WEBVTT 00:00:00.001 --> 00:00:02.880 Google runs millions of lines of Python code. 00:00:02.880 --> 00:00:07.440 The front-end servers that drive YouTube.com and YouTube's API 00:00:07.440 --> 00:00:09.940 are primarily written in Python, 00:00:09.940 --> 00:00:13.920 and they serve millions of requests per second. 00:00:13.920 --> 00:00:16.760 On this episode, you'll meet Dylan Trotter, 00:00:16.760 --> 00:00:19.420 who is working to increase the performance and concurrency 00:00:19.420 --> 00:00:20.900 of these servers powering YouTube. 00:00:20.900 --> 00:00:25.780 He just launched Grumpy, a Python implementation based on Go, 00:00:25.780 --> 00:00:27.880 the highly concurrent language from Google. 00:00:28.480 --> 00:00:33.180 This is Talk Python To Me, recorded January 12, 2017. 00:00:56.960 --> 00:01:01.240 Welcome to Talk Python To Me, a weekly podcast on Python, 00:01:01.240 --> 00:01:04.320 the language, the libraries, the ecosystem, and the personalities. 00:01:04.320 --> 00:01:06.420 This is your host, Michael Kennedy. 00:01:06.420 --> 00:01:08.420 Follow me on Twitter, where I'm @mkennedy. 00:01:08.420 --> 00:01:12.300 Keep up with the show and listen to past episodes at talkpython.fm, 00:01:12.300 --> 00:01:14.880 and follow the show on Twitter via at Talk Python. 00:01:15.660 --> 00:01:19.420 This episode has been sponsored by Hired and, as a new sponsor, 00:01:19.420 --> 00:01:21.580 pyimagesearch.com. 00:01:21.580 --> 00:01:26.320 They're announcing a Kickstarter campaign called Deep Learning for Computer Vision with Python, 00:01:26.320 --> 00:01:28.640 launching on Kickstarter right now. 00:01:28.640 --> 00:01:32.680 Thank both of these companies for supporting the show by checking out what they have to offer 00:01:32.680 --> 00:01:33.560 during their segments. 00:01:33.560 --> 00:01:35.800 Dylan, welcome to Talk Python. 00:01:35.800 --> 00:01:36.340 Thanks. 00:01:36.340 --> 00:01:36.920 Nice to be here. 00:01:37.060 --> 00:01:41.000 Yeah, I'm really excited to talk about Grumpy, actually. 00:01:41.000 --> 00:01:42.920 Grumpy, your Python project. 00:01:42.920 --> 00:01:43.580 It's going to be fun. 00:01:43.580 --> 00:01:47.080 Yeah, it's been pretty exciting a couple weeks since the release. 00:01:47.080 --> 00:01:48.500 So, yeah, I'm excited to talk about it, too. 00:01:48.500 --> 00:01:52.080 Yeah, it's definitely gotten a lot of attention in the open source world on GitHub. 00:01:52.600 --> 00:01:55.500 And we're going to dig into a lot of the details behind it. 00:01:55.500 --> 00:01:56.660 But let's start with you and your story. 00:01:56.660 --> 00:01:58.140 How did you get into programming in Python? 00:01:58.140 --> 00:02:01.780 I started programming, I guess, when I was in high school. 00:02:01.780 --> 00:02:07.020 I took, like, an intro programming course and kind of got the bug. 00:02:07.020 --> 00:02:10.060 And I just kind of took it from there. 00:02:10.060 --> 00:02:13.660 I was really into, like, programming little games and stuff like that back then. 00:02:13.660 --> 00:02:17.880 I did not do CS in university. 00:02:17.880 --> 00:02:23.280 I actually did physics, but I continued to work on programming in my own time a lot. 00:02:23.280 --> 00:02:30.780 And after or during university, actually, I got a gig at a sort of summer gig at a software company. 00:02:30.780 --> 00:02:35.680 And that gave me a leg up when I graduated, which was pretty lucky. 00:02:35.680 --> 00:02:41.680 And so I got a job at a visual effects company doing software there. 00:02:42.620 --> 00:02:48.800 So there was a bunch of different things, like a lot of sort of proprietary languages for the different packages. 00:02:48.800 --> 00:02:56.120 But Python sort of came out as a front runner in terms of integration with different visual effects packages and stuff like that. 00:02:56.120 --> 00:03:03.460 And so that's where I started to dig into Python, especially not so much on the sort of, like, effect side, 00:03:03.560 --> 00:03:07.980 but more on the pipeline data management kind of side of things. 00:03:07.980 --> 00:03:11.980 So there's a lot of asset management and stuff going on in visual effects studios. 00:03:11.980 --> 00:03:14.440 And Python's great for that sort of stuff. 00:03:14.440 --> 00:03:15.580 Yeah, that's really cool. 00:03:15.580 --> 00:03:22.300 I did a whole episode on Python and, like, game development studios and movies and production and stuff. 00:03:22.300 --> 00:03:27.080 I was really surprised how much Python glues all the tooling together for those folks. 00:03:27.440 --> 00:03:28.960 Yeah, it's really deep in there. 00:03:28.960 --> 00:03:35.980 In fact, when I was working in that area, that's when Python sort of started to come to the fore. 00:03:35.980 --> 00:03:44.540 And so, like, Maya, which is a big, like, modeling and animation package, built Python integration around that time. 00:03:44.540 --> 00:03:48.940 And Houdini is another one, similar use cases that integrate. 00:03:48.940 --> 00:03:53.580 Or actually, I think from pretty early on, Houdini had Python integration. 00:03:53.580 --> 00:03:58.500 So, yeah, it sort of became the de facto visual effects integration language. 00:03:58.500 --> 00:03:58.960 Okay. 00:03:58.960 --> 00:04:00.180 Yeah, yeah, very cool. 00:04:00.180 --> 00:04:01.920 And I think that's only growing. 00:04:01.920 --> 00:04:06.980 It seems like there's a couple areas where Python is sort of past critical mass. 00:04:06.980 --> 00:04:09.140 It's kind of like a black hole now. 00:04:09.140 --> 00:04:10.480 It's just sucking everything into it. 00:04:10.480 --> 00:04:10.840 Totally. 00:04:10.840 --> 00:04:11.340 Yeah, yeah. 00:04:11.340 --> 00:04:12.000 And that's a good thing. 00:04:12.000 --> 00:04:19.120 So you don't do visual effects anymore, although you kind of work in the video world these days. 00:04:19.120 --> 00:04:20.480 Why don't you tell everybody what you're up to? 00:04:20.480 --> 00:04:21.540 Yeah, sure. 00:04:21.540 --> 00:04:23.380 So I'm at YouTube now. 00:04:23.380 --> 00:04:25.700 I started there about seven years ago. 00:04:25.700 --> 00:04:28.040 It was kind of a big shift for me. 00:04:28.040 --> 00:04:35.000 Visual effects was a fun environment, but it was always kind of a dream to work at Google and stuff. 00:04:35.000 --> 00:04:36.800 So I took a gig at YouTube. 00:04:36.800 --> 00:04:40.540 And I've worked on a number of different teams there, actually. 00:04:40.540 --> 00:04:43.540 I've worked on sort of user-facing features. 00:04:43.540 --> 00:04:50.280 I was on the channels team for a long time, working on YouTube channels and stuff around that. 00:04:50.280 --> 00:04:54.540 And eventually got more into the infrastructure side. 00:04:54.540 --> 00:05:00.740 And so now I'm working on what's called, I guess, the application infrastructure group. 00:05:01.100 --> 00:05:10.020 And our team specifically looks after the application server that serves YouTube.com and YouTube APIs and those sorts of things. 00:05:10.020 --> 00:05:10.380 Excellent. 00:05:10.380 --> 00:05:18.980 So when we're all watching various things on YouTube, be it cat videos or something educational, we have you to thank for keeping those servers running. 00:05:19.660 --> 00:05:22.340 Yeah, well, me and a lot of other people. 00:05:22.340 --> 00:05:22.600 Yep. 00:05:22.600 --> 00:05:24.460 Yeah, yeah, I'm sure. 00:05:24.460 --> 00:05:26.800 Well, I'll thank you individually. 00:05:26.800 --> 00:05:27.720 So, yeah, awesome. 00:05:27.720 --> 00:05:30.240 Yeah, it sounds like a really fun place to work. 00:05:30.240 --> 00:05:31.320 Where's YouTube? 00:05:31.320 --> 00:05:33.400 Where's the center of the universe for YouTube? 00:05:33.400 --> 00:05:34.720 Is that Mountain View or somewhere else? 00:05:34.720 --> 00:05:38.580 San Bruno actually is where the main YouTube campus is. 00:05:38.580 --> 00:05:42.840 So there's a few different offices around the world. 00:05:42.840 --> 00:05:49.480 But the biggest group of the biggest sort of geographical concentration is in San Bruno. 00:05:49.480 --> 00:05:52.380 So there's a few buildings there that are YouTube. 00:05:52.380 --> 00:05:52.920 Okay. 00:05:52.920 --> 00:05:53.460 Yeah. 00:05:53.460 --> 00:05:53.920 Nice. 00:05:53.920 --> 00:05:55.000 That sounds like so much fun. 00:05:55.000 --> 00:05:59.060 So Python actually plays a really important role at YouTube these days. 00:05:59.060 --> 00:06:02.560 Let's talk about how it's used now and then how that kind of came to be. 00:06:02.560 --> 00:06:02.900 Sure. 00:06:02.900 --> 00:06:03.140 Yeah. 00:06:03.140 --> 00:06:10.960 So Python is what is running the main application server and a lot of the application code for 00:06:10.960 --> 00:06:18.940 the YouTube front end and for that serves like the website and the API service APIs that, 00:06:18.940 --> 00:06:21.540 you know, service your phone and those sorts of things. 00:06:21.540 --> 00:06:26.800 So it's sort of like the gateway for most user traffic. 00:06:26.800 --> 00:06:27.160 Right. 00:06:27.160 --> 00:06:31.540 And then maybe the Python code branches back into all sorts of Google services behind the 00:06:31.540 --> 00:06:34.220 scenes that are in a variety of technologies or something like that. 00:06:34.220 --> 00:06:34.440 Right. 00:06:34.440 --> 00:06:34.740 Yeah. 00:06:34.740 --> 00:06:38.960 There's a, there's a lot of different technologies and servers involved in the whole thing. 00:06:38.960 --> 00:06:39.120 Yep. 00:06:39.120 --> 00:06:39.400 Okay. 00:06:39.400 --> 00:06:43.000 So YouTube wasn't initially a Google creation, right? 