Code Custom C/C++ Adapter for Freezer Monitoring Lite

Remember how we created a Makefile in the root directory (/home/pi) in the previous tutorial Introduction to Custom C/C++ Adapters for Freezer Monitoring Lite. Now inside of the same folder, create another new blank file called "custom_c_adapter.cpp".

Save the file, and close the window. (You may need to type something and then delete it in order to save the blank file).

Reopen it using a C/C++ editor from your own computer. To do this, right click the file, go to Edit With > Edit With..., and point to the location of the editor executable on your local machine. We will be using Notepad++.

Your blank "custom_c_adapter.cpp" file should now be open in Notepad++. The first thing to do is add all the include statements needed for our program.

Paste the following:

#include <string>
#include <iostream>
#include <stdlib.h>
#include <time.h>
#include <stdio.h>

#include "sdk/esp_sdk.h"
#include "sdk/esp_error.h"
#include "sdk/esp_credentials.h"
#include "sdk/esp_project.h"
#include "sdk/esp_publisher.h"

using namespace std;

The first block of include statements are standard C and C++ libraries. The second block of include statements are Streaming Analytics specific SDK header files, located in the directory $STREAMING_HOME/lib/sdk. The SDK’s shared object libraries are located in $STREAMING_HOME/lib, and will be linked at build-time. These paths are defined in our Makefile.

We will be using two global variables, which can be used by any function in our .cpp file. We will declare a "g_error" of type EspError, and a "g_creds" of type EspCredentials. This declaration must be made ahead of all function declarations.

Add the following:

EspError *g_error = NULL;
EspCredentials *g_creds = NULL;

g_error will be used to hold any error data in case one of the SDK calls fail. g_creds will contain the user credentials needed to connect to the streaming project.

Our Custom C/C++ Adapter will contain three functions. Parameters omitted, they are print_error_and_exit(), exec(), and main().

The print_error_and_exit() function is called whenever a SDK call does not return 0, which indicates failure. It prints the error description along with what line it was called on, releases our global instances, and exits the program. The function itself takes two parameters: EspError *streamingError, and int line. It returns nothing. This function is optional to the functionality of our adapter, but will make debugging the program a lot easier.

Add the following after declaring global variables:

void print_error_and_exit(EspError * streamingError, int line) {
 printf("An error was encountered on line %d\n", line);
 printf("%s", esp_error_get_message(streamingError));
 printf("\nExiting\n");

 //EspCredentials freed
 if (g_creds) esp_credentials_free(g_creds, g_error);
 //stop SDK call
 esp_sdk_stop(g_error);
 //EspError freed
 if (g_error) esp_error_free(g_error);
 exit(1);
}

You are required to call esp_sdk_stop() in order to cleanly shut down the adapter. This is an SDK call.

This function takes one parameter, char const *cmd, and returns a string. It is used to output the value printed by whatever command is given in input. This function is useful because various scripts can be executed to poll external hardware sensors on our Raspberry Pi (or other remote device).

Add the following after the print_error_and_exit() function:

string exec(char const* cmd)
{
 FILE* pipe = popen(cmd, "r");
 if (!pipe) return "ERROR";
 char buffer[128];
 string result = "";
 while(!feof(pipe)) {
 if(fgets(buffer, 128, pipe) != NULL)
 result += buffer;
 }
 pclose(pipe);
 return result;
}

For example, if cmd = "echo 90" is passed in, the function will return "90". Additionally, if we had a python script called SensorPoll.py, we can pass in cmd = "sudo python SensorPoll.py". This will return the output of the sensor being polled.

The main() function is where a connection will be established between our code and the streaming lite project, and where the values are sent to the stream.

