Being pregnant is hard, but this tapas episode is good! Steph discovered and used a #yelling Slack channel and attended a remote magic show. Chris touches on TypeScript design decisions and edge cases.
Then they answer a question captured from a client Slack channel regarding a debate about whether I18n should be used in tests and whether tests should break when localized text changes.
This episode is brought to you by ScoutAPM. Give Scout a try for free today and Scout will donate $5 to the open source project of your choice when you deploy.
Gregg Tobo the Magician
Sean Wang - swyx - better twitter search
GitHub Pull Request File Tree Beta
Sam Zimmerman - CEO of Sagewell Financial on Giant Robots
TypeScript 4.1 feature
The Bike Shed: 269: Things are Knowable (Gary Bernhardt)
TSConfig Reference - Docs on every TSConfig option
This episode is brought to you by Studio 3T. Try Studio 3T's full suite of features for 30 days, no payment details needed.
Become a Sponsor of The Bike Shed!
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. There are a couple of new things in my world, so one of them that I wanted to talk about is the fact that being pregnant is hard. I feel like this is probably a known thing, but I feel like I don't hear it talked about as much as I'd really like, especially in sort of like a professional context. And so I just wanted to share for anyone else that may be listening, if you're also pregnant, this is hard.
And I also really appreciate my team. Going through the first trimester is typically where you experience a lot of morning sickness and fatigue, and I had all of that. And so I was at the point that most of my days, I didn't even start till about noon and even some days, starting at noon was a struggle. And thankfully, the thoughtbot client that I'm working with most of the teams are on West Coast hours, so that worked out pretty well.
But I even shared a post internally and was like, "Hey, I'm not doing great in the mornings. And so I really can't facilitate any morning meetings. I can't be part of some of the hiring intros that we do," because we like to have a team lead provide a welcoming and then closing for anyone that's coming for interview day. I couldn't do those, and those normally happen around 9:00 a.m. for Eastern Time. And everybody was super supportive of it. So I really appreciate all of thoughtbot and my managers and team being so great about this. Also, the client team they're wonderful.
It turns out growing a little human; I'm learning how hard it is and working full time. It's an interesting challenge. Oh, and as part of that appreciation because…so there's just not a lot of women that I've worked with. This may be one of those symptoms of being in tech where one, I haven't worked with tons of women, and then two, working with a woman who is also pregnant and going through that as well. So it's been a little bit isolating in that experience.
But there is someone that I follow on Twitter, @EmmaBostian. She's also one of the co-hosts for the Ladybug Podcast. And she has been just sharing some of her, like, I am two months sleep deprived. She's had her baby now, and she is sharing some of that journey. And I really appreciate people who just share that journey and what they're going through because then it helps normalize it for me in terms of what I'm feeling. I hope this helps normalize it for anybody else that might be listening too.
CHRIS: I certainly can't speak to the specifics of being pregnant. But I do think it's wonderful for you to use this space that we have here to try and forward that along and say what your experience is like and share that with folks and hopefully make it a little bit better for everyone else out there. Also, you snuck in a sneaky pro-tip there, which is work on the East Coast and have a West Coast team. That just sounds like the obvious correct way to go about this.
STEPH: That has worked out really well and been very helpful for me. I'm already not a great morning person; I've tried. I've really strived at times to be a morning person because I just have this idea in my head morning people get more stuff done. I don't think that's true, but I just have that idea. And I'm not the world's best morning person, so it has worked out for many reasons but yeah, especially in helping me get through that first trimester and also just supporting family and other things that are going on.
Oh, I also learned a pro-tip about Twitter. This is going to seem totally random, but it was relevant when I was searching for stuff on Twitter [laughs] that was related to tech and pregnancy. But I learned...because I wanted to be able to search for something that someone that I follow what they said but I couldn't remember who said it.
And so I found that in the search bar, I can add filter:follows. So you can have your search term like if you're looking for cake or pregnancy, or sleep-deprived and then look for filter:follows, and then that will filter the search results to everybody that you follow. I imagine that that probably works for followers too, but I haven't tried it.
CHRIS: I like the left turn you took us on there but still keeping it connected. On the topic of Twitter search, they apparently have a very powerful search, but it's also hidden, and you got to know the specific syntax and whatnot. But there is a wonderful project by Shawn Wang, AKA Swyx, on the internet, bettertwitter.netlify.com is the URL for it. I will share a link to his tweet introducing it. But it's a really wonderful tool that just provides a UI for all of these different filters and configurations. And both make discoverability that much better and then also make it easy to just compose one of these searches and use that.
