The Secret To High-Quality Code with Dr. Michaela Greiler and Liran Haimovitch

Liran Haimovitch, CTO of Rookout – an effortless debugging tool, and I talk about how to get to high-quality code.

We also talk about:
  • what are the challenges of moving fast
  • what does productivity mean
  • a lot about code reviews
  • and I also give you a glimpse of the research I’m currently doing.
Picture of Liran Haimovitch
About Liran Haimovitch
Liran Haimovitch, is the CTO of Rookout – an effortless debugging tool.
Book your awesomecodereview.com workshop!


Read the whole episode "The Secret To High-Quality Code with Dr. Michaela Greiler and Liran Haimovitch" (Transcript)

[If you want, you can help make the transcript better, and improve the podcast’s accessibility via Github. I’m happy to lend a hand to help you get started with pull requests, and open source work.]

Michaela: Hello and welcome to the software engineering unlocked podcast. I'm your host Dr. Michaela and today I have a special episode for you. Two weeks ago I talked with Liran haimovitch, the CTO of Rookout - an effortless debugging tool. Our conversation was so much fun and somebody on Twitter actually asked me if I could make it an episode, and I thought, well that's a brilliant idea. So, today I'm sharing my talk with Liran Haimovitch on the challenges and struggles for getting to high-quality software. Enjoy.

Maror:[00:00:00] Um, hi everyone. And welcome to our webinar today on the secret to high quality code. We're really excited to have you all here with us. Uh, so let me introduce you to Dr. Michaela and Liran Haimovitch, the stars of today's webinar. Dr. Michaela has been helping software teams build high quality software in an efficient and effective way for 10 years, and her mission is to lead teams to unlock their full potential through company workshops and team coaching sessions.

Liran is the co-founder and CTO of Rookout, which is a live data collection and debugging platform. He's an advocate of modern software methodologies like Agile, Lean and DevOps, and his passion is to understand how software actually works. So when he's not thinking of code, which is rarely, usually diving, hiking, or writing a new Rookout blog. Um, and so before we get started, I just want to remind you all that we do have time for questions at the end of the webinar; So please don't hesitate to leave questions in the question box and this will be recorded and we will be sending you the recording at the end. So. You're on Michaela, please take it away.

Michaela: [00:01:02] Thank you so much for your really nice and kind introduction. I'm really excited to talk with Liren today about, um, high quality code and get his whole perspective on this topic and pick his brain. So yeah, I'm really thrilled to be here.

Liran:[00:01:19] It's great to be here with you to discuss so many interesting topics.

Michaela:[00:01:24] Yeah, really cool. So in, in the beginning we discussed a little bit, like what should this webinar be about? And we thought like, let's come up with this idea that we are asking each other a little bit questions that, you know, are burning questions for ourselves or that we very often, you know, encounter. And, um, so I want to start with that theme and I want to ask you about the challenge that you see, or the challenges that you see that, uh, engineering teams face nowadays, with really moving fast. Right? So there's like this accelerate the book, for example, there, the DORA metrics, the many other metrics around code velocity. So it's apparently something that we want to do, right? We want to move fast. We want to be productive, but what are the challenges and how can we actually achieve that?

Liran:[00:02:12] So, I can say from my personal experience as well from pretty much everything I've read on the topic, the best way to move faster is to work in smaller units.

You mentioned DORA, the DORA metrics and accelerate, and they're constantly about, you know, roundtrip time for new features and the amount of new features that are being released and how can we build in it? How can we work in smaller unit to work? And the reason for that is because smaller units of work allow for much faster feedback cycles that allow you to learn much more; you get more feedback, you get, you learn every step of the way you learn more often, and you also delivered more value to the end customer on a more frequent basis. And in a way that's actually driving a lot more value. I guess the biggest challenge is actually, how, how do you do that? How can you keep moving ever faster? How can you deliver in smaller units while still keeping delivery efficient? And I found that one of the best way, the best way to start is quite often culture. And we talked a bit about you, eh, doing some recent research for that. So I would love to hear about that. What do you think about how best to build a better culture and how to promote a culture that deliver faster and deliver in smaller units of work?

Michaela:[00:03:34] Um, I'm a big advocate for, for great, uh, culture, right? Who isn't? Somehow everybody wants to work at the company that has a great culture, but unfortunately not everybody is. I'm currently actually doing a research project that's the one that I talked with you about a little bit is on productivity and work culture and the experiences of developers at their companies. And so I'm doing right now, a qualitative study, a grounded theory study where I'm really trying to deeply understand how are people experiencing their workplace and what factors are influencing their satisfaction, their happiness, their productivity, and what enables them to move fast, as you said, to be productive, to be their best selves.

And, you know, there are some factors around obviously release, for example, is one that's also covered by the DORA metric. Um, how has the release experienced? And here I'm not only talking about metrics because, I think on one hand, I'm extremely data-driven; whenever I was working with teams or am working with teams, also at Microsoft, we did a lot of the research that was very data-driven, but it was also, and this was very, very important for me, always qualitative as well. Right? So not only you're looking at the data, which gives you a very include complete picture, but you're understanding, trying to understand the whole experience. And so this research study is really looking at the whole life experience of developers. So on one hand we have like metrics like, um, you know, release cadence or from time to commit or from time to merge and so on. So what's very quantifiable, but on the other hand, you have um, the impressions and the perceptions of people around that. Right? So are they feeling better with it or, um, are they feeling worse. And the same is true for code reviews and so. A lot of the things can relate back to culture and culture is somehow enabler here, right?

