Why does the following operation not throw an error?

u =

         0    0.7071    0.7071
    0.7071         0    0.7071
    0.7071    0.7071         0

>> u(i,:)

ans =

         0    0.7071    0.7071

>> u2_mod I feel like the following is a source of error, because it could produce a behavior that is not wanted, but does not throw an error.

u2_mod =

         0
    0.6927
    0.6927

>> u(i,:) - u2_mod

ans =

         0    0.7071    0.7071
   -0.6927    0.0144    0.0144
   -0.6927    0.0144    0.0144

[Kaushik Lakshminarasimhan]

This is the documented behaviour of the '-' operator (see help minus). It will give you an output even if lengths along the Nth dimension do not match, as long as one of the operands is a scalar in that dimension.

Answer 1

[Veda Upadhye]

Hi Manuel,

The output you get is the expected output as explained by Kaushik. Both the matrices are of compatible sizes: https://www.mathworks.com/help/matlab/matlab_prog/compatible-array-sizes-for-basic-operations.html

Here, the second operand in this operation would get scaled according to the size of the first operand.

Hope this helped!

-Veda

Answer 2

[Manuel]

In my opinion this operation should at least show a warning message, since one does not always check for the right indices (row or colum vector) and it could completely mess up the code.

[Guillaume]

since one does not always check for the right indices (row or colum vector)

One should always validate the assumptions that he makes when he writes code. In particular, if one writes code that expects a vector to be in a particular direction (e.g. a column vector), the first line of the code should be:

validateattributes(expectedvector, {expectedclasses}, {'column'});

Alternatively one could ensure that the code works whatever the shape of input:

guaranteedtobecolumn = anyshapeinput(:);

"one does not always check" is just poor practice and will lead to bugs.

[Manuel]

I totally agree with you. Nevertheless, it would not hurt if a.m. operation trows an error since it is easily corrected...