Path: utzoo!attcan!uunet!pilchuck!dataio!bright
From: bright@Data-IO.COM (Walter Bright)
Newsgroups: comp.lang.c++
Subject: Re: zortech problem with lex
Message-ID: <2352@dataio.Data-IO.COM>
Date: 1 Mar 90 19:09:16 GMT
References: <6300008@ux1.cso.uiuc.edu> <2347@dataio.Data-IO.COM> <741@ns-mx.uiowa.edu>
Reply-To: bright@Data-IO.COM (Walter Bright)
Organization: Data I/O Corporation; Redmond, WA
Lines: 35

In article <741@ns-mx.uiowa.edu> csu025@umaxc.weeg.uiowa.edu.UUCP (Kent Williams) writes:
<<I know very little about lex/flex.
<If you don't mind my asking, do you USE any automated compiler writing
<tools?

In a word, no. lex is good for a lot of applications, but where maximum
performance is required, a hand-built scanner is necessary. It continually
surprises people that Zortech C compiles as fast as Turbo C, even though
ZTC is a 2 pass compiler and TC is 1 pass! The secret to speed is
in the scanner (not the symbol table!).

Other less significant reasons are:
1. When I started, in 1983, no lexes were available for MSDOS.
2. I don't care for copyright problems. I don't want to hire a lawyer
   to interpret the licensing agreement.
3. The compiler has been occasionally ported to other platforms for
   debugging purposes. Being constrained by the availability of a
   particular version of lex is unacceptable.
4. It turns out that it's *very* useful to have complete control over
   the compilation process from source code to EXE. Depending on
   some other vendor's program which generates a critical section of
   the compiler causes problems if the source to it is not available.
   (See the cfront people with their problems with identifier length
   in MSC!)

So-called 'compiler compilers' are also not terribly useful to me. They
are more appropriately called 'parser generators'. The parsing code itself
is a minor part of a real compiler. The 'compiler compiler' still requires
you to write symbol table, semantic analysis, optimization, and code
generation, which is over 95% of the compiler.

For programs where scanning and parsing are most of the job, like in
a tool like indent, lex and yacc are highly useful. Those tools are
also very nice for prototyping a new language, because it's easy to
drastically change the syntax of what you're compiling.