|IBM Parser Generator email@example.com (Olivier Lefevre) (1998-12-22)|
|From:||Olivier Lefevre <firstname.lastname@example.org>|
|Date:||22 Dec 1998 04:40:29 -0500|
|Organization:||Warburg Dillon Read|
The IBM Jikes (a Java compiler) team uses its own parser generator.
Work on the parser generator started in 1983. It
implements [Philppe] Charles's Ph.D. Dissertation on Automatic
Error Recovery and also includes several additional
compression algorithms he developed later. It has been
used inside IBM by various groups for over a decade.
One little-known feature of Jikes is that we maintain a
COMPLETE program representation -- every token (with
line and column numbers), every comment, and the AST
(abstract syntax tree) reflects the program exactly as
written. You can determine if a cast was explicitly written
or if it was required by language semantics. You can even
recover parentheses written in expressions, etc.
Apart from that it seems to be a traditional LALR(k) pg. Based
on that (probably incomplete) information, I'd be curious to know
whether they are doing anything that the other parsers out there
can't do. I guess the compression is likely to be the more original
Return to the
Search the comp.compilers archives again.