[antlr-interest] Scanning whether a dot is part of a keyword or of a namespace.

[Niek Sombekke]

Hi all,

The problem I have can be described as follows. I'm trying to write a lexer
to scan a .NET IL file. A number of keywords in IL can contain a dot.
E.gboth next instructions are valid.

  tokens
  {
    DOT_ASSEMBLY = ".assembly" ;
    ASSEMBLY = "assembly";
  }

On the other hand, it is possible to have a namespace System.Object, which
after scanning should deliver the tokentypes
    System - ID
    . - DOT
    Object - ID


While testing a bogus line, e.g. ".assembly System.Object", my scanner
delivers (depending on my grammar) either the tokens
    .assembly - DOT_ASSEMBLY
    System - ID
    .Object - ID
or
    . - DOT
    assembly - ASSEMBLY
    System - ID
    . - DOT
    Object - ID

Obviously, both are incorrect! In the first case, the '.' should be
seperated from 'Object'. In the second case, the '.' should be combined with
'assembly' making it a DOT_ASSEMBLY token.

How can I let ANTLR determine whether the dot is part of a keyword, or that
it seperates two identifiers within a namespace?


-- Niek
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.antlr.org/pipermail/antlr-interest/attachments/20060914/5ffb2268/attachment-0001.html