So we have like this practices around those areas, let's say release. These feedback loops that we have, release is actually a feedback loop. Code reviews are a feedback loop, right? Talking with product management is a feedback loop. How seamless, how smooth can we make them? And culture is really an enabler for that.

Why is it an enabler? Because if I'm allowed to say, if something is wrong or if I'm allowed to experiment, even experiment with some failures, right? Like I try something out, I try to work different with product management. It doesn't work out, what happens, right? What are the consequences of that? And that's why culture is so important because if people feel that they can experiment, if they feel that they can also express their opinion on it, they will drive more improvement. Right. Um, the research is really a lot about improvement; how much improvement can people um, drive and they normally know what's good.

Right. They know what's going bad. Um, or what's good. And so it's really about enabling them to act upon that. And that has a lot to do in here. The funny thing here is that metrics are really important to want on one hand, to enable people that we see and that we make it visible that there are problems. But on the other hand, metrics often also hold people back. Because if I'm, if I'm measured. By one metric. Right? Um, it means that if I'm trying something else, something new, that's not covered by this metric. It very well, it could very well be that I'm actually slowing down or I'm, The metrics outcomes going down while I'm trying something out. Right. And so the question is really about culture here again, how are people handling that? Right? Do I always have to perform to my OKR or KPIs or what not, right. The metrics and the goals that set around, or am I actually allowed to experiment here with things that might slow us down for a short time when I'm doing the improvements, because improvements are really hard to do without short term slowing down, right? Technical debt. How are you going to work on technical debt and still keep the features going? Right? Yeah. So this is what I am seeing here.

Liran:[00:07:41] That actually brings to mind the analogy from LEAN production, where you stop the line. When you see something is wrong and you say you have the give a individual engineers or individual employees, the permission to call it; to stop the line and spend the time and efforts to improve things, even at the cost of lost productivity in the short term, because it allows for continuous improvement.

Michaela:[00:08:09] Yeah, exactly. It's a really good analogy. Yeah.

Liran:[00:08:13] I think it's so important to create. Um, I remember I talked a lot about feedback, but you're right. It's critical that it's not just enough to have feedback, but it's super critical that the feedback experience is going to be positive. Even if this feedback is negative, it's important for people to be able to experience. Getting feedback -something positive, and in a way that if they're changing something or developing something, or if it's a bad product idea that the negative feedback should be, you know, about the, the feature about the task that this feature was bad, but the person who came up with it wasn't bad and they didn't necessarily make a bad choice by, uh, you know, going after this feature, and people should be glad about getting those so-called negative feedbacks and not attribute them personally. And that's super important to the culture, to the experience, kind of, how do you go about creating that? How do you go about building that environment where you get continuous feedback and experience is good.

Michaela:[00:09:23] Yeah. I mean, I'm thinking a lot about culture nowadays and, you know, the common sense is always all countries so hard, right? It's so hard to change. And if I'm in a, in a bad place, you know, it's a bad place. Um, and I think on one hand, that's probably true, but I don't hand now that I'm confronted so a lot with that, and I'm really working a lot with organizations and that they in displaced, I'm thinking about the small things that you can do, and culture really begins now really coming back to something very concrete. Um, I'm all about code review. So culture begins already in code reviews and for example, code review feedback.

And to, in my workshop, what I do, I work with people on how to give respectful feedback. And very often everybody thinks like, Oh, but I'm doing this right? Like we are not fighting in the code reviews or whatnot, or, you know, or it's only instances of that where we are mean, but it's more, it's about the collective awareness of not only do I fight with somebody or, you know, is it an unrespectful, but really is my mind about value? The value that I can provide to others. Is my mind about how can I actually, you know, improve the experience of my peers here. And I think this is something that's often not done, and this is something very small where a team can really start actively being more aware of that, more deliberate, more conscious about this. Um, and it starts already by understanding code is really, really hard and everybody has, um, everybody has a time pressure and you know, wants to deliver the features. And a lot of engineers say, well, you know, code review is good, but, uh, I actually have to deliver feature. So what's about the time that I have to spend on the code reviews somehow it's missing from my feature work and so on.

And so having really empathy around that and the experience of myself, but also off my, off my team, that's already creating culture, um, and being extremely, um, aware that feedback, even if it comes from a good colleague that you think like, we are all, you know, good friends and we are really on the same page that we still really take and this is now again, you know, slowing down, right. It's slowing down to make sure that I'm phrasing this feedback in a very respectful way, because we know that feedback can sting. Right. Um, and it can be misinterpreted. A lot of the feedback comes through a tool, which means it's an automated tool. I'm not directly talking to a person so sometimes I forget it and we are in this automated way of um, you know, looking through the algorithms, finding, you know, let's say edge cases or whatnot, finding problems. And so if you are in a very technical state of mind, and then we are hammering in our feedback and say, Oh, Variable name is wrong or, you know, or should be different. And then going, you know, taking one step back and thinking, is that creating a good culture? Or can I take this, you know, two more minutes and say, Oh, you know, what about, um, renaming this and even giving an explanation, you know, really expressing your, your mind. I think. Driving cultural change is definitely hard and, um, comes often also from the top, but there is a lot that teams can do for themselves. And even engineers themselves can ask themselves every day, like, what did I do positive today? Like I'm not only going somewhere in and they're expecting that culture would be great, but am I actually contributing to making a good culture here? Hmm.

