Any one experienced similar issue?

[a2012]

these two examples worked with me before on Matlab 2009 but now they are not working on Matlab 2013a:

x=sdpvar(4,5);

z=sdpvar(4,3);

 y=sdpvar(2,5);

 x(2:end)=0;

 x(:,[1;2;3])=z;

 x(3:4,:)+y

 Error in  +  (line 161)

                all_lmi_variables = [X.lmi_variables Y.lmi_variables]; 

x=sdpvar(4,5);

z=sdpvar(4,3);

y=sdpvar(2,5);

idx=ones(4,5);

idx(10)=0;

x(logical(idx))=0;

x(:,[1;2;3])=z;

x(3:4,:)+y

 

Error in  +  (line 161)

                all_lmi_variables = [X.lmi_variables Y.lmi_variables]; 

[Johan Löfberg]

Which YALMIP version? Works here...

[a2012]

last one. 

I have also this code that used to work for me before:

x=sdpvar(2,3)

x(2,:)=[1 2 3]

Error using sdpvar/subsasgn (line 273)

Error using sub2ind (line 52)

Out of range subscript.

[a2012]

these results are from Matlab 2013b not 2013a as I said before.

[Johan Löfberg]

Sorry I was running the develop version where this is already fixed. Change line 104 in sdpvar/subsasgn to I.subs{2} = 1:m;

[a2012]

thanks a lot. 

changing line 104 in subsasgn fixes the problem in x(2,:)=[1 2 3] .

but unfortunately it did not fix the two issues mentioned in the original post. 

what is happening in the two examples above is that X.lmi_variables (in plus function) will be defined as column vectors which causes horzcat issue in line 161.

[Johan Löfberg]

replace

[a2012]

Sorry the two examples in the original post still give the same error.

Error using horzcat

Dimensions of matrices being concatenated are not consistent.

Error in  +  (line 160)

                all_lmi_variables = [X.lmi_variables Y.lmi_variables];

[Johan Löfberg]

You cannot have replaced the file as line 160 doesn't look like that in the new @sdpvar/subsasgn.m I attached

[Johan Löfberg]

Sorry, got a bit confused here as there appears to be two completely different issues. I cannot reproduce the problem with the plus operator, will have to play around a bit

[Johan Löfberg]

This one should work in all matlab versions.

Thank you for alerting me.

FYI, the underlying problem is that MATLAB R2013 and later appears to return weird dimensions in union

R2012

 union([],[1 2 3])

union([1],[1 2 3])

ans =

     1     2     3


ans =

     1     2     3

R2014

union([],[1 2 3])

ans =

     1
     2
     3

K>> union([1],[1 2 3])

ans =

     1     2     3

[a2012]

thanks a lot. this fixes the problem.

:)

[Mark L. Stone]

Johan wrote "FYI, the underlying problem is that MATLAB R2013 and later appears to return weird dimensions in union"

As it turns out, it's not a bug, it's a feature.  Per help union:
    The behavior of union has changed.  This includes:
      -    occurrence of indices in IA and IB switched from last to first
      -    orientation of vector C
      -    IA and IB will always be column index vectors
      -    tighter restrictions on combinations of classes
 
    If this change in behavior has adversely affected your code, you may
    preserve the previous behavior with:
 
       [C,IA,IB] = union(A,B,'legacy')
       [C,IA,IB] = union(A,B,'rows','legacy')