Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!seismo!sundc!pitstop!sun!amdcad!ames!sri-spam!rutgers!psuvax1!gondor.psu.edu!schwartz
From: schwartz@gondor.psu.edu (Scott E. Schwartz)
Newsgroups: comp.lang.c
Subject: Re: Question on structures
Message-ID: <3048@psuvax1.psu.edu>
Date: Tue, 3-Nov-87 12:43:07 EST
Article-I.D.: psuvax1.3048
Posted: Tue Nov  3 12:43:07 1987
Date-Received: Sat, 7-Nov-87 05:32:43 EST
References: <1025@phoenix.Princeton.EDU> <2986@mit-vax.LCS.MIT.EDU> <5673@weitek.UUCP>
Sender: netnews@psuvax1.psu.edu
Reply-To: schwartz@gondor.psu.edu (Scott E. Schwartz)
Organization: Penn State University, University Park, PA
Lines: 34

In article <5673@weitek.UUCP> rober@weitek.UUCP (Don Rober) writes:
>>A simple solution (sligthly inefficient). Rewrite
>>	with a.b.c begin
>>		d = 10 ;
>>	end
>>as
>>	{
>>	 struct <whatever> *s = &a.b.c ;
>>	 s->d = 10 ;
>>	}
>A vastly superior solution in that there is nothing implicit happening.  The 
>typical Pascal error is that it is never clear what 'd' is a reference to.

Typical for who?  Anyone framiliar with nested block structured languages
should have no problem with this.  Certainly "never" is too strong a
qualifier.

>(for example, declare a local variable by the name of 'd' - which one is 
>referenced?) 

'd' refers to the field in the record.  That's part of the scoping rules.
Don't you ever have local variables whose names clash with names
of global variables?  (If not, I applaud you, but since C doesn't allow
nested procedures the temptation is very much less.)

> This C solution obviates the problem and is worth whatever the
>cost may be.

I disagree.  The 'cost' is clarity and efficiency.  I want both of these.
I mean, why declare, initialise and use a thoroughly useless local variable
when all you wanted was a (constant offset) field in a record anyway?


-- Scott Schwartz            schwartz@gondor.psu.edu