Liran:[00:12:49] I think it's so much more critical today as we are working remotely, because as you mentioned, we're often in that technical state of mind, whether it's on GitHub doing code review or on Slack or wherever, but behind what's actually happening is we're communicating with other human beings. We're not just no analyzing code and the testing stuff we're communicating to other human beings. And as we were, many of us are working remotely quite often, or most of the time. We can often forget there is a person on the other, on the other side of it. And sometimes we kind of forget to act with empathy, with compassion, and while we may be factually, correct we're not creating a good experience for the person on the other side of that communication.

Michaela:[00:13:35] Yeah. And actually about the factually correct thing, I have learned over my, you know, my time in the industry that the seldom leader case, seldom leader are very, very active, you know? And I think sometimes you forget about that and, this is two perspectives. Cultural views are a place where two perspectives really are, you know, they are, they are the benefit of it. And there are also the problem, right? That you are constantly having somebody that looks at the same thing and say, well, but I'm seeing something else, right? Like I'm seeing technical debt and you say, but it's fine. You know?

Liran:[00:14:15] The thing is engineers so often feel like they're factually correct. Even whether it's or not it's the case. I just, and quite often we're not actually factually correct. There is some degree of, eh, you know, afraid of common sense and various options, and there is not one single truth out there, but even if there is even if you there is, and you're convinced that there is in your, on that single spruce, it's still communication. And you must not forget that there are other things beyond fact, too, and more often than not, there are actually no facts and it's just your opinion, which might be very good and professional, but it's just an opinion.

Michaela:[00:14:59] Yeah, very true. And a lot of the time it's about strategies. And about the unknowns that you know, that the unknowns unknowns that we actually make guesses and decision and B we don't know if they are the best ones. And we even in hindsight, we cannot decide like if we could have, you know, if you would have done it differently, would we have a better outcome? We don't know. Right. So, yeah, it's, it's, it's really dealing with that and, and embracing that and maybe reminding maybe something that we can do also, always over time to build this culture is reminding us of that. Right. It's a little bit like we have to remind ourselves of the central things. We have to remind also ourselves that we are dealing with so many unknowns and that on one hand, you know, we are, I think at, at one point we have to go and say, Well, we don't know better. And maybe some people disagree here, but now is the time that we are, you know, buying in and going this way together. And I think this is also important for engineering teams, right? So in one of my code review workshops last week, for example, I give them a code base and it was, um, it seeded with errors, right? So it has issues and they asked the team to, to find those issues. And there are, you know, they are. Issues about readability, maintainability of the code, but there are also security issues. And so then we had a discussion about, you know, um, so there are a lot of issues. So how are you going to communicate to the person that wrote that code about those issues? Are you going to tell them all of them at once? You know, do we make like a plan around what should be, um, worked on first? And there was, for example, this discussion then between two senior engineers and they were saying, well, once by saying. Yeah. Um, so everybody agreed that, you know, sending them 300, 300 problems at one point is not the right thing to work with this junior. So, um, they were thinking, well, let's do it in, in phases, but it didn't, they couldn't agree. Like if security issues are more important than readability issues or not only readability, but making the code work. Right. So there was this discussion that, well, this is early stage, so it's, it's probably a prototype, so we should have to, you know, show. So let's do it. Make it correct first that it works and then work on the security issues that they had, like they had, there were injection box and cross site scripting products and so on. Right. But in the end, you know, like the whole team was discussing it. They couldn't really find a way forward. Right. There was one side that was very convinced that, well, these are really critical security bugs and they were really critical. And the other was like, well, but it's, you know, we use it internally right now. It's a prototype. So let's make the functionality work first. And so what, there was a back and forth, and I think this was a really nice example of, I couldn't tell, like, I couldn't say like they wanted me to be now the referee and say, Oh, you win. Right? Like the security team, Vince, we first do a security or we first do inability, but there is no right or wrong answer. It's just the strategy that you're going to do. And probably that you're not doing, you know, again, not sending all the issues at one is a good one. Um, but then in the end, it doesn't matter if you do one or the other, as long as the security bugs are not coming out right. In, in production. Um, yeah. And, and I think here it's really important to step back from this discussion at one point and say it's actually a nonsense discussion. Let's, you know, flip a coin and do one or the other. Um, yeah, this is what I think about this. So a lot of the things is really. It really depends. And then we have to make a decision and if we made the decision, this is the important thing. And then everybody has to buy in and not like, keep this resentment and say, Oh, the security or pre approach first. Right. And I think it's stupid. And so that's why I'm blocking here, which is culture. I think. Yeah.

Liran:[00:18:49] So actually it's not, it's interesting that you mentioned that because it's such a big topic. And I mean, so much effort goes into code reviews and often becoming the bottleneck, both for whoever has to do the code review. And, you know, spend the time and walk and provided feedback in both wherever need, wants to get, just to get this code out there. And they're just trying to, and, you know, they've just finished developing the feature. They just want to check off the it's been called of you and send it out there. So kind of what strategies should company follow to speed up their code reviews?

