339: What About Pictures?
The Bike Shed - Een podcast door thoughtbot - Dinsdagen
Categorieën:
Steph has a baby update and thoughts on movies, plus a question for Chris related to migrating Test Unit tests to RSpec. Chris watched a video from Google I/O where Chrome devs talked about a new feature called Page Transitions. He's also been working with a tool called Customer.io, an omnichannel communication whiz-bang adventure! Page transitions Overview Using yield_self for composable ActiveRecord relations A Case for Query Objects in Rails Customer.io Turning the database inside-out with Apache Samza | Confluent Datomic About CRDTs • Conflict-free Replicated Data Types Apache Kafka Resilient Management | A book for new managers in tech Mixpanel: Product Analytics for Mobile, Web, & More Become a Sponsor of The Bike Shed! Transcript: CHRIS: Golden roads are golden. Okay, everybody's got golden roads. You have golden roads, yes? That is what we're -- STEPH: Oh, I have golden roads, yes. [laughter] CHRIS: You might should inform that you've got golden roads, yeah. STEPH: Yeah, I don't know if I say might should as much but might could. I have been called out for that one a lot; I might could do that. CHRIS: [laughs] STEPH: That one just feels so natural to me than normal. Anytime someone calls it out, I'm like, yeah, what about it? [laughter] CHRIS: Do you want to fight? STEPH: Yeah, are we going to fight? CHRIS: I might could fight you. STEPH: I might could. I might should. [laughter] CHRIS: 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 have a couple of fun updates. I have a baby Viccari update, so little baby weighs about two pounds now, two pounds. I'm 25 weeks along. So not that I actually know the exact weight, I'm using all those apps that estimate based on how far along you are, so around two pounds, which is novel. Oh, and then the other thing I'm excited to tell you about...I'm not sure how I should feel that I just got more excited about this other thing. I'm very excited about baby Viccari. But the other thing is there's a new Jurassic Park movie coming out, and I'm very excited. I think it's June 10th is when it comes out. And given how much we have sung that theme song to each other and make references to what a clever girl, I needed to share that with you. Maybe you already know, maybe you're already in the loop, but if you don't, it's coming. CHRIS: Yeah, the internet likes to yell things like that. Have you watched all of the most recent ones? There are like two, and I think this will be the third in the revisiting or whatever, the Jurassic World version or something like that. But have you watched the others? STEPH: I haven't seen all of them. So I've, of course, seen the first one. I saw the one that Chris Pratt was in, and now he's in the latest one. But I think I've missed...maybe there's like two in the middle there. I have not watched those. CHRIS: There are three in the original trilogy, and then there are three now in the new trilogy, which now it's ending, and they got everybody. STEPH: Oh, I'm behind. CHRIS: They got people from the first one, and they got the people from the second trilogy. They just got everybody, and that's exciting. You know, it's that thing where they tap into nostalgia, and they take advantage of us via it. But I'm fine. I'm here for it. STEPH: I'm here for it, especially for Jurassic Park. But then there's also a new Top Gun movie coming out, which, I'll be honest, I'm totally going to watch. But I really didn't remember the first one. I don't know that I've really ever watched the first Top Gun. So Tim, my partner, and I watched that recently, and it's such a bad movie. I'm going to say it; [laughs] it's a bad movie. CHRIS: I mean, I don't want to disagree, but the volleyball scene, come on, come on, the volleyball scene. [laughter] STEPH: I mean, I totally had a good time watching the movie. But the one part that I finally kept complaining about is because every time they showed the lead female character, I can't think of her character name or the actress's name, but they kept playing that song, Take My Breath Away. And I was like, can we just get past the song? [laughs] Because if you go back and watch that movie, I swear they play it like six different times. It was a lot. It was too much. So I moved it into bad movie category but bad movie totally worth watching, whatever category that is. CHRIS: Now I kind of want to revisit it. I feel like the drinking game writes itself. But at a minimum, anytime Take My Breath Away plays, yeah. Well, all right, good to know. [laughs] STEPH: Well, if you do that, let me know how many shots or beers you drink because I think it will be a fair amount. I think it will be more than five. CHRIS: Yeah, it involves a delicate calibration to get that right. I don't think it's the sort of thing you just freehand. It writes itself but also, you want someone who's tried it before you so that you're not like, oh no, it's every time they show a jet. That was too many. You can't drink that much while watching this movie. STEPH: Yeah, that would be death by Top Gun. CHRIS: But not the normal way, the different, indirect death by Top Gun. STEPH: I don't know what the normal way is. [laughs] CHRIS: Like getting shot down by a Top Gun pilot. [laughter] STEPH: Yeah, that makes sense. [laughs] CHRIS: You know, the dogfighting in the plane. STEPH: The actual, yeah, going to war away. Just sitting on your couch and you drink too much poison away, yeah, that one. All right, that got weird. Moving on, [laughs] there's a new Jurassic Park movie. We're going to land on that note. And in the more technical world, I've got a couple of things on my mind. One of them is I have a question for you. I'm very excited to run this by you because I could use a friend in helping me decide what to do. So I am still on that journey where I am migrating Test::Unit test over to RSpec. And as I'm going through, it's going pretty well, but it's a little complicated because some of the Test::Unit tests have different setup than, say, the RSpec do. They might run different scripts beforehand where they're loading data. That's perhaps a different topic, but that's happening. And so that has changed a few things. But then overall, I've just been really just porting everything over, like, hey, if it exists in the Test::Unit, let's just bring it to RSpec, and then I'm going to change these asserts to expects and really not make any