WEBVTT captioned by sachac, checked by bhavin NOTE Rune 00:00:00.000 --> 00:00:05.119 Hello, EmacsConf. My name is Troy Hinckley, and this is my 00:00:05.120 --> 00:00:08.759 talk on Rune, a Rust implementation in Emacs. We strive to be 00:00:08.760 --> 00:00:11.839 bug compatible with Emacs, so you can use the same Elisp. 00:00:11.840 --> 00:00:14.879 It's still a fairly early stage experimental project, and 00:00:14.880 --> 00:00:17.081 we have some basic things implemented. NOTE The Emacs core 00:00:17.082 --> 00:00:19.946 Before I get started, I want to talk a bit more 00:00:19.947 --> 00:00:21.847 about what the core is. 00:00:21.848 --> 00:00:24.559 So the Emacs core, it includes the runtime, the interpreter, 00:00:24.560 --> 00:00:26.439 garbage collector, everything used to run the code. 00:00:26.440 --> 00:00:29.799 It includes the GUI. It includes all the data structures. 00:00:29.800 --> 00:00:31.919 If you look underneath all the Elisp data structures, 00:00:31.920 --> 00:00:33.599 there's C code underneath there, 00:00:33.600 --> 00:00:35.559 as well as the auxiliary functions 00:00:35.560 --> 00:00:39.239 of which there's about 1500. In making this talk, I don't 00:00:39.240 --> 00:00:40.919 want to give the impression that I'm saying the core is 00:00:40.920 --> 00:00:42.879 outdated or that needs to be replaced or that it can't be 00:00:42.880 --> 00:00:45.519 evolved on its own, because clearly it has continued to 00:00:45.520 --> 00:00:48.319 evolve. If we look in just the last few years, we can see that 00:00:48.320 --> 00:00:50.439 we've added native compilation, we've added tree-sitter 00:00:50.440 --> 00:00:52.759 support, we've added color emoji, and there's work right 00:00:52.760 --> 00:00:57.167 now to add a new garbage collector to Emacs as well. NOTE Why create this? 00:00:57.168 --> 00:01:01.071 Why create this project? Emacs has a long history. 00:01:01.072 --> 00:01:04.535 It has a lot of users. It needs to support a big community. 00:01:04.536 --> 00:01:06.837 Because of that, it has to be very conservative 00:01:06.838 --> 00:01:10.321 about what things it can allow into the project. 00:01:10.322 --> 00:01:11.639 Forks like this create an 00:01:11.640 --> 00:01:15.586 opportunity to experiment and try new approaches. 00:01:15.587 --> 00:01:18.799 This is particularly a good use case for Rust because the C core, 00:01:18.800 --> 00:01:20.849 it's pretty well tested. It's been around for a long time. 00:01:20.850 --> 00:01:22.959 A lot of the bugs have been ironed out, but when you're doing a 00:01:22.960 --> 00:01:26.439 new greenfield project, it's very easy to introduce new 00:01:26.440 --> 00:01:28.774 undefined behavior and memory unsafety 00:01:28.775 --> 00:01:32.376 and stuff like that. Rust protects us from most of that, 00:01:32.377 --> 00:01:34.937 but it also gives us the ability to be fast 00:01:34.938 --> 00:01:37.883 and has a strong ecosystem behind it. 00:01:37.884 --> 00:01:40.399 Rust is also really good at multi-threading. 00:01:40.400 --> 00:01:43.399 Their phrase in the community is fearless concurrency. 00:01:43.400 --> 00:01:45.559 They should be able to write concurrent programs without 00:01:45.560 --> 00:01:49.319 having to worry about data races. It's also really high 00:01:49.320 --> 00:01:51.839 performance. It has a really good regex engine. It's known 00:01:51.840 --> 00:01:55.864 for its non-copy I/O as well. NOTE How does this compare to other projects? 00:01:55.865 --> 00:01:57.479 How does this compare to other 00:01:57.480 --> 00:01:59.919 Rust and Emacs projects, whether there's been a couple? The 00:01:59.920 --> 00:02:02.799 first is Remacs. This project was the first. It took an 00:02:02.800 --> 00:02:05.519 outside-in approach. Basically you could take a C 00:02:05.520 --> 00:02:09.319 function and replace it with a Rust function and build it 00:02:09.320 --> 00:02:11.799 together as one executable. This is pretty easy to do 00:02:11.800 --> 00:02:14.639 because they could both talk over the C ABI. You could 00:02:14.640 --> 00:02:16.479 swap out functions once at a time. They made really good 00:02:16.480 --> 00:02:20.279 progress at first, but eventually they ran into the problem 00:02:20.280 --> 00:02:23.079 that as you get down to the really core parts of it, you can't 00:02:23.080 --> 00:02:25.919 just replace one function at a time anymore, because some of 00:02:25.920 --> 00:02:28.159 that functionality is connected to other things. Like for 00:02:28.160 --> 00:02:30.359 example, you can't replace the garbage collector without 00:02:30.360 --> 00:02:32.759 replacing the entire garbage collection system. So the 00:02:32.760 --> 00:02:36.279 progress really kind of slowed down. Another issue with it 00:02:36.280 --> 00:02:38.839 was, is that they were doing a one-to-one rewrite, so they 00:02:38.840 --> 00:02:41.079 weren't adding any new features or functionality, just 00:02:41.080 --> 00:02:43.879 taking the same code and replacing it in Rust, which doesn't 00:02:43.880 --> 00:02:46.801 add any advantages in and of itself. 