3 releases (breaking)

0.3.0 Oct 9, 2024
0.2.0 Oct 4, 2024
0.1.0 Sep 30, 2024

#388 in Machine learning

21 downloads per month

MIT license

19KB
511 lines

Egg Stats

Crates.io Version docs.rs Crates.io Version

About

This project makes it easy to visualize statistics about your egg runs. There are two components: egg-stats and egg-viz.

egg-stats is a library that adds a LoggingRunner that can wrap any existing egg::Runner and records customizable statistics to a csv file.

egg-viz is a web interface that let's you view the generated csv files to quickly explore these statistics. The web interface is still in development, and any feedback would be very useful.

Installation

You can add the egg-stats library to your project with:

cargo add egg-stats

You can install the visualizer with:

cargo install egg-viz

Usage

Generating data

LoggingScheduler is the key data-structure. Simply set it as the scheduler of your runner to generate data. You can wrap any other other egg::RewriteScheduler, and it will use it to actually schedule rule applications.

You need to set out_file to an open file, and set logging_enabled to true to see any data. There are a set of provided recorders to record common statistics. You can also easily implement your own. See the docs for more information.

Here is what an example looks like:

Runner::default()
        .with_scheduler(
            LoggingScheduler::from(scheduler)
                .with_out_file(
                    OpenOptions::new()
                        .write(true)
                        .create(true)
                        .truncate(true)
                        .open(path.as_ref())
                        .unwrap(),
                )
                .with_logging_enabled(true)
                .with_recorder(recorders::Timestamp::new(Instant::now()))
                .with_recorder(recorders::NumberENodes)
                .with_recorder(recorders::NumberEClasses)
                .with_recorder(recorders::BestProgram::new(AstSize, root)),
        );

Checkout examples for more complete examples.

Visualizing data

Simply run the following command to open the web interface, passing in the directory where your .csv files are located.

egg-viz <csv data dir>

image

Dependencies

~2–2.8MB
~53K SLoC