Convert Matlab string to wchar in C-Mex under Windows and Linux

I want to convert a Matlab string to a wchar string, such that I can use e.g. either the C-function _wopen? Or to a LPCWSTR under Windows?

Matlab's CHAR arrays use 2 bytes per element. Inside a C-Mex function I can use mxArrayToString to convert a Matlab CHAR array to a C char * vector. The documentation of mxArrayToString state, that "it supports multibyte character sets". Unfortunately this is not explained with any further details.

Currently under Windows I copy the contents of the mxChar array to a wchar_t vector and append a L'\0'. It seems to work, but I'm not really convinced that this is stable. Under Linux a wchar has a length of 4 bytes and the direct copy is not sufficient.

Are Matlab CHARs UTF16 encoded? How does mxArrayToString support multibyte charcter sets exactly? How do I convert mxChar to wchar correctly and efficiently under Windows and Linux?

[Kaustubha Govind]

Jan: If you don't receive a reply here, Tech Support might be able to give you an answer.

[Jan]

@Kaustubha: Thanks. There is a 2nd level question, which can be answered here only: How many people use Unicode strings already? Matlab decided for 2 byte CHARs a long time ago, but if nobody is able to use this correctly, this feature is simply a waste of memory. The CSSM thread http://www.mathworks.com/matlabcentral/newsreader/view_thread/301249 was a revealation: There are some undocumented functions. But the marvelous idea not to use 1-byte CHARs should be accompanied by a documented and stable interface.

Answer 1

[Alexander H]

Jan, thanks for your post.

Recently, I got entangled in passing Chinese word (non-latin character) from Matlab to mex. I tried wchar_t, mxChar, none of them works. While, mxArrayToString you mentioned works exactly as I want. Since I just want my mex function can deal with Chinese file path , so I didn't copy it to somewhere else. The code is simply as:

char *filename = mxArrayToString(prhs[0]);