Michaela:[00:19:27] So, um, I totally agree that code reviews can become a bottleneck and they're coming with a lot of pain points, but I think especially this. This mindset of, you know, cultivate is just another hurdle. Um, that's something that people have to work on. Right. So, um, we really have to understand and carve out also, what's the benefit of the code view? Why do we do it even here? Right. And if the feeling of the engineering as well, I just wants, uh, I want to look good to me and that's it. Um, then obviously it's a delay and it's a bottleneck and you know, the value. Probably isn't that high because even if the person gets good feedback, you know, if the person that receives it, doesn't actually want it, you know, what's the value of that. So I think that a lot of those is really for an organization and for a team to think about what do we want to get out of contribution? And there is a lot of imperative studies also that really show that the benefits like. Um, improved code base, readability, maintainability, um, less, you know, less issues, less bugs, defects in posts and pre-releases, um, all of those are happening. Culture-based, there's a lot of mentoring and learning happening. There is advantage knowledge sharing, but it's only if I'm open to it. And if I'm very clear about what I want to get out of here, because if I want. Let's say if I want to find it bug, it would be the best to ask a person that's familiar with the code to be on this code review and not, you know, a junior, but if I want to have this learning expect more in, in, you know, in the center, then obviously I ask somebody that maybe hasn't seen that code part before, so that they get familiar, that I have knowledge dissemination that I have more people that are familiar about this code base. And I think most organizations. They're not real bear. They, they hopped on this code review bag and because the hopped on, you know, to pull request, model to development and pull requests and code reviews are not the same. And so suddenly they wrote a pull request and they felt like, Oh, before I pull it in and look at the code and because I'm looking at the code, it's already a code review. And so now I'm doing code reviews and I want all of these benefits without actually investing I'm investing. And then here it comes back to this slow down. Right. So I have to probably slow down first. Really find out with my team, what is it that we want to get out of code reviews? How are we structuring our processes, our practices, and this has to do a lot, right? Like, depending on the risk profile of this code review, who should be on the code review would ask for feedback, how long should it take them? What issues are they looking for? All of that can actually be designed and very deliberately made. And then you're getting really a lot of benefits out. But if you're not doing that, yeah. Then you're in this state where. He just wanted, it looks good to me. Right? The other person knows you just want that, but still feel a little bit pressure, um, that they have to look at it because if it goes in, they're also responsible. And so there's this delay, um, and you don't want to spend time for it, but you have to, you know, and then you having an, I actually have a, if you look on my website, there's the code through your quadrant. Um, and this means like it's, it's, you have to access and it's the speed of the process and the value. And this means that you often have them. Organizations that are slowing speed. And low in value. Right? So they are low in, in speed. They're very slow. They're bottlenecks and they don't get value out or that they're fight fast because they're just giving out. Looks good to me, but they're not getting value out of this. Right. But even if you're waiting for look good to me, Like say half an hour or an hour or four hours, it's still slowing down your process. And the question is, was it worth it right? If people are not really taking the time to review. So in the end for me, it was probably a very long answer to your question, but it really comes down to what, why do you do code reviews? Right. And do you have to have an answer for that? And probably depending on the code change, the risk profile of the code change and the code change, you will have different answers to that, or this code review. I want that, you know, my junior engineer knows how that works. And so I'm sending it over or this, this code changes about how we are doing the checkout. So I definitely want, you know, two more eyes, um, to make sure that there are no, no defects going out. Right?

Liran:[00:23:47] Yeah. I mean, that sounds so complex. Can't we just automate this and install some tool and get it over with?

Michaela:[00:23:55] I definitely parts of it. And I think that a lot of people are, are, are doing stuff that tools should do for them. Right? So they, they, they are mocking on, you know, style issues they are talking about, you know, some, some things that actually study analysis tools could find. Um, or automated code review tools, whatever you want to call them. Uh, in the end it's, it's, linters that checkers steady analysis tools. Right. And they are actually much better than, than people. To find certain, certain errors and certain problems with your code, they can actually, you know, they can walk through your code and really find out, you know, if they're, if, if some code paths are not called and tell you, Oh, this is actually not going to call it. Or, you know, really also bugs study analysis bugs, but they are limited. So it's, it's not something, you know, they are not, you cannot comply, uh, replace the, the manual review. But you can replace a lot of that, you know, nitpicking, which is very unproductive in code reviews. Um, it doesn't matter. Like why would you have an engineer spend time on finding certain types of errors? If a tool could do it automatically? I'm I'm all for automation. I think it's so important to automate whatever you can automate here. Yeah. Are you using, are you using some automated tools in your pipeline?