The other thing that I'll recommend is, I think it's a Chrome plugin. I'm guessing is what I'm working with here like a browser extension, but it's called Twemex, T-W-E-M-EX. And there's a sidebar in Twitter now, which just seems wonderful and useful. So as I'm looking at a Swyx post here, or a tweet as they're called on Twitter because I know that vernacular, there's a sidebar which is specific to Shawn Wang.
And there's a search at the top so I can search within it. But it's just finding their most popular tweets and putting that on a sidebar. It's a very useful contextual addition to Twitter that I found just awesome. So that combination of things has made my Twitter experience much better. So yeah, we'll have show notes for both of those as well.
STEPH: Nice. I did not know about those. This may cause someone to laugh at me because maybe it's easier than I think. But I can never remember that advanced search that Twitter does offer; I have to search it every time. I just go to Google, and I'm like, advanced Twitter search, and then it brings up a site for me, and then I use that as the one that Twitter does provide. But yeah, from the normal UI, I don't know how to get there. Maybe I haven't tried hard enough. Maybe it's hidden.
CHRIS: It's like they're hiding it.
STEPH: Yeah, one of those. [laughs]
CHRIS: It's very costly. They have to like MapReduce the entire internet in order to make that search work. So they're like, well, what if we hide it because it's like 50 cents per query? And so maybe we shouldn't promote this too much.
CHRIS: And let's just live in the moment, everybody. Let's just swim in the Twitter stream rather than look back at the history. I make guesses about the universe now.
STEPH: [laughs] On a different note, I also discovered at thoughtbot in our variety of Slack channels that we have a yelling channel, and I had not used it before. I had not hung out there before. It's a delightful channel. It's a place that you just go, and you type in all caps. You can yell about anything that you would like to. And I specifically needed to yell about Gerrit, which is the replacement or the alternative that we're using for GitHub or GitLab, or Bitbucket, or any of those services.
So we're using Gerrit, and I've been working to feel comfortable with the UI and then be able to review CRs and things like that. My vernacular is also changing because my team refers to them as change requests instead of pull requests. So I'm floating back and forth between CRs and PRs. And because I'm in Gerrit world, I missed some of the updates that GitHub made to their pull request review screen. And so then I happened to hop in GitHub one day, and I saw it, and I was like, what is this? So that was novel.
But going back to yelling, I needed to yell about Gerrit because I have not found a way to collaborate with someone who has already pushed up changes. I have found ways that I can pull their changes which then took a little while. I found it in a sneaky little tab called download. I didn't expect it to be there. But then the actual snippet it's like, run this in your terminal, and this is then how you pull down the changes. And I'm like, okay, so I did that.
But I can't push to their existing changes because then I get like, well, you're not the owner, so we're going block you, which is like, cool, cool, cool. Okay, I kind of get that because you don't want me messing up somebody else's content or something that they've done. But I really, really, really want to collaborate with this person, and we're trying to do something together, and you're blocking me. And so I had to go to the yelling channel, and I felt better. And I'm yelling again. [laughs] Maybe I don't feel that great because I'm getting angry again talking about it.
CHRIS: You vented a little into the yelling channel; maybe not everything, though.
STEPH: Yeah, I still have more to vent because it's made life hard. Every time I wanted to push up a change or pull down someone else's changes, there are now all these CRs that then I just have to go and abandon, which is then the terminology for then essentially closing it and ignoring it, so I'm constantly going through.
And if I do want to pull in changes or collaborate, then there's a flow of either where I abandon mine, or I pull in their changes, but then I have to squash everything because if you push up multiple commits to Gerrit, it's going to split those commits into different CRs, don't like that. So there are a couple of things that have been pain points. And yeah, so plus-one for yelling channels, let people get it out.
CHRIS: Okay, so definitely some feelings that you are working through here. I'm happy to work together as a team to get through some of them. One thing that I want to touch on is you very quickly hinted at GitHub has got a bunch of new things that are cool. I want to talk about those. But I want to touch [laughs] on an anecdote. You talked about pushing something up to someone else's branch. You're like, oh, you know, I made some changes locally, and I'm going to push them up.
I had an interesting experience once where I was interacting with another developer. I had done some code review. They weren't quite understanding where I was. They had a lot of questions. And finally, I said, you know what? This will just be easier. Here, I pushed up a commit to your branch, so now you can see what I'm talking about. And I thought of this as a very innocuous act, but it was not interpreted that way. That individual interpreted it in a very aggressive sort of; it was not taken well.