00:02:46.802 --> 00:02:50.399 This spawned Emacs-NG, which was kind of the spiritual successor to 00:02:50.400 --> 00:02:52.746 Remacs, where they decided to add new functionality, 00:02:52.747 --> 00:02:55.808 the biggest one being a JavaScript runtime, 00:02:55.809 --> 00:02:58.230 as well as some new renderers to Emacs. 00:02:58.231 --> 00:03:01.314 This is no longer actively developed though. NOTE Multi-threading 00:03:01.315 --> 00:03:04.079 In this project, one of the big focuses we have is 00:03:04.080 --> 00:03:07.559 on multi-threading. The C core itself is, everything is 00:03:07.560 --> 00:03:09.959 designed around being single-threaded, all the data 00:03:09.960 --> 00:03:13.039 structures and everything like that. Rust has a great 00:03:13.040 --> 00:03:15.719 concurrency story. In Rust, everything is intended to be 00:03:15.720 --> 00:03:18.199 multi-threaded. That doesn't mean that everything has to 00:03:18.200 --> 00:03:20.719 run on multiple threads, but you can't write something that 00:03:20.720 --> 00:03:22.719 is limited to only running in a single-threaded 00:03:22.720 --> 00:03:25.799 environment. So this makes it really easy to use all the 00:03:25.800 --> 00:03:28.039 existing packages and build something that is 00:03:28.040 --> 00:03:30.480 concurrency safe. which is what we've done here, 00:03:30.481 --> 00:03:32.440 and that was relatively easy to do. NOTE Multi-threading elisp 00:03:32.441 --> 00:03:34.781 But adding it to Elisp is the hard part, 00:03:34.782 --> 00:03:36.502 because we've got to come up with a good model 00:03:36.503 --> 00:03:39.624 for Lisp, and Elisp is just a giant ball 00:03:39.625 --> 00:03:41.479 of mutable state. We need to find some 00:03:41.480 --> 00:03:44.566 way to tame that so we can make workable concurrency 00:03:44.567 --> 00:03:47.647 out of it. There's really two ways you can do this. NOTE No-GIL method 00:03:47.648 --> 00:03:49.268 One is what I call the no-GIL method. 00:03:49.269 --> 00:03:51.399 This is what Python is doing, where 00:03:51.400 --> 00:03:53.919 you take all of your data structures, you make them 00:03:53.920 --> 00:03:56.439 concurrency safe, and then you just leave it up to the 00:03:56.440 --> 00:03:58.119 programmer to decide what they're going to do with it. 00:03:58.120 --> 00:04:00.467 They've got to build safe abstractions on top of that. 00:04:00.468 --> 00:04:03.086 One of the big downsides with this is that 00:04:03.087 --> 00:04:05.247 it comes with a pretty high cost. 00:04:05.248 --> 00:04:07.799 The last benchmarks I've seen is that by making 00:04:07.800 --> 00:04:10.879 everything concurrency safe in Python, single-threaded 00:04:10.880 --> 00:04:15.799 code is about 20% slower in some benchmarks. 00:04:15.800 --> 00:04:19.079 Since most code is single-threaded, this has a big 00:04:19.080 --> 00:04:21.039 performance impact for most code that isn't taking 00:04:21.040 --> 00:04:23.719 advantage of the multi-threading. The other thing is this 00:04:23.720 --> 00:04:26.279 introduces a lot of nasty concurrency bugs because you can 00:04:26.280 --> 00:04:29.039 have anything mutating any part of the data from any thread, 00:04:29.040 --> 00:04:32.637 even if you can't have memory unsafety per se. NOTE Actors 00:04:32.638 --> 00:04:34.738 The other option is actors, 00:04:34.739 --> 00:04:36.639 which are a really known way to approach this, 00:04:36.640 --> 00:04:39.079 where you trade some of that flexibility that you get 00:04:39.080 --> 00:04:43.719 with fully concurrent for more control and. Code and 00:04:43.720 --> 00:04:45.839 functions are shared between all the different threads, 00:04:45.840 --> 00:04:50.599 but data has to be passed along channels between different 00:04:50.600 --> 00:04:51.251 actors. NOTE Multi-threading elisp (functions) 00:04:51.252 --> 00:04:52.919 We want the functions to be shared, and this 00:04:52.920 --> 00:04:55.159 should be pretty easy because we don't mutate functions 00:04:55.160 --> 00:05:00.119 like we do data, except when we do. In