344: Spinner Armageddon

Steph has an update and a question wrapped into one about the work that is being done to migrate the Test::Unit test over to RSpec. Chris got to do something exciting this week using dry-monads. Success or failure? This episode is brought to you by BuildPulse. Start your 14-day free trial of BuildPulse today. Bartender dry-rb - dry-monads v1.0 - Pattern matching alfred-workflows Raycast ruby-science Inertia.js Remix Become a Sponsor of The Bike Shed! Transcript: AD: Flaky tests take the joy out of programming. You push up some code, wait for the tests to run, and the build fails because of a test that has nothing to do with your change. So you click rebuild, and you wait. Again. And you hope you're lucky enough to get a passing build this time. Flaky tests slow everyone down, break your flow, and make things downright miserable. In a perfect world, tests would only break if there's a legitimate problem that would impact production. They'd fail immediately and consistently, not intermittently. But the world's not perfect, and flaky tests will happen, and you don't have time to fix all of them today. So how do you know where to start? BuildPulse automatically detects and tracks your team's flaky tests. Better still, it pinpoints the ones that are disrupting your team the most. With this list of top offenders, you'll know exactly where to focus your effort for maximum impact on making your builds more stable. In fact, the team at Codecademy was able to identify their flakiest tests with BuildPulse in just a few days. By focusing on those tests first, they reduced their flaky builds by more than 68% in less than a month! And you can do the same because BuildPulse integrates with the tools you're already using. It supports all of the major CI systems, including CircleCI, GitHub Actions, Jenkins, and others. And it analyzes test results for all popular test frameworks and programming languages, like RSpec, Jest, Go, pytest, PHPUnit, and more. So stop letting flaky tests slow you down. Start your 14-day free trial of BuildPulse today. To learn more, visit buildpulse.io/bikeshed. That's buildpulse.io/bikeshed. STEPH: What type of bird is the strongest bird? CHRIS: I don't know. STEPH: A crane. [laughter] STEPH: You're welcome. And on that note, shall we wrap up? CHRIS: Let's wrap up. [laughter] Hello and welcome to another episode of The Bike Shed, a weekly podcast from your friends at thoughtbot about developing great software. I'm Chris Toomey. STEPH: And I'm Steph Viccari. CHRIS: And together, we're here to share a bit of what we've learned along the way. So, Steph, what's new in your world? STEPH: Hey, Chris, I saw a good movie I'd like to tell you about. It was just over the weekend. It's called The Duke, and it's based on a real story. I should ask, have you seen it? Have you heard of this movie called The Duke? CHRIS: I don't think so. STEPH: Okay, cool. It's a true story, and it's based on an individual named Kempton Bunton who then stole a particular portrait, a Goya portrait; if you know your artist, I do not. But he stole a Goya portrait and then essentially held at ransom because he was a big advocate that the BBC News channel should be free for people that are living on a pension or that are war veterans because then they're not able to afford that fee. But then, if you take the BBC channel away from them, it disconnects them from society. And it's a very good movie. I highly recommend it. So I really enjoyed watching that over the weekend. CHRIS: All right. Excellent recommendation. We will, of course, add that to the show notes mostly so that I can find it again later. STEPH: On a more technical note, I have a small update, or it's more of a question. It's an update and a question wrapped into one about the work that is being done to migrate the Test::Unit test over to RSpec. This has been quite a journey that Joël and I have been on for a while now. And we're making progress, but we're realizing that we're spending like 95% of our time in the test setup and porting that over, specifically because we're mapping fixture data over to FactoryBot, and we're just realizing that's really painful. It's taking up a lot of time to do that. And initially, when I realized we were just doing that, we hadn't even really talked about it, but we were moving it over to FactoryBot. I was like, oh, cool. We'll get to delete all these fixtures because there are around 208 files of them. And so that felt like a really good additional accomplishment to migrating the test over. But now that we realize how much time we're spending migrating the data over for that test setup, we've reevaluated, and I shared with Joël in the Slack channel. I was like, crap. I was like, I have a bad idea, and I can't not say it now because it's crossed my mind. And my bad idea was what if we stopped porting over fixtures to FactoryBot and then we just added the fixtures to a directory that RSpec would look so then we can rely on those fixtures? And then that way, we're literally then ideally just copying over from Test::Unit over to RSpec. But it does mean a couple of things. Well, one, it means that we're now running those fixtures at the beginning of RSpec test. We're introducing another pattern of where these tests are already using FactoryBot, but now they have fixtures at the top, and then we won't get to delete the fixtures. So we had a conversation around how to manage and mitigate some of those concerns. And we're still in that exploratory. We're going to test it out and see if this really speeds us up referencing the fixtures. The question that's wrapped up in this is there's something different between how fixtures generate data and how factories generate data. So I've run into this a couple of times now where I moved data over to just call a factory. But then I was hitting these