xarvh/elm-gamepad

Painless Elm library to use & configure gamepads and game controllers

elm-gamepad's Issues

Documentation links are broken

Since there isn't an official way of previewing docs before publishing, I tested the links on https://elm-doc-preview.netlify.com/ which apparently interprets them differently than elm-package.

This is not going to get fixed until there will be a way to test links before publishing.

connection / disconnection getters?

? newlyConnectedGamepads : Blob -> List Int
? newlyDisconnectedGamepads : Blob -> List Int

Cannot remap gamepads with inertial sensors

Because the library can't distinguish between a number changing because the user deliberately activated a control, or just moved the gamepad.

"Quick and dirty" API

Elm Gamepad is mostly used for personal-use apps, apps that don't need to be usable by everyone on the internet.

It should be possible for Elm Gamepad to provide a minimal gamepad (a single stick and a single button) without any need for remapping.

Even if this wouldn't be reliable, it would allow people to put together quick-and-dirty apps without any need to wire in the remapping tool and persist remapping data.

Is this a good thing to add?
How would the API look like?
How would it be possible to make it clear that the API is not reliable?
unreliableGetGamepads : Blob -> List MinimalGamepad?

Restyling docs

Where did the documents about customizing the style end up? -_-

Gamepad.Remap should recognise when a gamepad is disconnected during remapping

If a gamepad is disconnected or replaced during remapping, Gamepad.Remap should inform the user and, if the right gamepad becomes available again, it should be able to resume the remapping.

Cull zombie gamepads?

Some gamepads persist even after being physically disconnected.
Remove them when their timestamp becomes too old?

Link bugzilla discussion in README.md

This is a good discussion of the current state of mapping support and its issues, would be useful to link it in the README.md
https://bugzilla.mozilla.org/show_bug.cgi?id=855364#c9

Clarify difference between id and index

Remap does not work with Standard Gamepads

Gamepad.Remap will complain that the gamepad is disconnected.

Fix in #9

Allow remapping single gamepad

A mapping should store also the gamepad index.

When searching for a mapping, first search one with the correct index.
Absent than, use any with the correct id.

Remap tool should have buttons to skip and cancel remapping

Extend README.md

• It should contain a minimal example without database and a full example with loading the db from flags.

• It should explain that gamepads won't be detected until they are touched.

Make it into a Port

It's easier to use if I just provide an elm package with decoders and maybe a guess function, and provide a function to cut & paste as port, so that the package can be published.

User should have the option to disable a gamepad

Skipping all controls should mark a gamepad as "disabled by the user"

New API with wasClicked and wasReleased

Port blob should provide current and previous states, each with its list of gamepad states and timestamps.
Advantages:

• Can detect button presses and button releases
• Can get rid of Destination* type.
• Can provide both time differences and time stamps
• Port signature becomes simpler
• Simpler JS

Encode database to JSON

Since a large, comprehensive gamepad database is not practical, size is not an issue any more.
Because of that, it's better to serialise the database in JSON format.
Advantages:

• Get ride of regex parsing
• Use a common format
• Moots #14

Fix ,,, tests

The triple comma ,,, is used in the serialised version of a db entry to divide the gamepad identifier from the button map.
Fuzzy testing will sometimes produce a gamepad identifier containing exactly that, making the test fail.

Is ,,, a legitimate separator?

Should I just tell the fuzzer not to produce that, and hope that no gamepad identifier will ever contain ,,,?

Is there a better way to serialise the database entry so that this cannot happen?

Move remapping-specific stuff in Gamepad.Remap.Advanced

"You need this stuff only if you plan to write your own gamepad remapping functions."
"Gamepad.Remap is actually written using only the stuff here (no special access to Gamepad opaque types"

Automatic remapping uses least squares on sigmoid to detect activated input

Use a robust non-linear regression against a sigmoid function to detect the input actually activated by the user during automated remapping.

