(I edited my question based on the first comment of @Bart Kiers - thank you!)

I have the following grammar:

SPACE : (' '|'\t'|'\n'|'\r')+ {$channel = HIDDEN;};
START : 'START:';
STRING_LITERAL  : ('"' .* '"')+;
rule    :  START STRING_LITERAL;

and I want to parse languages like: 'START: "abcd" START: "img src="test.jpg""' (string literals could be inside string literals).
The grammar defined above does not work if there are string literals inside a string literal because for the language 'START: "img src="test.jpg""' the lexer translates it into the following tokens: START('START:') STRING_LITERAL("img src=") test.jpg.
Is there any way to define a grammar which is fine for my problem?

• antlr
• antlr3

There are a couple of things wrong here:

• you cannot use fragment rules inside parser rules. You grammar will never create a START token;
• a . char (DOT-char) inside a parser rule matches any token, while inside a lexer rule, it matches any character;
• if you let .* match greedily (and you had defined a proper lexer rule that matches a string literal), the input START: "abcd" START: "img src="test.jpg"" would then have one large string in it: "abcd" START: "img src="test.jpg"" (the first and the last quote would be matched).

So, you cannot embed string literals inside string literals using the same quotes. The lexer is not able to determine if a quote is meant to close the string, or if it's the start of a (new) embedded string. You will need to change that in your grammar.