Odd behavior when using character sequence as an array index

I was trying to show my students how to use a vector as a lookup table to swap the case of alphabetic characters when I encountered this behavior that I do not understand.

Here is the lookup table:

>> swapCase = char(1:127)
ans =
 !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~

Now in that string I want to replace the upper case letters with lower case letters:

>> swapCase('A':'Z') = 'a':'z';

But it shows this error:

In an assignment  A(:) = B, the number of elements in A and B must be the same.

Hm? I didn't type swapCase(:) , I typed swapCase('A':'Z') . This looks like a bug of some sort. Either the Matlab parser is wrong or the error message is wrong.

It works fine if the argument is the numeric sequence 65:90 or the explicit string 'ABC...Z' . The only way I could get it to work with a character sequence is by placing it in square brackets:

>> swapCase(['A':'Z']) = 'a':'z';

But then that contradicts Matlab's principle of vector concatenation. What I mean is that 'A':'Z' and ['A':'Z'] and [['A':'Z']] should yield the same vector/string 'ABC...Z', but this is a situation where it does not.

Answer 1

[Philip Borghesani]

This is a known bug: Bug 405376

character indexing is a bit dangerous because of it and this which is not a bug just a special case:

a=1:5;
>> a(':')=-1

a =

      -1    -1    -1    -1    -1

[Jeffrey Meunier]

Got it. Thanks so much for your quick reply!