Liran:[00:25:15] So yeah, we actually adopted a GitHub advanced security. A few months ago at Rookout. And it was actually a pretty good tool for us. It allowed us to gain some insights, actually both brought us a lot of insights into some of the other code that broke out and kind of knowing where we might've pitfalls, but it also managed ha ha is helping us moving forward, knowing that it could work coder pushing through discussed meeting there. You know style checks and best practices, especially when it goes to more junior engineers or engineers or working in environments that are not a strength. Let's say, I know most of our full stack engineers spend most of their days between a, you know, react and node JS, but occasionally they dive into Golang. And then all of a sudden, they're not as fluent in, you know, what can go wrong and how should the code views. And some of those arrows can easily be caught by those automated static analysis tools. Also, it's a very useful tool personally, we've, we've recently developed support for Ruby and surely within that skeleton of project, we started with Rubocop, which is a very, very strict, eh, who ruby linter. And that's actually provided us with a lot of insight and kind of kept us very honest as we were developing the code, keeping functions, very short, creating a very orderly and well structured code. And that's kind of something that it's always a dilemma for me when starting a new project. Do you go ahead and spend a lot of time building the skeleton, building the CICD building, building linting. At the beginning of the, you wait for it later on, because you know that later on, doesn't always get by. And if you're adding a linter to an existing project, and then all of a sudden you're getting, you know, dozens of errors, then you might not be, get going around to fix them because it's too much work and that's always kind of a dilemma. But for that project, it was a very good experience for us in developing high-quality code.

Michaela:[00:27:25] Yeah, that's really nice. Yeah. There's also my experience. Like if you, if you add that to an already existing, quite substantial code base, right. It's just out of hands, right. You will have like all these red flags, orange, whatever, you know, depending on the tool that you have, like different severities of issues. And I always feel really bad because. I know that I'm not going to be able, like to go back and, you know, redo the past. Um, you can do it slowly by slowly, right? Like I'll voice called removal or at five by refactoring where you say, well, if I'm touching this code, I make it nice again or make it better and you can do it ongoing. Um, but yeah, I also feel like for, for existing code, it somehow has this, you failed here. Um, um, Psychological, uh, you know, by byproduct, but you'd be like, Oh, now I'm seeing what's all messy and you cannot really do it. Um, yeah. But yeah, it's good that it worked out. So apparently you could, could you remove all of the issues? Could you work through all of them?

Liran:[00:28:26] So we got through, and I think 95% of them ere, there were some, a few areas where we decided that. That code is not going to be the nicest code in the project. And that's okay. This code is mostly, you know, four, it wasn't was low maintenance, low complexity, just a lot of, you know, lung functions doing boring stuff. And we said, that's something we can live with without spending too much engineering efforts, kind of fixing it up and making it look the best.

Michaela:[00:28:59] Yeah. So in the interview and the research interviews that I'm doing right now, we talk a lot about technical debt as well, and how people deal with technical debt. And I'm asking different organizations, different teams, their strategies for technical debt. What are your strategies? We have like some, you know, some amount per sprint that you can use on that, or how do you, how do you even. Um, assess the value of working on this, uh, you know, technical, then you were talking like, Oh, we already okay with this part of the code base, but how do you assess that? And on a more strategical, systematic level, right?

Liran:[00:29:34] Yeah. So I guess that's a two part question. And on the one part, we do have a strategy and I can talk about it a bit, but I think it goes beyond that. I found that for you, you mentioned actually early on that nothing in very literal in tech is factual and most of it is opinions. And I think that's doubly true for a tech debt. And quite often one engineer joins a project and they decide that what's happening. Much of, many of the decisions have been taken before they joined our tech debt there have been wrong. And I would wager basic statement that it's. Probably the other way around. I mean, if the project is live, if it's generating value, if that piece of code was walking from when it was written up until now, then chances are the decision to, to do it. That way was actually correct. Or at least descent. And engineers often jump to say to, you know, define tech that because something is not in the latest design pattern or something is using an older technology or paradigm, or maybe simply because they don't understand something. So often the first thing you have to do when you think of tech debt is actually understand what's going on and truly think for yourself. I truly think about it. Is this truly affected? Oh, is this something you lack an understanding? And actually that's something we were seeing within Rueckert and with our customers that shook out is quite often used for once you have a better understanding of the code, because you can see how it's working and you can see inside of it, then you quite often realize that's not actually tech debt that's the, I just didn't understand how it was working. And once you get gain a better picture of how is it working, why is it working that way? And none of the sudden then it makes perfect sense. But obviously sometimes there is real product that there is real tech debt they, for the most part, we kind of manage tech debt on a, you know, quarterly on a quarterly roadmap. We have a very. Agile flexible quarterly roadmap while we manage our roadmap, eh, usual, all the rollout. And you also kind of add, you know, a handful of tasks for each team and full of mid-level mid large tasks for each team where they can, whether they should strive for a tech debt. And obviously, you know,tech debt usually comes last in priorities priorities. So it doesn't always get executed. A lot depends on the roadmap progress in general and especially on a. Eh, eh, new tasks that get pushed in from the sales team as well, working with customers. And there are always new requirements for improving performance, for a meeting new criteria for giving the best experience with possibly can for our customers. And those often override some of other stuff we have on the roadmap, but we do try to get at least some of the tech debt cleared every quarter, just to get a few low hanging fruits with high impact. Stuff that's been bothering us, that's bothering the team. And also we find that having those, you know, tasks in the queue engineers kind of find time in way to get to them, to get it out of the way.

Michaela:[00:32:56] Yeah. So what, what reminds me and what I wanted to ask you in that context is that the original or one of the very early on definitions of tech debt was code that didn't have tests right from my Confederacy would say, well, it's tech debt if you don't have tests, because then you really have a hard time refactoring and often, you know, There's also this new, I was actually, I did a podcast with him, uh, recently on, on, on my podcast and we were talking about it and then he, and he also sat like tech tech debt is the code that has been outlived, but a person that. Wrote it right. And that in our days in our, in our, um, very fast pace or, you know, um, tech industry where people will stay two years, maybe at the company, they write code and they actually never really see it in the maintainance pace. Right. So do you see it when they're writing it? Maybe when you're releasing it. And so a lot of the, you know, like a lot of the, the. Code becomes tech debt, because the knowledge is actually gone from the organization that, you know, wrote that that code or, you know, can maintainer or understand it. What's your perspective on that?

