const int in_b = out_b;
+
+          const int input_index = Offset(input_shape, in_b, in_h, in_w, in_d);
+          const int output_index =
+              Offset(output_shape, out_b, out_h, out_w, out_d);
+
+          output_data[output_index] = input_data[input_index];
+        }
+      }
+    }
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_DEPTH_TO_SPACE_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/elu.h

@@ -0,0 +1,37 @@
+/* Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ELU_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ELU_H_
+
+#include "tensorflow/lite/kernels/internal/cppmath.h"
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+
+namespace reference_ops {
+
+inline void Elu(const RuntimeShape& input_shape, const float* input_data,
+                const RuntimeShape& output_shape, float* output_data) {
+  const int flat_size = MatchingFlatSize(input_shape, output_shape);
+  for (int i = 0; i < flat_size; ++i) {
+    const float val = input_data[i];
+    output_data[i] = val < 0.0f ? TfLiteExpm1(val) : val;
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_ELU_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/exp.h

@@ -0,0 +1,38 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_EXP_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_EXP_H_
+
+#include <cmath>
+
+#include "ruy/profiler/instrumentation.h"  // from @ruy
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+namespace reference_ops {
+
+template <typename T>
+inline void Exp(const T* input_data, const size_t num_elements,
+                T* output_data) {
+  ruy::profiler::ScopeLabel label("Exp");
+  for (size_t idx = 0; idx < num_elements; ++idx) {
+    output_data[idx] = std::exp(input_data[idx]);
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_EXP_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/fill.h

@@ -0,0 +1,38 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FILL_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FILL_H_
+
+#include <cmath>
+
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+namespace reference_ops {
+
+template <typename T>
+void Fill(const RuntimeShape& value_shape, const T* value_data,
+          const RuntimeShape& output_shape, T* output_data) {
+  TFLITE_DCHECK_EQ(value_shape.DimensionsCount(), 0);
+  const int flat_size = output_shape.FlatSize();
+  for (int i = 0; i < flat_size; ++i) {
+    output_data[i] = *value_data;
+  }
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FILL_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/floor_div.h

@@ -0,0 +1,35 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FLOOR_DIV_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FLOOR_DIV_H_
+
+#include <cmath>
+#include <functional>
+
+#include "tensorflow/lite/kernels/internal/types.h"
+
+namespace tflite {
+namespace reference_ops {
+
+template <typename T>
+T FloorDiv(T input1, T input2) {
+  return std::floor(std::divides<double>()(static_cast<double>(input1),
+                                           static_cast<double>(input2)));
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FLOOR_DIV_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/floor_mod.h

@@ -0,0 +1,44 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#ifndef TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FLOOR_MOD_H_
+#define TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FLOOR_MOD_H_
+
+#include <cmath>
+#include <functional>
+
+namespace tflite {
+
+namespace reference_ops {
+
+template <typename T>
+T FloorMod(T input1, T input2) {
+  struct FloatMod {
+    float operator()(const float lhs, const float rhs) const {
+      return std::fmod(lhs, rhs);
+    }
+  };
+  using ModFunc = typename std::conditional<std::is_integral<T>::value,
+                                            std::modulus<T>, FloatMod>::type;
+  ModFunc mod_func;
+  T trunc_mod = mod_func(input1, input2);
+  return (trunc_mod != 0) && ((input2 < 0) != (trunc_mod < 0))
+             ? (trunc_mod + input2)
+             : trunc_mod;
+}
+
+}  // namespace reference_ops
+}  // namespace tflite
+
+#endif  // TENSORFLOW_LITE_KERNELS_INTERNAL_REFERENCE_FLOOR_MOD_H_

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/fully_connected.h

@@ -31,7 +31,7 @@ inline void FullyConnected(
     float* output_data) {
   const float output_activation_min = params.float_activation_min;
   const float output_activation_max = params.float_activation_max;
-  // TODO(benoitjacob): This really should be:
+  // TODO(b/62193649): This really should be:
   //     const int batches = ArraySize(output_dims, 1);
   // but the current --variable_batch hack consists in overwriting the 3rd
   // dimension with the runtime batch size, as we don't keep track for each
@@ -76,7 +76,7 @@ inline void FullyConnected(
   TFLITE_DCHECK_GE(output_shape.DimensionsCount(), 1);
 
   TFLITE_DCHECK_LE(output_activation_min, output_activation_max);
-  // TODO(benoitjacob): This really should be:
+  // TODO(b/62193649): This really should be:
   //     const int batches = ArraySize(output_dims, 1);
   // but the current --variable_batch hack consists in overwriting the 3rd
   // dimension with the runtime batch size, as we don't keep track for each
@@ -123,7 +123,7 @@ inline void FullyConnected(
 
