Observe:

void main() {
     int[3] a1 = [1, 3, 6];
     int[]  a2 = a1[] * 3;       // line 3, Error
     int[]  a3 = a1.dup[] *= 3;  // line 4, OK?
     int[]  a4 = (a1[] * 3).dup; // line 5, Error
}


Currently the operation in line 4 is accepted:

test.d(3,17): Error: array operation a1[] * 3 without destination memory not allowed
test.d(5,18): Error: array operation a1[] * 3 without destination memory not allowed

Is it a good idea to support something like line 5?

Bye,
bearophile

On Thursday, 9 October 2014 at 11:29:14 UTC, bearophile wrote:
> Observe:
>
> void main() {
>      int[3] a1 = [1, 3, 6];
>      int[]  a2 = a1[] * 3;       // line 3, Error
>      int[]  a3 = a1.dup[] *= 3;  // line 4, OK?
>      int[]  a4 = (a1[] * 3).dup; // line 5, Error
> }
>
>
> Currently the operation in line 4 is accepted:
>
> test.d(3,17): Error: array operation a1[] * 3 without destination memory not allowed
> test.d(5,18): Error: array operation a1[] * 3 without destination memory not allowed
>
> Is it a good idea to support something like line 5?
>
> Bye,
> bearophile

Why not? Looks fine to me, you've explicitly allocated some new memory for the result to sit in, the lack of which is the motivation for lines 3 an 5 being errors.

On Thursday, 9 October 2014 at 11:29:14 UTC, bearophile wrote:
> Observe:
>
> void main() {
>      int[3] a1 = [1, 3, 6];
>      int[]  a2 = a1[] * 3;       // line 3, Error
>      int[]  a3 = a1.dup[] *= 3;  // line 4, OK?
>      int[]  a4 = (a1[] * 3).dup; // line 5, Error
> }
>
>
> Currently the operation in line 4 is accepted:
>
> test.d(3,17): Error: array operation a1[] * 3 without destination memory not allowed
> test.d(5,18): Error: array operation a1[] * 3 without destination memory not allowed
>
> Is it a good idea to support something like line 5?

It's equivalent to:

    int[] tmp = a1[] * 3;
    int[] a4 = tmp.dup;

The first part is of course identical to line 3, so this should be an error, too. Normal rules for evaluation order require `a1[] * 3` to be evaluated before `(...).dup`, so where is it supposed to store the intermediate result?

Marc Schütz:

> It's equivalent to:
>
>     int[] tmp = a1[] * 3;
>     int[] a4 = tmp.dup;
>
> The first part is of course identical to line 3, so this should be an error, too. Normal rules for evaluation order require `a1[] * 3` to be evaluated before `(...).dup`, so where is it supposed to store the intermediate result?

So do we need a new different syntax to do it? :-)

Bye,
bearophile

On Thursday, 9 October 2014 at 18:01:27 UTC, John Colvin wrote:
> On Thursday, 9 October 2014 at 11:29:14 UTC, bearophile wrote:
>> Observe:
>>
>> void main() {
>>     int[3] a1 = [1, 3, 6];
>>     int[]  a2 = a1[] * 3;       // line 3, Error
>>     int[]  a3 = a1.dup[] *= 3;  // line 4, OK?
>>     int[]  a4 = (a1[] * 3).dup; // line 5, Error
>> }
>>
>>
>> Currently the operation in line 4 is accepted:
>>
>> test.d(3,17): Error: array operation a1[] * 3 without destination memory not allowed
>> test.d(5,18): Error: array operation a1[] * 3 without destination memory not allowed
>>
>> Is it a good idea to support something like line 5?
>>
>> Bye,
>> bearophile
>
> Why not? Looks fine to me, you've explicitly allocated some new memory for the result to sit in, the lack of which is the motivation for lines 3 an 5 being errors.

To clarify: 4 is good, 3 and 5 are not.