path: root/2024/captions/emacsconf-2024-gypsum--gypsum-my-clone-of-emacs-and-elisp-written-in-scheme--ramin-honary--answers.vtt

WEBVTT

00:00.069 --> 00:01.850
Troy Hinckley's project that I'm talking about. I was going

00:02.350 --> 00:22.139
to mention this in my presentation, but it's possible,

00:02.350 --> 00:22.139
theoretically, that Troy Hinckley, his project could be

00:02.350 --> 00:22.139
used as a scheme of limitation that actually runs my own

00:02.350 --> 00:22.139
version of Emacs. And although, you know, This is

00:25.478 --> 00:29.380
completely theoretical, and I don't know how difficult

00:25.478 --> 00:29.380
that would be. But if Troy Hinckley implemented enough of

00:30.781 --> 00:47.029
the R7-RS standard in Rust, it would theoretically be

00:30.781 --> 00:47.029
possible to run the Gypsum editor in Troy Hinckley's own

00:30.781 --> 00:47.029
editor. I thought that was kind of interesting, and I

00:48.270 --> 00:53.833
thought it was worth mentioning, at least in the questions

00:48.270 --> 00:53.833
and answers.

01:12.179 --> 01:14.080
I also mentioned this in the presentation. I wanted to see

01:14.940 --> 01:22.364
Robin Templeton's project presentation, but

01:14.940 --> 01:22.364
unfortunately it's going to be at like four in the morning

01:14.940 --> 01:22.364
for me. So I'm going to try and watch that tomorrow, but

01:22.984 --> 01:31.428
that's also going to be a very interesting project to keep an

01:22.984 --> 01:31.428
eye on if you're interested in Scheme. That's the project

01:33.149 --> 01:38.051
where you've got the Guylain interpreter running inside of

01:33.149 --> 01:38.051
the Emacs process. It's dynamically linked as a library.

02:04.699 --> 02:06.748
I'm ready for questions from anybody. You can ask or you can

02:07.431 --> 02:09.079
type. It's up to you.

02:32.319 --> 02:34.521
Okay, let me check the etherpad.

02:37.304 --> 02:38.245
Let's see here.

02:41.208 --> 02:42.830
I'm not sure if I'm doing that right.

02:46.373 --> 02:47.554
Let me check one more time. Oh, there it goes.

02:54.221 --> 02:55.702
Let's see, so this is...

03:00.151 --> 03:02.072
I didn't know about that first bit of history. Oh, I've heard

03:02.332 --> 00:03:09.369
RMS say that Scheme Guile is just a nicer Lisp, but I didn't

03:02.332 --> 03:09.776
know there were concrete talks attempts to use Guile for

03:02.332 --> 03:09.776
Emacs that early. Let's see, that was from janneke.

NOTE Q: I'm curious to know how the hell guile-emacs deals with all of the dynamically scoped modules out there. Is there any effort to automatically modularize and namespace stuff?

00:03:09.370 --> 00:03:19.241
I'm curious to know how the hell Guile Emacs deals with all the

03:14.318 --> 03:19.241
dynamically scoped modules out there. Is there any effort

03:20.181 --> 03:24.943
to automatically modularize and name? Let's see.

03:30.523 --> 03:35.806
That might be a better question for Robin Templeton. In my

03:36.727 --> 03:46.573
own project,

03:36.727 --> 03:46.573
there's no module system for Emacs Lisp. There is a module

03:46.693 --> 03:48.234
system for Scheme. And the Emacs Lisp interpreter runs in

03:49.695 --> 03:55.158
its own environment. the require system or whatever module

03:57.068 --> 04:11.736
system that Emacs has, once it's implemented, all of that

03:57.068 --> 04:11.736
would just happen inside of the Emacs Lisp environment,

03:57.068 --> 04:11.736
which is inside of the Scheme environment. And

04:12.437 --> 04:15.898
environments are objects in Scheme.

04:21.522 --> 04:24.103
I think a more difficult question is how to handle

04:26.420 --> 04:31.942
threading, and Scheme has very good threading built in, in

04:26.420 --> 04:31.942
Serphe-18[??].

04:34.283 --> 04:48.028
But I don't think it will be easy to write Emacs Lisp form

04:34.283 --> 04:48.028
bindings to the Scheme multi-threading implementation.

