[antlr-interest] can tree grammers have an output of AST?
Terence Parr
parrt at cs.usfca.edu
Mon Oct 16 10:13:30 PDT 2006
On Oct 15, 2006, at 11:30 PM, Hill, Robert wrote:
>> Ter's latest statement on translation philosophy is that trees
>> are built during parsing and then walked one or more times
>> without transformation (that started the latest thread on
>> translation approaches, after all).
>
> Hmm, I was (trying) to follow the thread , I must have missed that bit
> :(
Tree transformation is great if you are going from language L to L. I
think that managing Union tree grammars when moving from L to L' is
too much of a hassle. But, that is just my opinion.
> All I wanted to do was make successive passes only deal with the rules
> that are
> Relevent at that pass, so for instance if I optimise away unused
> paramters, I wouldn't have to deal with them on the next pass, instead
> of simply setUnused(true)..
An appropriate transformation.
>> ANTLR 3 was not designed
>> to do complex multi-pass transformations, nor was it designed
>> _not_ to support such transformations, for that matter.
>> However, features usually don't "just happen" in software, but
>> instead are the result of explicit design or redesign efforts.
>
> My misunderstanding, I just assumed that you could output either
> another
> AST or a template from the treeparser. Just seemed natural to me, I
> guess I was jumping the gun a bit ;)
Yes, it is my intention that you can output trees. I had to cut
that out of the first release so I could do a feature freeze and
focus on bugs. You will note that the wiki has:
http://www.antlr.org/wiki/display/ANTLR3/support+output%3DAST+in+tree
+grammars
from July 27.
Ter
More information about the antlr-interest
mailing list