Stephan Assmus <superstippi@xxxxxx> wrote: > On 2010-05-27 at 21:50:59 [+0200], Karsten Heimrich <host.haiku@xxxxxx> > > wrote: > > (1) I wonder if we shouldn't use const BString& string while > > introducing > > new API's? I know all the old API's use const char*, but memory an > > speed > > are not really a constraint here, no? > I am usually in favor of that since you introduced the COW BString > backend. > However in this case, these methods can actually unset the members by > passing > NULL. Granted, it's not different from passing "" or an empty > BString. I can > definitely change it, it wouldn't make a difference besides avoiding > string > copies. Since the target is in fact a BString, it cannot really be more expensive than passing a const char* - but it can be cheaper. But I think it's hard to find a consistent approach to this, since sometimes it's a lot cheaper to pass in const char* - maybe we should always just have both versions in our future APIs. Bye, Axel.