Open valeriupredoi opened 9 months ago
it appears the numcodecs pin has not been fixed yet in kerchunk=0.2.4, going to go chase the feedstock
Attention: Patch coverage is 84.61538%
with 14 lines
in your changes missing coverage. Please review.
Project coverage is 88.52%. Comparing base (
00d4b4a
) to head (730f6ec
). Report is 38 commits behind head on main.:exclamation: Current head 730f6ec differs from pull request most recent head d82b916
Please upload reports for the commit d82b916 to get more accurate results.
Files | Patch % | Lines |
---|---|---|
activestorage/netcdf_to_zarr.py | 80.82% | 14 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
OK new Kerchunk=0.2.4 works very well, and I also fixed the dreaded SegFault that was plaguing us until today :partying_face:
Description
This is a sandbox testing a slightly modded Kercunk that kerchunks inly a desired
Dataset
- the good folk at Kerchunk are already investigating means of implementing such an option in Kerchunk in https://github.com/fsspec/kerchunk/pull/424 - and some timing test results can be seen in https://github.com/fsspec/kerchunk/pull/424#issuecomment-1961482253 (after one setsdefault_cache_type
tofirst
as Martin recommended)Label: needs new Kerchunk
Update as of February 29
With the Kerchunk PR now merged https://github.com/fsspec/kerchunk/pull/424 and the issues related to the newer Kerchunk functionality now ironed out, this is ready for a merge when kerchunk release the new version that contains that PR. Test results are very promising:
So all the time spent in PyActiveStorage is sub-1s! This is for the
tests/test_compression_remote_reductionist.py::test_compression_and_filters_cmip6_forced_s3_from_local_bigger_file_v1
test.