You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Cython does support free-threading and you should be able to update your current code to work with the free-threaded build. Lots of other projects (numpy, pandas, scipy to name a few) are shipping free-threaded wheels on pypi using Cython's support.
I'm one of the authors of the free-threaded Python guide which has a lot more advice for porting libraries to support the free-threaded build. I'm happy to answer questions.
Ah, I thought I saw that they were using a prerelease Cython wheel from an alternate channel. If the standard conda-forge/pypi release supports it, then I'll give it a shot.
FYI, It doesn't look like there's CI for free threading though... at least, don't see it using the pyarrow nightly, unless I'm reading this wrong. Maybe in another issue perhaps, or change this issue title to be clearer (since it seems like the feature is done from the title).
Pyarrow just added nightly's for free threading, DuckDB hasn't shipped any (it's in a branch I'm working on), etc.
For my work, I've had to setup different workflows to pull the nightly's and test against the free threaded 3.14t.
Yeah, PyArrow is in the middle of the release process so then I can update. I figured it would be easier just to wait a couple weeks. (3.13t is being tested in CI too, at least with whatever version of PyArrow was uploaded - we don't exercise all that much of PyArrow so it seems to be OK.)
Activity
lidavidm commentedon Oct 8, 2025
Alternatively, we could explore nanobind etc...
ngoldbaum commentedon Oct 13, 2025
Cython does support free-threading and you should be able to update your current code to work with the free-threaded build. Lots of other projects (numpy, pandas, scipy to name a few) are shipping free-threaded wheels on pypi using Cython's support.
I'm one of the authors of the free-threaded Python guide which has a lot more advice for porting libraries to support the free-threaded build. I'm happy to answer questions.
lidavidm commentedon Oct 13, 2025
Ah, I thought I saw that they were using a prerelease Cython wheel from an alternate channel. If the standard conda-forge/pypi release supports it, then I'll give it a shot.
feat(python): support free-threading
feat(python): support free-threading
paultiq commentedon Oct 14, 2025
FYI, It doesn't look like there's CI for free threading though... at least, don't see it using the pyarrow nightly, unless I'm reading this wrong. Maybe in another issue perhaps, or change this issue title to be clearer (since it seems like the feature is done from the title).
Pyarrow just added nightly's for free threading, DuckDB hasn't shipped any (it's in a branch I'm working on), etc.
For my work, I've had to setup different workflows to pull the nightly's and test against the free threaded 3.14t.
Edit: Nevermind, I see you added #3574
feat(python): support free-threading (apache#3575)
lidavidm commentedon Oct 14, 2025
Yeah, PyArrow is in the middle of the release process so then I can update. I figured it would be easier just to wait a couple weeks. (3.13t is being tested in CI too, at least with whatever version of PyArrow was uploaded - we don't exercise all that much of PyArrow so it seems to be OK.)
paultiq commentedon Oct 14, 2025
Thx, I'll test against main then
I appreciate your work!
lidavidm commentedon Oct 14, 2025
We do publish nightlies as well: https://manage.fury.io/arrow-adbc-nightlies/python:adbc-driver-manager/
(As of this writing they haven't updated for yesterday's changes yet, but should soon.)