Fun with FastAPI

• FastAPI is the "new kid on the block" with regards to Python frameworks
• It is based on Starlette and brings several more cool features
• This demo is a rapid prototype created using Yahoo Finance APIs and fullstack FastAPI with Jinja2 templating engine
• The application can take in multiple stock symbols through the UI and fetch the financial data asynchronously using FastAPI Background Tasks
• The listing page offers basic query params based filtering
• The screenshots can be found over here

Table of contents:

• Application Screenshots
• Setup and Install
• Hello World demo
• Yahoo Finance
• Stocks Dashboard
  • Database Design
  • Creating Stocks
    • Pydantic
    • Dependency Injection
    • Background Tasks
  • Information Display
  • API documentation

Application Screenshots

• Adding multiple stocks
• Async data update using Background Tasks
• Data persistance and Template Rendering
• Filtering based on query params

Setup and Install

• Install Pyenv and set the local python version to 3.8.3
• Install Poetry
• Clone the repo
• Run poetry install
• Get inside the virtual environment using poetry shell

Hello World Demo

• Run the server using the command uvicorn hello_world:app --reload or my fancy make command make helloworld
• Go to http://localhost:8000/
• Go to http://localhost:8000/items/12
• Go to http://localhost:8000/items/12?query=apples%20are%20yummy

Yahoo Finance

• The financial data is extracted using the Yahoo Finance library
• A demo of the library is present in the file yfinance_demo.py where a sample response is documented as well

Stocks Dashboard

• Head to http://localhost:8000/docs for API documentation
• Head to http://localhost:8000 for list of all the stocks

Database Design

• Database is created using SQLite and SQLAlchemy as the ORM
• The database structure is as follows:

(supreme-enignma-rOVS3nom-py3.8) ➜  supreme-enigma git:(feature/database-design) ✗ sqlite3 stocks.db
SQLite version 3.24.0 2018-06-04 14:10:15
Enter ".help" for usage hints.
sqlite> .schema
CREATE TABLE stocks (
        id INTEGER NOT NULL, 
        symbol VARCHAR, 
        price NUMERIC(10, 2), 
        forward_pe NUMERIC(10, 2), 
        forward_eps NUMERIC(10, 2), 
        dividend_yield NUMERIC(10, 2), 
        ma50 NUMERIC(10, 2), 
        ma200 NUMERIC(10, 2), 
        PRIMARY KEY (id)
);
CREATE UNIQUE INDEX ix_stocks_symbol ON stocks (symbol);
CREATE INDEX ix_stocks_id ON stocks (id);
sqlite> 

Creating Stocks

• Pydantic to create struct or POJO equivalent for the incoming POST requests
• Dependency Injection used to make sure we have database connection while we are making a request
• Background Tasks for fetching the data from yfinance library

Pydantic

• Helps you structure your requests
• Handles controller level datatype validation
• Returns 422 response status code and descriptive payload in case of errors eg.

{
  "detail": [
    {
      "loc": [
        "body",
        "symbol"
      ],
      "msg": "field required",
      "type": "value_error.missing"
    }
  ]
}

Dependency Injection

• Executing something depends on executing something else
• Creating a stock depends on getting a database connection from the application
• This way we can use the db inside the controller layer
• Depends always have to come in the end of the function signature

Background Tasks

• Background tasks require the controller to be in the async mode
• The API is similar to Celery
• Inside the task, we are creating a new database session object and making persistent changes based on that

Information Display

• The information is displayed on the homepage of the application
• The data is rendered through a Jinja2 template which uses Semantic UI for styling and jQuery for bootstrapping javascript related work
• New stocks can be added by making Ajax based POST request
• The filters are used to generate query param based filtering for the application

API documentation

• The API documentation can be found at http://localhost:8000/docs
• The Insomnia collection is present over here