path: root/2023/captions/emacsconf-2023-cubing--speedcubing-in-emacs--vasilij-wasamasa-schneidermann--main.vtt

WEBVTT captioned by sachac, checked by sachac

NOTE Introduction

00:00:00.000 --> 00:00:08.359
Hello, everyone, and welcome to Speedcubing in Emacs.

00:00:08.360 --> 00:00:10.119
First of all, a little bit about myself.

00:00:10.120 --> 00:00:13.679
My name is Vasilij Schneidermann. Online, I go by wasamasa.

00:00:13.680 --> 00:00:18.039
I'm 31 years old. I work in information security,

00:00:18.040 --> 00:00:20.479
and I do consulting and hacking

00:00:20.480 --> 00:00:22.479
and stuff like figuring out

00:00:22.480 --> 00:00:25.279
how to break into other people's computers

00:00:25.280 --> 00:00:29.359
and how to secure their systems basically.

00:00:29.360 --> 00:00:31.439
You can reach me by email.

00:00:31.440 --> 00:00:36.639
I do have a self-hosted code repository thingy going on.

00:00:36.640 --> 00:00:40.399
I have a blog, and you can find me

00:00:40.400 --> 00:00:45.919
in some other places online, like IRC for example.

00:00:45.920 --> 00:00:48.679
So about the talk itself,

00:00:48.680 --> 00:00:52.839
I used to be into the Rubik's cube when I was in school.

00:00:52.840 --> 00:00:54.039
I forgot about it, though,

00:00:54.040 --> 00:00:56.279
because these cubes were not very good.

00:00:56.280 --> 00:01:02.279
Recently I did find some cheap looking cube at a shop.

00:01:02.280 --> 00:01:04.119
Did not pay terribly much for it.

00:01:04.120 --> 00:01:07.039
It was so, so much better than my old cube,

00:01:07.040 --> 00:01:08.639
it was unreal.

00:01:08.640 --> 00:01:11.479
This motivated me to get back into

00:01:11.480 --> 00:01:13.559
this really weird kind of hobby.

00:01:13.560 --> 00:01:17.999
For this, you need to be good at producing

00:01:18.000 --> 00:01:19.399
a truly random scramble

00:01:19.400 --> 00:01:22.319
and timing your attempts to get any better at it.

00:01:22.320 --> 00:01:23.719
There is, of course, existing software

00:01:23.720 --> 00:01:26.239
to do the scrambling for you and the recording

00:01:26.240 --> 00:01:28.079
and the timekeeping and such,

00:01:28.080 --> 00:01:31.239
but all the good options seem to be either web or mobile,

00:01:31.240 --> 00:01:33.239
for example the cstimer software

00:01:33.240 --> 00:01:35.399
or the twisty-timer app on Android.


NOTE Cubing in Emacs

00:01:35.400 --> 00:01:39.319
To my surprise, I did not find a single decent option

00:01:39.320 --> 00:01:41.959
inside Emacs, so this is basically a case study

00:01:41.960 --> 00:01:44.999
how to do better. For this, I wanted to make use of

00:01:45.000 --> 00:01:47.799
all the cool new Emacs features that appeared,

00:01:47.800 --> 00:01:50.879
like the SVG library; Transient,

00:01:50.880 --> 00:01:53.599
the library used for the Magit-style interfaces;

00:01:53.600 --> 00:01:56.439
and the recently added sqlite-mode.

00:01:56.440 --> 00:02:01.159
And most importantly it was about having fun.

NOTE Prior art

00:02:01.160 --> 00:02:02.759
So here's a full list of prior art,

00:02:02.760 --> 00:02:04.279
I will not go into detail about this,

00:02:04.280 --> 00:02:06.239
but basically we have things solving

00:02:06.240 --> 00:02:08.039
very different parts of this,

00:02:08.040 --> 00:02:10.759
but not all of it. For example: we have several,

00:02:10.760 --> 00:02:14.239
we have a timer. We have several solvers.

00:02:14.240 --> 00:02:16.039
We have some scramblers.

00:02:16.040 --> 00:02:19.359
We have some whole-cube simulators, including a 3D one.

00:02:19.360 --> 00:02:20.759
We have something for making it easier

00:02:20.760 --> 00:02:23.119
to enter your algorithms in the notation.

