WEBVTT 1 00:00:00.000 --> 00:00:05.192 [MUSIC] 2 00:00:05.192 --> 00:00:07.220 Welcome to the Haskell Weekly podcast. 3 00:00:07.220 --> 00:00:10.420 This show is about Haskell, a purely functional programming language. 4 00:00:10.420 --> 00:00:14.180 I'm your guest, Cameron Gera, and I'm an engineer here at ITProTV. 5 00:00:14.180 --> 00:00:18.040 And with me today is your host, Cody Goodman. 6 00:00:18.040 --> 00:00:21.450 He's one of our senior engineers here at ITProTV. 7 00:00:21.450 --> 00:00:23.890 That just means he's really really old, right Cody? 8 00:00:24.920 --> 00:00:27.200 Yep, I got the gray hairs to prove it. 9 00:00:27.200 --> 00:00:29.030 Programming can do that to you though. 10 00:00:29.030 --> 00:00:30.250 >> Well, I'm glad to have you on the show. 11 00:00:31.460 --> 00:00:35.840 I wanted to take some time today and kind of talk about refactoring Haskell. 12 00:00:35.840 --> 00:00:38.940 I stumbled upon article in Haskell Weekly this week, 13 00:00:38.940 --> 00:00:42.230 that was about continuous improvement with hlint code smells. 14 00:00:42.230 --> 00:00:49.801 So we use hlint here at ITProTV to kind of enhance and kind of unify our code. 15 00:00:49.801 --> 00:00:53.011 What are some things that you have experienced with hlint? 16 00:00:53.011 --> 00:00:57.515 And maybe some of the, I think right now, we're just gonna talk about hlint and 17 00:00:57.515 --> 00:01:00.822 then maybe dive more into refactoring a little bit later. 18 00:01:00.822 --> 00:01:01.902 How's that sound? 19 00:01:01.902 --> 00:01:03.679 >> Yeah, sounds good. 20 00:01:03.679 --> 00:01:08.928 So at ITPro, we've been using hlint ever since I've been here. 21 00:01:08.928 --> 00:01:15.732 Something that is kinda annoying about hlint is it'll bug you about things like, 22 00:01:15.732 --> 00:01:19.685 use parentheses here instead of dollar signs. 23 00:01:19.685 --> 00:01:22.905 And that may conflict with your personal style or 24 00:01:22.905 --> 00:01:27.617 what you think looks more aesthetically pleasing, but it also goes and 25 00:01:27.617 --> 00:01:30.890 it does more complex recommendations too. 26 00:01:30.890 --> 00:01:35.470 Like you can do an eta reduce here, you can use point-free here. 27 00:01:35.470 --> 00:01:40.800 It doesn't go to the extent of using composition, 28 00:01:40.800 --> 00:01:44.430 with composition, where you have three dots next to each other. 29 00:01:44.430 --> 00:01:48.270 But it provides sort of some common sense recommendations. 30 00:01:48.270 --> 00:01:51.910 It leads you to start refactoring code in a uniform way. 31 00:01:51.910 --> 00:01:55.435 And that's something really valuable across a code base. 32 00:01:55.435 --> 00:01:56.025 >> Yeah, I know. 33 00:01:56.025 --> 00:01:58.155 I feel like I have really enjoyed having hlint. 34 00:01:58.155 --> 00:01:59.602 Sometimes it was like, yeah, 35 00:01:59.602 --> 00:02:03.145 I was supposed to take away those parens that I accidentally left. 36 00:02:03.145 --> 00:02:08.176 Those are some of the smaller things it can do, but I also know we use map maybe 37 00:02:08.176 --> 00:02:12.600 and and it's like there's a cat. 38 00:02:12.600 --> 00:02:15.730 Maybe there's a function that does the same thing or I don't know. 39 00:02:15.730 --> 00:02:17.080 It's a cat maybe. 40 00:02:17.080 --> 00:02:19.723 Something like that that does a very similar thing so it just says, 41 00:02:19.723 --> 00:02:21.758 hey, use this function instead of this function. 42 00:02:21.758 --> 00:02:25.155 And it kinda cleans up and minimizes the code, 43 00:02:25.155 --> 00:02:31.163 which is nice to have that going into each file, and having some sense of pattern, 44 00:02:31.163 --> 00:02:35.740 and kind of just cleanliness around the code base, right? 45 00:02:35.740 --> 00:02:41.370 >> Yeah, you could have your Haskell looking like lists, but you have hlint, 46 00:02:41.370 --> 00:02:44.170 then you have some assurance that it's gonna be a lot more clear, and 47 00:02:44.170 --> 00:02:47.150 you can focus on what the code's doing, instead of all that noise. 