callbacks or after-save-hooks or weird things that were then preventing me from creating the record, even though fixtures was creating them just fine. And then Joël pointed out today that he was running into something similar where there were private methods that were getting called. And there were all sorts of additional code that was getting run with factories versus fixtures. And I don't have an answer. Like, I haven't looked into this. And it's frankly intentional because I was trying hard to not dive into understanding the mechanics. We really want to get through this. But now I'm starting to ponder a little more as to what is different with fixtures and factories? And I liked that factories is running these callbacks; that feels correct. But I'm surprised that fixtures doesn't, or at least that's the experience that I'm having. So there's some funkiness there that I'd like to explore. I'll be honest; I don't know if I'm going to. But if anybody happens to know what that funkiness is or why fixtures and factories are different in that regard, I would be very intrigued because, at some point, I might look into it just because I would like to know. CHRIS: Oh, that is interesting. I have not really worked with fixtures much at all. I've lived a factory life myself, and thus that's where almost all of my experience is. I'm not super surprised if this ends up being the case, like, the idea that fixtures are just some data that gets shoveled into the database directly as opposed to FactoryBot going through the model layer. And so it's sort of like that difference. But I don't know that for certain. That sounds like what this is and makes sense conceptually. But I think this is what you were saying like, that also kind of pushes me more in the direction of factories because it's like, oh, they're now representative. They're using our model layer, where we're defining certain truths. And I don't love callbacks as a mechanism. But if your app has them, then getting data that is representative is useful in tests. Like one of the things I add whenever I'm working with FactoryBot is the FactoryBot lint rake task RSpec thing that basically just says, "Are your factories valid?" which I think is a great baseline to have. Because you may add a migration that adds a default constraint or something like that to the database that suddenly all your factories are invalid, and it's breaking tests, but you don't know it. Like subtly, you change it, and it doesn't actually break a test, but then it's harder later. So that idea of just having more correctness baked in is always nice, especially when it can be automated like that, so definitely a fan of that. But yeah, interested if you do figure out the distinction. I do like your take, though, of like, but also, maybe I just won't figure this out. Maybe this isn't worth figuring it out. Although you were in the interesting spot of, you could just port the fixtures over and then be done and call the larger body of work done. But it's done in sort of a half-complete way, so it's an interesting trade-off space. I'm also interested to hear where you end up on that. STEPH: Yeah, it's a tough trade-off. It's one that we don't feel great about. But then it's also recognizing what's the true value of what we're trying to deliver? And it also comes down to the idea of churn versus complexity. And I feel like we are porting over existing complexity and even adding a smidge, not actual complexity but adding a smidge of indirection in terms that when someone sees this file, they're going to see a mixed-use of fixtures and factories, and that doesn't feel good. And so we've already talked about adding a giant comment above fixtures that just is very honest and says, "Hey, these were ported over. Please don't mimic this. But this is some legacy tests that we have brought over. And we haven't migrated the fixtures over to use factories." And then, in regards to the churn versus complexity, this code isn't likely to get touched like these tests. We really just need them to keep running and keep validating scenarios. But it's not likely that someone's going to come in here and really need to manage these anytime soon. At least, this is what I'm telling myself to make me feel better about it. So there's also that idea of yes, we are porting this over. This is also how they already exist. So if someone did need to manage these tests, then going to Test::Unit, they would have the same experience that they're going to have in RSpec. So that's really the crux of it is that we're not improving that experience. We're just moving it over and then trying to communicate that; yes, we have muddied the waters a little bit by introducing this other pattern. So we're going to find a way to communicate why we've introduced this other pattern, but that way, we can stay focused on actually porting things over to RSpec. As for the factories versus fixtures, I feel like you're onto something in terms of it's just skipping that model layer. And that's why a lot of that functionality isn't getting run. And I do appreciate the accuracy of factories. I'd much rather know is my data representative of real data that can get created in the world? And right now, it feels like some of the fixtures aren't. Like, how they're getting created seemed to bypass really important checks and validations, and that is wrong. That's not what we want to have in our test is, where we're creating data that then the rest of the application can't truly create. But that's another problem for another day. So that's an update on a trade-off that we have made in regards to the testing journey that we are on. What's going on in your world? CHRIS: Well, we got to do something exciting this week. I was working on some code. This is using dry-monads, the dry-rb space. So we have these result objects that we use pretty pervasively throughout the app, and often, we're in a controller. We run one of these command objects. So it's create user, and create user actually encompasses a ton of logic in our app, and that object returns a result. So it's either a success or a failure. And if it's a success, it'll be a success with that new user wrapped up inside of it, or if it's a failure, it's a specific error message. Actually, different structured error messages in different ways, some that would be pushed to the form, some that would be a flash message. There are actually fun, different things that we do there. But in the controller, when we interact with those result objects, typically what we'll do is we'll say result equals create user dot run, (result=createuser.run) and then pass it whatever data it needs. And then on the next line, we'll say results dot either, (results.either), which is a method on these result objects. It's on both the success and failure so you can treat them the same. And then you pass what ends up being a lambda or a stabby proc, or I forget what they are. But one of those sort of inline function type things in Ruby that always feel kind of weird. But you pass one of those, and you actually pass two of them, one for the success case and one for the failure case. And so in the success case, we redirect back with a notice of congratulations, your user was created. Or, in the failure case, we potentially do a flash message of an alert, or we send the errors down, or whatever it ends up being. But it allows us to handle both of those cases. But it's always been syntactically terrible, is how I would describe it. It's, yeah, I'm just going to leave it at that. We are now living in a wonderful, new world. This has been something that I've wanted to try for a while. But I finally realized we're actually on Ruby 2.7, and so thus, we have access to pattern matching in Ruby. So I get to take it for a spin for the first time, realizing that we were already on the correct version. And in particular, dry-monads has a page in their docs specific to how we can take advantage of pattern matching with the result objects that they provide us. There's nothing specific in the library as far as I understand it. This is just them showing a bunch of examples of how one might want to do it if they're working with these result objects. But it's really great because it gives the ability to interact with, you know, success is typically going to be a singular case. There's one success branch to this whole logic, but there are like seven different ways it can fail. And that's the whole idea as to why we use these command objects and the whole Railway Oriented Programming and that whole thing which I have...what is this word? [laughs] I feel like I should know it. It's a positive rant. I have raved; that is how our users kindly pointed that out to us. I have raved about the Railway Oriented Programming that allows us to do. But it's that idea that they're actually, you know, there's one happy path, and there are seven distinct failure modes, seven unhappy paths. And now, using pattern matching, we actually get a really expressive, readable, useful way to destructure each of those distinct failures to work with the particular bits of data that we need. So it was a very happy day, and I got to explore it. This is, again, a feature of Ruby, not a feature of dry-monads. But dry-monads just happens to embrace it and work really well with it. So that was awesome. STEPH: That is awesome. I've seen one or two; I don't know, I've seen a couple of tweets where people are like, yeah, Ruby pattern matching. I haven't found a way to use it. So I'm excited that you just shared a way that you found to use it. I'm also worried what it says about our developer culture that we know the word rant so well, but rave, we always have to reach back into our memory to be like, what's that positive word or something that we like? [laughs] CHRIS: And especially here on The Bike Shed, where we try to gravitate towards the positive. But yeah, it's an interesting point that you make. STEPH: We're a bunch of ranters. It's what we do, pranting ranters. I don't know why we're pranting. [laughs] CHRIS: Because it's that exciting. That's what it is. Actually, there was an interesting thing as we were playing around with the pattern matching code, just poking around in the console session with it, and it prints out a deprecation warning. It's like, warning: this is an experimental feature. Do not use it, be careful. But in the back of my head, I was like, I actually know how this whole thing plays out, Ruby 2.7, and I assure you, it's going to be fine. I have been to the future, at least I'm pretty sure. I think the version that is in Ruby 2.7 did end up getting adopted basically as it stands. And so, I think there is also a setting to turn off that deprecation warning. I haven't done it yet, but I mostly just enjoyed the conversation that I had with this deprecation message of like, listen, I've been to the future, and it's great. Well, it's complicated, but specific to this pattern matching [laughs] in Ruby 3+ versions, it went awesome. And I'm really excited about that future that we now live in. STEPH: I wish we had that for so many more things in our life [laughs] of like, here's a warning, and it's like, no, no, I've seen the future. It's all right. Or you're totally right; I should avoid and back out of this now. CHRIS: If only we could know how the things would play out, you know. But yeah, so pattern matching, very cool. I'll include a link in the show notes to the particular page in the dry-monads docs. But there are also other cool things on the internet. In an unrelated but also cool thing that I found this week, we use Tuple a lot within our organization for pair programming. For anyone who's not familiar with it, it's a really