Path: utzoo!attcan!uunet!husc6!think!ames!pasteur!agate!ALFA.berkeley.edu!bks
From: bks@ALFA.berkeley.edu (Brad Sherman)
Newsgroups: comp.lang.c
Subject: Style wars
Message-ID: <18621@agate.BERKELEY.EDU>
Date: 28 Dec 88 22:59:07 GMT
Sender: usenet@agate.BERKELEY.EDU
Reply-To: bks@ALFA.berkeley.edu (Brad Sherman)
Organization: University of California, Berkeley
Lines: 40


	There are other stylistic questions apart from curly brace placement
and degree of indentation.  (By the way, source code control systems may
thrash if cb and indent are used injudiciously %}.)
	If there are 100 functions in the program are there 100 source files?
If not, what "rules" should be followed to chunk functions? (I used to
put the I/O routines in files with the functions which needed them; then
we started porting to odd ?NIX machines, VMS, and DOS-networks, now it seems
"better" to hide all those #ifdef's in I/O dependent files?!).
	Does every source file have a declaration of *every* global
(perhaps via #include "global.h"), or should it be done on a need-to-know
basis at the top of the file?  Or only in the body of the function which
references it?
	Under what circumstances should sets of variables be glued together
into structures? Should functions know about variables which they don't
look at or modify?
	If 10% of the functions need to know about a variable, should it 
be made global? 50%? 90%?
	How are we to document soon to be heavily used code (i.e. sure to be
modified after some experimentation)? If I find a nicely troff'ed document
from 10/86 describing your program, but I see that 10 sccs entries (i.e.
modification dates) have occurred since, should I read the document or
studiously avoid it?

	" ... We might suggest that programmers adopt a systematic 
	stategy when they need to understand a program.  While
	reasonable, this is a totally impractical suggestion: a
	systematic stategy can only really be adopted when a program
	is smallish ...  If a programmer attempted to use the
	systematic stategy on a realistically large program, with
	for example, 50,000 lines, the programmer would probalbly
	never get to the enhancement."

	Pinto & Soloway "Providing the Requisite Knowledge Via Software
	Documentation," Proceedings of CHI, 1988. ACM, New York, p.257.

		-Brad Sherman (bks@ALFA.Berkeley.Edu)

------------
      "Hobgoblins have the consistency of small minds."