00:06:43.000 --> 00:06:44.540 It was created by some other folks. 00:06:44.540 --> 00:06:50.360 It was founded in, in 2005, I think by three guys. 00:06:50.780 --> 00:06:55.060 One of the one of the one of the one of the one of the ones that I joined in 2009 is Chad Hurley. 00:06:55.060 --> 00:06:59.000 I think at the time he was the president or something. 00:06:59.000 --> 00:07:01.820 He left shortly after I joined. 00:07:01.820 --> 00:07:07.280 But yeah, they, they built it in 2005 and it gained a lot of traction really early on. 00:07:08.080 --> 00:07:16.880 And I guess Google took an interest at some point in 2006 and, and ended up buying YouTube in 00:07:16.880 --> 00:07:17.940 November, 2006. 00:07:17.940 --> 00:07:18.380 Yeah. 00:07:18.380 --> 00:07:22.660 I'd say that was a great move for them because it's, it's such a central part of the internet. 00:07:22.660 --> 00:07:23.320 Yeah. 00:07:23.320 --> 00:07:25.960 I feel like it, it had YouTube. 00:07:25.960 --> 00:07:30.100 The idea was something that a lot of people probably had the idea. 00:07:30.100 --> 00:07:32.620 It was a thing that clearly should exist. 00:07:33.080 --> 00:07:39.900 But when you think of the infrastructure and the bandwidth costs and just the actual act 00:07:39.900 --> 00:07:44.660 of creating such, such a huge video network seems prohibitive. 00:07:44.660 --> 00:07:47.940 But, you know, once it came into existence, you know, I guess Google jumped on it. 00:07:47.940 --> 00:07:48.240 That's cool. 00:07:48.240 --> 00:07:54.760 I actually remember thinking what a simple idea it was and, and how like, it seemed so crazy 00:07:54.760 --> 00:07:58.520 at the time that I think the acquisition cost was like 1.6 billion or something like that. 00:07:58.520 --> 00:08:02.940 And, and I remember reading about that and I was like, good Lord, like, you know, you 00:08:02.940 --> 00:08:06.820 could, how, how is something so simple worth so much? 00:08:06.820 --> 00:08:11.380 But now that number seems so quaint compared to recent, recent stuff. 00:08:11.380 --> 00:08:13.420 So, yeah, yeah, of course, of course. 00:08:13.420 --> 00:08:18.560 I mean, a lot of companies go through the thinking, it's easy to go through the thinking of you 00:08:18.560 --> 00:08:19.880 could have just built that yourself. 00:08:19.880 --> 00:08:20.400 Yeah. 00:08:20.400 --> 00:08:20.640 Right. 00:08:20.640 --> 00:08:24.860 I mean, Facebook bought Instagram for like an insane amount of money and that's like a team 00:08:24.860 --> 00:08:25.900 of 12 people, right? 00:08:25.900 --> 00:08:28.160 For whatever it was like 19 billion or something. 00:08:28.160 --> 00:08:32.800 They, they could have easily paid 12 people to build another Instagram, but it's, it's also 00:08:32.800 --> 00:08:35.560 also got people's interests. 00:08:35.560 --> 00:08:36.780 It's got the users. 00:08:36.780 --> 00:08:37.960 It's got the momentum. 00:08:37.960 --> 00:08:40.300 And that, that's the thing I think people buy. 00:08:40.300 --> 00:08:40.960 Absolutely. 00:08:40.960 --> 00:08:41.940 That's what you're paying for. 00:08:41.940 --> 00:08:42.220 Yeah. 00:08:42.220 --> 00:08:44.100 But they didn't write it in Python at first, did they? 00:08:44.100 --> 00:08:47.500 No, they, the first implementation I believe was PHP. 00:08:47.500 --> 00:08:49.560 And I don't think that lasted very long. 00:08:49.560 --> 00:08:53.600 I think it was, most of that was rewritten in Python pretty early on. 00:08:53.600 --> 00:08:54.600 Well, before I was there. 00:08:54.600 --> 00:08:54.900 Sure. 00:08:54.900 --> 00:08:55.180 Sure. 00:08:56.020 --> 00:09:01.540 I suspect the way YouTube looks today with the growth of cloud computing and all the 00:09:01.540 --> 00:09:06.600 different APIs and services is probably super different from when, what you guys got back 00:09:06.600 --> 00:09:07.700 in 2006, right? 00:09:07.700 --> 00:09:08.440 Yeah. 00:09:08.440 --> 00:09:11.680 It's, I mean, I think, you know, the company has grown a lot. 00:09:11.780 --> 00:09:14.000 The, the use cases have grown a lot. 00:09:14.000 --> 00:09:16.320 It's just, I mean, it's kind of night and day. 00:09:16.320 --> 00:09:21.540 It's when I, when I first joined, you know, everyone was kind of in one floor of one building. 00:09:21.540 --> 00:09:24.640 And since then, you know, there's distributed all over the world. 00:09:24.640 --> 00:09:27.280 And so, yeah, it's, it's, it's changed a lot. 00:09:27.280 --> 00:09:27.540 Sure. 00:09:27.540 --> 00:09:28.440 Wow. 00:09:29.160 --> 00:09:29.560 Okay. 00:09:29.560 --> 00:09:37.240 So that brings us to today, to YouTube and what this project that you, was this something 00:09:37.240 --> 00:09:40.920 that you created this project called grumpy or where did this come from? 00:09:40.920 --> 00:09:41.340 Yeah. 00:09:41.340 --> 00:09:46.080 I talked about some of the challenges we were having with, you know, running Python at scale 00:09:46.080 --> 00:09:48.260 and on, on the blog post. 00:09:48.260 --> 00:09:56.720 And it basically, there's a few different aspects that, that affects our ability to run, you know, 00:09:56.720 --> 00:09:58.200 that many Python servers. 00:09:58.420 --> 00:10:05.880 The CPython runtime, well, it's, it's really great and it's highly optimized and, and it 00:10:05.880 --> 00:10:09.940 does a lot of things really well for our use case. 00:10:09.940 --> 00:10:14.460 It's never really been a focus for CPython as a project. 00:10:14.460 --> 00:10:22.120 You know, we thought, you know, maybe it makes sense to rethink how the runtime is built with 00:10:22.120 --> 00:10:27.040 a focus on concurrency and, and running large server applications. 00:10:27.040 --> 00:10:27.540 Yeah. 00:10:27.620 --> 00:10:31.820 And you're not, you guys are not the first people to have this idea of, well, maybe we 00:10:31.820 --> 00:10:37.440 could replace the CPython runtime interpreter with something else. 00:10:37.440 --> 00:10:44.000 There's like Jython, there's IronPython, there's PyPy, there's plugin JIT. 00:10:44.000 --> 00:10:48.860 So there's a lot of stuff happening there, but nobody's gone in the direction that you went 00:10:48.860 --> 00:10:49.320 in, right? 00:10:49.500 --> 00:10:49.800 Yeah. 00:10:49.800 --> 00:10:49.840 Yeah. 00:10:49.840 --> 00:10:53.520 It was, it was, it was an interesting, I mean, it's, you know, in a lot of ways it's kind 00:10:53.520 --> 00:10:53.940 of crazy. 00:10:53.940 --> 00:11:01.360 And the thing about Go, the Go runtime that Grumpy is based on is that it is kind of designed 00:11:01.360 --> 00:11:06.020 for very similar use cases to what we are interested in. 00:11:06.360 --> 00:11:14.480 So Go tends to be, tends to be used for writing highly concurrent server applications with, 00:11:14.480 --> 00:11:20.320 you know, a lot of like sort of message passing and things within the, within the application 00:11:20.320 --> 00:11:21.480 between threads. 00:11:21.480 --> 00:11:25.020 It seemed like kind of a good fit. 00:11:25.020 --> 00:11:30.800 And once I started to flesh things out and to build out some of the core functionality, 00:11:30.800 --> 00:11:35.700 some of the pieces started to fall into place and it started to look actually really compelling. 00:11:35.700 --> 00:11:37.780 And you're like, Hey, we could actually do this. 00:11:37.780 --> 00:11:39.100 We should stop for just a second. 00:11:39.100 --> 00:11:45.980 I don't think we've explicitly said your project is called Grumpy, which is a replacement for 00:11:45.980 --> 00:11:51.920 the CPython implementation with a entirely different Go implementation, right? 00:11:51.920 --> 00:11:52.680 Yeah, that's right. 00:11:52.680 --> 00:11:52.920 Yeah. 00:11:52.920 --> 00:11:53.200 Yeah. 00:11:53.200 --> 00:11:54.380 So, so very interesting. 00:11:54.380 --> 00:12:00.420 I think, you know, Go, obviously it makes sense for Google to be the ones experimenting 00:12:00.420 --> 00:12:01.240 with Go, right? 00:12:01.240 --> 00:12:03.120 Go comes from Google, doesn't it? 00:12:03.120 --> 00:12:03.540 It does. 00:12:03.540 --> 00:12:03.780 Yep. 00:12:03.780 --> 00:12:11.660 It was developed, I think, originally by, well, Rob Pike and I'm going to mix it up. 00:12:11.660 --> 00:12:15.740 It's either Ken Thompson or no, it's, yeah, it's Ken Thompson, I believe. 00:12:15.740 --> 00:12:16.660 Yeah. 00:12:16.660 --> 00:12:23.860 It was, it was developed for, I guess they had observed that I get similar to, you know, the, 00:12:23.860 --> 00:12:29.880 what the observations that we made about running Python programs for Python server programs. 00:12:29.920 --> 00:12:36.060 They had made sort of more general observations about writing server applications and how languages 00:12:36.060 --> 00:12:40.200 that existed didn't, didn't quite fit what our use cases. 00:12:40.200 --> 00:12:40.480 Yeah. 00:12:40.480 --> 00:12:45.400 Go is really quite, it's one of the newest languages out there that I would consider a mainstream 00:12:45.400 --> 00:12:46.040 language. 00:12:46.580 --> 00:12:51.780 It's not as mainstream as C, but it's definitely getting there and came out in 2012 in version 00:12:51.780 --> 00:12:52.840 one sort of officially. 00:12:52.840 --> 00:13:01.880 So it's born within this world of multi-core microservices, distributed cloud computing stuff, 00:13:01.880 --> 00:13:02.160 right? 00:13:02.160 --> 00:13:03.140 Yeah. 00:13:03.140 --> 00:13:03.600 Yeah. 00:13:04.140 --> 00:13:04.540 Okay. 00:13:04.540 --> 00:13:08.260 So let's dig into the, what is grumpy? 