And I think part of that was related to I think of Git commits as just these little ephemeral things where you're like, throw it out, feel free. This is just the easiest way for me to communicate this change in the context of the work that you're doing. I thought I was doing a nice favor thing here. That was not how it went. We had a good conversation after I got to the heart of where we both were emotionally on this thing. It was interesting. The interaction of emotion and tech is always interesting.
But as a result, I'm very, very careful with that now. I do think it's a great way as long as I've gotten buy-in from the person beforehand. But I will always spot check and be like, "Hey, just to confirm, I can just push up a commit to your branch, but are you okay with that? Is that fine with you?" So I've become very cautious with that.
STEPH: Yeah, that feels like one of those painful moments where it highlights that the people that you work with that you are accustomed to having a certain level of trust or default trust with those individuals, and then working with someone else that they don't have that where the cup is half-full in terms of that trust, or that this person means well kind of feelings towards a colleague or towards someone that they're working with.
So it totally makes sense that it's always good to check and just to be like, "Hey, I'd love to push up some changes to your branch. Is that cool?" And then once you've established that, then that just makes it easier. But I do remember that happening, and yeah, that was a bit painful and shocking because we didn't see that coming and then learned from it.
CHRIS: I do think it's an important thing to learn, though, because for me, in that moment, this was this throwaway operation that I thought almost nothing of, but then another individual interpreted it in a very different way. And that can happen, that can happen across tons of different things. And I don't even want to live in the idealized world where it's just tech; we're just pushing around zeros and ones; there's no human to this. But no, I actually believe it's a deeply human thing that we're doing here.
It's our job to teach the computers to be a little closer to us humans or something like that. And so it was a really pointed clarification of that for me where it was this thing that I didn't even think once about, no less twice, and yet someone else interpreted it in such a different way. So it was a useful learning situation for me.
STEPH: Yeah, I totally agree. I think that's a really wise default to have to check in with people before assuming that they'll be comfortable with something that we're comfortable with.
CHRIS: Indeed. But shifting back to what you mentioned of GitHub, a bunch of new stuff came in GitHub, and you were super excited about it. And then you went on to say other things about another system. [laughs] But let's talk about the great things in GitHub. What are the particular ones that have caught your eye? I've seen some, but I'm intrigued. Let's compare notes.
STEPH: So this is one of those where I hadn't seen GitHub in quite a while, and then I hopped in, and I was like, this is different. But some of the things that did stand out to me right away is that on the left-hand side, I can see all of the files that have been changed, and so that's a really nice tree where I just then immediately know.
Because that was one of the things that I often did going to a PR is that I would see what files are involved in this change because it was just a nice overview of what part of the applications am I walking through? Are there tests for this? Have they altered or added tests? And so I really like that about it. I'm sure there's other stuff. But that is the main thing that stood out to me. How about you?
CHRIS: Yeah, that sidebar file tree is very, very nice, which I find surprising because I don't use a file tree in my editor. I only do fuzzy finding to jump to files. But I think there's something about whenever GitHub had the file list; these are all the files that are changed. I'm like, this is just noise. I can't look at this and get anything out of it. But the file tree is so much more...there's a shape to it that my brain can sort of pattern match on. And it's just a much more discoverable way to observe that information. So I've really loved that. That was a wonderful one.
The other one that I was surprised by is GitHub semantic code analysis; stuff has gotten much, much better over time subtly. I didn't even notice this happening. But I was discussing something with someone today, and we were looking at it on GitHub, and I just happened to click on an identifier, and it popped up a little thing that says, "Oh, do you want to hop to the references or the definition of this?" I was like, that is what I want to do. And so I hopped to the definition, hopped to the definition of another thing, and was just jumping around in the code in a way that I didn't know was available. So that was really neat.
But then also, I was in a pull request at one point, and someone was writing a spec, and they had introduced a helper just like stub something at the bottom of a given spec file. And it's like, I feel like we have this one already. And I just clicked on the identifier. I think it might have actually been a matcher in RSpec, so it was like, have alert. And I was like, oh, I feel like we have this one, a matcher specific to flash message alerts on the page. And I clicked on it, and GitHub provided me a nice little inline dialog that showed me all of the definitions of have alert, which I think we were up to like four of them at that point.
So it had been copied and pasted across a couple of different files, which I think is totally fine and a great way to start, but they were very similar implementations. I was like, oh, looks like we actually already have this in a couple of places, maybe we clean it up and extract it to a common spec support thing, and ta-da, I was able to do all of that from the GitHub pull requests UI. And I was like, this is awesome. So kudos to the GitHub team for doing some nifty stuff. Also, can I get into the merge queue? Thank you.