Lisp, functions are 00:05:00.120 --> 00:05:03.239 just lists like anything else. So you can mutate them 00:05:03.240 --> 00:05:06.279 just like lists. Even if you're not talking about 00:05:06.280 --> 00:05:09.159 interpreted code, like if you have a native compiled 00:05:09.160 --> 00:05:11.959 function, you can still mutate the constants inside the 00:05:11.960 --> 00:05:14.839 function. For example, here we have a function returns a 00:05:14.840 --> 00:05:17.679 string. We take that string out, we mutate that string, and 00:05:17.680 --> 00:05:23.079 now the function returns a different string. In Rune, we 00:05:23.080 --> 00:05:24.999 enforce that all functions, their constants are 00:05:25.000 --> 00:05:27.199 immutable. You can't mutate the insides of a function. You 00:05:27.200 --> 00:05:29.239 can still swap out functions and redefine them, but you 00:05:29.240 --> 00:05:32.239 can't mutate the inside of a function. This enables them 00:05:32.240 --> 00:05:34.679 to be safely shared across threads. NOTE Caveats 00:05:34.680 --> 00:05:36.400 However, there are some caveats to this. 00:05:36.401 --> 00:05:38.159 For example, some functions actually do 00:05:38.160 --> 00:05:41.439 need to mutate their own data. The example that we run into is 00:05:41.440 --> 00:05:44.839 cl-generic. It uses a method cache. So it has to be able to 00:05:44.840 --> 00:05:47.639 update that cache. In this case, we just made a special 00:05:47.640 --> 00:05:50.799 case for this particular situation, but we don't know what 00:05:50.800 --> 00:05:53.159 more of these we're gonna run into the future where this is 00:05:53.160 --> 00:05:57.089 needed behavior to be able to mutate a function. NOTE Multi-threading elisp (data) 00:05:57.090 --> 00:05:59.810 Okay, so functions are pretty easy. 00:05:59.811 --> 00:06:00.919 They just can be shared between 00:06:00.920 --> 00:06:05.159 threads, but data can't be immutable, at least not into the 00:06:05.160 --> 00:06:08.759 model that Emacs currently has. We have two different 00:06:08.760 --> 00:06:12.039 ways to handle this. One is we require whenever you're 00:06:12.040 --> 00:06:14.399 calling some other code in a different thread, you have to 00:06:14.400 --> 00:06:17.039 send all the variables that it's going to need over to that 00:06:17.040 --> 00:06:19.159 thread. This is how you traditionally do inside actors. 00:06:19.160 --> 00:06:21.919 Any data that needs to go to a different actor needs to be sent 00:06:21.920 --> 00:06:25.519 over a channel. It's relatively easy implementation, but 00:06:25.520 --> 00:06:28.159 this is difficult in the Emacs case because everything is 00:06:28.160 --> 00:06:30.799 going to be accessing different variables. That means 00:06:30.800 --> 00:06:33.119 when you call something, you have to know ahead of time, all 00:06:33.120 --> 00:06:34.879 the different variables that are gonna be accessed inside 00:06:34.880 --> 00:06:38.248 that other thread and put those in when you call it. NOTE Copy values to other threads on demands 00:06:38.249 --> 00:06:40.959 The other option we're using is we're copying values to the 00:06:40.960 --> 00:06:43.439 other threads on demand. If you're running a thread, it 00:06:43.440 --> 00:06:45.759 tries to look up a variable. It doesn't have any value for 00:06:45.760 --> 00:06:48.759 that variable. It will go back and ask the main thread and it 00:06:48.760 --> 00:06:50.959 will copy that value into that thread and it can continue 00:06:50.960 --> 00:06:53.399 execution. This is nice because you can just launch some 00:06:53.400 --> 00:06:55.639 code and it'll take care of handling all the data transfer 00:06:55.640 --> 00:06:57.883 for you. NOTE Multi-threading elisp (buffers) 00:06:57.884 --> 00:07:00.359 But we don't want to be copying around is buffers, 00:07:00.360 --> 00:07:04.199 because they can be really large. In this case, we have a 00:07:04.200 --> 00:07:07.599 mutex. Each thread could only have one current buffer that 00:07:07.600 --> 00:07:12.279 it has an exclusive lock to. This comes with some 00:07:12.280 --> 00:07:16.079 trade-offs, big one being that if the user tries to access 00:07:16.080 --> 00:07:18.359 some buffer, they want to type something, and a background 00:07:18.360 --> 00:07:20.239 thread is holding onto that buffer, what do we do in that 00:07:20.240 --> 00:07:24.959 situation? And we still need to hold an exclusive lock, even 00:07:24.960 --> 00:07:26.359 if we're only going to read a buffer. If you have multiple 00:07:26.360 --> 00:07:29.159 readers, they