wonderful piece of technology that allows you to pair program pretty seamlessly, better video quality, all of those nice things that we want. But I found there was just the tiniest bit of friction in starting a Tuple call. I know I want to pair with this person. And I have to go up and click on the little menu bar, and then I have to find their name, then I have to click a button. That's just too much. That's not how...I want to live my life at the keyboard. I have a thing called Bartender, which is a little menu bar manager utility app that will collapse down and hide the icons. But it's also got a nice, little hotkey accessible pop-up window that allows me to filter down and open one of the menu bar pop-out menus. But unfortunately, when that happens, the Tuple window isn't interactive at that point. I can't use the arrow keys to go up and down. And so I was like, oh, man, I wonder if there's like an Alfred workflow for this. And it turns out indeed there is actually managed by the kind folks at Tuple themselves. So I was able to find that, install it; it's great. I have it now. I can use that. So that was a nice little upgrade to my workflow. I can just type like TC space and then start typing out the person's name, and then hit enter, and it will start a call immediately. And it doesn't actually make me more productive, but it makes me happier. And some days, that's what matters. STEPH: That's always so impressive to me when that happens where you're like, oh, I need a thing. And then you went through the saga that you just went through. And then the people who manage the application have already gotten there ahead of you, and they're like, don't worry, we've created this for you. That's one of those just beautiful moments of like, wow, y'all have really thought this through on a bunch of different levels and got there before me. CHRIS: It's somewhat unsurprising in this case because it's a very developer-centric organization, and Ben's background being a thoughtbot developer and Alfred user, I'm almost certain. Although I've seen folks talking about Raycast, which is the new hotness on the quick launcher world. I started eons ago in Quicksilver, and then I moved to Alfred, I don't know, ten years ago. I don't know what time it is anymore. But I've been in Alfred land for a while, but Raycast seems very cool. Just as an aside, I have not allowed myself... [laughs] this is another one of those like; I do not have permission to go explore this new tool yet because I don't think it will actually make me more productive, although it could make me happier. So... STEPH: I haven't heard of that one, Raycast. I'm literally adding it to the show notes right now as a way so you can find The Duke later, and I can find Raycast later [chuckles] and take a look at it and check it out. Although I really haven't embraced the whole Alfred workflow. I've seen people really enjoy it and just rave about it and how wonderful it is. But I haven't really leaned into that part of the world; I don't know why. I haven't set any hard and fast rules for myself where I can't play around with these technologies, but I haven't taken the time to do it either. CHRIS: You've also not found yourself writing thousands of lines of Vimscript because you thought that was a good idea. So you don't need as many guardrails it would seem. That's my guess. STEPH: This is true. CHRIS: Whereas I need to be intentional [laughs] with how I structure my interaction with my dev tools. STEPH: Instead, I'm just porting over fixtures from one place to another. [laughs] That's the weird space that I'm living in instead. [laughs] CHRIS: But you're getting paid for that. No one paid me for the Vimscript I wrote. [laughter] STEPH: That's fair. Speaking around process-y things, there's something that's been on my mind that Valeria, another thoughtboter, suggested around how we structure our meetings and the default timing that we have for meetings. So Thursdays are my team-focused day. And it's the day where I have a lot of one on ones. And I realized that I've scheduled them back to back, which is problematic because then I have zero break in between them, which I'm less concerned about that because then I can go for an hour or something and not have a break. And I'm not worried about that part. But it does mean that if one of those discussions happens to go over just even for like two or three minutes, then it means that someone else is waiting for me in those two to three minutes. And that feels unacceptable to me. So Valeria brought up a really good idea where I think it's only with the Google Meet paid version. I could be wrong there. But I think with the paid version of it that then you can set the new default for how long a meeting is going to last. So instead of having it default to 30 minutes, have it default to 25 minutes. So then, that way, you do have that five-minute buffer. So if you do go over just like two or three minutes with someone, you've still got like two minutes to then hop to the next call, and nobody's waiting for you. Or if you want those five minutes to then grab some water or something like that. So we haven't implemented it just yet because then there's discussion around is this a new practice that we want everybody to move to? Because I mean, if just one person does it, it doesn't work. You really need everybody to buy into the concept of we're now defaulting to 25 versus 30-minute meetings. So I'll have to let you know how that goes. But I'm intrigued to try it out because I think that would be very helpful for me. Although there's a part of me that then feels bad because it's like, well, if I have 30 minutes to chat with somebody, but now I'm reducing it to 25 minutes each time, I didn't love that I'm taking time away from our discussion. But that still feels like a better outcome than making somebody wait