Re: [Swig-devel] Java director:except status - problem with fragments

[William S F. <ws...@fu...>]

On 11/03/13 02:30, marvin persona wrote:
> I've followed the approaches you suggested, and have the changes to swig
> in place, but have not yet updated the tests, which should not take
> long.  But while updating the documentation I realized there are issues
> with the fragment approach.  Read documentation excerpt:
>
>  > To handle such exceptions correctly, it is necessary to translate
>  > them into C++ exceptions. This can be done in two different ways
>  > using the  %feature("director:except") directive.  In the simplest
>  > approach, simply attach a code block to a method to handle the
>  >mapping of java exceptions into C++ exceptions.
>
>  >// All the rules to associate a feature with an element apply
>  > %feature("director:except",fullname=1) MyClass::method(int x) {
>  >   jthrowable $error = jenv->ExceptionOccurred();
>  >   if ($error) {
>  >   if  (Swig::ExceptionMatches(...))
>  >     throw logic_error(Swig::GetExceptionMessage(...).message());
>  > ...
>
> But %features do not provide any way to signal that they require
> fragments, fragments are only for %typemaps.  So ExceptionMatches and
> GetExceptionMessage cannot be used.
>
Yes what a nuisance, that is bit of a gap in functionality with the 
whole %fragment design.

> Since your main concern (introduction of std::string dependency) is
> resolved, it seems like using methods that are always provided by
> director.swg might be fine.  An alternative that would be more work and
> may or may not make sense, would be to allow %features to specify
> fragments attributes in the same way as typemaps.
>
> I'd like to go ahead and just move the methods in to java/director.swg
> for now.  If you decide that adding fragment support to %feature is
> desirable, that can be done as a separate task.
Yes agreed. I might just look at implementing that.

William