each still need to take turns because we can't 00:07:29.160 --> 00:07:30.999 determine if at some point a thread is going to try and mutate 00:07:31.000 --> 00:07:33.879 the buffer. It has to be an exclusive lock. The other issue 00:07:33.880 --> 00:07:37.799 is buffer-locals. This is less of a implementation issue 00:07:37.800 --> 00:07:40.519 as much as it is a technical issue. Because you think about 00:07:40.520 --> 00:07:42.759 when we switch to a buffer, it has some buffer-local data and 00:07:42.760 --> 00:07:45.399 we have some thread-local data. As we go through, we're 00:07:45.400 --> 00:07:47.599 mutating everything. Those can get intertwined and 00:07:47.600 --> 00:07:49.719 pointing to each other. Then we switch away from that 00:07:49.720 --> 00:07:51.679 buffer. We need some quick way to be able to separate those 00:07:51.680 --> 00:07:54.279 out. The buffer-locals can go with the buffer-locals and 00:07:54.280 --> 00:07:56.439 the thread data can stay with thread data and make copies of 00:07:56.440 --> 00:07:58.719 anything that was pointing to the other side. But we don't 00:07:58.720 --> 00:08:02.839 have a good method to determine how to separate those two, 00:08:02.840 --> 00:08:05.359 like what data belongs to this and what data belongs to this, 00:08:05.360 --> 00:08:08.199 so that we can do that quickly. We haven't found a good 00:08:08.200 --> 00:08:09.599 solution to that yet, but it's something we're still 00:08:09.600 --> 00:08:11.902 working on. NOTE Would this actually be useful? 00:08:11.903 --> 00:08:13.079 The question is, would this actually be 00:08:13.080 --> 00:08:15.959 useful for doing real work inside Emacs? I would say, 00:08:15.960 --> 00:08:17.959 yes, there's a lot of things you can do with this. You could 00:08:17.960 --> 00:08:20.239 handle process output in the background. You can do syntax 00:08:20.240 --> 00:08:23.479 highlighting. You can do buffer search in parallel. You can 00:08:23.480 --> 00:08:26.679 do LSP. You can do fetching your mail in the background. You 00:08:26.680 --> 00:08:29.639 can have a window manager that doesn't block your window 00:08:29.640 --> 00:08:32.319 manager when Emacs is blocked. You could do 00:08:32.320 --> 00:08:34.479 something like a file system watcher that keeps up on files 00:08:34.480 --> 00:08:37.559 without blocking Emacs. This wouldn't be so great for 00:08:37.560 --> 00:08:39.159 building concurrent data structures or operating on 00:08:39.160 --> 00:08:42.199 shared data or building your own abstractions, because of the 00:08:42.200 --> 00:08:46.039 trade-offs that we've made here. Okay. That's talking 00:08:46.040 --> 00:08:46.918 about multi-threading. NOTE Precise garbage collection 00:08:46.919 --> 00:08:47.599 The other thing we're going to talk 00:08:47.600 --> 00:08:51.319 about is precise garbage collection. In Rune, we have a 00:08:51.320 --> 00:08:54.439 safe, precise garbage collection because of the Rust type 00:08:54.440 --> 00:08:58.119 system. Let's look at what the problem is with garbage 00:08:58.120 --> 00:09:00.479 collection in the first place. Really, the tricky part 00:09:00.480 --> 00:09:03.719 about garbage collection is rooting. How do we find out what 00:09:03.720 --> 00:09:06.159 the roots are? These are all the values that are on the 00:09:06.160 --> 00:09:08.679 stack or inside the registers. In this example here, we 00:09:08.680 --> 00:09:11.919 allocate an object. We call garbage_collect, that object's 00:09:11.920 --> 00:09:13.536 collected, and then we try and return it. 00:09:13.537 --> 00:09:16.536 It's no longer valid. NOTE How Emacs used to deal with roots 00:09:16.537 --> 00:09:19.039 Let's look at how Emacs used to deal with this 00:09:19.040 --> 00:09:22.559 problem way back in the day. There was a system called gcpro 00:09:22.560 --> 00:09:26.319 or GC Protect, which is basically designed that every time a 00:09:26.320 --> 00:09:28.919 value needed to survive past a garbage collection point, 00:09:28.920 --> 00:09:32.359 you had to try and protect it. In order to do this, you had 00:09:32.360 --> 00:09:35.439 to declare a struct, you had to put a macro around it to root 00:09:35.440 --> 00:09:37.999 the object, and then you had to unroot it when you were done-- 00:09:38.000 --> 00:09:41.559 past the garbage collection. Now the value is safe. You 00:09:41.560 --> 00:09:44.039 can see down here, I pulled these eight rules out from a 00:09:44.040 --> 00:09:46.919 really old version of the Emacs manual about all the things 00:09:46.920 --> 00:09:49.279 you had