I’m rendering a grid of isometric tiles with each tile being 64x32 in size. Some of the structures I’m rendering span over multiple tiles graphically and I’m wondering how these would be stored in a data structure, or even which data structure I should use. Currently I’ve got a 2D array where all tiles are floor-level, although I’m now trying to render a building and it’s presented a few questions:
- Are larger structures stored on multiple or single tiles?
- If they’re stored on single tiles how would collision detection work, as you could collide with a tile that’s not represented in the tile data for the collided tile?
- If they’re stored on multiple tiles and each tile just renders it’s piece of the structure, how would you render the upper parts of the structure (e.g. the upper floors of a building)?
If every floor also has to be represented in the tile data, wouldn’t this make a really tall impassible column on your level, when in reality you should be able to walk behind the building once the character is above the base level?
Any guidance would be helpful. 
Thanks