I have worked on a little FOSS project (I will ask later for code review), I have noticed some limitation on 2d arrays. I haven't found the answers on the forum

1. I can not initialize a 2d array in the body of the class,

class A
{
  Stuff[][] array2d = Stuff.ONE;
}

One must do that in the ctor

2. The following code yields
   Error: cannot implicitly convert expression 3 of type int to int[20][]

import std;
void main()
{
    writeln("Hello D");
    int[20][20] array2d;
    array2d = 3;
}

What I would expect, of course, is to automatically iterate through the entire 2d array and set each (i,j) pair to "3".

3. Similarly, it would be nice to have a foreach as follow:

foreach ( cell; array2d)// where i = 0 for all j then i = 1 for all j, etc

I guess I can build smthg with opApply but...
Could even be something like

foreach (i, j, cell; array2)

THose are just some remarks. Nice to have stuffs, if someone is looking that way. Thanks

On Saturday, 1 January 2022 at 00:19:56 UTC, chopchop wrote:

class A
{
  Stuff[][] array2d = Stuff.ONE;
}

If the size is statically known you probably can, however you'd be doing it as CTFE.

struct Chess {
  static int[8][8] grid = makeGrid();

  auto makeGrid() {
    int[8][8] tmp;
    //build grid stuff
    return tmp;
  }
}

I forget if class variables can be instantiated by default in a similar way or not (like working in Java).

import std;
void main()
{
    writeln("Hello D");
    int[20][20] array2d;
    array2d = 3;
}

Trying array2d array2d[] array2d[][] all result in the same error. I do agree being able to bulk fill or zeroize might be nice in some cases. But it's just a little more typing to get the same result.

I wonder if a template mixin would fix this.

foreach(ref y; array2d)
  y[]=3;

foreach ( cell; array2d)// where i = 0 for all j then i = 1 for all j, etc

So like this? Alternatively iota might also do the job.

foreach(i, ref j; array2d)
  j[]=i;

foreach (i, j, cell; array2)

Hmmm. Default behavior probably is better as is. Consider you can alias stuff to build larger stuff. So let's assume i am doing some chess stuff.

 alias cell=int;
 alias row=cell[8];
 alias grid=row[8];

 grid board;

Now let's assume i did something like... move piece and reference it wrong.

auto movePiece(int x,int y,int x2,int y2) {
  //do checks
  board[x2][y2]=board[x][y];

  //clear piece
  board=0; //should be board[x][y]=0;
}

With default behavior and being more explicit you'd catch a simple bug instead of tracing your code for what compiled correctly. it's one of the reasons assignment in if statements is disallowed as it's easy to mess up. Same for forcibly casting rather than assuming down-casting or other is fine.

On Saturday, 1 January 2022 at 00:19:56 UTC, chopchop wrote:

That's https://issues.dlang.org/show_bug.cgi?id=19178

I've started working on a fix, but haven't finished it yet.

On Saturday, 1 January 2022 at 11:57:41 UTC, Dennis wrote:

Thanks Dennis for your work.