00:02:23.120 --> 00:02:25.919
But nothing that does all of those things in one package,

00:02:25.920 --> 00:02:28.119
which kind of surprised me.

00:02:28.120 --> 00:02:32.039
So I present the `wca-prep` package.

NOTE The name

00:02:32.040 --> 00:02:35.559
So the name, I found it difficult

00:02:35.560 --> 00:02:39.959
to come up with a good name and so I looked

00:02:39.960 --> 00:02:42.559
and I saw, well there's this World Cube Association

00:02:42.560 --> 00:02:46.039
that holds these competitions where you compete.

00:02:46.040 --> 00:02:47.759
They do this for the Rubik's cube

00:02:47.760 --> 00:02:48.919
but also a few others,

00:02:48.920 --> 00:02:50.799
so there's like a standardized list

00:02:50.800 --> 00:02:52.639
of events they have for this.

00:02:52.640 --> 00:02:55.159
There is a standard notation for this

00:02:55.160 --> 00:02:56.519
and rules and everything.

00:02:56.520 --> 00:02:58.199
And the goal of my package is basically

00:02:58.200 --> 00:03:01.279
to help prepare myself for such a competition

00:03:01.280 --> 00:03:03.679
and in fact a week ago I went to my first one

00:03:03.680 --> 00:03:06.719
which was wild, but pretty cool.

00:03:06.720 --> 00:03:10.919
So for this reason I chose this name wca-prep,

00:03:10.920 --> 00:03:13.639
because it helps me prepare for this kind of competition

00:03:13.640 --> 00:03:16.519
and this limited the scope significantly,

NOTE What's in wca-prep

00:03:16.520 --> 00:03:18.999
I have a scrambler, visualization of the scramble,

00:03:19.000 --> 00:03:23.319
timer, and statistics.

00:03:23.320 --> 00:03:25.559
I excluded pretty much everything else I've seen.

00:03:25.560 --> 00:03:28.788
For this reason, I only tried to focus on

00:03:28.789 --> 00:03:32.199
some very basic puzzles I can solve comfortably,

00:03:32.200 --> 00:03:34.839
and did not want to do anything else

00:03:34.840 --> 00:03:36.439
that may complicate things significantly.

00:03:36.440 --> 00:03:40.479
No other kinds of puzzles, no simulation, no solving,

00:03:40.480 --> 00:03:43.919
no exotic events, and no specialized scrambles

00:03:43.920 --> 00:03:49.239
that are only good for practicing specific algorithms.

NOTE Demo

00:03:49.240 --> 00:03:54.199
So at this point the organizer should hopefully show

00:03:54.200 --> 00:03:57.999
a small video I've prepared, a one minute video showing how

00:03:58.000 --> 00:05:15.239
I actually use this to solve a cube and to time my solve.

NOTE Challenges: Representing the cube

00:05:15.240 --> 00:05:18.508
Okay, so building this thing, there were several challenges.

00:05:18.509 --> 00:05:20.508
The first one was how do I even represent

00:05:20.509 --> 00:05:22.468
the state of a Rubik's cube.

00:05:22.469 --> 00:05:25.508
For this there are many possible representations,

00:05:25.509 --> 00:05:27.708
no obvious best solution.

00:05:27.709 --> 00:05:29.628
I did not, well, what helped me was that

00:05:29.629 --> 00:05:31.988
I did not have to programmatically solve this thing,

00:05:31.989 --> 00:05:35.188
so I picked the easiest possible representation

00:05:35.189 --> 00:05:38.268
which is just an array of every single facelet.

00:05:38.269 --> 00:05:42.508
For a 3x3 cube you have 9 facelets on one side,

00:05:42.509 --> 00:05:47.268
so times 6 sides you would have 54 elements in this array.

00:05:47.269 --> 00:05:49.708
So with this representation, it's very simple,

00:05:49.709 --> 00:05:52.388
but it's kind of weird to do scrambles with this.

00:05:52.389 --> 00:05:54.908
But otherwise, it worked very, very well.

00:05:54.909 --> 00:05:57.268
In the future, I plan to learn some group theory,

00:05:57.269 --> 00:05:58.748
pick a better representation

00:05:58.749 --> 00:06:01.188
and do this in a much, much more elegant way

00:06:01.189 --> 00:06:07.868
without compromising speed too much.

