wcwidth and terminals [Re: Mg3a - a version of Mg2a developed on Cygwin]

[Thomas Wolff]

Am 24.02.2011 09:56, schrieb Corinna Vinschen:
> When on Cygwin, you might better use Cygwin's(*) wcwidth function.  It's
> based on the same code from Markus Kuhn, but it interacts with the
> setlocale function to make sure that the width returned for the CJK
> ambiguous width characters makes sense in the given locale.  Plus, it
> supports a Cygwin-specific locale modifier '@cjknarrow' which allows the
> user to modify this behaviour.  When using your own wcwidth, you're
> giving up on this feature.
On the other hand, not specific for cygwin, the wcwidth/wcswidht 
functions are in general based on installed locale data.
But you can never know whether the terminal you are running in actually 
uses those data.
On a Linux system, you may encounter a wide range of different 
environments, for example:
rxvt based on system locale which is often based on an outdated locale 
data installation (e.g. Unicode 3)
xterm with a hard-coded wcwidth
mlterm with some tweaks
remote operation (rlogin/ssh to/from other system) with inconsistent 
locale data base
...

For that reason, my editor mined implements an auto-detection of actual 
terminal width data (checking Unicode version and a number of 
terminal-specific odds).

For cygwin, it might be useful (although not standard) for wcwidth to 
consider whether it's running in a cygwin console or a terminal, so e.g. 
wcwidth (0x8080) should return 2 in mintty but 1 in a cygwin console.

Erik Blake wrote:
> And if you don't mind [L]GPL licensing, gnulib provides a source code
> replacement that guarantees wide character support on all modern porting
> platforms (particularly useful for mingw, which is sorely lacking on
> this front); and is currently working on introducing a wwchar_t type
> that is guaranteed to be UTF-32 even on cygwin (this is how coreutils
> gets wide character support for things like wc).  Portions of that
> gnulib code are incorporated into libunistring.  But from the sounds of
> your program's license, I'm not sure you can take advantage of gnulib or
> libunistring.
Maybe I should check whether my auto-detection code could be contributed 
to that project...

Thomas

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple