Support for pre-decoded rules

[beeme1mr]

Hey @diegoholiveira, thanks for the great library. We're using it as part of our flag evaluation engine in flagd. I was wondering if we could add support for pre-decoding the rules so that it doesn't need to happen on every call to Apply. JSON Logic engine, a JavaScript version of Json Logic, supports this, and the performance benefits are impressive. This approach would work well for use because changes to feature flag configuration happen relatively infrequently, while evaluations can happen extremely frequently.

If you're open to the idea, someone from the flagd community or I can work on a PR. Thanks!

[diegoholiveira]

@beeme1mr I'm super open to this idea. It would be super good if you or anyone from flagd could do this change because at the moment I'm kind of unavailable but I'll be able to implement this next year case you don't mind waiting.

@eltonjr this may benefit your services at Mercado Libre.

[TotalTechGeek]

@beeme1mr 😅 Heya, just a small heads up, because y'all are using engine.run in an older version of JSON Logic Engine, you're actually not currently getting those performance benefits.

engine.run uses an interpreter; which allows folks with strict Content Security Policies to use the engine.

Relevant Issue w/ an Experiment: handlebars-lang/handlebars.js#1934 (comment)

However, in v2 of JSON Logic Engine (and y'all should be able to seamlessly upgrade to the latest, which is like 3.0.5 right now) I added a closure "compiler" to engine.run that uses a WeakMap against your logic to cache the execution plan & optimize performance.

So as long as the object isn't being changed out / cloned, it can optimize it. If y'all want to get the full performance benefits, y'all will need to switch to precompiling the logic with engine.build

Also: I need to update that darned website 😅; I've been on a performance kick and trying to trim the codebase a bit.