[antlr-interest] lexical nondeterminism

[None None]

This might be in the FAQ, but I haven't seen it.  A
few months ago I started to look at antlr and had a
lexical problem.  I haven't had a chance to look at
again until today, but when I tried some of the
suggestions they didn't work.  If anyone could point
me in the correct direction it would be much
appreciated.

I have a file called dms.g that contains the
following:

class TestLexer extends Lexer;

options {
          k=10;
        }

protected LATLON : DMS "," DMS | FLOAT "," FLOAT;
protected DMS    : DIGIT DIGIT (DIGIT)? ":" DIGIT
DIGIT ":" DIGIT DIGIT "." DIGIT;
protected FLOAT  : (SIGN)? (DIGIT)+ ( "." (DIGIT)* )?;
protected SIGN   : "+" | "-";
protected DIGIT  : '0'..'9';


When I run "antlr dms.g" I get the following output

ANTLR Parser Generator   Version 2.7.5 (20050615)  
1989-2005 jGuru.com
dms.g:7: warning:lexical nondeterminism between alts 1
and 2 of block upon
dms.g:7:     k==1:'0'..'9'
dms.g:7:     k==2:'0'..'9'
dms.g:7:     k==3:'0'..'9'
dms.g:7:     k==4:'0'..'9'
dms.g:7:     k==5:'0'..'9'
dms.g:7:     k==6:'0'..'9'
dms.g:7:     k==7:'0'..'9'
dms.g:7:     k==8:'0'..'9'
dms.g:7:     k==9:'.','0'..'9'
dms.g:7:     k==10:'.','0'..'9'


It seems to me that by the 3rd or 4th token the ":"
the lexer should be able to tell which production (DMS
or FLOAT) was being constructed.  The thing puzzles me
is that if the DMS rule is changed to:

protected DMS    : DIGIT DIGIT DIGIT ":" DIGIT DIGIT
":" DIGIT DIGIT "." DIGIT;

antlr is happy and doesn't issue any errors!

Thanks in advance.

 


	
		
______________________________________________________ 
Yahoo! for Good 
Donate to the Hurricane Katrina relief effort. 
http://store.yahoo.com/redcross-donate3/