Path: utzoo!attcan!uunet!mcsun!ukc!stl!pgh
From: pgh@stl.stc.co.uk (P.G.Hamer)
Newsgroups: comp.lang.misc
Subject: Re: Type systems with dimensional units
Message-ID: <3978@stl.stc.co.uk>
Date: 28 Jan 91 10:00:51 GMT
References: <1991Jan21.225221.5598@shell.shell.com>
Sender: news@stl.stc.co.uk
Reply-To: "P.G.Hamer" <pgh@stl.stc.co.uk>
Organization: STC Technology Limited, London Road, Harlow, Essex, UK
Lines: 32

In the referenced article robison@brazil (Arch D. Robison) writes:
>I am looking for references on languages that incorporate ``units checking''
>into their type system. 

Tried to e-mail but it bounced.


Sorry, no good reference  - but I seem to remember an article in a very old
ACM comms (or just possibly BCS) - late 60's or early 70's.

I assume that you are going to allow mixed expressions (area = x * y), so 
I you will presumably want an intermediate representation that keeps powers
of the fundamental dimensions of space time (length, time, mass, ...).

As I remember my physics, it gets a bit dodgy in the area of electrical
units.

It also does not identify when dimensionless contants are missing. Superficially
this seems OK, but it can cause problems in several areas. First, it dousn't
spot things like missing multiples of pi in electrical formula (a fairly common
mistake), as the need for them is a function of the measurement system chosen
they seem sort of dimensional. Second, missing out 'cosmological' constants
(eg Plank's constant).  Third, things like revs (as in revs per second) and
counts (as in counts per second) are normally considered dimensionless -- but
should [usually?] balance in an equation. Fourth, some measurements don't have
very useful dimensions (eg one measure of viscosity is reported in seconds!).

Please post (or e-mail) any good references you find. As a general problem it 
seems quite difficult, although most problem domains would probably be well
served by a fairly simple solution using a few fundamental dimensions. 

Regards, Peter