00:13:08.260 --> 00:13:09.200 Let's dig in a little bit. 00:13:09.200 --> 00:13:13.220 Like how do I take, so I can take my Python code. 00:13:13.220 --> 00:13:20.240 I can write some, presumably some web app or something in a web service, and then I can run 00:13:20.240 --> 00:13:20.920 that on grumpy. 00:13:20.920 --> 00:13:22.580 Like what, what does grumpy do? 00:13:22.580 --> 00:13:24.760 How does it take my Python code and run it? 00:13:24.760 --> 00:13:29.200 So grumpy is, takes a little bit of a different tack than CPython. 00:13:29.200 --> 00:13:36.420 It's actually a trans compiler and a runtime, whereas you can kind of think of CPython as 00:13:36.420 --> 00:13:40.560 a, it's like a virtual machine bytecode interpreter and runtime. 00:13:40.560 --> 00:13:48.360 And in that sense, it's kind of like a combination of Cython and, you know, a bundling of Cython and 00:13:48.360 --> 00:13:50.600 CPython, except that it's all in Go. 00:13:50.600 --> 00:13:50.940 Right. 00:13:51.240 --> 00:13:58.320 So Cython takes a flavor of Python and then compiles it to C directly. 00:13:58.320 --> 00:14:04.500 And like C, Go is a statically typed compiled language. 00:14:04.500 --> 00:14:07.800 And so it's no longer interpreted. 00:14:07.800 --> 00:14:11.300 It's not even like JIT compiled like Java or .NET. 00:14:11.300 --> 00:14:12.980 It's full on compiled, right? 00:14:12.980 --> 00:14:13.760 That's correct. 00:14:13.760 --> 00:14:14.140 Okay. 00:14:14.140 --> 00:14:18.880 So the sort of runtime side of things is actually like the correspondence is like this Python 00:14:18.880 --> 00:14:19.700 C API. 00:14:20.020 --> 00:14:22.260 There's actually a Go grumpy API. 00:14:22.260 --> 00:14:31.580 And so what it's compiling is code that uses that API to mutate objects, to pull out a state 00:14:31.580 --> 00:14:32.860 and those sorts of things. 00:14:32.860 --> 00:14:40.960 And so whereas CPython or vanilla CPython uses a bytecode interpreter to actually drive those 00:14:40.960 --> 00:14:49.060 API calls, the Grumpy and Cython are actually generating code that drives those API calls. 00:14:49.060 --> 00:14:49.380 Okay. 00:14:49.380 --> 00:14:50.240 Yeah. 00:14:50.240 --> 00:14:51.040 Very, very cool. 00:14:51.040 --> 00:14:56.580 Now in your GitHub repo or the blog post, I don't remember where I got this, but you said 00:14:56.580 --> 00:15:00.900 it's intended to be a near drop-in replacement for CPython 2.7. 00:15:00.900 --> 00:15:02.360 How's that going? 00:15:02.360 --> 00:15:04.200 How far are you towards that goal? 00:15:04.200 --> 00:15:07.080 That's a pretty big set of APIs to cover. 00:15:07.300 --> 00:15:07.480 Yeah. 00:15:07.480 --> 00:15:10.960 I'm learning every day like how big Python is. 00:15:10.960 --> 00:15:14.660 Nobody told me about this weird case I'm going to have to support. 00:15:14.660 --> 00:15:15.720 Oh yeah, totally. 00:15:15.720 --> 00:15:16.560 Yeah. 00:15:16.560 --> 00:15:22.860 I mean, I've been the amount of sort of spelunking I've done in CPython internals is I did not 00:15:22.860 --> 00:15:23.800 expect all that. 00:15:24.360 --> 00:15:27.500 But yeah, so it's going pretty well. 00:15:27.500 --> 00:15:29.600 The core functionality is there. 00:15:29.600 --> 00:15:35.040 So like the basic semantics of the language in terms of attribute access and how types work 00:15:35.040 --> 00:15:42.160 and how method dispatch works, all of that functions basically fine. 00:15:42.160 --> 00:15:43.540 The basic types are all there. 00:15:43.700 --> 00:15:47.980 So lists and dictionaries and things all kind of work. 00:15:47.980 --> 00:15:53.080 Do those mostly map directly to the underlying Go structures? 00:15:53.080 --> 00:15:58.140 Like does a list in Python map to a slice in Go and things like that? 00:15:58.140 --> 00:16:00.820 Or do you have to do more complicated things to map it? 00:16:00.820 --> 00:16:01.840 It's more complicated. 00:16:01.840 --> 00:16:06.120 And the reason is that Python is so dynamic, right? 00:16:06.140 --> 00:16:10.280 Like method dispatch is so dynamic and attribute access. 00:16:10.280 --> 00:16:12.800 You can put attributes on just about anything. 00:16:12.800 --> 00:16:20.280 You know, if it was just this native Go types, then you wouldn't be able to put an attribute 00:16:20.280 --> 00:16:23.200 on a list or on a slice, right? 00:16:23.200 --> 00:16:23.700 Right. 00:16:23.700 --> 00:16:29.980 So it's actually, there's sort of wrapper types, basically structures that actually map very closely 00:16:29.980 --> 00:16:32.700 to CPython's object structures. 00:16:32.700 --> 00:16:33.700 Okay. 00:16:33.900 --> 00:16:38.740 Yeah, I can see that because you're working with a non-dynamic language and yet it has 00:16:38.740 --> 00:16:40.880 to support dynamic capabilities. 00:16:40.880 --> 00:16:43.980 So you got to somehow put a shim in there for that, right? 00:16:43.980 --> 00:16:44.680 That's right. 00:16:44.680 --> 00:16:45.020 Okay. 00:16:45.020 --> 00:16:50.700 I guess the biggest kind of gaps in terms of supporting or being a drop-in replacement are 00:16:50.700 --> 00:16:53.320 the standard library still needs a lot of work. 00:16:53.320 --> 00:17:00.880 So CPython has a lot of its standard library is actually written as C extension modules, which 00:17:00.880 --> 00:17:02.580 Grumpy does not support. 00:17:02.680 --> 00:17:06.720 So that's one area of significant divergence between the two words. 00:17:06.720 --> 00:17:08.440 And we could talk about that more. 00:17:08.440 --> 00:17:13.940 That's turned out to be sort of a big kind of beast to slay. 00:17:13.940 --> 00:17:20.240 The nice thing is that with, you know, all those other Python runtimes out there, there's 00:17:20.240 --> 00:17:24.700 actually, you know, you can find pure Python versions of most things. 00:17:24.860 --> 00:17:30.540 So like PyPy, for example, implements a number of libraries that are in Python that aren't implemented 00:17:30.540 --> 00:17:31.740 in CPython. 00:17:31.740 --> 00:17:32.100 Right. 00:17:32.100 --> 00:17:39.140 So you could say, start this transition or this backfilling of APIs by just moving to 00:17:39.140 --> 00:17:45.340 pure Python implementations that then get sent through Grumpy that actually get compiled 00:17:45.340 --> 00:17:46.600 or run on Go, right? 00:17:46.960 --> 00:17:47.160 Yep. 00:17:47.160 --> 00:17:48.020 That's exactly right. 00:17:48.020 --> 00:17:51.200 And maybe do some profiling and say, well, you know, people use lists a lot. 00:17:51.200 --> 00:17:55.120 Let's write that directly in Go or something like this, right? 00:17:55.120 --> 00:17:56.160 You can optimize later. 00:17:56.160 --> 00:17:56.700 Exactly. 00:17:56.700 --> 00:17:57.140 Yep. 00:17:57.180 --> 00:17:57.360 Okay. 00:17:57.360 --> 00:17:57.620 Yeah. 00:17:57.620 --> 00:18:00.440 I suspect that there's a long tail of like stuff. 00:18:00.440 --> 00:18:03.380 This doesn't really need to be optimized that last 5%. 00:18:03.380 --> 00:18:06.300 Whereas these are the few things that we really should focus on, right? 00:18:06.300 --> 00:18:06.580 Yeah. 00:18:06.580 --> 00:18:13.100 So right now, you know, I'm kind of focused on getting support for the whole, like I want 00:18:13.100 --> 00:18:17.860 to be able to run some common libraries that are written in Python. 00:18:17.860 --> 00:18:22.860 Some, I want some program, Python programs that are out there, like open source programs to 00:18:22.860 --> 00:18:24.740 be able to just use Grumpy. 00:18:24.860 --> 00:18:30.120 So like just getting it to the point where everything runs is the first step and then 00:18:30.120 --> 00:18:30.880 you make it fast. 00:18:30.880 --> 00:18:31.280 Okay. 00:18:31.280 --> 00:18:31.960 Yeah, of course. 00:18:31.960 --> 00:18:35.680 Making it work and then making it fast seems like the right order to me as well. 00:18:35.680 --> 00:18:40.520 So you said in your blog post that there's going to be some things that Grumpy will never 00:18:40.520 --> 00:18:44.820 support and then there's things that it doesn't support yet, but you're working towards. 00:18:44.820 --> 00:18:45.300 Yeah. 00:18:45.300 --> 00:18:50.800 So one of the things I mentioned already is the C extension support. 00:18:51.680 --> 00:18:58.700 The API for CPython is a bit different than the API for Grumpy because it's, well, for one 00:18:58.700 --> 00:19:02.960 thing it's a different language, but also the data structures are a little bit different. 00:19:02.960 --> 00:19:05.700 The function return values and things are a little bit different. 00:19:05.700 --> 00:19:14.680 And so there wasn't a good mapping between those APIs and it would be too constraining for, 00:19:14.940 --> 00:19:20.200 you know, to try to make Grumpy map perfectly to the C API. 00:19:20.200 --> 00:19:20.760 Sure. 00:19:20.760 --> 00:19:25.020 Have you looked at the CFFI stuff that PyPy was using? 00:19:25.020 --> 00:19:25.520 Right. 00:19:25.520 --> 00:19:29.480 So that's, I have not looked very closely at that. 00:19:29.480 --> 00:19:33.660 That is something that we've looked at internally for other reasons as well. 00:19:33.920 --> 00:19:38.620 But that is an interesting way to approach the problem. 00:19:38.620 --> 00:19:45.540 And, and potentially, you know, there are ways to bridge the two APIs that C and CFFI may 00:19:45.540 --> 00:19:46.200 be one of those. 00:19:46.200 --> 00:19:46.560 Yeah. 00:19:46.560 --> 00:19:46.840 Okay. 00:19:46.840 --> 00:19:51.280 Does go must have a C C integration option somewhere, right? 