matlab recognise french character
20 views (last 30 days)
Hervé on 14 Mar 2014
The problem comes from the fact that MS-Windows “DOS-shell” ("cmd.exe" window) does not use the same character encoding than MS-Windows and next than the 16 bits Matlab char type. The former uses (at least up to Windows7; I do not already check with Windows8) old DOS "code page" encoding; Windows use 8bits ISO-LATIN-1 (ISO 8859-1) and Matlab 16 bits char type uses "Unicode" character encoding (from the ~6.5 or ~7.0 version; before, Matlab used the native 8bits Windows encoding in a 8bits char type...).
It seems that the " dos " (or " ! ") Matlab command forgets to apply correct transcoding to the returned or displayed string. There is now a simple workaround:
b=unicode2native(task); % transcodes back to original 8bits DOS code
msgOK=native2unicode(b,'IBM850') % does the *correct* transcoding
'IBM850' stands for “Code Page 850”, which is the code page for French Windows (and old French DOS). For other Windows localization, the DOS command " CHCP " allows to retrieve the current code page (from within Matlab, type « !CHCP » in the Command Window with the exclamation mark in front).
It is important to note that this transcoding problem seems to only exist for the returned or displayed string, but not for the input string. If filenames or usernames contain accented character, you can directly type the string in the Matlab editor without the need to manually transcode it. This is illustrated in the following example that displays the list of the current directory files whose names contain an e_acute character "é":
[s,msgWrong]=dos('dir *é*.*'); % NO need to transcode "é" (e acute) in the input
msgOK=native2unicode(b,'IBM850') % but we need to correctly transcode the answer...
All of this sounds like a Matlab bug (that is there from many years). As I also recurrently face this problem, I have recently submitted a bug report to TMW in the hope that next Matlab release will fix this output string character transcoding error to avoid the need for the two lines workaround (which, however, cannot be used for the "!" syntax)...
More Answers (3)
Hervé on 16 Jan 2016
Note that Release R2015b of Matlab natively fixes this bug for "dos" and "!". We thus don't need this trick any more. (More over applying this method in (≥)R2015b would now give bad results). For information.
José-Luis on 16 Jan 2014
What character encoding are you using? Might be that French is not supported for it. You could always change it.