changes from there. But as I'm doing that, I'm seeing areas that I want to improve and things that I want to clear up, even if it's just extracting a variable name. Or, as I'm moving some of these over in Test::Unit, it's not clear to me exactly what the test is about. Like, it looks more like a method name in the way that the test is being described, but the actual behavior isn't clear to me as if I were writing this in RSpec, I think it would have more of a clear description. Maybe that's not specific to the actual testing framework. That might just be how these tests are set up. But I'm at that point where I'm questioning should I keep going in terms of where I am just copying everything over from Test::Unit and then moving it over to RSpec? Because ultimately, that is the goal, to migrate over. Or should I also include some time to then go back and clean up and try to add some clarity, maybe extract some variable names, see if I can reduce some lets, maybe even reduce some of the test helpers that I'm bringing over? How much cleanup should be involved, zero, lots? I don't know. I don't know what that...[laughs] I'm sure there's a middle ground in there somewhere. But I'm having trouble discerning for myself what's the right amount because this feels like one of those areas where if I don't do any cleanup, I'm not coming back to it, like, that's just the truth. So it's either now, or I have no idea when and maybe never. CHRIS: I'll be honest, the first thing that came to mind in this most recent time that you mentioned this is, did we consider just deleting these tests entirely? Is that on...like, there are very few of them, right? Like, are they even providing enough value? So that was question one, which let me pause to see what your thoughts there were. [chuckles] STEPH: I don't know if we specifically talked about that on the mic, but yes, that has been considered. And the team that owns those tests has said, "No, please don't delete them. We do get value from them." So we can port them over to RSpec, but we don't have time to port them over to RSpec. So we just need to keep letting them go on. But yet, not porting them conflicts with my goal of then trying to speed up CI. And so it'd be nice to collapse these Test::Unit tests over to RSpec because then that would bring our CI build down by several meaningful minutes. And also, it would reduce some of the complexity in the CI setup. CHRIS: Gotcha. Okay, so now, having set that aside, I always ask the first question of like, can you just put Derek Prior’s phone number on the webpage and call it an app? Is that the MVP of this app? No? Okay, all right, we have to build more. But yeah, I think to answer it and in a general way of trying to answer a broader set of questions here... I think this falls into a category of like if you find yourself having to move around some code, if that code is just comfortably running and the main thing you need to do is just to get it ported over to RSpec, I would probably do as little other work as possible. With the one consideration that if you find yourself needing to deeply load up the context of these tests like actually understand them in order to do the porting, then I would probably take advantage of that context because it's hard to get your head into a given piece of code, test or otherwise. And so if you're in there and you're like, well, now that I'm here, I can definitely see that we could rearrange some stuff and just definitively make it better, if you get to that place, I would consider it. But if this ends up being mostly a pretty rote transformation like you said, asserts become expects, and lets get switched around, you know, that sort of stuff, if it's a very mechanical process of getting done, I would probably say very minimal. But again, if there is that, like, you know what? I had to understand the test in order to port them anyway, so while I'm here, let me take advantage of that, that's probably the thing that I would consider. But if not that, then I would say even though it's messy and whatnot and your inclination would be to clean it, I would say leave it roughly as is. That's my guess or how I would approach it. STEPH: Yeah, I love that. I love how you pointed out, like, did you build up the context? Because you're right, in a lot of these test cases, I'm not, or I'm trying really hard to not build up context. I'm trying very hard to just move them over and, if I have to, mainly to find test descriptions. That's the main area I'm struggling to...how can I more explicitly state what this test does so the next person reading this will have more comprehension than I do? But otherwise, I'm trying hard to not have any real context around it. And that's such a good point because that's often...when someone else is in the middle of something, and they're deciding whether to include that cleanup or refactor or improvement, one of my suggestions is like, hey, we've got the context now. Let's go with it. But if you've built up very little context, then that's not a really good catalyst or reason to then dig in deeper and apply that cleanup. That's super helpful. Thank you. That will help reinforce what I'm going to do, which is exactly let's migrate RSpec and not worry about cleanup, which feels terrible; I'm just going to say that into the world. But it also feels like the right thing to do. CHRIS: Well, I'm happy to have helped. And I share the like, and it feels terrible. I want to do the right thing, but sometimes you got to pick a battle sort of thing. STEPH: Cool. Well, that's a huge help to me. What's going on in your world? CHRIS: What's going on in my world? I watched a great video the other day from the Google I/O. I think it's an event; I'm not actually sure, conference or something like that. But it was some Google Chrome developers talking about a new feature that's coming to the platform called Page Transitions. And I've kept an eye on this for a while, but it seems like it's more real. Like, I think they put out an RFC or an initial sort of set of ideas a while back. And the web community was like, "Oh, that's not going to work out so well." So they went back to the drawing board, revisited. I've actually implemented in Chrome Canary a version of the API. And then, in the video that I watched, which we'll include a show notes link to, they demoed the functionality of the Page Transitions API and showed what you can do. And it's super cool. It allows for