Why is D unpopular, redux.

[FeepingCreature]

On Tuesday, 24 May 2022 at 07:45:38 UTC, Timon Gehr wrote:
> On 24.05.22 07:31, Walter Bright wrote:
>> 
>>> https://issues.dlang.org/show_bug.cgi?id=2043
>> 
>> We have better tools to deal with 2043 today. Note that dgList 
>> exceeds the lifetime of b, and should fail to compile for that 
>> reason.
>> 
>> Let's try it (adding import and @safe and updating the code to 
>> D2):
>> ----
>> import std.stdio;
>> 
>> @safe:
>> 
>> void delegate()[] dgList;
>> 
>> void test() {
>> 
>>          foreach(int i; [1, 2, 3]) {
>>                  int b = 2;
>>                  dgList ~= { writeln(b); };
>>                  writeln(&b); // b will be the *same* on each 
>> iteration
>>          }
>> }
>> ----
>> 
>>  > dmd test.c -preview=dip1000
>> test.d(13): Error: reference to local variable `b` assigned to 
>> non-scope parameter `_param_0`
>> 
>> Looks like we can close it now!
>
> No. The correct behavior when a delegate literal outlives 
> captured variables is to allocate a closure. This case is not 
> an exception...

This is a bit of a controversial topic. Personally I agree that 
there should be a closure allocated every loop pass, but *every 
other language* allocates them for the whole stackframe, so... 
But then handling b is awkward, because the way that compilers 
usually square that circle is that they allocate a separate spot 
in the stackframe *per variable declaration*, so the loop 
variable only gets one distinct declaration. And it's hard 
otherwise, because if you allocate a separate frame every loop, 
then they can't reference each others' variables. Some languages 
solve this by making lambdas unable to write scope variables, but 
then obviously they're less useful. Anyway, so in this case the 
delegate can't help outlive b *even if* it's closure allocated.