crystal-community / bloom_filter Goto Github PK

Did you tested bloom_filter in a large database?

icr(0.24.1) > bloom = BloomFilter.new_optimal(300_000_000, 1.0e-6); nil
 => nil
icr(0.24.1) > bloom.bytesize
 => 1078319068_u32
icr(0.24.1) > bloom.hash_num
 => 20_u8

when I insert about 15-20 millions records, i receive 100% false positive.
based on https://hur.st/bloomfilter - settings from .new_optimal are correct.

Let's say now I have only 1000 elements, I choose hash_num = 4 to have false positive = 2% (just an assume number). Later I add more 1000 elements, can I expand the byte size to keep hash_num and false positive unchanged. Of course without building the filter from scratch.
Thank you.