[antlr-interest] Composite grammars - potential improvement

[Rampon Jerome]

Hi,


As a potential future incremental compilation speed-up (for ANTLR project developments)


I noticed the following with a large grammar

I do have split version with composite grammar
and not splitted one :

The generation of output JAVA Lexer is similar if antlr is invoked twice
on whatever case.


For the generation of output JAVA parser,
there are some differences in the order of delegated rules listing for the
composite version. It does not impact the results but
a kind of issue regarding incremental compilation and generation of an ANTLR project.


Probably some hash output listing which is not sorted alphabetically ...

For non composite version, only versionning and date of ANTLR differs
I use it with some diff <(tail -n +2 ANTLR output) ... to avoid over-compilation
in project builds


Not working with composite parsers ...


Jerome