[Numpy-discussion] NumPy-Discussion Digest, Vol 90, Issue 45

[Nathaniel Smith]

On Sun, Mar 16, 2014 at 4:37 PM, Colin J. Williams
<cjwilliams43 at gmail.com> wrote:
> I would like to see the case made for @.  Yes, I know that Guido has
> accepted the idea, but he has changed his mind before.

I'm not sure how to usefully respond to this, since, I already wrote a
~20 page document making the case for @? Maybe if you think the
arguments in it aren't good, it would be more helpful to explain which
ones and why?

> The PEP seems neutral to retaining both np.matrix and @.

I'm not sure what gives you this impression. The main point of the
whole first section of the PEP is to explain why the existence of
np.matrix causes problems and why a substantial majority of developers
hate it, and how adding @ will let us solve these problems. Whether we
actually get rid of np.matrix is a more complicated question (we'll
need sort of compatibility/transition strategy, it will depend on how
quickly python versions with @ support are adopted, etc.), but at the
very least the goal is that @ eventually replace it in all new code.

-n

-- 
Nathaniel J. Smith
Postdoctoral researcher - Informatics - University of Edinburgh
http://vorpus.org