In pseudo code, it has the following functionality:

1. Declare variables containing necessary values including hostname, stream, row values to be sent
2. Calls to start the SDK
3. Sets up previously instanced error object and credentials object
4. Creates project, stream, publisher, message writer, and relative row writer objects
5. Connects to project and publisher
6. Releases credentials object
7. Starts a do/while loop
   • Creates new row definition with the relative row writer
   • Define row operation as insert
   • Inserts row fields one by one
   • Ends row definition
   • Publishes and commits row

8. Calls to stop SDK

9. Releases error object

This document will explain step -by-step which SDK calls to use for each function, and how to construct a row for any schema. Further details will be provided in chart form for calls with less straight-forward parameters.

Start by setting up the main() function:

int main(){
  //Code will go here
}

1. Declare Adapter variables

   We are now coding the main() function, and the first thing to do is declare adapter variables. The adapter variables can be split into two groups. The first group is used to connect to our Streaming Lite project, and control the quantity and rate of rows sent. The second group is a direct reflection of the stream we are writing to, with one variable for each column.

   • Declare and Define Variables in the First Group

     First group of variables are hostname, port, stream, username, password, repeat, and interval. These variables can be named whatever you wish; they are either passed into the various SDK calls, or used for loop logic.

     Variable	Explanation
     hostname	String value containing the FQDN of your remote device
     port	Integer value of the port your Streaming Lite project is running on
     stream	String value of the stream name you wish to write to
     username	String value of username you started Streaming Lite with. If you didn’t start Streaming Lite with credentials, leave blank
     password	String value of password you started Streaming Lite with. If you didn’t start Streaming Lite with credentials, leave blank
     repeat	Integer value of many times you want to send a row of data into Streaming Lite. A value of -1 will cause the program to repeat infinitely
     interval	Integer value of the number of seconds in-between sending rows

     The implementation is shown here:

     C++

     Copy

     
     string val_host = "SAPraspberrypi2-B3";
     int val_port = 9230;
     string val_stream = "isFreezerTemperatureReading";
     string val_username ="";
     string val_password ="";
     int val_repeat = -1;
     int val_interval = 1;
     

     When this C/C++ Adapter executes, it will connect with the project running on port 9230, and write to the stream "isFreezerTemperatureReading". There are no credentials, and the program will continuously send one row every second.

   • Declare and Define Column Variables in the Second Group

     Recall that the schema of the stream we are writing to is:

     

     This means we will be using four variables. Once again, naming does not matter. In this document they will be:

     C++

     Copy

     
     string val_sensorId = "RaspberryPi";
     string val_Temperature_Command = "echo 90";
     int64_t val_readingDate; //generated right before column is set
     int64_4 val_id = 0; //generated in streaming lite
     

     • string val_sensorId is the name of our sensor
     • string val_temperature holds the command to run in order to obtain the value for our Temperature column
     • int64_t val_readingDate is a timestamp, and is only initialized here. A value will be assigned to it later in the code, directly before it is used
     • int64_t val_id is our primary key, but will be auto-generated inside the streaming project. Although we still declare it, the value of val_id is arbitrary and will be overwritten once inside the streaming project. The reason we need it in our Custom C/C++ Adapter is because the entire row must be constructed before it is sent into Streaming Lite.

   • Set Up a Flag

     The majority of SDK calls will return 0 upon successful completion. We will also set up a flag variable to detect any failures.

     C++

     Copy

     
     int errorflag = 0;
     

2. Create the Error Object

   This error object is passed into almost every subsequent SDK call as an argument. It works hand in hand with our flag variable: If we check that errorflag != 0 (meaning the SDK call has failed), our error object will contain the relevant error information.

   We have previously declared a global variable called g_error. Now we must instantiate the object. This is done using:

   C++

   Copy

   
   g_error = esp_error_create();
   

   We will need to remember to free the g_error object before the program exits.

3. Starting the SDK

   We must now start the SDK. This is done by calling esp_sdk_start(). We will need to remember to send a stop call to the SDK before the program exits.

   Additionally, we will pass the return value into errorflag, and check if it was successful. If it fails, g_error and __LINE__ are passed into print_error_and_exit() to output a description of the error. __LINE__ is a predefined macro that defines the current line number.

