Picking up all tiles/cells/objects from isometric tilemap under mouse selection area.

[stevenmare]

So I’ve a bit of an interesting problem with isometric tilemaps in Unity.
I have a two dimensional array of gameobjects. The position of the objects in this array determines where their associated sprites are rendered on an isometric tilemap.

I can then currently click on a position on the tilemap and get an associated gameobject in two ways.
Use

var worldPos = Camera.main.ScreenToWorldPoint(Input.MousePosition);
var tilePos = sampleTilemap.WorldToCell(worldPos);

To get the tile/cell position, then use that position to get the object from the array.

Give the object a Collider2d in world space and use

var worldPos = Camera.main.ScreenToWorldPoint(Input.MousePosition);
var hit = Physics2D.OverlapPoint(worldPos);

to get the collider that has reference to the associated gameobject.

But the question now comes is what to do if I want to mouse select an area to get multiple objects at once?
First solution is to give every gameObject an associated Collider2D shape and then use

var hits = Physics2D.OverlapArea(startPos, endPos);

But as the scene has over 10,000 trees that’s a lot of colliders to instantiate on scene load and should I be worried about memory usage?

Secondly I could get the tile/cell positions that lie under the selection area using the following and some algebra.

if (Input.GetMouseButtonDown(0))
{
    _startPos = InputController.MousePosition;
}

if (Input.GetMouseButton(0))
{
   _endPos = InputController.MousePosition;

   var startWorldPos = Camera.main.ScreenToWorldPoint(_startPos);
   var endWorldPos = Camera.main.ScreenToWorldPoint(_endPos);

//get corner tiles
//algebra to get all included tiles in selection area

To get the below image.

But is that a lot of work to be doing during one update frame?

If anyone has a more elegant solution or more widely used solution I would gladly appreciate it.

[Lo-renzo]

Your solution is probably fine and may not be worth optimizing. There’s no Unity API to do those tests for you.

You’re probably already doing this roughly:

1. Get the red RectInt of min/max cells.
   2. Test every cell whether it’s in the blue box.

I’ll describe an alternative testing for #2 that you can compare against your current solution:

1. Convert corners of blue box into grid/cellspace. This skips worldspace conversions later.
2. With those 4 points (ABCD), create two triangles: ABC, BCD.
3. Iterate your RectInt (red box) to test the centerpoint (+.5, +.5) of each cell against those two triangles. Compare the center point of each red box cell against two triangles.

How to cheaply test whether a point is in triangle? This is a simple way.

    public static bool PointInTriangle_Fast(Vector2 pt, // tested point
   Vector2 v1, Vector2 v2, Vector2 v3) // triangle points
    {
        static float GetSign(Vector2 p1, Vector2 p2, Vector2 p3)
        {
            return (p1.x - p3.x) * (p2.y - p3.y) - (p2.x - p3.x) * (p1.y - p3.y);
        }

        float s1 = GetSign(pt, v1, v2);
        float s2 = GetSign(pt, v2, v3);
        float s3 = GetSign(pt, v3, v1);

        bool neg = (s1 < 0) || (s2 < 0) || (s3 < 0);
        bool pos = (s1 > 0) || (s2 > 0) || (s3 > 0);

        return !(neg && pos);
    }

If the triangle test looks like less work than what you’re currently doing algebraically then maybe it’s better.