[Numpy-discussion] is __array_ufunc__ ready for prime-time?

[Chris Barker]

On Sat, Nov 4, 2017 at 6:47 AM, Marten van Kerkwijk <
m.h.vankerkwijk at gmail.com> wrote:

>
> You just summarized excellently why I'm on a quest to change `asarray`
> to `asanyarray` within numpy


+1 -- we should all be using asanyarray() most of the time. However a
couple notes:

asarray() pre-dates asanyarray() by a LOT. asanyarray was added to better
handle subclasses, but there is a lot of legacy code out there.

An legacy coders -- I know that I still usually use asarray without
thinking about it -- sorry!

Obviously, this covers only ndarray
> subclasses, not duck types, though I guess in principle one could use
> the ABC registration mechanism mentioned above to let those types pass
> through.
>

The trick there is that what does it mean to be duck-typed to an ndarray?
For may applications its' critical that the C API be the same, so
duck-typing doesn't really apply.

And in other cases, in only needs to support a small portion of the numpy
API. IS essence, there are an almost infinite number of possible ABCs for
an ndarray...

For my part, I've been known to write custom "array_like" code -- it checks
for the handful of methods I know I need to use, and tI test it against the
small handful of duck-typed arrays that I know I want my code to work with.

Klunky, and maybe we could come up with a standard way to do it and include
that in numpy, but I'm not sure that ABCs are the way to do it.


-CHB


-- 

Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

Chris.Barker at noaa.gov
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20171106/26494b9f/attachment.html>