   You will see this error-detection structure throughout our Custom C/C++ Adapter code:

   C++

   Copy

   
   errorflag = esp_sdk_start(g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
   

4. Create the Credentials object

   We have previously declared a global variable called g_creds. Now we must instantiate the object using esp_credentials_create(). It returns NULL if unsuccessful, which we must check for.

   C++

   Copy

   
   g_creds = esp_credentials_create(ESP_CREDENTIALS_USER_PASSWORD, g_error);
   if (NULL == g_creds) print_error_and_exit(g_error, __LINE__);
   

   ESP_CREDENTIALS_USER_PASSWORD is an enum of type ESP_CREDENTIALS_T. Other values can be found in the header file esp_credentials.h. The credentials object will need to be freed after we are done with it.

   

   We will now set the user and password for our credentials object as well, even though both values are blank. Make sure to convert our val_username and val_password variables into cstrings using c_str():

   C++

   Copy

   
   esp_credentials_set_user(g_creds, val_username.c_str(), g_error);
   esp_credentials_set_password(g_creds, val_password.c_str(), g_error);
   

   
   

5. Create and Connect to Project Object

   Create the project object. Since Streaming Lite only runs one project at a time, we will use esp_project_get_standalone(). Its parameters include our host and port name, our credentials object, the error object, and an EspProjectOptions type. Set the EspProjectOptions parameter as NULL for default values.

   C++

   Copy

   
   EspProject * project = esp_project_get_standalone(val_host.c_str(), val_port, g_creds, NULL, g_error);
   if (NULL == project) print_error_and_exit(g_error, __LINE__);
   

   

   Connect to the project by passing our project object into esp_project_connect(). Check for failure.

   C++

   Copy

   
   errorflag = esp_project_connect(project, g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
   

6. Free Credentials Object

   After the project object has been created, we are done with the credentials object. Free it using esp_credentials_free(). Check for failure.

   C++

   Copy

   
   errorflag = esp_credentials_free(g_creds, g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
   g_creds = NULL;
   

7. Retrieve the Stream Object

   After we have connected to the project object, we can now retrieve the project stream we want to write to. This is done by passing our project object, and stream name into esp_project_get_stream(). Make sure to convert val_stream into a c-string using c_str().

   C++

   Copy

   
   const EspStream * stream = esp_project_get_stream(project,
   val_stream.c_str(), g_error);
   if (NULL == stream) print_error_and_exit(g_error, __LINE__);
   

   

8. Create and Connect to the Publisher Object

   We now create the publisher object using esp_project_create_publisher(). Its lifetime is managed by the SDK, and does not require to be freed. We pass in our project object, the error object, and an EspPublisherOptions type. Set the EspPublisherOptions parameter as NULL for default values.

   C++

   Copy

   
   EspPublisher * publisher = esp_project_create_publisher(project, NULL, g_error);
   if (NULL == publisher) print_error_and_exit(g_error, __LINE__);
   

   

   Connect to the publisher you just created, by passing the publisher object into esp_publisher_connect():

   C++

   Copy

   
   errorflag = esp_publisher_connect(publisher, g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
   

9. Retrieve the Message Writer Object

   The message writer object builds a single row to be sent into the specific project and stream it was created for.

   We now retrieve the EspMessageWriter object, using our publisher and stream objects. This is achieved by calling esp_publisher_get_writer():

   C++

   Copy

   
   EspMessageWriter * message = esp_publisher_get_writer(publisher, stream, g_error);
   if (NULL == message) print_error_and_exit(g_error, __LINE__);
   

10. Retrieve the Relative Row Writer Object

    The relative row writer formats a row of data for the stream the EspMessageWriter writes to.