the sort of animations that you see in a lot of native mobile apps where you're looking at a ListView, you click on one of the items, and it grows to fill the whole screen. And now you're on the detail screen for that item that you were looking at. But there was this very continuous animated transition that allows you to keep context in your head and all of those sorts of nice things. And this just really helps to bridge that gap between, like, the web often lags behind the native mobile platforms in terms of the experiences that we can build. So it was really interesting to see what they've been able to pull off. The demo is a pretty short video, but it shows a couple of different variations of what you can build with it. And I was like, yeah, these look like cool native app transitions, really nifty. One thing that's very interesting is the actual implementation of this. So it's like you have one version of the page, and then you transition to a new version of the page, and in doing so, you want to animate between them. And the way that they do it is they have the first version of the page. They take a screenshot of it like the browser engine takes a screenshot of it. And then they put that picture on top of the actual browser page. Then they do the same thing with the next version of the page that they're going to transition to. And then they crossfade, like, change the opacity and size and whatnot between the two different images, and then you're there. And in the back of my mind, I'm like, I'm sorry, what now? You did which? I'm like, is this the genius solution that actually makes this work and is performant? And I wonder if there are trade-offs. Like, do you lose interactivity between those because you've got some images that are just on the screen? And what is that like? But as they were going through it, I was just like, wait, I'm sorry, you did what? This is either the best idea I've ever heard, or I'm not so sure about this. STEPH: That's fascinating. You had me with the first part in terms of they take a screenshot of the page that you're leaving. I'm like, yeah, that's a great idea. And then talking about taking a picture of the other page because then you have to load it but not show it to the user that it's loaded. And then take a picture of it, and then show them the picture of the loaded page. But then actually, like you said, then crossfade and then bring in the real functionality. I am...what am I? [laughter] CHRIS: What am I actually? STEPH: [laughs] What am I? I'm shocked. I'm surprised that that is so performant. Because yeah, I also wouldn't have thought of that, or I would have immediately have thought like, there's no way that's going to be performant enough. But that's fascinating. CHRIS: For me, performance seems more manageable, but it's the like, what are you trading off for that? Because that sounds like a hack. That sounds like the sort of thing I would recommend if we need to get an MVP out next week. And I'm like, what if we just tried this? Listen, it's got some trade-offs. So I'm really interested to see are those trade-offs present? Because it's the browser engine. It's, you know, the low-level platform that's actually managing this. And there are some nice hooks that allow you to control it. And at a CSS level, you can manage it and use keyframe animations to control the transition more directly. There's a JavaScript API to instrument the sequencing of things. And so it's giving you the right primitives and the right hooks. And the fact that the implementation happens to use pictures or screenshots, to use a slightly different word, it's like, okey dokey, that's what we're doing. Sounds fun. So I'm super interested because the functionality is deeply, deeply interesting to me. Svelte actually has a version of this, the crossfade utility, but you have to still really think about how do you sequence between the two pages and how do you do the connective tissue there? And then Svelte will manage it for you if you do all the right stuff. But the wiring up is somewhat complicated. So having this in the browser engine is really interesting to me. But yeah, pictures. STEPH: This is one of those ideas where I can't decide if this was someone who is very new to the team and new to the idea and was like, "Have we considered screenshots? Have we considered pictures?" Or if this is like the uber senior person on the team that was like, "Yeah, this will totally work with screenshots." I can't decide where in that range this idea falls, which I think makes me love it even more. Because it's very straightforward of like, hey, what if we just tried this? And it's working, so cool, cool, cool. CHRIS: There's a fantastic meme that's been making the rounds where it's a bell curve, and it's like, early in your career, middle of your career, late in your career. And so early in your career, you're like, everything in one file, all lines of code that's just where they go. And then in the middle of your career, you're like, no, no, no, we need different concerns, and files, and organizational structures. And then end of your career...and this was coming up in reference to the TypeScript team seems to have just thrown everything into one file. And it's the thing that they've migrated to over time. And so they have this many, many line file that is basically the TypeScript engine all in one file. And so it was a joke of like, they definitely know what they're doing with programming. They're not just starting last week sort of thing. And so it's this funny arc that certain things can go through. So I think that's an excellent summary there [laughs] of like, I think it was folks who have thought about this really hard. But I kind of hope it was someone who was just like, "I'm new here. But have we thought about pictures? What about pictures?" I also am a little worried that I just deeply misunderstood [laughs] the representation but glossed over it in the video, and I'm like, that sounds interesting. So hopefully, I'm not just wildly off base here. [laughs] But nonetheless, the functionality looks very interesting. STEPH: That would be a hilarious tweet. You know, I've been waiting for that moment where I've said something that I understood into the mic and someone on Twitter just being like, well, good try, but... [laughs] CHRIS: We had a couple of minutes where we tried to figure out what the opposite of ranting was, and we came up with pranting and made up a