[relay-compiler] @preloadable Support

[tobias-tengler]

This is based on an earlier implementation done by @tomgasson in #4110

Companion PR for @types/relay-runtime: DefinitelyTyped/DefinitelyTyped#68144

An example application using this can be found here

[alunyov]

Thank you for working on this!

[tobias-tengler]

@alunyov this would be ready to go from my side.

I'd like to add tests, but it seems like there is no existing test setup for generate_extra_artifacts. Is there something that could be pulled into OSS from the internal Relay version or what would be the best way to go about writing tests in your opinion?

[tobias-tengler]

@alunyov would you mind taking another look at this? It would be incredibly valuable for us to have this feature :)

[alunyov]

@tobias-tengler sorry for the delay. I'll try to get to this next week

[alunyov]

Overall, a very clean PR! Thank you for adding this.

Can you add a test for this? I think we have a way to add integration tests now: https://github.com/facebook/relay/tree/main/compiler/crates/relay-compiler/tests/relay_compiler_integration

[tobias-tengler]

@alunyov Added an integration test for each language option and fixed the eagerEsModules handling. Could you kindly give it another look? :)

[alunyov]

This looks great!

[alunyov]

@tobias-tengler lets fix these rust formatting issues, and I'll import this internally.

[tobias-tengler]

@alunyov fixed :) It's weird that my editor doesn't format correctly. I'm using VS Code, rust-analyzer and I'm working from within the compiler directory. Normally the editor should format this correctly or am I missing something? I often had to fix formatting by hand with this project...
The formatting job in the CI is using a (different) nightly version of Rust than the build. I added a rustup override for 1.72.0 for the compiler as that's what's being used to build the compiler in CI. With the nightly version from the lint job I can't build the project unfortunately.

[tobias-tengler]

Do you have an idea why this fails? https://github.com/facebook/relay/actions/runs/7452983745/job/20277398152?pr=4515

[alunyov]

Do you have an idea why this fails? https://github.com/facebook/relay/actions/runs/7452983745/job/20277398152?pr=4515

Did you run https://github.com/facebook/relay/blob/main/scripts/update-fixtures.sh to generate the fixture tests?

[alunyov]

@alunyov fixed :) It's weird that my editor doesn't format correctly. I'm using VS Code, rust-analyzer and I'm working from within the compiler directory. Normally the editor should format this correctly or am I missing something? I often had to fix formatting by hand with this project... The formatting job in the CI is using a (different) nightly version of Rust than the build. I added a rustup override for 1.72.0 for the compiler as that's what's being used to build the compiler in CI. With the nightly version from the lint job I can't build the project unfortunately.

Sorry about that... We have a weird internal rust configuration, that we somehow is trying to replicate in OSS.

cc @captbaritone - were there any discussion with internal Rust team on how to properly setup OSS rust projects that needs to be synced internally?

[tobias-tengler]

Oh f*** me 😂 I totally wrote them by hand, whoops.

[facebook-github-bot]

@alunyov has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[alunyov]

@tobias-tengler somehow the query text is missing in the operations - this is maybe related to the way the integration test prints these files...

[tobias-tengler]

This is the input. The local persister throws if the file doesn't already exists. This is just here to satisfy the persister.

[alunyov]

@tobias-tengler you'll also need to update integration test setup, to include the default extra generator to the test config

[facebook-github-bot]

@alunyov has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[facebook-github-bot]

@alunyov merged this pull request in 40fe615.

[alex-statsig]

Can / should this support mutations?

I've noticed mutations often add up to a large chunk of bundle sizes (sometimes due to including ...EntirePage to ensure data is all refetched). It seems like the AST shouldn't be needed immediately for mutations, and committing the mutation with appropriate variables could be decoupled (especially with persisted queries removing the need for the entire query text in the bundle).

It's not quite the same flow as preloading queries, but a similar idea I think to prioritize the part needed to issue the request (and defer the part to process the response). It seems especially applicable since mutations are not used more often than they are used.