Hi all, I'm writing to propose accepting my ByteStringDType NEP, with the intention of including it as a new feature in NumPy 2.6.
The latest NEP text is here: https://numpy.org/neps/nep-0058-bytestring-dtype.html The mailing list thread discussing the NEP is here: https://mail.python.org/archives/list/[email protected]/thread/O4BIZE4YIYAS2SYOMXIENVS7BP2NNNWI/ I don't think there are any remaining substantive discussion points. Joren was worried about the original proposal to make the scalar type be a Python bytes. The latest version as a new NumPy vbytes scalar because subtyping bytes is much easier than subtyping str and trying to use it to represent UTF-8 data. Joren was also worried about the typing difficulties inherent with StringDType and ByteStringDType's missing data support. I argued that limitations of the type system shouldn't limit runtime behavior of NumPy's features and that it's probably worth expanding the type system to support behavior keyed to floating point values like nan now that there is a concrete use-case in NumPy. I've spoken to Joren and I think he's fine with the current text and ByteStringDType prototype. The prototype currently lives in a draft PR ( https://github.com/numpy/numpy/pull/32772) which I plan to mark as ready to merge next week. I'm also hoping we can merge https://github.com/numpy/numpy/pull/32757, which deprecates implicit coercion of bytes to strings (e.g. np.array(["a", b"b"])). It provides shared machinery for dealing with this, which will let me avoid adding a bunch of special cases for ByteStringDType. If there are no substantive objections within 7 days from this email, then the NEP will be accepted; see NEP 0 for more details. Thanks for your time and consideration, -Nathan
_______________________________________________ NumPy-Discussion mailing list -- [email protected] To unsubscribe send an email to [email protected] https://mail.python.org/mailman3//lists/numpy-discussion.python.org Member address: [email protected]
