BansheeEngine/SceneView.txt

TODO:
 - Core thread gets stuck on shutdown when OpenGL is used...Somewhere in kernel

CONCRETE TASK:
 - Similar to how I have onRenderViewport callback in Renderer have another one that gets triggered from core thread
   - Hook up gizmo rendering there
 - Hook up gizmo manager to ScenePicking so gizmos are considered when picking
   - I'll likely need to update GizmoManager so I can query gizmo SceneObject based on gizmo index
 - Selection/ScenePicking/GizmoManager need to be started

IMMEDIATE:
 - SceneGrid is very ugly. Consider using default lines for now and come back with a better approach later.
   - Potentially enable line AA?
 - Picking code is completely untested and will likely need major fixing
 - Disable DX9 for editor as I will likely want to use geometry shaders for icon rendering, and possibly new AA line shader
   - Or just use MeshHeap and update the icon/lines every frame?
 - Test all the new DrawHelper3D methods

 GIZMO TODO:
  - IMPORTANT: Gizmo rendering happens in update() but it should happen whenever scene view is being rendered as the render target isn't set anywhere
  - Figure out how to deal with builtin components like Camera and Renderable (e.g. how will they have gizmos since they're not managed components?)

LATER:
 - Need a way to render text for gizmos and handles, and in scene in general

----------------------------------------------------------------------
Handles

SliderLine - position, direction, length
  - When initially activated it records position nearest so the line as the starting point
  - Further mouse dragging also finds nearest position to the line
  - Difference between those two results in a float value (how much to move along direction from position to reach new position)
  - Slider line has a capsule + sphere collider size of which can be set manually

SliderPlane - position, normal, size
  - Similar to line slider only the direction is determined dynamically as well as distance
  - Outputs a Vector2 (direction * distance moved)
  - A OOB is used as a collider

SliderDisc - position, normal, radius
  - When initially activated it records position nearest so the disc as the starting point
  - Further movement calculates the dynamic direction from the starting point to the current point on the plane the disc lies on
  - Distance along that direction is returned as amount of movement (similar to line slider)
  - Outputs a single float
  - A torus is used as a collider

Handles are always the same size regardless of the distance from camera. (Use same code as from gizmo rendering?)
These three types can be used for creating MoveHandle, RotationHandle, ScaleHandle
  - I can potentially move the colliders out of the sliders and add them to the *Handle classes as well
  - Handle classes will also handle the rendering (using the existing DrawHelper methods)

CONCRETE TASKS:
 - Need to add capsule, torus and OOB colliders
    - They need ray intersection code
 - Line, plane and disc need code for finding nearest point to a ray

Think about C# implementation.
Take into consideration local vs. global handles
Free move/rotate/scale handles need to exist as well
 - Scale is easy, just perform uniform scale. Use SliderPlane oriented towards camera
 - Move also use SliderPlane oriented towards camera
 - Rotation use SliderDisc oriented towards camera

See for inspiration: http://docs.unity3d.com/ScriptReference/Handles.html

----------------------------------------------------------------------
SelectionRenderer

Retrieve a list of selected objects from SelectionManager
Find ones with Renderable components
Retrieve Meshes, and world transforms from them
Draw that same mesh with either a wireframe or a grayed out shader with a slight depth bias

----------------------------------------------------------------------
SceneView editor flow:
  Hook up gizmo, handle and selection rendering methods to be executed after the scene is rendered
  Calculate mouse coords manually relative to the window and to the render texture GUI element
     - Don't use GUI events as we require more precise control (do we?)

  Detect mouse clicks on the scene render target
      Forward those mouse coordinates to HandleManager
      It checks if screen ray intersects any handles and returns the handle if it does
         If handle is found it is activated and method returns
         Otherwise we mark the coordinates as selection start

  Detect mouse drag on the scene render target
    - If we have an active handle
         Forward mouse coordinates to the active handle so it can do its thing
         return
    - Otherwise its assumed we are dragging a selection
         Update selection endpoint and send it to ScenePicking
         Use Selection to select picked objects if any
         return

  Detect mouse release on scene render target
     If we have an active handle
        Clear active handle
        return
     Otheriwse its assumed we are dragging a selection
        Do nothing
        return