diffhouse 1.1.0

Repository mining tool for structuring Git metadata at scale.

diffhouse: Repository Mining at Scale

Documentation

diffhouse is a Python solution for structuring Git metadata, designed to enable large-scale codebase analysis at practical speeds.

Key features are:

• Fast access to commit data, file changes and more
• Easy integration with pandas and polars
• Simple-to-use Python interface

Requirements

Python	3.10 or higher
Git	2.22 or higher

Git also needs to be added to the system PATH.

Limitations

At its core, diffhouse is a data extraction tool and therefore does not calculate software metrics like code churn or cyclomatic complexity; if this is needed, take a look at PyDriller instead.

Also note that revision data is limited to default branches only.

User Guide

This guide aims to cover the basic use cases of diffhouse. For the list of available repository objects and fields, check out the API Reference.

Installation

Install diffhouse through PyPi:

pip install diffhouse

Quickstart

from diffhouse import Repo

url = 'https://github.com/user/repo'

r = Repo(location = url, blobs = True).load()

for c in r.commits:
    print(c.commit_hash[:10], c.committer_date, c.author_email)

print(r.branches)
print(r.diffs[0].to_dict())

First, construct a Repo object and define its target repository via the location argument; this can be either a remote URL or a local path. Pass blobs = True to extract file data as well.

Calling Repo.load() will load all metadata into memory, which can then be accessed through the object's properties. See all properties

blobs = True requires a complete clone of the repository and therefore takes longer to execute. Omit this argument whenever possible.

Lazy Loading

For large repositories, calling load() can be slow and/or take up gigabytes of memory. It is recommended to use the lazy method via with instead:

with Repo(location = url, blobs = True) as r:
    c = list(r.stream_commits())

    for d in r.stream_diffs():
        if d.lines_added == 3:
            break

This brings two big benefits:

1. Object streaming functions are lazy generators, allowing for efficient memory use.
2. No processing power is spent on objects that are not explicitly requested.

See all streaming functions

Tabular Data

Commit, ChangedFile and Diff iterables can be passed directly to pandas and polars DataFrame constructors. No pre-processing is needed; table schemas will be inferred correctly.

import polars as pl

df = pl.DataFrame(r.changed_files)
print(df.schema)

diffhouse stores datetime values as ISO 8601 strings to preserve time zone offsets. When converting these to datetime objects in a DataFrame, use the parser's UTC option.