RailsConf 2018 Recap

I went to RailsConf last week and it was an amazing experience. DHH's keynote reminded me why I love Rails. Eileen's keynote made me super pumped for Rails 6. And all the talks were a delightful reminder of why I love programming and why I love the Ruby community. Here are some deeper thoughts and notes, divided up by talk:

Note: I'll post links to talks that I reference as soon as they are up.

DHH's keynote

This talk really hit home for me since I was very recently battling with Play for 4 months. One of the things that I do really love about Rails is that I can focus on solving the problem I actually want to solve, not problems that have been solved before (like... authentication). And, while I do agree that junior developers and people just starting out should not have to know SQL, I do think that knowledge of SQL is still useful if you want to be a good Rails developer. Relying solely on ActiveRecord is a mistake.

Crash Course in RSpec: stubs and doubles and mocks -- oh my!

This workshop managed to be both good and not quite what I wanted. I had hoped by the title that there would be a big emphasis on stubbing, but it was more of a footnote. It was a good crash course though and if you don't have much prior RSpec experience, check out Nicole's tutorial.

Interviewer Skills

Jennifer Tu of Cohere gave an excellent workshop on interviewer skills that I have about 4 pages of notes from that I will try to sum up here. One of the first things she brought up was that a team should have specific goals in mind when interviewing:

  • What values does the team have?
  • What characteristics does the candidate have?
  • What actions does the candidate take in certain situations?
  • What makes someone successful on my team?

For each attribute that the interviewers want the candidate have, they should ask questions that dig into how a candidate behaves. For example, if your team values kind feedback, instead of asking "Do you give kind feedback?" or "Are you nice when responding to pull requests?", ask "Have you ever given feedback to someone whose code was not good? What did you do? Why?". If you value independent learning, ask:

  • How do you learn something new?
  • Do you have an example of a time when you ran into code you didn't understand?
  • Share a time when you had a problem dumped into your lap but you had no idea what to do.

Make sure to wrap a question in context to ensure the candidate fully understands what you are asking. For example, one of the attendees wanted independent thinkers and people who would question decisions. They were currently asking this question:

You get a user request to add a blue button. How do you add a blue button?

However, in the context of an interview, someone who would normally question a decision like that might resonable think that the interviewer just wants to know if they know how to add a button to a page in HTML. What they should ask is:

We get a lot of feature requests and they aren't always valid. What would you do if you got a feature request to add a blue button?

Allow interviewees to show the skill if you can. Theoretical scenarios often just end up only showing red flags. Play acting is the better option. For example, if you want to know if someone gives kind feedback, give them some bad code and have them review it. If you want to know how they handle conflict, play act with the two interviewers coming up with conflicting ideas and ask them how they would resolve it.

It is the job of the interviewer to give the candidate the opportunity to show off. Interrupt (politely) if needed. You will be doing them a favor! Here are some possible polite interruptions:

  • I like where you are going with this but....
  • I'm sorry to interrupt, but I'm really curious about...
  • This is interesting, but I really want to hear more about...

You should also be sure to set an agenda and share rubrics with other interviewers ahead of time.

Pairing: A Guide To Fruitful Collaboration

André Arko gave this talk on the best way to pair and, as someone who has paired incorrectly for a while, it was quite interesting. So the basis of pairing is two devs, one machine.

I can't not use this gif even though pairing is not this.

Anyway! You should be actively collaborating. The best way to think of pairing is to think of it as one little meeting. If done right, it should push you to be a better dev and away from bad habits. Above all, pairing needs trust. If you are condescending, that breaks the trust of your pair and makes you a lousy pair. One good way to pair is to have the driver write a test, codes until the test passes, writes a new test, then switch driver to the other person, who then repeats the process. Never say "let me do this quickly by myself." That is not pairing! Help the driver solve the problem and stay on the same page, so you both understand. There's a lot more to this talk, but I think you should watch it yourself 😃

The Practical Guide to Building An Apprenticeship

Megan Tiu built out the apprenticeship program at CallRail and so we get to learn from her experience! To start an apprenticeship program, you need:

  • plan (what are they going to do?)
  • cash (pay them!)
  • buy-in (convince the boss!)

You can sell it by noting that apprenticeship programs:

  • eliminate onboarding costs (you get to teach a newbie developer your way of doing things)
  • eliminate recruiting costs (why pay a recruiter $10K when you can give it to your apprentice)
  • easier to hire seniors (who love to mentor)

Here's what you want to know about your plan:

  • How long will the program be? (suggestion: 3-4 months)
  • How many apprentices do you want to have? (ensure there are enough seniors to mentor them)
  • What should they know prior to starting? (do you expect them to have a basic working knowledge of Rails?)
  • What should they learn?
  • How will they learn it? (through tickets, a big project, pairing, etc)

For hiring your apprentices, you want an application (basic questions to get to the heart of what they are about), a code challenge, and a final interview. If possible, do end-to-end anonymization until they get to the final interview. You also want to ensure you have a rubric prior to starting this process. After you hire them, try giving lessons on foundation concepts, then give them small changes (bugs/internal code). Then rotate them around to different teams, including customer facing product. And don't forget to set early expectations!

Eileen's keynote

Eileen Uchitelle totally pumped me up. She discussed the various ways she is looking to make Rails more scalable by default. One of the things that really stuck with me was when she mentioned how so many companies are doing these things individually... so why not make them part of the overall framework and share the knowledge!

The Code Free Developer Interview

Can you tell I am into interviewing? This was a talk by Pete Holiday, also from CallRail. Here are the problems with coding during interviews:

  • don't replicate real work
  • disadvantage people without free time (code challenges)
  • live coding is very stressful, even for experienced people
  • difficult to develop and maintain a good code challenge
  • many passive candidates won't do the takehome (I've done this before)

So what's the solution? The primary solution is to just talk to candidates.

  • Ask all the candidates a consistent set of questions
  • Define a rubric ahead of time
  • Write down thoughts right after the interview

That's it! But there's more. Here are three possible techniques for a code-free interview:

1. Dig into their experience. Let them direct you to what they feel is most important. Ask questions like:

  • What was your role in the project?
  • How does the feature work?
  • What's the worst technical debt? Why hasn't the team fixed it? How would you fix it?
  • Has it had any bugs/outages in production? What happened? How did the team fix it?

2. Have them do a code review. If you choose this, make sure you are not using production code (they will have no context), are actively reducing complexity, and include realistic bugs without making it a bug hunt. One good option is to have a completely contrived situation with a simple application and a pull request to that simple app. Another is to fork an open source repository and create a contrived PR. The pull request should include no detail in the commit message, unsquashed commits, non-idiomatic code, overly complex, bad variable names, and actual bugs.

3. Try doing a collaborative system design. For this, you want to hypothetically build a tool, platform, or a project. You don't want any code or pseudocode and you should be working with the candidate. The general idea should be easy to understand and either related to the skills you're hiring for or well known. This can be forever-long, so it needs to be timeboxed. Let the candidate lead and build complexity if it's needed. For example:

Let's say we want to build Facebook. Get rid of the boilerplate (we already have users) and then ask "How do we implement status updates?". Once they get there, we can go deeper and ask about privacy controls, then granular privacy controls, and past that potential performance problems.

I loved this talk because I think code-free developer interviews should be the norm and have also been advocating for it at companies that I have been at.

Plays Well With Others: Improv For Nerds

H. Wade Minter gave this workshop and I don't have any notes on it because it was an improv class. But! One of the big things I took from it was our last activity. To remove bias from ideas, we did the following:

  • each wrote down an idea on how to improve RailsConf for next year
  • exchanged that idea with another person
  • each paired up with someone else, compared ideas, and gave each idea a number of points (total points for the two ideas could not be higher than 7)
  • exchanged ideas with a different person
  • wash, rinse, repeat until we have compared ideas 5 times

At that point, we had seen about 10 different ideas (plus our own) and the best idea could have a total score of 35 with the worst having a score of 0. Our top idea had a score of about 26, with a good number being around 22. We had a couple of bad ideas in the double digits (I'm looking at you, bacon table). This definitely seems like a good practice for any organization with a decent number of people.

And that's it...

I did some more, but I don't have any notes! I also sat and watched my friend Sam Phippen pair for an hour and a half, so if you want to learn yourself some RSpec, watch here!