Liran:[00:34:07] So there's actually at Rookout we've kind of we've wrote and talked a lot about understandability. It's exactly what you mentioned. It's about knowledge it's about if you're able to understand the software, the code fairly well, then. You're the new C you can do a lot. I mean, you can get stuff done. I mean, I think the most obvious example of that is, you know, those simple exercises you get on introduction to computer sciences readFile from disk, sort an array, eh, those kinds of stuff. And you know, those exercises you can usually do right now as a senior engineer in 10 minutes, 20 minutes. And you're done, but if you were to get the same task within the context of a very large system, especially one you're not intimately familiar with, then all of the sudden the same tasks can take you weeks. And then you're going to start complaining about tech debt and lack of knowledge and documentation. And if at the same time, or to give that same task to the two, a person who was one of the founding team of that system, they're still going to get it done in, I know maybe not 10 minutes, but 30 minutes. And preserving knowledge is super critical. And at the same time, we need better tooling. We need better tooling that would allow us to work with systems that are complex, that we're not intimately familiar with. Obviously testing is there, as you mentioned, testing is a form of tech that because testing is a Godwin. That allows it to operate in a, in an area where you're not familiar with. It allows you to easily debug the code that I, you to see it in action. Even though it's a developing environment, you can see the code running, you could see it in action. You can make changes in a control mirror. We know what you're changing, you know what you're going to impact, but that's at the same time, I think tests can be incredibly expensive. Even more so if you're not already familiar with the code, so it's kind of, you know, a conundrum you're saying the code is not very good because it's effective because it doesn't have any tests and it doesn't understand it, but then it's going to be very hard for you to add tests to the code. You're not honest. You don't understand. I think observability tools, by the way, can provide you with some insights into how the code is walking. Right. Eh, but at the end of the day, nothing beats just debugging the code, stepping through it, see what, see the actual types and values of variables, seeing the inputs and outputs of the system and seeing it in action. That's the best way to understand the code.

Michaela:[00:36:45] Yeah. I think the too, when I was at the university of Victoria in Canada, um, I was doing a research sabbatical there. And they, they developed a tool. I think at that point was called driver. You will not find it because it was a research tool, right. Not really popular and bad. The tool itself was really cool because it helped you understand cold. Um, my, my research area was called code comprehension. And so really helping teams and engineers understand code. And so this tool was made in a way that you could Deepak and it showed, you know, the traces and the values as you just described. Right. But this is like, 15 years back pretty long time. So it was very novel at that point. Right. And so this was really used to understand coach. I think debugging is definitely one of the ways how we understand code, right. That we really go through it and try to understand what's going on a really interesting resource maybe in that, um, in that regard is also a book that's coming out from a friend of mine, Felienne Hermans it's called "The program has brain" and it talks a lot about cognitive load and code reading. Um, there's actually a workshop that I'm going to attend today about code reading, um, from her. And, um, yeah, and I think this is really, this is really interesting because it again goes into these different versions of cognitive load and also confusion that you have with code and confusion can come, come from different sources. One is lack. Of your own knowledge, right? So being a junior or, you know, being a senior engineer, you have a different knowledge base. So you can actually go back to your longterm memory and quickly access how to load the file, or you know, how to save a file, how to close a file and so on. And then as a junior, you have to really think actively think about this with means that you're. Your, your processing power more or less, right? It's reduced because you have activity after actively think about this. And we can think around four to seven things. So if you're already thinking about those things, you know, there are only two more things that you can add. And as a senior example, you, you. You have that in your long-term memory. So you have seven things that you can think about. And the interesting aspect here is also the same with what you said about the code base. If I'm familiar with the code base, I can load parts of that from my long-term memory. And I don't have to use my short term memory. I don't have to use the processor for that. Right. Um, and so there's definitely, there's exactly what you're seeing here. Um, maybe something else that I want to add here is you said knowledge dissemination. And code reviews are really good for that. Right? So that's why I'm saying the organization has to understand the benefits as a whole, right? And suddenly if you understand that, well, if I can actually have, have a, a larger part of my team, be more familiar with a larger part of the code base, that's actually extremely valuable and it will. You know, it will speed up your development process quite a bit. And there are also studies on that around code reviews where we really see that, um, it teams that have code reviews in place. They already have a Vitor understanding of the code base than teams that don't have. Right. They're only known only know what they're working on. Um, and so why it's slowing you down to do the reviews. It really speeds you up. Once you have to work in this pace, right. Or in this place off the copies, or if somebody leaves, you have other engineers that are also familiar with. With that. And so I think there are other benefits that are really, really, um, really important here. Yeah.

Liran:[00:40:19] Yeah. I Michaela, I'm hearing here speak, it's obvious. You're an advocate of code reviews and you're passionate about it and you're making great, great arguments about why it's so important and how the part, the value fit. But don't people ever come to you and say, I don't know, it's slowing me down. It's making stuff complex. I mean, I don't want to do pull request. I don't want to do code reviews. I just want to skip the whole things and kind of what do you send them?

