flare

Feature Flag Loading and Runtime Evaluation

Flare is a dynamic gating library that evaluates conditions based on input contexts. The package allows you to set up gates using various conditions and evaluate these conditions based on a given input. The package provides a promise-based API for use in any JavaScript or TypeScript project. Both the browser and node environments are supported.

Installation

yarn add @radekstepan/flare

Usage

The Flare class is the main export of this package. It is used to create a new Flare instance. You need to provide a Data array or a Promise resolving to a Date array in the constructor. The Data array contains maps of Gate definitions that are compiled and stored for evaluation.

import Flare from "@radekstepan/flare";

const gates = Promise.resolve([{
  gate1: {
    conditions: [{...}], // Define conditions here
    eval: '{...}' // Expression to evaluate
  },
  //...more gates
}]);

const flare = new Flare(gates);

See the schema definition in packages/utils/ for how to structure the gates and their conditions. Alternatively, see the the test fixtures in the packages/main/ project or the recipes/ folder for examples.

You can evaluate a specific gate or all gates using the evaluate and evaluateAll methods respectively. Both methods return a Promise resolving to an object with gate names as keys and the result of their evaluation as values.

flare.evaluate("gate1", context)
  .then(result => console.log(result));

flare.evaluateAll(context)
  .then(result => console.log(result));

Recipes

Benchmark

This benchmark was conducted using a script located in the packages/benchmark directory, executed on a system equipped with a Ryzen 5 3600 processor and running Ubuntu for Windows Subsystem for Linux (WSL). The script utilizes autocannon, running it against a Node HTTP server that evaluates Flare gates on each request.

The benchmark results displayed below indicate the performance metrics for varying numbers of gates (n gates), with each gate having 10 conditions and each condition having 1,000 values.

Gates n	0	1	10	20	50	50 (M2 Pro)	50 (Ryzen 5 7600X)
Avg Requests/sec	64,236	19,800	8,306	5,877	2,710	5,102	7,037
Avg Latency (ms)	0.01	0.03	1.10	1.16	3.35	1.24	1.12
Std Dev Latency (ms)	0.07	0.19	0.37	0.49	1.00	0.67	0.36
Boot Time (ms)	1	7	34	61	141	112	77

Please note that the "Boot Time" value denotes the time it takes to launch an HTTP server, load all Gates from a JSON file, compile all the expressions, and convert the list of values for each gate condition into sets.

Furthermore, the methods flare.evaluate and flare.evaluateAll are non-blocking. They return a promise that resolves once the input Gates have finished loading.