Re: wfstream::get() function and multibyte encoding

[Paolo Carlini <pcarlini at suse dot de>]

Hi,

Joseph D. Wagner wrote:

> The get() function on a wfstream object, at least the way I am using it, still fetches only 1 byte.  I would have though get() on wfstream would fetch sizeof(wchar_t).  I am aware that different overloaded versions of get() can fetch multiple characters, but is there a switch, setting, or option to tell the get() function to fetch sizeof(wchar_t) instead of just a single byte?
What actually happens when get() is invoked depends on many different 
factors and, definitely, you cannot say in general that it fetches a 
given number of bytes. Neither there is a switch, option or setting to 
this effect.

In general, if you are using an unbuffered stream synchronized with 
stdio (the default for cin, cout, etc...) each individual get boils down 
to the system call getwc, whose behavior depends on the global locale 
(as set by setlocale). If, on the other hand, you are using a wfstream, 
then the conversion machinery prescribed by the Standard, and based on 
the codecvt facet, is in effect and you can change it via imbue, 
basically. In this case, I/O actually happens via a buffer which is 
converted from/to the external representation by the codecvt facet.

Remember that for all those options to work well and correctly wrt the 
Standard you need gcc3.4.0.

Paolo.