How to Write a Test That Is Always Green No Matter What the Output Is

It's gardening season! Stephanie swaps seeds with friends and talks about her Chicago garden. Joël recently started experimenting with a dedicated bookmark manager.

They discuss the aspirational (and sometimes dogmatic) sides of TDD and explore when to test: first or after.

How does that affect the tests?
How does that affect the code?
How does that affect workflow?
Are you a "better" programmer because you 100% TDD?

---

This episode is brought to you by Airbrake. Visit Frictionless error monitoring and performance insight for your app stack.

---

• Cassidy William's Productivity tools
• raindrop.io Bookmark Manager
• Simplifying Tests by Extracting Side-Effects

Transcript:

JOËL: Hello and welcome to another episode of The Bike Shed, a weekly podcast from your friends at thoughtbot about developing great software. I'm Joël Quenneville.

STEPHANIE: And I'm Stephanie Minn. And together, we're here to share a bit of what we've learned along the way.

JOËL: So, Stephanie, what is new in your world?

STEPHANIE: It's gardening season here in Chicago. So right now, it is like mid-April as we're recording this, and we are just starting to get some warm weather. And this is usually the time that I do my garden planning for the season. And the other week, I went over to a friend's place, and we did a bit of a seed share. So we just each have collected fruit and vegetable seeds and herbs and all that.

And a really fun way to collect more things to grow is to share with your friends. Seeds are super cheap, but I feel like you could just have like an infinite amount for all of the things that you might want to grow. And so it's really nice to be able to, yes, spread that gardening love around and share with your friends.

JOËL: I'm imagining something like people trading collectible trading cards but the plant version.

STEPHANIE: Yeah, exactly. The fun thing that we did, my friend and I, because, you know, you usually get a little envelope with between 10 and 50 or more seeds, and they're super tiny. Some of them are really teeny tiny, like with broccoli, for example, it's like I can't even explain. It's less than a millimeter, I swear. It's very easy to just lose them, so you want to keep them contained.

But because we are sharing, we don't have a second envelope for the other person to take home with them. And so we actually made our own little envelopes with some origami paper that she had. And we folded it and stapled it and made it very cute. And so I came home with a bunch of these very adorable handmade envelopes with all of my new seeds.

JOËL: Are you mostly doing vegetables, or are these flowers?

STEPHANIE: Yeah, so we mostly focus on vegetables for our garden. And we do like to sprinkle some flower seeds in our yard. But that is more just like throw some seeds out there, and whatever happens to them happens. But with the vegetables, we put a little bit more effort because we usually try to have a good yield.

So in past years, that has meant starting seeds indoors because, in Chicago, we have a shorter growing season than some warmer climate places. And the late summer vegetables like tomatoes, peppers those usually take a little bit longer. So if you want to get a good yield, you might want to start them inside a little early before it's warm enough for them to go outside.

JOËL: So, do you have a garden plot out in your yard, or do you have a community garden plot? How does that work?

STEPHANIE: I am really grateful to have a bit of backyard space. And we have three raised beds that we built that cover...I think each one is 3 feet by 10 feet, so quite a good amount of space. Yeah, we're able to grow a lot of food. Our highlights include shishito peppers. That's one that I really like to grow myself a lot because I usually don't see them in stores as frequently. We grow really great eggplants. Tomatoes, obviously, is a pretty popular beginner-friendly vegetable plant. And we like to grow a lot because then we can process it all and can some of it so we can have nice tomato sauce that's homegrown year round.

JOËL: Hmm, sounds delicious. Do you experiment with the different varieties?

STEPHANIE: We do. That's also a way that the seed sharing is really helpful because maybe I'll get some varieties of certain vegetables like cucumbers or whatever, but maybe my friend has a different kind. And I think we try to do a mix of growing the varieties that we know we like and then experimenting with some ones that are new to us.

JOËL: It's hard to beat fresh vegetables in the summer.

