turbo-json-checker

Crates.ioturbo-json-checker
lib.rsturbo-json-checker
version2.0.1
sourcesrc
created_at2021-08-17 15:03:37.871005
updated_at2021-08-20 20:45:38.294816
descriptionA pushdown automaton low memory JSON bytes stream checker returning the JSON root-type followed by its start and end index in the Reader
homepagehttps://github.com/irevoire/turbo-json-checker
repositoryhttps://github.com/irevoire/turbo-json-checker
max_upload_size
id438500
size42,898
Tamo (irevoire)

documentation

README

turbo-json-checker

This is a fork of oxidized-json-checker that returns the json type and the json starting and ending index in the Reader.

This is a pure Rust version of the JSON_checker library.

This is a Pushdown Automaton that very quickly determines if a JSON text is syntactically correct. It could be used to filter inputs to a system, or to verify that the outputs of a system are syntactically correct.

You can use it with the std::io::Read Rust trait to checked if a JSON is valid without having to keep it in memory.

Performances

I ran some tests against jq to make sure the library when in the bounds. I used a big JSON lines files (8.3GB) that I converted to JSON using jq -cs '.' 😜

You can find those Wikipedia articles on the benchmark repository of Paul Masurel's Tantivy.

jq type

How many times does jq takes when it comes to checking and determining the type of a JSON document? Probably too much, and also a little bit of memory: 12GB!

$ time cat ../wiki-articles.json | jq type
"array"

real    1m55.064s
user    1m37.335s
sys     0m21.935s

ojc

How many times does it takes to ojc? Just a little bit less! It also consumes 0kb of memory.

$ time cat ../wiki-articles.json | ojc
Array

real  0m56.780s
user  0m47.487s
sys   0m12.628s

ojc with SIMD

How many times does it takes to ojc already? 56s, that can't be true, we are in 2020... What about enabling some SIMD optimizations? Compile the binary with the nightly feature and here we go!

$ cargo build --release --features nightly
$ time cat ../wiki-articles.json | ojc
Array

real    0m15.818s
user    0m10.892s
sys     0m10.721s
Commit count: 51

cargo fmt