   TFLITE_DCHECK_LE(output_activation_min, output_activation_max);
   TFLITE_DCHECK_EQ(output_offset, 0);
-  // TODO(benoitjacob): This really should be:
+  // TODO(b/62193649): This really should be:
   //     const int batches = ArraySize(output_dims, 1);
   // but the current --variable_batch hack consists in overwriting the 3rd
   // dimension with the runtime batch size, as we don't keep track for each
@@ -176,7 +176,7 @@ inline void ShuffledFullyConnected(
   TFLITE_DCHECK_GE(input_shape.DimensionsCount(), 1);
   TFLITE_DCHECK_GE(weights_shape.DimensionsCount(), 2);
   TFLITE_DCHECK_GE(output_shape.DimensionsCount(), 1);
-  // TODO(benoitjacob): This really should be:
+  // TODO(b/62193649): This really should be:
   //     const int batches = ArraySize(output_dims, 1);
   // but the current --variable_batch hack consists in overwriting the 3rd
   // dimension with the runtime batch size, as we don't keep track for each

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/integer_ops/add.h

@@ -34,55 +34,24 @@ inline void CheckArithmeticParams(const ArithmeticParams& params) {
   TFLITE_DCHECK_LE(-params.input2_offset, std::numeric_limits<int8_t>::max());
 }
 
-// Element-wise add that can often be used for inner loop of broadcast add as
-// well as the non-broadcast add.
-inline void AddElementwise(int size, const ArithmeticParams& params,
-                           const int8_t* input1_data, const int8_t* input2_data,
-                           int8_t* output_data) {
+inline void ElementWise(
+    int size, const ArithmeticParams& params, const int8_t* input1_data,
+    const int8_t* input2_data, int8_t* output_data,
+    void (*check_arithmetic_params)(const ArithmeticParams&),
+    int8_t (*binary_func)(int8_t, int8_t, const ArithmeticParams&)) {
   CheckArithmeticParams(params);
-
   for (int i = 0; i < size; ++i) {
-    const int32_t input1_val = params.input1_offset + input1_data[i];
-    const int32_t input2_val = params.input2_offset + input2_data[i];
-    const int32_t shifted_input1_val = input1_val * (1 << params.left_shift);
-    const int32_t shifted_input2_val = input2_val * (1 << params.left_shift);
-    const int32_t scaled_input1_val =
-        MultiplyByQuantizedMultiplierSmallerThanOneExp(
-            shifted_input1_val, params.input1_multiplier, params.input1_shift);
-    const int32_t scaled_input2_val =
-        MultiplyByQuantizedMultiplierSmallerThanOneExp(
-            shifted_input2_val, params.input2_multiplier, params.input2_shift);
-    const int32_t raw_sum = scaled_input1_val + scaled_input2_val;
-    const int32_t raw_output =
-        MultiplyByQuantizedMultiplierSmallerThanOneExp(
-            raw_sum, params.output_multiplier, params.output_shift) +
-        params.output_offset;
-    const int32_t clamped_output =
-        std::min(params.quantized_activation_max,
-                 std::max(params.quantized_activation_min, raw_output));
-    output_data[i] = static_cast<int8_t>(clamped_output);
+    output_data[i] = binary_func(input1_data[i], input2_data[i], params);
   }
 }
 
-inline void Add(const ArithmeticParams& params,
-                const RuntimeShape& input1_shape, const int8_t* input1_data,
-                const RuntimeShape& input2_shape, const int8_t* input2_data,
-                const RuntimeShape& output_shape, int8_t* output_data) {
-  CheckArithmeticParams(params);
-
-  const int flat_size =
-      MatchingElementsSize(input1_shape, input2_shape, output_shape);
-
-  AddElementwise(flat_size, params, input1_data, input2_data, output_data);
-}
-
-inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
-                               const RuntimeShape& input1_shape,
-                               const int8_t* input1_data,
-                               const RuntimeShape& input2_shape,
-                               const int8_t* input2_data,
-                               const RuntimeShape& output_shape,
-                               int8_t* output_data) {
+inline void BroadcastBinaryFunction4DSlow(
+    const ArithmeticParams& params, const RuntimeShape& input1_shape,
+    const int8_t* input1_data, const RuntimeShape& input2_shape,
+    const int8_t* input2_data, const RuntimeShape& output_shape,
+    int8_t* output_data,
+    void (*check_arithmetic_params)(const ArithmeticParams&),
+    int8_t (*binary_func)(int8_t, int8_t, const ArithmeticParams&)) {
   NdArrayDesc<4> desc1;
   NdArrayDesc<4> desc2;
   NdArrayDescsForElementwiseBroadcast(input1_shape, input2_shape, &desc1,
@@ -105,40 +74,70 @@ inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
     for (int y = 0; y < extended_output_shape.Dims(1); ++y) {
       for (int x = 0; x < extended_output_shape.Dims(2); ++x) {
         for (int c = 0; c < extended_output_shape.Dims(3); ++c) {
-          const int32_t input1_val =
-              params.input1_offset +
-              input1_data[SubscriptToIndex(desc1, b, y, x, c)];
-          const int32_t input2_val =
-              params.input2_offset +
-              input2_data[SubscriptToIndex(desc2, b, y, x, c)];
-          const int32_t shifted_input1_val =
-              input1_val * (1 << params.left_shift);
-          const int32_t shifted_input2_val =
-              input2_val * (1 << params.left_shift);
-          const int32_t scaled_input1_val =
-              MultiplyByQuantizedMultiplierSmallerThanOneExp(
-                  shifted_input1_val, params.input1_multiplier,
-                  params.input1_shift);
-          const int32_t scaled_input2_val =
-              MultiplyByQuantizedMultiplierSmallerThanOneExp(
-                  shifted_input2_val, params.input2_multiplier,
-                  params.input2_shift);
-          const int32_t raw_sum = scaled_input1_val + scaled_input2_val;
-          const int32_t raw_output =
-              MultiplyByQuantizedMultiplierSmallerThanOneExp(
-                  raw_sum, params.output_multiplier, params.output_shift) +
-              params.output_offset;
-          const int32_t clamped_output =
-              std::min(params.quantized_activation_max,
-                       std::max(params.quantized_activation_min, raw_output));
-          output_data[Offset(extended_output_shape, b, y, x, c)] =
-              static_cast<int8_t>(clamped_output);
+          output_data[Offset(extended_output_shape, b, y, x, c)] = binary_func(
+              input1_data[SubscriptToIndex(desc1, b, y, x, c)],
+              input2_data[SubscriptToIndex(desc2, b, y, x, c)], params);
         }
       }
     }
   }
 }
 
