Design Overview

Plywood consists of three logical parts: an expression language for describing data queries, a set of externals for connecting to databases, and a collection of useful helper functions.

Expression Language

At its core Plywood contains an expression language (DSL) that is used to describe data queries.

Here is an example:

var ex = ply()
  .apply('Count', $('wiki').count())
  .apply('TotalAdded', '$wiki.sum($added)')
    $('wiki').split('$page', 'Page')
      .apply('Count', $('wiki').count())
      .sort('$Count', 'descending')

Plywood's language (called plywood, with a lowercase "P") is heavily inspired by Hadley Wickham's split-apply-combine principle and the D3 API.

Plywood expressions were designed with the following ideas in mind:

For more information about expressions check out the API reference.


While Plywood can crunch numbers internally using native JavaScript (this is useful for unit tests) its true utility is in being able to pass queries to databases. As of this writing only Druid and MySQL externals exist but more will be added.

The externals act as query planners and schedulers for their respective databases. In the case of the Druid External it also acts as a Polyfill, filling in key missing functionality in the native API.

Here is an example of a Druid external:

  engine: 'druid',         
  source: 'wikipedia',  // The datasource name in Druid
  timeAttribute: 'time',    // Druid's anonymous time attribute will be called 'time'

  requester: druidRequester // a thing that knows how to make Druid requests


A varied collection of helper functions are included with Plywood with the idea of making the task of building a query layer as simple as possible.

One notable example of a helper is the SQL parser which parses PlyQL, a SQL-like language, into plywood expressions allowing those to be executed via the Plywood externals. This is how Plywood can provide a SQL-like interface to Druid.