meta: uv plugin viability, and whether reqstool-python-poetry-plugin is still needed

[jimisola]

Summary

Should reqstool add a reqstool-python-uv-plugin, and is reqstool-python-poetry-plugin
still worth maintaining? Opened to record findings, not to propose a concrete build — the
evidence below points at "not yet" for the first and "needs real usage data" for the
second.

A uv plugin isn't buildable the way hatch's and poetry's are — yet

reqstool-python-hatch-plugin and reqstool-python-poetry-plugin both work the same way:
a custom build-time hook, registered through the backend's plugin system, that scans source
for decorators during hatch build / poetry build and writes annotations.yml.

uv is two different things, and only one of them is relevant here:

• uv as package/dependency manager (uv add, uv sync, uv build) — this is
  orthogonal to which build backend a project declares. A project can use uv for
  everything and still declare build-backend = "hatchling.build" in pyproject.toml.
  In that case uv build just invokes hatchling under PEP 517, and
  reqstool-python-hatch-plugin already works, unmodified, for uv users today. No new
  plugin needed for this case.
• uv_build, uv's own build backend — this is what a reqstool-python-uv-plugin would
  actually need to hook into, and it currently has no build-hook or plugin mechanism at all.
  astral-sh/uv#14561 is the open upstream
  request for one. uv-dynamic-versioning, the closest existing analog, states outright
  that it "doesn't work with the uv build backend right now" for the same reason.

So: nothing to build until uv#14561 (or equivalent) lands upstream. Worth revisiting when it
does, but starting now would mean building against a backend with no extension point.

Is poetry still needed?

Signal, not a conclusion — I don't have download or dependent-repo data to settle this:

• reqstool-python-poetry-plugin's own README carries > [!WARNING] Poetry plugin support is currently untested. Functionality may be broken. Contributions welcome. — written by
  the project itself, not inferred.
• Its commit history has no feature or fix commits since early on; the only recent activity
  is the same automated CI/dependency-bump/migration commits every repo got.
• Its issue history is entirely closed maintenance issues (Renovate config, docs, dynamic
  versioning investigation) — nothing recent asking for it to work.

reqstool-python-hatch-plugin by contrast has no such warning and is the same age.

I couldn't get PyPI download stats for either package to compare actual usage — pypistats.org
wasn't reachable from here. That's the number I'd want before deciding to deprecate: whether
anyone downloads the poetry plugin at all, versus whether it's simply never been fixed
because no one asked.

Suggested next steps

• No uv plugin work until there's an upstream extension point to build against. Track
  Build/Metadata Hook astral-sh/uv#14561.
• Before any decision on poetry: pull download counts for
  reqstool-python-poetry-plugin vs reqstool-python-hatch-plugin from pypistats.org (or
  PyPI's BigQuery dataset) to see whether "untested, may be broken" has actually cost anyone
  anything.

[jimisola]

Download data landed (pypistats.org rate-limits hard — most calls 429'd, these got through
before it did):

Package	last_day	last_week	last_month
reqstool-python-poetry-plugin	0	4	15
reqstool-python-decorators	8	1,105	1,916
reqstool-python-hatch-plugin	—	—	~581 (shields.io, sources the same data, one read got through before it hit the same limit)

Poetry plugin downloads are about 1% of decorators' and 2.5% of hatch-plugin's. That
doesn't prove zero real users — 15/month could be one CI pipeline reinstalling on every run
— but it's consistent with the README's own warning rather than contradicting it. Combined
with no feature or fix commits since inception, I'd read this as: low but not obviously
zero usage, and nobody has complained loudly enough to get it fixed.

Doesn't change the recommendation — still "decide with data, not vibes" — except now there
is some data. Worth a repeat pull after a longer window (30/90 day) before actually pulling
the plug, in case this month is unusually quiet.