04:48.548 --> 04:50.749
Emacs Lisp was just not cut out for that kind of thing. So I

04:51.710 --> 04:59.894
think each Emacs Lisp, you could, I suppose, have multiple

04:51.710 --> 04:59.894
threads each running their own Emacs Lisp environment.

05:01.375 --> 05:02.956
Scheme would make that very simple to do.

05:06.018 --> 05:16.744
And then there'd just be a question of how you would get those

05:06.018 --> 05:16.744
different interpreters to communicate with each other,

05:06.018 --> 05:16.744
perhaps using the same protocol that's used by the Emacs

05:06.018 --> 05:16.744
server. But I haven't thought that far ahead yet.

NOTE Q: Would it be possible to support a GUI toolkit other than GTK?

05:23.646 --> 05:28.709
Would it be possible to support a GUI toolkit other than the

05:23.646 --> 05:28.709
GTK? Like, how is it still supports Lucid? Yes, this is

05:31.291 --> 05:33.232
absolutely a goal of the project. I'm trying to keep the back

05:33.873 --> 05:38.416
end separate as possible. The scheme has what you call

05:39.817 --> 05:42.478
parameters. And these are like global variables that are

05:43.199 --> 05:46.221
still somewhat thread safe. And every call to the GUI goes

05:47.484 --> 05:51.225
through a parameter. So the Emacs, the interpreter and the

05:52.125 --> 05:59.367
editor logic is all in one module. And then that module calls

05:59.987 --> 06:04.309
out into a separate GUI module. And then you can implement

06:04.989 --> 06:07.690
different GUI modules. So you could have one for GTK3, one

06:08.430 --> 06:13.171
for GTK4, if you want to write the extern C bindings around Qt

06:13.843 --> 06:20.725
or full tick, that would certainly be possible as well. It

06:21.185 --> 06:32.168
would be nice maybe to have an SDL implementation based

06:21.185 --> 06:32.168
maybe on Chikiti or some kind of immediate mode GUI,

06:21.185 --> 06:32.168
something like that. But definitely GTK3 through Guile GI

06:33.808 --> 06:38.750
is the reference implementation. Things start there. But

06:41.298 --> 06:43.959
I'm very interested in supporting other GUIs, yes. Let's

06:45.199 --> 00:06:45.256
see.

NOTE Q: Do you plan to provide improvements to Elisp as a language, or is the focus on a compatibility layer to facilitate doing all new extensions, etc. in Scheme?

00:06:45.257 --> 00:06:45.879
Question, do you plan to provide improvements to ELisp

06:47.540 --> 06:56.342
as a language or focus on a compatibility layer to

06:47.540 --> 06:56.342
facilitate all new extensions in Scheme? Yeah, the second

06:57.142 --> 06:57.962
one. I want to move off to Scheme. I would like for this

07:03.384 --> 07:05.264
project to try and keep up to date with each new release of

07:05.666 --> 07:10.789
Emacs and Emacs Lisp. That's a difficult moving target to

07:11.850 --> 07:14.552
follow, I realize. But to the greatest extent possible, any

07:15.152 --> 07:23.397
new features to Emacs Lisp will be pulled in from GNU Emacs.

07:25.419 --> 07:29.041
If we happen to be able to implement something cool in

07:25.419 --> 07:29.041
Scheme, and be able to port it over to Emacs Lisp, then sure,

07:29.437 --> 07:36.543
it'd be nice to be able to upload or to submit that upstream to

07:29.437 --> 07:36.543
the GNU Emacs. But I think I would prefer to have new features

07:38.584 --> 07:43.708
written in Scheme. I would like this gypsum to be more of a

07:43.989 --> 07:52.075
Scheme app platform that just happens to be able to also run

07:43.989 --> 07:52.075
Emacs Lisp. That's how I see it. Of course, this will be a

07:54.577 --> 07:56.699
community project. I'm open to debate about that if anybody

07:58.809 --> 08:02.012
wants to convince me otherwise.

08:08.439 --> 08:11.683
Why is being able to interpret all of that EL a useful goal?

