Should we keep Query::get_component and friends?

[nicopap]

Query has methods to access individual components of the query for a specific entity.

But is it really useful? One can simply use Query::get and access the component they want.

Those methods is code we need to take care of. This is not only a risk in terms of soundness (since it touches fairly hairy parts of the ECS) but it's also a burden. For example, you need to implement a get_component method when implementing dynamic queries. We also need to document them and test them, add similar API to QueryState, add component (panicking version), etc.

If we remove them, we don't have to worry about any of what we mentioned above.

Examinees

• Query::get_component
• Query::get_component_mut
• Query::get_component_unchecked

Alleged usefulness

• For generic APIs where you have some Q: WorldQuery and don't know what components it has/what position they might be in.

That's it? I've implemented a lot of bevy plugins and personally never came around a use case for get_component, but if you can give a compelling use-case I'll add it to the list.

[james7132]

A comment I made on Discord that is worth reposting here: More often than not, I've seen people just use Query::get and just discard the unused parts, and the compiler optimizes out the unused fetches. It's definitely less ergonomic, but the fetches in Query::get are 100% in alignment with our safety checks used in iteration.

[B-Reif]

I would imagine any use case for these methods would be better-served by dynamic queries.