48 00:02:47.150 --> 00:02:51.590 >> Right, and I think it just kind of gets 49 00:02:51.590 --> 00:02:54.820 you back to coding efficiently, which is really nice. 50 00:02:54.820 --> 00:02:57.558 So okay, so we've got hlint, we use hlint. 51 00:02:57.558 --> 00:03:02.551 We've kind of experienced hlint, and this article kind of talks 52 00:03:02.551 --> 00:03:08.340 about four differently code smells that hlint has helped them with. 53 00:03:08.340 --> 00:03:13.180 One being along functions, another one being functions with many arguments. 54 00:03:13.180 --> 00:03:16.993 We have long type list and then we have lots of imports. 55 00:03:16.993 --> 00:03:21.969 So something that, those things seem to kind of maybe clean up and 56 00:03:21.969 --> 00:03:26.230 minimize this expansive language that Haskell can be. 57 00:03:26.230 --> 00:03:29.568 Kind of minimizes it a little bit to make maybe beginners and 58 00:03:29.568 --> 00:03:34.610 intermediate people kind of be able to come in and understand what's going on. 59 00:03:34.610 --> 00:03:36.212 Have you experienced any of that? 60 00:03:36.212 --> 00:03:39.362 I don't think we've really come across any of those four in our code base that 61 00:03:39.362 --> 00:03:40.840 I've seen. 62 00:03:40.840 --> 00:03:45.093 >> So a lot of imports, especially when you're just trying to hack something 63 00:03:45.093 --> 00:03:47.193 together, you can end up with that. 64 00:03:47.193 --> 00:03:49.496 So you leave a debug.trace or something. 65 00:03:49.496 --> 00:03:52.760 Hlint's gonna catch that and you can remove it. 66 00:03:52.760 --> 00:03:57.097 And then, like you said, for newcomers who are newer to Haskell, they might, 67 00:03:57.097 --> 00:04:00.531 if you're not using qualified imports, they'll be saying, 68 00:04:00.531 --> 00:04:02.630 where's this function to find that? 69 00:04:02.630 --> 00:04:05.590 If you have a ton of imports you're not using, it's going to be really hard for 70 00:04:05.590 --> 00:04:07.070 them to track it down. 71 00:04:07.070 --> 00:04:08.120 >> Yeah, I agree with you, Cody. 72 00:04:08.120 --> 00:04:12.370 I think it does help intermediate, only being in Haskell myself for 73 00:04:12.370 --> 00:04:17.650 a little over a year now, hlint really just kind of helped me 74 00:04:17.650 --> 00:04:21.200 shape my styles when I'm writing Haskell. 75 00:04:21.200 --> 00:04:23.040 I know you've written Haskell a lot longer than I have. 76 00:04:23.040 --> 00:04:26.790 So you probably have a little more frustrations with hlint, because it's 77 00:04:26.790 --> 00:04:31.080 like, I want you to do this when you really prefer to do it maybe another way. 78 00:04:31.080 --> 00:04:35.600 But I think overall, it helps all of us, any of the engineers here at ITProTV come 79 00:04:35.600 --> 00:04:41.350 into our code base and understand, this is what we do, this is the style we have, 80 00:04:41.350 --> 00:04:45.530 this is set across all our files and we know what's what. 81 00:04:45.530 --> 00:04:48.740 And I think that's really, really valuable. 82 00:04:48.740 --> 00:04:53.039 >> Right, and I think something actually useful when you're learning Haskell is to 83 00:04:53.039 --> 00:05:00.190 use hlint, follow all the suggestions and not only follow them but ask why. 84 00:05:00.190 --> 00:05:05.319 Why is it useful to transform this like that, because there's a lot of experienced 85 00:05:05.319 --> 00:05:09.951 developers or at least one who came up with these different rewrite rules. 86 00:05:09.951 --> 00:05:11.741 And they were there for a reason and 87 00:05:11.741 --> 00:05:14.723 they were born out of the experience that developer had. 88 00:05:14.723 --> 00:05:17.630 >> Hm, yeah, which I think is really cool. 89 00:05:17.630 --> 00:05:23.470 One thing that I kind of wish hlint did, which is probably just expecting too much, 90 00:05:23.470 --> 00:05:29.570 is kind of being able to take a function and kind of maybe analyze the way in 91 00:05:29.570 --> 00:05:33.720 which it's being performed, taking, okay what are the inputs, what's the output? 