[Monotone-devel] utf8 vs. string (for args and other stuff)

[Julio M. Merino Vidal]

Hi,

As part of the help-rewrite branch, I'm modifying the command class  
and the related functions (defined in commands.hh).  While I already  
have most of the "internal" work done (a rewrite of the class and  
algorithms), I'm now finding problems when integrating those changes  
into the public functions because of string/utf8 inconsistencies.

I see that some places in the code define the user's arguments (argv)  
as a vector<utf8>, whereas other places use vector<string>.   
Conversions between the two happen in several places, as if they were  
always convertible.  Based on:

http://www.venge.net/mtn-wiki/I18nL10n

I assume the arguments fall into the first category, "user's display/ 
entry charset".  What's the correct way to represent them?  string?   
Should those vector<utf8> args instances be vector<string> instead?   
If not, why are there two different representations?

At last, consider the commands class defined in cmd.hh.  All its  
fields are strings at the moment.  If I understood it correctly, it's  
ok for desc() to return a string because it will be in the user's  
charset; however, could desc_ be stored internally as utf8 (because  
its contents come from monotone's untranslated source code)?  Which  
could be the correct type for, e.g., the name field?  utf8?

Thank you.

--
Julio M. Merino Vidal <address@hidden>