Ideally, all pieces could be rounded. just doing the corners is more efficient, and will probably look almost as good.

See cubing/twisty.js#54 for rounded stickers.

• Two incoming R moves in short succession should be merged.
• Simultaneous L R' should animate at the same time. (This matters especially because M' is L R' with a reorientation.)

• Sliding puzzles
  • Continuous sliding mechamisms: https://mathsbot.com/manipulatives/rekenrek (similar to an abacus)

The gold standard for this right now is http://www.speedsolving.com/wiki/extensions/algdb/vcube/visualcube.php

Other puzzles have notation that is partially compatible. twisty.js will need some abstraction to cover them.

http://cube.crider.co.uk/algtrans.html

I believe the new twizzle app would be nicer if it supported side-by-side equal-sized views, one of the front and one of the back. The current small back view (rear view mirror?) is useful for debugging, but having a back view of the same size would make the interface work much more nicely for actual interactive solves.

Puzzles

• Images for stickers/faces
• Photorealistic textures & lighting
• Rounded cubies/stickers
• Motion blur
• Show internals
• Camera settings

Interface

• Theme colors

i.e. arbitrary custom stickering images like Werner's applet supports.

(This should have been filed a long while ago?)

https://github.com/IjzerenHein/node-web-bluetooth

The code should accept the "empty" puzzle with no sets (and an empty solved state and moves that do nothing).

In my UI I'm allowing users to turn on and off various parts of the puzzles (corners/edges/centers) and if they turn everything off, kpuzzle throws. It should not.

Here's a list of other software whose features we should look at:

• Ryan Heise's revolutionary Rubik's Cube Simulator with keyboard input.
• Werner Randelshofer's highly configurable virtual cubes
  • http://randelshofer.ch/rubik/index.html
  • https://www.randelshofer.ch/cubetwister/doc/cubetwister-complete-guide.html
  • See cubing/twisty.js#47.
• Josef Jelinek's simple and configurable AnimCube (with some credit to Karl Hornell and Lars Petrus).
• @larspetrus's Roofpig https://github.com/larspetrus/Roofpig
• Michiel van der Blonk's (?) FMC helper page
• Herbert Kociemba's Cube Explorer
• Michael Gottlieb's QCube https://mzrg.com/js/qcube-v2.html
• https://scarygami.github.io/scary-cube/
• https://github.com/hakatashi/smart-cube-timer
• VisualCube https://www.speedsolving.com/wiki/extensions/algdb/vcube/visualcube.php
  • Server rendering: #47
• UltimateMagicCube

See more at https://github.com/cubing/cube-code#javascript

Ideally, we'd have a nice system that makes it nice to do speedsolves, but provides features like logging of solvs (look at how you do your solves, and how this has changed!), etc.

The ksolve parser does not accept ksolve files that start with a # comment.

Half turns should be slow/longer than quarter turns.
It should be possible to specify the time for each base move.

(I know what I want to do. Will detail later.)

e.g. L R' is interpreted as M' together with an orientation change (see cubing/cuble.js#4) if they fire within 0.1s of each other.

Demo of showing only certain pieces: http://www.randelshofer.ch/rubik/virtual_cubes/rubik/instructions/instructions.html
In particular, there's a core!

If someone is implementing a new puzzle/feature we didn't anticipate, it would be nice if they could hook into a simple plugin API to control:

• parsing algs
• puzzle state
• puzzle display

1. Pick the settings that actually matter.
2. Try to make as many of these dynamic as possible (i.e. allow changing after the initial constructor).

• Actually, it might be good to move as many possible settings out of the constructor as possible, and then provide convenience functions for the most common setup pattern.
  • For example, control.speed can always be set with setSpeed after the twistyScene is constructed.
  • It might be cheap enough to construct/remove/hide things like hint stickers "on the fly". Certainly better than creating a new twistyScene.
• full color/sticker settings (color lists for faces?), image maps
• custom coloring for e.g. different cube steps (OLL, 6x6x6 centers)
• transparency (param)
• bgcolor (transparent default) - is this really necessary?
• BLD mode
• Each puzzle: Interface for random scramble?
• animation/transition curve
• Config file (thus, standard specs)
• Callback interface
• notations?
• See CubeTwister
• Playback interface (one level abstracted, but still on same div?)
• camera settings (location, scale?, orthoQ, etc.)
• rear view
• Mouse support
• Cube panning/rotation
• static rendering support (for WCA scramblers, etc.)
• debug mode
• bandaging?

In principle (and as suggested by three.js demos), it's not hard to capture mouse/touch events and use them to animate moves on the cube. I think the main issue here is handling the in-between animation while someone is moving a layer.

Should include:

• rendering the (partial) move, with live updating
• good mobile support
  • needs to interact with touch support from the touch branch.

For each line:

• Step name
• AUF/rotations
• actual moves
• move count(s)

Title pretty much says it all. Ideally this gets done in as twisty-independent a way as possible.

Implement a decent set of parameters that make it posible and worthwhile to switch out AnimCube and Randelshofer's appplets for three.js on cube pages. Ideally, this should be able to handle tons of twisties on a page (achievable by flattening the twisties per cubing/twisty.js#12 while not active).

