BloomFilter capacity limitation

[crazylazyowl]

Good day!

I have a question about the BloomFilter support.

I tried migrating my project from Redis to KVRocks and encountered an issue where it’s not possible to create a BloomFilter with a capacity greater than 2^32. I checked the code [click] and noticed that when parsing the BF.RESERVE command, you are using ParseInt<uint32_t> for the capacity argument.

As a result:

127.0.0.1:6666> bf.reserve x 0.001 20000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 10000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 9000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 8000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 7000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 6000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 5000000000
(error) ERR Bad capacity
127.0.0.1:6666> bf.reserve x 0.001 4000000000
OK
(0.65s)

Is this a deliberate limitation? Or we can consider such behaviour as a bug?

Thank you!

[PragmaTwice]

As you can see, the parse_capacity is typed uint32_t, which means it cannot be larger than 2^32.

It's expected.

[crazylazyowl]

So, it's not a bug and you won't change it to 2**64?

[PragmaTwice]

If you check https://kvrocks.apache.org/community/data-structure-on-rocksdb#bloom-filter, you can get more idea about how it's stored on disk.

I'm not the guy who design this part, but from the design you can see it's 32bit based. If you think that it's not enough, maybe you can provide your use case so that we can have a discussion : )

[crazylazyowl]

Thank you :) I'll check it