When looking at the code quality of an application and trying to decide on a plan of action, it is not always enough to get a snapshot of the quality at a given time. You also need some visibility into the application history to take the right decision.
For example, let's say that you have defined a requirement that each application should have a coverage by unit tests of 60%. And let's say that one of your application has 50% coverage. Without historical data, you can only conclude that the team does not write sufficient unit tests and that it should become the next focus of attention. But let's now say that a couple of months ago the coverage for this application was 20%... then you would probably conclude that although the requirement is not made yet, the team is doing a very good job and that no special focus needs to be put on unit tests. This is why having access to historical information is important.
Historical information is provided through the Time Machine and the Tendencies.
The Time Machine
The Time Machine is made of two generic widgets that can be added to any dashboard.
The Timeline widget provides the ability to display a chart containing historical data of up to 3 metrics. Passing your mouse over the timeline will display the different values.
This widget can be customized:
History Table Widget
The History Table widget provides the capability to display a table with the historical data for up to 10 metrics:
This widget can be customized:
A default Time Machine dashboard comes out-of-the-box. This is a combination of Timeline and History Table widgets:
What are Tendencies?
The tendencies are arrows that are displayed next to metrics in the dashboards. Those arrows show the trend for the measure.
How to Read Tendencies?
There are five different levels of tendency
Black arrows are used to represent tendencies on the quantitative metrics (the ones that does not reflect quality, for example number of lines of code).
Red or green arrows are used to represent tendencies on the qualitative metrics (the ones that reflects quality, for example code coverage). Red is used when the quality decreases, green when it increases.
How are Tendencies Calculated?
To compute the tendencies, using a simple difference between the last two measures of each metric would not be accurate enough. Therefore a more advanced algorithm is implemented: the least squares method. It is a linear regression analysis that helps remove the noise in order to determine a trend on discrete measures. In other words, all the measure in the last XX days are taken into account, then the set of measures is checked to see if it makes sense (by testing the correlation rate), and an estimated slope is computed and displayed using arrows.
It is possible to configure the tendencies by logging in as an administrator and going to Settings > Configuration > General Settings > Differential Views.
- To configure the number of days used to compute tendencies, set the Tendency period property to the desired number of days. Default is 30.
Whatever the reason (wrong quality profile, issue with analysis, etc.), it is possible to remove a snapshot to polish the Time Machine widgets. Note that the last snapshot can never be deleted.
At project level, go to Configuration > History.
For every snapshot, it is possible to manually:
- Add, rename or remove a version
- Add, rename or remove an event
Removing Useless Data
When you run a new analysis of your project, some data that were previously available will no longer be (for example the source code of the previous analysis, measures at file level, etc.). These data will automatically be removed at the end of a new analysis.
Removing Unnecessary Analyses
It is very useful to frequently analyze a project to see how quality evolves. It is also very useful to be able to see the trends over weeks, months, years. But when you look back in time, you do not really need the same level of details as you would for one week ago. The Database Cleaner aims to delete some rows in the database to save space and to improve overall performances. Here is its default configuration:
- All data older than 5 years old are deleted
- For each project:
- only one analysis per month is kept over 1 year
- only one analysis per week is kept over 1 month
- only one analysis per day is kept over 1 day
- All closed issues more than 30 days old are deleted
- History at package/directory level is not kept
- Snapshots that include an event (new project version, new quality profile version, etc.) are not deleted
These settings can be changed at Settings > Configuration > General Settings > Database Cleaner.