feat(@graphql-codegen/core): Non-async codegen function interface

[stocaaro]

Is your feature request related to a problem? Please describe.

I'm working on augmenting the aws-amplify library to generate data interfaces for lambda functions to consume, which requires running codegen during the infrastructure build step in CDK. CDK doesn't support awaiting async calls related to concerns about keeping definitions deterministic.

Would it be possible to support running codegen synchronously / without the Promise wrapper?

Reviewing the implementation, it appears codegen is async to support async profiler, cache and plugin implementations. Our current usage doesn't include async behavior for any of these, but there's no way for me to decouple the async wrappers baked into the library behavior.

Our customers schema is composed during CDK execution, which we want to codegen so their projects can depend upon an updated schema directly in their lambdas.

Describe the solution you'd like

I've experimented with making/testing this change to codegen, offering a codegenSync variant implementation with the needed types.

This would either support profiler, plugin and cache's that are sync only, or strip these features out if they only make sense as async behavior.

If an option like this would be considered for merge into the library, I could turn my prototype into a draft PR.

Describe alternatives you've considered

• I have explored using execa to perform this async behavior in a separate process.
• I have explored introducing async behavior into the cdk constructs that compose the schema.

Any additional important details?

No response

[saihaj]

Why can you not run codegen before running CDK build?

[stocaaro]

We looked into that. The graphql schema we're running codegen against isn't finalized until the cdk artifacts are being composed, which supports having multipe schema's combined into a single API surface.

We have been running codegen after the cdk build was complete, however, now we want to allow customers to consume a generated introspection schema from within a lambda, which would require running codegen after the cdk build starts, but before it is complete.