STEPHANIE: Yeah. I'm very excited, especially because during the fall and winter seasons here in Chicago, our local food is a little less exciting. It still can be good, but it's been a lot of root vegetables and the like when we try to eat seasonally in the other season. So I'm really looking forward to stuff that's just juicy and fresh, and it's just one of my biggest joys during the summer. What about you, Joël, what's new in your world?

JOËL: I've recently started experimenting with a dedicated bookmark manager. This is not because I have been to too many bookstores and have all the free bookmarks they give you. These are the digital bookmarks to websites, and I've been really bad at managing those. I mostly just memorize the keywords I need to Google to get access to that website, which is a terrible way of doing things.

And then I've got a mix of a few different browsers, which I don't sync, and have a couple of bookmarks. I use a little bit of Pocket, which is a tool by Mozilla. It's all right, but the search capabilities are not very good. So sometimes I'll know it's in there, but I can't find it.

STEPHANIE: I'm so glad you brought up this topic because I am in a similar boat where I read a lot of things on the internet and have just thrown them all into my top-level bookmarks hierarchy. And that has not really been working for me, either. So I'm really curious to find out how you've been solving this problem.

JOËL: So recently, I volunteered to be a mentor for first-time speakers at the upcoming RailsConf in Atlanta. And someone was asking me about designing slides, and we were talking a little bit about when should you use maybe a bulleted list on a slide versus when there are other options available. I knew that I had read years ago a fantastic resource on slide design. But try as I could, I could not Google this and get the page that I was looking for.

This was shared to me by somebody else as part of a conference preparation group years ago, and so I reached out to this person. I was like, "Hey, so do you happen to remember that link you shared with me five years ago?" And this person says, "I do remember it. I don't have the link either."

STEPHANIE: I've literally been in this exact same situation where I remembered that there was an article that I read, and I remembered exactly who shared it with me or who I talked about it with, and when I couldn't find it, trying to reach out to them and also not being able to find it through them.

JOËL: So the story ends well because I was able to log into an old Slack group...

STEPHANIE: Wow.

JOËL: That had been created for the speakers at this conference and dig through the history. And luckily, I still had access to the group. I was still in that private channel for the speakers. And I found the link, and I was able to share it with others. So that was great. But then I started thinking; I can't keep living this way. I need something better.

STEPHANIE: It's true. Even though we are expert Googlers as developers, sometimes the search just doesn't get you the thing you're looking for.

JOËL: So, about this time, I'm scrolling Twitter as one does. And I saw a tweet from Cassidy Williams talking about some of the productivity tools that she's been using this year did a longer article about it. And I started reading it, and a tool that she mentioned there is Raindrop.io, which is an all-in-one bookmark manager. And I'm like, oh, that is exactly, I think, the missing piece of technology in my life right now. So I went up and signed up for it, and so far, it's been pretty good. I'm experimenting with it.

But I've consolidated a lot of the links that were in my head or in some of these other places, put it in there, categorized them a little bit, tagged them. And hopefully, this becomes a better way so that when I want to reference a link for someone else either in a conversation or as a resource or even for myself, maybe when I'm writing an article, I'm like, oh, I know I read something that would act as a good resource here. I can go to Raindrop and get that article without any of these other shenanigans I had to do this time.

STEPHANIE: Amazing. What is special about Raindrop as opposed to just your native browser bookmark capabilities?

JOËL: It has some deeper structuring capabilities in terms of not everything has to be hierarchical. It has tags as well as categories. And I think most importantly, for me, it has search, which seems to be pretty good at surfacing things. It also has some somewhat smart capabilities where it will automatically figure out if the thing that you've linked is an article, or a document, or a video, something like that. So you can filter by these inferred types as well. It has the ability to sync across devices, which browsers can do if you're signed up for them.

