conv.m compatibility

[Moritz Borgmann]

At 5:03 Uhr -0700 2008-04-14, Rolf Schirmacher wrote:
>  > | In matlab b controls whether y is a
>>  | column or row vector.
>>  | Example (matlab):
>>  | >> a=ones(1,5);
>>  | >> b=ones(1,5);
>>  | >> size(conv(a,b))
>>  | ans =
>>  |      1     9
>>  | >> size(conv(a.',b))
>>  | ans =
>>  |      1     9
>>  | >> size(conv(a,b.'))
>>  | ans =
>  > |      9     1
>  > | In octave y is always a row vector.
>  >
>  > Are you sure this is not a bug in Matlab?  Nearly all other polynomial
>>  operations return row vectors, and this feature does not seem to be
>>  documented.
>>
>  > jwe
>
>I am not sure about our current understanding whether Matlab might have a
>bug or octave by not following its conventions... at least I am currently
>looking through some wired code that depends on this behaviour ("feature"?)
>of the other brand, and so I want to bring this issue to your attention
>again: Should we adopt to Matlabs behaviour or not?

see my recent thread

http://www.nabble.com/conv-does-not-preverve-orientation-of-input-td16342836.html

Matlab's behavior is even weirder than the description given in the 
2005 bug report cited by you, see my original post.

After the discussion, I tend to retract my original patch in favor of 
another one that doesn't change functionality, but makes the output 
orientation explicit in the documentation - and fix the code that 
makes assumptions about the orientation of the output.

-M