-
Notifications
You must be signed in to change notification settings - Fork 84
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
zarr-python
v3 compatibility
#516
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There's much less change here than I might have thought.
kerchunk/hdf.py
Outdated
@@ -496,6 +516,8 @@ def _translator( | |||
if h5obj.fletcher32: | |||
logging.info("Discarding fletcher32 checksum") | |||
v["size"] -= 4 | |||
key = str.removeprefix(h5obj.name, "/") + "/" + ".".join(map(str, k)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is the same as what _chunk_key
did? Maybe make it a function with a comment saying it's a copy/reimplementation.
By the way, is h5obj.name
not actually a string, so you could have done h5obj.name.removeprefix()
?
kerchunk/hdf.py
Outdated
shape=h5obj.shape, | ||
dtype=dt or h5obj.dtype, | ||
chunks=h5obj.chunks or False, | ||
fill_value=fill, | ||
compression=None, | ||
compressor=None, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So here, you could reintroduce the compressor
filters = filters[:-1]
compressor = filters[-1]
but obviously it depends on whether there are indeed any filters at all.
It would still need back compat, since filters-only datasts definitely exist.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
yeah the big issue is that v3 cares about what type of operation it is, and v2w doesnt so moving them around doesnt necessarily fix that bug
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So there needs to be a change upstream?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes this: zarr-developers/zarr-python#2325
kerchunk/hdf.py
Outdated
for k, v in self.store_dict.items(): | ||
if isinstance(v, zarr.core.buffer.cpu.Buffer): | ||
key = str.removeprefix(k, "/") | ||
new_keys[key] = v.to_bytes() | ||
keys_to_remove.append(k) | ||
for k in keys_to_remove: | ||
del self.store_dict[k] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is the hacky bit and could use some explanations. Even when requesting "v2", zarr makes Buffer objects, and the keys are also wrong?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah so two issues here:
- the keys we get from hdf are for example
/depth/.zarray
when then need to bedepth/.zarray
- we cant jsonify buffers, which is how the internal
MemoryStore
in v3 stores its data. So we need to convert the buffers to bytes to be serialized
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
OK - would appreciate comments on the code saying this.
Also worth noting... zarr 3 doesn't support numcodecs codecs out of the box. There is a pr to help this zarr-developers/numcodecs#524 (see also here for updated version zarr-developers/numcodecs#597) but it would mean a change to codec names which causes an incompatibility. For the initial icechunk examples we handle this in virtualizarr but long term it probably belongs here to work standalone. |
When this zarr-developers/zarr-python#2425 goes in it should unblock this to work full with zarr python v3. We will also need to create both numcodecs and zarr v3 codec versions of all the custom kerchunk codecs so that a given dataset can be loaded in either v2 or v3 contexts (say if you kerchunk a grib file, then want to convert those references to an icechunk store) |
Is that a subclass in each case, specifying that it is to be bytes->bytes? |
Sorry the numcodecs versions exists as they are today. Yes the v3 version would be basically subclassing zarr.abc.codec using the numcodecs implementations of the kerchunk codecs. Although grib decoder should really be bytes to array i think |
This is required upstream: fsspec/filesystem_spec#1734 |
Also of note: To get this to work with zarr 3, we pass an fsspec |
Grib works as long as read in zarr 2 format. To be used with zarr 3 the codec needs to be ported over to the new abc.Codec class from zarr |
OK, only failures now are in test__grib_idx.py , which use datatree. I think this is to do with a leading "/" in the paths... |
🎉 exciting progress! DataTree support is not yet supported fully for Zarr-Python 3, see pydata/xarray#9984. I think a conditional xfail would be the way to go here. |
I don't know how much effort we want to spend finding the deadlock triggered by the HDF5 embed test that is now commented out. |
I'd been bashing my head against those grib tests all day and couldn't figure out why, even in a debugger and doing things manually, none of the DataTree nodes had any data... In the process of messing with this I uncovered what may be a |
Might want to add this small update as well, since it better captures the intent of the version check so far as I can tell: mpiannucci#6 |
!!! Please file this as an issue. Maybe it helps with the datatree thing too. |
This comment was marked as outdated.
This comment was marked as outdated.
Great to see this progress! The preceding |
@TomNicholas , I am skipping that whole test module - all is green |
Wow! Okay please let us know when you release this then ❤️ |
@martindurant Not a huge deal, but I noticed that this PR didn't make it in: mpiannucci#6 Should I cut it against |
Actually, it looks like that's now dead code. Following this PR, kechunk only works with zarr3, for better or worse. |
So far the only tested file type is HDF5. Thats the only module that currently works with new zarr python in some way