STEPH: [laughs] There it is. That is very cool. I didn't know I could do that from the pull request screen. I've seen it where if I'm browsing code that, then I can see a snippet of where everything's defined and then go there, but I hadn't seen that from the pull request. I did find the changelogs for GitHub that talk about the introduction of having the tree, so we'll be sure to include a link in the show notes for that too. But yes, thank you for letting me use our podcast as a yelling channel. It's been delightful. [laughs]
Hi, friends, and now a quick break to hear from today's sponsor, Scout APM.
Scout APM is an application performance monitoring tool that's designed to help developers find and fix performance issues quickly. With an intuitive user interface, Scout will tie bottlenecks to source code so you can quickly pinpoint and resolve performance abnormalities like N+1 queries, slow database queries, and memory bloat.
Scout also recently implemented external service monitoring, adding even more granularity when it comes to HTTP requests and API calls. So give Scout a try today with a free 14-day trial and experience first-hand why developers worldwide call Scout their best friend.
And as an added bonus for Bike Shed listeners, Scout will donate $5 to the open-source project of your choice when you deploy. To learn more, visit scoutapm.com/bikeshed. That's scoutapm.com/bikeshed.
CHRIS: Well, speaking of podcasts, actually, there was an interesting thing that happened where the CEO of Sagewell Financial, the company of which I am the CTO of, Sam Zimmerman is his name, and he went on the Giant Robots Podcast with Chad a couple of weeks ago. So that is now available. We'll link to that in the show notes. I'll be honest; it was a very interesting experience for me. I listened to portions of it. If we're being honest, I searched for my name in the transcript, and it showed up, and I was like, okay, that's cool. And it was interesting to hear two different individuals that I've worked with either in the past or currently talking about it.
But then also, for anyone that's been interested in what I'm building over at Sagewell Financial and wants to hear it from someone who can probably do a much better job of pitching and describing the problem space that we're working in, and all of the fun challenges that we have, and that we're hopefully living up to and building something very interesting, I think Sam does a really fantastic job of that. That's the reason I'm at the company, frankly.
So yeah, if anyone wants to hear a little bit more about that, that is a very interesting episode. It was a little weird for me to listen to personally, but I think everybody else will probably have a normal experience listening to it because they're not the CTO of the company. So that's one thing.
But moving on, I feel like today's going to be a grab bag episode or tapas episode, lots of small plates, as we were discussing as we were prepping for this episode. But to share one little thing that happened, I've been a little more removed from the code of late, something that we've talked about on and off in previous episodes. Thankfully, I have a wonderful team that's doing an absolutely fantastic job moving very rapidly through features and bug fixes and all those sorts of things.
But also, I'm just not as involved even in code review at this point. And so I saw one that snuck through today that, I'm going to be honest, I had an emotional reaction to. I've talked myself down; we're fine now. But the team collectively made the decision to move from a line length of 80 characters to a line length of 120 characters, and I had some feelings.
STEPH: Did you fire everybody? [laughs]
CHRIS: No. I immediately said, doesn't really matter. This is the whole conversation around auto-formatting tools is like we're just taking the decision away. I personally am a fan of the smaller line length because I like to have multiple files open left to right. That is my reason for it, but that's my reason. A collective of the developers that are frankly working more in the code than I am at this point decided this was meaningful. It was a thing that we could automate. I think that we can, you know, it's not a thing that we have to manage. So I was like, cool. There we go.
The one thing that I did follow up on I was like, okay; y'all snuck this one in, it's fine, I'm fine with it. I feel fine; everything's fine. But let's add that to the git-blame-ignore-revs file, which is a useful thing to know about. Because otherwise, we have a handful of different changes like this where we upgrade Prettier, and suddenly, the manner in which it formats the files changes, so we have to reformat everything at once. And this magical file that exists in Git to say, "Hey, ignore this revision because it is not relevant to the semantic history of the app," and so it also takes that decision out of the consideration like yeah, should we reformat or not? Because then it'll be noisy. That magical file takes that decision away, and so I love that.
STEPH: I so love the idea because you took vacation recently twice. So I love the idea of there was a little coup and people are applauding, and they're like, while Chris is on vacation, we're going to merge this change [laughs] that changes the character line. And yeah, that brings me joy. Well, I'm glad you're working through it. Sounds like we're both working through some hard emotional stuff. [laughs]
CHRIS: Life's tricky, is all I'm going to say.
STEPH: I am curious, what prompted the 80 characters versus 120? This is one of those areas that's like, yeah, I have my default preference like you said. But I'm more intrigued just when people are interested in changing it and what goes with it. So do you remember one of the reasons that 120 just suited their preferences better?