00:19:51.280 --> 00:19:51.840 It does. 00:19:51.840 --> 00:19:52.000 Yeah. 00:19:52.000 --> 00:19:52.260 Yeah. 00:19:52.260 --> 00:19:52.480 Okay. 00:19:52.480 --> 00:19:55.940 And the other thing you said is not going to support is things like eval. 00:19:55.940 --> 00:20:03.680 And again, this is like, it is possible to implement something that's a little bit hokey to support 00:20:03.680 --> 00:20:04.840 eval or exec. 00:20:04.840 --> 00:20:06.100 Shell out and compile. 00:20:06.100 --> 00:20:07.320 Oh yeah, exactly. 00:20:07.320 --> 00:20:11.740 I mean, like that's, well, I mean, it's funny you think about it and like, that's, that's actually 00:20:11.740 --> 00:20:13.680 what Python is doing, right? 00:20:13.680 --> 00:20:18.100 It's like, except that it's a bytecode compiler and then it's executing in a VM. 00:20:18.360 --> 00:20:26.080 If you instead are actually doing a, you know, an actual static compilation and then executing 00:20:26.080 --> 00:20:26.580 that. 00:20:26.580 --> 00:20:32.300 It's not conceptually that much different, except that the tool chain that you have to use to 00:20:32.300 --> 00:20:33.860 do the compilation and stuff is much heavier. 00:20:33.860 --> 00:20:38.060 So it's going to be slower and it just, it kind of doesn't make a lot of sense. 00:20:38.060 --> 00:20:43.660 I think I could see maybe supporting it for, you know, debugging use cases and things like 00:20:43.660 --> 00:20:44.000 that. 00:20:44.140 --> 00:20:50.500 I don't think I kind of want to avoid having to worry too much about like, you know, 00:20:50.500 --> 00:20:52.320 making that performant or whatever. 00:20:52.320 --> 00:20:52.960 Yeah, sure. 00:20:52.960 --> 00:20:56.060 I, I, for one would, don't think I would miss it. 00:20:56.060 --> 00:20:56.820 I think it's fine. 00:20:56.820 --> 00:20:57.240 Yeah. 00:20:57.240 --> 00:21:04.480 The other thing about exec and eval is there's very few cases I've ever come across in all 00:21:04.480 --> 00:21:07.620 my years of programming Python where exec or eval was a good idea. 00:21:08.380 --> 00:21:14.220 So actually like, I kind of think that it's an unnecessary aspect of that language. 00:21:14.220 --> 00:21:14.700 Yeah. 00:21:14.700 --> 00:21:15.300 That's interesting. 00:21:15.300 --> 00:21:20.300 And you know, it is kind of keeping with go in the sense that go is very strict about 00:21:20.300 --> 00:21:23.680 conventions and some of the best practices that it believes. 00:21:23.680 --> 00:21:29.000 Like for example, if you have an import of a package and you're not using that package, 00:21:29.000 --> 00:21:30.420 that's a compilation error, right? 00:21:30.420 --> 00:21:31.260 Things like that. 00:21:31.260 --> 00:21:31.680 Right. 00:21:31.680 --> 00:21:32.160 Absolutely. 00:21:32.160 --> 00:21:32.580 Yep. 00:21:32.580 --> 00:21:32.820 Yeah. 00:21:32.820 --> 00:21:35.560 So eval skipping eval seems like that's all right. 00:21:35.560 --> 00:21:51.940 This portion of talk Python to me is brought to you by hired hired is the platform for 00:21:51.940 --> 00:21:56.560 top Python developer jobs, create your profile and instantly get access to 3,500 companies 00:21:56.560 --> 00:21:58.120 who will work to compete with you. 00:21:58.120 --> 00:22:02.380 Take it from one of hired users who recently got a job and said, I had my first offer on 00:22:02.380 --> 00:22:06.280 Thursday after going live on Monday and I ended up getting eight offers in total. 00:22:06.280 --> 00:22:09.740 I've worked with recruiters in the past, but they've always been pretty hit and miss. 00:22:09.740 --> 00:22:12.580 I tried LinkedIn, but I found hired to be the best. 00:22:12.580 --> 00:22:14.640 I really liked knowing the salary upfront. 00:22:14.640 --> 00:22:17.040 Privacy was also a huge seller for me. 00:22:17.040 --> 00:22:18.260 Sounds awesome. 00:22:18.260 --> 00:22:18.720 Doesn't it? 00:22:18.720 --> 00:22:20.740 Well, wait until you hear about the signing bonus. 00:22:20.740 --> 00:22:24.180 Everyone who accepts the job from hired gets a thousand dollars signing bonus. 00:22:24.180 --> 00:22:26.820 And as talk Python listeners, it gets way sweeter. 00:22:27.140 --> 00:22:31.220 Use the link hired.com slash talk Python to me and hired will double the signing bonus 00:22:31.220 --> 00:22:32.060 to $2,000. 00:22:32.060 --> 00:22:33.840 Opportunities knocking. 00:22:33.840 --> 00:22:37.600 Visit hired.com slash talk Python to me and answer the door. 00:22:37.600 --> 00:22:49.120 Then you said there's a set of things that you're going to support, but it doesn't yet. 00:22:49.120 --> 00:22:50.760 What are those? 00:22:50.760 --> 00:22:55.800 We talked a little bit about some of this stuff, but like the standard library is not there yet. 00:22:56.080 --> 00:23:00.460 There's a subset of the standard libraries is available currently. 00:23:00.460 --> 00:23:04.440 Can you give us like a percentage of what, how far you are down that path? 00:23:04.440 --> 00:23:09.320 I mean, everyone listening, this, this whole project has been like, I don't get up for three 00:23:09.320 --> 00:23:09.800 or four weeks. 00:23:09.800 --> 00:23:12.460 So it's not like you should have implemented at all. 00:23:12.460 --> 00:23:14.200 It's just curious, like how far you've gotten. 00:23:14.200 --> 00:23:16.120 It's really hard to put a percentage on it. 00:23:16.120 --> 00:23:19.540 I guess, I mean, I probably could like, you know, compare lines of code or something, but 00:23:19.540 --> 00:23:25.120 I think that what's going to happen is you're going to get sort of a core set of libraries 00:23:25.120 --> 00:23:29.060 that run all the other libraries and everything will just kind of fall into place. 00:23:29.500 --> 00:23:33.420 So I think it's, it's sort of more important to count those core libraries. 00:23:33.420 --> 00:23:38.660 And you know, that's, that's things like types and collections and operator and all those 00:23:38.660 --> 00:23:38.980 things. 00:23:38.980 --> 00:23:42.300 And some of those are already there. 00:23:42.300 --> 00:23:46.640 I mean, I feel like, yeah, it's hard to put a number on it. 00:23:46.640 --> 00:23:47.140 Yeah, sure. 00:23:47.140 --> 00:23:52.680 Maybe it's one of those things where it's, it seems like you're not very far and then all 00:23:52.680 --> 00:23:55.220 of a sudden it kind of unlocks and things go really quick. 00:23:55.420 --> 00:23:56.280 That's the dream. 00:23:56.280 --> 00:24:02.320 That's a good way to like, it's a, it's a optimistic view of the future. 00:24:02.320 --> 00:24:03.040 That's right. 00:24:03.040 --> 00:24:09.380 If you're going to clone the repo off GitHub and try things out, like you may be disappointed 00:24:09.380 --> 00:24:12.060 that, that your favorite libraries aren't there. 00:24:12.480 --> 00:24:18.200 There's a good chance that if you have a program that's, that's at all, you know, complex that 00:24:18.200 --> 00:24:20.680 there are some libraries that are missing for you. 00:24:20.680 --> 00:24:23.800 I'd say, I don't know, maybe 20% or something like that. 00:24:23.800 --> 00:24:24.260 Okay. 00:24:24.260 --> 00:24:25.220 Well, that's good. 00:24:25.220 --> 00:24:27.580 And you said you also want to support all the built-ins. 00:24:27.580 --> 00:24:28.240 That's right. 00:24:28.240 --> 00:24:28.720 Yeah. 00:24:28.880 --> 00:24:30.440 That's obviously a good idea. 00:24:30.440 --> 00:24:30.900 Yeah. 00:24:30.900 --> 00:24:32.440 Those are important. 00:24:32.440 --> 00:24:37.960 And again, you know, there's a bunch of stuff that just hasn't a bunch of like functions 00:24:37.960 --> 00:24:45.340 like map and, and reduce and things like that, that I haven't got around to, haven't needed 00:24:45.340 --> 00:24:50.440 to support them yet, but they're actually pretty straightforward to implement by and large. 00:24:50.620 --> 00:24:54.440 So, so I think we're, we're pretty far along on, on that stuff. 00:24:54.440 --> 00:25:01.620 So how much of your focus on Grumpy is going to be to make this a project that you guys could 00:25:01.620 --> 00:25:09.280 use for your specific use cases at Google and then make that a skeleton or base and people can 00:25:09.280 --> 00:25:14.020 come along and add other features and contribute to the open source project to make it more broad 00:25:14.020 --> 00:25:19.720 versus how much are you trying to make this like we're trying to re entirely replace CPython. 00:25:19.720 --> 00:25:24.320 So I think that we, I want to see, I like, okay. 00:25:24.320 --> 00:25:25.420 So, so I put it this way. 00:25:25.420 --> 00:25:33.020 I'm interested in, you know, solving some of these concurrent use cases that don't have a great answer 00:25:33.020 --> 00:25:34.020 in CPython. 00:25:34.020 --> 00:25:35.880 That's the primary focus. 00:25:35.880 --> 00:25:41.880 But I, again, it might be my optimism is showing again, but like, I feel like once you kind of have 00:25:41.880 --> 00:25:48.920 some of those use cases locked down, now people start to use it for, for things you didn't expect 00:25:48.920 --> 00:25:49.500 right away. 00:25:49.500 --> 00:25:57.000 I know that like scientific computing is an area where Python has a really well-established 00:25:57.000 --> 00:26:01.640 libraries and, and NumPy is, is sort of crucial to some of this stuff. 00:26:01.640 --> 00:26:05.140 And that's got, that's C, you know, involves C extensions. 00:26:05.680 --> 00:26:11.640 And I think in the near term, I don't see Grumpy being useful for numerical analysis. 00:26:11.640 --> 00:26:17.740 And, you know, that's kind of compounded by Go doesn't have too many sort of inroads in that 00:26:17.740 --> 00:26:19.160 direction either. 