[antlr-interest] Re: ANTLR+Eclipse

[Thomas Brandon]

The Antlr eclipse plugin implements such functionality. It does 
syntax highlighting and produces an outline of the grammar. I'm not 
100% familiar with the code but it appears to be using the builtin 
Eclipse support to do the syntax highlighting and an Antlr parser to 
do the outline view. The reconciling strategy handles document 
changes by simply reparsing the whole document upon any change. This 
also means the whole source model must be discarded each parse. 
There does not appear to be any support for tracking changes at a a 
lower-level which would become more critical if other, more 
complicated, tools are consuming the source model. I'm not sure how 
good the performance would be if this method was used for lower-
level features such as syntax highlighting or code completion which 
require more frequent updates. The Netbeans project implemented 
incremental lexing on top of an Antlr based lexer 
(lexer.netbeans.org), perhaps you could port something like that to 
Eclipse. Incremental parsing in Antlr seems to be a more difficult 
prospect (though Antlr 3 may make this easier). Batch parsing on top 
of incremental lexing may be good enough for code-completion and the 
like.

Tom.
--- In antlr-interest at yahoogroups.com, "Tiller, Michael \(M.M.\)" 
<mtiller at f...> wrote:
> I was wondering if anybody had ever tried to use ANTLR to 
implement a
> scanner in Eclipse.  I'm not talking about a plug-in to Eclipse to
> facilitate grammar development, I'm talking about the reverse.
> Specifically, a tool that can scan the document associated with an
> editor in Eclipse and pull out information about the locations of
> keywords, comments, etc. and then integrated this into the Eclipse
> framework for identifying tokens, doing completion, etc.
> 
> It seems like all that would be required would be to interface the 
lexer
> to the Eclipse representation of a document (so you could scan the
> characters in the document) and then somehow tuck the results away 
so
> that subsequent calls to specialized token scanners, etc. wouldn't 
even
> look at the document but would instead just return information 
based on
> walking the AST.
> 
> So, has anybody done this?  It seems like this would be a great
> cornerstone for plug-in development because you could do real 
parsing
> rather than relying on the (what appear to be) rather crude 
primitives
> in Eclipse for doing this kind of thing.
> 
> --
> Mike



 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/antlr-interest/

<*> To unsubscribe from this group, send an email to:
    antlr-interest-unsubscribe at yahoogroups.com

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/