to keep track of when you were trying to use this 00:09:49.280 --> 00:09:52.319 system. All right, so there was a special handling for 00:09:52.320 --> 00:09:54.639 nested GC protects. You had to make sure the memory was 00:09:54.640 --> 00:09:58.239 initialized. You had to make sure that traps couldn't occur 00:09:58.240 --> 00:10:00.839 between allocating and when GC protect would happen. It 00:10:00.840 --> 00:10:03.319 can be tricky because you don't always know when a function 00:10:03.320 --> 00:10:06.879 that's getting called could potentially call garbage 00:10:06.880 --> 00:10:10.719 collection. So if you got something wrong, you also 00:10:10.720 --> 00:10:12.719 might not catch it for a long time because garbage 00:10:12.720 --> 00:10:15.719 collection may only get called one out of 99 times. The other 00:10:15.720 --> 00:10:18.999 99 times is just fine. That one time it happens and you 00:10:19.000 --> 00:10:22.559 can't reproduce the issue. When you do get this wrong and 00:10:22.560 --> 00:10:24.439 some, something doesn't get rooted and it gets 00:10:24.440 --> 00:10:26.319 overwritten, it generally doesn't show up right where the 00:10:26.320 --> 00:10:28.799 problem is. It gets showed up way later when you actually try 00:10:28.800 --> 00:10:31.479 and access the value and the value is invalid. You've got 00:10:31.480 --> 00:10:33.639 to track it back to where that thing did not get properly 00:10:33.640 --> 00:10:37.359 rooted. It's a huge source of bugs and very hard to 00:10:37.360 --> 00:10:38.712 maintain. NOTE Conservative stack scanning 00:10:38.713 --> 00:10:40.119 Emacs decided to go with a different path, 00:10:40.120 --> 00:10:42.399 which we call conservative stack scanning. Basically, 00:10:42.400 --> 00:10:45.239 the garbage collector just looks at the stack and all the 00:10:45.240 --> 00:10:47.959 registers and any data inside there that looks like it could 00:10:47.960 --> 00:10:52.279 be a pointer, it treats it as a pointer. This is nice because 00:10:52.280 --> 00:10:54.711 you get really easy root tracking, 00:10:54.712 --> 00:10:56.113 but it also comes with some trade-offs, 00:10:56.114 --> 00:11:00.156 mostly that your objects are no longer movable. NOTE Movable objects 00:11:00.157 --> 00:11:03.079 Why would we want movable objects in Emacs? 00:11:03.080 --> 00:11:05.839 There's a couple of different reasons. One is compaction. 00:11:05.840 --> 00:11:08.199 You can take all your heap, you can pack that on down because 00:11:08.200 --> 00:11:11.239 you can coalesce all your objects together. Another is that 00:11:11.240 --> 00:11:13.239 it's easy to implement generational garbage collection. 00:11:13.240 --> 00:11:16.039 You can just copy everything out of your minor heap into your 00:11:16.040 --> 00:11:21.839 older heap. Really, Emacs is kind of uniquely ideal for 00:11:21.840 --> 00:11:24.279 generational collection, because the typical way we 00:11:24.280 --> 00:11:27.799 interact with Emacs is as a series of commands. You execute 00:11:27.800 --> 00:11:29.959 some command, you'd execute the next command, you execute 00:11:29.960 --> 00:11:33.199 a command. It could be happening every key press, it could be 00:11:33.200 --> 00:11:36.759 happening with M-x. However long that command is, that is 00:11:36.760 --> 00:11:40.959 the ideal length for the minor collection generation, the 00:11:40.960 --> 00:11:43.399 first generation. Because once you're done with that 00:11:43.400 --> 00:11:45.879 generation, anything that's still existing is going to be 00:11:45.880 --> 00:11:49.079 around for a very long time. So that works out really well 00:11:49.080 --> 00:11:52.279 for Emacs. We want to make this a generational collector. 00:11:52.280 --> 00:11:56.199 The other thing is with object layout. We use a lot of lists 00:11:56.200 --> 00:12:00.559 inside Emacs Lisp. Every time you go to the cdr, you've 00:12:00.560 --> 00:12:03.039 got to be chasing a pointer around the heap and following 00:12:03.040 --> 00:12:05.439 that. That can potentially result in cache misses and 00:12:05.440 --> 00:12:08.239 all sorts of other things like that. So it can take a long 00:12:08.240 --> 00:12:12.159 time. It can be quite slow. But if you have the ability to move 00:12:12.160 --> 00:12:16.559 objects, you can just relocate an entire list and lay it out 00:12:16.560 --> 00:12:19.168 in an array right next to each other inside memory. 00:12:19.169 --> 00:12:22.479 So iterating over it is just as fast as iterating over an array. 00:12:22.480 --> 00:12:25.421 But you can only do that if you have movable objects. 