- Generally I'll be editing one or more layers of tilemap and use the tile indices in my engine to sort out collisions if at all possible.
- If it makes sense for the project, I'll place markers for enemy spawn positions or whatever as point objects or rects or polys for areas, polylines for patrol routes - but I generally don't "dress them up" like I've seen others do; I just set the name of the point
- I'm afraid this depends on what the engine is meant to do :)
- General purpose = support as much as possible.
- Definitely need rect tile layer support and support for getting the information out of different types of objects
- Infinite map mode still has an origin set in the file IIRC which should be used for positioning the tilemap and avoiding offsets. You'd have to keep the map as chunks in-engine.
- Here's an example from arco because it's what I have on hand. It's not necessarily "representative" other than, no real concrete convention between games.
In general people's expectations from tilemaps seem incredibly varied. For KAG and Trench Run we just loaded from png image maps and this was popular with the community because its very easy to edit a png image; but it doesn't allow sub-tile offsets. Some games will just use one layer and will set up all the other stuff with hardcoded stuff in engine. Some games will use all the advanced object prototype stuff. Expect to get wildly different responses if the net falls wide enough :)
Hope it helps