for three to five minutes if something else goes over. So have you ever run into something like that? How do you manage back-to-back meetings? Do you intentionally schedule a break in between or? CHRIS: I do try to give myself some buffer time. I stack meetings but not so much so that they're just back to back. So I'll stack them like Wednesdays are a meeting-heavy day for me. That's intentional just to be like, all right, I know that my day is going to get chopped up. So let's just really lean into that, chop the heck out of Wednesday afternoons, and then the rest of the week can hopefully have slightly longer deep work-type sessions. And, yeah, in general, I try and have like a little gap in between them. But often what I'll do for that is I'll stagger the start of the next meeting to be rather than on the hour or the half-hour, I start it on the 15th minute. And so then it's sort of I now have these little 15-minute gaps in my workflow, which is enough time to do one or two small things or to go get a drink or whatever it is or if things do run over. Like, again, I feel what you're saying of like, I don't necessarily want to constrain a meeting. Or I also don't necessarily want to go into the habit of often over-running. I think it's good to be intentional. Start meetings on time, end meetings on time. If there's a great conversation that's happening, maybe there's another follow-up meeting that should happen or something like that. But for as nonsensical of a human as I believe myself to be, I am rather rigid about meetings. I try very hard to be on time. I try very hard to wrap them up on time to make sure I go to the next one. And so with that, the 15-minute staggering is what I've found works for me. STEPH: Yeah, that makes sense. One-on-ones feels special to me because I wholeheartedly agree with being very diligent about like, hey, this is our meeting time. Let's do a time check. Someone says that at the end, and then that way, everybody can move on. But one on ones are, there's more open discussion space, and I hate cutting people off, especially because it might not be until the last 15 minutes that you really got into the meat of the conversation. Or you really got somewhere that's a little bit more personal or things that you want to talk about. So if someone's like, "Yeah, let me tell you about my life goals," and you're like, "Oh, no, wait, sorry. We're out of time." That feels terrible and tragic to do. So I struggle with that part of it. CHRIS: I will say actually, on that note, I'm now thinking through, but I believe this to be true. Everyone that reports to me I have a 45-minute one-on-one with, and then my CEO I set up the one-on-one. So I also made that one a 45-minute one-on-one. And that has worked out really well. Typically, I try and structure it and reiterate this from time to time of, like, hey, this is your space, not mine. So let's have whatever conversation fits in here. And it's fine if we don't need to use the whole time, but I want to make sure that we have it and that we protect it. Because I often find much like retro, I don't know; I think everything's fine. And then suddenly the conversation starts, and you're like, you know what? Actually, I'm really concerned now that you mentioned it. And you need that sort of empty space that then the reality sort of pop up into. And so with one on one, I try and make sure that there is that space, but I'm fine with being like, we can cut this short. We can move on from one-on-one topics to more of status updates; let's talk about the work. But I want to make sure that we lead with is there anything deeper, any concerns, anything you want to talk through? And sort of having the space and time for that. STEPH: I like that. And I also think it speaks more directly to the problem I'm having because I'm saying that we keep running over a couple of minutes, and so someone else is waiting. So rather than shorten it, which is where I'm already feeling some pain...although I still think that's a good idea to have a default of 25-minute meetings so then that way, there is a break versus the full 30. So if people want to have back-to-back meetings, they still have a little bit of time in between. But for one on ones specifically, upping it to 45 minutes feels nice because then you've got that 15-minute buffer likely. I mean, maybe you schedule a meeting, but, I don't know, that's funky. But likely, you've got a 15-minute buffer until your next one. And then that's also an area that I feel comfortable in sharing with folks and saying, "Hey, I've booked this whole 45 minutes. But if we don't need the whole time, that's fine." I'm comfortable saying, "Hey, we can end early, and you can get more of your time back to focus on some other areas." It's more the cutting someone off when they're talking because I have to hop to the next thing. I absolutely hate that feeling. So thanks, I think I'll give that a go. I think I'll try actually bumping it up to 45 minutes, presuming that other people like that strategy too, since they're opting in [laughs] to the 45 minutes structure. But that sounds like a nice solution. CHRIS: Well yeah, happy to share it. Actually, one interesting thing that I'm realizing, having been a manager at thoughtbot and then now being a manager within Sagewell, the nature of the interactions are very different. With thoughtbot, I was often on other projects. I was not working with my team day to day in any real capacity. So it was once every two weeks, I would have this moment to reconnect with them. And there was some amount of just catching up. Ideally, not like status update, low-level sort of thing, but sort of just like hey, what have you been working on? What have you been struggling with? What have you been enjoying? There was more like I needed bigger space, I would say for that, or it's not