CHRIS: Frankly, again, I was not super involved in the discussion or what led them to it.
CHRIS: My guess is 120 is used...I think 80 is a pretty common one. I think 120 is another of the common ones. So I think it's just a thing that exists out there in the mindshare. But also, my guess is they made the switch to 120 and then reformatted a few files that had like, ah, this is like 85 characters, and that's annoying. What does it look like if we bump it up?
And so 120 provided a meaningful change of like, this is a thing that splits to four lines if we have an 80 character thing, or it's one line if it's 120 characters, which is a surprising thing to say, but that's actually the way it plays out in certain cases because the way Prettier will break lines isn't just put stuff on the next line always. It's got to break across multiple lines, actually. All right, now that we're back in the opinion space, I have a strong one.
STEPH: This is The Bikeshed. We can live up to that name. [laughs]
CHRIS: So I do want an additional configuration in Prettier Ruby. This is the thing I'll say. Maybe I can chase down Kevin Newton and see if he's open to this. But when Prettier does break method call with arguments going into it but no parens on that method call, and it breaks out to multiple lines, it does the dangling indent thing, which I do not like. I find it distasteful; I find it noisy, the shape of the code. I'm a big fan of the squint test. I know that from Sandi Metz, I believe, or maybe it's Avdi Grimm. I associate it with both of them in my mind.
But it's just a way to look at the code and kind of squint, and you see the shape of it, and it tells you something. And when the lines break in that weird way, and you have these arbitrary dangling indents, the shape of the code is broken up. And I don't feel so strongly. I actually regularly stop myself from commenting on pull requests on this because it's very easy. All you need to do is add explicit parens, and then Prettier will wrap the line in what I believe is a much more aesthetically pleasing, concise, consistent, lots of other good adjectives here that are definitely just my preferences and not facts about the world.
But so what I want is, Prettier, hey, if you're going to break this line across multiple lines, insert the parens. Parens are no longer optional for breaking across multiple lines; parens are only optional within a given line. So if we're not breaking across lines, I want that configuration because this is now one of those things where I could comment on this. And if they added the optional parens, then Prettier would reform it in a different way. And I want my auto formatter don't give me ways to do stuff. Like, constrain me more but also within the constraints of the preferences that I have, please, thank you.
STEPH: I love all the varying levels there [laughter] of you want a thing, but you know it's also very personal to you and how you're walking that line and hopping back and forth on each side. I also love the idea. We have the idea of clean code. I really want something that's called distasteful code now [laughs] where you just give examples of distasteful code, yes. Well, I wish you good luck in your journey [laughs] and how this goes and how you continue to battle.
I also appreciate that you mentioned when you're reviewing code how you know it's something that you really want, but you will refrain from commenting on that. I just appreciate when people have that filter to recognize, like, is this valuable? Is it important? Or, like you said, how can we just make this more of the default so then we don't even have to talk about it? And then lean into whatever the default the team goes with.
CHRIS: Well, thank you. I very much appreciate that because, frankly, it's been very difficult.
STEPH: I do have something I want to yell about but in a very positive way or pranting as we determined or, you know, raving, the actual real term that wonderful listeners pointed out to us.
CHRIS: Prant for life. That's my stance.
STEPH: We had a magic show at thoughtbot. It was all remote, but the wonderful Gregg Tobo, the magician, performed a magic show for us where we all showed up on Zoom. And it was interactive, and it was delightful, and it was so much fun. And so if you need something fun for your team that you just want to bring folks together, highly recommend. I had no idea I was going to enjoy a magic show this much, but it was a lot of fun. So I'll be sure to include some links in the show notes in case that interests anyone. But yeah, magic. I'm doing jazz hands. People can't see it, but magic.
I like how you referred earlier, saying that today is more of like a tapas episode. And I'm realizing that all of my tapas are related to being pregnant, yelling, and magic shows, and I'm okay with that. [laughs] But on that note, what else is on your tapas plate?
CHRIS: Actually, a nice positive one that came into the world...I always like when we get those. So this is interesting because I was actually looking back at the history, and I had Gary Bernhardt on The Bike Shed back in Episode 269. We'll include a link in the show notes. But we talked a bunch about various things, including TypeScript. And I was lamenting what I saw as a pretty big edge case in TypeScript.
But now we get to the fun part of the story, TypeScript, as of version 4.1, which came out like the week that I recorded with Gary Bernhardt, which was interesting to look at the timeline here. TypeScript has added a new configuration. So a new strictness dial that you can configure in your tsconfig called noUncheckedIndexedAccess. So if you have an array and you are getting an element out of it by index, TypeScript will say, "Hey, you got to check if that's undefined," because to be clear, very much could be undefined. And I was so happy to find this.
