[antlr-interest] Parsing complex languages like C++

[Paul Johnson]

software.au at gmail.com wrote:
> Does anyone know how to parse complex languages like C++?  Can it only
> be done with significant hacks to an LALR or LR parser?  Can it be
> done at all recursive descent?  Any type of algorithm that works for
> this language, and other complex/difficult to parse languages?

g++ is still, I think, LALR(1). They switched from Bison to a hand-coded 
recursive-descent parser sometime around 3.3, and it made a vast 
improvement in the language conformance.

I can't immediately see that using LR makes it intrinsically difficult 
to parse C++, but I haven't tried it. The gcc team did blame a lot of 
bugs on Bison, but I think that's probably just down to Bison's 
shortcomings. In particular, I think a much tighter coupling between 
Bison and Flex would make it a lot more useful and generally applicable.

Bison has been GLR for a couple of years at least; it supports unlimited 
(?) lookahead.

Paul