    Retrieve the relative row writer by passing our message writer object into esp_message_writer_get_relative_rowwriter():

    C++

    Copy

    
    EspRelativeRowWriter * row = esp_message_writer_get_relative_rowwriter(message, g_error);
    if (NULL == row) print_error_and_exit(g_error, __LINE__);
    

11. Loop
    After setting up all required objects, we will now be able to construct and send our row. This will be done inside a loop, governed by our val_repeat and val_interval variables. This loop runs inside main() and is what controls the quantity and frequency of data the Custom C/C++ Adapter sends.

    • Set Up the Loop

      Since our loop has to run at least once, we will use a do/while loop. Copy this into your main():

      C++

      Copy

      
      do
      {
        if ( val_repeat != -1)
          val_repeat --;
      
          //***more code here***
      
          if ((val_repeat == -1 || val_repeat > 0) && val_interval > 0)
            sleep( val_interval);
          }
      }
      while ((val_repeat == -1 || val_repeat > 0) && val_interval > 0);
      

      There are two if statements inside the loop. The first if statement checks for our infinity condition. If the value of val_repeat is -1, no decrement occurs and the loop runs forever.

      The second if statement pauses execution using the sleep() function for a val_interval number of seconds. This controls the frequency of loop execution.

      The code to construct and send a row will be written in between these two if statements.

    • Starting a New Row Definition and Setting the Operation

      We will now start a new row definition by passing our EspRelativeRowWriter object into esp_relative_rowwriter_start_row(). This is the first call to do when starting a row. Insert this after the first if statement.

      C++

      Copy

      
      errorflag = esp_relative_rowwriter_start_row(row, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
      

      We now need to set the operation for the row. This is done using esp_relative_rowwriter_set_operation(). Once again, we pass in our EspRelativeRowWriter object.

      C++

      Copy

      
      errorflag = esp_relative_rowwriter_set_operation(row, ESP_STREAM_OP_INSERT, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
      

      ESP_STREAM_OP_INSERT is an enum of type ESP_STREAM_TYPE_T, located inside the header file esp_stream.h.

      

    • Constructing One Row

      The C/C++ SDK provides a series of functions which set one field of your row at a time. Every function will set a field with a specific data type, and these functions cannot search by field name. Therefore, calling one will set the next available matching field of that data type, without regard for field names. A full list of the functions for each data type is listed below. Parameters have been omitted; you can find all their declarations inside the header file esp_publisher.h. Each function generally has three parameters: our relative row writer object, the value we are writing, and our error object.

      

      • Column 1: SensorId

        Following the schema of the stream we are writing to, we will now set the first field, called SensorId. The field is a type string, and the value we wish to write is in our variable val_sensorId. To set the field, we use esp_relative_rowwriter_set_string(). Make sure to convert from string to cstring using c_str().

        C++

        Copy

        
        errorflag = esp_relative_rowwriter_set_string(row, val_sensorId.c_str(), g_error);
        if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
        

        

      • Column 2: Temperature

        The second field is called Temperature, and will contain our temperature sensor data. This is where the function exec() must be called, to run the command in our string variable val_Temperature_Command.
        We will start by initializing a new variable called string temperature_s, using exec()’s return value. Since our command is "echo 90", temperature_s will get a value of 90.

        C++

        Copy

        
        string temperature_s = exec(val_Temperature_Command.c_str());
        

        Next, we will set the Temperature field by calling esp_relative_rowwriter_set_float(). We use the functions c_str() and atof() to cast temperature_s into a double.

        C++

        Copy

        
        errorflag = esp_relative_rowwriter_set_float(row, atof(temperature_s.c_str()), g_error);
        if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
        

        

      • Column 3: Reading Date

        The next field is ReadingDate, a timestamp for the row. Earlier, we only declared the variable val_readingdate. We will now give it a value, using the time(0) function. This is the number of seconds since epoch.

        C++

        Copy

        
        val_readingDate = time(0);
        

        To set the column, we use esp_relative_rowwriter_set_msdate(). We pass in val_readingdate, multiplied by 1000. This is because esp_relative_rowwriter_set_msdate() reads values as milliseconds, not seconds.

