Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!mcsun!ukc!dcl-cs!aber-cs!emerald!pcg
From: pcg@emerald.cs.aber.ac.uk (Piercarlo Grandi)
Newsgroups: comp.lang.c++
Subject: Re: Define's and Devices.
Message-ID: <PCG.89Nov27225140@emerald.cs.aber.ac.uk>
Date: 27 Nov 89 22:51:40 GMT
References: <478@hades.OZ>
Sender: pcg@aber-cs.UUCP
Organization: Coleg Prifysgol Cymru
Lines: 78
In-reply-to: ing@hades.OZ's message of 25 Nov 89 00:39:45 GMT

In article <478@hades.OZ> ing@hades.OZ (Ian Gold) writes:


	   The first problem is the "#define syndrome".  Assume I have a keyboard
   defined as

	   enum keys { ENTER = 0x13,
		       ESC   = 0x27,
		       A_CHR = 0x41,
		       B_CHR = 0x42,
			   .
			   .
			   .
		     };

   In C I would declare this globally so everybody knows about it.  What I
   would like to do in C++ is something like

	   class keys
	   {
	   public:
		   enum keys { ENTER = 0x13 
				   .
				   .
				   .
			     };
	   };

	   class program : keys
	   {
	   public:
		   void program :: function(void) { return keys::ENTER; }
	   }

   and allow it to be inherited only by the classes that need it.
   The problem is that the above example is NOT C++.


Actually it is C++ 2.0; in C++ 1.2 etc... you can instead use
static consts, in some compilers at least [in 1.2 you cannot
initialize a class static, but a const often can...].

   I realise
   that this is much slower than using #defines and would only be useful if
   one uses multiple inheritance.

Without discussing this, neither assertion is necessarily true. Use enums,
not #defines, in C++ 2.0.


	   My second problem is that of devices.  Say I have three CPU's
   2 keyboards and half a VDU.  Somewhere I must declare these as being the
   only devices avaliable.  IS THE FOLLOWING CODE THE "CORRECT" WAY TO HANDLE
   THE SITUATION.

	   class program
	   {
		   CPU 		cpu1;
		   CPU 		cpu2;
		   CPU		cpu3;
		   KEYBOARD	keyboard1;
		   KEYBOARD	keyboard2;
		   VCU		vdu(0.5);	
		   boolean		program();
	   public:
		   go() { keyboard1.reset(); 
			  keyboard2.reset(); 
			  vdu.reset();
			  program();         }
	   };

This is debatable. Function/control decomposition need not be
mapped onto data abstraction. This may well be excessive
application of Object Orientedness.
--
Piercarlo "Peter" Grandi           | ARPA: pcg%cs.aber.ac.uk@nsfnet-relay.ac.uk
Dept of CS, UCW Aberystwyth        | UUCP: ...!mcvax!ukc!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk