Measuring time distribution of engineering test and production code

Abstract

Testing software is an important factor in application development. Without proper testing, software can be unreliable and show failures with drastic consequences. There have been attempts to measure how much time is spent on testing but these methods are not accurate as they depend on incomplete or subjective data. An Eclipse plugin called ‘WatchDog’ enables us to measure how much time a developer spends on reading and writing test and production code. Three different companies and one open source project took part in an experiment where a total of nine developers were monitored from two to seven months. After the results were collected, these developers were interviewed to explain the measurements. According to this data set, we noticed that time spent on test code can greatly differ as values from 0% to 67% were measured compared to time spent on production code. Values for reading code vary from 22% to 61% of time spent compared to writing code. This thesis goes into depth on this topic to look for explanations of the values that were measured.