STEPHANIE: Nice. I like that it has that search functionality that you mentioned because I think I'm definitely in the boat of just scrolling through all of my untagged, unorganized bookmarks. And it's really tough to find what I'm looking for, especially if the meta title also doesn't quite tell me exactly the keywords that I'm needing to be scanning for in that moment. So I will definitely have to give it a try.

JOËL: I believe you can get full-text search if you pay for the premium version (I'm currently trying the free version.), which in theory, could mean that it searches the contents of the article. It's not clear with that. But I do know they save a snapshot of the text of the article.

STEPHANIE: That's really interesting because then it's almost like a search engine but scoped to the things that you have saved.

JOËL: Yes.

STEPHANIE: Nice.

JOËL: I'll see how that goes, and maybe six months from now, I can talk a little bit about what the experience has been using that.

STEPHANIE: Yeah, six months from now, you can tell us all about how you have no issues or qualms with how you've been managing bookmarks because everything is working perfectly well for you. [laughs]

JOËL: JK, I've dropped this whole bookmark thing.

STEPHANIE: That's true. That's also the flip side of trying out a new tool, [laughs], isn't it?

MID-ROLL AD:

Debugging errors can be a developer’s worst nightmare...but it doesn’t have to be. Airbrake is an award-winning error monitoring, performance, and deployment tracking tool created by developers for developers that can actually help cut your debugging time in half.

So why do developers love Airbrake? It has all of the information that web developers need to monitor their application - including error management, performance insights, and deploy tracking!

Airbrake’s debugging tool catches all of your project errors, intelligently groups them, and points you to the issue in the code so you can quickly fix the bug before customers are impacted.

In addition to stellar error monitoring, Airbrake’s lightweight APM helps developers to track the performance and availability of their application through metrics like HTTP requests, response times, error occurrences, and user satisfaction.

Finally, Airbrake Deploy Tracking helps developers track trends, fix bad deploys, and improve code quality.

Since 2008, Airbrake has been a staple in the Ruby community and has grown to cover all major programming languages. Airbrake seamlessly integrates with your favorite apps to include modern features like single sign-on and SDK-based installation. From testing to production, Airbrake notifiers have your back.

Your time is valuable, so why waste it combing through logs, waiting for user reports, or retrofitting other tools to monitor your application? You literally have nothing to lose. Head on over to airbrake.io/try/bikeshed to create your FREE developer account today!

STEPHANIE: So our topic for today is something that we've had in our topic backlog for a while, and I'm excited to talk about it. It's TDD, which I think is a very well-known, potentially controversial topic of discussion in the world of software development. And specifically, we wanted to talk about when TDD is useful and when you actually might also have some value in writing tests afterwards. And in preparation for this topic today, I actually have been TDDing most of my client work this week.

JOËL: What? You're telling me you don't TDD 100% of the time? Are you even a real developer?

STEPHANIE: I am a real developer, and I do not TDD 100% of the time. I'm just going to say it. It's on record.

JOËL: You know what? Me too.

STEPHANIE: Wow, I'm glad we could clear the air on that one. [laughs]

JOËL: What percent of the time would you say that you do TDD? In this case, test first as opposed to maybe testing after you've written the code or maybe not testing at all.

STEPHANIE: Hmm, that's an interesting question. A part of me wants to answer it in my ideal workflow terms. But I think that is less interesting than reality, which is I will usually at least try to test first if I'm feeling like I am up for it. So maybe the percentage is, I don't know, I really couldn't tell you, but I'm just going to throw out 40% of the time [laughs] because that seems pretty, I don't know, reasonable. Sometimes you wake up, and you're just like, I'm not going to do it today. [laughs] And other days, you wake up, and you're like, you know? It sounds like a fun exercise to do for this particular feature.

So yeah, if I TDD 40% of the time, then I think maybe I write tests after another 40% or 50%. And then [laughs] I'm hesitant to say this on the air, but sometimes you code, and you don't write tests for it and would not recommend it for the majority of your work. But I'm just going to be real here that sometimes it happens.

