415: Codebase Calibration

Episode 415 · February 6th, 2024 · 30 mins 54 secs

About this Episode

Stephanie has a delightful and cute Ruby thing to share: Honeybadger, the error monitoring service, has created exceptionalcreatures.com, where they've illustrated and characterized various common Ruby errors into little monsters, and they're adorable. Meanwhile, Joël encourages folks to submit proposals for RailsConf.

Together, Stephanie and Joël delve into the nuances of adapting to and working within new codebases, akin to aligning with a shared mental model or vision. They ponder several vital questions that every developer faces when encountering a new project: the balance between exploring a codebase to understand its structure and diving straight into tasks, the decision-making process behind adopting new patterns versus adhering to established ones, and the strategies teams can employ to assist developers who are familiarizing themselves with a new environment.

Honeybadger's Exceptional Creatures
RailsConf CFP coaching sessions
Support and Maintenance Episode


 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's new in your world?

STEPHANIE: I have a delightful and cute Ruby thing to share I'd seen just in our internal company Slack. Honeybadger, the error monitoring service, has created a cute little webpage called exceptionalcreatures.com, where they've basically illustrated and characterized various common Ruby errors into little monsters [laughs], and I find them adorable. I think their goal is also to make it a really helpful resource for people encountering these kinds of errors, learning about them for the first time, and figuring how to triage or debug them.

And I just think it's a really cool way of, like, making it super approachable, debugging and, you know, when you first encounter a scary error message, can be really overwhelming, and then Googling about it can also be equally [chuckles] overwhelming. So, I just really liked the whimsy that they kind of injected into something that could be really hard to learn about. Like, there are so many different error messages in Ruby and in Rails and whatever other libraries you're using. And so, that's kind of a...I think they've created a one-stop shop for, you know, figuring out how to move forward with common errors.

And I also like that it's a bit of a collective effort. They're calling it, like, a bestiary for all the little creatures [laughs] that they've discovered. And I think you can, like, submit your own favorite Ruby error and any guidance you might have for someone trying to debug it.

JOËL: That's adorable. It reminds me a little bit of HTTP status codes as cat memes site. It has that same energy. One thing that I think is really interesting is that because it's Honeybadger, they have stats on, like, frequency of these errors, and a lot of these ones are tied to...I think they're picking some of the most commonly surfaced errors.

STEPHANIE: Yeah, there's little, like, ratings, too, for how frequently they occur, kind of just like, I don't know, Pokémon [laughs] [inaudible 02:31]. I think it's really neat that they're using something like a learning from their business or maybe even some, like, proprietary information and sharing it with the world so that we can learn from it.

JOËL: I think one thing that's worth specifying as well is that these are specific exception classes that get raised. So, they're not just, like, random error strings that you see in the wild. They don't often have a whole lot of documentation around them, so it's nice to see a dedicated page for each and a little bit of maybe how this is used in the real world versus maybe how they were designed to be used. Maybe there's a line or two in the docs about, you know, core Ruby when a NoMethodError should be raised. How does NoMethodError actually get used, you know, in real life, and the exceptions that Honeybadger is capturing. That's really interesting to see.

STEPHANIE: Yeah, I like how each page for the exception class, and I'm glad you made that distinction, is kind of, like, crowdsourced guidance and information from the community, so I think you could even, you know, contribute to it if you wanted. But yeah, just a fun, little website to bring you some delight when you're on your next head-smacking, debugging adventure [laughs].

JOËL: And I love that it brings some joy to the topic, but, honestly, I think it's a pretty good reference. I could see myself linking to this anytime I want to have a deeper discussion on exceptions. So, maybe there's a code review, and maybe I want to suggest that we raise a different error than the one that we're doing. I could see myself in that GitHub comment being like, "Oh, instead of, you know, raising an exception here, why don't we instead raise a NoMethodError or something like that?" And then link to the bestiary page.

STEPHANIE: So, Joël, what's new in your world?

JOËL: So, just recently, RailsConf announced their call for proposals. It's a fairly short period this year, only about three-ish weeks long. So, I've been really encouraging colleagues to submit and trying to be a resource for people who are interested in speaking at conferences. We did a Q&A session with a fellow thoughtboter, Aji Slater, who's also a former RailsConf speaker, about what makes for a good talk, what is it like to submit to a call for proposals, you know, kind of everything from the process from having an idea all the way to stage presence and delivering. And there's a lot of great questions that got asked and some good discussion that happened there.

STEPHANIE: Nice. Yeah, I think I have noticed that you are doing a lot more to help, especially first-time speakers give their first conference talk this year. And I'm wondering if there's anything you've learned or any hopes and dreams you have for kind of the amount of time you're investing into supporting others.

