BUG: Avoid intp conversion regression in Cython 3 #25094
Merged
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
seberg
added a commit
to seberg/numpy
that referenced
this pull request
Nov 9, 2023
This is the minimal backport version of numpygh-25094, which simply aligns the Cython 2 and Cython 3 definitions.
In the Cython 3 `__init__.pxd` `intp` was defined to `Py_ssize_t` which is *correct*. However, Cython has special rules for `Py_ssize_t`. It uses the `PyNumber_AsIndex` (`operator.index`) function for conversion which rejects for example floats. Of course I prefer that behavior, but: 1. There is no great reason why `np.intp_t` should be all that special. 2. Changing it silently changes downstream (and `np.random`) behavior. It would be awesome if Cython would allow transitioning to using `operator.index()` style coercion since you can still explicitly convert via `int()`. Now, why the new definition to `signed long`!? `ssize_t` doesn't work because it is *also* special, using `PyLong_AsSize_t`. It seems to me that adding signed might give this type a bit of preference in promotion, so it is is just for good sport. (It is impossible to get Cython promotion fully right.)
d54da80 to
e26e75e
Compare
This was referenced Nov 9, 2023
|
PRs that make you go um ... Is the long here an actual C long? |
|
No, it tells Cython to make it behave like a C long, but it is whatever it is on the C level (which is an EDIT: Cython has |
|
LGTM. Is there a corresponding issue that this closes? |
|
gh-25061 was it as it is really a more broad issue, but already closed in the narrow version. I also ran into this in sklearn and the mtrand.pyx changes (which is undone here). |
|
In it goes, thanks Sebastian. |
charris
pushed a commit
to charris/numpy
that referenced
this pull request
Nov 11, 2023
This is the minimal backport version of numpygh-25094, which simply aligns the Cython 2 and Cython 3 definitions.
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.
In the Cython 3
__init__.pxdintpwas defined toPy_ssize_twhich is correct. However, Cython has special rules forPy_ssize_t. It uses thePyNumber_AsIndex(operator.index) function for conversion which rejects for example floats.Of course I prefer that behavior, but:
np.intp_tshould be all that special.np.random) behavior.It would be awesome if Cython would allow transitioning to using
operator.index()style coercion since you can still explicitly convert viaint().Now, why the new definition to
signed long!?ssize_tdoesn't work because it is also special, usingPyLong_AsSize_t. It seems to me that adding signed might give this type a bit of preference in promotion, so it is is just for good sport. (It is impossible to get Cython promotion fully right.)