JOËL: It's always a trade-off in terms of the work you put in versus the value you're getting out of it. And sometimes, you get very little value out of a test.

STEPHANIE: Yeah, that's real. It totally depends on what you're doing.

JOËL: I think one thing that's interesting for us, because we're consultants, so we move from one project to another, is that some projects are set up in a way that they're very test friendly. It's easy to have a testing workflow with them. And then others are just incredibly painful to test because of the way the system has been architected.

And I think a TDD purist would then tell us that this is a symptom of high coupling or other architectural problems; that's probably true. But also, you don't have time to re-architect the entire system, and so then it becomes a question of trade-offs. Can I test some things easily today? Can I refactor a few things that will make this local change somewhat easier to test? And then, where is it not worth the effort to make something testable?

STEPHANIE: Yeah, I've definitely struggled with that, where a part of me wanted to test something very thoroughly or even do test-driven development and then ran into some obstacles along the way and having to be realistic about that effort.

The other thing I was referring to around it depending is also the actual code you're working on. So maybe if you're just writing a script or something to automate some dev workflow, it's okay for that not to be tested. And I also do think that the decision to TDD is very dependent on whether you are writing net new code, or refactoring, or having to deal with legacy code.

JOËL: That definitely makes a difference. For me, when I'm refactoring in the purest sense, changing structure without changing behavior, in theory, I should not be writing tests for that because there should already be existing tests, and I'm not changing behaviors. So the test suite should prove that my changes did not change behavior. In practice, oftentimes, there is not the coverage that needs to be there. I don't know about you; I feel like I often don't trust the code enough, where I'm a little bit scared to do a refactor if there isn't test coverage. How about you?

STEPHANIE: I've been running into that issue a lot on my current client project where I've been making an intentional effort to add test coverage before I make any changes because that forces me to really understand how things work because either I read a piece of code and I just can't tell at all. Or I learn later on that I thought I understood something based on the class or the method names, but it turns out that there was actually some nuance in there or side effects or what have you that belied my understanding [laughs] of what it was doing.

And after a few times of that lack of trust that you talked about popping up, I was like, okay, I think, at least for me, the way that I can feel good about the work that I'm doing is to set myself up for success in that way.

JOËL: Do you ever find that the code that you write in a test-driven approach tends to end up different than code that you might write with a test-after approach?

STEPHANIE: Yeah, I think this can actually be answered at a few different levels but let me start with talking about how I like to practice TDD. If I'm given a user story, I usually try to work outside in. So I will write a feature or acceptance test and that involves testing how the user would interact with our application.

At that point, I will usually go with the most naive implementation to get the test passing, and so it probably won't look pretty. In a recent case, I was adding a new parameter to a controller, and I just put everything in the controller to get the test green. [laughs] And then, at that point, is when I gave that code a second pass and looked for areas to extract where I could.

JOËL: That refactor step and the red, green refactor cycle is really important.

STEPHANIE: Yeah, absolutely. I think that is where I find TDD to be the most valuable from that higher-level perspective. And I know that there are different schools of thought on this. But that helps ensure that at least I have written the code to make the feature work the way that I was hoping. And I use TDD less for driving design decisions just because I like to have something to react to that is helpful for me rather than having a blank slate of, okay, let me write a test with an idea about how an object's interface will look. And so that's what works for me.

So I do think it's kind of a mix of like from an acceptance test level; I am at least writing code that I know works, but the shape of the code for me is less determined by how I test.

JOËL: So when you're looking at code that you've written six months down the line, you generally can't tell the difference whether it was test-driven or written first and tested after.

STEPHANIE: That sounds right. That's just how my process works. In fact, I think recently we, to go on a quick tangent, we talked about writing conference talks, and I think I even mentioned for me the process is looking at the thing and then revising. And I think that the design driving element of TDD that a lot of people like is a bit less effective for me personally.

