[HACKDAY] added populate method to Promise to chain off any load / find call

[zgavin]

not 100% on this being the right way to go, but it compiles and actually works 🙃

[stephenh]

I appreciate the creativity here :-) but yeah, I'm not comfortable with this level of prototype hacking, both a) outside of graphql-service, and b) these populate methods only allow to promises of entities and not all promises.

[zgavin]

was thinking maybe we just provide the method definitions in their own interface and a helper function that would add the implementation to Promise.prototype. Then consumers could apply it globally on an opt in basis.

eg we'd define something like this:

export interface PromisePopulate<T> {
  populate<T extends Entity, H extends LoadHint<T>>(this: Promise<T>, hint: H): Promise<Loaded<T, H>>;
  ...
}
export function addPopulateToPromise() { Promise.prototype.populate = ... }

and then codebases that use joist could add this somewhere to get the behavior if they wanted:

declare global { interface Promise<T> extends PromisePopulate<T> {} }
addPopulateToPromise();

[stephenh]

And this is what we originally wanted to be author.publisher.load("images") but doesn't jive with TS?

Can you just the return type of Reference.load to Promise<...> & { populate: ... }, and then monkey-patch populate directly onto that single promise instead of the global Promise prototype?

I guess I wouldn't be surprised if, to do something like:

const promise = em.load(....);
promise.populate = ....
return promise;

We can't use async functions b/c the compiler/runtime rewriting of the function might create it's own promise, and not use the promise that we monkey-patched with a populate key.

[zgavin]

iirc adding a Loaded or LoadHint anywhere in the definition of Reference or Collection resulted in circular reference compiler errors