00:26:19.400 --> 00:26:25.140 So, but on the other hand, you know, some of the static, the advantages of like being 00:26:25.140 --> 00:26:33.240 statically compiled and, and type inferencing and compiling down to native operations, that is 00:26:33.240 --> 00:26:37.100 potentially useful for, you know, scientific computing and those sorts of things. 00:26:37.100 --> 00:26:42.720 So, so I kind of see, you know, I want to focus on our, our immediate use cases, but I have this 00:26:42.720 --> 00:26:44.100 kind of idea that there's. 00:26:44.100 --> 00:26:48.800 You know, more opportunities out there once that's, once things are kind of working. 00:26:48.800 --> 00:26:49.460 Okay. 00:26:49.460 --> 00:26:49.700 Yeah. 00:26:49.700 --> 00:26:49.740 Yeah. 00:26:49.740 --> 00:26:51.680 That, that seems like a good roadmap to me. 00:26:51.680 --> 00:26:52.680 It makes a lot of sense. 00:26:52.680 --> 00:26:59.640 So let's talk about the execution engine, which effectively, effectively is the execution engine 00:26:59.640 --> 00:27:01.420 of Go versus CPython. 00:27:01.420 --> 00:27:06.560 So CPython, the Python code gets converted to bytecode. 00:27:06.560 --> 00:27:13.340 Those bytecode instructions are sent to like a super large force for loop switch sort of thing. 00:27:13.340 --> 00:27:15.800 And those are interpreted and run. 00:27:15.800 --> 00:27:17.560 How does Go work? 00:27:17.960 --> 00:27:25.420 Go has a runtime, which is to say that there's code that is running, that's managing things 00:27:25.420 --> 00:27:32.020 like Go routines, which are the equivalent of threads and, and Go programs and garbage collection 00:27:32.020 --> 00:27:33.200 and things like that. 00:27:33.200 --> 00:27:39.780 But much of what is actually happening throughout a Go program is just, is actually, you know, 00:27:39.780 --> 00:27:41.320 low level machine instruction. 00:27:41.320 --> 00:27:47.520 So the Go program, much like a C program is compiled down to a machine code and actually 00:27:47.520 --> 00:27:48.440 executed natively. 00:27:48.440 --> 00:27:48.840 Right. 00:27:48.840 --> 00:27:49.940 That makes sense. 00:27:49.940 --> 00:27:55.420 So you say Go has a garbage collection, which is, is awesome. 00:27:55.420 --> 00:27:57.080 Do you know what kind it is? 00:27:57.080 --> 00:28:02.880 Is it reference counting or is it like mark and sweep or what, what kind of garbage is it 00:28:02.880 --> 00:28:03.480 deterministic? 00:28:03.480 --> 00:28:04.980 How's the garbage collector work and go? 00:28:04.980 --> 00:28:07.280 This is not my area of expertise. 00:28:07.280 --> 00:28:07.780 Nor mine. 00:28:07.980 --> 00:28:10.340 But it is not reference counted. 00:28:10.340 --> 00:28:15.380 So I believe that it is a, and actually this has changed significantly. 00:28:15.380 --> 00:28:22.240 I believe in 1.7, they significantly re retro or sort of retrofitted the garbage collector. 00:28:22.900 --> 00:28:31.440 It mostly just around the way that garbage for particular Go routines is managed garbage that 00:28:31.440 --> 00:28:33.700 is sort of local to particular Go routines. 00:28:33.700 --> 00:28:40.280 And, but it's, it's sort of a traditional, otherwise it's pretty, it's kind of a traditional garbage 00:28:40.280 --> 00:28:45.400 collector that much similar to what Java has. 00:28:46.080 --> 00:28:47.920 But it's actually much simpler. 00:28:47.920 --> 00:28:52.900 Java has a, a number of different algorithms it supports and a lot of tuning parameters. 00:28:52.900 --> 00:28:59.880 Go's garbage collection is fairly, is much simpler and is targeted for the use case of, 00:28:59.880 --> 00:29:03.620 you know, handling requests in a server application and those sorts of things. 00:29:03.620 --> 00:29:04.000 Yeah. 00:29:04.000 --> 00:29:04.520 It makes sense. 00:29:04.600 --> 00:29:07.760 I suspect they highly parallelize that thinking of Go as well. 00:29:07.760 --> 00:29:13.720 One thing you said that's nice about executing ultimately on Go is you said the deployment story 00:29:13.720 --> 00:29:15.080 is a little bit simpler. 00:29:15.080 --> 00:29:23.620 You know, Python, you do, when you deploy a Python program, you are actually including your like 00:29:23.620 --> 00:29:27.460 PY files or at least your PYC files in the deployment. 00:29:28.360 --> 00:29:33.740 And so you have to have some way to sort of package them together and ship them off to 00:29:33.740 --> 00:29:36.040 production or wherever you're running your program. 00:29:36.040 --> 00:29:36.420 Right. 00:29:36.420 --> 00:29:39.880 And beyond that, also the dependencies and the runtime, right? 00:29:39.880 --> 00:29:41.780 So you got to have all of those things. 00:29:41.780 --> 00:29:42.540 That's right. 00:29:42.540 --> 00:29:44.020 Which can make it really tricky. 00:29:44.020 --> 00:29:48.820 And there's things like PY2 AMP, PY2 XE, CX freeze, the Bware project. 00:29:48.820 --> 00:29:53.720 There's a lot of project trying to make that something you can ship around, but it's not simple. 00:29:53.720 --> 00:29:54.680 Yeah, that's right. 00:29:54.800 --> 00:29:59.560 And, you know, so I'm sure people who have run Python in production have run into, you 00:29:59.560 --> 00:30:04.700 know, version mismatches, things like that, using the system Python version, which was, 00:30:04.700 --> 00:30:07.800 you know, different than the one they were developing on and so on. 00:30:07.800 --> 00:30:13.660 The nice thing about statically compiled programs in general is that you, you produce a binary 00:30:13.660 --> 00:30:17.400 and you just, you can put that just about anywhere and it'll run. 00:30:18.000 --> 00:30:21.020 And that's very true for Go programs. 00:30:21.020 --> 00:30:24.720 There's few dependencies in most cases. 00:30:24.720 --> 00:30:30.700 Most of the runtime is actually compiled or is actually linked into the executable. 00:30:30.700 --> 00:30:31.600 Yeah, that's really cool. 00:30:31.600 --> 00:30:35.620 What's the size of like a Hello World compiled output? 00:30:35.620 --> 00:30:35.980 Do you know? 00:30:35.980 --> 00:30:38.460 I have not looked at the size myself. 00:30:38.700 --> 00:30:44.060 I think I saw some comment somewhere that said it was something like three megabytes. 00:30:44.060 --> 00:30:49.320 So it's, it's pretty substantial, but you know, that, that includes a lot of overhead for 00:30:49.320 --> 00:30:54.420 the runtime that, that wouldn't increase significantly if your program grew. 00:30:54.420 --> 00:30:54.780 Right. 00:30:54.780 --> 00:30:55.240 Absolutely. 00:30:55.240 --> 00:30:58.680 Like, you know, the next 10,000 lines add 10 K or something. 00:30:58.680 --> 00:30:58.980 Right. 00:30:58.980 --> 00:30:59.440 Exactly. 00:30:59.440 --> 00:30:59.700 Yeah. 00:30:59.940 --> 00:31:04.180 I think three megs is totally fine to get a good deployment story, stability. 00:31:04.180 --> 00:31:06.640 You run what you shipped, all those things. 00:31:06.640 --> 00:31:11.380 Like if this was 1994, three megs would be a problem, but it's not today, right? 00:31:11.380 --> 00:31:12.700 Yeah, that's right. 00:31:12.700 --> 00:31:13.220 Nice. 00:31:13.220 --> 00:31:20.400 So what sort of optimizations do you think are possible if you run Python on Go, if rather 00:31:20.400 --> 00:31:22.840 than as an interpreted system? 00:31:22.840 --> 00:31:26.660 This is not an area I've dug into significantly yet. 00:31:27.160 --> 00:31:33.560 My thinking is that if you can determine that a particular, for example, a particular integer 00:31:33.560 --> 00:31:40.560 counter in a function is only ever an integer type and it only, you know, uses integer operations 00:31:40.560 --> 00:31:46.000 like increment or, or whatever, then there's no need to go through the whole Python method 00:31:46.000 --> 00:31:50.080 dispatch and creating new integer objects. 00:31:50.080 --> 00:31:56.540 Every time you increment that counter, you can actually just use a native integer and increment 00:31:56.540 --> 00:31:57.960 using native operations. 00:31:57.960 --> 00:32:02.500 So that's a, that's a really simple example, but not, not uncommon. 00:32:02.500 --> 00:32:08.900 I think once you kind of broaden that to a whole program optimization, that's when things 00:32:08.900 --> 00:32:12.520 start to get interesting because then you can think about like, well, if you know that a 00:32:12.520 --> 00:32:18.860 function is only ever called with particular parameters or parameters of a particular type, 00:32:18.940 --> 00:32:25.060 then you can make some assumptions and again, use native, maybe use native data types. 00:32:25.060 --> 00:32:25.420 Sure. 00:32:25.420 --> 00:32:27.480 What about type annotations? 00:32:27.480 --> 00:32:35.220 And I know that's more a Python three thing, but would you be able to, or interested in having some flavor that 00:32:35.220 --> 00:32:39.400 takes type annotations and then uses that for certain types of optimizations? 00:32:39.400 --> 00:32:40.120 Yeah. 00:32:40.120 --> 00:32:46.160 I thought about this and, and I'm a little ambivalent because, you know, type annotations, the way that 00:32:46.160 --> 00:32:52.540 they are sort of used today, they're not intended to actually, you know, raise or anything if they're 00:32:52.540 --> 00:32:53.420 not respected. 00:32:53.920 --> 00:33:01.280 it's mostly for analysis before you ship your program to like, you know, make the linting, the 00:33:01.280 --> 00:33:02.920 linters job easier and things like that. 00:33:02.920 --> 00:33:03.420 Right. 00:33:03.420 --> 00:33:09.720 And so, when, once it actually in CPython, once your type annotations, once you're actually 00:33:09.720 --> 00:33:12.600 running your program, the type annotations basically have no effect. 