JOËL: Hmm. Would you say that TDD does not impact the shape of the code that you end up creating in response to the tests? Or when you're talking about design, are you mostly thinking in terms of the interface that you would have in the test itself, like, what arguments the constructor takes or things like that?

STEPHANIE: I think I was talking more about the latter, the interface, the construction arguments. When I do test afterwards, I also will notice the way the setup of my test how that is feeling. And if it is feeling a bit unwieldy or is a bit complicated, that will cue me to maybe take another pass at the code itself. So that's actually one way that testing after can signal to me a way that I might want to change my code.

JOËL: Okay, so you're getting some of those pressures that you get from testing, but you respond to them in a like second path?

STEPHANIE: Yeah, I think so. I'm curious how you TDD and whether you notice changes in how your code looks.

JOËL: I think there are a couple of things that TDD does in my workflow that are really nice. One is it keeps me focused in terms of getting the work done because you're just following from one failure to another. It also keeps me focused in terms of scope. It's really easy for my engineering brain to be like, oh, we could totally do this thing and all that, and it's like, no, that's not needed to solve the problem at hand.

Because in TDD, you try the smallest solution that will solve your problem, and then you will refactor it to make it maybe nicer to work with. But you try not to add new behavior that's not required in order to pass the test, and that can be a really helpful forcing function for me.

STEPHANIE: That's interesting because I was just thinking about how sometimes, at least with the outside-in approach that I was talking about, I will find that the scope of the ticket is too big as I make changes to get the desired quality of the code that I want.

Like I mentioned, the naive implementation, like, sure, maybe everything is in a controller, but as soon as I'm starting to do that second pass, and I want to maybe change another class and to make it work for my needs, I will notice it start to sprawl a little bit. And that is usually a signal to me that, like, oh, maybe what I need first is just refactoring the objects that I'm hoping to use to get the desired implementation. And that ends up being a separate PR that I do first to then set myself up for making the change.

JOËL: The classic make the change easy before you then go and make the easy change.

STEPHANIE: Right. But that does mean that that initial feature test that I wrote won't ever be green. So I do have to kind of like back out of making that change and just be like, okay, today is not the day [laughs] that I'm going to get this feature working.

JOËL: There are some times where I'm in a situation like that, and I will kind of recognize, oh, there's a refactor step that's happening right now as a sort of subtask. And so, I will make that refactor change that I need to and then commit only those files that were a part of that refactor and may be included as part of the PR with the feature change or maybe push it up and make it its own PR.

But depending on what the refactor is, oftentimes, I can kind of do it sort of all more or less continuously but decide once I've done that refactor step, okay, commit time but only those files for the smaller set of changes, and then keep moving with that outside-in approach.

One thing I have noticed about the style of code that I tend to produce when I TDD versus when I don't is how I will tend to decouple things. And so because coupled code is really annoying to test in isolation, TDD sort of forces me to do more dependency injection, passing objects to others. It will often force me or maybe not force me, but it gives me that wholesome pressure to maybe separate HTTP requests from more of the business logic in my code, which otherwise I might completely intermix because it's just so convenient. Even certain things like class methods, I might tend to overuse them or use them more if I'm not test driving than if I were.

STEPHANIE: When you talk about coupling, I'm curious, do you end up mocking a lot in the tests that you are writing to drive your development?

JOËL: No, but if I'm testing after, I probably will. Mocking, I think, is a sign of coupling generally. In tests where you're just passing objects to each other, generally, you can get away with passing in a test double or something, whereas if you're hard-coding dependencies, you often have to mock.

STEPHANIE: Got it. That makes a lot more sense now. I think that does require a bit of thought upfront about what kinds of objects you might need and what they would provide for you in the thing that you're testing.

JOËL: Yes. There's definitely a phase where let's say; I'm testing some kind of third-party integration; I'm just kind of trying to do it all in one object that has a mix of business logic and some HTTP request stuff. It gets really annoying as we're adding...maybe the first feature is okay. I use WebMock, and I stub out a request, and it's good. And then the second one, I feel like I'm kind of duplicating that. And then the third one, I've got to deal with retries.