We turned it on in our codebase. It found the error in the place that we actually had an error and then found a few others that I think probably had errored at some times. But it was just one of those for me very nice things to be able to dial up the strictness and enforce correctness within our codebase, and so I was very happy about it. Other folks may say that seems like too much work. And, you know, I get that, I get that take. I'm definitely on the side of I'm willing to go through the effort to have enforced correctness, but you know, that's a choice.
STEPH: Yeah, that's thoughtful. I like that, how you said you can dial up the strictness so then as you are introducing TypeScript, then people have that option. There is an argument there in the back of my head that's like, well, if you're introducing types, then you want to start more strict because then you're just creating problems for yourself down the road. But I also understand that that can make things very difficult to then introduce it to teams in existing codebases. So that seems like a really nice addition where then people can say, "Yeah, no, I really want the strictness. This is why I'm here," and then they can turn that on.
CHRIS: So TypeScript in the configuration has strict mode, so you say strict true. And that is a moving target with each new version of TypeScript. But it's their sort of [inaudible 28:14] set of things that are part of strict, but apparently, this one's not in it. So now I'm like, wait, can I have a stricter? Can I have a strictest option? Can I have dial it to 11, please? [laughs] Really rough me up and make sure my code is correct.
But it is the sort of thing like when we turn any of these on; it will find things in our codebase. Some of them, we have to appease the compiler even though we know the code to be correct. But the code is not provably correct as it sits in our file. So I am, again, happy to make that exchange. And I like that TypeScript as a project gives us configurability. But again, I am on team where's the strictest button? I would like to push that as hard as I can and live that life.
STEPH: Yeah, I like that phrasing that you just said about provably correct. That's nice.
CHRIS: That's the world I want to live in, everything you own in the box to the left, which is probably correct.
STEPH: [laughs] That's how that song goes.
CHRIS: Yeah. This is a reference to move errors to the left, which I think I've referenced before. But now that I'm just referencing Beyoncé and not the actual article, it's probably worth referencing the article, but the idea of, like, if a user hits an error, that's not great. So let's move it back to QA, that's a little further to the left in sort of the timeline.
But what if we could move it to an automated test in CI? But what if we could move it into your editor? What if we could move it even further to the left? And so, a type system tends to be sort of very far ratcheted up to the left. It's as early as possible that you can catch these. So again, to reference Beyoncé, everything you own in a box to the left.
STEPH: [singing] Everything you own in the box to the left.
CHRIS: Thank you for doing the needful work there.
And now a quick break to hear from today's sponsor, Studio 3T.
When you're developing applications, it can often be a chore to work with your underlying data. Studio 3T equips you with a complete set of tools to work with MongoDB data. From building queries with drag and drop, to creating complex aggregation pipelines; Studio 3T makes it easy.
And now, there's Studio 3T Free, a free edition of Studio 3T, which delivers an essential core of tools. This means you can get started, for free, with Studio 3T Free, and when you're ready, you can upgrade and enjoy even more features through Studio 3T Pro and Studio 3T Ultimate. The different editions unlock more tools and additional integrations with MongoDB, SQL, Oracle, and Sybase.
You can start today by downloading Studio 3T Free, which also includes a 30-day free trial of all the features of Studio 3T Ultimate, so you can try out some of the enterprise features as well. No credit card required. To start your trial, head to studio3t.com/free that's studio3t.com/free.
STEPH: I have a question for you that I'd really love to get your opinion on because I myself I’m waffling back and forth where someone brought up some really great points about a concern or just a question they had brought up around testing and i18n specifically. And I agree with the things that they're saying, but yet, there's also a part of me that doesn't, and so I'm Stephanie divided. And so, I'm trying to figure out where I stand on this.
So let me dive in and give you some context; I'm going to share the statement/question that they had asked. So here we go. "One of my priorities has been I should be able to review a test without having to reference any other code. References to i18n means that I have to go over to YAML and make sure the right keys have the right values, and that seems error-prone.
In some cases, a lack of a hit in the YAML defers to defaults. If the intent is to override the name of model attribute and error messages and it is coded incorrectly, the code fails silently without translating and uses the humanized attribute name, and that would go undetected. If libraries change structure, it might also fail silently as well, so to me, the only failsafe way is to be fully explicit in test."
So this goes with the idea that if you're writing tests and then you're testing text, but it’s on the screen or perhaps an email, that you're actually going to assert against that string that is shown to the user instead of referencing the i18n keys. And then that also backs up this person's idea that you really want to not have to jump around. If you're reading a test, everything you really need to know about that test should live very close by.