08:12.464 --> 08:14.626
Sure, there is a lot of code written in Elisp. Can we

08:15.206 --> 08:17.749
consider... Oh, it's still being written. Please go ahead

08:18.390 --> 08:19.491
and finish writing.

NOTE Q: Can we consider a translator like utility to convert elisp to scheme, once guile-emacs becomes a reality?

08:29.673 --> 08:35.576
Can we consider a translator like utility to convert eLisp

08:29.673 --> 08:35.576
to Scheme once Guile-Emacs has become a reality?

08:36.716 --> 08:37.076
Certainly. For the time being, I just wanted to get the

08:38.717 --> 08:42.639
interpreter running. So the actual, the Guile-Emacs Lisp,

08:44.520 --> 08:58.666
the one that was written in 2011 that I didn't write, that

08:44.520 --> 08:58.666
actually does compile to, I think it's the tree

08:44.520 --> 08:58.666
intermediate representation It's one of the intermediate

08:59.076 --> 09:03.697
languages that Guile uses to compile Guile scheme itself.

09:04.817 --> 09:09.299
So the Emacs lisp that was written before actually does

09:04.817 --> 09:09.299
that. It actually compiles and makes use of the entire Guile

09:09.339 --> 09:20.761
compiler tool chain and actually produces like JIT

09:09.339 --> 09:20.761
compilable binaries, which is really cool. Like I said,

09:23.342 --> 09:25.943
that's the one that I had trouble getting to work properly.

09:29.209 --> 09:30.890
Maybe we can follow that architecture. I'm not sure how to do

09:33.052 --> 09:45.102
that, but I would like to be able to do some kind of

09:33.052 --> 09:45.102
translating, keeping in mind that we want to have this be

09:33.052 --> 09:45.102
portable, do various schemes. And so Guile makes this very

09:45.988 --> 09:50.289
easy, but other schemes don't. Gambit might do this pretty

09:51.549 --> 09:53.530
well as well. It compiles to C and then compiles C down to a

09:53.950 --> 10:01.471
dynamically linkable library. So yeah, I think probably

10:03.372 --> 10:09.373
the most portable, I'm just thinking out loud right now,

10:10.652 --> 10:21.715
most portable implementation will just be able to

10:10.652 --> 10:21.715
translate Emacs Lisp directly to Scheme, which is not what

10:10.652 --> 10:21.715
the old Guile Emacs Lisp implementation does. That goes to

10:21.755 --> 10:26.777
TreeIL, so it's very, very Guile-specific, can't be

10:21.755 --> 10:26.777
ported. But yeah, if we could somehow get Emacs Lisp

10:28.359 --> 10:42.045
translated to Scheme and then compiled, say, in Shea Scheme

10:28.359 --> 10:42.045
or Gambit or MIT Scheme or one of those other compilers, that

10:28.359 --> 10:42.045
would be very cool. And I would absolutely love to do that.

10:44.906 --> 10:49.948
And I would very quickly accept any code into the code base

10:44.906 --> 10:49.948
that would do that.

NOTE Q: Why is being able to interpret all of \`init.el\` an useful goal?

10:54.390 --> 10:56.291
Oh, and to answer the question about init.el,

10:59.207 --> 11:17.215
It's just because people spend a lot of time on their configs

10:59.207 --> 11:17.215
and it would be nice if, you know, you're starting to use this

10:59.207 --> 11:17.215
new editor and want it to be similar to Emacs users, just the

10:59.207 --> 11:17.215
Emacs community in general and people who are familiar with

10:59.207 --> 11:17.215
using Emacs. It would be more useful to everybody in the

11:17.715 --> 11:25.379
Emacs community if this were more compatible with GNU

11:17.715 --> 11:25.379
Emacs. And so that's why that's, I think that's an important

11:25.679 --> 11:27.960
goal.

11:34.465 --> 11:35.467
Question is not yet. Great. Oh, here comes another

11:38.471 --> 11:39.613
question.

NOTE Q: What is the plan to handle elisp packages that depend on 3rd party/external libraries? (libgit/magit or rg/ripgrep)?

