Optimise tile loading

[BeebBenjamin]

To reduce the load on our tiles server for a potential public exposure of the plug-in we need to refactor the ThreeCanvas to only load in tiles that have been loaded into Mirador i.e. that are visible on screen for that zoom level and then load them in as they are also added to Mirador.

I would need to grab the following:

1. the current size of the picture
2. what tiles are present on screen, their positions etc
3. the intersection of the whole image on screen

Then from the ThreeCanvas

1. offest camera based on the size and zoom of picture
2. change the textures and mesh tiles (their sizes, numbers, and positions) depending on the zoom level and pan

[BeebBenjamin]

Notes

• viewer.world controls all the tiles.
• _items refer to the layers, so if there are 7 layers loaded, there will be 7 items in the _items array.
• each item in _items then has a source object containing critical tile information, and a tileMatrix containing information on the tiles loaded onto the screen (this object will update as you pan).
• the index of tileMatrix corresponds exactly to the zoom level you are at, so for an image with 5 possible scales, 5 will be maximum resolution/size of the tiled image.
• In theory the tile data could be grabbed from the _tile_cache but in reality I think the data is deleted once it has been drawn to the HTML canvas.
• What I think we need to do, is add a handler for when the tiles are drawn that grabs the data and feeds it to the ThreeCanvas, otherwise it is lost. How this might perform in real time is another question.

[BeebBenjamin]

First draft of the implementation of this is ready, but has some bugs to iron out. One issue is when the renderer is activated at the max tileLevel, the previous tileLevel has not loaded in yet and the layer is completely transparent to the default Mirador image below.

[BeebBenjamin]

On QA for testing

[BeebBenjamin]

Looks like it works for now, so closing.