And I really agree with that initial statement; I want everything that's very close to the test, especially if it's anywhere in that expectation line, I really want it close, so I can understand what's the expectation, what's under test, what are the inputs, what's the expected outcome. So I wholeheartedly support that idea.
But yet, I am in the camp that I then will use YAML keys instead of providing that exact string because I do look at i18n as a helpful abstraction, and I want to trust that i18n is doing its job. And so that way, I don't have to provide that string that's there because then we're also choosing, okay, well, which language are we going to always use for our test?
So this is the part where I feel divided. So I'm going to walk you through some of the reasons that I really support this idea and other reasons that I still use the i18n keys and then get your take on it. So there is a part of me that when I'm using the i18n YAML keys, it does make me sad because it reduces the readability in tests. Sometimes the keys are really well named where maybe it's a mailer.welcomemessage. And I'm like, okay, I understand the gist. I don't need to go see the actual string.
I also think they highlighted a really good use case where if you're overriding behavior and it could default to something else, your test is still going to pass, and you don't actually know. So I could see the use case there where if you are overriding, then you want to be explicit about the string that you expect back. I also think there are some i18n messages that are fairly complex, and where then I really would like to see the string.
So if you are formatting a date or a time or you're passing in just a lot of variables, then there's a chance that I do want to see how did that actually get generated for the person who's going to be reading it versus just maybe it's garbage text that came out? And I want to validate that the message that we think we're crafting is actually the one that the user is going to see.
The case against actually being explicit, my biggest one is because then I do see i18n as a helpful abstraction. And I want to trust this abstraction that it's doing its job and it's doing it well. Because then if I do use explicit strings, it makes me sad if I change text from like hello to welcome, and now I have a failing test. I don't like that idea either.
So I'm torn between these two worlds of it is very nice to have everything that you need in a test to be able to understand what is the expectation, but then I also lean into this abstraction and reference the i18n keys. So, Chris, with all of that, that was a bit of a whirlwind, [laughs] what are your thoughts? How do you test this stuff?
CHRIS: Honestly, I'm surprised that you've got that much division in your own answer because for me, this is very obvious there's one...no, I'm kidding. This is obviously complicated. Similar to you, I think I'm going to have to give a grab bag of answers because I don't have a singular thought of like it is concretely this or that.
I tend to go for explicit strings and tests all the way to...so like the readability of a test, and the conciseness of a test is interesting. I will often see developers extract. Say they're creating a user with a specific email, and then they log in with that email later, and then they expect something else. And so the email is referenced a few times, and they'll extract that into a variable called email. And I personally will tend to not do that. I will inline the literal string like firstname.lastname@example.org, and I'll do it in a few places. And I'm fine with that duplication because I like the readability of any given line that you're reading. So I will make that trade-off within tests.
This is the thing I think we've talked about before, but the idea of DRY in tests is like I want to be careful applying that idea, Don't Repeat Yourself, to break apart the acronym. Those abstractions I will use them less than tests. And so I want the explicitness, I want the readability, I want to tell a little story, all of that feels true.
That said, to flip it around, one of the things that I'm hearing...so I think I'm hearing a part of this that is around well, we can fail silently because we fail symmetrically in both the implementation and our test. Then an assertion may actually match even though it's matching on a fallback. I think that's a configurable thing. I would actually want my test to raise if I'm referencing an i18n key that is not defined.
Now, granted, that's different for languages. And maybe this becomes a more complex story of like in production; in a different locale, it will fail because we don't have 100% parity across all our locale files. But fundamentally, I want to make sure that at least exists in our base, which I think typically would be en-US as the locale. I want to make sure all keys are looked up and found, and it's an error otherwise in our test. So that's a feeling. But am I misunderstanding that part of the story or how that configuration typically works?
STEPH: No, I think you've got it. But just to make sure we're on the same page, so if you reference a key that doesn't exist, then it is going to fail. So at least you have your test failure is going to let you know that you've referenced something that doesn't exist. But if you are referencing, like if you want to override the defaults that Rails or i18n has provided for a model and say for an error message, if you reference that, but you want to override it, but then you've forgotten, that does exist.
So you're not going to get the failure; you're going to get a different message. So it's probably not a terrible experience for the user. It's not going to crash. They're going to see something, but they're not going to see the custom message that you intended them to see.
CHRIS: Gotcha. Okay, well, just to name it, the thing that I was describing, I don't know that that would be the configuration for every system. So I would strongly encourage any system where i18n just has a singular behavior which is we fall back to the key. I want my test to absolutely tell me if that's happening. And that should be a failure of the test. But to the discoverability documentation bit, I do wonder if tooling can actually help answer the question.
