Re: About Linguistic2::getHyphenationPositions()

On 01.06.2015 09:45, Stephan Bergmann wrote:
> On 06/01/2015 01:38 AM, Matteo Casalin wrote:
> >      while converting some sal_uInt16 to sal_Int32 in
> > cui/source/dialogs/hyphen.cxx, I noticed that getHyphenationPositions()
> > returns a sequence of short/sal_Int16. The surrounding code suggests
> > that sal_Int32 would be more appropriate, but I see that this function
> > is listed in some .idl files and I can't say if changing it would break
> > some published API.
>
> Looking at cui/source/dialogs/hyphen.cxx, you probably mean method 
> getHyphenationPositions of UNO interface 
> css.linguistic2.XPossibleHyphens.  Which is indeed published, so cannot 
> be changed, for backwards compatibility.
>
> So if there is really convincing reason to change this to e.g. 
> sequence<long>, either introduce a css.linguistic2.XPossibleHyphens2 
> etc. or discuss what the implications would be of incompatibly changing 
> css.linguistic.XPossibleHyphens.
>
> Otherwise, best document that getHyphenationPositions unfortunately has 
> a poor return type for historic reasons, and ensure that its 
> implementations do reasonable things (e.g., do not include too large 
> positions in the returned sequence but emit a SAL_WARN instead).

i don't see much need to change this; ideally it should use "long" but
the interface is for hyphenation of a single *word* - those are usually
quite short.  so the interface doesn't have to "work" for a 64k "word",
it's enough if it doesn't crash.