word instead of going with praising or raving. No, that's what it is, raving. [laughs] STEPH: No, raving. I will never forget now, raving. [laughs] CHRIS: So, I mean, we've done this before. STEPH: That's true. Although they were nice, I don't think they tweeted. I think they sent in an email. They were like, "Hey, friends." [laughter] CHRIS: Actually, we got a handful of emails on that. [laughter] STEPH: Did you know the English language? CHRIS: Thank you, kind Bikeshed audience, for not shaming us in public. I mean, feel free if you feel like it. [laughs] But one other thing that came up in this video, though, is the speaker was describing single-page apps are very common, and you want to have animated transitions and this and that. And I was like, single-page app, okay, fine. I don't like the terminology but whatever. I would like us to call it the client-side app or client-side routing or something else. But the fact that it's a single page is just a technical consideration that no user would call it that. Users are like; I go to the web app. I like that it has URLs. Those seem different to me. Anyway, this is my hill. I'm going to die on it. But then the speaker in the video, in contrast to single-page app referenced multi-page app, and I was like, oh, come on, come on. I get it. Like, yes, there are just balls of JavaScript that you can download on the internet and have a dynamic graphics editor. But I think almost all good things on the web should have URLs, and that's what I would call the multiple pages. But again, that's just me griping about some stuff. And to name it, I don't think I'm just griping for griping sake. Like, again, I like to think about things from the user perspective, and the URL being so important. And having worked with plenty of apps that are implemented in JavaScript and don't take the URL or the idea that we can have different routable resources seriously and everything is just one URL, that's a failure mode in my mind. We missed an opportunity here. So I think I'm saying a useful thing here and not just complaining on the internet. But with that, I will stop complaining on the internet and send it back over to you. What else is new in your world, Steph? STEPH: I do remember the first time that you griped about it, and you were griping about URLs. And there was a part of me that was like, what is he talking about? [laughter] And then over time, I was like, oh, I get it now as I started actually working more in that world. But it took me a little bit to really appreciate that gripe and where you're coming from. And I agree; I think you're coming from a reasonable place, not that I'm biased at all as your co-host, but you know. CHRIS: I really like the honest summary that you're giving of, like, honestly, the first time you said this, I let you go for a while, but I did not know what you were talking about. [laughs] And I was like, okay, good data point. I'm going to store that one away and think about it a bunch. But that's fine. I'm glad you're now hanging out with me still. [laughter] STEPH: Don't do that. Don't think about it a bunch. [laughs] Let's see, oh, something else that's going on in my world. I had a really fun pairing session with another thoughtboter where we were digging into query objects and essentially extracting some logic out of an ActiveRecord model and then giving that behavior its own space and elevated namespace in a query object. And one of the questions or one of the things that came up that we needed to incorporate was optional filters. So say if you are searching for a pizza place that's nearby and you provide a city, but you don't provide what's the optional zip code, then we want to make sure that we don't apply the zip code in the where clause because then you would return all the pizza places that have a nil zip code, and that's just not what you want. So we need to respect the fact that not all the filters need to be applied. And there are a couple of ways to go about it. And it was a fun journey to see the different ways that we could structure it. So one of the really good starting points is captured in a blog post by Derek Prior, which we'll include a link to in the show notes, and it's using yield_self for composable ActiveRecord relations. But essentially, it starts out with an example where it shows that you're assigning a value to then the result of an if statement. So it's like, hey, if the zip code is present, then let's filter by zip code; if not, then just give us back the original relation. And then you can just keep building on it from there. And then there's a really nice implementation that Derek built on that then uses yield_self to pass the relation through, which then provides a really nice readability for as you are then stepping through each filter and which one should and shouldn't be applied. And now there's another blog post, and this one's written by Thiago Silva, A Case for Query Objects in Rails. And this one highlighted an approach that I haven't used before. And I initially had some mixed feelings about it. But this approach uses the extending method, which is a method that's on ActiveRecord query methods. And it's used to extend the scope with additional methods. You can either do this by providing the name of a module or by providing a block. It's only going to apply to that instance or to that specific scope when you're using it. So it's not going to be like you're running an include or something like that where all instances are going to now have access to these methods. So by using that method, extending, then you can create a module that says, "Hey, I want to create this by zip code filter that will then check if we have a zip code, let's apply it, if not, return the relation. And it also creates a really pretty chaining experience of like, here's my original class name. Let's extend with these specific scopes, and then we can say by zip code, by pizza topping, whatever else it is that we're looking to filter by. And I was initially...I saw the extending, and it made me nervous because I was like, oh, what all does this apply to? And is it going to impact anything outside of this class? But the more I've looked at it, the more I really like it. So I think you've seen this blog post too. And I'm curious, what are your thoughts about this? CHRIS: I did see this blog post come through. I follow that thoughtbot blog real close because it turns out some of the best writing on the