92 00:05:33.720 --> 00:05:36.400 Okay, let's see how they wrote the code to make this work. 93 00:05:36.400 --> 00:05:40.514 I wish it would be able to maybe look at it and say, hey, 94 00:05:40.514 --> 00:05:43.198 you should probably break this up or 95 00:05:43.198 --> 00:05:48.664 you should kind of shift things around because it's too conglomerate. 96 00:05:48.664 --> 00:05:50.368 There's too many- >> Too complex. 97 00:05:50.368 --> 00:05:51.128 >> Too complex, right. 98 00:05:51.128 --> 00:05:53.470 Sorry, conglomerate doesn't really make sense there. 99 00:05:53.470 --> 00:05:55.060 But yeah, too complex. 100 00:05:55.060 --> 00:05:56.670 >> Those functions are starting a business. 101 00:05:57.820 --> 00:06:01.010 >> Jeez. But say we have anonymous functions, 102 00:06:01.010 --> 00:06:03.320 instead of anonymous functions like being, hey, 103 00:06:03.320 --> 00:06:06.680 let's pull this out and those kind of things may be nice. 104 00:06:06.680 --> 00:06:12.090 Because I know this week, you and I both had come across some code in our code 105 00:06:12.090 --> 00:06:17.670 base that was, it worked but it was very thick. 106 00:06:17.670 --> 00:06:19.238 It's hard to fully comprehend. 107 00:06:19.238 --> 00:06:26.916 >> Right, and it actually wasn't even that it was that large either. 108 00:06:26.916 --> 00:06:32.293 It was kind of the, I think the fact that in one place we took advantage of partial 109 00:06:32.293 --> 00:06:37.270 application, where you have to read from right to left to understand it, 110 00:06:37.270 --> 00:06:41.790 and then we had a lambda inside of there which is left to right. 111 00:06:41.790 --> 00:06:44.660 I think that was sort of the thing that made it hard to follow. 112 00:06:44.660 --> 00:06:49.570 And then when you add or reduce that, you get the advantage of 113 00:06:49.570 --> 00:06:55.140 it all being right to left typically, and that's really valuable. 114 00:06:55.140 --> 00:06:58.540 Yeah, so if hlint could have maybe identified that, 115 00:06:58.540 --> 00:07:00.310 we had conflicting styles, right. 116 00:07:00.310 --> 00:07:03.980 And it could have been, hey, let's reevaluate this. 117 00:07:03.980 --> 00:07:07.309 If you want to stick to point free, okay, stick to point free, but 118 00:07:07.309 --> 00:07:09.522 don't be jumping all these things in here. 119 00:07:09.522 --> 00:07:14.513 Don't mix partial application with lambdas, that doesn't make for 120 00:07:14.513 --> 00:07:16.710 easy readability. 121 00:07:16.710 --> 00:07:20.300 So I think, for us, that would kinda be maybe a nice thing to have. 122 00:07:20.300 --> 00:07:25.640 But nice thing is Haskell makes re-factoring pretty easy. 123 00:07:25.640 --> 00:07:28.730 Cuz the problem we had was kind of complex. 124 00:07:28.730 --> 00:07:34.860 We had a document that had a list of, in our case, it was users. 125 00:07:34.860 --> 00:07:39.989 And we wanted to kind of create a map that had one of 126 00:07:39.989 --> 00:07:45.610 those users tied to another value on that document. 127 00:07:45.610 --> 00:07:50.100 And so, say we had ten users, we wanted to have a map with ten entries. 128 00:07:50.100 --> 00:07:53.974 And then we were actually doing that over a list of those documents. 129 00:07:53.974 --> 00:07:57.590 So it was kind of just a lot going on and 130 00:07:57.590 --> 00:08:02.990 so we have some helper functions that we used to kind of group things, 131 00:08:02.990 --> 00:08:06.675 but that didn't really fall into, this category wasn't there. 132 00:08:06.675 --> 00:08:12.491 And so to begin with, I just was like, hey let's, I was working with Jason. 133 00:08:12.491 --> 00:08:15.430 We're like, hey, let's just figure out how to make this work. 134 00:08:15.430 --> 00:08:17.420 We know what the type signature looks like. 135 00:08:17.420 --> 00:08:21.037 Let's just let the compiler happen, but obviously, we put that out for PR. 136 00:08:21.037 --> 00:08:24.505 We're kind of like, this seems rough, there was code smells, right? 137 00:08:24.505 --> 00:08:26.305 There could be something better here. 138 00:08:26.305 --> 00:08:29.225 So I wish, I mean, that would have been kinda cool, 139 00:08:29.225 --> 00:08:32.730 if hlint could have done something like that for us, say, hey, reevaluate. 