Adds tests (and fixes) for the dtype invariant. #739
Merged
+54
−20
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
That is, np.array(array).dtype == np.dtype(eltype(array)).
For this to hold, we needed to restrict to only creating dtypes for
primitives, tuples and named tuples. We removed support for arbitrary
structs (which are not supported by our implementation of the array
interface and buffer protocol).
We also worked around a feature/bug/quirk of numpy in that if you do
numpy.dtype(descr) where descr is a list of (name,type) field
descriptors of a struct, then the dtype you get is not the same as the
dtype of an array constructed from something whose array interface has
that same descr. In particular, if any item in descr is struct padding
like ("", "|V4"), then on conversion to a dtype the name is replaced
with e.g. "f2". Going the array route, the padding gets ignored and
does not feature in the resulting dtype. The fix here is to compute a
different representation of the same information for the dtype - namely
the dict of names, types and offsets way.
Member
Author
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
That is, np.array(array).dtype == np.dtype(eltype(array)).
For this to hold, we needed to restrict to only creating dtypes for primitives, tuples and named tuples. We removed support for arbitrary structs (which are not supported by our implementation of the array interface and buffer protocol).
We also worked around a feature/bug/quirk of numpy in that if you do numpy.dtype(descr) where descr is a list of (name,type) field descriptors of a struct, then the dtype you get is not the same as the dtype of an array constructed from something whose array interface has that same descr. In particular, if any item in descr is struct padding like ("", "|V4"), then on conversion to a dtype the name is replaced with e.g. "f2". Going the array route, the padding gets ignored and does not feature in the resulting dtype. The fix here is to compute a different representation of the same information for the dtype - namely the dict of names, types and offsets way.