Re: bsxfun for arrays

[Jordi Gutiérrez Hermoso]

2011/5/13 Jordi Gutiérrez Hermoso <address@hidden>:
> While working on #33292, I came across a behaviour of bsxfun I don't 
> understand:
>
>     x = rand(1,1,3); y= rand(2,2,3); u = bsxfun(@(t) t, x,y); sum(u(2,:))
>
> That above outputs zero at changeset b6eaecf83bb2 because all of
> u(2,:) is zero. I don't get it. Why wasn't the "vertical vector" x
> copied four times?

To expand on this, the following does output what I would expect:

      x = rand(1,1,3); y= rand(2,2,3); u = bsxfun(@(t,s) t+s-s, x,y);
sum(u(2,:))

but removing that +s-s, I get a different output.

- Jordi G. H.

>
> I had someone try this in Matlab, and they reported that it errored
> out with  "Error using ==> bsxfun: Invalid output dimensions" which I
> find even more mystifying. Who specified anything about output
> dimensions?
>
> TIA,
> - Jordi G. H.
>