@marioslab/buckety-buckeroo

buckety-buckeroo

Got streaming data and want to create an approximate histogram for it? Want to also render the bins to a canvas without bells and whistles? That's what buckety-buckeroo does for you.

The implementation is based on the paper Streaming Parallel Decision Trees.

Demo

Online demo Source

Usage

You can include buckety-buckeroo directly on your page via.

<script src="https://cdn.jsdelivr.net/npm/@marioslab/buckety-buckeroo@1.0.0/dist/iife/buckety-buckero.min.js">

Every function and "class" is exposed on the global buckety object.

If you manage your project dependencies with NPM, add buckety-buckeroo as follows:

npm install @marioslab/buckety-buckeroo

Then import all exported functions and classes in your source files via:

import * as buckety from "@marioslab/buckety-buckeroo"

Create a histogram and specify the number of buckets it should have:

let histogram = new buckety.Histogram(100);

Then add values:

histrogram.addValue(432.32);
histrogram.addValue(-0.21);
histrogram.addValue(64);

You can calculate approximate quantiles:

let tenPercentQuantile = histogram.quartile(0.1);

Or perform your own calculation by iterating the buckets:

let mean = 0;
for (let bucket of histogram.buckets) {
	mean += bucket.value;
}
mean /= histogram.buckets.length;

Finally, you can draw the histogram to a canvas.

<canvas></canvas>
<script>
	let canvas document.querySelector("canvas");
	let histogram = new buckety.Histogram(20);
	let histogramChart = new buckety.HistrogramChart(canvas, { histrogram: histogram });
	histrogramChart.render();
</script>

See the demo above for an interactive example.

Will you add feature $X

No, but I might accept pull requests.

Development

Install Git, Node.js, and Visual Studio Code. Then:

git clone https://github.com/badlogic/buckety-buckeroo
cd buckety-buckeroo
npm run dev
code.

Press F5, modify code, press CTRL+S, see the page auto-update with your changes, set some breakpoints.