Introduce Polars for dumping and loading data

[hirosassa]

fixes #304

In this PR, I introduced polars into gokart.
Currently the implementation is super rough and I didn't check the correctness.
Feel free to leave your comment for this software architecture / implementation.

[hiro-o918]

hmm, it is a little confused to fallback to polars when import error occurred.

In my opinion, we need to introduce some global feature flags to switch dataframe frameworks.
For backward compatibility, we set pandas in default, then add a configure function like the follwoing.

DATAFRAME_FRAMEWORK = 'pandas'
def setup_dataframe_framework(framework: Literal['pandas', 'polars']):
   if framework == 'polars':
     try
         import  polars
     except ImportError:
          raise RuntimeError(...)
    DATAFRAME_FRAMEWORK = framework

According to the flag, we can switch the implementation.

In addition, we can probably set processor class, though I do not check this work...
This would make each CsvFileProcessor simple

if DATAFRAME_FRAMEWORK == 'pandas'
  CsvFileProcessor  =  PandasCsvFileProcessor 
elif DATAFRAME_FRAMEWORK == 'polars'
  CsvFileProcessor  =  PolarsCsvFileProcessor 

[hiro-o918]

import os
from typing import Protocol, Type


class IFeature(Protocol):
    def run(self) -> None: ...


class Feature1:
    def __init__(self): ...
    def run(self):
        print('feature1')


class Feature2:
    def __init__(self): ...
    def run(self):
        print('feature2')


Feature: Type[IFeature]
if os.environ.get('FEATURE') == '1':
    Feature = Feature1
elif os.environ.get('FEATURE') == '2':
    Feature = Feature2
else:
    raise ValueError("Invalid FEATURE environment variable value. Please set it to '1' or '2'.")


Feature().run()

❯ uv run foo.py
Traceback (most recent call last):
  File "gokart/foo.py", line 27, in <module>
    raise ValueError("Invalid FEATURE environment variable value. Please set it to '1' or '2'.")
ValueError: Invalid FEATURE environment variable value. Please set it to '1' or '2'.
❯ FEATURE=1 uv run foo.py
feature1
❯ FEATURE=2 uv run foo.py
feature2

Switching class by an environment variable works!

[hirosassa]

@hiro-o918 Thanks for the comment. I applied your suggestion. I think it looks fine!

[hirosassa]

comments added

[hirosassa]

skip if polars is not installed

[hirosassa]

skip if polars is installed

[hirosassa]

add the test run with polars extra

[hirosassa]

labels with extra branching
ref: tox-dev/tox#2406

[hirosassa]

if the polars label is specified, add polars to dependencies

[hiro-o918]

I prefer to raise an exception instead of ignoring the ValueError.
users never find ValueError('GOKART_DATAFRAME_FRAMEWORK_POLARS_ENABLED is not set. Use pandas as dataframe framework.') since it is ignored on L31.

DATAFRAME_FRAMEWORK = os.getenv('GOKART_DATAFRAME_FRAMEWORK', 'pandas')

if GOKART_DATAFRAME_FRAMEWORK == 'polars'
  try:
       import polars
  except ImportError:
       raise ValueError('please install polars to use polars as a framework of dataframe for gokart')

[hirosassa]

Thanks for your suggestion. fixed in 79839e1