00:12:25.422 --> 00:12:28.399 I'll point out here too, that with conservative stack scanning, 00:12:28.400 --> 00:12:31.599 it's not that all objects are immovable. It's only ones that 00:12:31.600 --> 00:12:35.519 are pointed to from the stack or from the registers that have 00:12:35.520 --> 00:12:38.828 to become immovable. NOTE How Rust makes precise GC easy 00:12:38.829 --> 00:12:41.039 Let's look at how Rust makes precise 00:12:41.040 --> 00:12:44.439 garbage collection easy. Here I have some Rust code to 00:12:44.440 --> 00:12:47.279 show kind of how the lifetime system works and what we call 00:12:47.280 --> 00:12:49.879 XOR mutability, where we can only have one mutable 00:12:49.880 --> 00:12:52.879 reference or multiple immutable references to the same 00:12:52.880 --> 00:12:56.199 thing. Here we declare a vector, we take a reference to the 00:12:56.200 --> 00:12:59.199 first element of the vector, and then we mutate the vector. 00:12:59.200 --> 00:13:02.239 Now this could potentially resize the vector and move it to a 00:13:02.240 --> 00:13:04.919 different location in memory, so that reference is no 00:13:04.920 --> 00:13:07.759 longer valid. The nice thing is, Rust catches this for 00:13:07.760 --> 00:13:10.479 us. It says, hey, this is no longer valid. This reference 00:13:10.480 --> 00:13:14.519 can't survive past when you mutated it. Okay? That's 00:13:14.520 --> 00:13:17.559 exactly what we want for a garbage collector. You can see 00:13:17.560 --> 00:13:19.879 here, we take this in a garbage collection context, we 00:13:19.880 --> 00:13:23.359 create a new context object, we add an object, we call 00:13:23.360 --> 00:13:26.759 garbage_collect, then we try and access that object. It's no 00:13:26.760 --> 00:13:29.199 longer accessible, and Rust will prevent us from trying to 00:13:29.200 --> 00:13:34.839 access that variable. So, how do we solve this? We have a 00:13:34.840 --> 00:13:39.759 root macro. We declared this root macro, it lets us take the 00:13:39.760 --> 00:13:41.759 object and let it live past garbage collection, and 00:13:41.760 --> 00:13:45.319 everything works out. The nice thing is, this root macro 00:13:45.320 --> 00:13:47.799 will get dropped when it's out of scope, so we don't have to 00:13:47.800 --> 00:13:51.519 worry about the un-gc-protect step of this. Statically, 00:13:51.520 --> 00:13:55.799 Rust will verify and tell us any object that needs to be 00:13:55.800 --> 00:13:58.279 rooted. If we try and access it, it'll tell us it's invalid. 00:13:58.280 --> 00:14:00.999 We have this root macro and then we can access it. So in 00:14:01.000 --> 00:14:03.759 that way, we have safe, precise garbage collection without 00:14:03.760 --> 00:14:07.479 any chance of introducing undefined behavior, which is 00:14:07.480 --> 00:14:09.999 really, really powerful. It's really easy because the 00:14:10.000 --> 00:14:13.226 type system will catch it all for us. NOTE Other Rust niceties: proc macro 00:14:13.227 --> 00:14:15.147 There's some other Rust niceties I want to kind of 00:14:15.148 --> 00:14:16.799 talk through that are useful, but 00:14:16.800 --> 00:14:21.079 are not, you know, star features. One is proc macros. You 00:14:21.080 --> 00:14:23.679 can see up on the top, you can see how you declare a function 00:14:23.680 --> 00:14:27.359 inside the C core. All right. You have to use the macro. You 00:14:27.360 --> 00:14:29.141 have to put the list type, the function type, 00:14:29.142 --> 00:14:30.963 the struct type, the different types of arguments 00:14:30.964 --> 00:14:33.225 or different number of arguments, the doc string, 00:14:33.226 --> 00:14:36.023 and then you can put your argument listing down inside there. 00:14:36.024 --> 00:14:37.984 On the Rust side, we just write this like we would 00:14:37.985 --> 00:14:40.044 any other Rust function. And then we put 00:14:40.045 --> 00:14:41.285 the defun proc macro on there 00:14:41.286 --> 00:14:44.186 and it takes care of everything for us behind the scenes. 00:14:44.187 --> 00:14:46.407 A couple of cool additional things we can do with this 00:14:46.408 --> 00:14:48.727 is that we don't have to make everything just an object. 00:14:48.728 --> 00:14:49.759 We can actually make things 00:14:49.760 --> 00:14:54.239 more specific types. Here we have symbols. As well as 00:14:54.240 --> 00:14:56.279 you can see subfeature, it's an optional parameter, and we 00:14:56.280 --> 00:15:00.919 just make it an option inside Rust and it automatically make 00:15:00.920 --> 00:15:03.599 it an optional inside Elisp. 