Why does strcmp return 0 in in the matlab script while 1 on consol

Hi,

I wonder what mistake I am doing!

When I run the following code on console I get correct output (flag=1) while the same code evaluates to (flag=0) when run in matlab script file. I am trying with Hindi unicode characters. During debug watch I get same values as expected in the variable but evaluation is not as expected. I tried with isequal() function also. but it results same behavior.

-------------

previous='ग';

character='ा';

if strcmp(previous,'ग') && strcmp(character,'ा')

     flag=1

else

     flag=0

end

-------------------------

Answer 1

[Walter Roberson]

In R2014a on OS-X, I find that there is a difference between these two lines:

previous='ग';      %works
previous='ग'       %assigns char(26)

Experimenting, it appears that any literal character with char position beyond 255 gets converted to char(26) if it appears in any context other than an assignment terminated with a semi-colon.

This is, of course, a bug. I have no information about which versions it affects.

One thing you need to be very careful with is that when you Save a .m file and you are using the English version of MATLAB (at least up through R2014a in OS-X), then the file is saved as 8-bit characters not as UTF-8, so any literal character beyond char(255) will become a sequence of characters rather than a single character. For example, char(300) would become the pair [char(1), char(44)] in the file. I gather that the non-English versions of MATLAB (or possibly later versions of MATLAB) offer an option to save the source file as UTF-8; my version does not offer that.

[Rajesh M]

but how do I compare two characters from Hindi? Its strange that the comparison works on console but not in .m file. Is there any way to come out of it?

Also I tried using double. It is more strange. When write on the consol I get

    char_1=double('रै')

    char_2=double('ग')

    char_1 =

        2352        2376

    char_2 =

        2327

But the same in the .m file evaluates to

char1 = 63 63
char2 = 63

Why this behavior? an why 63 always. I can understand two components but value 63 is not solicited. I should get 2352 2376 for 'रै' and 2327 for 'ग' respectively.

My basic problem is to compare two strings of Hindi.

statements like

if current_char==char1 && previous==char2 results in error asking for scalar operands for it.

[Walter Roberson]

Compare their unicode positions. For example, 'ग' is 2327 so

if strcmp(previous,char(2327));disp('matches');end

You would of course assign the symbols to variables like

char_GA = char(2327);

if strcmp(previous, char_GA); disp('matches'); end

[Walter Roberson]

In OS-X R2014a, at the console, using the English version of MATLAB,

>> char_1=double('रै')

char_1 =

      26    26

The problem affects characters in quotes at the console in any context except

variable='string';

The 63 that you see in the file happens when you save the file without telling it to save it as UTF-8 (which is not an option on R2014a OS-X English version. If you close the file and open it again you will see that all of the Hindi characters have been replaced by '?'.

The problem you have is not with comparing Hindi strings that are already in variables: the problem is with getting particular characters into code. And the way to do that in a portable manner is to never use the quoted characters at all, and to instead use their unicode positions. For example, instead of coding

strcmp(previous, 'मतलब')

code

h_TA = char(2340);
h_BA = char(2348);
h_MA = char(2350);
h_LA = char(2354);

h_MATALABA = [h_MA h_TA h_LA h_BA];

if strcmp(previous, hMATALABA)

[Rajesh M]

Thanks