WEBVTT 00:00.000 --> 00:15.200 Good. Wonderful. Okay. So, this is even with a Y. Yeah? Snacker? Yeah, it's actually 00:15.200 --> 00:20.880 pronounced like that. Okay, great. Even Snacker. And he's going to, like, I find, I find the 00:20.880 --> 00:26.040 title of this talk really amazing. The title of this talk is tips and tricks to fix your 00:26.040 --> 00:33.080 nicks, and it's like, so poetic. So, yeah, I was planning to say it one last time to it 00:33.080 --> 00:38.080 for the chemical effect, but yeah. Sorry. No, I'm going to, I'm not going to say anything anymore. 00:38.080 --> 00:47.520 I'm going to leave you the stage. A round of applause, please for Yvonne. And our last talk 00:47.520 --> 00:55.600 for today, go ahead. Yeah. So, hello everyone. And welcome in tips and tricks to fix your nicks. 00:55.600 --> 01:05.840 I say it. So, the truth is, I already gave that talk a couple of months ago in French, 01:05.840 --> 01:13.440 in capital Dileb, in Toulouse, and it was totally semi-improvised. And this is a revised version 01:13.440 --> 01:19.320 of it that was a bit shrink to fit the, the, the Devron schedule, because that's the last 01:19.360 --> 01:28.120 talk of the day. And the motivation behind this talk is, I guess, as true for every technology, 01:28.120 --> 01:34.520 there is stuff that are super obvious for people to use to, to the scene. We're talking 01:34.520 --> 01:44.440 about nicks, that are super cryptic to the newcomers. And in nicks, if you copypaste some 01:44.440 --> 01:48.360 snippet from the official wiki into your configuration, most of the time it works, it 01:48.360 --> 01:54.680 is magic, it's super nice. But sometimes it breaks, and when it breaks, it's terrible. And 01:54.680 --> 02:00.120 I feel that's one of the things where it's true with nicks, that's, there's things that 02:00.120 --> 02:04.680 are obvious for some people that are not for those. So, maybe some of you will not learn 02:04.680 --> 02:10.360 anything new, and maybe some of you will, and that's my hope. And in the end, there are a few links 02:10.360 --> 02:15.320 because it's a bit too short to make a lot of demo. And it's not exhaustive. If you have more 02:15.400 --> 02:20.040 idea, we can discuss, is it always track, or I would try at some point to make that into, 02:20.040 --> 02:28.680 I know a wiki update, a blog post, something like that. Anyway, I'm involved. I urge first 02:28.680 --> 02:34.840 about nicks, I guess, a decade ago, but I really start to use it, like daily driving it, 02:35.880 --> 02:41.720 why it was the pandemic, the last one. Yeah, we were in first them, so that's the best 02:41.800 --> 02:48.760 spot to catch the last cool versus. And anyway, no, nicks is paying my rent every month, 02:50.120 --> 02:58.040 just nicks, so it's a super cool technology. And yeah, what we'll talk about today is 02:58.040 --> 03:03.320 fitting to two different parts that are really the two parts that explain all nicks work, 03:03.320 --> 03:08.920 which is first, you're evaluating a bunch of nicks code to producing the relation file, 03:09.000 --> 03:14.120 and then you're done, you're actually realizing buildings of the derivation 03:14.120 --> 03:20.600 fighting to actual binaries, crypto, whatever it puts. And there's two separated steps, 03:20.600 --> 03:26.920 actually, really present in nicks, in frogs, like licks, in even gicks, all are the same with that. 03:26.920 --> 03:32.360 And depending if your error is happening in the first or second stage of this process, 03:32.360 --> 03:39.720 it's really different experience. So first, the obvious things, the beginning expression, 03:39.720 --> 03:47.160 first part. Yeah, nicks is a really cool language, personally, I like the syntax, 03:47.160 --> 03:51.160 I think it's pretty decent language, there's really weird color case, but it's not the point of 03:51.160 --> 04:00.120 this talk. There is issue with nicks, first, a really dynamic core and language where you can 04:00.200 --> 04:05.480 do a lot of things that will not be possible in more static language, indeed, 04:05.480 --> 04:10.840 and it's functional. And functional language, I have a feature, which I personally find really 04:10.840 --> 04:15.720 cool and that is by default for every function, that is there is partial application, 04:16.600 --> 04:24.440 which means that if you forgot an argument to your function, it would work, it will just 04:24.520 --> 04:29.320 raise a neural way later in your program, in your expression, which is actually bad, 04:29.320 --> 04:35.880 because the stack trace will not show you the place where you made zero, actually, but somewhere 04:35.880 --> 04:45.720 totally different. And six sets, a bit tricky is you to work with, I have, yeah, and so 04:45.720 --> 04:52.760 that what they just say, the stack trace is often bad. My immediate advice for that, advice for that 04:52.920 --> 04:59.080 is if you have the choice to not use several argument for a function, if you're crafting 04:59.080 --> 05:04.200 your nicks piece of code, and you can, rather, make a function that can naturally be set, 05:04.200 --> 05:10.680 rather than several named argument, a positional argument, sorry, to choose that should be set, 05:10.680 --> 05:17.240 at least if you forgot a knitem, you would have a proper error in the right place, use only 05:18.200 --> 05:22.360 a function that takes several argument, which is a function that takes a function, a functional 05:22.360 --> 05:31.800 program, only when you really need it, not just often. And yeah, and I made runtime types, 05:32.600 --> 05:40.200 libraries that catch this problem, and no, that's not, yeah, this is a flying copy of my website. 05:41.160 --> 05:46.280 Yeah, I presented two years ago at first, them, so, not really interesting, but yeah, 05:47.640 --> 05:54.120 self-advertising. A lot of scheme are similar, mechanism built in, so, yeah. 05:55.560 --> 06:00.200 Next thing is, often when you start with nicks, you're like, it's just a configuration language, 06:00.200 --> 06:05.720 so I do not have to really tune my editor, like, it's real programming language, 06:05.800 --> 06:12.040 like that way you should really do that, use a language several protocol, use stuff like 06:12.040 --> 06:18.360 pre-commit, or pre-push, oops, all this kind of things, there's a lot of static analyzer 06:18.360 --> 06:23.480 for nicks, like dead nicks, etc. There really makes your experience way better, because it's 06:23.480 --> 06:31.400 really easy to, I don't know, for good, parenthesis is nicks, and you really have no clue, 06:31.400 --> 06:39.800 and I think a cool editor really helps you, that's abuse, but more saying, and next there is a 06:39.800 --> 06:45.560 really cool REPL in nicks, that's when I start using it to debug my program interactive, 06:45.560 --> 06:52.040 it's really changed my life, and I think everyone should do that, and that will be the point 06:52.040 --> 07:00.040 of the first demo, and the next thing is, when you ask someone of two, or to debug an 07:00.120 --> 07:07.400 nicks script for a while, I mean for more than 20 years, the answer was just traced, which is 07:07.400 --> 07:14.760 a bit bad, and no, it's not that new, but no, we have a debugger, and so I will just demo it, 07:14.760 --> 07:19.960 because I think that's really cool features, that's a lot of people are not where it exists, 07:20.920 --> 07:26.600 and yes, so that's demo time, yeah, so I have two terminal window, 07:28.440 --> 07:33.240 on the top, it's some nicks expression, really simple, or my variable start with a z, 07:33.240 --> 07:39.000 because it will be shown in the end of all variables in my REPL, it's convenient, and I have 07:39.080 --> 07:47.160 the REPL, and there is, I think I can load a nicks expression, it's called nicks REPL, 07:48.520 --> 07:56.200 demo, yeah, I load it, I load only one variable, which is the final variable, the z, zc, 07:57.640 --> 08:04.920 and I can evaluate this value, it's 42, it's super interesting, but no, I can trace 08:05.880 --> 08:12.680 and trace, it's super obvious if you ever used functional programming, and if not, I'm really sorry for you, 08:13.960 --> 08:20.920 within trace takes two argument, of course, the first one is what will be traced, and the second one 08:20.920 --> 08:27.880 will be what built-in trace is written as a value, and so two things could be two different number, 08:28.040 --> 08:36.040 but most of the time, what I do is I do in built-in trace x, and I define x as the 08:36.040 --> 08:43.480 thing, I want both to return and