Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-eddie.UUCP
Path: utzoo!linus!security!genrad!mit-eddie!barmar
From: barmar@mit-eddie.UUCP (Barry Margolin)
Newsgroups: net.lang
Subject: Re: Strong Typing and Ignorance
Message-ID: <1016@mit-eddie.UUCP>
Date: Thu, 8-Dec-83 05:13:48 EST
Article-I.D.: mit-eddi.1016
Posted: Thu Dec  8 05:13:48 1983
Date-Received: Fri, 9-Dec-83 06:31:33 EST
References: <2892@utcsrgv.UUCP> <137@csd1.UUCP> <1457@rlgvax.UUCP>
Organization: MIT, Cambridge, MA
Lines: 44

	... the assignment of an integer to a subrange-typed variable
	requires runtime checking.  Has any work been done on proving
	such assignments correct given constraints on the values assumed
	by the integers in question?

I don't know about research, but isn't that what the whole idea
of types is supposed to obviate?  Also, unless some runtime check
has been done the constraints cannot be assumed to be true, since
the values of most variables are determined by something external
to the program, namely user input.  This is also the only time
you should ever run into an assignment from a superset to a
subset, i.e. the assignment of the input variable to an internal
variable of a subtype.  If the program explicitly checks that the
user's input is valid then an ideal optimizing compiler would
recognize that a check was not needed in the assignment code
(hmm, did I just repeat what the original submission was saying?)
On the other hand, another way to do this would be to have a
handler for the condition signaled if the range-checking failed.
The latter doesn't require as much smarts from the compiler, and
also has a chance of producing better code in some types of type
conversion.  For instance, a good way in PL/I to turn a user
input into a numeric value is:

	on conversion begin;
	     /* Do something about bad input */
	     ...
	end;

	fixed_bin_var = fixed (char_var);

	revert conversion;

This could produce better code than something like

	if valid_fixed_bin (char_var) then
	     fixed_bin_var = fixed (char_var);
	else /* Do something about bad input */
	     ...

because you only have to examine the char_var variable once.
-- 
			Barry Margolin
			ARPA: barmar@MIT-Multics
			UUCP: ..!genrad!mit-eddie!barmar