surprising to me that you're bumping into 30 minutes not being quite long enough. Whereas regularly, in the one on ones that I have now, we end up cutting them short or shifting out of true one-on-one mode into more general conversation and chatting about Raycast or other tools or whatever it is because we are working together daily. And we're pairing very regularly, and we're all on the same project and all sorts of in sync and know what's going on. And we're having retro together. We have plenty of places to have the conversation. So the one-on-one again, still, I keep the same cadence and the same time structure just because I want to make sure we have the space for any day that we really need that. But in general, we don't. Whereas when I was at thoughtbot, it was all the more necessary. And I think for folks listening; I could imagine if you're in a team lead position and if you're working very closely with folks, then you may be on the one side of things versus if you're a little bit more at a distance from the work that they're doing day to day. That's probably an interesting question to ask, and think about how you want to structure it. STEPH: Yeah, I think that's an excellent point. Because you're right; I don't see these individuals. We may not have really gotten to interact, except for our daily syncs outside of that. So then yeah, there's always like a good first 10 minutes of where we're just chatting about life and catching up on how things are going before then we dive into some other things. So I think that's a really good point. Cool, solving management problems on the mic. I dig it. In slightly different news, I've joined a book club, which I'm excited about. This book club is about Ruby. It's specifically reading the book Ruby Science, which is a book that was written and published by thoughtbot. And it requires zero homework, which is my favorite type of book club. Because I have found I always want to be part of book clubs. I'm always interested in them, but then I'm not great at budgeting the time to make sure I read everything I'm supposed to read. And so then it comes time for folks to get together. And I'm like, well, I didn't do my homework, so I can't join it. But for this one, it's being led by Joël, and the goal is that you don't have to do the homework. And they're just really short sections. So whoever's in charge of leading that particular session of the book club they're going to provide an overview of what's covered in whatever the reading material that we're supposed to read, whatever topic we're covering that day. They're going to provide an overview of it, an example of it, so then we can all talk about it together. So if you read it, that's wonderful. You're a bit ahead and could even join the meeting like five minutes late. Or, if you haven't read it, then you could join and then get that update. So I'm very excited about it. And this was one of those books that I'd forgotten that thoughtbot had written, and it's one that I've never read. And it's public for anybody that's interested in it. So to cover a little bit of details about it, so it talks about code smells, ways to refactor code, and then also common patterns that you can use to solve some issues. So there's a lot of really just great content that's in it. And I'll be sure to include a link in the show notes for anyone else that's interested. CHRIS: And again, to reiterate, this book is free at this point. Previously, in the past, it was available for purchase. But at one point a number of years ago, thoughtbot set all of the books free. And so now that along with a handful of other books like...what's Edward's DNS book? Domain Name Sanity, I believe, is Edward's book name that Edward Loveall wrote when he was not a thoughtboter, [laughs] and then later joined as a thoughtboter, and then we made the book free. But on the specific topic of Ruby Science, that is a book that I will never forget. And the reason I will never forget it is that book was written by the one and only CTO Joe Ferris, who is an incredibly talented developer. And when I was interviewing with thoughtbot, I got down to the final day, which is a pairing session. You do a morning pairing session with one thoughtbot developer, and you do an afternoon pairing session with another thoughtbot developer. So in the morning, I was working with someone on actually a patch to Rails which was pretty cool. I'd never really done that, so that was exciting. And that went fine with the exception that I kept turning on Caps Lock on their keyboard because I was used to Caps Lock being CTRL, and then Vim was going real weird for me. But otherwise, that went really well. But then, in the afternoon, I was paired with the one and only CTO Joe Ferris, who was writing the book Ruby Science at that time. And the nature of the book is like, here's a code sample, and then here's that code sample improved, just a lot of sort of side-by-side comparisons of code. And I forget the exact way that this went, but I just remember being terrified because Joe would put some code up on the screen and be like, "What do you think?" And I was like, oh, is this the good code or the bad code? I feel like I should know. I do not know. I'm not sure. It worked out fine, I guess. I made it through. But I just remember being so terrified at that point. I was just like, oh no, this is how it ends for me. It's been a good run. STEPH: [laughs] CHRIS: I made it this far. I would have loved to work for this nice thoughtbot company, but here we are. But yeah, I made it through. [laughs] STEPH: There are so many layers to that too where it's like, well if I say it's terrible, are you going to be offended? Like, how's this going to go for me if I speak my truths? Or what am I going to miss? Yeah, that seems very interesting (I kind of like it.) but also a terrifying pairing