00:33:12.600 --> 00:33:21.500 And so I'm a little hesitant to say that grumpy should use these in a more, in sort of a more 00:33:21.500 --> 00:33:27.760 strict way, because I think that might have affect programs compatibility and stuff like 00:33:27.760 --> 00:33:27.980 that. 00:33:27.980 --> 00:33:28.220 Yeah. 00:33:28.220 --> 00:33:29.260 It will absolutely do that. 00:33:29.260 --> 00:33:29.620 Wouldn't it? 00:33:29.620 --> 00:33:29.840 Yeah. 00:33:29.840 --> 00:33:31.580 There's some real advantages there. 00:33:31.580 --> 00:33:36.000 If you, if you do make them strict, then you say that a type, an argument is an integer, 00:33:36.000 --> 00:33:40.660 then yeah, it makes the optimizer's job way easier because it can, you know, it doesn't have 00:33:40.660 --> 00:33:43.480 to do any inferencing to determine that relationship. 00:33:43.480 --> 00:33:49.440 Obviously it would break the sort of contract with type annotations that these are just for editors 00:33:49.440 --> 00:33:53.720 and linters and to help you, but not actually meant to affect the runtime. 00:33:53.720 --> 00:33:54.500 That's right. 00:33:54.500 --> 00:33:59.800 On the other hand, if, if you could make some part of code that's like really critical go, 00:33:59.800 --> 00:34:05.320 you know, 10 or a hundred times faster by putting a type annotation that's strict, you know, 00:34:05.320 --> 00:34:06.960 you might be willing to make that trade off. 00:34:06.960 --> 00:34:10.600 So I have no, I don't know which way would be the right way to go either, but it's interesting 00:34:10.600 --> 00:34:11.180 to think about. 00:34:11.180 --> 00:34:11.500 Yeah. 00:34:11.780 --> 00:34:15.700 I'm very curious how that sort of evolves. 00:34:15.700 --> 00:34:16.040 Yeah. 00:34:16.040 --> 00:34:16.460 Yeah. 00:34:16.460 --> 00:34:16.640 Yeah. 00:34:16.640 --> 00:34:18.060 I'm going to keep an eye on it. 00:34:18.060 --> 00:34:18.380 That's cool. 00:34:18.380 --> 00:34:20.220 So let's talk about when you launched. 00:34:20.220 --> 00:34:22.260 So this should be pretty fresh in your mind, right? 00:34:22.260 --> 00:34:23.420 Yeah. 00:34:23.420 --> 00:34:25.000 It's not a very old project. 00:34:25.000 --> 00:34:27.160 It's about a week, week and a day. 00:34:27.160 --> 00:34:27.480 Yeah. 00:34:28.060 --> 00:34:28.380 Yeah. 00:34:28.380 --> 00:34:37.700 So we, well, I guess we migrated the code to GitHub in mid-December and I spent some time 00:34:37.700 --> 00:34:42.560 over the next month kind of cleaning it or the next few weeks cleaning up the code and adding 00:34:42.560 --> 00:34:50.920 some functionality for the build system that we were not able to use, obviously the internal 00:34:50.920 --> 00:34:53.440 Google build system in the open source project. 00:34:53.540 --> 00:34:59.260 So I had to build some of that out and then I guess January 4th. 00:34:59.260 --> 00:34:59.780 Yeah. 00:34:59.780 --> 00:35:01.080 I guess it was the 4th. 00:35:01.080 --> 00:35:03.900 That's eight days ago just for the day of the recording. 00:35:03.900 --> 00:35:04.100 Yeah. 00:35:04.100 --> 00:35:04.580 Yeah. 00:35:04.580 --> 00:35:05.380 We did. 00:35:05.380 --> 00:35:11.960 We, we sort of coordinated an open source blog post with, with the actual making the GitHub 00:35:11.960 --> 00:35:18.020 repo public and got a little bit of traction on hacker news and, and yeah, it was kind of 00:35:18.020 --> 00:35:21.140 astonishing how great the reception was. 00:35:21.140 --> 00:35:21.380 Yeah. 00:35:21.380 --> 00:35:22.500 It's going like crazy. 00:35:22.660 --> 00:35:27.580 Like when I took notes to, for this conversation, like four or five days ago, I had said there 00:35:27.580 --> 00:35:28.860 were 5,000 stars in GitHub. 00:35:28.860 --> 00:35:31.420 Now, maybe that was three days ago. 00:35:31.420 --> 00:35:35.880 Now there's 6,000, almost 6,317 contributors. 00:35:35.880 --> 00:35:40.400 That's, that's a pretty serious uptake for a project that's been out for eight days. 00:35:40.400 --> 00:35:40.860 Yeah. 00:35:40.860 --> 00:35:46.180 I, I think the thing that kind of blew me away most was the number of pull requests that 00:35:46.180 --> 00:35:46.800 I got. 00:35:46.800 --> 00:35:52.440 I mean, right on day one, people were digging into the code and, you know, doing like it, 00:35:52.440 --> 00:35:56.960 it, the code, there are tricky parts to the code and it's not necessarily obvious how you 00:35:56.960 --> 00:35:58.800 ought to write certain features. 00:35:59.060 --> 00:36:04.140 And people, you know, really dug in and started filling out some of this functionality that's 00:36:04.140 --> 00:36:09.260 missing and started talking about, you know, well, how are we going to support programs or 00:36:09.260 --> 00:36:15.100 libraries, Python, third party Python libraries out of the box and stuff like that. 00:36:15.100 --> 00:36:16.540 So it's been great. 00:36:16.540 --> 00:36:20.740 I've had a really good time working with some, some of these people that have been contributing. 00:36:20.740 --> 00:36:21.100 Yeah. 00:36:21.100 --> 00:36:21.400 Yeah. 00:36:21.400 --> 00:36:22.700 I would say that's really cool. 00:36:22.700 --> 00:36:28.340 You talked about the code a little bit, looking on GitHub, GitHub thinks it's 77% go 00:36:28.340 --> 00:36:31.740 code, 22% Python code and a bit of a make file. 00:36:31.740 --> 00:36:32.160 Yeah. 00:36:32.160 --> 00:36:32.780 That's about right. 00:36:32.780 --> 00:36:33.860 Yeah, that's about right. 00:36:33.860 --> 00:36:38.380 And, and a lot of that Python code is actually just tests, and benchmarks and things. 00:36:38.380 --> 00:36:41.300 So it's, it, most of it is, is go. 00:36:41.300 --> 00:36:47.360 And, and actually, I guess the standard libraries, which most of which are copied from, from 00:36:47.360 --> 00:36:51.760 other places like CPython, there's pretty substantial amount of Python, but that's not like, you know, 00:36:51.760 --> 00:36:55.960 I don't think about too much about that code since we don't have to write it or maintain it. 00:36:55.960 --> 00:36:56.520 Yeah, absolutely. 00:36:56.520 --> 00:36:59.320 So how do you ensure compatibility in this? 00:36:59.320 --> 00:37:01.760 Like, are you running the standard CPython test? 00:37:01.760 --> 00:37:03.160 That's something that we're working to. 00:37:03.160 --> 00:37:05.620 So that's sort of milestone number one. 00:37:05.700 --> 00:37:11.080 I haven't published a roadmap document yet, but getting to the point where we can run 00:37:11.080 --> 00:37:16.460 the unit test library is going to be a huge milestone because it means we can then run 00:37:16.460 --> 00:37:19.720 the unit tests that are written for CPython. 00:37:19.720 --> 00:37:21.960 That would be a huge milestone just on compatibility. 00:37:21.960 --> 00:37:22.900 Exactly. 00:37:22.900 --> 00:37:30.620 Before, we get there, we've been writing small tests to that, you know, demonstrate, 00:37:30.620 --> 00:37:32.700 um, compatibility concerns and stuff like that. 00:37:32.700 --> 00:37:36.460 And then running those in both Python and Crumpy. 00:37:36.460 --> 00:37:36.900 Okay. 00:37:36.900 --> 00:37:37.540 Yeah. 00:37:37.540 --> 00:37:37.940 Very cool. 00:37:37.940 --> 00:37:44.700 So let's talk about why you chose Go because are there three sort of officially blessed languages 00:37:44.700 --> 00:37:45.140 at Google? 00:37:45.140 --> 00:37:47.640 There's Python, there's Go and Java. 00:37:47.640 --> 00:37:48.100 Is that? 00:37:48.100 --> 00:37:48.780 And C++. 00:37:48.780 --> 00:37:49.980 Is that the story these days? 00:37:49.980 --> 00:37:50.280 Yeah. 00:37:50.280 --> 00:37:50.460 Right. 00:37:50.460 --> 00:37:50.740 Of course. 00:37:50.740 --> 00:37:51.200 And C++. 00:37:51.200 --> 00:37:51.720 So four. 00:37:51.720 --> 00:37:53.420 So why did you choose Go? 00:37:53.420 --> 00:37:55.740 Like you could have tried Jython or something, right? 00:37:55.740 --> 00:38:00.060 Jython is something that we, we've looked into. 00:38:00.960 --> 00:38:03.940 Jython is a really great mature product. 00:38:03.940 --> 00:38:11.840 it's our experience that it's better to start a project on Jython, than to migrate 00:38:11.840 --> 00:38:12.620 to Jython. 00:38:12.620 --> 00:38:18.480 There's a number of compatibility issues, not so much like the kinds of compatibility issues 00:38:18.480 --> 00:38:24.920 like, oh, on an, on CPython, this function returns, a different type or something like 00:38:24.920 --> 00:38:31.700 that more that there are certain constraints of running in the JVM that make certain programs 00:38:31.700 --> 00:38:34.960 not work very well or, or those sorts of things. 00:38:34.960 --> 00:38:37.980 So like performance issues that sort of crop up in those sorts of things. 00:38:37.980 --> 00:38:45.020 It sounds like running on the JVM was not the best concurrency server story as it might've 00:38:45.020 --> 00:38:49.780 been running on Go because Go is more focused on concurrency from the beginning and things 00:38:49.780 --> 00:38:50.240 like that. 00:38:50.240 --> 00:38:52.180 That might be more important to you guys. 00:38:52.180 --> 00:38:52.580 Yeah. 00:38:52.580 --> 00:38:53.640 I think that was part of it. 00:38:53.680 --> 00:38:59.260 I mean, like lightweight Go routines are definitely a big advantage to Go. 00:38:59.260 --> 00:39:02.940 So Java has native threads, which have large stacks. 00:39:02.940 --> 00:39:09.520 And so it has sort of a different performance characteristics for concurrent workloads. 00:39:09.660 --> 00:39:15.600 and so you have to kind of write programs, parallel programs in a slightly different way 00:39:15.600 --> 00:39:23.960 for Java, but also, for real time server applications, the JIT actually can be a liability. 