NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/software/4908

Internet Message Format | 1992-12-12 | 2.9 KB

Path: sparky!uunet!zaphod.mps.ohio-state.edu!swrinde!emory!wupost!spool.mu.edu!agate!dog.ee.lbl.gov!horse.ee.lbl.gov!torek
From: [e-mail] (Chris Torek)
Newsgroups: comp.software-eng
Subject: what is `code? (was C code layout)
Date: 12 Dec 1992 23:17:01 GMT
Organization: Lawrence Berkeley Laboratory, Berkeley CA
Lines: 54
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]>
NNTP-Posting-Host: 128.3.112.15

(first a short note on style)

In article <[e-mail]> [e-mail]
writes:
>... Standards are fine and students should learn that they will probably
>be required to code to standards when they get a real job. Why rob them
>of the LEARNING experience of trying different styles when they are
>students?

If any one teacher decrees a certain style, several others will decree
others or not decree one at all, so this particular complaint seems
unjustified (at least for students for whom programming is a large part
of their coursework---in a `Pascal for Psych Majors' class a single
particular style requirement may have a more lasting effect, and hence
requires more care).

>WHAT! A makefile is NOT CODE.

A makefile certainly *is* code.  The Unix `make' utility is programmed
in a peculiar mix of declarative and imperative operations.  The
declarative sections describe interdependencies (`foo.o may change
whenever foo.c or global.h changes'), and the imperatives are fed to
the shell for processing (`to produce foo.o, execute the following
commands').

(Most `make' clones, including GNUmake and Pmake, have added typical
imperative-language directives, so that one can, for instance, say `for
i in x y z:  declare that $i depends on g and when rebuilding $i use
...' or `if the file ../Makefile.inc exists, read it'.  In a sense this
is a loss: an entirely-declarative language has the advantage of
allowing arbitrary implementation changes.  On the other hand, through
clever use of macros and inclusion, one can now reduce many makefiles
entirely to declarations; the Berkeley net.2 distribution has many
programs whose makefile simply names the program and, if necessary, the
source files.  [The default is to assume that program foo is built from
foo.c.  All C-language inclusion dependencies are derived
semi-automatically.])

One mini-language I created recently consists entirely of declarations,
such as:

    The name `esp' represents a device which
      attaches to an sbus interface,
      and provides a scsi interface.
    The code for an `esp' is built from
      the C source file sparc/sbus/esp.c,

After reading several sets of these files, the `compiler' for this
language emits, among other things, a makefile.  This makefile is still
code, but it is no longer *source* code.  The source is the set of
declarations.
--
In-Real-Life: Chris Torek, Lawrence Berkeley Lab CSE/EE ([phone])
Berkeley, CA        Domain:    [e-mail]