session. CHRIS: I think it went well because I think the code...I'd been following thoughtbot's work, and I knew who Joe was and had heard him on podcasts and things. And I kind of knew roughly where things were, and I was like, that code looks messy. And so I think I mostly got it right, but just the openness of the question of like, what do you think? I was like, oh God. [laughs] So yeah, that book will always be in my memories, is how I would describe it. STEPH: Well, I'm glad it worked out so we could be here today recording a podcast together. [laughs] CHRIS: Recording a podcast together. Now that I say all that, though, it's been a long time since I've read the book. So maybe I'll take a revisit. And definitely interested to hear more about your book club and how that goes. But shifting ever so slightly (I don't have a lot to say on this topic.) but there's a new framework technology thing out there that has caught my attention. And this hasn't happened for a while, so it's kind of novel for me. So I tend to try and keep my eye on where is the sort of trend of web development going? And I found Inertia a while ago, and I've been very, very happy with that as sort of this is the default answer as to how I build websites. To be clear, Inertia is still the answer as to how I build websites. I love Inertia. I love what it represents. But I'm seeing some stuff that's really interesting that is different. Specifically, Remix.run is the thing that I'm seeing. I mentioned it, I think, in the last episode talking about there was some stuff that they were doing with data loading and async versus synchronous, and do you wait on it or? They had built some really nice levers and trade-offs into the framework. And there's a really great talk that Ryan Florence, one of the creators of Remix.run, gave about that and showed what they were building. I've been exploring it a little bit more in-depth now. And there is some really, really interesting stuff in Remix. In particular, it's a meta-framework, I think, is the nonsense phrase that we use to describe it. But it's built on top of React. That won't be true for forever. I think it's actually they would say it's more built on top of React Router. But it is very similar to Next.js for folks that have seen that. But it's got a little bit more thought around data loading. How do we change data? How do we revalidate data after? There's a ton of stuff that, having worked in many React client-side API-heavy apps that there's so much pain, cache invalidation. How do you think about the cache? When do you fetch from the network? How do you avoid showing 19 different loading spinners on the page? And Remix as a framework has some really, I think, robust and well-thought-out answers to a lot of that. So I am super-duper intrigued by what they're doing over there. There's a particular video that I think shows off what Remix represents really well. It's Ryan Florence, that same individual, the creator of Remix, building just a newsletter signup page. But he goes through like, let's start from the bare bones, simplest thing. It's just an input, and a form submits to the server. That's it. And so we're starting from web 2.0, long, long ago, sort of ideas, and then he gradually enhances it with animations and transitions and error states. And even at the end, goes through an accessibility audit using the screen reader to say, "Look, Remix helps you get really close because you're just using web fundamentals." But then goes a couple of steps further and actually makes it work really, really well for a screen reader. And, yeah, overall, I'm just super impressed by the project, really, really intrigued by the work that they're doing. And frankly, I see a couple of different projects that are sort of in this space. So yeah, again, very early but excited. STEPH: On their website...I'm checking it out as you're walking me through it, and on their website, they have "Say goodbye to Spinnageddon." And that's very cute. [laughs] CHRIS: There's some fundamental stuff that I think we've just kind of as a web community, we made some trade-offs that I personally really don't like. And that idea of just spinners everywhere just sending down a ball of application logic and a giant JavaScript file turning it on on someone's computer. And then immediately, it has to fetch back to the server. There are just trade-offs there that are not great. I love that Remix is sort of flipping that around. I will say, just to sort of couch the excitement that I'm expressing right now, that Remix exists in a certain place. It helps with building complex UIs. But it doesn't have anything in the data layer. So you have to bring your own data layer and figure out what that means. We have ActiveRecord within Rails, and it's deeply integrated. And so you would need to bring a Prisma or some other database connection or whatever it is. And it also doesn't have more sort of full-featured framework things. Like with Rails, it's very easy to get started with a background job system. Remix has no answer to that because they're like, no, no, this is what we're doing over here. But similarly, security is probably the one that concerns me the most. There's an open conversation in their discussion portal about CSRF protection and a back and forth of whether or not Remix should have that out of the box or not. And there are trade-offs because there are different adapters that you can use for auth. And each would require their own CSRF mitigation. But to me, that is the sort of thing that I would want a framework to have. Or I'd be interested in a framework that continues to build on top of Remix that adds in background jobs and databases and all that kind of stuff as a complete solution, something more akin to a Rails or a Laravel where it's like, here we go. This is everything. But again, having some of