Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!samsung!brutus.cs.uiuc.edu!ux1.cso.uiuc.edu!ux1.cso.uiuc.edu!m.cs.uiuc.edu!sunb6!voss
From: voss@sunb6.cs.uiuc.edu
Newsgroups: comp.lang.c++
Subject: Re: zortech problem with lex
Message-ID: <24800002@sunb6>
Date: 28 Feb 90 17:39:23 GMT
References: <6300008@ux1.cso.uiuc.edu>
Lines: 52
Nf-ID: #R:ux1.cso.uiuc.edu:6300008:sunb6:24800002:000:2589
Nf-From: sunb6.cs.uiuc.edu!voss    Feb 27 20:30:00 1990


> maxsmith@athena.mit.edu writes:
> Lex generated scanners are slower than hand crafted ones.  Doubly so for
> yacc and its output and yacc it self can't produce a C++ parser, something
> to do with the language not conforming to LALR(1) and all that.

I saw that comment about C++ not being LALR(1) in the g++ distribution,
and at the time believed it.  However, my biggest thrill at OOPSLA '89
was sitting at the same table as Bjarne Stroustrup for a few hours drinking
free beer.  He said that he IS USING A LALR(1) GRAMMER for C++ 2.0.  
He appeared sober & honest, therefore I assume that C++ is LALR(1), but 
non-trivial.
=============================================================================
	First, let me make it clear that I am by no means an expert
at writing compilers (<grin> yet).  I am a graduate student in CS, 
currently taking my first (much overdue) compiler course.  We are using 
the text _Compilers_Principles_Techniques_and_Tools_ by Aho, Sethi, and Ullman.
The programming portion of the class involves using lex/flex and 
yacc/byacc/bison to implement a subset of Pascal.  Now my questions:

> meissner@osf.org
> Second of all, lex-ing is fairly simple.  It took me a day or so, to
> write from scratch the lexer for Data General C.  It would have taken
> more time to come up to speed with lex (ignoring the problem that we
> didn't have lex at the time) than to just write the code.  IMHO if you
> think writing a lexer is too hard, you have no business writing
> compilers.

	No, I don't think writing a lexer is hard.  I did it many times
before I even heard of lex (though I didn't know I was writing "lexers.")
However, I can now write and debug a lexical analyzer in half a day max
using flex, and anyone knowing lex/flex could easily maintain my code.
Therefore I doubt I will ever write another "lexer" by hand.

	I think the surprise is that a "commercial compiler writer"
would not be familiar with flex/lex.  If you don't mind my asking,
when did you study compilers, and what text did you use?

==============================================================================
> schmidt@zola.ics.uci.edu writes:
> Actually, GNU C uses a reserved word recognizer function generated
> automatically by my GPERF program.  But I understand your general
> point.
then goes on to give performance figures showing GPERF is much faster
than flex/lex.

	I am not familiar with GPERF.  What did you give up to
get the speed advantage?  Is GPERF available via anonymous ftp? Where?

						thanks,
						Bill Voss
Email: billvoss@uiuc.edu or voss@cs.uiuc.edu