Crate reproducible 

reproducible helps Rust projects generate reproducible accuracy and benchmark reports.

Quick Start

use reproducible::prelude::*;
use reproducible::metrics;

// 1. Define the functions we want to evaluate

let fn_add = |inputs: &[f64]| vec![inputs[0] + inputs[1] + 5.0 * f64::EPSILON];
let fn_sub = |inputs: &[f64]| vec![inputs[0] - inputs[1] - 10.0 * f64::EPSILON];

// 2. Compose the report

let report = Report::new()

    // Add columns for metrics and statistics
    .with_column(
        Column::<f64>::accuracy("Mean Relative Error (eps)")
            .with_metric(metrics::rel_err_eps)
    )
    .with_column(
        Column::<f64>::accuracy("Max Absolute Error")
            .with_metric(metrics::abs_err)
            .with_stat(ColumnStat::Max)
    )
    .with_column(Column::<f64>::perf("Latency"))

    // Add rows corresponding to each function you want to evaluate
    .with_row(Row::new("math/add", fn_add).with_test_cases(vec![
        TestCase { inputs: vec![1.0, 2.0], expected: vec![3.0] },
    ]))
    .with_row(Row::new("math/sub", fn_sub).with_test_cases(vec![
        TestCase { inputs: vec![1.0, 2.0], expected: vec![-1.0] },
    ]));

// 3. Render to Markdown

println!("{}", report.render_markdown());

Output:

| Function | Mean Relative Error (eps) | Max Absolute Error | Latency |
|----------|---------------------------|--------------------|---------|
| math/add | 1.33                      | 8.88e-16           | 9.3 ns  |
| math/sub | 10.00                     | 2.22e-15           | 9.7 ns  |

Modules

benchmark

Benchmark ingestion helpers.

columns

metrics

parser

CSV parsing utilities for accuracy test datasets.

prelude

render

Markdown table rendering for reports.

report

Fluent report composition API.

rows

stats

Statistical utilities for report aggregation.

Macros

as_array

Wrap a function that takes a fixed array into one that takes a slice.

row

Shorthand for creating a Row from a function, using its name as the row label.

wrap_args

Wrap a function that takes positional arguments into one that takes a slice.

Functions

to_pretty_json

Serialize report rows to pretty JSON for machine-friendly artifacts.