I have this struct:

struct Deb {
    string name;
    ...
    Unit[string] tags; // set of tags

    Deb copy() const {
        Deb deb;
        ...
        foreach (key, value; tags) // XXX
            deb.tags[key] = value;
        return deb;
    }

    void clear() {
        name = "";
        ...
        tags.clear;
    }

...
}

I am populating an AA with these structs:

Deb[string] debForName;

I'm using this approach (pseudo-code):

Deb deb;
foreach (datum; data) {
    populateDeb(datum, deb);
    debForName[deb.name] = deb.copy; // YYY
    deb.clear;
}

(1) XXX Is there a nicer way to copy an AA?
(2) YYY Is there a nicer way to copy a struct? Use a copy constructor or implement a dup or idup?

On 3/9/20 9:23 AM, mark wrote:
> I have this struct:
> 
> struct Deb {
>      string name;
>      ...
>      Unit[string] tags; // set of tags
> 
>      Deb copy() const {
>          Deb deb;
>          ...
>          foreach (key, value; tags) // XXX
>              deb.tags[key] = value;
>          return deb;
>      }
> 
>      void clear() {
>          name = "";
>          ...
>          tags.clear;
>      }
> 
> ....
> }
> 
> I am populating an AA with these structs:
> 
> Deb[string] debForName;
> 
> I'm using this approach (pseudo-code):
> 
> Deb deb;
> foreach (datum; data) {
>      populateDeb(datum, deb);
>      debForName[deb.name] = deb.copy; // YYY
>      deb.clear;
> }
> 
> (1) XXX Is there a nicer way to copy an AA?

aa.dup. However, this isn't going to work in a const member function (though technically there shouldn't be anything wrong with it).

> (2) YYY Is there a nicer way to copy a struct? Use a copy constructor or implement a dup or idup?

I would name it dup instead of copy for consistency with D. A copy constructor is pretty heavy for a struct to do a complete duplication of the AA. You should have to opt-in to that.

-Steve

On Monday, 9 March 2020 at 14:45:15 UTC, Steven Schveighoffer wrote:
> On 3/9/20 9:23 AM, mark wrote:
>> I have this struct:
[snip]
> I would name it dup instead of copy for consistency with D. A copy constructor is pretty heavy for a struct to do a complete duplication of the AA. You should have to opt-in to that.
>
> -Steve

Thanks, I'd already realised I ought to rename it .dup.

Also, I only actually need to create it from the keys since I'm using the tags AA as a set so every value is the same. I did try tags.dup but got this:

src/deb.d(25,24): Error: cannot implicitly convert expression dup(this.tags) of type const(void[0])[string] to void[0][string]

On 3/9/20 11:49 AM, mark wrote:
> On Monday, 9 March 2020 at 14:45:15 UTC, Steven Schveighoffer wrote:
>> On 3/9/20 9:23 AM, mark wrote:
>>> I have this struct:
> [snip]
>> I would name it dup instead of copy for consistency with D. A copy constructor is pretty heavy for a struct to do a complete duplication of the AA. You should have to opt-in to that.
>>
> 
> Thanks, I'd already realised I ought to rename it .dup.
> 
> Also, I only actually need to create it from the keys since I'm using the tags AA as a set so every value is the same. I did try tags.dup but got this:
> 
> src/deb.d(25,24): Error: cannot implicitly convert expression dup(this.tags) of type const(void[0])[string] to void[0][string]
> 

Right, because it's inside a const member function, the tags member is const.

Right now, it looks like dup does not work even for value types (e.g. const int[int] -> int[int]) let alone types with indirection (it should also work for string keys). If there's not an issue already, someone should file it.

-Steve