Opened 16 years ago

Last modified 16 years ago

#1146 closed defect (fixed)

Problem with [:] in arrays and matrixes (from Mathcore)

Reported by: Jan Brugård Owned by: Jan Brugård
Priority: high Milestone:
Component: Version:
Keywords: Cc: Jan Brugård,

Description

See the model:

function z
  input Real B[:,:];
  output Real A[size(B, 1),size(B, 2)]=zeros(size(B, 1), size(B, 2));
end z;

The error messages are:

Check of z failed
Error: Wrong type or wrong number of arguments to zeros(size(B, 1), size(B, 2))'.

Functions like these are used in Modelica.Math.Matrices.singularValues.

There are problems with the colon dimensions [:] in many classes and I think a lot of tickets are written on different models that doesn't work because of this.

Change History (5)

comment:2 by Adrian Pop, 16 years ago

OpenModelica doesn't support defining the output
for a function at the same time you are defining
the output component.

The setting of the output needs to be moved into an algorithm section.

comment:3 by Jan Brugård, 16 years ago

Doing that does not help as far as I can see. I changed the model according your comment:

function singularValues "Compute singular values and left and right singular vectors"
  extends Modelica.Icons.Function;
  input Real A[:,:] "Matrix";
  output Real sigma[min(size(A, 1), size(A, 2))] "Singular values";
  output Real U[size(A, 1),size(A, 1)] "Left orthogonal matrix";
  output Real VT[size(A, 2),size(A, 2)] "Transposed right orthogonal matrix ";
  annotation(preferedView="info", Documentation(info="<HTML>
<h3><font color=\"#008000\">Syntax</font></h3>
<blockquote><pre>
         sigma = Matrices.<b>singularValues</b>(A);
(sigma, U, VT) = Matrices.<b>singularValues</b>(A);
</pre></blockquote>
<h3><font color=\"#008000\">Description</font></h3>
<p>
This function computes the singular values and optionally the
singular vectors of matrix A. Basically the singular
value decomposition of A is computed, i.e.,
</p>
<blockquote><pre>
<b>A</b> = <b>U</b> <b><font face=\"Symbol\">S</font></b> <b>V</b><sup>T</sup>
  = U*Sigma*VT
</blockquote></pre>
<p>
where <b>U </b>and <b>V</b> are orthogonal matrices (<b>UU</b><sup>T</sup>=<b>I,
</b><b>VV</b><sup>T</sup>=<b>I</b>). <b><font face=\"Symbol\">S
</font></b> = diag(<font face=\"Symbol\">s</font><sub>i</sub>) 
has the same size as matrix A with nonnegative diagonal elements 
in decreasing order and with all other elements zero
(<font face=\"Symbol\">s</font><sub>1</sub> is the largest element). The function
returns the singular values <font face=\"Symbol\">s</font><sub>i</sub>
in vector <tt>sigma</tt> and the orthogonal matrices in
matrices <tt>U</tt> and <tt>V</tt>.
</p>
<h3><font color=\"#008000\">Example</font></h3>
<blockquote><pre>
  A = [1, 2,  3,  4;
       3, 4,  5, -2;
      -1, 2, -3,  5];
  (sigma, U, VT) = singularValues(A);
  results in:
     sigma = {8.33, 6.94, 2.31}; 
  i.e.
     Sigma = [8.33,    0,    0, 0;
                 0, 6.94,    0, 0;
                 0,    0, 2.31, 0]
</pre></blockquote>
<h3><font color=\"#008000\">See also</font></h3>
<a href=\"Modelica:Modelica.Math.Matrices.eigenValues\">Matrices.eigenValues</a>
</HTML>"), Icon(coordinateSystem(extent={{-100,100},{100,-100}}, preserveAspectRatio=true, initialScale=0.1, grid={10,10})), Diagram(coordinateSystem(extent={{-100,100},{100,-100}}, preserveAspectRatio=true, initialScale=0.1, grid={10,10})));
protected 
  Integer info;
  Integer n=min(size(A, 1), size(A, 2)) "Number of singular values";
algorithm 
  U:=zeros(size(A, 1), size(A, 1)) "Left orthogonal matrix";
  VT:=zeros(size(A, 2), size(A, 2)) "Transposed right orthogonal matrix ";
  (sigma,U,VT,info):=Matrices.LAPACK.dgesvd(A);
  assert(info == 0, "The numerical algorithm to compute the
singular value decomposition did not converge");
end singularValues;

but still get the same error:

[81] 16:21:34 Validation of class Modelica.Math.Matrices.singularValues
Check of Modelica.Math.Matrices.singularValues failed
Error: Wrong type or wrong number of arguments to zeros(size(A, 1), size(A, 1))'.

comment:4 by Adrian Pop, 16 years ago

Ok, this is an additional problem with zeros()
function that doesn't support multiple dimensions
I think. I'll check that and come back with another
comment.

comment:5 by Per Östlund, 16 years ago

Fixed in revision 5093, see testcase FillSize. The problem was that elabBuiltinFill didn't handle cases where it couldn't constant evaluate the arguments to fill (zeros is just a special case of fill).

Note: See TracTickets for help on using tickets.