Overview

This is a django based application to create and run the data-pipelines. The application runs Prefect tasks. Eventually, the requests made via API are converted to corresponding Prefect tasks. The tasks' status, history etc. can be monitored in the Prefect cloud by running - prefect orion start and opening the url shown in the prompt.

Requirements

• Once the code is cloned from the git, install the requirements from requirements.txt file by running pip install -r requirements.txt
• The project uses rabbitmq, which can be installed from the official website

Background tasks

This application needs several background tasks to be running. Following are the must-to run processes before running the application. Run all these processes parallely in different terminals.

1. python manage.py runserver - This starts the django server, and it listens to the API requests. This can be considered an entry-point to our program.
2. python manage.py process_tasks --queue create_pipeline - Runs the create-pipeline background task. Request from the runserver is received by this task.
3. python manage.py runscript worker_demon.py - Runs the rabbitmq - worker demon.

Demo Request

A demo request to the shepherd API consists the following in the request body.

  "pipeline_name": "Skip_merge_anonymize on Res.271",
  "res_id" : 271,
  "db_action":"create",
  "transformers_list" : [{"name" : "skip_column", "order_no" : 1, "context": {"columns":["format"]}},
                        {"name" : "merge_columns", "order_no" : 2, "context": {"column1":"title", "column2":"price", "output_column":"title with price", 
                        "separator":"|"}},
                        {"name" : "anonymize", "order_no" : 3, "context": {"to_replace" : "Sir", 
  "replace_val": "Prof", "column": "author"}}]
}

1. pipeline_name - Name of the pipeline that needs to be created.
2. res_id - Resource ID i.e. ID of the resource that needs transformation. This can be considered input data for our pipeline.
3. db_action - Takes either create or update. This tells us whether to create a new resource in our db out of transformed data or to update the existing resource with the transformed data.
4. transformers_list - List of json objects.
   1. name - Name of the task that needs to be performed.
   2. order_no - The order number of the task. In the above example - skip_column is followed by merge_columns which is followed by anonymize tasks as the order numbers of the corresponding tasks are 1, 2 and 3 respectively.
   3. context - Necessary inputs to perform the task. This is task specific.

The final HTTP request looks like following.

POST http://127.0.0.1:8000/transformer/res_transform
Content-Type: application/json
{
    "pipeline_name": "Skip_merge_anonymize on Res.271",
    "res_id" : 271,
    "db_action":"create",
    "transformers_list" : [{"name" : "skip_column", "order_no" : 1, "context": {"columns":["format"]}},
                        {"name" : "merge_columns", "order_no" : 2, "context": {"column1":"title", "column2":"price", "output_column":"title with price", 
                        "separator":"|"}},
                        {"name" : "anonymize", "order_no" : 3, "context": {"to_replace" : "Sir", 
  "replace_val": "Prof", "column": "author"}}]
}

Adding new tasks to the pipeline

Following are the steps to be followed to add a new task to the pipeline.

1. Define your task name and the context (i.e. necessary information to perform the task).
2. Write the task(i.e. your Python function) in prefect_tasks file as a prefect task. Note: Prefect task is a Python function annotated with @task. Make sure you have the same arguments passed to your function as other tasks defined in the file.

Let's understand this through an example. Suppose you need to add a task named - add_prefix which adds a given prefix to all the values in the specified column So, the task name would be - add_prefix. To define the context, let's define the necessary inputs first.

• We will be needing the column name. So, the context should contain a key named 'column'
• We will also be needing a string which acts as prefix. So the second key in the context should be - 'prefix'

Finally, our context should look something like this.

"context" : {"column": "<column_name_here>", "prefix":"<prefix_string_here>"}

Now we should add the task in prefect_tasks file. Go to the file, and add the task.

@task
def add_prefix(context, pipeline, task_obj):
    column = context['column']
    prefix_string = context['prefix']
    # Rest of your logic here..

A request to create a pipeline with this task should look something like,

POST http://127.0.0.1:8000/transformer/res_transform
Content-Type: application/json
{
    "pipeline_name": "Test_prefixing",
    "res_id" : 271,
    "db_action":"create",
    "transformers_list" : [
        {"name" : "add_prefix", 
        "order_no" : 1, 
        "context": {"column": "Planets", "prefix": "The"}}
        ]
}

Flow of the code

As there are many background tasks involved, it might be a bit confusing at first. Here is how control flows once request is made.

API end-point...>Pipeline creation...>Rabbitmq worker...>Model to pipeline...>Prefect tasks...>Model to pipeline...>Utils