+inline int8_t AddFunc(int8_t x, int8_t y, const ArithmeticParams& params) {
+  const int32_t input1_val = params.input1_offset + x;
+  const int32_t input2_val = params.input2_offset + y;
+  const int32_t shifted_input1_val = input1_val * (1 << params.left_shift);
+  const int32_t shifted_input2_val = input2_val * (1 << params.left_shift);
+  const int32_t scaled_input1_val =
+      MultiplyByQuantizedMultiplierSmallerThanOneExp(
+          shifted_input1_val, params.input1_multiplier, params.input1_shift);
+  const int32_t scaled_input2_val =
+      MultiplyByQuantizedMultiplierSmallerThanOneExp(
+          shifted_input2_val, params.input2_multiplier, params.input2_shift);
+  const int32_t raw_sum = scaled_input1_val + scaled_input2_val;
+  const int32_t raw_output =
+      MultiplyByQuantizedMultiplierSmallerThanOneExp(
+          raw_sum, params.output_multiplier, params.output_shift) +
+      params.output_offset;
+  const int32_t clamped_output =
+      std::min(params.quantized_activation_max,
+               std::max(params.quantized_activation_min, raw_output));
+  return static_cast<int8_t>(clamped_output);
+}
+
+// Element-wise add that can often be used for inner loop of broadcast add as
+// well as the non-broadcast add.
+inline void AddElementwise(int size, const ArithmeticParams& params,
+                           const int8_t* input1_data, const int8_t* input2_data,
+                           int8_t* output_data) {
+  ElementWise(size, params, input1_data, input2_data, output_data,
+              CheckArithmeticParams, AddFunc);
+}
+
+inline void Add(const ArithmeticParams& params,
+                const RuntimeShape& input1_shape, const int8_t* input1_data,
+                const RuntimeShape& input2_shape, const int8_t* input2_data,
+                const RuntimeShape& output_shape, int8_t* output_data) {
+  CheckArithmeticParams(params);
+
+  const int flat_size =
+      MatchingElementsSize(input1_shape, input2_shape, output_shape);
+
+  AddElementwise(flat_size, params, input1_data, input2_data, output_data);
+}
+
+inline void BroadcastAdd4DSlow(const ArithmeticParams& params,
+                               const RuntimeShape& input1_shape,
+                               const int8_t* input1_data,
+                               const RuntimeShape& input2_shape,
+                               const int8_t* input2_data,
+                               const RuntimeShape& output_shape,
+                               int8_t* output_data) {
+  BroadcastBinaryFunction4DSlow(params, input1_shape, input1_data, input2_shape,
+                                input2_data, output_shape, output_data,
+                                CheckArithmeticParams, AddFunc);
+}
+
 }  // namespace reference_integer_ops
 }  // namespace tflite
 

code/components/tfmicro/tensorflow/lite/kernels/internal/reference/integer_ops/conv.h

@@ -101,7 +101,7 @@ inline void ConvPerChannel(
                 // long as the filter size (filter_y * filter_x * in_channel)
                 // does not exceed 2^16, which is the case in all the models
                 // we have seen so far.
-                // TODO(jianlijianli): Add a check to make sure the
+                // TODO(b/174275578): Add a check to make sure the
                 // accumulator depth is smaller than 2^16.
                 acc += filter_val * (input_val + input_offset);
               }