JOËL: What I'd like to see is a lot of people submitting proposals; that's always a great thing. And, a proposal, even if it doesn't get accepted, is a thing that you can resubmit. And so, having gone through the effort of building a proposal and especially getting it maybe peer-reviewed by some colleagues to polish your idea, I think is already just a really great exercise, and it's one that you can shop around. It's one that you can maybe convert into a blog post if you need to. You can convert that into some kind of podcast appearance. So, I think it's a great way to take an idea you're excited about and focus it, even if you can't get into RailsConf.

STEPHANIE: I really like that metric for success. It reminds me of a writer friend I have who actually was a guest on the show, Nicole Zhu. She submits a lot of short stories to magazines and applications to writing fellowships, and she celebrates every rejection. I think at the end of the year, she, like, celebrates herself for having received, you know, like, 15 rejections or something that year because that meant that she just went for it and, you know, did the hard part of doing the work, putting yourself out there. And that is just as important, you know, if not more than whatever achievement or goal or the idea of having something accepted.

JOËL: Yeah, I have to admit; rejection hurts. It's not a fun thing to go through. But I think even if you sort of make it to that final stage of having written a proposal and it gets rejected, you get a lot of value out of that journey sort of regardless of whether you get accepted or not. So, I encourage more people to do that.

To any of our listeners who are interested, the RailsConf call for proposals goes through February 13th, 2024. So, if you are listening before then and are inspired, I recommend submitting. If you're unsure of what makes for a good CFP, RailsConf is currently offering coaching sessions to help craft better proposals. They have one on February 5th, one on February 6th, and one on February 7th, so those are also options to look into if this is maybe your first time and you're not sure. There's a signup form. We'll link to it in the show notes.

STEPHANIE: So, another update I have that I'm excited to get into for the rest of the episode is my recent work on our support and maintenance team, which I've talked about on the show before. But for any listeners who don't know, it's a kind of sub-team at thoughtbot that is focused on helping maintain multiple client projects at a time. But, at this point, you know, there's not as much active feature development, but the work is focused on keeping the codebase up to date, making any dependency upgrades, fixing any bugs that come up, and general support. So, clients have a team to kind of address those things as they come up.

And when I had last talked about it on the podcast, I was really excited because it was a bit of a different way of working. I felt like it was very novel to be, you know, have a lot of different projects and domains to be getting into. And knowing that I was working on this team, like, short-term and, you know, it may not be me in the future continuing what I might have started during my rotation, I thought it was really interesting to be optimizing towards, like, completion of a task. And that had kind of changed my workflow a bit and my process.

JOËL: So, now that you've been doing work on the support and maintenance team for a while and you've kind of maybe gotten more comfortable with it, how are you generally feeling about this idea of sort of jumping into new codebases all the time?

STEPHANIE: It is both fun and more challenging than I thought it would be. I tend to actually really enjoy that period of joining a new team or a project and exploring, you know, a codebase and getting up to speed, and that's something that we do a lot as consultants. But I think I started to realize that it's a bit of a tricky balance to figure out how much time should I be spending understanding what this codebase is doing? Like, how much of the application do I need to be understanding, and how much poking around should I be doing before just trying to get started on my first task, the first starter ticket that I'm given?

There's a bit of a balance there because, on one hand, you could just immediately start on the task and kind of just, you know, have your blinders [chuckles] on and not really care too much about what the rest of the code is doing outside of the change that you're trying to make. But that also means that you don't have that context of why certain things are the way they are. Maybe, like, the way that you want to be building something actually won't work because of some unexpected complexity with the app.

So, I think there, you know, needs to be time spent digging around a little bit, but then you could also be digging around for a long time [chuckles] before you feel like, okay, I finally have enough understanding of this new codebase to, like, build a feature exactly how a seasoned developer on the team might.

JOËL: I imagine that probably varies a little bit based on the task that you're doing. So, something like, oh, we want to upgrade this codebase to Ruby 3.3, probably requires you to have a very different understanding of the codebase than there's a bug where submitting a comment double posts it, and you have to dig into that. Both of those require you to understand the application on very different levels and kind of understand different mental models of what the app is doing.

STEPHANIE: Yeah, absolutely. That's a really good point that it can depend on what you are first asked to work on. And, in fact, I actually think that is a good guidepost for where you should be looking because you could develop a mental model that is just completely unrelated [chuckles] to what you're asked to do. And so, I suppose that is, you know, usually a good place to start, at least is like, okay, I have this first task, and there's some understanding and acceptance that, like, the more you work on this codebase, the more you'll explore and discover other parts of it, and that can be on a need to know kind of basis.