12:08.539 --> 12:17.742
Okay, what is the plan to handle elisp packages that depend

12:08.539 --> 12:17.742
on third-party or external libraries like git or magit

12:08.539 --> 12:17.742
or ripgrep? So that's going to be tricky. It depends on how

12:21.523 --> 12:26.224
these external packages are linked into emacs. If it's

12:26.844 --> 12:33.646
going to be a dynamic library like Robin Templeton's

12:26.844 --> 12:33.646
project which you load the libgit library into the Emacs

12:35.289 --> 12:41.931
process, that is going to be extremely difficult. So if you

12:44.032 --> 12:52.975
have an external library like, I don't know, libgit or

12:44.032 --> 12:52.975
what's the GUI thing? Cabal. No, not Cabal. Cairo, libcairo

12:57.736 --> 13:01.398
to do SVG graphics and so on.

13:04.483 --> 13:17.480
You can do that very easily with Guile, but then on top of

13:04.483 --> 13:17.480
that, implementing Emacs list bindings to it, I mean,

13:04.483 --> 13:17.480
you've got two layers there, and that makes things pretty

13:04.483 --> 13:17.480
difficult. So it's possible. And to some degree, maybe

13:21.935 --> 13:30.842
necessary for example, Cairo, if we want to do SVG graphics

13:21.935 --> 13:30.842
the way that Emacs Lisp does, we're going to have to have

13:21.935 --> 13:30.842
that. So that would be necessary. We would have to have those

13:32.643 --> 13:33.944
two layers. Yes, let's do that. But if it's like for Magit,

13:38.047 --> 13:50.596
you can just call out to your git process, and then you're

13:38.047 --> 13:50.596
just using the regular process APIs that Emacs Lisp has. And

13:51.451 --> 13:58.475
that can be, already we, like Guile has some very good

13:51.451 --> 13:58.475
implementations for process management. And so it would

13:59.055 --> 14:05.438
just be a matter of wrapping up those in the Emacs lisp form

13:59.055 --> 14:05.438
bindings. So yeah, dynamic libraries, I wanna try to avoid.

14:12.222 --> 14:20.366
And I would prefer to do things more through, you know,

14:12.222 --> 14:20.366
launching a child process in the Emacs process. and then

14:20.956 --> 14:24.798
communicating over the standard in, standard out

14:20.956 --> 14:24.798
channels.

14:29.460 --> 14:40.386
That's the easier way to do things, I think, because then you

14:29.460 --> 14:40.386
can just use the process library that Emacs already has, and

14:29.460 --> 14:40.386
you can just reuse all of that code.

14:43.969 --> 14:49.912
I'm not sure how ripgrep works, unfortunately, but I

14:43.969 --> 14:49.912
believe that's also a process, a child process. So, we can

14:50.412 --> 14:53.774
just reuse all of the Emacs Lisp code that does that already.

14:54.014 --> 15:05.979
We just need to make sure that the process management

14:54.014 --> 15:05.979
implementation and scheme is properly bound to Emacs Lisp,

14:54.014 --> 15:05.979
and it works the same as GNU Emacs does. Once that's all set,

15:06.360 --> 15:13.383
then these porcelains, like around git, should fall into

15:06.360 --> 15:13.383
place. without too much difficulty, hopefully.

NOTE Q: Not really a question, but how about Schemacs as a name?

15:21.112 --> 15:22.593
How about Schemax as a name? I like the name. I like that name.

15:28.937 --> 15:32.920
I haven't really looked into like, is that already used or is

15:28.937 --> 15:32.920
that going to be confusing? But certainly something we can

15:33.380 --> 15:35.021
discuss.

15:38.243 --> 15:39.264
Another thing I should mention,

15:42.157 --> 15:48.278
I should probably set up a server or something like Discord

15:42.157 --> 15:48.278
or something like that. Discourse, not Discord.

15:51.619 --> 15:56.220
Discourse, the open source one, where we could actually

15:51.619 --> 15:56.220
chat about this stuff. For the time being, ActivityPub,

15:56.540 --> 16:05.562
mostly Mastodon, is how I communicate with people in real

15:56.540 --> 16:05.562
time, that or email. So if you want to get a hold of me, check

16:09.809 --> 16:15.571
the notes for this presentation and just send me an email.

