Tuesday, January 19, 2010

Craftsman Swap – Day 1 at Relevance

My first day at Relevance actually began Sunday night when Chad Humpries and Kris Singleton were kind enough to entertain me for the evening. I had dinner with Chad at Piedmont, an excellent local Italian place. The food was great and the conversation interesting. Chad talked to me about his inspiration for Micronaut, which is now Rspec2, and went over some of its features. He’s on the hook for giving me a 10 minute overview of the framework and some of its innovative features. Pretty excited about that. He also offered to meet me for breakfast the following morning while we walked to the Relevance office for a late night tour.

Kris met Chad and I at the office as he’d kindly offered to shuttle me to and from his place for a night of gaming. We played two rounds of Dominion (the seaside expansion) and a train game I cannot remember the name of, but that was after a glass of bourbon. We played to about ten. It was a great way to finish out the evening.

I met Chad the next morning at Dos Perros a little Mexican place next door to Relevance. About half way through the meal I look up to see a familiar face outside the window. Holy crap! Its Brian Marick!

He’s in town doing his own craftsman swap – so to speak. I don’t know if he’d call it that, but his plan seems similar to mine – do some work with the top-notch Relevance crew. It’s always a good day when Brian walks through your door. How exciting.

We all made it up stairs and hung around a bit before the 8:45 all-hands standup. As stand ups go it was pretty much what I expected from an Agile shop. What did you do yesterday? what’s the plan for today? what’s blocking you? Brain and I were set up with our pairing partners as well, which looked to have been discussed before hand. No problem.

Rob Sanheim and I (my pair for the day) started off with an architectural overview of the project I was joining. The project itself is really interesting, but I can’t say much about it. NDAs and all. We had the usual sort of arrowed diagrams describing the major components of the system and their interactions. I was interested in how their testing strategy played out across the tiers of the application, so we drew some lines to show the boundaries between the test suites and where they overlapped. It was a useful experiment. It turned out that in a past life I had worked quite a lot in the domain, so most of it made sense to me at a high level. Good enough to get started.

The work environment at Relevance is really nice. Everyone sits together in one large room. The desks are huge with ample room for whatever you think you’ll need for the day. Each pairing station has two monitors, two keyboards, and two mice. Most of the shops I’ve visited do it that way and it works. The office itself had a really warm, relaxing atmosphere. Exposed wood beams, pine plank flooring and lots of windows.

We settled in and began our pairing session by reviewing the story in Mingle. It was larger than I usually like a story to be and Rob agreed that it was a bigger card than they usually played. We discussed breaking the story down. A significant refactoring was necessary, so I pitched doing that first as an “internal” dev-facing card. Rob seemed to like the idea initially, but thought it would be a hard sell for the customer – no value. The second pitch was to break the story by its two major functional components – configuration and execution. That didn’t make sense from a value standpoint either – the configuration part had no value to the customer without being able to execute it. All that discussion took about 5 minutes and in the end we decided to take on the whole card.

The card itself broke down like this. It had the usual title and a paragraph or so describing the story. That was followed up by four cucumber features that loosely described the acceptance criteria. These were written by a developer beforehand. The bottom of the card contained a Q/A section that listed responses to developer questions and anything outstanding. The whole card was sent to the customer for approval before we began work. I liked that a lot.

The part of the system we’d be changing is a Rails webapp. First things first, we made sure the test suite passed. I noticed Rob was using something like Autospec, but it wasn’t. It is called Watchr and it looks pretty awesome. My first big knowledge win for the week. It works like Autospec in that it automatically re-runs the test suite whenever something changes, but it gives you a lot more control over what's getting executed. Going to be a big win for me on my client project back home.

We had spent about an hour on the card when we were pulled into the (I think daily) project standup meeting with the customer. It was efficient and well run. We discussed all of the outstanding issues from last week and how they were being addressed. Ran through the story work for the day. It was pretty much your standard stad-up meeting.

15 minutes later we’re back to work. The communal stereo offered an eclectic mix of tunes to set the mood – pretty much any genre you can imagine. By lunch we’d managed to translate most of the first acceptance test from the story into something we could execute.

Relevance brings in lunch every day for its employees. Well that’s nice, but what’s the value? Is catering lunch every day just a perk or does it add something to Relevance’s culture? I think it does. First, everyone eats together every day. A lot of interesting discussions went on over lunch and everyone was there to participate. Second, we all got back to work at the same time. That is a fantastic feature if you’re pairing a lot – and Relevance certainly is.

The rest of the afternoon was standard Rails fare – a few migrations, some Refactoring, fixing specs. We did get a little hung up in our acceptance test where we wanted to assert the presence of a few fields in the view. We got something working with rspec’s include matcher, but the failure output was terrible. It dumped the entire response body. We blew a good 20 minutes trying to clean it up to no avail. We're using Micronaut (RSpec2) so have_tag wasn't available and webrat's selector doesn't work with assertions. We benched the cleanup when it became apparent it was going to be harder than we thought. It wasn’t that important.

Our day ended with an 1.5 hour company retrospective. I was told these happen every two weeks here at Relevance. We don’t do these at Obtiva and I really liked the idea. We do a lot more on-site work than Relevance does, at least it seems that way. An all-hands company retrospective is a great idea for us and I’m definitely going to take back to Obtiva with me.

Overall, it was a pretty awesome first day. The way they put together their projects really resonates with my own personal style. All the code we worked with is exactly what I expected to see. Very high quality, easy to comprehend and simple to test. I had enough confidence to take the driver’s seat after a few hours and I’m happy to say I was able to contribute a couple useful ideas on day 1. That speaks not to my own ability but to the quality of my pairing partner and the expressiveness of the system he was helping me to learn. Well done.

Looking forward to day 2.