Benchmark envelop-only vs deep field matching

[embano1]

What is your idea?

Discussed with @baldawar that it would be nice to know, e.g. in README.md whether it's preferable to pattern match on envelop (first-level) fields only or whether there's no significant impact on a match at a lower-level, e.g. Detail.someField.

Would you be willing to make the change?

Maybe, need advise on Java benchmarking.

Additional context

n/a

[baldawar]

This requires a new benchmark test within https://github.com/aws/event-ruler/blob/main/src/test/software/amazon/event/ruler/Benchmarks.java that compares performance at different depths.

If the results are noteworthy, then we should also update the performance section within the readme https://github.com/aws/event-ruler/blob/main/README.md#performance

[timbray]

To be honest I doubt there's much difference. But worth checking.

[baldawar]

It'll be helpful to have the test focus on depth at different levels more than just envelope vs detail. I've seen some code regress horribly as depth increases because of the data-structures in use while traversing downwards into JSON.

[timbray]

The matching cost shouldn't differ in the slightest, but the flattening can obviously be affected… I think it'll probably be O(N) in something like the total number of field-steps.

[embano1]

Interesting findings in Quamina :)