Hello.

In D1, this code fails:

void foo(S)(ref S s){
  foreach(ref k; s.tupleof){
    k = 1;
  }
}

struct K{
  int g;
}

void main(){
  K k;
  foo(k);
  assert(k.g == 1);
}


test.d(5): Error: no storage class for value k

(referring to 'k = 1;')

Is this an expected error, and is there a good way to get the semantics to actually work?

On 3/7/10 19:11, Ellery Newcomer wrote:
> Hello.
>
> In D1, this code fails:
>
> void foo(S)(ref S s){
> foreach(ref k; s.tupleof){
> k = 1;
> }
> }
>
> struct K{
> int g;
> }
>
> void main(){
> K k;
> foo(k);
> assert(k.g == 1);
> }
>
>
> test.d(5): Error: no storage class for value k
>
> (referring to 'k = 1;')
>
> Is this an expected error, and is there a good way to get the semantics
> to actually work?

It's a bug: http://d.puremagic.com/issues/show_bug.cgi?id=2411
You can use the following code as a workaround:

foreach (i, dummy ; s.tupleof)
	s.tupleof[i] = 1;

On 03/07/2010 12:23 PM, Jacob Carlborg wrote:
> On 3/7/10 19:11, Ellery Newcomer wrote:
>> Hello.
>>
>> In D1, this code fails:
>>
>> void foo(S)(ref S s){
>> foreach(ref k; s.tupleof){
>> k = 1;
>> }
>> }
>>
>> struct K{
>> int g;
>> }
>>
>> void main(){
>> K k;
>> foo(k);
>> assert(k.g == 1);
>> }
>>
>>
>> test.d(5): Error: no storage class for value k
>>
>> (referring to 'k = 1;')
>>
>> Is this an expected error, and is there a good way to get the semantics
>> to actually work?
>
> It's a bug: http://d.puremagic.com/issues/show_bug.cgi?id=2411
> You can use the following code as a workaround:
>
> foreach (i, dummy ; s.tupleof)
> s.tupleof[i] = 1;

Totally awesome! Thanks a float.infinity!

On Sun, Mar 7, 2010 at 19:53, Ellery Newcomer <ellery-newcomer@utulsa.edu>wrote:

> On 03/07/2010 12:23 PM, Jacob Carlborg wrote:
>
>> It's a bug: http://d.puremagic.com/issues/show_bug.cgi?id=2411 You can use the following code as a workaround:
>>
>> foreach (i, dummy ; s.tupleof)
>> s.tupleof[i] = 1;
>>
>
> Totally awesome! Thanks a float.infinity!
>

I remember spending *weeks* (well, OK, *hours*) trying to find a way to do such a mapping after trying the first foreach, and getting some recursive template to work... Only to find this static foreach trick, looking at Phobos code (for D2) some month after. It sent ripples in all my code...

As a side note, it works for any tuple: those created by .tupleof, but also the variadic template parameters, the expansion of a std.typecons.Tuple (D2), those created by templates, etc. So you can do:

void reinitialize(T...)(ref T ts) // Variadic template
{
    foreach(i, Type; T) ts[i] = T[i].init; // iterating on the components of
T.
}

int a = 1;
char c = 'a';
string s = "abc";

reinitialize(a,c,s);

assert(a = 0); // int.init
assert(c =  char.init); // is there a literal for char.init? '' doesn't
work.
assert(s="");


Philippe

Philippe Sigaud:
> assert(c =  ); // is there a literal for char.init? '' doesn't
> work.

I think it's '\xFF' but I don't know if it may change in future, so using char.init is probably better, more readable, more explicit in its purpose, and less prone to typing bugs.

Bye,
bearophile

On Mon, Mar 8, 2010 at 22:11, bearophile <bearophileHUGS@lycos.com> wrote:

> Philippe Sigaud:
> > assert(c =  ); // is there a literal for char.init? '' doesn't
> > work.
>
> I think it's '\xFF' but I don't know if it may change in future, so using char.init is probably better, more readable, more explicit in its purpose, and less prone to typing bugs.
>
> Thanks. It's an invalid char, as NaN for floats?