Installation

Jetro ships as three artifacts:

Artifact	What it is	Audience
`jetro` (crate)	Rust library — query/transform JSON in-process	Rust developers
`jetro-py`	Python bindings (PyPI)	Python users
`jetrocli`	Standalone CLI `jetrocli` for shell use	Anyone with JSON in a terminal

Rust library

Add to Cargo.toml:

[dependencies]
jetro = "0.5"

The simd-json feature is on by default and gives a ~4× cold-start win by parsing bytes directly into Val (no serde_json::Value intermediate). To fall back to the legacy serde-only path:

[dependencies]
jetro = { version = "0.5", default-features = false }

Quick sanity check:

use jetro::Jetro;

fn main() -> anyhow::Result<()> {
    let bytes = br#"{"books":[{"title":"Dune","year":1965}]}"#;
    let j = Jetro::from_bytes(bytes)?;
    let titles: serde_json::Value = j.collect("$.books.map(@.title)")?;
    println!("{}", titles);  // ["Dune"]
    Ok(())
}

Long-lived engine

If you process many documents with overlapping queries, keep a JetroEngine around. It holds shared plan and VM caches:

use jetro::JetroEngine;

let eng = JetroEngine::default();
for doc in docs {
    let v = eng.collect(&doc, "$.users.filter(active).count()")?;
    println!("{}", v);
}

Plan-cache default capacity is 256 entries; it evicts wholesale when full.

Python bindings

pip install jetro-py

import jetro

doc = {"books": [{"title": "Dune", "year": 1965}]}
print(jetro.collect(doc, "$.books.map(@.title)"))   # ['Dune']

The Python wheel embeds the same Rust core, so query syntax is identical.

CLI (jetrocli)

Install via Homebrew:

brew install mitghi/jetrocli/jetrocli

Or build from source:

git clone https://github.com/mitghi/jetrocli
cd jetrocli && cargo install --path .

Use it like jq:

echo '{"x":[1,2,3]}' | jetrocli '$.x.sum()'
# 6

cat data.json | jetrocli '$.users.filter(@.active).map(@.email)'

Building from source

git clone https://github.com/mitghi/jetro
cd jetro
cargo build --release         # build everything
cargo test                    # full suite
cargo bench -p jetro-core     # micro-benchmarks

Workspace layout:

jetro/             facade crate (re-exports + public API)
jetro-core/        engine: parser, planner, executor, builtins, runtime
jetro-core/fuzz/   cargo-fuzz harness (feature-gated)

Verifying your install

Run the tour from the next chapter against your install. If every query produces the printed output, you're ready.

The Jetro Book