Re: syntaxic exploration

[Niklas Holsti <niklas.holsti@tidorum.invalid>]

On 17-12-23 18:23 , Jeffrey R. Carter wrote:
> On 12/22/2017 11:11 PM, Niklas Holsti wrote:
>>
>> I think the aggregate form (null array) was suggested some time ago,
>> in analogy with the existing (null record). It looks good to me, but
>> as I remember, there was some objection. One problem is that if the
>> index type has exactly one value, then a null array with that index
>> type cannot exist (because then A'First = A'Last for any such array A).
>
> Actually, a null array of such a type can exist if it's a string type,
> because you can use the string literal "" for such types. AIUI, 'Last is
> undefined for such a value.

RM 4.2(9) says "for a null string literal, the upper bound is the 
predecessor of the lower bound". I would understand this to mean 
applying 'Pred to the lower bound, which will raise Constraint_Error if 
the type has only one value.

If I try this with GNAT:

    type One_T is (Unique);
    type Str_T is array (One_T range <>) of Character;
    S : constant Str_T := "";

I get compilation errors:

    nullstr.adb:11:26: null string literal not allowed for type
       "Str_T"  defined at line 9
    nullstr.adb:11:26: static expression fails Constraint_Check

where line 11 is the one with the "" literal.

So one can try to make a null array of this type with "", but it will fail.

> Allowing (null array) for non-string array
> types with an index type with a single value would presumably work the
> same.

I agree, it should also raise Constraint_Error.

-- 
Niklas Holsti
Tidorum Ltd
niklas holsti tidorum fi
       .      @       .