How can I set an ANTLR token value from an input parameter? - antlr4

Question

Suppose that I have a very simple grammar defined in ANTLR 4:
input : String Separator String ;
String : 'a'..'z' ;
Separator : ',' ;
For this grammar, the separator is fixed; it will always be a comma. Is there a way to make the separator variable? That is, I'd like to define the separator using an input parameter, which is set by the code that calls the lexer-parser. I can define a getter and setter like so:
#lexer::members
{
String sep = ",";
public void setSep(String sep)
{
this.sep = sep;
}
private String getSep()
{
return sep;
}
}
But how do I change the value of the separator in the lexer rule? This is close, but wrong:
Separator : ',' { setText(getSep()); } ;

Answer 1

After looking at some other questions, I decided to try and solve this with semantic predicates. Here's my complete solution:
grammar InputCombinedGrammar;
#parser::members
{
String sep = ",";
public void setSep(String sep)
{
this.sep = sep;
}
private String getSep()
{
return sep;
}
}
input : String { getSep().equals(_input.LT(1).getText()) }? Separator String EOF ;
String : 'a'..'z' ;
Separator : . ;
Two items to note:
The separator will match on any character, not just commas.
The semantic predicate uses lookahead to compare the next token to the separator. If it matches, then the rule proceeds. If not, then it will throw an error.
This solution is trusting the semantic predicate to only use the correct separator. I'm pretty happy with this solution, but I'd like to see others.

Answer 2

I would handle it inside the lexer:
#lexer::members {
...
}
input : String Separator String EOF ;
Separator : { sep.equals(_input.LA(1).getText()) }? . ;
String : 'a'..'z' ;
If you do it inside the parser, all rules defined after the Separator : . ; can never be a single character since it would get caught by the Separator rule.