Michaela:[00:40:50] So, honestly, I don't, I don't have a lot of people that have this complete mindset. I have a lot of people that would say I really would like to skip code reviews because I don't have time to do them because my reward system around my recognition and what I'm expected to do is something completely else. And then I have to look at code reviews and, and this is not part of it. Right. So I recall one person, I was just talking with them. Like we could go around that. Right. It was part of the research again, and they were talking about it. How, how is it? It's really difficult. They actually love code reviews and they learn quite a lot and they would have much more, um, much more benefit and would feel better about them. If this would be actively part of their job description and their expectations. But it's in very many, many organization. It's. It's window dressing. It's like, yeah, we want you to do code reviews and it's really mandatory and they have to do them. But on the other hand, there is no time to actually do them. Right. And I think that's, that's what I see very often. I definitely see people that haven't had good experience with code reviews that don't maybe see the benefits out of that. But I also have, on the other hand, I think this is why I'm such a strong advocate for that. I have people, really, a lot of people that have seen the benefits and that have done code reviews in the right way. And, and, you know, with good processes around and with a good culture around that, that they say I would never, ever work anywhere else without code reviews, because it's a, it's a mentoring tool. It's a learning tool. I'm learning so much more. I'm so connected to my team. Right. And not working in a silo anymore, but this needs a certain time of code reviews. You cannot like work on a feature for a month and then throw over like thousands of lines of code or whatever volunteers look at it and give me feedback, right? Like this is not gonna work, right. This is, uh, this is definitely a frustrating experience for everybody. And in this case, I say, get rid of it. You're not getting anything out of it other than frustration. Um, but also be honest to yourself that you're actually not really doing code reviews, right? You're throwing pieces of un understandable codes to somebody else that can spend maybe half an hour, an hour to look through thousand lines of code. What are they going to say to you? Nothing. Right. And so maybe it's really to be about, be honest and say, if I want that, I need to slow down, understand how to design the process. Um, maybe even get help for doing that. Right. And then, and then really do it right. And have so many people that really love code reviews and so many teams that are striving through that. Um, and yeah, so definitely if you know, if they don't bring any value, then it's really, I think it's very often the process that's just completely screwed up and the culture around it.

Liran:[00:43:39] Yeah. And do you find that companies struggled to understand which pull request was deserved called the code reviews versus which don't, what policies did they have in place to know to solve them out? Sometimes I know sometimes just adding a log line and then you need to go through the same code review process, or at least by definition, it's the same workflow as if you're adding a big feature. So kind of how the companies go around managing the different kind of poll requests.

Michaela:[00:44:08] So I think this is really a part of, I cannot generalize, um, because for some organization it's definitely valuable to go. Through a pull request or a code review for every line of code that they're doing, even if it's a log line. Right. Um, but then this is a certain type of company and they have certain goals around it and it's beneficial. I definitely see also, um, you know, organizations that have some code review guidelines in place and it says we have to look at every line and it's a log line. It makes no sense here. Um, very often here, people haven't thought about again, you know, what are our goals with code reviews? And if you think about the goals and it's a logline in, you know, a web website that I can update within minutes because I have a fast pipeline, why would I go through a code review here? Right? Why would I slow that down? What's the benefit here? Um, so I think that organizations that are more vague about their contribution to practices and process, and really take the time they understand that. Um, and it has to do with risk profiles. Can engineers, do they have like guidelines to work around? Have we thought about this as an engineering team? What are our values? Um, when I was working with Microsoft, we had like, there were, there's not one code review. Policy, right. It really depends. Office has a different policy than windows. And then even in, in office, you have like different teams that have different policies and so on. And so the teams really thought about, um, some teams would say, well, for us, every line is reviewed. And then other teams would say, uh, well, vs keeping, for example, refactorings if you do a refactoring and you can show it, it's, uh, uh, And refactoring that has no side effects, then you can just put it in or some teams would do the review after they've pushed it, for example. Right. So after, after committed, after pushing and after merging, they're doing the so. The, the policies really differ. And I'm not saying that, you know, even if the differ for some teams, they were really good for some teams do or not. Um, it really depends how, how honest and how in there and reflected people were around their code reviews. Um, but you can definitely be design and, you know, even have automatic things that help you to decide whether or not something should have a code review. Right. You could buy somebody if you think about conventional commits. Where you have certain aspects in the commit message even right? Even those systems are in place. If you, that you could do here, where it could incorporate some of the risk of something, or you have against that again, as it tools around somehow assess the risk. And that helps you to decide whether or not you need a code review or in what depth you need a code review here, how many people should be under code review. Right. So, so many questions, uh, yeah. Yeah. Touch on what you meant.

Liran:[00:47:07] Uh, that's. Exactly. Yeah. That's that's perfect. I think we're almost running out of time here. So maybe Mo join us, throwing a few questions from the audience?

Maror:[00:47:17] Okay. So actually, a few questions did come up. If you guys are ready for it, um, Michaela, we'll start with you. Can peer programming, replace code reviews?