        C++

        Copy

        
        errorflag = esp_relative_rowwriter_set_msdate(row, val_readingDate * 1000, g_error);
        if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
        

        

      • Column 4: Id

        The last field is Id. Recall that this is automatically generated in our Streaming Lite project. Therefore, we are arbitrarily sending a value of 0, from our val_id variable. To set the column, we use esp_relative_rowwriter_set_long():

        C++

        Copy

        
        errorflag = esp_relative_rowwriter_set_long(row, val_id, g_error);
        if (errorflag!=0) print_error_and_exit(g_error, __LINE__);
        

        

    • Ending the Row Definition

      After every field in our row has been populated, we can now end the row definition. This is done by passing our relative row writer object into esp_relative_rowwriter_end_row().

      ```C++

      errorflag = esp_relative_rowwriter_end_row(row, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, LINE);
      ```

    • Publishing and Committing Data

      We now publish the data we have packaged inside our message writer object. This is done by passing our message writer object and publisher object into esp_publisher_publish()

      C++

      Copy

      
      esp_publisher_publish() errorflag = esp_publisher_publish(publisher, message, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
      

      We now issue a commit call to the publisher, which causes it to process all its input queues and commit the data

      C++

      Copy

      
      errorflag = esp_publisher_commit(publisher, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
      

      Since no errors were detected, at this point we can print a line to notify the user that the row has published correctly:

      C++

      Copy

      
      printf("Messages published successfully\n");
      

12. Stopping the SDK

    We have finished coding the loop. Jumping out of the loop now, we have to do two things before our program ends. After the loop, we will need to send a stop call to the SDK, and release our error object.

    Similar to how we started it, we will now call esp_sdk_stop() to stop the SDK. Insert this line after our loop body:

    C++

    Copy

    
    errorflag = esp_sdk_stop(g_error);
    if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
    

13. Releasing the Error Object

    The last line in our main() function is a call to release our error object:

    C++

    Copy

    
    esp_error_free(g_error);
    

    Save and close your .cpp file. You should now have two files inside of your "custom_c_adapter" folder, named "custom_c_adapter.cpp" and "Makefile".

For the question below, select all of the correct answers, and click Validate.

custom_c_adapter.cpp

#include <string>
#include <iostream>
#include <stdlib.h>
#include <time.h>
#include <stdio.h>
#include "sdk/esp_sdk.h"
#include "sdk/esp_error.h"
#include "sdk/esp_credentials.h"
#include "sdk/esp_project.h"
#include "sdk/esp_publisher.h"

using namespace std;

EspError * g_error = NULL;
EspCredentials * g_creds = NULL;

void print_error_and_exit(EspError * streamingError, int line) {
      printf("An error was encountered on line %d\n", line);
      printf("%s", esp_error_get_message(streamingError));
      printf("\nExiting\n");

      //free EspCredentials object if it was created
      if (g_creds) esp_credentials_free(g_creds, g_error);

      //stop SDK
      esp_sdk_stop(g_error);

      //free EspError object if it was created
      if (g_error) esp_error_free(g_error);

      exit(1);
   }

// Execute a command and return result as string.
// In this example the command should only generate one number.
string exec(char const * cmd) {
   FILE * pipe = popen(cmd, "r");
   if (!pipe) return "ERROR";
   char buffer[128];
   string result = "";
   while (!feof(pipe)) {
      if (fgets(buffer, 128, pipe) != NULL)
         result += buffer;
   }
   pclose(pipe);
   return result;
}

int main() {
   string val_host = "SAPraspberrypi2-B3";
   int val_port = 9230;
   string val_stream = "isFreezerTemperatureReading";
   string val_username = "";
   string val_password = "";

   int val_repeat = -1;
   int val_interval = 1;

   string val_sensorId = "RaspberryPi";
   string val_Temperature_Command = "echo 90";
   int64_t val_readingDate; //will be generated right before row is
   sent
   int64_t val_id = 0; //generated in streaming lite

   int errorflag = 0;

