MaulingMonkey's typescript gamepad API for consistent cross-browser and cross-gamepad binding, polling for input, etc.

• GitHub: github.com/MaulingMonkey/mmk.gamepad
• Demo: maulingmonkey.com/mmk.gamepad/demo/
• Docs: maulingmonkey.com/mmk.gamepad/docs/
• NPM: @maulingmonkey/gamepad
• License: Apache 2.0

The vanilla browser gamepad APIs are unusable without a lot work, except for maybe if you stick to vanilla Microsoft 360/XB1 gamepads.

1. Noop fallbacks for unsupported browsers so you don't have to do your own feature testing.
2. Edge has conflicting gamepad browser navigation that needs to be disabled, if you don't want basic input like pressing (B) to navigate away.
3. Many browsers poorly standardize gamepads. Even ignoring outright buggy "standard" layout gamepads, I've seen dpads exposed as -1/8..+1 axises, int -> float bitcasts of accelerometer data, gamepads that start initialized to non-resting states... testing hardware you don't have is hard, I guess.
4. Basic gamepad APIs don't handle basic things like deadzone functionality, events (although Edge has some keydown events for gamepads)

1. All mmk.gamepad functions should still "work" as if there were no gamepads connected.
2. This browser may still have compatability bugs even providing the "no gamepads" interface.
3. This browser implements a gamepad API, but no gamepads work.
4. You may need to specify <meta http-equiv="X-UA-Compatible" content="IE=9" /> or better to avoid JS errors.
5. Edge has some built-in gamepad navigation behavior. mmk.gamepad will disable this by default.

1. This gamepad displays an incorrect or generic name which we cannot work around (e.g. "Xbox 360 Controller" for XB1 controllers, or "xinput")
2. This has a custom .mapping !== "standard"

4. Not Chromium - which still defines the gamepad API, but doesn't appear to actually export info about connected gamepads even when Chrome will on the same system.

5. Note that the baseline gamepad API lies about this gamepad being in the standard layout already in Chrome 66 on Ubuntu 18.04 LTS. tryRemapStdLayout will fix these to use the correct layout anyways.

6. Some triggers/axises of this gamepad currently misinitialize to non-zero values (e.g. -1, 0.5), although they'll correct themselves after some use. Could be worked around with some more stateful mapping (e.g. treating known bad values as 0.0 until the right stick / triggers are sufficiently exercised.)

See demo/demo.ts

// In case you want to display a hint to switch browsers ;)
if (!mmk.gamepad.isSupported()) console.warn("Gamepads not available in this browser");

// The rough equivalent of navigator.getGamepads(), with fallbacks.
var gamepads = mmk.gamepad.getRawGamepads() // Can return undefined elements (for e.g. disconnected controllers)

var snapshot = mmk.gamepad.cloneGamepads(gamepads); // Deep copy of basic fields
snapshot[0]  = mmk.gamepad.cloneGamepad(snapshot[0]); // Hey why not

// Works despite the lack of "connectedgamepad" and "disconnectedgamepad" events:
mmk.gamepad.addRawConnectedListener   ( function(gamepad) { console.log("Connected gamepad:",   gamepad); } );
mmk.gamepad.addRawDisconnectedListener( function(gamepad) { console.log("Disconnected gamepad:", gamepad); } );
// (note that addRawConnectedListener will immediately invoke your callback on already connected gamepads!)

var standard = mmk.gamepad.tryRemapStdLayout(snapshot[0]);
if (standard) { console.assert(standard.mapping === "standard"); } // Well, that was easy!
else          { ... } // Fallback to non-standard mapping prompts etc...

• Grab NPM via node.js
• Install per project

  npm i @maulingmonkey/gamepad

Negative values should always be left/up/forward/counterclockwise, with positive values corresponding to their opposites (right/down/backwards/clockwise). This might be unintuitive for some axises on an individual basis (yes, joystick forward is negative!), but this has a couple of advantages. Firstly, it keeps everything nice and consistent for remapping purpouses (thumbstick up is negative too, and you might want to use that as your throttle!) Secondly, it reduces the importance of properly categorizing your axises ("should I map this as a throttle, or as a slider?") as this won't impact how the values are actually mapped.

Additionally, axises should almost always span the entire [-1..+1] range. While you might want to remap this to [0..+1] in your game (possibly with a button for reversing), consistency at the lowest layer will make such remaps more straightforward. The exception to this would be any unidirectional, spring-loaded axises - any spring-loaded resting position should be 0. It's possible such axises should be considered buttons instead - e.g. gamepad triggers are considered buttons in the "standard" mapping.