Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!caip!think!nike!ucbcad!ucbvax!DECWRL.DEC.COM!haynes
From: haynes@DECWRL.DEC.COM
Newsgroups: net.lang.mod2
Subject: Re: Single Pass Modula-2 Compiler
Message-ID: <8610142109.AA23717@vulcan.DEC.COM>
Date: Tue, 14-Oct-86 19:28:07 EDT
Article-I.D.: vulcan.8610142109.AA23717
Posted: Tue Oct 14 19:28:07 1986
Date-Received: Mon, 20-Oct-86 22:14:03 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The ARPA Internet
Lines: 27


	Re: Declaration after use is an abomination, 
	    and should not bear consideration.

Yeah? And so's your mother! (Proof by assertion refuted by childish ad
hominem attack. Now that I've satisified net protocol...)

What's wrong with declaration "after use"? Especially for procedures
and the destinations of pointer types, declaration after "first
mention" is often a natural programming style. It does make single pass
compilation more difficult, sometimes impossible, and has performance
implications for any compiler that has to translate the language that
allows it. However declaration after use is not the obvious abomination
you would try to hand wave us into believing.

If claim that declaration before use results in more correct, easier to
maintain programs, I will reply that the easiest to maintain large
programs I ever worked on had all procedures declared in strict
alphabetical order. That sure made a given procedure easy to find.

One pass compilers are a boon to compiler writers and people who have
to run on small machines. For them a one pass compiler is often the
only possibility. For others who have large machines, and want an easy
to use language, but don't want to sacrifice correctness or efficient
object code, a multi-pass compiler is often the right thing.

	-- Charles