JOËL: So, I'm thinking that if you are doing something like a Ruby upgrade or even a Rails upgrade, a lot of what you care about the app is going to be on a more mechanical level. So, you want to know what gems you're using. You want to know what different patterns are being used, maybe how callbacks are happening, any particular features that are version-specific that are being used, things like that.

Whereas if you're, you know, say, fixing a bug, you might care a lot more about some of the product-level concerns. What are we actually trying to do here? What is the expected user experience? How does this deviate from that? What were the underlying mental models of the developers? So, there's almost, like, two lenses you can look at the code. Now, I almost want to make this a two-dimensional thing, where you can look at it either from, like, a very kind of mechanical lens or a product lens in one axis.

And then, on the other axis, you could look at it from a very high-level 10,000-foot view and maybe zoom in a little bit where you need, versus a very localized view; here's where the bug is happening on this page, and then sort of zoom out as necessary. And I could see different sorts of tasks falling in different quadrants there of, do I need a more mechanical view? Do I need a more product-focused view? And do I need to be looking locally versus globally?

STEPHANIE: Wow. I can't believe you just created a Cartesian graph [laughs] for this problem on the fly. But I love it because I do think that actually lines up with different strategies I've taken before. It's like, how much do you even look at the code before deciding that you can't really get a good picture of it, of what the product is, without just poking around from the app itself?

I actually think that I tend to start from the code. Like, maybe I'll see a screenshot that someone has shared of the app, you know, like a bug or something that they want me to fix, and then looking for that text in the code first, and then trying to kind of follow that path, whereas it's also, you know, perfectly viable to try to see the app being used in production, or staging, or something first to get a better understanding of some of the business problems it's trying to solve.

JOËL: When you jump into a new codebase, do you sort of consciously take the time to plan your approach or sort of think about, like, how much knowledge of this new codebase do I need before I can, like, actually look at the problem at hand?

STEPHANIE: Ooh, that's kind of a hard question to answer because I think my experience has told me enough times that it's never what I think it's going to [laughs] be, not never, but it frequently surprises me. It has surprised me enough times that it's kind of hard to know off the bat because it's not...as much as we work in frameworks that have opinions and conventions, a lot of the work that happens is understanding how this particular codebase and team does things and then having to maybe shift or adjust from there.

So, I think I don't do a lot of planning. I don't really have an idea about how much time it'll take me because I can't really know until I dive in a little bit. So, that is usually my first instinct, even if someone is wanting to, like, talk to me about an approach or be, like, "Hey, like, how long do you think this might take based on your experience as a consultant?" This is my first task. Oftentimes, I really can't say until I've had a little bit of downtime to, in some ways, like, acquire the knowledge [chuckles] to figure that out or answer that question.

JOËL: How much knowledge do you like to get upfront about an app before you dive into actually doing the task at hand? Are there any things, like, when you get access to a new codebase, that you'll always want to look at to get a sense of the project before you look at any tickets?

STEPHANIE: I actually start at the model level. Usually, I am curious about what kinds of objects we're working with. In fact, I think that is really helpful for me. They're like building blocks, in order for me to, like, conceptually understand this world that's being represented by a codebase. And I kind of either go outwards or inwards from there. Usually, if there's a model that is, like, calling to me as like, oh, I'll probably need to interact with, then I'll go and seek out, like, where that model is created, maybe through controllers, maybe through background jobs, or something like that, and start to piece together entry points into the application.

I find that helpful because a lot of the times, it can be hard to know whether certain pages or routes are even used at all anymore. They could just be dead code and could be a bit misleading. I've certainly been misled [chuckles] more than once. And so, I think if I'm able to pull out the main domain objects that I notice in a ticket or just hear people talk about on the team, that's usually where I gravitate towards first. What about you? Do you have a place you like to start when it comes to exploring a new codebase like that?

JOËL: The routes file is always a good sort of overview of, like, what is going on in the app. Scanning the models directory is also a great start in a Rails app to get a sense of what is this app about? What are the core nouns in our vocabulary? Another thing that's good to look for in a codebase is what are the big types of patterns that they tend to use?

The Rails ecosystem goes through fads, and, over time, different patterns will be more popular than others. And so, it's often useful to see, oh, is this an app where everything happens in service objects, or is this an app that likes to rely on view components to render their views? Things like that. Once you get a sense of that, you get a little bit of a better sense of how things are architected beyond just the basic MVC.

