Add NDBuffer.empty #3191
Merged
Add NDBuffer.empty #3191
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
github-actions
bot
added
the
needs release notes
TomAugspurger
force-pushed
the
tom/ndarray-empty
branch
from
7a9767b to
9357820
Compare
github-actions
bot
removed
the
needs release notes
TomAugspurger
force-pushed
the
tom/ndarray-empty
branch
from
9357820 to
e91b630
Compare
TomAugspurger
force-pushed
the
tom/ndarray-empty
branch
from
e91b630 to
3f6d18f
Compare
In zarr-developers#2904, I've found that `np.empty(...)` can be quite a bit faster than the `np.full` we use in places (https://github.com/zarr-developers/zarr-python/blob/baabf08d07e8518e3d37bd83c493a1d46ea7ac1d/src/zarr/core/buffer/cpu.py#L149. Though note that `np.empty` and `np.zeros` are about the same for the cases where we use that.) For the common case of chunk-aligned reads, the memset used by `np.full` or `np.zeros` should be unnecessary, because the codec pipeline will overwrite the memory anyway (or overwritten with `fill_value` if the store is missing the key). This adds a new method `NDBuffer.empty`, mirroring `np.empty`. To preserve backwards compatibility, it's *not* abstract. It delegates to the less efficient `NDBuffer.create`. But I've implemented it for our `gpu` and `cpu` buffers.
TomAugspurger
force-pushed
the
tom/ndarray-empty
branch
from
3f6d18f to
0ed6bc1
Compare
TomAugspurger
commented
Jul 1, 2025
| dtype: npt.DTypeLike, | ||
| order: Literal["C", "F"] = "C", | ||
| ) -> Self: | ||
| return super(cpu.NDBuffer, cls).empty(shape=shape, dtype=dtype, order=order) |
There was a problem hiding this comment.
This is to test the base core.NDBuffer implementation. This class inherits from cpu.NDBuffer so we use super(cpu.NDBuffer, cls) to get its parent's impelentation (the base class).
d-v-b
approved these changes
Jul 3, 2025
|
Just a note: this might cause issues for people running the |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #3191 +/- ##
=======================================
Coverage 94.48% 94.49%
=======================================
Files 78 78
Lines 8609 8623 +14
=======================================
+ Hits 8134 8148 +14
Misses 475 475
🚀 New features to boost your workflow:
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In #2904, I've found that
np.empty(...)can be quite a bit faster than thenp.fullwe use in places(
zarr-python/src/zarr/core/buffer/cpu.py
Line 149 in baabf08
np.emptyandnp.zerosare about the same for the cases where we use that.)For the common case of chunk-aligned reads, the memset used by
np.fullornp.zerosshould be unnecessary, because the codec pipeline will overwrite the memory anyway (or overwritten withfill_valueif the store is missing the key).This adds a new method
NDBuffer.empty, mirroringnp.empty. To preserve backwards compatibility, it's not abstract. It delegates to the less efficientNDBuffer.create. But I've implemented it for ourgpuandcpubuffers.