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
Regression in v2.18.0 when setting an item in an object array #1874
I have tried with a simpler environment in Python 3.12 and 3.11 on Windows. I set up a virtual environment (via python -m venv) and installed zarr and msgpack; pip freeze reports:
Ok was eventually able to reproduce. The issue goes away when reverting the change in PR: #1800
@dcherian would you be able to take a closer look at the error here?
Edit: Possibly the easiest change is just to revert that code in v2. In v3 breaking changes would be expected (though maybe we could warn about this in v2?). Though not sure if there is a reason we want that optimization in v2 also (if so maybe it needs a closer look)
Activity
dcherian commentedon May 13, 2024
FWIW I couldn't reproduce this on python 3.11
aplowman commentedon May 13, 2024
I have tried with a simpler environment in Python 3.12 and 3.11 on Windows. I set up a virtual environment (via
python -m venv) and installed zarr and msgpack;pip freezereports:for both Python 3.11.3 and Python 3.12.3. This works fine, but if I upgrade Zarr to 2.18.0, I see the exception in both Python versions.
Edited to add traceback from Python 3.12.3:
jhamman commentedon May 13, 2024
@aplowman - would you mind adding a complete traceback to your report above?
jakirkham commentedon May 13, 2024
Does this happen for you on Linux?
FWIW tried reproducing on macOS with Conda and was unable to
aplowman commentedon May 13, 2024
I can reproduce the issue on Google Colab, which is Ubuntu: https://colab.research.google.com/drive/1IO3WFOJM5rGJOvve_j-UZahk_t3hke0J?usp=sharing
jakirkham commentedon May 13, 2024
Thanks for the example! 🙏
Looking through the code paths in the traceback in
git blame, the changes appear to be all 2yrs old or older. Maybe I'm missing something?Am wondering what would have changed that would lead to this issue
The only relevant thing I can think of is this PR ( zarr-developers/numcodecs#417 ), but that landed over 1yr ago
Are others aware of changes that occurred here that might be relevant?
jakirkham commentedon May 13, 2024
Ok was eventually able to reproduce. The issue goes away when reverting the change in PR: #1800
@dcherian would you be able to take a closer look at the error here?
Edit: Possibly the easiest change is just to revert that code in v2. In v3 breaking changes would be expected (though maybe we could warn about this in v2?). Though not sure if there is a reason we want that optimization in v2 also (if so maybe it needs a closer look)
dcherian commentedon May 13, 2024
What did it take to reproduce? Does #1875 fix it? If not, I agree that it's best to revert.
rabernat commentedon May 14, 2024
Since this is about a singleton array (
shape=0) I would be very optimistic that #1875 fixes this. Is it really so hard to reproduce?Fix zarr-developers#1874
dcherian commentedon May 14, 2024
Sorry I misread the reproducer. Now fixed on #1875
aplowman commentedon May 14, 2024
I'm not sure #1875 fully resolves this issue. I am able to set the item now, but I cannot then retrieve it: https://colab.research.google.com/drive/1hTmv23yTQbTgv2tY0uOBW-Kd1YyasBlu?usp=sharing . The item retrieval stack trace is below.
jakirkham commentedon May 16, 2024
@aplowman if you have a moment, could you please test PR ( #1875 )?
aplowman commentedon May 16, 2024
Works for me! Thanks all.