Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!purdue!ames!apple!agate!ucbvax!hplabs!otter!ijd
From: ijd@otter.hpl.hp.com (Ian Dickinson)
Newsgroups: comp.lang.misc
Subject: Re: multi-dimensional switch statements
Message-ID: <2400025@otter.hpl.hp.com>
Date: 25 Jul 89 09:27:50 GMT
References: <26584@agate.BERKELEY.EDU>
Organization: Hewlett-Packard Laboratories, Bristol, UK.
Lines: 57

> /otter:comp.lang.misc / mwm@mica.berkeley.edu (Mike Meyer) /
> The basic idea is to be able to code a state table as is, so that the
> table is obvious. A two-dimensional switch statement is the analogy
> that comes to mind. A typical example might look like:
> 
> 
> switch (row:bumpx, col: bumpy) {
> 
> 	case 0:		case 3:		case 5:			case 7:
> 
> case 2:	z = x++ + y;	z = 2 * x + y;	z = 3 * x++ + y;	z = x;
> case 4:	z = x + y;	z = x + y++;	z = 3 * x + y++;	z = y;
> case 7:	z = x++ + y++;	z = x + 2 * y;	z = 8 * x++ + y++;	z = x;
> 
> }

From your example, it looks as though you don't need anything more than
pattern matching on the left-hand side of an equation:

Eg, Prolog
state_machine( 0, 2, Z ) :-  Z is <expression>
state_machine( 0, 4, Z ) :-  Z is <expression>
state_machine( 0, 7, Z ) :-  Z is <expression>
state_machine( 3, 2, Z ) :-  Z is <expression>
... etc


or, SML
fun stateMachine 0 2  = <expression>
  | stateMachine 0 4  = <expression>
  | stateMachine 0 7  = <expression>
  | stateMachine 3 2  = <expression>
... etc

Compiler technology for both languages is maturing, so they could each do
a pretty good job of optimising the tests.  I have left out the details of
the <expressions> since good programming practice in each language avoids
scungy imperative operators like ++.  Adding the extra parameters to do
it properly would obscure the point.

Once upon a different life, I would have preferred to use Prolog.  However,
having been writing SML code for some time now, I like the mental hygiene
of keeping the expressions type-correct, and having the compiler enforce
that for me.  In the above example, you wouldn't have to keep to integers
for the case labels,  but each parameter (effectively, every column in the
above) would have to be the same type.  In return, the compiler will warn
when you have incomplete coverage of that type.

If I missed the point of the question, please explain further.
Cheers,
Ian.


| Ian Dickinson,  HP Labs, Information Systems Centre,    Bristol, England  |
| net: ijd%otter@hplabs.hp.com        |                                     |
|  or:       ijd@hplb.uucp            |  ?-  mind(X),  body(X),  spirit(X). |
| These opinions are all my own work. |                                     |