alt Hacker NewsThank you. Having a native bytes type is non-negotiable for any performance intensive application that cannot afford the overhead of base64 encoding. And yes, Rkyv also implements this idea of indexing serialized data. The main differences are:
1) Rkyv uses a binary tree vs Lite³ B-tree (B-trees are more cache and space efficient).
2) Rkyv is immutable once serialized. Lite³ allows for arbitrary mutations on serialized data.
3) Rkyv is Rust only. Lite³ is a 9.3 kB C library free of dependencies.
4) Rkyv as a custom binary format is not directly compatible with other formats. Lite³ can be directly converted to/from JSON.
I have not benchmarked Lite³ against Rust libraries, though it would be an interesting experiment.
Replies
1) when did they downgrade? I've stared for hours at that particular code...
2) no you just don't get to move data freely.
3) I don't believe JSON has any place in a system that needs C because it can't handle Rust.
4) JSON can't handle non-tree structures, it's further very limited in expressivity. Rkyv is more of a code gen akin to ASN.1
Happy benchmarking, feel free to use the rkyv benchmark tooling and ensure you have enough link time optimization going on.
That second point is huge – Rkyv does have limited support for in-place mutation, but it is quite limited!
If you added support for running jq natively, that would be very cool. Lite³ brings the B-trees, jq brings the query parser and bytecode, combined, you get SQLite :P