140 00:08:32.730 --> 00:08:40.055 >> Yeah, and I think it would definitely be possible, although difficult, for 141 00:08:40.055 --> 00:08:46.440 hlint to say, your code is from right to left because of usage of composition here. 142 00:08:46.440 --> 00:08:48.950 Then in this lambda, it is left to right. 143 00:08:48.950 --> 00:08:51.400 I think that's something that could be added. 144 00:08:51.400 --> 00:08:54.420 Then you were kind of hinting at complexity. 145 00:08:54.420 --> 00:08:59.890 And there is an existing calculation to 146 00:08:59.890 --> 00:09:03.850 figure that out, which is called cyclomatic complexity. 147 00:09:03.850 --> 00:09:06.660 And that's pretty interesting. 148 00:09:06.660 --> 00:09:08.260 >> Does hlint have that? 149 00:09:08.260 --> 00:09:09.480 >> I don't think so, 150 00:09:09.480 --> 00:09:13.070 but it'd be interesting to start an issue about that and have a discussion. 151 00:09:13.070 --> 00:09:15.090 >> Yeah, it really would. 152 00:09:15.090 --> 00:09:18.150 Maybe we'll see if that could be put into place. 153 00:09:18.150 --> 00:09:19.860 >> Sure. >> No, but I think that is really awesome. 154 00:09:21.220 --> 00:09:22.833 Don, you were gonna say something [CROSSTALK]. 155 00:09:22.833 --> 00:09:27.875 >> Yeah, yeah, I was gonna say that our specific problem, 156 00:09:27.875 --> 00:09:33.861 what helped make sense of it to me, is that we had that set of users and 157 00:09:33.861 --> 00:09:38.470 then we could easily turn that into a map. 158 00:09:38.470 --> 00:09:43.929 But from that map, we needed to, the map's values 159 00:09:43.929 --> 00:09:48.340 had another list of users and we needed to basically expand the map out. 160 00:09:49.350 --> 00:09:51.755 And that was what kind of made it difficult. 161 00:09:51.755 --> 00:09:55.059 And you might reach for a fold in that situation but 162 00:09:55.059 --> 00:09:59.228 something that's often times easier is making a lot of maps and 163 00:09:59.228 --> 00:10:02.784 unioning them together, which is where we ended up. 164 00:10:02.784 --> 00:10:05.054 >> Good old union with this. 165 00:10:05.054 --> 00:10:09.455 Also the map had, the key was a user ID, but 166 00:10:09.455 --> 00:10:12.886 the value within that was a list. 167 00:10:12.886 --> 00:10:16.665 And so unioning that with semi group allowed us to 168 00:10:16.665 --> 00:10:19.998 get exactly the result we were expecting. 169 00:10:19.998 --> 00:10:23.860 And it cleaned it up because, and we felt like when Jason and 170 00:10:23.860 --> 00:10:27.894 I wrote it we felt like, there could be a better way to do this. 171 00:10:27.894 --> 00:10:33.630 We knew that there was some sort of, just kind of something wrong. 172 00:10:33.630 --> 00:10:35.350 We felt like we could have, 173 00:10:35.350 --> 00:10:39.540 we changed things from set to a list because we deal with a set. 174 00:10:39.540 --> 00:10:44.990 And then we felt like we could have left that as a set and done the operation that 175 00:10:44.990 --> 00:10:48.475 we needed to, and which was what we ended up getting to, which was really wrong. 176 00:10:48.475 --> 00:10:53.120 But in this case hlint didn't necessarily save the day. 177 00:10:53.120 --> 00:10:56.930 But in a lot of other cases it has, I'm very thankful for hlint. 178 00:10:56.930 --> 00:11:02.020 >> I think you're getting at a sort of unique selling proposition of Haskell, 179 00:11:02.020 --> 00:11:07.620 which is that you can not only sort of write the code that just works quickly. 180 00:11:07.620 --> 00:11:10.940 The compiler will help you do it, and then once you write it, 181 00:11:10.940 --> 00:11:12.100 even if you're not proud of it, 182 00:11:12.100 --> 00:11:15.690 you know that when you wanna go refactor it, you have a lot of guarantees. 183 00:11:15.690 --> 00:11:20.510 It's way easier to actually hack the code, get it out there real quick, and 184 00:11:20.510 --> 00:11:24.140 then have the confidence you can come back later without breaking things and 185 00:11:24.140 --> 00:11:25.550 improve upon it. 186 00:11:25.550 --> 00:11:28.520 >> No, and I really value that aspect of Haskell. 