Gottlob Frege wrote:

On 1/22/07, *Tobias Schwinger* <tschwinger@isonews2.com <mailto:tschwinger@isonews2.com>> wrote:

John Christopher wrote: > Hello, > Does Boost contain facilities to convert std::string into std::wstring and > the other way around?

Should there?

std::string a; std::wstring b( a.begin(),a.end()); std::string c(b.begin(),b.end());

All string algorithms in Boost are templates so they should work fine with both std::string and std::wstring.

http://www.boost.org/doc/html/string_algo.html

Regards, Tobias

I think the real question is whether the OP was hoping for Unicode conversions of some kind in the process. ie what 'type' (sematically) are the wchar_t's and char's? UTF16? UTF32? (what size is the wchar_t?) etc.

I'm not sure that's what the OP really wanted: If we want to convert back and forth we are probably just using the (compatible) ASCII subset (a typical example: COM calls). A proper conversion to UTF-8 when narrowing the character type can be worse than overflow if the rest of the code can't handle UTF-8. However, Boost.Unicode sure would be a nice addition.

Also, does the last line: std::string c(b.begin(), b.end()) give at least a compiler warning about downconverting from wchar_t to char, and thus losing bits?

Diagnostic messages are implementation-defined. The code I posted does not warn with gcc4 (all warnings enabled), for instance.

(Which would be a hint that maybe you need some Unicode conversion or something else.)

Yeah, like most compiler warnings, it means "know what you're doing" ;-). Regards, Tobias

John Christopher wrote:

Hi,

...

...

std::string a; std::wstring b( a.begin(),a.end()); std::string c(b.begin(),b.end());

for a while now, I have started wondering if I was not shooting myself in

The reason I asked is that while I have been using the foot without knowing... I am no Unicode expert and Boost facilities in that area would certainly give me some confidence. Boost is nice also because codes are visited by skilled and experienced programmers and I am glad to benefit from that... :) More generally, I am still struggling on which string class I should use: std::string, std::wstring, some custom class...

Not claiming to be a expert in this field, I typedef std::wstring string; for my project if I need Unicode. I expect the iterators of some Unicode string class to (at least) give me wide characters, regardless of the underlying in-memory encoding. The 'typedef' (placed in some global header) allows me to easily change the string class, if I have to (e.g. for saving memory or bandwidth). std::string is basically just a container for bytes, that is length() will start lying as soon as there are characters that are represented by more than one byte (like with UTF-8). Regards, Tobias