This issue does not exist to argue the merits of Resources as Components, for some of the rationale we refer to this HackMD (this really should've been an RFC, huh?). This issue exists to track the progress on implementation and to reference in code when it is only partially working. If there are steps you feel like should be added, or if there are problems with this approach, please let me know.

Implementation Plan

This is a tracking issue for the Resources as Components. This outline is roughly copied from #17485.

• Test robustness, to be less reliant on specific entity counts (Improve test robustness #20248, Adjust benches to be more robust #20207) . We deprecate world.iter_entities() and world._iter_entities_mut() as part of that effort Deprecate iter_entities and iter_entities_mut #20260.
• Reintroduce iter_entities() Revive iter_entities() when you only have immutable World access ❤️‍🩹  #22864.
• Implement resources as components through a custom #[derive(Resource)] macro Store Resources as components on singleton entities #20934.
• Gut ReflectResource Store Resources as components on singleton entities #20934.
• Rename non-send resources to non-send data Store Resources as components on singleton entities #20934.
• Remove resource/component redundancies in component registration Remove component/resource redundancies #22930.
• Add component features to the resource macro.
• Change clear_all to also remove non-send data clear_all() clears all (non-send data) #22911 .
• Eliminate resources from Access, Remove resources from Access #22910 .
• Remove resources from bevy_scene Remove resources from scenes #22968 .
• Create a cleaner component_scope / resource_scope implementation.