The scikit-tda synthetic-dataset generator, and the only recipe here with a pull request open against a public channel.

tadasets

tadasets generates the synthetic point clouds the field tests against — tori, swiss rolls, dsphere, the infinity sign — the shapes whose homology you already know, which is what makes them useful for checking that an implementation returns the right barcode.

It is the smallest package in this set and it was blocking two closures at once, which is why it is the one that got submitted. A missing tadasets kept the scikit-tda-recipe closure from resolving from channels, and it was a declared runtime dependency of petls-pytorch until petls-pytorch-recipe‘s fork demoted it to a benchmark extra. That demotion is now merged upstream, so this recipe backs only the scikit-tda gap.

The recipe is verified green on osx-arm64 — the sdist checksum matches, setuptools builds, the import and pip check pass. It is a pure-Python universal wheel, so noarch: python makes the artefact byte-identical on linux-64.

A pull request is open at conda-forge/staged-recipes #34367. When it lands, this note stops being a build and becomes history.