Programmer at work: statistics from a two year project (gitstats)

By Chris Done

Recently I discovered a tool called gitstats, which generates some interesting statistics from commits made to a Git project. I thought it would be fun to reflect on a particular project I have been working on.¹

The project

For roughly the past two years I have been working on a web application.² Actually, gitstats tells me that it has been exactly 456 days.³ I am the only developer on this project. I thought I’d write up on some simple statistics from this project.

The application is a research paper submission and publication system, among other things.

Scope

The project program code spans several languages, including:

• Haskell: 125 modules, 17,837+ lines of code
• JavaScript: 92~ modules, 21,854+ lines of code
• CSS: 4,385 lines of code
• Java: 734 lines of code
• HTML, SQL: negligable raw code, most generated by JS or Haskell

For features and bugs, there have been 68+ pushes to production, resulting in an internal tag version of 2.6.5, with an average of 37 commits per tag (release).

I tried out this Ruby script that generates stats of inserts/deletions, and the output is:

Git scores (in LOC):
unknown           :  +3469917      -3150559

I don’t know if any of these details help to paint a picture of the scope of the project, but that’s it.

Commits over the year

By year

I have produced substantially more code in 2011 than in 2010, which feels somewhat accurate. I started in august of 2010, so this seems to add-up.

By month

The pattern you might expect happens across the months year; the most work occurs in Autumn where tensions and expectations are high for things to be completed before the end of the year. Winter is expectedly very quiet due to the holidays and effective low demand. Summer, equally, is a very lazy and unproductive time⁴, partly due to the heat and partly due to holidays.

Commits during the day

This result confirmed my suspicions, that I’m most productive in the morning before lunch, and then a little around 5ish. The table below confirms it; I make most commits just before lunch or just after lunch, because I did all the work in the morning. After lunch I’m always very sluggish and lethargic. Somehow it appears I still manage to finish commits I started in the morning. I very rarely work in the evening, though it does happen.

Commits during the week

In this case I was surprised. I’d always thought that my productivity was high on average, but very sporadic. I’m aware that some days once in a while I just don’t feel like doing anything/can’t get in the groove.

However, from the chart below it seems that, on average, there is a distinct, and perfectly reasonable pattern.

• I start mondays with a surprising amount of energy; I hit the ground running, clocking 491 commits in total for monday, or 7.5 commits per monday on average.
• I go through tuesdays and wednesdays churning out code with presumably very good flow, clocking 561 and 563 commits respectively, or 8.6 commits each day.
• But then on thursday I drop down to less than monday’s great start, to a mere decently good⁵ 422 (or 6.4 for the day), implying that, somehow, for me thursdays are either less tense and urgent; I feel more relaxed, or it could be that I burn out by thursday after the bursts on monday, tuesday and wednesday and start pacing myself.
• Finally on fridays, I drop down to 329. Clearly by friday I am ready for the weekend, and so is everyone else. Nothing gets pushed to production on a friday. There is little urgency about it, more time for discussion and reflection for next week. Only 5 commits on average for this day.
• Sometimes I do some work on weekends. I thought I never did anything on weekends but there is some activity.

Are commits a good indication of work done?

I think so. It’s true for sure that I can work all day and not commit anything, but on average I break my tasks into smaller tasks that can fit into a few hour’s worth of work and one or two commits.

A commit is a complete piece of work. This is a much better indicator of productivity than merely being at the office, sat at the computer.

I will also say that my stats are very similar to a colleague of mine, on a project of about the same size, who exhibits similar patterns of days of the week, but different hours of the day (he’s more active in the afternoon).

What did I learn from this information?

One thing I’ve learned is that my performance is fairly consistently spread in an uneven way. It would be interesting to pace myself better over the week so that I maintain a pace like monday across the board. I’m not sure if this would increase my morale or weaken it.

It might be wise for me to eat smaller meals throughout the day, rather than a big lunch. This seems to really slow me down; I’m not sure whether it’s the full stomach or just a necessary recharge. Also, if I can’t change this pattern, it might be good for me to start scheduling non-coding/hard stuff to that period of the day, such as cleaning up tickets and organizing/planning.

Mildly curious how everyone else works

I would be interested to see people’s work patterns based on this gitstats tool for their activity over the year, over the week, and during the day. Especially if you have a year or more worth of data.