internet is on there. But I saw this...also, as an aside, I like that we've got two Derek Prior references in one episode. Let's see if we can go for three before the end. But one thing that did stand out to me in it is I have historically avoided scopes using scope like ActiveRecord macro thing. It's a class method, but like, it's magic. It does magic. And a while ago, class methods and scopes became roughly equivalent, not exactly equivalent, but close enough. And for me, I want to use methods because I know stuff about methods. I know about default arguments. And I know about all of these different subtleties because they're just methods at the end of the day, whereas scopes are special; they have certain behavior. And I've never really known of the behavior beyond the fact that they get implemented in a different way. And so I was never really sold on them. And they're different enough from methods, and I know methods well. So I'm like, let's use the normal stuff where we can. The one thing that's really interesting, though, is the returning nil that was mentioned in this blog post. If you return nil in a scope, it will handle that for you. Whereas all of my query objects have a like, well, if this thing applies, then scope dot or relation dot where blah, blah, blah, else return relation unchanged. And the fact that that natively exists within scope is interesting enough to make me reconsider my stance on scopes versus class methods. I think I'm still doing class method. But it is an interesting consideration that I was unaware of before. STEPH: Yeah, it's an interesting point. I hadn't really considered as much whether I'm defining a class-level method versus a scope in this particular case. And I also didn't realize that scopes handle that nil case for you. That was one of the other things that I learned by reading through this blog post. I was like, oh, that is a nicety. Like, that is something that I get for free. So I agree. I think this is one of those things that I like enough that I'd really like to try it out more and then see how it goes and start to incorporate it into my process. Because this feels like one of those common areas of where I get to it, and I'm like, how do I do this again? And yield_self was just complicated enough in terms of then using the fancy method method to then be able to call the method that I want that I was like, I don't remember how to do this. I had to look it up each time. But including this module with extending and then being able to use scopes that way feels like something that would be intuitive for me that then I could just pick up and run with each time. CHRIS: If it helps, you can use then instead of yield_self because we did upgrade our Ruby a while back to have that change. But I don't think that actually solves the thing that you're describing. I'd have liked the ampersand method and then simple method name magic incantation that is part of the thing that Derek wrote up. I do use it when I write query objects, but I have to think about it or look it up each time and be like, how do I do that? All right, that's how I do that. STEPH: Yeah, that's one of the things that I really appreciate is how often folks will go back and update blog posts, or they will add an addition to them to say, "Hey, there's something new that came out that then is still relevant to this topic." So then you can read through it and see the latest and the greatest. It's a really nice touch to a number of our blog posts. But yeah, that's what was on my mind regarding query objects. What else is going on in your world? CHRIS: I have this growing feeling that I don't quite know what to do with. I think I've talked about it across some of our conversations in the world of observability. But broadly, I'm starting to like...I feel like my brain has shifted, and I now see the world slightly differently, and I can't go back. But I also don't know how to stick the landing and complete this transition in my brain. So it's basically everything's an event stream; this feels true. That's life. The arrow of time goes in one direction as far as I understand it. And I'm now starting to see it manifest in the code that we're writing. Like, we have code to log things, and we have places where we want to log more intentionally. Then occasionally, we send stuff off to Sentry. And Sentry tells us when there are errors, that's great. But in a lot of places, we have both. Like, we will warn about something happening, and we'll send that to the logs because we want to have that in the logs, which is basically the whole history of what's happened. But we also have it in Sentry, but Sentry's version is just this expanded version that has a bunch more data about the user, and things, and the browser that they were in. But they're two variations on the same event. And then similarly, analytics is this, like, third leg of well, this thing happened, and we want to know about it in the context. And what's been really interesting is we're working with a tool called Customer.io, which is an omnichannel communication whiz-bang adventure. For us, it does email, SMS, and push notifications. And it's integrated into our segment pipeline, so events flow in, events and users essentially. So we have those two different primitives within it. And then within it, we can say like, when a user does X, then send them an email with this copy. As an aside, Customer.io is a fantastic platform. I'm super-duper impressed. We went through a search for a tool like it, and we ended up on a lot of sales demos with folks that were like, hey, so yeah, starting point is $25,000 per year. And, you know, we can talk about it, but it's only going to go up from there when we talk about it, just to be clear. And it's a year minimum contract, and you're going to love it. And we're like, you do have impressive platforms, but okay, that's a bunch. And then, we found Customer.io, and it's month-by-month pricing. And it had a surprisingly complete feature set. So overall, I've been super impressed with Customer.io and everything that they've afforded. But now that I'm seeing it, I kind of want to move everything into that world where like, Customer.io allows non-engineer team members to interact with that event stream and then make things happen. And that's what we're doing all the time. But I'm at that point where I think I see the thing that I want, but I have no idea how to get there. And it might not even be tractable