Question: Sync getter?

[yaacovCR]

What kind of improvement might you envision for a synchronous API with the in-memory option? or is that already an option? Or not worth the complexity of returning a promise or value and having to check?

[Julien-R44]

Honestly yes, naively I'd say it would require quite some work and add some complexity to the codebase, so I'm not sure it's worth it.

What's your use case? Are you worried about the perf-hit caused by creating promises, or do you just need to use bentocache in a synchronous context?

[yaacovCR]

The performance hit. But not per se as a user of this library, just asking as a learning opportunity and to compare approaches.

My context is that I work here and there on open source development within the graphql server reference implementation and within the code base there is a lot of effort on staying sync when we can despite the more complex return types, so I was surprised to be honest that this library didn't.

One exception of course is DataLoader, that will stay async so that one synchronous hit amidst a batch does not cause the batch to split. If one is using bentocache inside a dataloader, all will be well.

But considering the general purpose of this library, I was a bit surprised. I guess it's also about that tradeoff of performance vs usability...

Sorry for the late reply and thanks for taking a look at my question!

[yaacovCR]

Let me put a practical spin on it. I want to create a graphql executor that utilizes query planning and I want to cache the plans, but I don't want every operation to be async. I see now that you have L1 drivers that are synchronous and L2 that are async, but I can't use bentocache for plan caching because even if the plan is in L1 the call to bentocache is async. I can use my own L1 cache and only call bentocache if it's empty, but then I am recreating the superb functionality of this library. :(