search

pangeo-data / rechunker

Disk-to-disk chunk transformation for chunked arrays.
https://rechunker.readthedocs.io/
MIT License
163 stars 25 forks source link

Add CI #5

Closed TomAugspurger closed 4 years ago

TomAugspurger commented 4 years ago

We may need to merge this before we can verify that it's actually working.

TomAugspurger commented 4 years ago

Oh it's running (and failing) on my branch. Will fix that up first and merge. https://github.com/TomAugspurger/rechunker/runs/696938602?check_suite_focus=true

TomAugspurger commented 4 years ago

I think either is fine, I'm just trying to debug an issue on the CI. For some reason Python is complaining about non-JSON serializable objects, I think a scalar numpy.int64 value. Will figure it out.

On Thu, May 21, 2020 at 6:47 PM Ryan Abernathey notifications@github.com wrote:

@rabernat commented on this pull request.

In rechunker/algorithm.py https://github.com/pangeo-data/rechunker/pull/5#discussion_r428972533:

@@ -3,7 +3,10 @@

try: from math import prod +

  • numpy_prod = False

Apparently prod is not in math until python 3.8. Maybe we should just always use numpy here? Math has type annotations.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/pangeo-data/rechunker/pull/5#pullrequestreview-416563795, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAKAOIXBJ52ZHW66JM5Y6ADRSW4RBANCNFSM4NG7C3QA .

TomAugspurger commented 4 years ago

All passing on my branch.

I'm not really sure why the conversion from NumPy to python ints was needed, but somewhere Zarr was trying to JSON serialize them (I think the metadata like shape / chunks).