[Scilab-Dev] SEP #95 : ismatrix(), isrow(), iscolumn(), IsAScalar()

[Samuel Gougeon]

Hello,

This SEP has been posted under the review 
https://codereview.scilab.org/#/c/11455/ :
https://codereview.scilab.org/cat/11455%2C2%2CSEP/SEP_095_ismatrix_isrow_iscolumn_issquare.odt

The case of issquare() is discussed in the separated thread
http://mailinglists.scilab.org/SEP-95-issquare-tt4026683.html

The full rationale of this SEP says:
/These functions are some elementary functions that exist in Matlab to 
check if a matrix or a //
//vector is a matrix, a row, a column or a square matrix /

No more comments about this a-priori Matlab worship.
About these 4 functions: difficult to me thinking that their 
implementation is a priority,
or even really needed. Let's consider that they could be useful, and 
start with the most
provocative one:

*IsAScalar()*:
  * is replacing size(A,"*")==1withIsAScalar(A)worthwhile??
     About it: where are the capitals? I have already forgotten... Happily,
     there are 5 help maintained pages in en, fr, ru, pt and ja to get them.
     Ouf.
     Pushing Scilab on this side, it should finally fall...
     By the way, won't matlabers -- that are expected to love it -- shout
     "Is it isscalar(), or something else ?"
     It would rather deserve a class action against it, wouldn't it? At 
least.
     By the way, the posted help page says that IsAScalar() could be applied
     only to decimals. I agree, it will never be too weak :-)
     How is it possible to still weaken it?
     Sorry for this bad mood...

For the next 3 functions that are proposed:*isrow(), iscolumn() and 
ismatrix()*, the
following features are discussed:
  * case of []  :
     The answer should be conventional: either the answer is %T for the 
3 functions,
     or it is %F for the 3.
     Matlab answers look unconsistent. They are:
isrow([])  => 0
iscolumn([]) => 0  but
    ismatrix([]) => 1
     Could anyone explain how and why, while any other non-empty
     column or row is detected as a matrix, [] is also claimed to be a 
matrix()
     but neither a row nor a column..

     The present Scilab implementation gives:
     isrow([])  => %F    iscolumn([]) => %F   but    ismatrix([]) => %T
     So, it is Matlab-compatible. It is likely its overal business.

  * case of a scalar (says %pi):
     Matlab returns:
isrow(pi)    => 1
     iscolumn(pi) => 1
     ismatrix(pi) => 1
I definitely agree with Matlab, despite the Matlab-mania of Scilab is 
weakening...:
     isrow(%pi) => %F
     iscolumn(%pi) => %F
     ismatrix(%pi) => %F

  * case of an hypermatrix with singleton dimension(s):
     - Matlab returns always 0 (false) for any input hypermatrix, for 
the 3 functions,
        even when it has singleton dimension, say: ismatrix(rand(2,1,2)) 
=> 0
     - This behaviour is debattable. Scilab could be more comprehensive 
at no
       expenses:

  * Here are some suggested implementations (without overhead):
      function r = ismatrix(A)
          s = size(squeeze(A))
          s = length(s(s>0))
          r = s>0 & s<3
      endfunction
        giving:
     ismatrix([])          // %F
     ismatrix(%pi)         // %T
     ismatrix([1,1])       // %T
     ismatrix([1;1])       // %T
     ismatrix(rand(2,3))   // %T
     ismatrix(rand(2,3,2)) // %F
     ismatrix(rand(2,1,3)) // %T

      function r = isrow(A)
          s = size(A)
          s2 = s(find(s(2:$)>=1))
          r = s(1)==1 & (length(s2)==1 | prod(s2)==1)
      endfunction
        giving:
     isrow([])          // %F
     isrow(%pi)         // %T
     isrow([1,1])       // %T
     isrow([1;1])       // %F
     isrow(rand(2,3))   // %F
     isrow(rand(1,3,2)) // %F
     isrow(rand(1,1,3)) // %T
     isrow(rand(2,1,1)) // %F

         and finally:

      function r = iscolumn(A)
          s = size(A)
          r = s(1)>=1 & length(find(s(2:$)>1))==0
      endfunction
        giving:
     iscolumn([])          // %F
     iscolumn(%pi)         // %T
     iscolumn([1,1])       // %F
     iscolumn([1;1])       // %T
     iscolumn(rand(2,3))   // %F
     iscolumn(rand(1,1,3)) // %F
     iscolumn(rand(3,1,1)) // %T

Comments of any forthcoming user would be very appreciable.
Hope to read them soon,

Regards
Samuel Gougeon

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/dev/attachments/20130513/b68366fe/attachment.htm>