Testing

• Don't skip it because you are in a hurry
• Gives you confidence to change your code
• Write tests at the correct level
• Most your tests should be unit tests
• You do need functional / integration tests
• Check your code coverage
  • go test -cover

Testing Repositories

• gocheck​
  • gopkg.in/check.v1
  • way better than the built in standard library
  • provides test suites, asserts, and checkers
• github.com/juju/testing
  • stand alone test library using gocheck
  • isolation and cleanup base suites
  • additional checkers

Asynchronous Go

• Learn to use channels properly
• Use sync.Mutex when necessary
• Be careful with goroutine lifetimes

Parallel executing is hard

• Just because you think something should be running before something else doesn't mean it will
• Run tests with the race flag
  • go test -race
• Avoid using time.Sleep and time.After in production code
  • timing issues between goroutines can be very hard to track down
  • consider a Clock interface, like the one defined in github.com/juju/utils/clock

Channels

• Channels will block
  • even buffered will channels block
• Never use naked channel reads or writes in production code

Naked Read / Write

func something(events <-chan event) {

    // prelude

    next := <-events

    // do something with next

}

Select Read / Write

func something(events <-chan event, done <-struct{}) {

    // prelude

    select {

        case next := <-events:

            // do something with next

        case <-done:

            return

    }

}

Developing

• Focus
  • Thrashing is bad, very bad
• If you can, pair early on development
  • This shares knowledge and validates approach
• Write tests
• Code review
• Feature flags are good, feature branches not so much

Developing a feature

• Break the task down into pieces
• Each piece should be no more than 2-4 days of work
  • preferably 2
• Measure the development process
  • if someone is on a task that should only take two days, but is taking longer, go talk with them, or if it is you, seek help
  • either the approach is wrong, or sizing was misjudged and needs to be re-evaluated
• Burn down charts of feature development can help motivation