00:06:07.869 --> 00:06:10.708
Yes. Once I had the representation,

00:06:10.709 --> 00:06:13.628
the scrambling itself should not be too hard.

00:06:13.629 --> 00:06:17.748
For this, it's important to consider that basically

00:06:17.749 --> 00:06:19.148
if you do a face turn

00:06:19.149 --> 00:06:22.428
you end up swapping some facelets with other facelets,

00:06:22.429 --> 00:06:26.028
that's the easiest way to think about this.

00:06:26.029 --> 00:06:29.268
To determine which one goes into which one's position,

00:06:29.269 --> 00:06:32.470
it was pretty confusing to figure this out.

00:06:32.471 --> 00:06:34.308
For this I went through a few papers,

00:06:34.309 --> 00:06:36.028
and I found one which suggested

00:06:36.029 --> 00:06:37.948
to just build a cube out of paper,

00:06:37.949 --> 00:06:40.028
number every facelet, and turn it

00:06:40.029 --> 00:06:44.348
and keep track of which facelet moved into which position.

00:06:44.349 --> 00:06:47.508
And programmatically, the `cl-rotatef` macro

00:06:47.509 --> 00:06:49.388
was very, very useful for doing this kind of

00:06:49.389 --> 00:06:51.628
in-place swapping you need for this operation.

00:06:51.629 --> 00:06:54.868
So in the future, group theory would hopefully

00:06:54.869 --> 00:06:57.988
make this a bit less awkward.

00:06:57.989 --> 00:07:00.108
Here's a photo of this paper cube I made

00:07:00.109 --> 00:07:03.868
along with a real cube. As you can see

00:07:03.869 --> 00:07:07.348
mathematically speaking, they are the same thing,

00:07:07.349 --> 00:07:09.268
they just look very, very different.

NOTE Scrambling

00:07:09.269 --> 00:07:14.308
So the scramble algorithm itself,

00:07:14.309 --> 00:07:19.428
I pondered how this would even be done. In the competitions,

00:07:19.429 --> 00:07:21.588
They do this in a very, very elaborate way.

00:07:21.589 --> 00:07:22.748
They generate a random cube,

00:07:22.749 --> 00:07:25.388
they try to solve it, and if it's solvable

00:07:25.389 --> 00:07:28.548
they use these solution moves

00:07:28.549 --> 00:07:30.828
to turn into a scramble basically.

00:07:30.829 --> 00:07:34.948
And they also make sure to canonicalize the moves,

00:07:34.949 --> 00:07:38.548
so if you have subsequent moves that can be simplified,

00:07:38.549 --> 00:07:40.588
they do simplify these as much as possible.

00:07:40.589 --> 00:07:41.228
For example,

00:07:41.229 --> 00:07:43.748
if you have two subsequent rotations in one direction,

00:07:43.749 --> 00:07:46.668
it's turned into a different kind of rotation,

00:07:46.669 --> 00:07:49.388
so 90 and 90 equals 180.

00:07:49.389 --> 00:07:53.308
And the other Elisp scramblers I looked at,

00:07:53.309 --> 00:07:55.108
they generate random moves.

00:07:55.109 --> 00:07:57.508
Some of them do canonicalize. Not all of them.

00:07:57.509 --> 00:08:00.908
This one tries to do the best low-fi thing,

00:08:00.909 --> 00:08:02.388
that is, generating random moves,

00:08:02.389 --> 00:08:04.028
canonicalizing and repeating

00:08:04.029 --> 00:08:09.548
until enough have been generated.

NOTE Visualization

00:08:09.549 --> 00:08:13.148
For the visualization I had to figure out

00:08:13.149 --> 00:08:14.508
something else too complicated.

00:08:14.509 --> 00:08:17.228
For this, I tried to figure out

00:08:17.229 --> 00:08:19.868
where every facelift would end up in the puzzle view

00:08:19.869 --> 00:08:21.428
when you would unfold it.

00:08:21.429 --> 00:08:25.668
And for this, I did not consider the facelet orientation.

00:08:25.669 --> 00:08:29.268
This may be important later for some other puzzles

00:08:29.269 --> 00:08:31.148
where you can end up with very twisted faces,

00:08:31.149 --> 00:08:33.028
but for simple cubes, it's not a problem.

00:08:33.029 --> 00:08:36.308
My initial prototype used colored text,

00:08:36.309 --> 00:08:38.748
but later, I used the SVG library.

