Improve the possibilities of collecting stats

[stevana]

Ideas on how we can improve stats collecting:

• http://www.quviq.com/getting-better-statistics-from-state-machine-tests/
• http://quviq.com/documentation/eqc/

[stevana]

The Cleanroom people have some neat tools, e.g. JUMBL, for analysing Markov chains. For more information see:

• http://jumbl.sourceforge.net/
• https://sourceforge.net/projects/jumbl/files/jumbl_user.pdf/download
• https://www.researchgate.net/publication/265184529_Computations_for_Markov_Chain_Usage_Models

[stevana]

Some important progress was made in the blog post "An in-depth look at quickcheck-state-machine", see especially the "Labelling" section.

Some of those ideas are now part of the Test.StateMachine.Labelling module and used in the process registry example.

The two papers that seems to introduce the labelledExamples combinator in QuickCheck:

• How well are your requirements tested? [PDF]
• Understanding Formal Specifications through Good Examples [PDF, video]

Seem to somehow be able to produce the minimal labelled examples without the user having to tag everything manually though? So perhaps there's more work to be done on automatically tagging?

[stevana]

There's also features/3 and call_features/{1,2} in the eqc statem API, not sure if they are part of the labelled examples machinery though.

[stevana]

Another talk by John Hughes that explains labelledExamples is called Building on developers' intuitions to create effective property-based tests.