So now I've got to go back to the first one and add some two or three WebMocks because now we've got exponential backoff code that's happening here. And this new feature broke the old tests. And it just becomes this really annoying thing to do. And then I might start thinking, okay, how do I separate these two things? I have one place where I test the HTTP logic, the exponential backoff, the what to do if I get a 404 from the API. And then, separately, I can just have the business logic and test all of those branches there without having to touch any of the HTTP stuff.

I think you could get there from a few different paths. So you could get there by sort of following a lot of classic design principles, things like SOLID, because they kind of converge on that general idea as well. You could even get there if you took more of a functional programming approach where you are really good at separating side-effectful code from, I'm going to use the term loosely here, pure functions.

I've heard some people make the distinction between IO versus non-IO in code and how that affects the types of tests that you write for them. And separating those two is a thing that you might do, even if you weren't writing tests at all, if that's a design principle that you know to follow.

STEPHANIE: Yeah, that's a great point. I was thinking, as you were talking about your approach for handling that potential feature with talking to a third party, that I've heard that particular task or problem in software development used as an example for a lot of those different techniques or strategies that you mentioned. And I suppose TDD really is just a tool, and it doesn't replace your experience or intuition.

And earlier, when we were talking about times that you don't do TDD, I will have to say that if I am doing something that I've done many times before, I feel confident enough that I don't need to lean on that red, green refactor cycle. At that point, it's more muscle memory. And maybe I do forget a step along the way, but I have the experience to know how to debug that or to see the error and know exactly what it was that I did wrong. And in that case, I am tapping into something different than using TDD.

JOËL: I think definitely, for a lot of things now, there are patterns that I have learned where even if I weren't TDDing, I might do a third-party integration using this pattern because I've done it via TDD enough to know that this is a structure that I find works very well in terms of the coupling of things. And then maybe if I want to fill in some tests afterwards, then I'll thank my past self that I'm using a pattern that plays nicely with that. One thing that I do notice happens sometimes is that when people add tests after the fact, they will add tests that are green but that don't necessarily fail if the code breaks. Have you ever seen that?

STEPHANIE: I have seen that before. In fact, I just saw it recently where we had a false positive test. And I made a change expecting the test to fail, and it didn't, which is not great because the value that tests have are when they fail, you want to be alerted when something goes wrong. Just because they're green doesn't mean that everything works. It just means that they didn't detect a problem. And in this particular case, I don't know if the developer who wrote this test had TDDed or not. But I did notice that in the test, we were mocking a method, and that ended up being the cause of the false positive.

JOËL: I'm always a little bit skeptical of mocks because I feel like I've seen so many either brittle tests or tests that will succeed all the time come out of mocks. I don't know if you've ever heard the term tautological test or a test that is a tautology.

STEPHANIE: No, I haven't. What does that mean?

JOËL: In its sort of most basic sense, it's a test that is always green no matter what the output is. Some people think of it more in terms of self-referential tests, like, oh, a thing equals itself, which, yes, it does, and those tend to be always green. But it's not always self-referential. It can be some other subtle ways. Typically this happens when mocking or specifically if you mock the system under test. It's very easy to write a test that is now going to always be green, no matter how the code changes.

A fun fact about the word tautology is it comes from discrete math, which is the topic of my RailsConf talk. If you write out a truth table that shows all the possible inputs and whether or not something will be true or false, depending on what the inputs are, the output column is all true in a tautology, which tells you that no matter what the inputs are, you're going to get true out of that method or function or equation. And so, if this was a Boolean expression in Ruby, you could replace that by hardcoding true and get the same result.

