Hi all,
for code completion I need a way to force a lexer to match somehow even if it
is not complete. Sounds weird, I know. What I have in mind is this: Imagine you
have defined special identifiers which can be quoted (double quote char). When
the user has started typing something like this:
select "abc
and now invokes code completion I want to show all possible identifiers. But
with a lexer rule like this:
ANSI_QUOTED_ID:
DOUBLE_QUOTE
( options {greedy=false;} :
ESCAPE_SEQUENCE
| .
)* DOUBLE_QUOTE
;
I only get an error node plus a not so helpful error message:
1 errors found
-end of input-(1) : error 3 : , at offset 0, at <EOF> : cannot match to any
predicted input...
The AST looks like:
(1, 0, nil) nil
(1, 0, SELECT_SYM55 [418]) select
(0, 0, <invalid> [0]) Tree Error Node
(1, 0, (null) [-1]) <EOF>
However, what I would like is to get ANSI_QUOTED_ID to match the entire rest of
the input and return this as the ID (plus the error, so I know it was
incorrect). This way I know we are looking for a (possibly quoted) identifier
and, also important, get the text written already. Using EOF or an empty
alternative for the closing DOUBLE_QUOTE gives me various ambiguities, so this
is not a solution either.
Scanning the text in the editor backwards is another option, but requires more
processing so I would prefer a (simple) solution in the grammar, if possible.
Mike
--
www.soft-gems.net
_______________________________________________
antlr-dev mailing list
[email protected]
http://www.antlr.org/mailman/listinfo/antlr-dev
