Why does assigning a categorical with empty and non empty values result in a error in MATLAB 8.2(R2013b)?

I tried to execute the following commands in MATLAB R2013b

 

     >> D_num = 10*(1:10)';
     >> C_num = [1:3, NaN, NaN, 1]';
     >> D = nominal(D_num);
     >> C = nominal(C_num);
     >> D(10+(1:4), :) = C(1:4, :)

and I receive the following error:

Subscript indices must either be real positive integers or logicals.

 Error in categorical/subsasgn>convertCodesLocal (line 100)
         newlyAssigned(unique(bcodes)) = true;

 Error in categorical/subsasgn (line 51)
             [bcodes,anamesNew] = convertCodesLocal(bcodes,bnames,anames,a.isProtected);

I could execute the above code with no issue in MATLAB R2013a. How do I resolve this?

Answer 1

[MathWorks Support Team]

There is an unexpected behavior in MATLAB 8.2 R2013b in the way that assignment into categorical arrays is handled. This error is due to assigning category arrays with 'NaN' and nonempty values into another category array.

The workaround to this issue is to pre-define 'NaN' as a category. For example:

     >> D_num = 10*(1:10)';
     >> C_num = [1:3, NaN, NaN, 1]';
     >> D = nominal(D_num,num2str(D_num),unique(D_num));
     >> C = nominal(C_num,{'1','2','3','NaN'},[1 2 3 NaN]); 
     >> D(10+(1:4), :) = C(1:4, :)

This will return the expected result as shown:

D = 

       10 
       20 
       30 
       40 
       50 
       60 
       70 
       80 
       90 
       100 
       1 
       2 
       3 
       NaN