STEPHANIE: Yeah, that's what I was imagining, a function that just returns true. [laughs]

JOËL: And that's effectively what you can accidentally write when you're creating a test that is a tautological test is one where you could have just replaced the entire thing with expect true to be true, and it would have the same effect. And, like you said, tests only have value when they fail. And a test that never fails has no value. So TDD has this red, green refactor cycle. I feel like you could probably come up with a cute slogan like that for a testing-after style. So maybe I guess you'd start off you write some code, then you write a test that theoretically passes for it.

So you start green, but then you want to make sure you see that test fail, so you got to go red and then comment out the code or something. Then comment it back in to see that it goes back to green to make sure that not only does this test fail when the code is broken but also that bringing this test back is what makes it pass, which is an important distinction. So maybe it's a green, red, green, and then maybe refactor. Because one thing that I admired in the style that you were talking about earlier is that even when you test after, you include a refactor step. The test at the end is not the final step in your workflow.

STEPHANIE: Yeah, that's a really good point. When you said green, red, green, I was thinking of a Christmas garland [laughs] or something like that. But yeah, I do think that stuff gets skipped sometimes. If you are testing after, you're backfilling tests for code you wrote, and at that point, you think you know how it will work, and so you're writing your tests kind of colored with that in mind.

I like the injecting commenting something out or changing an input or something that you know should make the test fail, just so that you can confirm that you didn't just write a test that expects true to equal true or give you a false positive like that, then go back to green. And as you were saying that, it did make me think like, oh, well, that's like a whole extra step as opposed to TDD where we do just have red, green refactor. We don't have that extra step. But I think the effort is just like put in at a different point in time.

JOËL: Agreed. It's important that you see the code fail and that you see it pass after the change. The order has changed a little bit, but those two kinds of core elements are present. Kind of by default, you have no choice when you're doing TDD. You have the ability to skip that if you're testing after, but ideally, you incorporate those in a robust test after workflow as well.

STEPHANIE: Yeah. And I know I mentioned times when I've done something enough or used a pattern enough that maybe I'll just go ahead and implement it and then backfill with tests. And I also recognize that in those moments, I could have done something wrong, that there is some amount of wanting to check that the test failed. And I imagine there is some kind of balance to achieve there between the speed that you get by having that experience and knowing the direction you want to take things and applying a pattern that you've done a lot with being like, oh, we're all human, and sometimes we make mistakes.

JOËL: In a situation where you feel like you're coding something that you've coded up 100 times before, you're very familiar with this. Do you find that a test after workflow is faster for you?

STEPHANIE: Hmm, that's an interesting question.

JOËL: Because I think that's often a motivation. It's like, I don't want to bother, like, I just have the idea. I know what to do. Let me just write that code and get it done.

STEPHANIE: I think if I were introducing a new route or controller action or whatever, I don't need to go through the cycle of writing a test and it failing because I haven't added the action to the controller yet. It's like I know that that is the next logical step, and so maybe I might skip it there. But if I'm at the point where I'm working with business or domain logic, I think that's where is the value of test writing first because it's like, I passed the framework and passed my tools. And now, I'm working at working through the logic of the business problem itself.

JOËL: So you're working in maybe slightly larger iterations of that red, green refactor cycle.

STEPHANIE: Yeah, that's a good way to describe it.

JOËL: I was recently working on a gem and tried to TDD it from scratch and went with micro iterations. And it was actually really fun, and there was a flow to it. And this is a greenfield side project. And it helped me stay focused. I think it did give me a decent design. I really enjoyed it.

STEPHANIE: Nice. Was there something satisfying about seeing that green each time and kind of doing that bit of mechanical labor? And I can see how that can feel almost meditative.

JOËL: Yes. And I think also because this was a problem that I didn't fully know how I was going to solve, TDD helped really focus me on solving sub-parts of that problem, things that I can hold in my head and solve in a minimalistic way and then iterate on.

STEPHANIE: I like that a lot. You were using that technique, and that really helped for the task at hand, which was, in this case, a bit smaller in scope. I think the way you and I have been talking about TDD has been very realistic and very reasonable. And I'm curious what you think about people who kind of use it as the pinnacle of how you should write code.