Michaela:[00:47:26] Okay. Um, pair programming. So I, yeah, this is a, this is a very often, uh, asked question and my answer is no, it's very similar to, you know, can automated contributes, replaced code reviews. I think they are very complimentary here again. So if you have pair programming, um, You'd probably have different code review practices again. Right. So very often we talk about code reviews and then code reviews are that thing that everybody does the same, which is completely not true, could be, it can be so many things, right. If I'm looking over the shoulder with somebody and looking at the code at the same time, it's an over shoulder over the shoulder code review. And so pair programming could actually be one kind of code review, but then you have to ask your, you know, for yourself or your organization again, Do we need more? Do we need like some gatekeeping around that so that we have another person do we need in fairness right around that? If I have two people that are pairing very often, then you have like this knowledge silo, again, that those two people know about the code, but maybe I want other people in that, so we'll add them. So, um, code review can be, uh, a complimentary strategy to pairing, but I definitely say it should look different, right? For team, the task pairing code review should look different than for a team that does know pairing. Yeah. Okay.

Maror:[00:48:50] Very cool. Um, Liran, I think this one's for you, what's the relevance of code reviews for compliance.

Liran:[00:48:58] So I think we found that there are a few, few key elements in that I think compliance kind of often requires that, eh, some peer reviews, every change, and I think it goes back to what Michaela said about the purpose of code reviews. And compliance for the most part would be focusing on first and foremost general security review, but even more. So it's an often a question of trust and governance that you essentially know what code is going into the system in a way. I think it's very different from most of America is been talking about today, about, you know, in-depth review, understanding the code and eh, me ensuring that you have all the right pieces in place. It's more about cursory examination that you make sure that you're not, you're not changing anything. You shouldn't be changing that the person is making that commit within the assigned task is working on and within the assigned scope, if there are any changes to security, sensitive area that you go through additional scrutiny. But if it doesn't eh, you know, It's more about ensuring that whoever is made the task, did what he was supposed to do rather than the quality of the work he did. So that's a very different thing. And it's very important again, to kind of. Define the purpose of the code review. Is it just about understanding the scope of the task and the scope of the change, or is it about deeply evaluating it? Giving feedback, mentoring, sharing, knowledge and so on and so forth.

Maror:[00:50:43] Okay. Um, and Michaela, if people wanted to learn more about code reviews, where, where would they be able to go to do that?

Michaela:[00:50:50] Okay. Yeah. Um, obviously I can see my website, right? I'm writing quite a bit about code reviews, which would be (Awesome Code Reviews)[awesome.Codereviews.com]. Or you can also go through my link. That's my name, michaelagreiler dot com, which is a little bit more difficult for me. We can put it somewhere, but I'm Awesome. Code reviews, dot com should have them work as well. And, um, yeah, I also have like a GitHub. A project that's about code reviews, um, where I'm listing a lot of different resources that I find on the web. So it's not only from me, but also what I started recently doing is best practices from different organizations. So there are articles where you see like how, um, You know, for example, the Google does code reviews, or how is, you know, VMware doing code reviews and other, um, resources that I found really valuable as well. I also have like code review checklist there on my github profile. Um, so it it's, uh, the github thing. And then my, my handle is M G R E I L E R. And so, yeah, there, you can find also quite some stuff, um, that, um, that comes from everywhere that I found this valuable.

Maror:[00:52:03] That is a wealth of information that everyone should definitely take advantage of. Um, and I will make sure to send out your Twitter handle for them too, so they get it. Um, and on the topic of learning more in Liran, where can you learn more about Rookout?

Liran:[00:52:19] So you can learn more about Rookout first and foremost, that's rookout.com, which is our awesome website. We've just launched a new website. And so feel free to check it out. Also, you can reach out to me on Twitter at Liran underscore last, and I'll be happy to chat with you and share more about what we're doing.

Maror:[00:52:39] Amazing. Okay. So then we have one last question here, um, and it looks Michaela like it's for you. The question is, do we need additional manual reviews or testing if we have a study analysis tools or is that enough?

Michaela:[00:52:52] Okay. Um, I think I touched it a little bit on that. So I definitely think it's complimentary again. Right. So if you have, like, I definitely recommend to have studying analysis, test tools, have static analysis tools, security tools, because they are much more systematic and they they're defining more issues. They are less error prone. You're not overlooking something, right. Especially for things that are systematic. As I said, Um, for example, security testing tools are really good or, you know, security analysis tools are really good for injection box, um, where, you know, people would have a hard time and it's just unproductive for them to look at that. Um, you know, in, in the, in terms of what a tool could do here, but then for example, broken off, um, authentication or just the flow of things that is really beyond the scope of tools right now. Right. So if you're, for example, sending out. Let's say that you're somebody is requesting a password reset, right? So the whole, uh, workflows through that can be very, very broken and there are no tools that, right. An alpha example can check for that. So that definitely has to be done manually by, by person and very similar in the code review sense. Right. So, um, there are really good static analysis tools, but there's always things that just the tool cannot do for you. So they are complimentary, I would say. Okay,

Maror:[00:54:18] thanks. So that's all we have time for today, unfortunately. Um, but hopefully we can also down again cause it's been great. Um, so thank you everyone for joining us, we will be sending a follow-up email with the recording and Michaela and Liran's contact information for whoever wants to get in touch with them. And thank you Michaela. And thank you again.

Michaela:[00:54:38] Yeah. Thank you so much. It was really fun.

Liran:[00:54:41] Thank you. Thank you.

Copyright 2022 Doctor McKayla