either. There's the wonderful Turning the Database Inside Out talk, which describes how everything is an event stream. And what if we actually were to structure things that way and do materialized views on top of it? And the actual UI that you're looking at is a materialized view on top of the database, which is a materialized view on top of that event stream. So I'm mostly in this, like, I want to figure this out. I want to start to unify all this stuff. And analytics pipes to one tool that gets a version of this event stream, and our logs are just another, and our error system is another variation on it. But they're all sort of sampling from that one event stream. But I have no idea how to do that. And then when you have a database, then you're like, well, that's also just a static representation of a point in time, which is the opposite of an event stream. So what do you do there? So there are folks out there that are doing good thinking on this. So I'm going to keep my ear to the ground and try and see what's everybody thinking on this front? But I can't shake the feeling that there's something here that I'm missing that I want to stitch together. STEPH: I'm intrigued on how to take this further because everything you're saying resonates in terms of having these event streams that you're working with. But yet, I can't mentally replace that with the existing model that I have in my mind of where there are still certain ideas of records or things that exist in the world. And I want to encapsulate the data and store that in the database. And maybe I look it up based on when it happens; maybe I don't. Maybe I'm looking it up by something completely different. So yeah, I'm also intrigued by your thoughts, but I'm also not sure where to take it. Who are some of the folks that are doing some of the thinking in this area that you're interested in, or where might you look next? CHRIS: There's the Kafka world of we have an event log, and then we're processing on top of that, and we're building stream processing engines as the core. They seem to be closest to the Turning the Database Inside Out talk that I was thinking or that I mentioned earlier. There's also the idea of CRDTs, which are Conflict-free Replicated Data Types, which are really interesting. I see them used particularly in real-time application. So it's this other tool, but they are basically event logs. And then you can communicate them well and have two different people working on something collaboratively. And these event logs then have a natural way to come together and produce a common version of the document on either end. That's at least my loose understanding of it, but it seems like a variation on this theme. So I've been looking at that a little bit. But again, I can't see how to map that to like, but I know how to make a Rails app with a Postgres database. And I think I'm reasonably capable at it, or at least I've been able to produce things that are useful to humans using it. And so it feels like there is this pretty large gap. Because what makes sense in my head is if you follow this all the way, it fundamentally re-architects everything. And so that's A, scary, and B; I have no idea how to get there, but I am intrigued. Like, I feel like there's something there. There's also Datomic is the other thing that comes to mind, which is a database engine in the Clojure world that stores the versions of things over time; that idea of the user is active. It’s like, well, yeah, but when were they not? That's an event. That transition is an event that Postgres does not maintain at this point. And so, all I know is that the user is active. Maybe I store a timestamp because I'm thinking proactively about this. But Datomic is like no, no, fundamentally, as a primitive in this database; that's how we organize and think about stuff. And I know I've talked about Datomic on here before. So I've circled around these ideas before. And I'm pretty sure I'm just going to spend a couple of minutes circling and then stop because I have no idea how to connect the dots. [laughs] But I want to figure this out. STEPH: I have not worked with Kafka. But one of the main benefits I understand with Kafka is that by storing everything as a stream, you're never going to lose like a message. So if you are sending a message to another system and then that message gets lost in transit, you don't actually know if it got acknowledged or what happened with it, and replaying is really hard. Where do you pick up again? While using something like Kafka, you know exactly what you sent last, and then you're not going to have that uncertainty as to what messages went through and which ones didn't. And then the ability to replay is so important. I'm curious, as you continue to explore this, do you have a particular pain point in mind that you'd like to see improve? Or is it more just like, this seems like a really cool, novel idea; how can I incorporate more of this into my world? CHRIS: I think it's the latter. But I think the thing that I keep feeling is we keep going back and re-instrumenting versions of this. We're adding more logging or more analytics events over the wire or other things. But then, as I send these analytics events over the wire, we have Mixpanel downstream as an analytics visualization and workflow tool or Customer.io. At this point, those applications, I think, have a richer understanding of our users than our core Rails app. And something about that feels wrong to me. We're also streaming everything that goes through segment to S3 because I had a realization of this a while back. I'm like, that event stream is very interesting. I don't want to lose it. I'm going to put it somewhere that I get to keep it. So even if we move off of either Mixpanel or Customer.io or any of those other platforms, we still have our data. That's our data, and we're going to hold on to it. But interestingly, Customer.io, when it sends a message, will push an event back into segments. So it's like doubly connected to segment, which is managing this sort of event bus of data. And so Mixpanel then gets an even richer set there, and the Rails app is like, I'm cool. I'm still hanging out, and I'm doing stuff; it's fine. But the fact that the Rails app is fundamentally less aware of the things that have happened is really interesting to me. And I am not running into issues with it, but I do feel odd about it. STEPH: That touched on a theme that is interesting