   //creates error object. Must be freed after use
   g_error = esp_error_create();

   //esp_sdk_start initializes the sdk, must matched with same number of stop calls
   errorflag = esp_sdk_start(g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

   //sets up credentials with authentication using digest and signature verification
   //other types are available as well
   g_creds = esp_credentials_create(ESP_CREDENTIALS_USER_PASSWORD, g_error);
   if (NULL == g_creds) print_error_and_exit(g_error, __LINE__);

   esp_credentials_set_user(g_creds, val_username.c_str(), g_error);
   esp_credentials_set_password(g_creds, val_password.c_str(), g_error);

   //retrieves standalone project (Streaming Lite) object to connect to
   EspProject * project = esp_project_get_standalone(val_host.c_str(), val_port, g_creds, NULL, g_error);
   if (NULL == project) print_error_and_exit(g_error, __LINE__);

   //connect to project
   errorflag = esp_project_connect(project, g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

   // Once we have retrieved the project we do not need the EspCredentials object
   // Since we are not going to reuse them, need to free them
   errorflag = esp_credentials_free(g_creds, g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);
   g_creds = NULL;

   //Retrieves the stream object
   const EspStream * stream = esp_project_get_stream(project, val_stream.c_str(), g_error);
   if (NULL == stream) print_error_and_exit(g_error, __LINE__);

   //Creates a publisher for this project. Lifetime is managed by SDK
   EspPublisher * publisher = esp_project_create_publisher(project, NULL, g_error);
   if (NULL == publisher) print_error_and_exit(g_error, __LINE__);

   //Connects to publisher
   errorflag = esp_publisher_connect(publisher, g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

   //retrieve an EspMessageWriter to build a single row
   EspMessageWriter * message = esp_publisher_get_writer(publisher, stream, g_error);
   if (NULL == message) print_error_and_exit(g_error, __LINE__);

   //formats data for the stream the EspMessageWriter writes to
   EspRelativeRowWriter * row = esp_message_writer_get_relative_rowwriter(message, g_error);
   if (NULL == row) print_error_and_exit(g_error, __LINE__);

   do {

      if (val_repeat != -1) val_repeat--;

      //starts new row definition. First call when starting a row
      errorflag = esp_relative_rowwriter_start_row(row, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      //sets operation for row
      errorflag = esp_relative_rowwriter_set_operation(row, ESP_STREAM_OP_INSERT, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      // column 1 a string (SensorID)
      errorflag = esp_relative_rowwriter_set_string(row, val_sensorId.c_str(), g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      // column 2 a float (Temperature)
      string temperature_s = exec(val_Temperature_Command.c_str());
      errorflag = esp_relative_rowwriter_set_float(row, atof(temperature_s.c_str()), g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      //column 3 a msdate (ReadingDate)
      val_readingDate = time(0);
      errorflag = esp_relative_rowwriter_set_msdate(row, val_readingDate * 1000, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      //column 4 a long (Id)
      errorflag = esp_relative_rowwriter_set_long(row, val_id, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      //ends row definition
      errorflag = esp_relative_rowwriter_end_row(row, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      //Publishes data currently in EspMessageWriter. On successful completion data in EspMessageWriter is cleared
      errorflag = esp_publisher_publish(publisher, message, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      //causes publisher to process all its input queues and commit the data to its log stores
      errorflag = esp_publisher_commit(publisher, g_error);
      if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

      printf("Messages published successfully\n");

      // wait for interval..
      if ((val_repeat == -1 || val_repeat > 0) && val_interval > 0)
         sleep(val_interval);
   }
   while ((val_repeat == -1 || val_repeat > 0) && val_interval > 0);

   //stopping the sdk
   errorflag = esp_sdk_stop(g_error);
   if (errorflag != 0) print_error_and_exit(g_error, __LINE__);

   //release error object
   esp_error_free(g_error);
}