Alambic

One of the first things to setup when considering to use Alambic is the quality model.

In a nutshell, the quality model states what is important to you (your definition of quality, in this very context) and how to measure it. As an example, let's say we want to check the activity of our community. We decide to measure it through a set of metrics retrieved from mailing lists, forums and configuration management (git):

polarsys_quality_model_adapted.png

So the chain looks like:

On the other hand from the bottom up measures are converted to indicators using a 4-thresholds scale. Indicators are then aggregated up from the leaves to the root of the quality model tree. As an example:

The Activity's value is the average of all its children, which is in this case 3 out of 5. Other attributes are computed the same way, and values are aggregated up to the root of the quality model.

Terminology

Quality attributes

Quality attributes represent features of quality you want to achieve, like maintainability (for product quality) or activity (for community quality). Since everybody has a different meaning for such characteristics, it is vital to precisely define them to setup a semantic foundation for all participants. It is a good idea to rely on existing models for these definitions, like ISO 9126 or ISO 250xx.

Indicators

Indicators range from 1 (bad) to 5 (excellent). They are entirely subjective, since they will vary depending on the project's context and on people's understanding of what is good or not. In other words they are a localised judgement on the raw numbers collected from the data sources.

Scales

Scales define what is bad, good or excellent regarding the collected numbers. They are a way of adapting raw numbers to the local context.

Scales provide thresholds, so they are defined as an array of 4 integers as in: [15, 30, 50, 80]. In this case, percentages from 0 to 15% would be classified 1/5, from 16 to 30% would be 2/5, from 31 to 50% would be 3/5, from 51 to 80% would be 4/5, and over 81% would be 5/5.

alambic_presentation_metrics.jpeg

It should be noted that:

Metrics

Metrics are the base measures used to assess the quality model. Examples of metrics include Number of emails sent on the developer mailing list during last week or Number of commits during last week.