to me, the idea that I hadn't really considered it in those terms. But yeah, our application provides the tool in which people can interact with. But then we outsource the behavior analysis of our users and understanding what that flow is and what they're going through. I hadn't really thought about those concrete terms and where someone else owns the behavior of our users, but yet we own all the interaction points. And then we really need both to then make decisions about features and things that we're building next. But that also feels like building a whole new product, that behavior analysis portion of it, so it's interesting. My consulting brain is going wild at the moment between like, yeah, it would be great to own that. But that the other time if there's this other service that has already built that product and they're doing it super well, then let's keep letting them manage that portion of our business until we really need to bring it in-house. Because then we need to incorporate it more into our application itself so then we can surface things to the user. That's the part where then I get really interested, or that's the pain point that I could see is if we wanted more of that behavior analysis, that then we want to surface that in the app, then always having to go to a third-party would start to feel tedious or could feel more brittle. CHRIS: Yeah, I'm definitely 100% on not rebuilding Mixpanel in our app and being okay with the fact that we're sending that. Again, the thing that I did to make myself feel better about this is stream the data to S3 so that I have a version of it. And if we want to rebuild the data warehouse down the road to build some sort of machine learning data pipeline thing, we've got some raw data to work with. But I'm noticing lots of places where we're transforming a side effect, a behavior that we have in the system to dispatching an event. And so right now, we have a bunch of stuff that we pipe over to Slack to inform our admin team, hey, this thing happened. You should probably intervene. But I'm looking at that, and we're doing it directly because we can control the message in Slack a little bit better. But I had this thought in the back of my mind; it's like, could we just send that as an event, and then some downstream tool can configure messages and alerts into Slack? Because then the admin team could actually instrument this themselves. And they could be like; we are no longer interested in this event. Users seem fine on that front. But we do care about this new event. And all we need to do as the engineering team is properly instrument all of that event stream tapping. Every event just needs to get piped over. And then lots of powerful tools downstream from that that can allow different consumers of that data to do things, and broadly, that dispatch events, consume them on the other side, do fun stuff. That's the story. That's the dream. But I don't know; I can't connect all the dots. It's probably going to take me a couple of weeks to connect all these dots, or maybe years, or maybe my entire career, something like that. But, I don't know, I'm going to keep trying. STEPH: This feels like a fun startup narrative, though, where you start by building the thing that people can interact with. As more people start to interact with it, how do we start giving more of our team the ability to then manage the product that then all of these users are interacting with? And then that's the part that you start optimizing for. So there are always different interesting bits when you talk about the different stages of Sagewell, and like, what's the thing you're optimizing for? And I'm sure it's still heavily users. But now there's also this addition of we are also optimizing for our team to now manage the product. CHRIS: Yes, you're 100%. You're spot on there. We have definitely joked internally about spinning out a small company to build this analytics alerting tool [laughs] but obviously not going to do that because that's a distraction. And it is interesting, like, we want to build for the users the best thing that we can and where the admin team fits within that. To me, they're very much customers of engineering. Our job is to build the thing for the users but also, to be honest, we have to build a thing for the admins to support the users and exactly where that falls. Like, you and I have talked a handful of times about maybe the admin isn't as polished in design as other things. But it's definitely tested because that's a critical part of how this application works. Maybe not directly for a user but one step removed for a user, so it matters. Absolutely we're writing tests to cover that behavior. And so yeah, those trade-offs are always interesting to me and exploring that space. But 100%: our admin team are core customers of the work that we're doing in engineering. And we try and stay very close and very friendly with them. STEPH: Yeah, I really appreciate how you're framing that. And I very much agree and believe with you that our admin users are incredibly important. CHRIS: Well, thank you. Yeah, we're trying over here. But yeah, I think I can wrap up that segment of me rambling about ideas that are half-formed in my mind but hopefully are directionally important. Anyway, what else is up with you? STEPH: So, not that long ago, I asked you a question around how the heck to manage themes that I have going on. So we've talked about lots of fun productivity things around managing to-dos, and emails, and all that stuff. And my latest one is thinking about, like, I have a theme that I want to focus on, maybe it's this week, maybe it's for a couple of months. And how do I capture that and surface it to myself and see that I'm making progress on that? And I don't have an answer to that. But I do have a theme that I wanted to share. And the one that I'm currently focused on is building up management skills and team lead skills. That is something that I'm focused on at the moment and partially because I was inspired to read the book Resilient Management written by Lara Hogan. And so I think that is what has really set the idea. But as I picked up the book, I was like, this is a really great book, and I'd really like to share some of this. And then so that grew into like, well, let's just go ahead and make this a theme where I'm learning this, and I'm sharing this with everyone else. So along that note, I figured I would share that here. So we use