Re: why use flex?

> Travers> The only real complaint I have is the *still* lacking direct
> Travers> C++ support. If one wants that, one has to use to modified
> Travers> Flex++/Bison++ utilities.
>
> Could someone enlighten me what kind of real advantages using
> generated scanner and parser classes could have over the traditional
> yyparse () / yylex () interface? Does it have something to do with
> re-entrant parsers? Or does it make "sub-parser calls" easier
> (i.e. when the input can contain multiple languages that cannot be
> feasibly handled by one parser, like SDL (ITU-T Z.100,Z.105) with
> ASN.1 definitions embedded between the ALTERNATIVE/ENDALTERNATIVE
> keywords)?


Excellent question which I forgot to answer: It has more to do with
encapsulating multiple parsers and making them easier to become
re-entrant. Currently, one has to ask Flex/Bison to emit their
scanners and parsers with a prefix, which isn't that bad, but it makes
it difficult to, say, switch lexers at run-time for a source code
highlighting system. One has to add a case statement at every point
one needs to call the lexer for C-style lexical scanners. With an
object encapsulation, the scanner objects can just be switched without
a problem.


(Of course this is a strawman argument: One could write a wrapper
class by hand for each lexical scanner, but it would be nice if Flex
and Bison would do it for me. <grin>)


--
Travers Naran Computer Programmer, P/T Meddler in Time & Space
New Westminster, British Columbia, Canada, North America, Earth...
E-mail: tnaran@direct.ca, naran@vcn.bc.ca
WWW: <http://www.loonie.net/~tnaran>

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