Require numpy >= 1.24#250
Conversation
With older `numpy` versions ufuncs without parameters do not work. There are several such functions in `pyerfa`, so those `numpy` versions were not properly supported anyways.
mhvk
left a comment
mhvk
left a comment
There was a problem hiding this comment.
Thanks, this is indeed overdue. Only one request: could you add an entry in CHANGES.rst? Those entries will be put in the release notes (and really we should have a release soon, so this is another good reason to do so).
| ==================== | ||
| - Let multi-output functions now return a ``namedtuple`` so that | ||
| the various results can be accessed by attribute. [gh-176] | ||
| - ``pyerfa`` now requires Python 3.10 or later. [gh-185] |
There was a problem hiding this comment.
Wasn't this PR about numpy? In itself fine to mention python too!
There was a problem hiding this comment.
This PR only updates the numpy requirement, but I noticed that I had forgotten to add the PR number to the erfa.cal2jd() bugfix entry, and then I thought if I am updating the change log I might as well add the entry for the Python version requirement too.
There was a problem hiding this comment.
Ah, sorry, I had missed the numpy addition further down, so all is OK.
| ==================== | ||
| - Let multi-output functions now return a ``namedtuple`` so that | ||
| the various results can be accessed by attribute. [gh-176] | ||
| - ``pyerfa`` now requires Python 3.10 or later. [gh-185] |
There was a problem hiding this comment.
Ah, sorry, I had missed the numpy addition further down, so all is OK.
3 participants
With older
numpyversions ufuncs without parameters do not work. There are several such functions inpyerfa, so thosenumpyversions are not being properly supported anyways. SPEC 0 allows dropping support fornumpy< 2, so we will still be very generous with our requirements.