[antlr-interest] Lexer and Parser class packaging

[Martin Probst]

> Having the "anonymous" @header automatically map to one or the  
> other (parser, in the current form) seems counter-intuitive, while  
> only having the explicit specifications gives the system a more  
> "symmetric" feel.

It's indeed counter intuitive.

If I say "At a station, you'll usually find a bus stop", it's usually  
taken that I imply all sorts of stations that make sense in the  
context, i.e. train stations, tramway stations, subway stations, etc.  
If I want to limit the statement to "subway stations", then I'll  
specify that by being explicit. This is simply a convention in human  
language, and it's also logical.

I think ANTLR should follow that and put stuff that shows up in  
"@header {}" in both the lexer and parser, and it should do the same  
for "@member {}". While of course putting something in "@member {}"  
without a qualifier is nearly always an error, it seems a lot more  
sensible to me to require explicit specification if you want uh,  
explicit specification ;-)

@parser::header ==> to parser
@lexer::header ==> to lexer
@header ==> to both

All @ statements without a qualification should go into all generated  
products. It's also not obvious why the Parser should take such  
precedence over a lexer or over any tree parser.

Regards,
Martin