these more advanced concepts and patterns to build really, really delightful UIs without having to change out the fundamental way that you're building things. STEPH: Interesting. Yeah, I think you've answered a couple of questions that I had about it. I am curious as to how it fits into your current tech stack. So you've mentioned that you're excited and that it's helpful. But given that you already have Rails, and Inertia, and Svelte, does it plug and play with the other libraries or the other frameworks that you have? Are you going to have to replace something to then take advantage of Remix? What does that roadmap look like? CHRIS: Oh yeah, I don't expect to be using Remix anytime soon. I'm just keeping an eye on it. I think it would be a pretty fundamental shift because it ends up being the server layer. So it would replace Rails. It would replace the Inertia within the stack that I'm using. This is why as I started, I was like, Inertia is still my answer. Because Inertia integrates really well with Rails and allows me to do the sort of it's not progressive enhancement, but it's like, I want fancy UI, and I don't want to give up on Rails. And so, Inertia is a great answer for that. Remix does not quite fit in the same way. Remix will own all of the request-response lifecycle. And so, if I were to use it, I would need to build out the rest of that myself. So I would need to figure out the data layer. I would need to figure out other things. I wouldn't be using Rails. I'm sure there's a way to shoehorn the technologies together, but I think it sort of architecturally would be misaligned. And so my sense is that folks out there are building...they're sort of piecing together parts of the stack to fill out the rest. And Remix is a really fantastic controller and view from their down experience and routing layer. So it's routing, controller, view I would say Remix has a really great answer to, but it doesn't have as much of the other stuff. Whereas in my case, Inertia and Rails come together and give me a great answer to the whole story. STEPH: Got it. Okay, that's super helpful. CHRIS: But yeah, again, I'm in very much the exploratory phase. I'm super intrigued by a lot of what I've seen of it and also just sort of the mindset, the ethos of the project as it were. That sounds fancy as I say it, but it's what I mean. I think they want to build from web fundamentals and then enhance the experience on top of that, and I think that's a really great way to go. It means that links will work. It means that routing and URLs will work by default. It means that you won't have loading spinner Armageddon, and these are core fundamentals that I believe make for good websites and web applications. So super interested to see where they go with it. But again, for me, I'm still very much in the Rails Inertia camp. Certainly, I mean, I've built Sagewell on top of it, so I'm going to be hanging out with it for a while, but also, it would still be my answer if I were starting something new right now. I'm just really intrigued by there's a new example out there in the world, this Remix thing that's pushing the envelope in a way that I think is really great. But with that, my now…what was that? My second or my third rave? Also called the positive rant, as we call it. But yeah, I think on that note, what do you think? Should we wrap up? STEPH: Let's wrap up. CHRIS: The show notes for this episode can be found at bikeshed.fm. STEPH: This show is produced and edited by Mandy Moore. CHRIS: If you enjoyed listening, one really easy way to support the show is to leave us a quick rating or even a review on iTunes, as it really helps other folks find the show. STEPH: If you have any feedback for this or any of our other episodes, you can reach us at @_bikeshed or reach me on Twitter @SViccari. CHRIS: And I'm @christoomey. STEPH: Or you can reach us at [email protected] via email. CHRIS: Thanks so much for listening to The Bike Shed, and we'll see you next week. ALL: Byeeeeeeee!!!!!!!!! ANNOUNCER: This podcast was brought to you by thoughtbot. thoughtbot is your expert design and development partner. Let's make your product and team a success.Sponsored By:BuildPulse: Flaky tests take the joy out of programming. You push up some code, wait for the tests to run, and the build fails because of a test that has nothing to do with your change. So you click rebuild and you wait. Again. And you hope you're lucky enough to get a passing build this time. Flaky tests slow everyone down, break your flow and make things downright miserable. In a perfect world, tests would only break if there's a legitimate problem that would impact production. They'd fail immediately and consistently, not intermittently. But the world's not perfect, and flaky tests will happen, and you don't have time to fix them all today. So how do you know where to start? BuildPulse automatically detects and tracks your team's flaky tests. Better still, it pinpoints the ones that are disrupting your team the most. With this list of top offenders, you'll know exactly where to focus your effort for maximum impact on making your builds more stable. In fact, the team at Codecademy was able to identify their flakiest tests with BuildPulse in just a few days. By focusing on those tests first, they reduced their flaky builds by more than 68% in less than a month! And you can do the same because BuildPulse integrates with the tools you're already using. It supports all the major CI systems, including CircleCI, GitHub Actions, Jenkins, and others. And it analyzes test results for all popular test frameworks and programming languages, like RSpec, Jest, Go, pytest, PHPUnit, and more. So stop letting flaky tests slow you down. Start your 14-day free trial of BuildPulse today. To learn more, visit buildpulse.io/bikeshed. That's buildpulse.io/bikeshed.Support The Bike Shed