STEPHANIE: I like that you mentioned fads because I think I can definitely tell, you know, how modern an app is or kind of where it might be stuck in time [chuckles] a little bit based on those patterns and libraries that it's heavily utilizing, which I actually find to be an interesting and kind of challenging position to be in because how do you approach making changes to a codebase that is using a lot of patterns or styles from back in the day? Would you continue following those same patterns, or do you feel motivated to introduce something new or kind of what might be trendy now?

JOËL: This is the boring answer, but it's almost never worth it to, like, rewrite the codebase just to use a new pattern. Just introducing the new pattern in some of the new things means there are now two patterns. That's also not a great outcome for the team. So, without some other compelling reason, I default to using the established patterns.

STEPHANIE: Even if it's something you don't like?

JOËL: Yes. I'm not a huge fan of service objects, but I work in plenty of codebases that have them, and so where it makes sense, I will use service objects there. Service objects are not mutually exclusive with other things, and so sometimes it might make sense to say, look, I don't feel like I can justify a service object here. I'll do this logic in a view, or maybe I'll pull this out into some other object that's not a service object and that can live alongside nicely. But I'm not necessarily introducing a new pattern. I'm just deciding that this particular extraction might not necessarily need a service object.

STEPHANIE: That's an interesting way to describe it, not as a pattern, but as kind of, like, choosing not to use the existing [chuckles] pattern. But that doesn't mean, like, totally shifting the architecture or even how you're asking other people to understand the codebase. And I think I'm in agreement. I'm actually a bit of a follower, too, [laughs], where I want to, I don't know, just make things match a little bit with what's already been created, follow that style. That becomes pretty important to me when integrating with a team in a codebase.

But I actually think that, you know, when you are calibrating to a codebase, you're in a position where you don't have all that baggage and history about how things need to be. And maybe you might be empowered to have a little bit more freedom to question existing patterns or bring some new ideas to the team to, hopefully, like, help the code evolve. I think that's something that I struggle with sometimes is feeling compelled to follow what came before me and also wanting to introduce some new things just to see what the team might think about them.

JOËL: A lot of that can vary depending on what is the pattern you want to introduce and sort of what your role is going to be on that team. But that is something that's nice about someone new coming onto a project. They haven't just sort of accepted that things are the way they are, especially for things that the team already doesn't like but doesn't feel like they have the energy to do anything better about it.

So, maybe you're in a codebase where there's a ton of Ruby code in your ERB templates, and it's not really a pattern that you're following. It's just a thing that's there. It's been sort of the path of least resistance for a long time, and it's easier to add more lines in there, but nobody likes it. New person joins the team, and their naive exuberance is just like, "We can fix it. We can make it better."

And maybe that's, you know, going back and rewriting all of your views. That's probably not the best use of their time. But it could be maybe the first time they have to touch one of these views, cleaning up that one and starting a conversation among the team. "Hey, here are some patterns that we might like to clean up some of these views instead," or "Here are maybe some guidelines for anything new that we write that we want to do to keep our views clean," and sort of start moving the needle in a positive direction.

STEPHANIE: I like the idea of moving the needle. Even though I tend to not want to stir the pot with any big changes, one thing that I do find myself doing is in a couple of places in the specs, just trying to refactor a bit away from using lets. There were some kind of forward-thinking decisions made before when RSpec was basically going to deprecate using the describe block without prepending it with their module, so just kind of throwing that in there whenever I would touch a spec and asking other people to do the same.

And then, recently, one kind of, like, small syntax thing that I hadn't seen before, and maybe this is just because of the age of the codebases in which I'm working, the argument forwarding syntax in Ruby that has been new, I mean, it's like not totally new anymore [laughs], but throwing that in there a little every now and then to just kind of shift away from this, you know, dated version of the code kind of towards things that other people are seeing and in newer projects.

JOËL: I love harnessing that energy of being new on a project and wanting to make things better. How do you avoid just being, you know, that developer, though, that's new, comes in, and just wants to change everything for the sake of change or for your own personal opinions and just kind of moves things around, stirs the pot, but doesn't really contribute anything net positive to the team? Because I've definitely seen that as well, and that's not a good first contribution or, you know, contribution in general as a newer team member. How do we avoid being that person while still capitalizing on that energy of being someone new and wanting to make a positive impact?

STEPHANIE: Yeah, that's a great point, and I kind of alluded to this earlier when I asked, like, oh, like, even if you don't like an existing syntax or pattern you'll still follow it? And I think liking something a different way is not a good enough reason [chuckles]. But if you are able to have a good reason, like I mentioned with the RSpec prepending, you know, it didn't need to happen now, but if we would hope to upgrade that gem eventually, then yeah, that was a good reason to make that change as opposed to just purely aesthetic [laughs].

