Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!rutgers!sri-spam!ames!ucbcad!ucbvax!olsen.UUCP!nagler
From: nagler@olsen.UUCP (Robert Nagler)
Newsgroups: comp.lang.modula2
Subject: Re: Is Modula-2 a One-pass language?
Message-ID: <8708252010.AA00419@klaus.olsen.uucp>
Date: Tue, 25-Aug-87 18:57:44 EDT
Article-I.D.: klaus.8708252010.AA00419
Posted: Tue Aug 25 18:57:44 1987
Date-Received: Thu, 27-Aug-87 02:53:49 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The ARPA Internet
Lines: 41


    >I sent a message on 22nd July to info-modula-2 about one-pass compilation.
    >I was surprised to receive only three replies (from Chuck Bilbe,

I think Randy mentioned a minor problem in receiving replies, because
your mail address didn't work.  I tried sending a reply directly to you, 
but it bounced and I just gave up. 

In answer to the question posed:
    >I would be interested to hear of any other difficulties that a one-pass
    >compiler would have in compiling the language defined by the Report.

I would first like to change the system to a guilty until proven
innocent version as follows:
    Why do we need to remove the loose declaration order of Modula-2?

The answers to this question are:
    1) The compiler will go faster.
    2) "Use before declaration is an abomination" -- Randy Bush.

In response to answer (1), I would like ask how much faster the
compiler will go in comparison to the average compilation/link
sequence of a typical M2 program.  My initial impression would be
10 to 20% faster, but then I am not a compiler writer.  The speed
improvement on a 10 second compile doesn't seem worth it in 
comparison to the amount of time it takes a programmmer to verify 
that a module will be fed into the compiler in the "correct" order.

Answer (2) is an interesting opinion which Dr. Wirth allowed us to
explore.  After using C and Pascal, I find the loose order of
declaration in M2 allows the programmer to put the important concepts
at the top of the module where they are seen first.  I wonder if (2)
is more applicable to systems which require programmers to use
cross-referenced listings as opposed to ones which supply
multi-window text editors.

In any event, I do request that the question be rephrased
to account for the large body of Modula-2 code which may or may
not follow this upward-incompatible restriction.

Rob