[antlr-interest] Antlr vs Spirit

[Stuart Dootson]

I've used both Spirit and Antlr in anger...

My feeling is that I'd prefer to use Spirit for smaller (rather than
larger) parsing applications (for example, parsing a log file rather
than a source file). This is mainly because a) compile times can get
quite large, and b) I always have this nagging feeling in the back of
my mind that my compiler will raise an internal compiler error if I
add too many more grammar rules :-) (this is mainly because I've used
Visual C++ 6 too much - I use VC++ 7.1 now, which is a lot better in
this respect). Another point is that you need to be using a relatively
modern compiler (GCC 3.3+, VC7.1+, something like that).

The main problem I have with Spirit is that there is no separate lexer
- this complicates grammar design as you have to think about
tokenisation in the same space as the grammar.

The main features I like about Spirit are closure support, ease of
integration with C++ projects and extensibility.

BTW - Spirit *does* support trees - tree construction at least - see
http://www.boost.org/libs/spirit/doc/trees.html.

So, in conclusion - I may use Spirit for small parsing tasks, but I'll
stick with Antlr for larger tasks. If I had to implement an Ada parser
(for example - it's something I've done before), I'd break out Antlr.
Also - if I were advising a parsing newbie (not theoretical - I'm
doing that at the moment), I'd get them to start with Antlr. Spirit
requires more background knowledge of things not related to parsing
than Antlr.

Stuart Dootson