Re: sort indexing

[Nicholas Jankowski]

On Feb 15, 2016 4:10 PM, "Daniel J Sebald" <address@hidden> wrote:
>
> On 02/15/2016 02:39 PM, Doug Stewart wrote:
>>
>>
>> I have a question about indexing.
>>
>> a=randi(9,5)
>> [b i]=sort(a)
>> c=a(i)
>>
>> I would think that c should be the same as b, but it is not.
>>
>> The index array i has all the correct information in it as can be seen with
>>
>> for k=1:columns(a)
>> w(:,k)=a(i(:,k),k);
>> endfor
>> now b-w is equal to 0
>>
>> Is there  some technical reason that if you try and use the index array
>> i on an matrix
>> of the same dimensions, that it can't work?
>> I would think that it should apply each col of the i to a  in a(i)
>> as I did in the loop.
>>
>> I know it can be vectorized:
>>
>>   a(sub2ind (size(a), i, repmat(1:4, rows(a), 1)))
>>   a(i+(0:columns(a)-1)*rows(a))
>>
>> but I just think that octave should be smart enough to just do a(i)
>>
>> Doug
>
>
> Doug,
>
> I looks like a() is being vectorized when accessed as a(i):
>
> octave:1> a = [0:4] + 5*[0:4]'
> a =
>
>     0    1    2    3    4
>     5    6    7    8    9
>    10   11   12   13   14
>    15   16   17   18   19
>    20   21   22   23   24
>
> octave:2> i = a + 1
> i =
>
>     1    2    3    4    5
>     6    7    8    9   10
>    11   12   13   14   15
>    16   17   18   19   20
>    21 22 23 24 25
>
> octave:3> a(i)
> ans =
>
>     0    5   10   15   20
>     1    6   11   16   21
>     2    7   12   17   22
>     3    8   13   18   23
>     4    9   14   19   24
>
> The Octave documentation states that indexing multidimensional arrays when indexed by a scalar are treated in column-major order.  Hence in the above example because of the way I defined i, a(i) is transposed.
>
> https://www.gnu.org/software/octave/doc/interpreter/Index-Expressions.html
>
> Dan
>

Anyone checked Matlab compatibility yet?