Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!mit-eddie!think!ames!ptsfa!ihnp4!alberta!auvax!rwa
From: rwa@auvax.UUCP (Ross Alexander)
Newsgroups: comp.lang.c
Subject: Re: type-indexed arrays (was: enum - enum ?)
Message-ID: <187@auvax.UUCP>
Date: Sun, 14-Jun-87 19:42:12 EDT
Article-I.D.: auvax.187
Posted: Sun Jun 14 19:42:12 1987
Date-Received: Sun, 21-Jun-87 02:11:42 EDT
References: <139@starfire.UUCP> <516@haddock.UUCP> <20540@sun.uucp> <1226@crash.CTS.COM>
Organization: Athabasca U., Alberta, Canada
Lines: 82
Summary: you miss my point, i think

In article <1226@crash.CTS.COM>, ford@crash.CTS.COM (Michael Ditto) writes:
> In article <184@auvax.UUCP> rwa@auvax.UUCP (Ross Alexander) writes:
> >	int	array[ short ];
> >
> >implies that array[] has members indexed by -1, -2, -3, and so on
> 
> No.  A subscript must simply be some sort of integral type which the copmiler
> can perform multiplication on or stick into an index register (i.e. it must
> be something you can add to a pointer).

yes, yes, we all know that.  float and double subscript values are
strictly for BASIC, right :-) ??  what i was driving at was "how much
storage is the compiler going to ALLOCATE for array[ short ] ?", and
the corollory question is "what is the value of the index of the
first and last allocated elements of array[]?".

by analogy with Pascal, one might expect the first allocated element
to have a negative index value, since 'short' is a signed type, and
is really "MIN_SHORT_VALUE ..  MAX_SHORT_VALUE" when you think of it
as a subrange type (subrange of int).  on 'typical' architectures,
this might be -0x8000 ..  0x7fff or whatever.  this was my meaning
(poorly expressed, i grant you) when i said 'indexed by -1, -2, -3,
and so on'.  at any rate, the array will have (some) elements indexed
by negative values, and (some) by non-negative values.

now, getting back to the can of worms (smile, gentle people), say i
declare

	int	char_count[ char ];

(I want to count character frequencies in some input, or something),
and further let us say that chars are signed, so this declaration is
equivalent to

	int	char_count[ MIN_CHAR .. MAX_CHAR ];

with MIN_CHAR == -128, and MAX_CHAR == 127.  (As an aside, this is
precisely the case locally).

now, you suggest that the token 'char_count' would have the value '&
char_count[ MIN_CHAR ]'.  (actually, it's a pointer constant, but
never mind).  then i can say that '* ( char_count + 3 )' references
an int three int's further on in memory than '& char_count[ MIN_CHAR
]'.  ok?  and of course, 'char_count[ 3 ]' is the third int after
'char_count[ 0 ]', which to my way of thinking is in the middle of
the array.  but we get into a small problem when we observe that '* (
char_count + 3 )' and 'char_count[ 3 ]' are formally equivalent, but
talk about entirely different places (whoops!!).  this _reductio ad
absurdem_ leads me to think that fooling with the current semantics
of the pointer constant 'char_count' is not a good idea, i.e., it
should still mean '& char_count[ 0 ]' regardless of the fact that
happens to be the middle of the array.  And you yourself observed
how much grief that is going to cause.

has anybody got a bright idea?  i am (reluctantly) forced to think
that fiddling with the semantics of C array declarations is not wise.

as any aside,  any one have a good way to introduce sets into C?  we
already have enum's,  i really would like to have sets and set
operations.  i know that you can fake it for small sets ( <= number
of bits in a 'long' ) very easily, but it gets ugly after that.  and
writing

	enum days { sunday, monday, /* ... */, friday, saturday };
	long days_off;
	int x;

	days_off = ( 1 << sunday ) | ( 1 << saturday );
	if ( ( days_off & ( 1 << x ) ) != 0 ) { /* do something */ };


is a little less intuitive than ( i create syntax freely here, the
semantics are pretty much the same )

	set days_off[ days ];

	days_off = [ sunday, saturday ];
	if ( days_off[ x ] ) { /* do something */ };

i don't like the notation especially, but...

...!ihnp4!alberta!auvax!rwa		Ross Alexander, Athabasca University