00:39:23.960 --> 00:39:31.940 It becomes difficult to, you know, reproduce certain kinds of, certain kinds of issues, 00:39:31.940 --> 00:39:39.420 debug certain kinds of problems and consistent because consistency of, how requests are handled, 00:39:39.420 --> 00:39:42.160 is really important in these kinds of applications. 00:39:42.160 --> 00:39:48.180 And, and the JIT can make, you know, identical requests behave very differently depending on 00:39:48.180 --> 00:39:50.120 where in the life cycle the program is. 00:39:50.120 --> 00:39:50.420 Sure. 00:39:50.420 --> 00:39:50.780 Yeah. 00:39:50.780 --> 00:39:51.600 That makes a lot of sense. 00:39:51.600 --> 00:39:54.460 Being statically typed, you get a little more predictability. 00:39:54.460 --> 00:39:55.040 Absolutely. 00:39:55.040 --> 00:39:59.660 Well, not, sorry, not statically typed, compiled to like machine instructions rather than digits. 00:39:59.660 --> 00:39:59.960 Yeah. 00:39:59.960 --> 00:40:00.340 Yeah. 00:40:00.340 --> 00:40:00.600 Yeah. 00:40:00.600 --> 00:40:01.040 That's right. 00:40:01.040 --> 00:40:01.240 Yep. 00:40:01.240 --> 00:40:02.520 Hey everyone. 00:40:02.520 --> 00:40:06.500 Let me take just a moment and tell you about a new sponsor with a cool and timely offer. 00:40:06.900 --> 00:40:11.300 This portion of talk Python to me is brought to you by deep learning for computer vision 00:40:11.300 --> 00:40:16.560 with Python, a new book from pi image search.com launching on Kickstarter right now. 00:40:16.560 --> 00:40:21.560 Have you ever wondered how Facebook can not only detect your face in an image, but also recognize 00:40:21.560 --> 00:40:22.820 and tag you as well. 00:40:22.820 --> 00:40:24.140 It's not magic. 00:40:24.140 --> 00:40:30.060 Facebook uses specialized machine learning algorithms called deep learning in pi image search wants 00:40:30.060 --> 00:40:32.920 to pull back the curtain and show you how these algorithms work. 00:40:32.920 --> 00:40:37.220 Their new book is designed from the ground up to help you reach expert status. 00:40:37.220 --> 00:40:41.880 Even if you've never worked with machine learning or neural networks before inside deep learning 00:40:41.880 --> 00:40:46.320 for computer vision with Python, you'll find super practical walkthroughs, hands-on tutorials 00:40:46.320 --> 00:40:50.940 with lots of code and a no fluff teaching style that is guaranteed to cut through all the cruft 00:40:50.940 --> 00:40:53.580 and help you master deep learning for visual recognition. 00:40:53.580 --> 00:40:59.520 To learn more about this book and back the Kickstarter campaign, just head to pi image search.com 00:40:59.520 --> 00:41:00.560 slash Kickstarter. 00:41:01.560 --> 00:41:01.960 Yeah. 00:41:01.960 --> 00:41:03.980 So how do you run apps on, on Grumpy? 00:41:03.980 --> 00:41:09.080 Like if I have Python code and I want to make it, make it go, how do I make it go on Grumpy? 00:41:09.080 --> 00:41:14.760 This is sort of a hot topic right now in the issue tracker on GitHub because like the 00:41:14.760 --> 00:41:20.400 build system that I have and is strictly focused on, you know, getting the internal libraries working. 00:41:20.400 --> 00:41:25.860 And so it doesn't have good support for building a program that's outside that directory structure 00:41:25.860 --> 00:41:29.000 or using libraries that are in your Python path or anything like that. 00:41:29.160 --> 00:41:32.680 And so we're debating kind of how exactly it should be supported. 00:41:32.680 --> 00:41:38.580 So right now, if you want to run a program or compile a library, you have to kind of drop 00:41:38.580 --> 00:41:45.160 it into that directory structure and the make system will pick up on it and, and, try 00:41:45.160 --> 00:41:46.400 to compile it into go. 00:41:46.400 --> 00:41:53.940 But, ideally, you know, you have some kind of Python path style construct where it can find 00:41:53.940 --> 00:41:58.000 Python code and build it in a sort of standard way. 00:41:58.000 --> 00:42:00.280 That's something that we're working towards. 00:42:00.280 --> 00:42:01.640 Okay, cool. 00:42:01.640 --> 00:42:08.180 Now, if people want to contribute to Grumpy, there's like three major areas that, that make it up. 00:42:08.180 --> 00:42:11.480 You want to talk about those three areas so they maybe can use it as a roadmap? 00:42:11.700 --> 00:42:19.280 You can kind of think of it as the trans compiler, which is the tool called Grump C and that takes 00:42:19.280 --> 00:42:23.460 Python code and it actually uses it's written in Python and it uses the AST module. 00:42:23.460 --> 00:42:24.500 So it's kind of cheating. 00:42:25.380 --> 00:42:33.600 Another milestone will be when Grump C can compile Grump C and, that takes, the Python 00:42:33.600 --> 00:42:36.060 code and spits out some go code. 00:42:36.060 --> 00:42:41.680 And then you're going to, the second part is the Grumpy runtime, which is kind of the parallel 00:42:41.680 --> 00:42:42.840 of the C API. 00:42:42.840 --> 00:42:48.160 The trans compiled go code will depend on that runtime. 00:42:48.160 --> 00:42:53.660 So it imports the runtime and uses the, constructs and functions and things in the runtime. 00:42:53.660 --> 00:42:58.500 And so that's another sort of component that's written strictly in go. 00:42:58.500 --> 00:43:02.180 And that's where all the sort of data structures and things are defined. 00:43:02.180 --> 00:43:09.280 and finally there's the standard library that is a mostly written in or actually exclusively 00:43:09.280 --> 00:43:15.320 written in Python, but also has some uses some of the Grumpy native extensions to actually 00:43:15.320 --> 00:43:19.280 interface directly with go packages and, and functions and things. 00:43:19.280 --> 00:43:23.440 so those are, so there's sort of the three areas and there's a lot of work to do in, 00:43:23.440 --> 00:43:24.800 in all of those different areas. 00:43:24.800 --> 00:43:29.480 I'd say like the standard library is, is the biggest chunk of work to do at this point. 00:43:29.480 --> 00:43:34.100 Presumably you guys chose go because of the concurrency story, right? 00:43:34.100 --> 00:43:40.680 And if you have Python code running on go, you want to leverage that concurrency. 00:43:40.680 --> 00:43:42.760 Do you have to use a different API? 00:43:43.400 --> 00:43:44.820 this is Python two seven. 00:43:44.820 --> 00:43:47.040 So you don't have things like async or wait. 00:43:47.040 --> 00:43:50.880 How do I interact with the concurrency model of go? 00:43:50.880 --> 00:43:57.340 Currently, the way that go routines are made available is through the threading library. 00:43:57.340 --> 00:44:00.960 So the standard Python threading library, you create a thread and start it. 00:44:00.960 --> 00:44:09.040 And that actually starts a go routine instead of a native thread that will work pretty seamlessly with existing code. 00:44:09.040 --> 00:44:15.580 I don't foresee huge problems there in terms of like the differences between those kinds of threads. 00:44:15.780 --> 00:44:22.220 And again, like, you know, go has the concept of channels, which are sort of a message passing mechanism. 00:44:22.420 --> 00:44:31.420 And whereas in Python, you have a queue, the queue data structure, and this isn't actually implemented, but I plan to implement a queue using channels. 00:44:31.420 --> 00:44:36.200 And so you should be able to just write Python concurrent Python code like you always have. 00:44:36.200 --> 00:44:48.200 But I think to really take advantage of sort of the concurrency model, you probably, eventually I'd, I'd like to implement the async and await Python construct. 00:44:48.200 --> 00:44:50.420 I think that would be a huge win. 00:44:50.420 --> 00:44:52.880 Yeah, that would be, that would be a huge win. 00:44:52.880 --> 00:45:02.180 And it seems to me like using the threading API is much more coarse grained concurrency than go is really built for. 00:45:02.180 --> 00:45:06.120 And while it would work, it's not, not taking full advantage. 00:45:06.120 --> 00:45:20.400 The idea with go is you can start a go routine or starting a go routine is extremely lightweight and passing messages back and forth is the way to sort of share state rather than with sharing memory. 00:45:20.400 --> 00:45:21.400 Or sharing objects. 00:45:21.400 --> 00:45:34.780 So I think that programs that are written with sort of heavyweight threads in mind aren't necessarily going to be the best possible way to express that functionality. 00:45:34.960 --> 00:45:44.900 And so, you know, long-term I could see, you know, maybe, well, actually because you can access native go constructs. 00:45:44.900 --> 00:45:49.960 For example, you will be able to, in a grumpy program, use go channels directly. 00:45:50.360 --> 00:45:52.580 You know, that has upsides and downsides. 00:45:52.580 --> 00:45:55.840 It starts to diverge from the Python language and those sorts of things. 00:45:55.840 --> 00:46:00.460 Yeah, but it's not unlike, Iron Python or Jython or those things, right? 00:46:00.460 --> 00:46:05.700 Where you can reach down into the underlying JVM or CLR or something like that. 00:46:05.700 --> 00:46:06.100 That's right. 00:46:06.100 --> 00:46:06.340 Yep. 00:46:06.340 --> 00:46:06.880 Absolutely. 00:46:07.740 --> 00:46:08.020 Okay. 00:46:08.020 --> 00:46:13.700 So if you're going towards async and await, what's the story on Python three? 00:46:13.700 --> 00:46:17.260 Since I feel like the threading concurrency story is a lot better in Python three. 