Basecamp at thoughtbot. And so, I've been sharing some Basecamp posts around what I'm learning in each chapter. But to bring some of that knowledge here as well, some of the cool stuff that I have learned so far...and this is just still very early on in the book. There are a couple of different topics that Laura covers in the first chapter, and one of them is humans’ core needs at work. And then there's also the concept of meeting your team, some really good questions that you can ask during your first one-on-one to get to know the person that then you're going to be managing. The part that really resonated with me and something that I would like to then coach myself to try is helping the team get to know you. So as a manager, not only are you going out of your way to really get to know that person, but how are you then helping them get to know you as well? Because then that's really going to help set that relationship in regards of they know what kind of things that you're optimizing for. Maybe you're optimizing for a deadline, or for business goals, or maybe it's for transparency, or maybe it would be helpful to communicate to someone that you're managing to say, "Hey, I'm trying some new management techniques. Let me know how this goes." [chuckles] So there's a healthier relationship of not only are you learning them, but they're also learning you. So some of the questions that Laura includes as examples as something that you can share with your team is what do you optimize for in your role? So is it that you're optimizing for specific financial goals or building up teammates? Or maybe it's collaboration, so you're really looking for opportunities for people to pair together. What do you want your teammates to lean on you for? I really liked that question. Like, what are some of the areas that bring you joy or something that you feel really skilled in that then you want people to come to you for? Because that's something that before I was a manager...but it's just as you are growing as a developer, that's such a great question of like, what do you want to be known for? What do you want to be that thing that when people think of, they're like, oh, you should go see Chris about this, or you should go see Steph about this? And two other good questions include what are your work styles and preferences? And what management skills are you currently working on learning or improving? So I really like this concept of how can I share more of myself? And the great thing about this book that I'm learning too is while it is geared towards people that are managers, I think it's so wonderful for people who are non-managers or aspiring managers to read this as well because then it can help you manage whoever's managing you. So then that way, you can have some upward management. So we had recent conversations around when you are new to a team and getting used to a manager, or maybe if you're a junior, you have to take a lot of self-advocacy into your role to make sure things are going well. And I think this book does a really good job for people that are looking to not only manage others but also manage themselves and manage upward. So that's some of the journeys from the first chapter. I'll keep you posted on the other chapters as I'm learning more. And yeah, if anybody hasn't read this book or if you're interested, I highly recommend it. I'll make sure to include a link in the show notes. CHRIS: That was just the first chapter? STEPH: Yeah, that was just the first chapter. CHRIS: My goodness. STEPH: And I shortened it drastically. [laughs] CHRIS: Okay. All right, off to the races. But I think the summary that you gave there, particularly these are true when you're managing folks but also to manage yourself and to manage up, like, this is relevant to everyone in some capacity in some shape or form. And so that feels very true. STEPH: I will include one more fun aspect from the book, and that's circling back to the humans' core needs at work. And she references Paloma Medina, a coach, and trainer who came up with this acronym. The acronym is BICEPS, and it stands for belonging, improvement, choice, equality, predictability, and significance. And then details how each of those are important to us in our work and how when one of those feels threatened, then that can lead to some problems at work or just even in our personal life. But the fun example that she gave was not when there's a huge restructuring of the organization and things like that are going on as being the most concerning in terms of how many of these needs are going to be threatened or become vulnerable. But changing where someone sits at work can actually hit all of these, and it can threaten each of these needs. And it made me think, oh, cool, plus-one for being remote because we can sit wherever we want. [laughs] But that was a really fun example of how someone's needs at work, I mean, just moving their desk, which resonates, too, because I've heard that from other people. Some of the friends that I have that work in more of a People Ops role talk about when they had to shift people around how that caused so much grief. And they were just shocked that it caused so much grief. And this explains why that can be such a big deal. So that was a fun example to read through. CHRIS: I'm now having flashbacks to times where I was like, oh, I love my spot in the office. I love the people I'm sitting with. And then there was that day, and I had to move. Yeah, no, those were days. This is true. STEPH: It triggered all the core BICEPS, all the things that you need to work. It threatened all of them. Or it could have improved them; who knows? CHRIS: There were definitely those as well, yeah. Although I think it's harder to know that it's going to be great on the way in, so it's mostly negative. I think it has that weird bias because you're like, this was a thing, I knew it. I at least understood it. And then you're in a new space, and you're like, I don't know, is this going to be terrible or great? I mean, hopefully, it's only great because you work with great people, and it's a great office. [laughs] But, like, the unknown, you're moving into the unknown, and so I think it has an inherent at least questioning bias to it. STEPH: Agreed. On that note, shall we wrap up? CHRIS: Let's wrap up. 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: Byeeeeeee!!!!!!! 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.Support The Bike Shed