And as I was describing the wonderful experience I had on GitHub the other day, viewing code as just static characters in a file is both true and also, I think increasingly, a limited view of it. We have editors, and we have code hosting tools that can understand semantically our code a little bit better. There's got to be like 20 Different VS Code plugins that, when you hover on an i18n reference, it will do the lookup for you. That feels like a thing that exists, and if it doesn't, well, now I’ve nerd-sniped myself, and I got a weekend project. JK, I'm definitely not building that this weekend.
But that feels like can we use that to solve this? Maybe not. But that's just another thought of where we have these limitations where it's static, like those abstractions can be useful. But if we can very quickly dereference them, then the cost of the abstraction or that separation becomes smaller, and so the pain is reduced. And I wonder if that's a way to sort of offset it.
STEPH: If I can poke at that a little bit more, because I think you're touching on something that I haven't expressed or thought through explicitly, but it's the idea of, like, why do I like the abstraction? What is it that's drawing me towards using these keys? And I think it's because most of the cases, I don't care. I don't care what the string is, and so that feels nice. Like, I understand that, yes, we're referencing something. If that key didn't exist, I'm going to see a failure.
So I know that there's text there, and that's why I do lean into referencing the keys instead of the text because it feels good to not have to care about that stuff. And if we do make changes to the text, then it suddenly doesn't fail, and then I have to go update a test because we added a period or added a comma. I think that's the path of more sadness for me. And my goal is always a path of least sadness. So I think that's why I lean into it [laughs], I'm guessing. Is that why you lean into it as well? Or what do you like about referencing the keys over the explicit text?
CHRIS: No, I think I share your inclination there, and the reason that you're in favor of it, and I think the consistency like if we're going to use i18n, then we should lean in because it's a non-trivial thing to do like porting to i18n projects, and they're tricky. Getting it right from the first step is also tricky. If you're going to do it, then let's lean in, and thus let's use that abstraction overall. But yeah, same ideas as you.
STEPH: Cool. I think that helps validate where I'm at in terms of how I rationalize about this where ultimately, I do like leaning into that abstraction. And as you'd mentioned, some of those porting projects, I haven't been on one specifically, but I've seen that they are a lot of work. And so, if we have that in our system, then we want to continue to use it.
It does reduce some of the readability. Like you said, maybe there's a VS Code plugin or some way that then we can help people be able to see if they want that full context in the test and not have to jump over to YAML. But yeah, otherwise, unless it's overriding default behavior or complex, then that's what I'm going to go with is with the keys.
But I really appreciate this person's very thoughtful question and approach to testing because, normally or typically, I fully agree with I want full context in the test. And this one was one of those outliers that came up for me, and I had to really think through all the feelings and the reasons that I have for those feelings. 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@example.com via email.
CHRIS: Thanks so much for listening to The Bike Shed, and we'll see you next week.
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
When you're developing applications, it can often be a chore to work with your underlying data. Studio 3T equips you with a complete set of tools to work with MongoDB data. From building queries with drag and drop, to creating complex aggregation pipelines; Studio 3T makes it easy.
And now, there's Studio 3T Free, a free edition of Studio 3T which delivers an essential core of tools. This means you can get started, for free, with Studio 3T Free and when you're ready, you can upgrade and enjoy even more features through Studio 3T Pro and Studio 3T Ultimate. The different editions unlock more tools and additional integrations with Mongo DB, SQL, Oracle and Sybase.
You can start today by downloading Studio 3T Free, which also includes a 30 day free trial of all the features of Studio 3T Ultimate, so you can try out some of the enterprise features as well. No credit card required.
Scout APM is leading-edge application performance monitoring designed to help Rails developers quickly find and fix performance issues without having to deal with the headache or overhead of enterprise-platform feature bloat.
With a developer-centric UI and tracing logic that ties bottlenecks to source code, you can quickly pinpoint and resolve performance abnormalities -- like N+1 queries, slow database queries, memory bloat, and more.
Scout's real-time alerting and weekly digest emails let you rest easy knowing Scout's on watch and resolving performance issues before your customers ever see them.
Scout has also launched its new error monitoring feature add-on for Python applications. Now you can connect your error reporting and application monitoring data on one platform.
See for yourself why developers worldwide call Scout their best friend and try our error monitoring and APM free for 14-days, no credit card needed!
And as an added bonus for Bikeshed listeners: Scout will donate $5 to the open-source project of your choice when you deploy.
Learn more at scoutapm.com/bikeshed.