00:15:03.600 --> 00:15:05.181 This makes them really easy to write. 00:15:05.182 --> 00:15:06.439 I can't take credit for this is because this is 00:15:06.440 --> 00:15:09.119 something that I saw inside Remacs and I stole from them, but 00:15:09.120 --> 00:15:11.439 it makes the functions really easy to call from each other 00:15:11.440 --> 00:15:14.559 and really easy to write as well. NOTE sum types 00:15:14.560 --> 00:15:18.523 Another thing that's really nice is sum types. 00:15:18.524 --> 00:15:21.039 In the C core, if I wanted to get a 00:15:21.040 --> 00:15:23.759 string out of an object, I would first need to check that it's 00:15:23.760 --> 00:15:28.319 a string and then dereference it as a string. But if it's not a 00:15:28.320 --> 00:15:30.679 string, I may introduce undefined behavior. So in 00:15:30.680 --> 00:15:32.799 complicated code, I have to make sure that I have always 00:15:32.800 --> 00:15:34.959 checked what type it is before I try and dereference that 00:15:34.960 --> 00:15:37.879 type. We don't have to worry about any of that inside Rust 00:15:37.880 --> 00:15:41.319 because we can untag a value and we can use their some types, 00:15:41.320 --> 00:15:44.399 basically create an enum and we can match on what the 00:15:44.400 --> 00:15:47.639 different values can be. Then we only get out the types 00:15:47.640 --> 00:15:50.359 that are viable or are actually there. So we never 00:15:50.360 --> 00:15:52.159 accidentally get something out of an object that we didn't 00:15:52.160 --> 00:15:54.239 mean to, or dereference it as something that doesn't 00:15:54.240 --> 00:15:56.879 really exist. We can just match on it and we can get out the 00:15:56.880 --> 00:16:01.040 values that we need, which is really, really powerful. NOTE Regex 00:16:01.041 --> 00:16:03.639 So there's some other Rust niceties as well working with here. 00:16:03.640 --> 00:16:07.799 One is the regex engine inside Rust is really fast, high 00:16:07.800 --> 00:16:10.959 performance. We are using that for the Elixir regex 00:16:10.960 --> 00:16:14.879 engine to give it high performance and worst-case 00:16:14.880 --> 00:16:16.051 guarantees. NOTE Parsers 00:16:16.052 --> 00:16:18.599 The other is that Rust has a lot of really good 00:16:18.600 --> 00:16:21.559 parsers for things like JSON that are no copy parsers that 00:16:21.560 --> 00:16:24.719 are high performance. We can use those inside Rune as 00:16:24.720 --> 00:16:27.209 well. NOTE Other changes: GUI first, terminal second 00:16:27.210 --> 00:16:29.439 There's a handful of other changes we're working on 00:16:29.440 --> 00:16:33.119 that are not Rust-specific, but we'd like to see. The first is 00:16:33.120 --> 00:16:36.759 being GUI first, terminal second. This means two things. 00:16:36.760 --> 00:16:40.039 First is that we have all of our key bindings. Right now 00:16:40.040 --> 00:16:43.279 inside Emacs, C-i and TAB are bound to the same key 00:16:43.280 --> 00:16:45.039 binding by default, because that's how it works inside the 00:16:45.040 --> 00:16:48.119 terminal. In the GUI, you shouldn't have that limitation. 00:16:48.120 --> 00:16:52.559 The second is that the GUI should not block when Lisp is 00:16:52.560 --> 00:16:55.199 blocked. It should be independent of that. Your GUI can 00:16:55.200 --> 00:16:58.918 still continue to operate when Lisp is running. NOTE Off-screen cursor 00:16:58.919 --> 00:17:01.279 The other is the ability to have an off-screen cursor 00:17:01.280 --> 00:17:02.699 so that you can be typing on something, 00:17:02.700 --> 00:17:04.319 you can scroll up and down and the point 00:17:04.320 --> 00:17:06.719 doesn't have to follow you where you lose your place where 00:17:06.720 --> 00:17:09.399 you were before. You don't have to intentionally set a mark. 00:17:09.400 --> 00:17:11.199 You can just scroll and then start typing and it'll go back up 00:17:11.200 --> 00:17:13.879 to where it was before, like it works in most applications. 