How array concatenation works... internally

[Dnewbie via Digitalmars-d-learn]

Hi,

I'd like to understand how array concatenation works internally, 
like the example below:


//DMD64 D Compiler 2.072.2
import std.stdio;
void main(){
string[] arr;
arr.length = 2;
arr[0] = "Hello";
arr[1] = "World";
writeln(arr.length);
arr = arr[0..1] ~ "New String" ~ arr[1..2];
writeln(arr.length);
foreach(string a; arr){
writeln(a);
}
}
http://rextester.com/DDW84343

The code above prints:
2
3
Hello
New String
World


So, It changes the "arr" length and put the "New String" between 
the other two. It's very fast with some other tests that I made.


Now I'm curious to know what's happening under the hood. It's 
related to memcpy?


On Phobos "array.d" source I've found this:

/// Concatenation with rebinding.
void opCatAssign(R)(R another)
{
auto newThis = this ~ another;
move(newThis, this);
}

But now I'm having problem to find how I can reach this "move" 
function, since I couldn't find any "move" on the "std" folder.


Thanks in advance.

Re: How array concatenation works... internally

[Dnewbie via Digitalmars-d-learn]

On Monday, 23 April 2018 at 23:15:13 UTC, Dnewbie wrote:

It's related to memcpy?


By the way... It's related to realloc and memcpy?

Re: How array concatenation works... internally

[Steven Schveighoffer via Digitalmars-d-learn]

On 4/23/18 7:15 PM, Dnewbie wrote:

Hi,

I'd like to understand how array concatenation works internally, like 
the example below:


//DMD64 D Compiler 2.072.2
import std.stdio;
void main(){
     string[] arr;
     arr.length = 2;
     arr[0] = "Hello";
     arr[1] = "World";
     writeln(arr.length);
     arr = arr[0..1] ~ "New String" ~ arr[1..2];
     writeln(arr.length);
     foreach(string a; arr){
     writeln(a);
     }
}
http://rextester.com/DDW84343

The code above prints:
2
3
Hello
New String
World


So, It changes the "arr" length and put the "New String" between the 
other two. It's very fast with some other tests that I made.


What it has done is built a completely new array, with the new 3 
elements. The old array is still there. You can witness this by keeping 
a reference to the old array:


auto arr2 = arr;
arr = arr[0 .. 1] ~ "New String" ~ arr[1 .. 2];
writeln(arr2);
writeln(arr);

Now I'm curious to know what's happening under the hood. It's related to 
memcpy?


There's not much copying going on here, each string is stored in the arr 
as a pointer and length pair. So you are just making a copy of those.



On Phobos "array.d" source I've found this:

     /// Concatenation with rebinding.
     void opCatAssign(R)(R another)
     {
     auto newThis = this ~ another;
     move(newThis, this);
     }


This is different. The builtin arrays are not part of phobos, they are 
defined by the compiler. std.array.Array is a different type.


If you want to know more about the array runtime, I suggest this 
article: https://dlang.org/articles/d-array-article.html


But now I'm having problem to find how I can reach this "move" function, 
since I couldn't find any "move" on the "std" folder.


Move is std.algorithm.move: 
https://dlang.org/phobos/std_algorithm_mutation.html#.move


-Steve

Re: How array concatenation works... internally

[Dnewbie via Digitalmars-d-learn]

On Monday, 23 April 2018 at 23:27:17 UTC, Steven Schveighoffer 
wrote:

...
If you want to know more about the array runtime, I suggest 
this article: https://dlang.org/articles/d-array-article.html

...


Thanks for replying and this article is what I was looking for.