cuckoofilter

There are smaller alternatives to Bloom filters, if size is the only issue. A few examples: - https://github.com/0xcb/Golomb-coded-map - https://github.com/efficient/cuckoofilter - https://github.com/FastFilter/xor_singleheader/

(Co-author of the paper.) You are right. However, it is a bit hard to find the threshold. Even the original author of the cuckoo filter didn't get it right, see https://github.com/efficient/cuckoofilter/issues/9 - for this reason we also lowered the maximum load to 0.94. However, I found (running the benchmarks) that sometimes that is not sufficient (insertion still fails). So, where is the cliff exactly? Is it 0.9? Or lower? And, as you say, it also depends on the hash function. We have used the Murmur3 finalizer, which is statistically very good. What if you use a weaker hash function? For a regular cuckoo hash table, you can just re-build it. But for a cuckoo filter rebuild is not easily possible as you don't have the keys. You are also right regarding augmenting the cuckoo filter (e.g. "cuckoo hashing with a stash"), but it slows thin

https://github.com/efficient/libcuckoo There's the related cuckoo filter as well: https://github.com/efficient/cuckoofilter

Just to prevent rampant over-generalization in an article whose subject is the topic of bias: We've implemented various special-purpose sorts at least three times in the last five years. Here's one example: https://github.com/efficient/cuckoofilter/blob/master/src/pa... Yes, this is all in a very high-performance (sometimes insanely so) context, but it does happen. Most of them were like this - unrolled special-purpose versions derived from a sorting network. Some were for GPU.

There is also a C++ implementation by the original authors: https://github.com/efficient/cuckoofilter

Cuckoo filters also support deletion. https://github.com/efficient/cuckoofilter (To the other person who replied: In order to delete from a counting bloom filter, or a cuckoo filter, or any other data structure that doesn't store the original item, you have to know the item that you're deleting was present. Otherwise, you have the problem you noted.)