[darcs-users] [patch318] tune the patch parser (and 10 more)

[Ganesh Sittampalam]

I'll address one specific point quickly, as I want to take time to reply 
to the rest properly.

On Wed, 4 Aug 2010, Jason Dagit wrote:

> So, if I implement the above space leak fix, do you plan to accept the 
> patches?  I ask because if you are not planning to accept them I won't 
> bother with the fix :)

I think there are two issues to consider here: whether your use of try is 
ok, and whether it's a good idea to expose try in our parser code at all, 
for fear that someone will come along later and make seemingly innocuous 
use of it causing a serious problem.

IMO the latter point is one that should be discussed a bit more widely. My 
general impression - and I'm by no means an expert - about the existence 
of try in parser libraries is that it can be quite problematic. Here it 
would be changing the interface of the parsing library away from one which 
pretty much enforces a single pass scan. On the other hand, the benefits 
in compositionality are nice, as we can see from your patches. So I'm 
torn, but I think other people should get involved in the decision, and 
that we should perhaps investigate what parser libraries out there offer 
it.

Ganesh