om-nomnomnom is a parser for the OpenMetrics (text) exposition format written using the nom parsing framework. This crate is designed to ingest OpenMetrics data that an exporter might generate. If you're looking to instrument your Rust application, this is the wrong crate.

[dependencies]
om-nomnomnom = "0.1"

Compiler support: requires nightly

Usage is fairly simple:

    let om_data : &str = indoc! {"
        # TYPE a gauge
        # HELP a help
        a 1
        # EOF
    "};

    let families = om_nomnomnom::parse(om_data)?;

    for (name, family) in families.iter() {
        println!("{} ({:?})", name, family.metric_type);
    }

To iterate over the samples and pull out all the values:

    let family = families
        .values()
        .first()
        .expect("Empty exposition?");
    let values = family
        .samples
        .iter()
        .map(|sample| sample.number)
        .collect::<Vec<f64>>();

om-nomnomnom focuses on correctness more than performance. Even so its performance is on par with other Rust implementations and well ahead of the reference parser written in Python.

The speed / correctness tradeoff can be further configured with the following features:

• naive_wide_char_support
• no_interleave_metric
• enforce_timestamp_monotonic
• hash_fnv

• Serialization
• Convenience structs for each family type