Am Monday 17 August 2009 schrieb Matthew Woehlke:
> Thomas Lübking wrote:
> > Performance ain't crucial here (and QChar::isLetter() just tests some
> > flags), but i just tested and QChar grants ::isLetter to really a lot of
> > stuff (esp. as any utf-8 char seems to be split an the first part becomes
> > a letter, though it's /no/ [A-Za-z] :-(
> 
> Ah... doesn't that make QString::mid broken? (Or would that be QChar?)
> Though that shouldn't be a problem if e.g. "駄" gets translated to "G_q_".
 
"駄" (except the """s") has QString::length() == 3
all are represented by some weird "?" in a diamond, no substring matches the 
identifier regexp
駄[0].isLetter() == true
駄[1].isLetter() == false
駄[2].isLetter() == false

"ü" has QString::length() == 2
all are represented by some weird "?" in a diamond, no substring matches the
ü[0].isLetter() == true
ü[1].isLetter() == false

I don't know whether this makes QString or QChar "broken"* but the suggested 
solution guarantees a valid dbus path -regarding the object name- which in 
doubt ends up like "_________" ;-P

Thomas

*i'd preferably force the whole world to restrict to ASCII - that's two more 
letters than the romans needed to control an empire for a millenium >-)