> Since the << operator is already used in BString, the suggested > BStringBuilder API cannot be used here, but I would personally prefer > methods over the << operator anyway. > Other than that, I think it makes sense to have two different APIs for > that, even though I wouldn't mind if both of them end up in BString > eventually. BStringBuilder would be a different class, like ostringstream is in plain C++. > > Other than that, as Ingo pointed out on the original thread > (//www.freelists.org/post/haiku/Proposals-for-the-LocaleKit, > 44), we still need a dedicated solution for numbering. > It might be a good idea to look at the solutions in Qt, and Java for > this. The latter uses a choice formatter that the translator specified > similar to what Ingo suggested: > {1,choice,0#zero,1#one,2#many} > See > http://download.oracle.com/javase/1.4.2/docs/api/java/text/MessageFormat.html ... which uses ICU as a backend. Something we may use too :) (IIRC somewhere around hte PluralFormat class.)