00:08:38.749 --> 00:08:41.588
It turned out to be easy enough to use, actually.

00:08:41.589 --> 00:08:46.108
Currently, I have hard-coded face-color mappings,

00:08:46.109 --> 00:08:49.108
but I plan to replace this so that theming is possible.

00:08:49.109 --> 00:08:51.588
For example, if you happen to have a cube

00:08:51.589 --> 00:08:54.689
that does not have the same color mappings as I do,

00:08:54.690 --> 00:08:56.468
then you should be able to fix this.

NOTE UI with Transient

00:08:56.469 --> 00:09:01.428
Next challenge was to build

00:09:01.429 --> 00:09:03.948
a beautiful intuitive UI with Transient.

00:09:03.949 --> 00:09:06.868
The reason why I chose this is

00:09:06.869 --> 00:09:10.348
because it would be self-documenting and Magit-style,

00:09:10.349 --> 00:09:12.348
and everyone knows how Magit works basically.

00:09:12.349 --> 00:09:15.308
Since Transient has become part of Emacs,

00:09:15.309 --> 00:09:17.228
there is really no reason to not try it out.

00:09:17.229 --> 00:09:21.668
The problem was documentation is difficult to understand.

00:09:21.669 --> 00:09:23.388
It's very abstract and high level,

00:09:23.389 --> 00:09:25.868
and it's hard to figure out. "Okay,

00:09:25.869 --> 00:09:26.788
I want to do something,

00:09:26.789 --> 00:09:28.908
how am I supposed to do this?"

00:09:28.909 --> 00:09:33.348
I did find transient-showcase, which has lots of examples,

00:09:33.349 --> 00:09:35.628
but they don't really feel finished

00:09:35.629 --> 00:09:39.068
and not realistic enough.

00:09:39.069 --> 00:09:40.748
When I tried to use the package,

00:09:40.749 --> 00:09:42.908
I got plenty of unhelpful error messages

00:09:42.909 --> 00:09:44.108
when using it incorrectly.

00:09:44.109 --> 00:09:45.948
I did manage to figure it out,

00:09:45.949 --> 00:09:50.588
but I plan to find more actual examples of it,

00:09:50.589 --> 00:09:53.428
to have an executable reference basically

00:09:53.429 --> 00:09:55.628
and try to improve my use of it.

NOTE Book-keeping with SQLite

00:09:55.629 --> 00:10:01.548
For the book-keeping, I used SQLite.

00:10:01.549 --> 00:10:04.548
This is a very recent addition to Emacs,

00:10:04.549 --> 00:10:07.308
it only appeared in the current major version.

00:10:07.309 --> 00:10:09.388
It's still very early days.

00:10:09.389 --> 00:10:13.028
I found some oddities, one of them turned out to be

00:10:13.029 --> 00:10:14.828
a bug in the transaction macro.

00:10:14.829 --> 00:10:17.588
Like basically, if you do an SQL transaction

00:10:17.589 --> 00:10:20.188
and an error happens, then every helper I found

00:10:20.189 --> 00:10:20.948
does a rollback on an error.

00:10:20.949 --> 00:10:26.748
But this one did not. It actually committed on an error,

00:10:26.749 --> 00:10:29.868
and this was very weird to figure out.

00:10:29.869 --> 00:10:34.308
I reported a bug. Eli was nice enough to send me a patch.

00:10:34.309 --> 00:10:35.428
We did some patch review,

00:10:35.429 --> 00:10:37.988
and he ended up fixing it properly.

00:10:37.989 --> 00:10:45.668
So yes, there's still a lot to be done there, and yeah,

00:10:45.669 --> 00:10:46.908
the API is very basic.

00:10:46.909 --> 00:10:48.908
You don't have convenience helpers

00:10:48.909 --> 00:10:51.308
like fetch the first row or fetch the first value

00:10:51.309 --> 00:10:54.429
or anything, but they're easy enough to write yourself.

00:10:54.430 --> 00:10:56.369
And the biggest challenge with this bookkeeping part

00:10:56.370 --> 00:10:58.028
was figuring out a decent schema,

00:10:58.029 --> 00:11:00.148
like how to organize data correctly

00:11:00.149 --> 00:11:02.348
so that it would not be awkward to manipulate.

