Ish Nagy

cutting down the testcases to the bare minimum (3%fpp and 1G items), the test now completes in little over 10minutes on my machine. would this be an acceptable running time?...

> Bloom filter functions (BloomFilterAggregate / BloomFilterMightContain) seem like internal ones so I wouldn't expect serialized filters to be shared between different application runs, but it seems some native accelerators...

> I like the fixed bloom filter logic in this PR but I'm not sure we need such convoluted logging in tests. the test should be silent now, by default....

> In addition, there are many duplicated code after this PR. I'm not sure this is the best way to extend the existing framework. Do you think we can minimize...

> Before verifying the any performance benefit, [...] Mind you, any performance benefit won't be visible from the sketch project only. If anything, it can be somewhat slower, since the...

One additional thing regarding performance. The first versions of the new test code did both the inserts and the queries in a serial order. This didn't do any good for...

I looked into how we could reduce the code duplication and I see two high level options. #### a quick-n-dirty monkey patch where `BloomFilterImplV2` extends `BloomFilterImpl` and overrides the methods...

@dongjoon-hyun , @peter-toth , the requested changes to reduce code repetition are done, please have a look, when you have the time.

should I resolve the started conversations myself, when I think I'm done, or will they be resolved by the reviewer who started them, when they are confirmed to be acceptable?