187 00:11:28.520 --> 00:11:32.980 I think it's, for us coming from a JavaScript background, 188 00:11:32.980 --> 00:11:37.354 most of us, we kind of dealt with, don't touch that code, 189 00:11:37.354 --> 00:11:40.773 you don't know what the side effects will be. 190 00:11:40.773 --> 00:11:43.220 >> It's a ball of marbles, it's a ball of marbles. 191 00:11:43.220 --> 00:11:48.000 But being able to say, okay, we're in a compiled language that's strongly typed. 192 00:11:48.000 --> 00:11:52.495 That value is just tremendous, because any time you need to come back and 193 00:11:52.495 --> 00:11:54.810 re-factor, you have confidence. 194 00:11:54.810 --> 00:11:59.280 I think that's something that everyone should get on board with, and join and 195 00:11:59.280 --> 00:12:03.920 love Haskell, because it really gives you that confidence and that peace. 196 00:12:03.920 --> 00:12:08.355 I remember, a year, a year and a half ago, when I was writing Java Script, and then 197 00:12:08.355 --> 00:12:14.300 I'd go reopen some old controller or model or action that I wanted to mess with. 198 00:12:14.300 --> 00:12:18.380 And I would just have this sense of stress, of I'm touching something. 199 00:12:18.380 --> 00:12:19.850 Is this gonna work? 200 00:12:19.850 --> 00:12:21.840 Is our test gonna catch this? 201 00:12:21.840 --> 00:12:25.800 Cuz that's all you have in Javascript, you don't necessarily, in user testing. 202 00:12:26.860 --> 00:12:28.100 But that's all you got. 203 00:12:29.180 --> 00:12:33.500 Haskell, we have strongly typed languages which makes that kind of refactor nice and 204 00:12:33.500 --> 00:12:36.900 then we have tooling like hlint and Brittany. 205 00:12:36.900 --> 00:12:42.570 Things that kind of keep us within lines, within guards. 206 00:12:42.570 --> 00:12:45.687 Whereas like JavaScript, it's the wild, wild West. 207 00:12:45.687 --> 00:12:50.225 You have, eslint and stuff like that, but I think, 208 00:12:50.225 --> 00:12:55.190 hlint is just leaps and bounds above eslint or any other. 209 00:12:55.190 --> 00:12:57.466 Well, in my opinion, I could be wrong, and 210 00:12:57.466 --> 00:13:00.980 I'm sure somebody on the internet may want to tell me I'm wrong. 211 00:13:00.980 --> 00:13:06.875 And that's okay, but in my opinion, hlint, for me in Haskell is great. 212 00:13:09.744 --> 00:13:14.201 >> It leads you to, hlint, those different rewrites that it does, 213 00:13:14.201 --> 00:13:18.360 how you can glean understanding of why that does happen. 214 00:13:18.360 --> 00:13:23.540 Just the act of refactoring something can lead you to 215 00:13:23.540 --> 00:13:27.350 understand the nature of the problem you're working on as a whole. 216 00:13:28.530 --> 00:13:33.430 That's an economic factor of deciding to rewrite something that's 217 00:13:33.430 --> 00:13:36.540 a code smell that I think a lot of people don't take into account. 218 00:13:38.040 --> 00:13:39.790 >> Yeah, that's fair. 219 00:13:39.790 --> 00:13:45.610 Well, Cody, I think we've had quite the show here, quite the talk on 220 00:13:45.610 --> 00:13:49.400 hlint and even refactoring a little bit in Haskell and others. 221 00:13:49.400 --> 00:13:50.780 Could be even more to talk about, 222 00:13:50.780 --> 00:13:54.950 but we wanna respect our listener's time and sign off. 223 00:13:54.950 --> 00:13:57.740 So I'll shoot it over to you. 224 00:13:57.740 --> 00:13:59.310 >> All right, thanks, Cameron. 225 00:13:59.310 --> 00:14:01.900 Thanks for being on the show with me today. 226 00:14:01.900 --> 00:14:05.180 Thanks for listening to the Haskell Weekly podcast. 227 00:14:05.180 --> 00:14:07.150 This has been episode eight. 228 00:14:07.150 --> 00:14:13.090 If you like our show, find out more at our website, HaskellWeekly.news. 229 00:14:13.090 --> 00:14:14.480 Thanks again for listening. 230 00:14:14.480 --> 00:14:16.703 I've been your host, Cody Goodman. 231 00:14:16.703 --> 00:14:17.694 We'll see you again next week. 232 00:14:17.694 --> 00:14:20.467 >> Adios.