You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It looks like we should start preparing the transition to __numpy_ufunc__ (see e.g. numpy/numpy#4709 (comment)) since it will replace __array_prepare__ etc. in the long term.
@mhvk@mdboom - are either of you interested in trying to investigate this?
@astrofrog - I started implementing the __numpy_ufunc__ stuff quite a while ago, but ran into an issue with two parameter ufunc's (numpy/numpy#4160). Since that is resolved, I can revive this. This will have to wait a bit, though (till I return from holidays, i.e., at least a week).
I usually try to have only one issue per issue... (hence like issue2pr a lot!). Anyway, in this case the astropy PR seems fine but it is held up by numpy bugs -- indeed, the only three remaining numpy-1.9-blockers are from us...
@mhvk That's my preference as well--I'm definitely an issue2pr fan as well. But in cases where the original issue and the PR are separate it's a lot easier for me to trace the issue (both programmatically and by hand) if both are kept open. There are also many cases where a PR is closed without merging, and if the original issue it was meant to address is closed already then it's easy to lose track of the issue.
Activity
mhvk commentedon May 18, 2014
@astrofrog - I started implementing the
__numpy_ufunc__stuff quite a while ago, but ran into an issue with two parameter ufunc's (numpy/numpy#4160). Since that is resolved, I can revive this. This will have to wait a bit, though (till I return from holidays, i.e., at least a week).astrofrog commentedon May 18, 2014
@mhvk - great, thanks!
Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
mhvk commentedon Jul 8, 2014
Am closing this one, since there is an actual PR.
Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
embray commentedon Jul 11, 2014
@mhvk I try to leave issues open until and unless a PR addressing it is actually merged.
mhvk commentedon Jul 11, 2014
I usually try to have only one issue per issue... (hence like
issue2pra lot!). Anyway, in this case the astropy PR seems fine but it is held up by numpy bugs -- indeed, the only three remaining numpy-1.9-blockers are from us...embray commentedon Jul 11, 2014
@mhvk That's my preference as well--I'm definitely an
issue2prfan as well. But in cases where the original issue and the PR are separate it's a lot easier for me to trace the issue (both programmatically and by hand) if both are kept open. There are also many cases where a PR is closed without merging, and if the original issue it was meant to address is closed already then it's easy to lose track of the issue.Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
Add __numpy_ufunc__, adjust tests, remove numpy-dev bugs (closes astr…
mhvk commentedon Feb 9, 2017
Added
upstream fix requiredsince this is still not in numpy. Also changed title to__array_ufunc__-- the likely successor.[-]Transition Quantity to use ``__numpy_ufunc__``[/-][+]Transition Quantity to use ``__array_ufunc__``[/+]bsipocz commentedon Jun 27, 2017
I believe #2583 implemented this, reopen if that's not the case.