Filtering out bots from GitHub data

bots.py

# List of common bots on GitHub
# Doesn't include ones that would already be filtered out by the is_bot function
# but it won't hurt to also include them in here

GITHUB_BOTS = [
    'netlify',
    'linear-app',
    'codeclimate',
    'renovate',
    'renovate-approve',
    'github-actions',
    'vercel',
    'googlebot',
    'codesandbox-ci',
    'sizebot',
    'tensorflow-jenkins',
    'tensorflowbutler',
    'google-ml-butler',
    'google-cla',
    'coveralls',
]

is_bot.py

import pandas as pd
from typing import Optional

from bots import GITHUB_BOTS


def is_bot(
    row: pd.Series,
    author_col: str,
    bot_col: Optional[str] = None,
    bot_col_val: Optional[str] = None,
) -> bool:
    # Check if the bot column actually says it's a bot
    # e.g. pr_author_typename == 'Bot'
    if (
        (bot_col is not None)
        and (row[bot_col] is not None)
        and (row[bot_col] == bot_col_val)
    ):
        return True

    # If the bot column doesn't hold any evidence, rely on the author name
    author = row[author_col]
    return (
        '-bot' in author
        or '[bot]' in author
        or 'dependabot' in author
        or author in GITHUB_BOTS
    )

test_is_bot.py

import pandas as pd
from is_bot import is_bot


def test_is_bot_checks_author_col_for_mention_of_bot():
    assert is_bot(pd.Series({'pr_author': 'bertie-bot'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'bertie-bott'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'bertie [bot]'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'bertie[bot]'}), author_col='pr_author')
    assert is_bot(
        pd.Series({'pr_author': 'facebook-github-bot'}), author_col='pr_author'
    )


def test_is_bot_catches_all_types_of_dependabot():
    assert is_bot(pd.Series({'pr_author': 'dependabot'}), author_col='pr_author')
    assert is_bot(
        pd.Series({'pr_author': 'dependabot-preview'}), author_col='pr_author'
    )
    assert is_bot(
        pd.Series({'pr_author': 'dependabot-preview'}), author_col='pr_author'
    )


def test_is_bot_catches_specific_bot_names():
    assert is_bot(pd.Series({'pr_author': 'netlify'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'linear-app'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'codeclimate'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'renovate'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'renovate-approve'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'github-actions'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'vercel'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'googlebot'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'google-cla'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'codesandbox-ci'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'sizebot'}), author_col='pr_author')
    assert is_bot(
        pd.Series({'pr_author': 'tensorflow-jenkins'}), author_col='pr_author'
    )
    assert is_bot(pd.Series({'pr_author': 'tensorflowbutler'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'google-cla'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'google-ml-buter'}), author_col='pr_author')
    assert is_bot(pd.Series({'pr_author': 'coveralls'}), author_col='pr_author')


def test_is_bot_does_not_filter_out_incidental_botlike_substrings():
    assert not is_bot(pd.Series({'pr_author': 'hannah abbott'}), author_col='pr_author')
    assert not is_bot(
        pd.Series({'pr_author': 'i-like-to-renovate'}), author_col='pr_author'
    )


def test_is_bot_prioritises_bot_col():
    assert is_bot(
        pd.Series(
            {'pr_author': 'human being sounding name', 'pr_author_typename': 'Bot'}
        ),
        author_col='pr_author',
        bot_col='pr_author_typename',
        bot_col_val='Bot',
    )

    # bot_col_val overrides the fact that the author name is not bot-like
    assert is_bot(
        pd.Series(
            {'pr_author': 'human being sounding name', 'pr_author_typename': 'Bot'}
        ),
        author_col='pr_author',
        bot_col='pr_author_typename',
        bot_col_val='Bot',
    )

    # bot_col_val is not correct
    assert not is_bot(
        pd.Series(
            {'pr_author': 'human being sounding name', 'pr_author_typename': 'User'}
        ),
        author_col='pr_author',
        bot_col='pr_author_typename',
        bot_col_val='Bot',
    )