00:46:17.260 --> 00:46:17.720 Yeah. 00:46:17.720 --> 00:46:19.460 I'd love to support Python three. 00:46:19.460 --> 00:46:22.480 The long-term goal is definitely to support it. 00:46:22.480 --> 00:46:29.840 The reason for 2.7 is that we have a large, YouTube had a large existing Python code base and 00:46:29.840 --> 00:46:31.800 that was a 2.7. 00:46:31.800 --> 00:46:37.400 So that was the main reason for choosing 2.7 out of the gate, but certainly long-term, 00:46:37.400 --> 00:46:39.820 I'd like to see all Python three supported. 00:46:39.820 --> 00:46:40.280 Right. 00:46:40.280 --> 00:46:41.540 Oh, that'd be, that'd be fantastic. 00:46:41.540 --> 00:46:42.600 I'd like to see that as well. 00:46:42.600 --> 00:46:45.200 I mean, it certainly makes sense if you're working on the YouTube team. 00:46:45.200 --> 00:46:51.780 YouTube has a tremendously large and widely adopted deployment of Python two seven. 00:46:51.780 --> 00:46:55.760 Like you want to, you know, work where you can have the biggest impact locally, right? 00:46:55.760 --> 00:46:56.280 Absolutely. 00:46:56.280 --> 00:46:56.840 Yeah. 00:46:56.840 --> 00:47:02.400 So reading the tea leaves, does this mean that Grumpy might someday run YouTube? 00:47:02.400 --> 00:47:04.120 I want to hedge a little bit on that. 00:47:04.120 --> 00:47:10.940 I think there's a sort of a long road ahead before Grumpy's ready to handle the kinds of 00:47:10.940 --> 00:47:13.740 large applications that we run on YouTube. 00:47:13.740 --> 00:47:18.700 So I wouldn't want to speculate about the long-term outcomes there. 00:47:18.700 --> 00:47:18.880 Sure. 00:47:18.880 --> 00:47:19.280 Yeah. 00:47:19.280 --> 00:47:19.440 Yeah. 00:47:19.440 --> 00:47:19.760 Of course. 00:47:19.760 --> 00:47:23.620 You know, let me just imagine, let's imagine a world where it did. 00:47:23.620 --> 00:47:30.280 That would be, probably the first few weeks that it switched to Grumpy would be a little 00:47:30.280 --> 00:47:31.280 bit nerve wracking, right? 00:47:31.280 --> 00:47:31.680 Yeah. 00:47:31.680 --> 00:47:33.540 It would definitely. 00:47:33.540 --> 00:47:36.360 If YouTube goes down and it's your fault, that's going to be a problem. 00:47:36.360 --> 00:47:37.200 Yeah, exactly. 00:47:37.200 --> 00:47:39.240 I don't want to be that guy. 00:47:39.240 --> 00:47:40.380 Exactly. 00:47:40.380 --> 00:47:40.860 Exactly. 00:47:40.860 --> 00:47:42.520 Here's the four pages we're giving you. 00:47:42.520 --> 00:47:43.100 No, just kidding. 00:47:43.100 --> 00:47:47.460 But it would, if, if someday that, that came to be, that would be a really cool outcome of 00:47:47.460 --> 00:47:47.860 this project. 00:47:47.860 --> 00:47:48.500 Yeah, absolutely. 00:47:48.500 --> 00:47:50.080 that's, that's sort of the dream. 00:47:50.080 --> 00:47:50.440 Excellent. 00:47:50.440 --> 00:47:51.020 Okay. 00:47:51.020 --> 00:47:52.820 So maybe that's, that's a good place to leave it. 00:47:52.820 --> 00:47:56.200 Let me ask you just a couple of questions before we let you out of here. 00:47:56.200 --> 00:47:58.820 If you're going to write some code, what editor do you use? 00:47:58.820 --> 00:47:59.480 Vim. 00:47:59.480 --> 00:47:59.880 Vim. 00:47:59.880 --> 00:48:00.460 All right. 00:48:00.460 --> 00:48:00.680 Yeah. 00:48:00.680 --> 00:48:01.360 Very cool. 00:48:01.360 --> 00:48:06.200 And there's over 96,000 packages on PyPI these days. 00:48:06.200 --> 00:48:09.020 And I'm sure you've come across some that are kind of unique. 00:48:09.020 --> 00:48:11.180 You're like, Hey, have you heard about this package? 00:48:11.180 --> 00:48:11.740 It's pretty cool. 00:48:11.740 --> 00:48:12.360 You should check it out. 00:48:12.360 --> 00:48:13.800 You got any, coming to mind? 00:48:13.800 --> 00:48:14.540 You know, it's funny. 00:48:14.540 --> 00:48:20.560 I mean, because I do a lot of my, most of my development inside Google, you know, we 00:48:20.560 --> 00:48:25.640 kind of have a different set of tools we tend to use. 00:48:26.020 --> 00:48:30.580 I don't have a ton of, experience with a lot of PyPI packages. 00:48:30.580 --> 00:48:30.940 Yeah. 00:48:30.940 --> 00:48:33.240 So it's a little bit more a dark matter. 00:48:33.240 --> 00:48:38.040 We out here in the larger universe don't get to see a lot of the cool stuff you guys get 00:48:38.040 --> 00:48:38.280 to use. 00:48:38.280 --> 00:48:39.560 I'm sure it's pretty neat though. 00:48:39.560 --> 00:48:40.220 Absolutely. 00:48:40.220 --> 00:48:41.140 All right. 00:48:41.140 --> 00:48:41.800 Awesome. 00:48:41.800 --> 00:48:43.360 So how about a final call for action? 00:48:43.360 --> 00:48:44.900 Like how can people get started grumpy? 00:48:44.900 --> 00:48:48.180 What can they do if they, if this resonates with them, things like that? 00:48:48.180 --> 00:48:48.680 yeah. 00:48:48.720 --> 00:48:52.860 I mean, we're, we're super interested in, in seeing where the project goes. 00:48:52.860 --> 00:48:57.700 I, I don't have, like I said, I would like to see, where grumpy can be useful 00:48:57.700 --> 00:49:01.460 besides just, you know, large concurrent server applications. 00:49:01.460 --> 00:49:04.120 Community feedback around that is great. 00:49:04.120 --> 00:49:08.220 I, people have been filing, issues asking about, you know, support for different things. 00:49:08.220 --> 00:49:12.160 And that's been really illuminating seeing where people are thinking about where this might 00:49:12.160 --> 00:49:12.520 be useful. 00:49:12.520 --> 00:49:13.880 So that's huge. 00:49:13.880 --> 00:49:20.820 if, you have the time and the inclination, try it out, just clone the repo and type make 00:49:20.820 --> 00:49:25.820 run and, and try out Python and go and, report any issues. 00:49:25.820 --> 00:49:27.320 That's really useful to us. 00:49:27.320 --> 00:49:30.700 And, and obviously there's a ton of work to do. 00:49:30.700 --> 00:49:36.240 we talked about some of the different things and, you know, contributions, via PR, 00:49:36.240 --> 00:49:38.680 pull requests on GitHub are really appreciated. 00:49:38.680 --> 00:49:42.460 It's been kind of amazing how much people effort people have put in already. 00:49:42.460 --> 00:49:45.700 So that's been, really exciting for us. 00:49:45.700 --> 00:49:45.920 Yeah. 00:49:45.920 --> 00:49:47.480 It's, it's a cool project. 00:49:47.480 --> 00:49:54.100 And I think if we have yet another powerful, flexible runtime that has some different trade 00:49:54.100 --> 00:49:56.760 offs that we can make for Python, that's great for everyone. 00:49:56.760 --> 00:50:00.200 So congratulations on your project and thanks for sharing it with everyone. 00:50:00.200 --> 00:50:00.460 Yeah. 00:50:00.460 --> 00:50:01.360 Thanks very much, Michael. 00:50:01.360 --> 00:50:01.780 You bet. 00:50:01.780 --> 00:50:02.380 Talk to you later. 00:50:02.380 --> 00:50:06.660 This has been another episode of talk Python to me. 00:50:06.660 --> 00:50:09.000 Today's guest has been Dylan Trotter. 00:50:09.000 --> 00:50:12.600 And this episode has been sponsored by hired and pie image search. 00:50:12.600 --> 00:50:14.620 Thank you both for supporting the show. 00:50:14.620 --> 00:50:17.800 Hired wants to help you find your next big thing. 00:50:17.800 --> 00:50:22.500 Visit hired.com slash talk Python to me to get five or more offers with salary and equity 00:50:22.500 --> 00:50:26.380 presented right up front and a special listener signing bonus of $2,000. 00:50:27.700 --> 00:50:31.460 Struggling to get started with neural networks, deep learning and image recognition. 00:50:31.460 --> 00:50:33.900 Pie image search.com can help with that. 00:50:33.900 --> 00:50:39.080 To learn more about their new book, deep learning for visual recognition with Python and back the 00:50:39.080 --> 00:50:39.900 Kickstarter campaign. 00:50:39.900 --> 00:50:43.560 Just head to pie image search.com slash Kickstarter. 00:50:44.500 --> 00:50:46.520 Are you or a colleague trying to learn Python? 00:50:46.520 --> 00:50:51.180 Have you tried books and videos that just left you bored by covering topics point by point? 00:50:51.180 --> 00:50:57.180 Well, check out my online course Python Jumpstart by building 10 apps at talkpython.fm/course 00:50:57.180 --> 00:50:59.800 to experience a more engaging way to learn Python. 00:50:59.800 --> 00:51:04.580 And if you're looking for something a little more advanced, try my Write Pythonic Code course 00:51:04.580 --> 00:51:07.120 at talkpython.fm/pythonic. 00:51:08.520 --> 00:51:10.180 Be sure to subscribe to the show. 00:51:10.180 --> 00:51:12.380 Open your favorite podcatcher and search for Python. 00:51:12.380 --> 00:51:13.620 We should be right at the top. 00:51:13.620 --> 00:51:19.740 You can also find the iTunes feed at /itunes, Google Play feed at /play and direct 00:51:19.740 --> 00:51:22.920 RSS feed at /rss on talkpython.fm. 00:51:22.920 --> 00:51:28.020 Our theme music is Developers, Developers, Developers by Corey Smith, who goes by Smix. 00:51:28.020 --> 00:51:31.040 Corey just recently started selling his tracks on iTunes. 00:51:31.040 --> 00:51:34.720 So I recommend you check it out at talkpython.fm/music. 00:51:34.720 --> 00:51:38.480 You can browse his tracks he has for sale on iTunes and listen to the 00:51:38.480 --> 00:51:40.080 full length version of the theme song. 00:51:40.080 --> 00:51:42.140 This is your host, Michael Kennedy. 00:51:42.140 --> 00:51:43.420 Thanks so much for listening. 00:51:43.420 --> 00:51:44.640 I really appreciate it. 00:51:44.640 --> 00:51:46.760 Smix, let's get out of here. 00:52:08.440 --> 00:52:09.020 Don't forget. 00:52:09.020 --> 00:52:11.080 you