00:11:02.349 --> 00:11:05.748
And with this, you can finally build a package

00:11:05.749 --> 00:11:07.388
that remembers its state properly

00:11:07.389 --> 00:11:10.468
and don't have to run into foot guns

00:11:10.469 --> 00:11:12.628
with Lisp-style serialization, deserialization.

NOTE Conclusion

00:11:12.629 --> 00:11:18.188
So yes, that concludes it so far.

00:11:18.189 --> 00:11:22.188
So what did I learn from this exercise?

00:11:22.189 --> 00:11:24.508
Well, there are still plenty of packages

00:11:24.509 --> 00:11:25.588
for Emacs to be written.

00:11:25.589 --> 00:11:28.908
If you think everything you can think of

00:11:28.909 --> 00:11:31.348
or you need has already been written, well, guess what?

00:11:31.349 --> 00:11:31.788
No.

00:11:31.789 --> 00:11:34.044
These are still plenty of specialized things

00:11:34.045 --> 00:11:36.788
that could need your help.

00:11:36.789 --> 00:11:39.788
These cubes do not require advanced mathematics,

00:11:39.789 --> 00:11:41.148
contrary to what you may think.

00:11:41.149 --> 00:11:44.708
Yes, you can apply advanced mathematics to them

00:11:44.709 --> 00:11:47.468
if you want to, but you don't have to.

00:11:47.469 --> 00:11:50.988
What surprised me about this is basically group theory.

00:11:50.989 --> 00:11:52.068
I've heard of it before.

00:11:52.069 --> 00:11:53.828
It seemed to be a meme, basically,

00:11:53.829 --> 00:11:56.468
because it has been like mostly Haskell people

00:11:56.469 --> 00:11:58.188
being very excited about this

00:11:58.189 --> 00:12:02.508
and it seemed kind of, like, divorced from reality, basically.

00:12:02.509 --> 00:12:05.948
But this puzzle, it actually proves that yes,

00:12:05.949 --> 00:12:06.948
it has its use.

00:12:06.949 --> 00:12:08.428
It definitely has.

00:12:08.429 --> 00:12:11.388
You just have to find the right problem matching it,

00:12:11.389 --> 00:12:13.468
and yeah.

00:12:13.469 --> 00:12:15.388
So yeah, once I understand it better,

00:12:15.389 --> 00:12:18.548
the topic, I expect to write better code.

00:12:18.549 --> 00:12:24.468
These new Emacs features, they work well enough.

00:12:24.469 --> 00:12:25.908
There are some rough edges.

00:12:25.909 --> 00:12:27.428
They definitely need more testing.

00:12:27.429 --> 00:12:30.668
So please, please, everyone,

00:12:30.669 --> 00:12:34.548
if you write Elisp, please try SQLite or Transient

00:12:34.549 --> 00:12:36.708
or anything else that looks cool and shiny.

00:12:36.709 --> 00:12:38.468
Report bugs.

00:12:38.469 --> 00:12:41.588
Find ways to improve them, anything. And yeah,

00:12:41.589 --> 00:12:44.868
I'm sure that if we do this,

00:12:44.869 --> 00:12:47.668
then Emacs will continue to get even better.

00:12:47.669 --> 00:12:51.788
So yeah, what's next for this package?

00:12:51.789 --> 00:12:55.988
Well, I could... There are lots of obvious UI improvements

00:12:55.989 --> 00:12:57.348
and testing to be done.

00:12:57.349 --> 00:12:59.708
I basically want to reach feature parity

00:12:59.709 --> 00:13:02.428
with the twisty-timer app, which this is very much inspired by.

00:13:02.429 --> 00:13:06.668
I want nice-looking stats like graphical ones

00:13:06.669 --> 00:13:08.788
instead of just a simple list of times.

00:13:08.789 --> 00:13:11.228
And I want support for more puzzles, of course,

00:13:11.229 --> 00:13:12.548
not just the simple cubes,

00:13:12.549 --> 00:13:14.588
but as I progress learning these puzzles,

00:13:14.589 --> 00:13:18.068
I want to have Emacs supporting me for this.

00:13:18.069 --> 00:13:22.428
But generally, it's a very open-ended package.

00:13:22.429 --> 00:13:26.628
And this concludes the talk.

00:13:26.629 --> 00:13:30.909
Thank you very much.