[darcs-users] neat paper

[Kenneth Knowles]

On Mon, Mar 29, 2004 at 01:18:31AM +0100, Ganesh Sittampalam wrote:
> 
> > [Adam Megacz]
> > > Imagine parse-tree level source code diffing... ;)
> 
> > It wouldn't be such a daunting task with a decent plugin architecture
> > for different languages.  (Especially for languages that have
> > open-sourced licenced grammars)  Anyone know of research/hacking on this
> > subject?  (Ignore this question if the paper answers it - I'll get to
> > that today)
> 
> The Intentional Programming project at Microsoft some years back had a
> syntax-tree based VC system, but it'd dead now, sadly. I can't think of
> anyone else that's done it off the top of my head, but it's not all that
> hard. The biggest issue is how to deal with incomplete code - i.e. code
> that wouldn't actually parse validly - some people want to check that kind
> of thing into VC. There's work on incremental parsing for editors that is
> likely to be applicable, though.

Well, falling back to text-only mode when something fails to parse wouldn't be
the end of the world, but admittedly solving that would be better for research.
Modifying the grammar so that error tokens could carry their verbatim contents,
and having those text-diffed seems like it might work.  I'm sure the researchers
have thought of this and more...

Kenn