Repository navigation
BUG: numpy.char.chararray is deprecated #19216
Description
xref: numpy/numpy#30605
this causes about 800 tests to fail, most of which correspond to a single failure mode where we check an object type with isinstance and can easily be handled (I'll open a patch in a sec).
The rest of them are not necessarily difficult to handle, but patching may require breaking changes since we may need to change the return type of public functions (possibly indirectly) in cases where we deliberately instantiate this class.
Related issue:
Linked PRs:
Activity
neutrinoceros commentedon Jan 26, 2026
all remaining failures are coming from a single line in
io.fitswhere we intentionally create anp.char.chararrayview. Not sure how to change it yet. See https://github.com/astropy/astropy/pull/19217/changes#r2726958048[-]BUG: numpy.char.charray is deprecated[/-][+]BUG: numpy.char.chararray is deprecated[/+]saimn commentedon Jan 26, 2026
See also #3862, it's not only about
isinstancechecks, io.fits relies on chararray to automatically strip trailing spaces from string columns.neutrinoceros commentedon Jan 26, 2026
Yes. #19217 is only meant to eliminate the dominant failure mode that'll hit CI in a couple days. I'm hoping we can tackle other long term issues after that
pllim commentedon Feb 3, 2026
Until we can figure out a proper solution (which seems non-trivial), should we just ignore the DeprecationWarning for now? It is very disruptive. I have issued a memo on astropy-dev asking downstream to ignore it.
neutrinoceros commentedon Feb 3, 2026
It looks non trivial indeed. I'd be happy to dive further in, I've just been reluctant to do it while this PR is unresolved
iamtanishqjain commentedon Feb 4, 2026
neutrinoceros commentedon Feb 4, 2026
thanks for offering but we're already working on it.
iamtanishqjain commentedon Feb 9, 2026
pllim commentedon Feb 12, 2026
Can I re-request #19216 (comment) again for the warning to be temporarily ignored? We could easily miss remote data test problems (e.g., #19294) unless we also run Extra CI, which is optional.
neutrinoceros commentedon Feb 12, 2026
I would still prefer #19217 as the temporary patch, but it doesn't seem to be going anywhere, so if that's the only way forward, let's do it.
pllim commentedon Feb 13, 2026
pllim commentedon Feb 13, 2026
#19303 works and does not change any functionality. Turns out I have to ignore the warning in not 1, but 3 places, but I do have to ignore it in functional code, not just the test. I think reverting this should be easy enough when a more serious fix is ready to go?
neutrinoceros commentedon Feb 13, 2026
@saimn
pllim commentedon Feb 20, 2026
I am going to keep this open until we have a backport version, so we don't forget.
pllim commentedon Feb 26, 2026
Manual backport happened. This is done. Thanks, everyone!
#19335
Human gold patch for astropy__astropy-19335
Human gold patch for astropy__astropy-19335