Migrate to zarr-python 3 #49
Conversation
|
@d-v-b I think this is all good for review now. The only thing broken is the docstests, because memory stores have a different name each time they're printed (because the memory address changes). I'm not too sure the best way to fix this... |
I think the best solution would be for zarr-python to stop using the memory address in the repr for memory stores :) |
|
is it OK if I make commits against |
|
Yeah go for it! |
|
This is currently blocked by zarr-developers/zarr-python#3038 |
dstansby
force-pushed
the
zarr-python-3
branch
3 times, most recently
from
a1b1960 to
41b579a
Compare
|
In good news using zarr main after zarr-developers/zarr-python#3038 was merged works 🎉 . Guess we just need to wait for a new zarr release now, and this should be good to go, but this is now ready for review. |
docs/index.md
Outdated
| @@ -35,15 +37,9 @@ print(spec.model_dump()) | |||
| 'dtype': '|u1', | |||
| 'fill_value': 0, | |||
| 'order': 'C', | |||
| 'filters': None, | |||
| 'filters': [], | |||
There was a problem hiding this comment.
For whoever reviews this: I updated filters to be an empty list instead of None if there are no filters. This matches zarr-python 3 behaviour. I would be easy enough to revert to previous behaviour though, and set this back to None. Would appreciate thoughts on this.
There was a problem hiding this comment.
if the goal is to accurately model what's in zarr-python 2 metadata documents, then filters: [] is invalid
There was a problem hiding this comment.
I will definitely fix this to revert to filters=None then 👍
There was a problem hiding this comment.
fwiw I think the spec is wrong here, and it should allow filters: [], but it is what it is 🤷
This switches exisiting code from
zarr-python2 tozarr-python3, which is part of fixing #33.