RE: C++ lexer (GCC 3.1.1) requires knowledge of other C dialects

["Gary Funck" <gary at intrepid dot com>]

> -----Original Message-----
> From: Robert Dewar [mailto:dewar@gnat.com]
> Sent: Thursday, August 01, 2002 3:21 AM
> To: gary@Intrepid.Com; per@bothner.com
> Cc: gcc@gcc.gnu.org
> Subject: Re: C++ lexer (GCC 3.1.1) requires knowledge of other C
> dialects
>
>
> <<There are tricks to solve most of these problems, but a
> recursive-descent parser is just so much easier to understand, to debug,
> and to handle special quirks that I much prefer it.  Yes, the actual
> grammar isn't as cleanly separated, and the source is slightly more
> verbose, but I much prefer it.
> >>
>
> When we use the phrase "recursive descent" what we mean in practice is
> not a straight LL(1) parser. If we wanted a straight LL(1) parser, then
> we could use an appropriate parser generator (though why would we do this
> when LRK PG's are clearly preferable).
>
> The point about recursive descent is that it is hand written code that can
> use any techniques it likes to parse the source. It can do a bit of backup
> or extra look ahead if it wants, it can keep information on the side and
> roam around the tree etc. This is particularly use for two purposes:
[...]

Although this is a bit of philosophical debate (regarding the merits of
handbuilt vs. program-generated parsers), I'll point out that the ANTLR
parser generator (http://www.antlr.org) generates LL(k) recursive descent
parsers with additional features which include: (1) syntactic and semantic
predicates which allow for arbitrary checks before an alternative is taken,
(2) approximate linear lookahead which accepts a larger class of LL(k)
grammars, (3) automated error recovery, and support manual, arbitrary error
recovery, (4) parsers built in C, C++, and Java, (5) builtin support for
building AST's (Abstract Syntax Trees).