JOËL: That's one where there is pretty much a single right answer to. If you plan to keep staying up to date with versions of RSpec, you will eventually need to do all these code changes because, you know, they're deprecating the old way. Getting ahead of that gradually as we touch spec files, there's kind of no downside to it.

STEPHANIE: That's true, though maybe there is a person who exists out there who's like, "I love this old version of RSpec, and I will die on this hill that we have to stay on [laughs] it."

But I also think that I have preferences, but I'm not so attached to them. Ideally, you know, what I would love to receive is just, like, curiosity about like, "Oh, like, why did you make this change?" And just kind of share my reasoning. And sometimes in that process, I realize, you know, I don't have a great reason, and I'll just say, "I don't have a great reason. This is just the way I like it. But if it doesn't work for you, like, tell me, and I'll consider changing it back. [chuckles]"

JOËL: Maybe that's where there's a lot of benefit is the sort of curiosity on the part of the existing team and sort of openness to both learn about existing practices but also share about different practices from the new teammate. And maybe that's you're coming in, and you have a different style where you like to write tests, maybe without using RSpec's let syntax; the team is using it. Maybe you can have a conversation with the team. It's almost certainly not worth it for you to go and rewrite the entire test suite to not use let and be like, "Hey, first PR. I made your test better."

STEPHANIE: Hundreds of files changed, thousands [laughs] of lines of code. I think that's actually a good segue into the question of how can a team support a new hire or a new developer who is still calibrating to a codebase? I think I'm curious about this being different from onboarding because, you know, there are a lot of things that we already kind of expect to give some extra time and leeway for someone who's new coming in. But what might be some ways to support a new developer that are less well known?

JOËL: One that I really like is getting them involved as early as possible in code review because then they get to see the patterns that are coming in, and they can be involved in conversations on those. The first PR you're reviewing, and you see a bunch of tests leaning heavily on let, and maybe you ask a question, "Is this a pattern that we're following in this codebase? Did we have a particular motivation for why we chose this?"

And, you know, and you don't want to do it in a sort of, like, passive-aggressive way because you're trying to push something else. It has to come from a place of genuine curiosity, but you're allowing the new teammate to both see a lot of the existing patterns kind of in very quick succession because you see a pretty good cross-section of those when you review code.

And also, to have conversations about them, to ask anything like, "Oh, that's unusual. I didn't know we were doing that." Or, "Hey, is this a pattern that we're doing kind of just local to this subsystem, or is this something that's happening all the way? Is this a pattern that we're using and liking? Is this a thing that we were doing five years ago that we're phasing out, but there's still a few of them left?" Those are all, I think, great questions to ask when you're getting started.

STEPHANIE: That makes a lot of sense. It's different from saying, "This is how we do things here," and expecting them to adapt or, you know, change to fit into that style or culture, and being open to letting it evolve based on the new team, the new people on the team and what they might be bringing to the table.

I like to ask the question, "What do you need to know?" Or "What do you need to be successful?" as opposed to telling them what I think they need [laughs]. I think that is something that I actually kind of recently, not regret exactly, but I was kind of helping out some folks who were going to be joining the team and just trying to, like, shove all this information down their throats and be like, "Oh, and watch out for these gotchas. And this app uses a lot of callbacks, and they're really complex."

And I think I was maybe coloring their [chuckles] experience a little bit and expecting them to be able to drink from the fire hose, as opposed to trusting that they can see for themselves, you know, like, what is going on, and form opinions about it, and ask questions that will support them in whatever they are looking to do. When we talked earlier about the four different quadrants, like, the kind of information they need to know will differ based off of their task, based off of their experience. So, that's one way that I am thinking about to, like, make space for a new developer to help shape that culture, rather than insisting that things are the way they are.

JOËL: It can be a fine balance where you want to be open to change while also you have to remain kind of ruthlessly pragmatic about the fact that change can be expensive. And so, a lot of changes you need to be justified, and you don't want to just be rewriting your patterns for every new employee or, you know, just to follow the latest trends because we've seen a lot of trends come and go in the Rails ecosystem, and getting on all of them is just not worth our time.

STEPHANIE: And that's the hard truth of there's always trade-offs [laughs] in software development, isn't that right?

JOËL: It sure is. You can't always chase the newest shiny, as fun as that is.

STEPHANIE: On that note, shall we wrap up?

JOËL: Let's wrap up.

STEPHANIE: 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: Byeeeeeeee!!!!!!!


Did you know thoughtbot has a referral program? If you introduce us to someone looking for a design or development partner, we will compensate you if they decide to work with us.

More info on our website at: tbot.io/referral. Or you can email us at referrals@thoughtbot.com with any questions.

Support The Bike Shed