Flex/Bison: Bad token management?

Question

I've a problem in my lexer and in my parser.

First, in my lexer I've a line like that:

"if"    beginScope(stOTHER); return IF;

And in my parser:

stmt: IF '(' exp ')' stmts
...
stmts: stmt
       | '{' stmt_list '}'
       | '{' '}'

In a code like that:

if(sth) {
    dosth;
}

if(other) {
    doothersth;
}

beginScope will be called two times, because ( I think ) Bison don't know where is the end of the if statement, so when it found the IF token, he takes that as the end of the if statement, and read it a second time to start the other if statement...

Please help me...

Can you call beginScope() for "{" in your Flex file? That way, you can endScope() when you see "}". Just have these single-character expressions return a token like OPENBRACE and CLOSEBRACE to Bison. — chrisaycock
No I can't, because beginScope needs a type in parameter, and there is not always a "{" as you can see — gnidmoo
Move the beginScope operation into your parser actions. Use a mid-rule action if necessary. The lexer must not have side effects, for exactly the reason you have discovered. — zwol

vitaut vitaut · Accepted Answer · 2013-02-28T06:16:04

As Zack mentioned in the comments, you should call beginScope from a parser action:

stmt: IF { beginScope(stOTHER); } '(' exp ')' stmts

Flex/Bison: Bad token management?

1 Answers