16:16.752 --> 16:18.012
Any question at all is fine. If you want to contribute code,

16:19.633 --> 16:25.495
if you want to just learn how to contribute code, send me any

16:19.633 --> 16:25.495
questions. It's fine. I'm happy to answer them. And we can

16:30.256 --> 16:31.757
talk about the name as well.

NOTE Q: Why is it not feasible for the Emacs layer that interprets Emacs Lisp (the core in C) ot have a Scheme interpreter, instead of using Guile?

16:45.931 --> 16:54.215
Okay, why is it not feasible for the Emacs layer that

16:45.931 --> 16:54.215
interprets Emacs Lisp, the core in C, have a Scheme

16:45.931 --> 16:54.215
interpreter instead of using Guile? Let's see, I have to,

16:55.496 --> 16:57.257
okay. Emacs layer interprets Emacs Lisp, the core in C, have

16:57.737 --> 17:05.942
a Scheme interpreter instead of using Guile. Okay, so that,

17:07.362 --> 17:13.906
the question xlarsx is asking, xlars, x, So Lars is asking,

17:14.744 --> 17:28.093
is it not feasible for there to be an

17:14.744 --> 17:28.093
Emacs layer that interprets Emacs Lisp have a scheme

17:14.744 --> 17:28.093
interpreter? This is Robin Templeton's project. And

17:30.815 --> 17:32.156
they're presenting later today. So check the roster and be

17:32.697 --> 17:41.303
sure to see that presentation because that's exactly what

17:32.697 --> 17:41.303
Robin Templeton is doing. That's not what I'm doing though.

17:44.419 --> 17:46.459
I'm trying to create something in Scheme. But yes, there is

17:48.280 --> 17:54.921
an attempt to get an Scheme interpreter to run inside of

17:48.280 --> 17:54.921
Emacs itself. And it has its own method of binding to Emacs

17:55.181 --> 18:05.323
Lisp functions and translating data like Lisp structures

17:55.181 --> 18:05.323
between Guile Scheme and Emacs Lisp. Robin will explain all

18:05.943 --> 18:08.284
of that in their presentation.

18:28.519 --> 18:33.020
OK, I think I've got through all the questions on Etherpad.

18:33.620 --> 18:35.500
But I'm going to hang out here for a bit longer. And yeah, feel

18:37.621 --> 18:46.182
free to do a video chat with me or send me more questions on

18:37.621 --> 18:46.182
Etherpad or here in the big blue button. And so I'm just going

18:47.002 --> 18:48.082
to hang out. And thanks for asking all your questions. And

18:51.663 --> 18:56.024
yeah, I look forward to working with all of you if you're

18:51.663 --> 18:56.024
interested. take it easy. Thanks so much for the talk and

18:59.935 --> 19:08.180
looking forward to seeing some of your progress as this

18:59.935 --> 19:08.180
moves forward, exciting space. We'll go ahead and leave the

19:09.261 --> 19:14.925
room open for you and thanks for offering to hang out and chat

19:09.261 --> 19:14.925
with other people that come by. Feel free to throw something

19:15.025 --> 19:18.287
in the chat if you want to remind people you're still here.

19:19.557 --> 19:25.143
Meanwhile, on the stream, we have moved along to our next

19:19.557 --> 19:25.143
talk on Rust, and that is just getting started. But again,

19:25.283 --> 19:30.549
we're continuing to record this, and I'll just keep an eye on

19:25.283 --> 19:30.549
it to stop the recording. Thank you. Thank you. It was

19:33.352 --> 19:33.853
awesome.

21:47.935 --> 21:50.558
So it seems like it's slowed down here for the Q&A. I don't see

21:50.638 --> 21:53.741
anybody else on BBB, so I'm going to go ahead and stop the

21:50.638 --> 21:53.741
recording. We can start it back up. I would say, yes, there's

21:55.282 --> 21:58.906
a lot of things you can do with this. You can handle

21:58.926 --> 22:00.627
processing. Yeah, I'm going to try and join over the chat for

22:02.029 --> 22:07.614
the next talk. I'm not sure if I can do both big blue buttons at

22:08.635 --> 22:11.538
the same time. You should be able to just watch your mute

22:13.206 --> 22:19.998
settings and mute tab settings and whatever all you have to

22:13.206 --> 22:19.998
avoid bleed through. Okay.