From Maisqual Private Wiki

Jump to: navigation, search

This is the milestone for the month of August, 2012.


[edit] Architecture

[edit] Models

We don't want to be dependent on a specific model for quality and practices assessment, which would introduce a bias in the correlation[1]. For that reason, we decided to gather and analysis many different base measures, and to build different practices and quality models from them afterwards.

[edit] Gathering data

[edit] Sampling

Detection of practices and evolution of quality needs a fine sampling frequency. Example frequencies are per-version, or time-based (e.g. per month, per week, per day). For small samples, we get more precise information but face a huge-volume problem. Furthermore, the estimated delay for changes to occur after an action is around a few weeks. We decided to build two indexes, one for rough picturing of evolution of the project (per version), and another for fine analysis (per week).

[edit] Tools

Retrieval of data heavily depends on the tools used by the project.

[edit] Output

We defined a standard format for our data.

Projects that do not provide some information are included with NA's as missing values.

[edit] Analysis of data

Once all metrics are put in a standardised format, we are able to apply a (mostly) automated process of analysis to all projects.

These processes are described in this section.

[edit] Types of data

The sources for our input data are the following:

  • Code metrics, extracted via SQuORE.
  • Mailing lists and Forums, as means of communication between the team members.
  • Configuration Management
  • Process-Oriented

[edit] Code metrics

It is important to understand what are the metrics gathered, and their so-far known consequences. They have been described on the public maisqual wiki.

There are two axes for the code metrics: the metrics for all files that belong to a build, and the evolution of the application metrics across versions.

[edit] Tools

R scripts have been written for automatic generation of graphics

[edit] Projects

[edit] Apache Ant

We wrote an article to describe analysis of Ant with R.

[edit] GCC

[edit] References

  1. We basically want to correlate actions and results, not the way we measure quality.
Personal tools