[antlr-interest] Hand made lexer 600(!!!!!) times faster
Ian Kaplan
iank at bearcave.com
Thu Feb 24 13:47:45 PST 2005
Summary: writing a scanner is not a big deal
ANTLR has a couple of core application areas. One is a way to
quickly bring up a scanner and parser for little languages or simple
parsing tasks. In many of these performance is not a huge
consideration.
Another use is for production quality parsers for demanding
applications and for applications where performance is a concern.
In these cases I don't see why it is a big deal to write your own
scanner. I just implemented a custom scanner for an ANTLR based
query laguage parser. It took a couple of days to do the scanner
and related data structures (reserved word tables and support data
structures).
I know I keep writing this and you're all probably tired of reading
it. But as far as I'm concerned the power of ANTLR is as a parser
generator. I can bring up a parser much faster with ANTLR and the
parser is more reliable. I can also change the parser more
rapidly. The scanner generation is a nice feature, but I don't feel
that it is core to why I use ANTLR. I feel the same way about tree
generation (I do my own tree generation too).
Ian
More information about the antlr-interest
mailing list