antlr4 share rules across modes

Question

I want to write a lexer that has multiple modes. But the modes are mostly similar. The only difference is that I reference the same character by a different name. I have it working, the problem is I have to copy the entire lexer change all the names, add types, and add one line for each mode.

Here is the general problem I want to solve. I want a comma to have high precedence outside a '[' ']'. I want it to have a low precedence inside the '[', ']'. To do that I push and pop modes with the '[' and ']'. But since the only thing I am changing is the precedence I have to copy all the rules into each mode and given them different names.

One additional thing, once inside a '[' there would be no way to get back to the high precedence comma. So when the grammar moves into a '{', '}' section the comma reverts back to high precedence.

In order to accomplish this I have the initial default mode plus CONJUNCTION (high precedence) and NON_CONJUNCTION (low precedence). I copy all the rules from the default mode and rename them to C_ or NC_. Then I assign their type back to the type of the default mode.

I would rather accomplish this without coping, renaming, and assigning types to all the rules from the default mode.

Here is my lexer:

lexer grammar DabarLexer;

OPEN_PAREN : '(' -> pushMode(NON_CONJUNCTION) ;
CLOSE_PAREN : ')' -> popMode;
OPEN_BRACKET : '[' -> pushMode(NON_CONJUNCTION) ;
CLOSE_BRACKET : ']' -> popMode ;
OPEN_CURLY : '{' -> pushMode(CONJUNCTION) ;
CLOSE_CURLY : '}' -> popMode ;
SPACE : ' ' ;
HEAVY_COMMA : ',';
ENDLINE : '\n' ;
PERIOD : '.' ;
SINGLE_QUOTE : '\'' ;
DOUBLE_QUOTE : '"' ;
INDENTION   : '\t' -> skip;

fragment SYMBOL : HEAVY_COMMA | OPEN_BRACKET | CLOSE_BRACKET | OPEN_PAREN | CLOSE_PAREN | OPEN_CURLY | CLOSE_CURLY | SPACE | ENDLINE | PERIOD | SINGLE_QUOTE | DOUBLE_QUOTE | INDENTION ;
ESCAPE : '\\' SYMBOL ;

fragment NON_SYMBOL : ~[(){}',; \n.\t"\[\]] ;
IDENTIFIER : (NON_SYMBOL | ESCAPE)+ ;
LITERAL : (SINGLE_QUOTE (NON_SYMBOL | ESCAPE)+ SINGLE_QUOTE) | DOUBLE_QUOTE (NON_SYMBOL | ESCAPE)+ DOUBLE_QUOTE ;

mode CONJUNCTION ;
C_HEAVY_COMMA : ',' -> type(HEAVY_COMMA);

C_OPEN_PAREN : '(' -> type(OPEN_PAREN), pushMode(NON_CONJUNCTION) ;
C_CLOSE_PAREN : ')' -> type(CLOSE_PAREN), popMode;
C_OPEN_BRACKET : '[' -> type(OPEN_BRACKET), pushMode(NON_CONJUNCTION) ;
C_CLOSE_BRACKET : ']' -> type(CLOSE_BRACKET), popMode ;
C_OPEN_CURLY : '{' -> type(OPEN_CURLY), pushMode(CONJUNCTION) ;
C_CLOSE_CURLY : '}' -> type(CLOSE_CURLY), popMode ;
C_SPACE : ' ' -> type(SPACE);
C_ENDLINE : '\n' -> type(ENDLINE);
C_PERIOD : '.' -> type(PERIOD);
C_SINGLE_QUOTE : '\'' -> type(SINGLE_QUOTE);
C_DOUBLE_QUOTE : '"' -> type(DOUBLE_QUOTE);
C_INDENTION   : '\t' -> type(INDENTION),skip;