This might overcome both #41 and #25 but needs to be tested extensively as it would be a much more complicated algorithm, more prone to unexpected behaviour.

https://scipy-cookbook.readthedocs.io/items/robust_regression.html
https://github.com/scipy/scipy/blob/v1.3.0/scipy/optimize/_lsq/least_squares.py#L240-L939

The Example looks horrid

• Needs some redesign and restyling
• Code needs to be more clear and better commented
• Should tell the user if there are gamepads that can be configured

First `dt` is negative

It looks like performance.now() returns a timestamp that's a few milliseconds ahead of the first timestamp giveno to window.requestAnimationFrame's callback; the result is that the first dt passed to the subscription is negative, ie about -10ms instead than the usual +16ms.

Add a way for devs to preview the mapping tool without a physical pad

Allow multiple origins to map to the same destination

For example: map both A and DpadUp buttons to the same "Jump Up" action.

This is made trivially possible by the advanced configuration menu that will be added to fix #41

(Also, since we're here: turn the Origin type into a record and un-opaque it, since it's not exposed at all)

PS4 triggers

looking at the advanced example, the "Left Trigger / Left Analog Lever" flips to True when i pull the trigger on the controller. however, "Left Trigger (analog)" stays at 0.

if i pull the right trigger on the controller, both "Right Trigger (analog)" and "Left Trigger (analog)" show the value from the right trigger.

(lots left to do, but i think i'll get this integrated into a game. thanks!)

Create comprehensive testing suite

There are too many stupid edge cases that the remapping tool needs to be aware of.
They should be covered via tests.

Should the `gamepad` argument always be last?

The gamepad getters take the gamepad as first argument, and the control name as second.
This is done because it was convenient in many situations in my own code.
However there were as many cases that would have benefited from having gamepad last.

It would be interesting to collect more use cases.

Link to html5gamepad

Docs should link to http://html5gamepad.com/

Firefox produces -1 when analog triggers are in rest position

https://bugzilla.mozilla.org/show_bug.cgi?id=1434408#c1

This screws up with the origin estimation function.

Fix: have the origin estimation consider two consecutive frames instead than just 1?

Docs: clarify API layers

Make more immediately clear that the library has an API for quick usage and another API that gives more control.

Only English and French translations present.

This will have to wait until #41 is addressed.

Rewrite Remap entirely

Gamepad.Remap is a pain to wire in and still does nothing.
It's much better to provide the full TEA machinery for a top-of-the-notch remapping UI that can be plugged in any app.

Pros

• Much easier API, less work to add remapping capability
• Better UI

Cons

• Adds a dependency on elm/html

Readme should contain gamepad schema

Find an image with the named buttons and slap it in README.md

The library allows using unconfigured controls

It shouldn't.

Change call order?

Per https://w3c.github.io/gamepad/#usage-examples it might be better to poll right after the call to requestAnimationFrame()

animationFrameTimestamp does not produce an absolute timestamp

Gamepad.animationFrameTimestamp produces time via the window.requestAnimationFrame callback, which is given a DOMHighResTimeStamp.

This seems to produce a time relative to window init instead than an absolute time, which a Posix date should always be.

Further, the docs sell Gamepad.animationFrameTimestamp as a drop-in replacement for Browser.Events.onAnimationFrame, which is not true.

Fix: the port module should use Date.now(), per https://github.com/elm/browser/blob/1.0.0/src/Elm/Kernel/Browser.js#L265 as the timestamp value.

Support multiple languages

The remap tool is giving instruction to the user in English, which severely limit the tool's accessibility.

A simple way to address this would be to add a remapViewWithLanguage function that allows to specify a alternate text content for each of the remap messages.

Remap picks the gamepad

Currently, Remap.init must be provided with the index of the gamepad to remap.

There is little value in letting the user code manage this.

The most user friendly behaviour seems to be to just let the gamepad user press a button to indicate which gamepad is to be remapped, and Remap should implement it.