• AnimCube
• Randelshofer
• VisualCube
  • Not a Java applet, but has URL-based configuration

3x3x3

• Fridrich
  • Cross
  • F2L
  • F2L with certain slots unsolved/multislotting?
  • OLL
  • PLL
• MGLS
  • ELS
  • CLS
• LL substeps
• Roux
  • L6E
  • CMLL
• Petrus
  • 2x2x2
  • 2x2x3
• ZBLS
• ZZ

Big Cubes

• Centers
• Reduction

Cursory research suggests that it should be a value from 0 to 100, but the GiiKer cube appears to report 55 when the app shows it as 100% charged.

Define an interface for each alg type, then define AlgorithmJSON as their union.

Official

• Pyraminx (any ordedimensions + edge twisting?)
• Megaminx (any dimensions)
• Skewb (any dimensions)
• Clock (NxNxN?)
• Square-1 (any number of layers?)

Maybe

• Pillowed Cubes?
• Octahedron?
• Square-2
• pyra crystal
• UFO/cheese/anything dihedral
• KxLxM cuboids

Sliding

• 15 Puzzle
• Astrolabacus
• Black Hole

Shoot for the Moon!

• Investigate UMC
• Magic?
• Sticker-sliding cube
• Heli cube
• Anything on gelatinbrain / Jaap's page?

Ideally, I wouldn't need to be the one to implement all of these.

https://github.com/spite/ccapture.js/

See #47 for the server-side rendering issue. It may that there is a simpler solution for this on the client.

We may not be able to prevent this, and this hurts our ability to mimic gelatinbrain moves.

• Ryan Heise's revolutionary Rubik's Cube Simulator with keyboard input.
• Werner Randelshofer's highly configurable virtual cubes
  • See cubing/twisty.js#47.
• Josef Jelinek's simple and configurable AnimCube (with some credit to Karl Hornell and Lars Petrus).
• Michiel van der Blonk's (?) FMC helper page

Curious why 222 has only 7 pieces?

https://github.com/cubing/kpuzzle.js/blob/master/src/puzzle_definitions.ts#L6

I'm looking for a way to set the current cube state like so: setCubeState("uuuuuuuuu fffffffff rrrrrrrrr bbbbbbbbb lllllllll ddddddddd"). The specific format/protocol doesn't matter. Is there a way to do this using twisty.js? I found a setPosition method, but this only seems to change the "pointer" pointing to the current move within an algorithm.
Is such a feature even desired? It would probably break some assumptions like v.setPosition(0); v.isIdentity() == true.

• qcube
• randelsflat
• cube explorer

We have a good parser generator, but we could get some benefits from writing our own:

• Better error messages for parsing errors
• Streaming/incremental parsing
• Very efficient

Working Draft:

Classes

Classes implemented using prototypes.
All setters should also have a getter.

Twisty

Twisty.animation.setSpeed(ratio);
Twisty.animation.playForwards();
Twisty.animation.playBackwards();
Twisty.animation.pause();
Twisty.animation.playSingleMoveForwards();
Twisty.animation.playSingleMoveBackwards();
Twisty.animation.skipToStart();
Twisty.animation.skipToEnd();
Twisty.animation.setPosition(ms, "animateQuickly" | "immediate")
Twisty.setup.setSolvedState(TwistyState)
Twisty.setAlg(TwistyAlg)

Twisty3x3x3 (prototype: Twisty)

TwistyState

A TwistyState is a ksolve-like extension of the idea of a permutation.

TwistyState.fromAlg()
TwistyState.invert() @returns TwistyState
TwistyState.compose(TwistyState) @returns TWistyState

AnimatingTwistyState

TwistyState.fromAlg()
TwistyState.invert() ?

Alg

An alg is a collection of moves and other algs.
Can be either a straight move sequence, or a group/communtator/conjugate.

Alg.addMoves(?)
Alg.timestampToToken(ms) @returns AlgToken
Alg.tokenToTimestamp(AlgToken) @returns ms
Alg.reparse("<original alg> R U R'")

Alg3x3x3 (prototype: Alg)

AlgToken

DOMAlg ?

// Something for highlighting

TwistyView

new TwistyView(rendererType, div);
  (rendererType == canvas, CSS div, etc.)
  // support shadow DOM where possible?
TwistyView.addPuzzle(puzzle);
TwistyView.setCameraLocation(coords)
TwistyView.fullscreen();
  // Perhaps the client can just use the normal HTML5 API on the div?
// 2-axis View rotation support support

Convenience

Twisty.emulateAnimCubeApplet()
Twisty.emulateRandelshoferApplet()
Twisty.emulateRoofpig

TwistySolver

TwistySolver.solveState(); @returns Alg

TODO

• Sticker Dragging
• Keyboard input

Twizzle should display the face name (U, F, BL, etc.) on hover over a face. Otherwise it's not at all clear what face is what (especially for puzzles that have no established conventions for face names).

Move the relevant code from alg.cubing.net into twisty.js.

Right now if you click a move in twizzle, the move gets appended to the algorithm and the move is executed, but there is no coalescing of moves (say, U then U into U2). If you enable automatic coalescing, then clicking the second U causes the cube to "jump" back to the state before the first U and execute a U2; this is jarring and incorrect. This should be improved to behave similarly to the original twizzle (currently at twizzle.cubing.net).

I'm not sure if this is needed, but it might be nice, and it's not hard to provide (notwithstanding issue cubing/cuble.js#7).

Putting a dot (a short pause) in an algorithm in twizzle does not behave properly. This is because the MoveIndexer returns a BlockMove, and a Pause is not a BlockMove. A lot of signatures need to be changed to permit the move indexer to return a Unit (likely), and then the call sites need to check to see if what was returned was actually a BlockMove or a non-state-change algorithm element such as a pause.