Re: Compiler Books? Parsers?

Chris F Clark wrote:
> In the lexical case, the parser assumes that the tokens the lexer has
> handed to it are correct. However, when one delimiter is missed, the
> tokens the lexer hands to the parser are not the correct ones--they
> are gibberish (lexically correct gibberish). And if one looked at the
> lexical problem in terms of finding a correctly parsable program with
> the *minimum editing distance*, one would immediately see that the
> correct program would have a different set of tokens. However, I've
> never seen a lexer/parser that attempted to find the program with the
> minimum editing distance from the original source.


This sounds much like robust *natural* language parsing where you have
malformed sentences all the time still trying to understand the meaning
of them. I have seen approachs using propagation of constraints
describing the structure of valid sentences. This incoporates weak
constraints that only claim certain properties *should* be satiesfied. I
can't remember how parsing actually was done and what variables were
constraint over what domain, but I could find it out if anyone is
interested.


Oliver

Return to the comp.compilers page.
Search the comp.compilers archives again.