jiter

Fast iterable JSON parser.

Documentation is available at docs.rs/jiter.

jiter has three interfaces:

• JsonValue an enum representing JSON data
• Jiter an iterator over JSON data
• PythonParse which parses a JSON string into a Python object

JsonValue Example

See the JsonValue docs for more details.

use jiter::JsonValue;

let json_data = r#"
    {
        "name": "John Doe",
        "age": 43,
        "phones": [
            "+44 1234567",
            "+44 2345678"
        ]
    }"#;
let json_value = JsonValue::parse(json_data.as_bytes(), true).unwrap();
println!("{:#?}", json_value);

returns:

Object(
    {
        "name": Str("John Doe"),
        "age": Int(43),
        "phones": Array(
            [
                Str("+44 1234567"),
                Str("+44 2345678"),
            ],
        ),
    },
)

Jiter Example

To use Jiter, you need to know what schema you're expecting:

use jiter::{Jiter, NumberInt, Peek};

let json_data = r#"
    {
        "name": "John Doe",
        "age": 43,
        "phones": [
            "+44 1234567",
            "+44 2345678"
        ]
    }"#;
let mut jiter = Jiter::new(json_data.as_bytes());
assert_eq!(jiter.next_object().unwrap(), Some("name"));
assert_eq!(jiter.next_str().unwrap(), "John Doe");
assert_eq!(jiter.next_key().unwrap(), Some("age"));
assert_eq!(jiter.next_int().unwrap(), NumberInt::Int(43));
assert_eq!(jiter.next_key().unwrap(), Some("phones"));
assert_eq!(jiter.next_array().unwrap(), Some(Peek::String));
// we know the next value is a string as we just asserted so
assert_eq!(jiter.known_str().unwrap(), "+44 1234567");
assert_eq!(jiter.array_step().unwrap(), Some(Peek::String));
// same again
assert_eq!(jiter.known_str().unwrap(), "+44 2345678");
// next we'll get `None` from `array_step` as the array is finished
assert_eq!(jiter.array_step().unwrap(), None);
// and `None` from `next_key` as the object is finished
assert_eq!(jiter.next_key().unwrap(), None);
// and we check there's nothing else in the input
jiter.finish().unwrap();

Benchmarks

There are lies, damned lies and benchmarks.

In particular, serde-json benchmarks use serde_json::Value which is significantly slower than deserializing to a string.

For more details, see the benchmarks.

Part of the Pydantic Stack

The Pydantic Stack is everything you need to ship production-grade AI agents:

• Pydantic AI - Type-safe agent framework
• Pydantic Logfire - AI-first, full-stack observability
• Logfire AI Gateway - Unified LLM proxy