[Numpy-discussion] Asking for opinions: Priops

Friedrich Romstedt friedrichromstedt@gmail....
Thu Sep 23 05:54:16 CDT 2010


Hi Sebastian,

Thanks for your reply!

2010/9/22 Sebastian Walter <sebastian.walter@gmail.com>:
> [...] I think the issue is not so much numpy but rather the
> way Python implements operator overloading using methods like __add__
> and __radd__.  Hence, your suggestion seems to be a Python Enhancement
> Proposal and should be discussed without any references to numpy or
> bugs related to numpy.set_numeric_ops.

Yeah, I agree, but i refrained from spamming the list with two
separate mails, although I see now it would have been better, and
would have brought more eyes on it.

For the PEP, I will look into this and will check the mailing lists.

For the technical things, I think the implementation of the operations
must lie in the classes, and thus __add__ etc. are in principle okay.
But as described in the new README, there is need to organise this
functions.  priop (note the rename, "priops" turned out to be always
spelled "priop" in the source code) could define this new layer.

> Maybe you could also have a look at Go's interfaces (Googles
> programming language) which seems to be somewhat related to your
> approach.

I will try.  Can you provide an URL?

> On a more practical note: Why exactly do you use set_numeric_ops? You
> could also
> 1) use numpy.ndarrays with dtype=object

This is too slow.  And it's eating up memory because of the Python
objects stored with all their bells and whistles.

> 2) or create new numpy.ndarray -like class and set  __array_priority__ > 2
> both approaches work well for me.

I wanted to avoid exactly this because I think priop is a better
approach, via set_numeric_ops().

The new URL is:

http://github.com/friedrichromstedt/priop (just for the rename, which
may be discussable).

It contains now also an implementation, which was much less hard than
expected ...

Friedrich


More information about the NumPy-Discussion mailing list