both to evaluate, this is also the only way to do logs, 08:44.280 --> 08:49.480 when you want to do logs, logs up to return value, and you have to ask for this value for the 08:49.480 --> 08:56.200 logs to be printed, I will demonstrate that just in a moment, so I reload my file, 08:58.600 --> 09:07.720 and if I do zc, no, I have a trace, and I was result, but you know, maybe if you're coming from 09:07.720 --> 09:15.160 asker, it's like is there a tank or is there a value that I've been compute, if I re-ask for zc, 09:15.320 --> 09:21.560 I don't have to trace any more, I just have to trace at the time one, I ask for zc, only one tank, 09:21.560 --> 09:26.280 so when you're going to just lazy, it's maybe obvious, it's maybe not, and so early PL is a really 09:26.280 --> 09:34.760 good way to have a good understanding of that, then I want to show off the new things that is called 09:35.160 --> 09:39.960 oops, sorry, the new things that is called, and I have a LSP, it's super convenient, 09:39.960 --> 09:48.120 within Brexit, so I have a breakpoint and I have, yeah, I don't need built in actually, 09:48.120 --> 10:03.240 thank you LSP, so yeah, and if I reload and I do zc, it's just doing new things, maybe I can, 10:03.960 --> 10:10.040 I don't know clothes and relinquish your RAPL, and we try, it's not working, why, because it's 10:10.040 --> 10:17.320 obvious, I have to use the debugger, optional flag, because if you don't use it, the debugger is not 10:17.320 --> 10:23.720 enabled by default, which is great, which is sometimes a bit annoying when your nix command is 10:23.720 --> 10:31.080 called, deep into some strapting or CI, anyway, no, I have a debugger, and if I do, I load my 10:31.160 --> 10:41.560 file, I compute zc, no, I have breakpoint, and if I check, no, I have access to za and b, okay, I've 10:41.560 --> 10:51.080 lost a value, and I have the same thing, first time I have a trace, so if I do continue, I don't know 10:51.080 --> 10:58.920 my command, c, yes, it's continue, I have the result, and no, you see, I add the result, without the 10:58.920 --> 11:05.560 trace, because zd was already compute, so yeah, you get set on this thing, it's really, really, 11:05.560 --> 11:11.960 ended to use RAPL, to use this kind of buildings, to use the debugger, to understand actually, 11:11.960 --> 11:21.080 all an xx% is evaluated, and to not be blind, I really recommend it, and yeah, that was a short 11:21.080 --> 11:27.880 demo about it, and second part is debugging builds, so when your success would be evaluated, 11:27.960 --> 11:32.520 wall nix expression, you produce derivation file, I know your building, there's derivation 11:32.520 --> 11:39.320 file inside the sandbox, those demo are a bit artren without a reliable internet connection, 11:39.320 --> 11:43.240 and I was a bit afraid by it for the internet connection, so it's everything is pre-recorded, 11:44.760 --> 11:51.080 so what is the problem, the problem is this slide is wrong, no, because nix actually is incremental, 11:51.160 --> 11:58.120 there is an atomic measure in nix, which is derivation, if you already build a derivation, 11:58.120 --> 12:04.600 we're not durable exact derivation, so you can make your nix build more incremental by making 12:04.600 --> 12:10.440 more derivation, more derivation, more we're not reviewing everything each time, but inside the 12:10.440 --> 12:16.760 derivation, if you have an issue, you will always rebuild the wall, script inside the derivation, 12:16.840 --> 12:22.360 inside the sandbox from scratch, and that's not interactive at all, if you use to do occur, 12:22.360 --> 12:28.520 you're like, oh, I will spawn a shell in my do occur and just mess everything, you can do that 12:28.520 --> 12:35.160 easily with nix for obvious reason, maybe not so obvious, but yeah, that's something that often 12:35.160 --> 12:44.440 you can also already frustrated with, because your feedback loop is super slow, and also, 12:44.520 --> 12:50.760 then another problem I didn't wrong hit on the slide, but nix is actually when you're running, 12:50.760 --> 12:56.600 building a lot of derivation in parallel, in arbitrary order, so sometimes you have an issue, 12:56.600 --> 13:02.280 you run your command, you don't have the same error message, because it's