Re: Ideas for auto BSX

[Jordi Gutiérrez Hermoso]

Argh, forgot to CC the list.

2011/9/30 Michael D Godfrey <address@hidden>:
> On 09/30/2011 12:11 PM, Jordi Gutiérrez Hermoso wrote:
>
> Perhaps ".-" would do? I've never seen
> anyone use this.
>
> This may be a good idea., even though the uniformity of the present
> implementation is also good.
>
> (I vaguely remember using something
> like that when we implemented this kind of expansion in a design
> system.  Long ago, though....)
>
> The point that the current implementation can lead to hard to
> find bugs is obviously worth thinking about.

This is the numpy discussion that motivated me:

   http://www.mail-archive.com/address@hidden/msg14705.html

They're absolutely right. It's a clunky interface, and there's no
reason why we have to mimic the same clunky interface in Octave if we
can do better.

To be fair, they do have a different cultural slant in numpy, where
they deeply treasure their "broadcasting" (most of that broadcasting
in numpy is what we slightly erroneously always refer to as
"vectorisation" in Octave). So perhaps every numpy user expects this
to be on but Octave users won't expect it.

I think making this a global option (and now we quibble over the
default, on by default, or off by default?) is going to be the only
solution.

- Jordi G. H.