Fix v3 codec pipeline #4
Conversation
virtualizarr/zarr.py
Outdated
| @@ -4,6 +4,7 @@ | |||
| import numcodecs | |||
| import numpy as np | |||
| import ujson # type: ignore | |||
| from zarr.core.metadata.v3 import parse_codecs | |||
There was a problem hiding this comment.
Will this break compatibility with non-v3 zarr-python? In fact what's the best way to approach that in general? I don't mind updating virtualizarr to require bleeding edge zarr-python.
There was a problem hiding this comment.
We can try to import it and if not the pipeline function can just error?
There was a problem hiding this comment.
Sure - so let's move the import inside the pipeline function.
|
I have the codecs resolving correctly, but currently data isn’t loading right so going to take a beat to look into why |
| @@ -213,4 +217,5 @@ def _num_codec_config_to_configurable(num_codec: dict) -> dict: | |||
| Convert a numcodecs codec into a zarr v3 configurable. | |||
| """ | |||
| num_codec_copy = num_codec.copy() | |||
| return {"name": num_codec_copy.pop("id"), "configuration": num_codec_copy} | |||
| name = "numcodecs." + num_codec_copy.pop("id") | |||
There was a problem hiding this comment.
This is subject to zarr-developers/numcodecs#597 being installed.. but it works
|
Im going to merge this in, because without it the icechunk branch isnt super useful |
docs/releases.rstapi.rst