TDAtaframe

TDA brought to dataframes.

Documentation

Repo-level docs live in docs/README.md.

• Architecture
• Data Model
• Development Guide

Install

TDAtaframe uses the libtorch libraries provided by the Python torch package. You do not need to install libtorch system-wide.

Install with

pip install tdataframe

The Rust backend currently targets torch==2.7.0, matching tch-rs 0.20.0. The package supports Python 3.12 and 3.13. If pip builds from source, it will compile against the torch package in the build environment.

On first install, it is normal for compilation to take a few minutes.

Verify Packaging

Run the same install path that CI uses:

python3 scripts/wheel_install_test.py

That script builds a wheel through pip's isolated PEP 517 path, installs the wheel into a fresh virtualenv, runs pip check, imports the Rust extension against PyTorch's bundled libtorch, and then runs the test suite from the installed wheel.

For cross-machine checks, GitHub Actions runs the wheel test on Linux x64, macOS x64, macOS arm64, and Windows x64. If you have the GitHub CLI installed and authenticated, this command triggers CI and saves the logs locally:

python3 scripts/ci_watch.py