00:17:13.880 --> 00:17:16.304 And this can be optional. NOTE Image flow 00:17:16.305 --> 00:17:18.079 The other is image flow. We want it 00:17:18.080 --> 00:17:20.879 so that you can easily flow images and you can have large 00:17:20.880 --> 00:17:23.159 images and scroll past them without jumping and you can flow 00:17:23.160 --> 00:17:24.439 text around images. NOTE Testing 00:17:24.440 --> 00:17:29.799 How are we testing this project? Because there's a lot of 00:17:29.800 --> 00:17:33.159 things that you could get wrong here. One thing we're doing 00:17:33.160 --> 00:17:38.039 is we're using ERT. Emacs ships with over 7,000 built-in 00:17:38.040 --> 00:17:42.879 tests--Elisp tests. We are using this test suite to test 00:17:42.880 --> 00:17:45.079 our project as well. We can kind of use this as a dashboard 00:17:45.080 --> 00:17:47.679 of saying how close are we to getting to parity with GNU 00:17:47.680 --> 00:17:52.319 Emacs. The other thing that we have is a tool called elprop, 00:17:52.320 --> 00:17:55.279 which is an external utility that basically tests for 00:17:55.280 --> 00:17:58.719 correctness. Because really, the correctness of Rune is 00:17:58.720 --> 00:18:00.999 whatever Emacs does, because there's no official spec on 00:18:01.000 --> 00:18:04.079 how things should behave. To do this, we can go look at 00:18:04.080 --> 00:18:07.159 the Rust function signature. We know what the arguments 00:18:07.160 --> 00:18:09.319 are, we know how many they are, and we know what types they 00:18:09.320 --> 00:18:11.679 should be. Given that information, we can generate a 00:18:11.680 --> 00:18:15.279 whole bunch of random functions feeding those types in. And 00:18:15.280 --> 00:18:18.959 then we send a copy over to Emacs, we send a copy over to Rune. 00:18:18.960 --> 00:18:21.679 They each evaluate it and they return the result and we make 00:18:21.680 --> 00:18:23.519 sure the results are the same. Then you do that for 00:18:23.520 --> 00:18:26.199 thousands of different implementations of the function. 00:18:26.200 --> 00:18:29.039 And it helps us find corner cases really easy without having 00:18:29.040 --> 00:18:31.639 to handwrite a whole bunch of different cases to test things 00:18:31.640 --> 00:18:36.344 and say, where are these two functions different? NOTE Status 00:18:36.345 --> 00:18:39.359 So the current status: we already have a multi-threaded Elixir 00:18:39.360 --> 00:18:42.999 interpreter and bytecode engine inside there. There's no 00:18:43.000 --> 00:18:45.679 actual text editor in there yet, but the primitives are 00:18:45.680 --> 00:18:48.679 there. Like you can insert text, move point around, 00:18:48.680 --> 00:18:52.039 delete text, do different things like that. But we don't 00:18:52.040 --> 00:18:53.679 have a GUI hooked up to different key bindings to actually 00:18:53.680 --> 00:18:58.159 type on. There's just a REPL to operate in. We have about 00:18:58.160 --> 00:19:01.279 250 of the 1500 built-in functions already implemented 00:19:01.280 --> 00:19:04.119 inside there. There's a lot of low-hanging fruit inside this 00:19:04.120 --> 00:19:07.246 area to still be implemented. NOTE Next directions 00:19:07.247 --> 00:19:07.719 The next directions we're 00:19:07.720 --> 00:19:11.959 working on is we're optimizing the GC. We want to make it 00:19:11.960 --> 00:19:13.839 generational. Like I said, right now, it's just a simple 00:19:13.840 --> 00:19:17.359 semi-spaced copying GC. We want to add a proper GUI. We need 00:19:17.360 --> 00:19:19.599 to implement text properties, overlays, process and job 00:19:19.600 --> 00:19:22.738 control, all that goodness right there. NOTE How to get involved 00:19:22.739 --> 00:19:25.378 How can you get involved? This is hosted on GitHub. 00:19:25.379 --> 00:19:26.424 You can come on over. 00:19:26.425 --> 00:19:28.639 If you have any ideas about how to implement something or 00:19:28.640 --> 00:19:30.639 something you'd like to see done, go ahead and just open an 00:19:30.640 --> 00:19:32.799 issue so we can have a discussion about it. We've had lots of 00:19:32.800 --> 00:19:34.599 interesting discussions with different people coming in 00:19:34.600 --> 00:19:37.639 to the GitHub repo. If you're interested in contributing, 00:19:37.640 --> 00:19:40.439 the easiest way is probably to run elprop, pick some 00:19:40.440 --> 00:19:43.279 function, run elprop on it. I promise it won't take long to 00:19:43.280 --> 00:19:45.639 find some issues, some discrepancy between Emacs and Rune, 00:19:45.640 --> 00:19:48.959 and that lets you dive into the Rust code and figure out, and 00:19:48.960 --> 00:19:50.879 the C code, and figure out what the difference is between the 00:19:50.880 --> 00:19:53.119 two. or come along and help implement your favorite 00:19:53.120 --> 00:19:55.679 functionality. This has been a really interesting project 00:19:55.680 --> 00:19:58.359 so far, and we've had a handful of different contributors on 00:19:58.360 --> 00:20:01.799 it who just kind of want to learn Rust or get more into 00:20:01.800 --> 00:20:06.000 systems-level programming. Thank you.