Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site gatech.UUCP
Path: utzoo!linus!philabs!cmcl2!floyd!clyde!akgua!gatech!spaf
From: spaf@gatech.UUCP (Gene Spafford)
Newsgroups: net.lang
Subject: Re: Strong Typing and Ignorance
Message-ID: <2723@gatech.UUCP>
Date: Thu, 8-Dec-83 01:47:58 EST
Article-I.D.: gatech.2723
Posted: Thu Dec  8 01:47:58 1983
Date-Received: Sat, 10-Dec-83 01:37:37 EST
References: <2892@utcsrgv.UUCP> <137@csd1.UUCP>, <1457@rlgvax.UUCP>
Organization: Georgia Tech School of ICS, Atlanta
Lines: 48


		From: guy@rlgvax.UUCP
		Newsgroups: net.lang
		Subject: Re: Strong Typing and Ignorance
		Message-ID: <1457@rlgvax.UUCP>

		[...]
		procedure foo(var i, j : integer);

		var k : 2 .. 12;

		    begin
			{ ASSERT 1 <= i <= 6, 1 <= j <= 6 }
			k = i + j;
		    end

		if the assertion in question is true, 2 <= i+j <= 12 so the
		assignment will always be valid.


Yes, that assertion will make the result true in all cases, but it is not
a completely correct assertion.  That is, input of the values -37 and 40
for i and j would also produce a correct result, as would 0 and 3 (for 
example).  Depending on how integer arithmetic is defined and implemented,
it is entirely possible that the values -40000 and -25532 could also
produce an answer in range (assuming 16 bit, 2's complement, with no
error triggered for underflow).

The question concerns the exact type, and associated semantics, used in
the evaluation of the expression on the right hand side.  Is it done in
the mode of the LHS?  Or is it done in the mode of the "most precise"
(whatever that is) mode on the RHS?  In general, I see the assignment
operator as a function implying type conversion, and the operation
on the right is done according to whatever semantics apply to those
variables -- without regard to the LHS.  That way, definition of
a subrange is a restriction on the returned value of the assignment
function, akin to the restrictions in some languages about assigning
reals to integers, or defining the range on a trig function.

Look at it this way -- if the assignment statement was
	i := i +j;
would that assertion still be valid?

-- 
Off the Wall of Gene Spafford
School of ICS, Georgia Tech, Atlanta GA 30332
CSNet:	Spaf @ GATech		ARPA:	Spaf.GATech @ CSNet-Relay
uucp:	...!{akgua,allegra,rlgvax,sb1,unmvax,ulysses,ut-sally}!gatech!spaf