[Numpy-discussion] dtype subarray comparison

[Mark Wiebe]

On Thu, Oct 21, 2010 at 2:47 PM, Pauli Virtanen <pav at iki.fi> wrote:

> Thu, 21 Oct 2010 14:25:24 -0700, Mark Wiebe wrote:
> [clip]
> > That sounds like it could be tricky to handle correctly, especially
> > since it should probably work to compare Fortran-order arrays with
> > non-Fortran ones.
> >
> > Does anything actually depend on this behavior?  Maybe it's something
> > that could change?
>
> I believe nothing in Numpy or Scipy depends on that. The main question is
> how much of 3rd party code is likely to use this feature.
>
> I suppose this quirk is not widely relied on, and as far as I know, the
> behavior has not been documented. For instance, Travis's Guide to Numpy
> says (on subdtype): "If a field whose dtype object has this attribute is
> retrieved, then the extra dimensions implied by the shape are tacked on to
> the end of the retrieved array."
>
> IMO, it should be possible to change this with little prior notice.
>
> A can of worms :)


I found the commit where Travis introduced the behaviour:

http://github.com/numpy/numpy/commit/2fd136eb7977dd5172eb92d173a7c39921b0cf98

http://projects.scipy.org/numpy/changeset/2116

I didn't find any discussion in the mailing list or a trac ticket, so
hopefully Travis could elaborate on it?

-Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20101021/a6335276/attachment.html>