Re: 3.6.1: undocumented changes to randperm?

[Juan Pablo Carbajal]

On Fri, Mar 2, 2012 at 1:33 PM, Ben Abbott <address@hidden> wrote:
> On Mar 2, 2012, at 7:14 AM, Juan Pablo Carbajal wrote:
>
>> Hi,
>>
>> I am comparing the output of
>> randperm(5,3)
>>
>> in 3.4.3 and 3.6.1
>>
>> In 3.4.3 this creates a matrix with 3 random permutations
>>   4   1   2   3   5
>>   4   3   2   5   1
>>   3   2   1   4   5
>>
>> In 3.6.1 this gives a permutation of 5 integers of length 3.
>>
>> I am ok with the change, but searching the output of
>>> news
>> in 3.6.1 I do not find anything about randperm
>>
>> Is this an undocumented change?
>>
>> Thanks
>
> The changeset is below.
>
>        http://hg.savannah.gnu.org/hgweb/octave/rev/2896c083576a
>
> The current implementation is compatible with Matlab. Compatibility fixes are 
> treated like bugs, which we normally don't mention in the NEWS file.
>
> Ben

I see.
however the NEWS files is about changes that are visible to the users.
My code falling apart after an update was clearly visible. Maybe bug
fixes that change interface should be at least mentioned as such? Like

The following functions changed their interface to be more Matlab compatible:

 * randperm
...

Just an idea, as I said I am Ok with the change. But I suggested
colleagues to update to octave 3.6.1 (and I was heard, amazing!) and
then I got a stream of complaints. :D

Thanks


-- 
M. Sc. Juan Pablo Carbajal
-----
PhD Student
University of Zürich
http://ailab.ifi.uzh.ch/carbajal/