just fail at another place, 13:02.280 --> 13:08.520 and it will not never, I mean, it's this rigid system, but it don't fail or 13:08.600 --> 13:13.960 really agree when there is more than one thing to fix, and sometimes you really are to see what's 13:14.680 --> 13:19.400 the past between what you're actually trying to do and what's these derivations that are failing, 13:21.800 --> 13:28.760 some of his first advice is to do something else, since default nix would put, which is quite 13:28.760 --> 13:36.040 arbitrary, you have the dash l option, l caps, but anyway nix with monitor is really great, 13:36.200 --> 13:43.640 and there is a modified version of it, that's even used to do things more in parallel, but 13:44.920 --> 13:50.920 it's not the point of his talk, and give you two, I guess, I have some same 13:51.960 --> 14:02.520 next trick, but I will show that later, as Kinima play pre-record nix input monitor, 14:03.400 --> 14:09.720 yeah, so the nix input monitor, it looks like that, you have this kind of 14:11.080 --> 14:18.600 things where you have no more nix l caps, it would put, but you see, at some point it will appear, 14:19.720 --> 14:24.680 some dependency graph where you see what you are actually building, maybe that's not the best example, 14:26.360 --> 14:29.640 it shows some kind of tree when you can see, okay, I'm building this derivation, 14:29.640 --> 14:36.120 because it's in this derivation and starts to, it's really great to, and it's really easy to use, 14:36.120 --> 14:45.800 you just replace nix binomial command 9, please use it, it's not new, but yeah, use that too, 14:47.400 --> 14:56.760 yeah, what's next, it's super easy to inspect the derivation to actually see what's inside, 14:57.160 --> 15:02.280 and it's super nice to understand what's the relationship between two derivations, you have y-dependes, 15:02.280 --> 15:11.240 and you have nix tree for that, sorry, so y-dependent gives you this kind of outputs, 15:11.240 --> 15:23.880 and nix tree too much window, nix tree gives you something like that, yeah, I here, I'm inspecting 15:24.600 --> 15:31.480 the developer shell of my willy-twork, of the environment I used to build this light, 15:32.200 --> 15:40.360 yeah, it's super fast, we are almost there, yeah, and I'm out of time, so just show it 15:40.360 --> 15:46.360 very quickly and then the torque is finished, so yeah, my derivation is to plus more, it's just six 15:46.360 --> 15:53.240 gigabyte to build slides, a pretty reasonable, yeah, I can see, use text live, what's text, 15:53.240 --> 16:00.040 space in text live, and you can see there's maybe stuff you can remove from your derivations, 16:00.040 --> 16:09.240 kind of since, so yeah, blah, blah, blah, blah, and yeah, and yeah, and there's two other things, 16:09.240 --> 16:15.640 I didn't have the time to show, the first one is the nickel, rip on out, and you can basically 16:15.800 --> 16:20.520 go interactively inside your derivation, environment, and test everything, and you can also write 16:20.520 --> 16:26.600 any test for your nix code, yeah, there's two links in the slides, I will be online, this is 16:26.600 --> 16:32.520 this course, I suppose that takes pain off to use, there's also sources of two is big point, 16:32.520 --> 16:38.200 okay, and there's really old, three-block posts, quite old, that takes pain off to write, 16:38.200 --> 16:43.240 you need test for your nix expression, which is something, I don't see a lot of people doing, 16:43.240 --> 16:56.440 but I think it's nice to do, and that's all, thank you, all right, so we are out of time, 16:56.440 --> 17:02.760 so we don't really have time for Q&A anymore, but this was great, great talk, I think it's really 17:02.760 --> 17:07.080 important for people to get to know these tools because it's really hard to get started with nix, 17:07.080 --> 17:13.160 and knowing that stuff really does help, so thank you very much for the talk, thank you very 17:13.480 --> 17:20.600 much for everyone who took part in this in some way, it was a lot of fun, maybe we'll see you 17:20.600 --> 17:27.240 at the Booth tomorrow or something, hopefully people will have good conversations following this, 17:27.240 --> 17:31.720 and have a nice evening, and enjoy the rest of the first, and thank you, bye.