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
{{ message }}
Repository navigation
[Question] Publish a PyPI release from master to remove the numpy<1.24 cap? #495
Would it be plausible or reasonable to publish a new release to PyPI from the current master? Or maybe just the fix that was made November 2025 as a patch release to PyPI?
Summary
I maintain xnatbidscli at the NIMH Data Science and Sharing Team. It's a CLI that downloads neuroimaging data from XNAT and converts it to BIDS. Its physioconvert subcommand uses phys2bids as a library (phys2bids.io.load_acq/load_txt/load_mat/load_gep/load_smr and phys2bids.phys2bids.phys2bids()) to write BIDS physio files next to the converted MRI data.
Additional detail
The latest phys2bids PyPI release, 2.10.0 (May 4, 2023), declares numpy >=1.10, <1.24. That cap was added as a temporary fix (see #434). It makes phys2bids 2.10.0 hard to install alongside today's scientific Python packages:
numpy 1.23.x has no wheels for Python 3.12 or newer. Installing phys2bids 2.10.0 on 3.12+ means building numpy 1.23 from source, which generally fails.
nibabel 5.4+ requires numpy>=1.25, and pandas 2.2+ requires numpy>=1.26 on Python 3.12+. Any environment that needs one of these can't install phys2bids 2.10.0 with pip.
Because of this, xnatbidscli can only offer physioconvert on Python 3.11 for pip users. On 3.11, I hold numpy at 1.23.5 and nibabel below 5.4; on 3.12+ I leave phys2bids out entirely.
Your master's setup.cfg now declares numpy >=1.10 with no upper cap. setup.cfg was last changed sometime after 2.10.0 (Nov 14, 2025, "Update support to newer python versions" maybe?). master is 94 commits ahead of the 2.10.0 tag. That includes the Python 3.13 fix in #491 (May 2026), so your code seems ready; only the PyPI release is missing.
For what it's worth, before switching to the Python 3.11-only setup, I ran phys2bids 2.10.0 on numpy 1.26.4, using a dependency override in our uv config, and saw no problems in the loaders or the conversion workflow I used.
Next Steps
Could I help in any way, not knowing your repo or practices well? And thanks for maintaining phys2bids!
Hi @ericearl , let me check a couple of things for this, but I think it's time to make a new release. Is it urgent or can it wait for a couple of days?
Activity
smoia commentedon Sep 28, 2026
Hi @ericearl , let me check a couple of things for this, but I think it's time to make a new release. Is it urgent or can it wait for a couple of days?
ericearl commentedon Sep 28, 2026
It's totally not urgent, I made a release today, but I can patch in the update with your new version whenever that's ready. Thanks @smoia !