Introduction

Overview

Skip is an open-source framework that enables software developers to build and run reactive services and systems.

Skip services expose reactive resources: data and computation outputs which can be queried over HTTP by non-reactive clients or subscribed to by reactive clients to receive updates in real time.

Skip services are written in a declarative style, defining a reactive computation graph over some input data. Under the hood, the Skip framework efficiently evaluates and updates reactive computations, delivering up-to-date and correct results without incurring heavy recomputation costs or requiring any bug-prone explicit handling of dependency tracking or change propagation.

Core concepts

The main components of the Skip Runtime, the reactive computation engine at the core of the Skip system, are collections and mappers.

Collections and mappers together form a graph of data and computation dependencies that the engine can use to efficiently maintain reactive computations as inputs and data are updated.

Collections are the core data structure over which reactive computations operate, they are the vertices of the Skip reactive computation graph. A collection consists of elements, each of which associates one or more values to a key. Keys and values can be arbitrary JSON data.

Collections can either be eager, meaning they are always kept up-to-date, or lazy, meaning that values are only computed or updated in response to queries for a particular key.

Mappers are values that express a computation from elements of one collection to elements of another collection. Mappers form the edges of the Skip reactive computation graph, specifying transformations and compositions of data to produce intermediate results and outputs.

Crucially, mappers must be deterministic and side-effect free, so that the reactive runtime can re-execute them as needed to maintain results that are both up-to-date and guaranteed from-scratch consistent (that is, exactly the same as the result that would be computed if the equivalent non-reactive computation were executed from scratch on the current input).