datalib
Datalib is a JavaScript data utility library. It provides facilities for data loading, type inference, common statistics, and string templates. While datalib was created to power Vega and related projects, it is also a standalone library useful for data-driven JavaScript applications on both the client (web browser) and server (e.g., node.js).
For documentation, see the datalib API Reference.
Use
Datalib provides a set of utilities for working with data. These include:
- Loading and parsing data files (JSON, TopoJSON, CSV, TSV).
- Summary statistics (mean, deviation, median, correlation, histograms, etc).
- Group-by aggregation queries, including streaming data support.
- Data-driven string templates with expressive formatting filters.
- Utilities for working with JavaScript functions, objects and arrays.
Datalib can be used both server-side and client-side. For use in node.js, simply npm install datalib
or include datalib as a dependency in your package.json file. For use on the client, install datalib via bower install datalib
or include datalib.min.js on your web page. The minified JS file is built using rollup (see below for details).
Example
// Load datalib.
var dl = require('datalib');
// Load and parse a CSV file. Datalib does type inference for you.
// The result is an array of JavaScript objects with named values.
// Parsed dates are stored as UNIX timestamp values.
var data = dl.csv('https://vega.github.io/datalib/data/stocks.csv');
// Show summary statistics for each column of the data table.
console.log(dl.format.summary(data));
// Compute mean and standard deviation by ticker symbol.
var rollup = dl.groupby('symbol')
.summarize({'price': ['mean', 'stdev']})
.execute(data);
console.log(dl.format.table(rollup));
// Compute correlation measures between price and date.
console.log(
dl.cor(data, 'price', 'date'), // Pearson product-moment correlation
dl.cor.rank(data, 'price', 'date'), // Spearman rank correlation
dl.cor.dist(data, 'price', 'date') // Distance correlation
);
// Compute mutual information distance between years and binned price.
var bin_price = dl.$bin(data, 'price'); // returns binned price values
var year_date = dl.$year('date'); // returns year from date field
var counts = dl.groupby(year_date, bin_price).count().execute(data);
console.log(dl.mutual.dist(counts, 'bin_price', 'year_date', 'count'));
Build Process
To use datalib in the browser, you need to build the datalib.js and datalib.min.js files. We assume that you have npm installed.
- Run
npm install
in the datalib folder to install dependencies. - Run
npm run build
. This will invoke rollup to bundle the source files into datalib.js, and then uglify-js to create the minified datalib.min.js.
Webpack 1
If you are using Webpack 1, you need to enable a JSON-loader. To do so, first npm install --save json-loader
, then add the loader to your webpack config:
{
module: {
loaders: [{
test: /\.json$/,
loader: 'json-loader'
}]
}
}