Re: [boost] Discovering array size

From: Howard Hinnant (hinnant_at_[hidden])
Date: 2003-12-03 09:37:55

On Dec 3, 2003, at 7:31 AM, AlisdairM wrote:

> The following seems such an obvious utility, I was surprised I could 
> not
> find it in the boost distribution.  Could someone point me to the
> equivalent if I am missing it.  Alternatively, is this worth cleaning 
> up
> for a submission?
>
> Basic problem is to know the size of the following array:
>
> int[] aInts = { 1, 2, 3 );
<snip>
> I suspect there may be a way to make this a compile-time constant, but
> can't figure out the automatic type deduction yet.

Two similar "type traits" were proposed for the type traits std::tr1 
lib:

http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/papers/2003/n1519.htm

template <class T, unsigned I = 0> struct dimension {
   static   const std::size_t value                    = 
implementation_defined;
   typedef  std::size_t                                value_type;
   typedef  integral_constant<value_type,value>        type;
   operator type()const;
};

This is a compile-time version of your "size" which has also been 
generalized to pick out the ith dimension, defaulting to the 0th (or 
outer) dimension as your size function extracts:

assert(dimension<int>::value == 0);
assert(dimension<int[2]>::value == 2);
assert(dimension<int[2][4]>::value == 4);
assert(dimension<int[][4]>::value == 4);
assert((dimension<int, 1>::value) == 0);
assert((dimension<int[2], 1>::value) == 0);
assert((dimension<int[2][4], 1>::value) == 4);
assert((dimension<int[][4], 1>::value) == 4);

The second traits in this area was rank (the number of dimensions):

template <class T> struct rank {
   static   const std::size_t value                    = 
implementation_defined;
   typedef  std::size_t                                value_type;
   typedef  integral_constant<value_type,value>        type;
   operator type()const;
};

assert(rank<int>::value == 0);
assert(rank<int[2]>::value == 1);
assert(rank<int[][4]>::value == 2);

The reaction of the lwg at the last standards meeting to these two 
"array traits" can be found here:

http://anubis.dkuug.dk/jtc1/sc22/wg21/docs/papers/2003/n1541.pdf

>  The LWG wasn’t sure whether this was useful; the few people who could 
> use it reliably for metaprogramming would probably find it just as 
> easy to write it themselves.

I disagree with this stance and it is on my to-do list to motivate 
dimension and rank for the Spring meeting.  As you appear to have a 
need for at least "dimension", I would appreciate it if you could 
expound on your desire for this functionality.  Any arguments you might 
have for why this functionality belongs in the std::tr1::type_traits 
lib will be incorporated into my own comments and presented to the lwg.

If I'm mistaken, and no one here would like to see dimension and rank 
in the std::tr1::type_traits lib, then please let me know that too.  I 
/believe/ these would be useful.  But I do not /know/ they would.  If 
I'm mistaken, it wouldn't be the first time, and I might as well find 
out sooner than later. :-)

-Howard