JOËL: I think that's really interesting because TDD is a really wonderful technique, and I wish more people used it. But it's kind of taken on a mystique of its own where if you do TDD or claim to do TDD 100% of the time, now, all of a sudden, you've put yourself on another level. And I think people even who choose for pragmatic reasons not to TDD all the time maybe feel a little bit of guilt or at least feel the need to explain themselves to other people to say, "Hey, I didn't TDD this here. Well, let me explain to you why that's okay, and I'm not a bad programmer."

STEPHANIE: Yeah, I think we even alluded a little bit to that earlier in the show, and I could hear my hesitancy to be like, oh, I guess I'm going to say this and have all these people hear it. But I think that's a good point that it's okay for you not to do it 100% of the time. That doesn't make you any less of a programmer. Also, the way we've been talking about it also makes it sound like one of those things where it's like you do have to learn the rules before you can break them. And so there is value in learning it and doing it.

And then you also, after having done it enough, know when you want to use it or when you don't. My advice for folks who haven't really done it before or don't quite see the value of it is just to try it and then decide for yourself. I think at the end of the day, we should all feel empowered to be able to decide how we work best.

JOËL: It's also really valuable, I think, to maybe pair with someone who is really good at that and to get to see what their workflow is like. Oftentimes, there's almost a hump of getting into it where you are more productive without TDDing because you're not comfortable with the flow or with the techniques. And it takes a lot of expertise to get over that hump where maybe at the expert end of things, you are more productive with it. And on the less expert end, it just becomes a chore that takes up all of your time or ends up giving you results that aren't that great anyway. And so, how do you cross that chasm?

STEPHANIE: Yeah, that's a really, really great point because, in some ways, when you first get started, it will feel slow. You are unlearning the ways that you have known to code before and trying to do it in a different way. And I really like your advice about trying to pair with someone who has expertise or has been practicing it for a long time. That was my first real introduction to it too. At this point, I had been a few years into my career and hadn't really tried it because it seemed very daunting, and seeing someone else verbalize their process and seeing their workflow was really helpful for me to get on board.

JOËL: I think what I would like is for TDD to be a tool that is aspirational for a lot of people. If you're new to the technique and you've paired with somebody who's really good at it, and you see the flow that they have and be like, wow, that's really good. I would love to get that incorporated in the way that I work. Rather than a sort of measuring stick for how elite of a programmer you are. There's no sense in shaming people over the tools they use.

STEPHANIE: Right. Because it should also be accessible, and if you make people feel bad about it, and then it's not accessible to folks.

JOËL: On that note. Shall we wrap up?

STEPHANIE: Yeah, let's wrap up. Show notes for this episode can be found at bikeshed.fm.

JOËL: This show has been produced and edited by Mandy Moore.

STEPHANIE: If you enjoyed listening, one really easy way to support the show is to leave us a quick rating or even a review in iTunes. It really helps other folks find the show.

JOËL: If you have any feedback for this or any of our other episodes, you can reach us @_bikeshed, or you can reach me @joelquen on Twitter.

STEPHANIE: Or reach both of us at hosts@bikeshed.fm via email.

JOËL: Thanks so much for listening to The Bike Shed, and we'll see you next week.

ALL: Byeeeeeee!!!!!!

ANNOUNCER: This podcast is brought to you by thoughtbot, your expert strategy, design, development, and product management partner. We bring digital products from idea to success and teach you how because we care. Learn more at thoughtbot.com.

Sponsored By:

• Airbrake: Deploy fearlessly and fix bugs faster with Airbrake Error & Performance Monitoring. Airbrake notifiers are available for all major programming languages and frameworks, and install in minutes, with an open-source SDK-based install and near-zero technical debt. Spend less time tracking down bugs and more time developing. Visit Frictionless error monitoring and performance insight for your app stack.

Support The Bike Shed