I'm missing something about the way associative arrays are allocated.
In the example below, case 1 and case 2 seem to be the same, a type indexed by a long.
Case 2 fails with a Range violation.

Can someone explain the difference?
I found a work around (or the correct way) in case 3.

struct Project
{
	string  date;
	string  name;
}

		// case 1
		string[long] a1;
		a1[10] = "testing a1";

		foreach (f; a1)
			writeln ("name ", f);

		// case 2
		Project[long] a2;
		a2[10].name = "testing a2";

		foreach (f; a2)
			writeln ("name ", f.name);

		// case 3
		Project*[long] a3;
		a3[10] = new Project;
		a3[10].name = "testing a3";

		foreach (f; a3)
			writeln ("name ", f.name);

On Friday, 1 January 2021 at 01:43:50 UTC, Chris Bare wrote:
> 		a1[10] = "testing a1";

this actually constructs a new thing since it is a straight x = y assignment.

> 		a2[10].name = "testing a2";

But this looks up something first. It doesn't construct a2[10], it looks it up first to fetch the name member... hence the range violation since it isn't constructed yet.


> 		a3[10] = new Project;

and this again will construct since it is a straight x = y again.



So you could also do

a2[10] = Person()

then do the lookup of name

On 12/31/20 8:54 PM, Adam D. Ruppe wrote:
> On Friday, 1 January 2021 at 01:43:50 UTC, Chris Bare wrote:
>>         a1[10] = "testing a1";
> 
> this actually constructs a new thing since it is a straight x = y assignment.
> 
>>         a2[10].name = "testing a2";
> 
> But this looks up something first. It doesn't construct a2[10], it looks it up first to fetch the name member... hence the range violation since it isn't constructed yet.
> 
> 
>>         a3[10] = new Project;
> 
> and this again will construct since it is a straight x = y again.
> 
> 
> 
> So you could also do
> 
> a2[10] = Person()
> 
> then do the lookup of name

or a2[10] = Person(null, "testing a2")

-Steve