fragment C_SYMBOL : ( HEAVY_COMMA | C_OPEN_BRACKET | C_CLOSE_BRACKET | C_OPEN_PAREN | C_CLOSE_PAREN | C_OPEN_CURLY | C_CLOSE_CURLY | C_SPACE | C_ENDLINE | C_PERIOD | C_SINGLE_QUOTE | C_DOUBLE_QUOTE | C_INDENTION ) ;
C_ESCAPE : '\\' C_SYMBOL -> type(ESCAPE);

fragment C_NON_SYMBOL : ~[(){}',; \n.\t"\[\]] ;
C_IDENTIFIER : (C_NON_SYMBOL | C_ESCAPE)+ -> type(IDENTIFIER);
C_LITERAL : ((C_SINGLE_QUOTE (C_NON_SYMBOL | C_ESCAPE)+ C_SINGLE_QUOTE) | C_DOUBLE_QUOTE (C_NON_SYMBOL | C_ESCAPE)+ C_DOUBLE_QUOTE) -> type(LITERAL);



mode NON_CONJUNCTION ;
LIGHT_COMMA : ',' ;

NC_OPEN_PAREN : '(' -> type(OPEN_PAREN), pushMode(NON_CONJUNCTION) ;
NC_CLOSE_PAREN : ')' -> type(CLOSE_PAREN), popMode;
NC_OPEN_BRACKET : '[' -> type(OPEN_BRACKET), pushMode(NON_CONJUNCTION) ;
NC_CLOSE_BRACKET : ']' -> type(CLOSE_BRACKET), popMode ;
NC_OPEN_CURLY : '{' -> type(OPEN_CURLY), pushMode(CONJUNCTION) ;
NC_CLOSE_CURLY : '}' -> type(CLOSE_CURLY), popMode ;
NC_SPACE : ' ' -> type(SPACE);
NC_ENDLINE : '\n' -> type(ENDLINE);
NC_PERIOD : '.' -> type(PERIOD);
NC_SINGLE_QUOTE : '\'' -> type(SINGLE_QUOTE);
NC_DOUBLE_QUOTE : '"' -> type(DOUBLE_QUOTE);
NC_INDENTION   : '\t' -> type(INDENTION),skip;

fragment NC_SYMBOL : ( LIGHT_COMMA | NC_OPEN_BRACKET | NC_CLOSE_BRACKET | NC_OPEN_PAREN | NC_CLOSE_PAREN | NC_OPEN_CURLY | NC_CLOSE_CURLY | NC_SPACE | NC_ENDLINE | NC_PERIOD | NC_SINGLE_QUOTE | NC_DOUBLE_QUOTE | NC_INDENTION ) ;
NC_ESCAPE : '\\' NC_SYMBOL -> type(ESCAPE);

fragment NC_NON_SYMBOL : ~[(){}',; \n.\t"\[\]] ;
NC_IDENTIFIER : (NC_NON_SYMBOL | NC_ESCAPE)+ -> type(IDENTIFIER);
NC_LITERAL : ((NC_SINGLE_QUOTE (NC_NON_SYMBOL | NC_ESCAPE)+ NC_SINGLE_QUOTE) | NC_DOUBLE_QUOTE (NC_NON_SYMBOL | NC_ESCAPE)+ NC_DOUBLE_QUOTE) -> type(LITERAL);

Artilleryman

Answer 1

Your current solution is very similar to the solution I use. For example, take a look at the TemplateComment mode of the grammar I use for for StringTemplate 4 support in ANTLRWorks 2. One helpful thing I implemented in ANTLR 4 a while back is it won't create duplicate token types for a rule in this form.

// No TemplateComment_NEWLINE token type is created here, because the
// type(NEWLINE) action means this rule produces tokens of a specific type.
TemplateComment_NEWLINE : NEWLINE -> type(NEWLINE), channel(HIDDEN);

Guidon

I reviewed your link. The solutions are they same. Definitely using the type allows you to limit how much is duplicated. It allows the parser one type even though the parser mentions several. The same goes for the parser. I do not have the cred to choose your answer. – Artilleryman