Python Metaprogramming

This repo contains two examples of metaclass usecases in Python: abstract base classes and dataclasses. I tried to replicate as closely as possible the features of the ABC and dataclasses standard Python modules while still remaining as simple as possible for educational purposes.

Installation

The base package has no dependencies and can be installed by running

pip install .

in the directory containing setup.py. It can then be imported with the metademo module.

Automated testing

If you wish to run the automated tests, then the pytest module is required. The tests can be executed by running

pytest

in the root or /tests directory.

Project structure

├── metademo
│   ├── __init__.py
│   ├── myabc.py          # Abstract base class implementation
│   └── mydataclass.py    # Dataclass implementation
├── setup.py
└── tests                 # Automated tests
    ├── test_myabc.py
    └── test_mydataclass.py

Usage example

Abstract Base Class

from metademo.myabc import MyABC, myabstractmethod
from typing import List, Dict

class DataReader(MyABC):
  @myabstractmethod
  def read_input(self, data_path: str) -> List[Dict]:
    pass
    
class CSVReader(DataReader):
  def read_input(self, data_path: str) -> List[Dict]:
    print(f'Reading csv file {data_path}...')

class DatabaseReader(DataReader):
  def read_input(self, data_path: str) -> List[Dict]:
    print(f'Reading data from table {data_path}...')

csv_reader = CSVReader()
db_reader = DatabaseReader()
abstract_reader = DataReader()    # Throws exception because there are unimplemented abstract methods

Dataclasses

from metademo.mydataclass import MyDataClass

class Person(MyDataClass):
    name: str
    age: int

    def greet(self):
        return f'Hello, {self.name}!'

victor = Person('Victor', 21)   # Automatically generated init method
victor.greet()                  # Hello, Victor!

TODO

General

• Add CI automated testing and linting with Github Actions
• Rename metademo/ directory to src/

Abstract Base Class

• Throw error if abstract and concrete type annotation signature do not match

Dataclass

• Add repr method
• Add type hints to __init__ params (check branch feat/type-annotations for current attempts at this featrue)
• Add default values to __init__ parameters