bevy_entitiles 0.2.2

A tilemap library for bevy. With many algorithms built in.
docs.rs failed to build bevy_entitiles-0.2.2
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Visit the last successful build: bevy_entitiles-0.12.0

Bevy EntiTiles 🗺️

A tilemap library for bevy. With many algorithms built in.

This repo is under maintenance as long as this message exists. ( Hope this message can bring you peace of mind. Yeah, that's childish :p )

Notice that the following progress may not up to date. Check the README.md in dev brach to get the latest progress!

Currently Working On

  • Tilemap Serializing

Future Goals

  • Wave Function Collapse ( Optimization; Collapse map patterns )
  • Pathfinding
  • Tilemap-Link
  • Runtime Mesh & Texture Baking
  • Tilemap Serializing
  • Chunk Unloading
  • Wang Tilling
  • Tilemap Mask
  • Frustum Culling ( Optimization )
  • Physics
  • Tilemap Editor

Delayed

  • SSAO ( Personal capability )
  • Volumetric Clouds / Fog ( Personal capability )
  • Lighting ( Personal capability )
  • Realtime Shadow ( Personal capability )

Known Issues

  • The success probability of the wfc algorithm significantly decreased after switching to LookupHeap. ( And that's why I didn't switch to LookupHeap )
  • Program panics when load a tilemap twice.

Features

Flag Funtionality
algorithm Implementation of algorithms
debug Show some debug info including aabbs for chunks and tilemaps, path finding results etc.
debug-verbose Show verbose process of functions. This feature is only useful when debug algorithms. But it's useless for most of you.
physics_rapier Physics support for bevy_rapier
physics_xpbd Physics support for bevy_xpbd, like setting colliders etc.
serializing Save and load the tilemap from files. Also contains tools for upgrading files.

Performance

Platform: 10600KF

Frustum Culling

Bevy 0.11.3, crate 0.1.1, 1000x1000 tiles

Pathfinding

The pathfinding algorithm is very fast.

Notice this tests are done with synchronized pathfinding. Which means whole algorithm will figure the path out in one frame. But since 0.2.1, we support the asynchronized one. The algorithm can complete a part of the pathfinding and continue it in the next frame. This will make it even smoother.

Size Time(avg of 3 tests) ms
100x100 12.00
500x500 295.67
1000x1000 1384.33

Bevy 0.12, crate 0.2.1, using pathfinding example

Wave Function Collapse

In the following case, each tile has at least one corresponding color gap with its neighboring tiles.

Video at docs/vids/wfc.mp4

Size Time(avg of 3 tests) ms Time(avg of 3 tests) ms
10x10 33.312 16.264 (3)
20x20 490.950 96.009 (3)
30x30 2,280.121 335.697 (6)
50x50 18,838.542 2,095.428 (8)
100x100 (Not measurable) 32,309.045 (16)

Column 1: Bevy 0.11.3, crate 0.2.0, NoneWeighted; Column 2: Bevy 0.12, crate 0.2.1, NoneWeighted, max_retrace_factor = number in parentheses

Special Thanks

  • SSAO & Volumetric Clouds / Fog inspired by this video.

  • bevy_ecs_tilemap. I took this crate as the reference and learnt the basis of bevy rendering.

Versions

Bevy ver EntiTiles ver
0.12.x 0.2.x
0.11.x 0.1.x