On 10Feb2004, Christian T. Steigies <address@hidden> wrote:
 I wrote a little function to generate maximum length pseudo random
 sequences, the function returns a vector. I am doing some matrix
 multiplication with this vector and notice that in octave 2.1.50 the vector
 is treated as row vector, whereas in 2.1.53 it is treated as a column
 vector. Simple example:

 octave:1> a(1)=1
 a = 1
 octave:2> a(2)=2
 a =

 1
 2

 octave:3> version
 ans = 2.1.50
 octave:4> [nr,nc]=size(a)
 nr = 2
 nc = 1


 octave:1> a(1)=1
 a = 1
 octave:2> a(2)=2
 a =

 1 2

 octave:3> version
 ans = 2.1.53
 octave:4> [nr,nc]=size(a)
 nr = 1
 nc = 2

 Is this expected behaviour and if so is it going to stay that way? A quick
 search didn't show anything in the changelog, but maybe I was looking for
 the wrong thing.
Yes, this change is permanent, for compatibility with Matlab. In the
past, you could use the builtin variable prefer_column_vectors to
choose what Octave would do for things like
clear a; a(1)=1; a(2)=2
but now Octave will only do the Matlabcompatible thing and create row
vectors.
Other than the potential compatibility problems, the existence of a
variable like prefer_column_vectors made it harder to write code that
would work no matter which preference other users might have. So it
was best to eliminate the variable.
BTW, if you are creating vectors by doing things like
clear a;
for i = 1:N
a(i) = something();
end
then you should probably rethink your method, because this type of
loop the vector A to be resized N times, which will be slow.
jwe

Octave is freely available under the terms of the GNU GPL.
Octave's home on the web: http://www.octave.org
How to fund new projects: http://www.octave.org/funding.html
Subscription information: http://www.octave.org/archive.html
