Rapid Feedback

A ManagementPattern use to get out from FearCulture.

Workers need rapid feedback to improve quickly and BecomeCompetent.

Workers who underperform usually do so because they do not have enough feedback.

A bit of citation of evidence would make this statement a bit more credible

I read about this in a ManagementBook?. I will see if I can find the title.

Some ExtremeProgrammers have argued that running all their tests takes 2 hours or more. That clearly is the opposite of RapidFeedback. RapidFeedback means that the time it takes to compile must be comparable to the time it takes to run the tests. I do not want to wait for a compilation longer than 1 minute, so this rules out CeePlusPlus.

So far the only exception of RapidFeedback I've found is when running UnitTests against the database. I usually do thing ModelFirst. Then I create a DatabaseLayer? that implementes something like a DAO. Before I create the DatabaseLayer?, I must create the actual database and I want that to be in ConcurrentVersionsSystem too, therefore I simply create a big script that creates the database, the stored procedures (if there are any) and adds the basic data. Then I even run the most basic UnitTests in the same database script.

If that script is called install.sql, then another script is called uninstall.sql which simply removes the database. install.sql is a requirement for installing the application and a test that the database is compliant. Once install.sql is run, the DatabaseLayer? UnitTests come into action. Usually install.sql takes a lot of time: Almost as much as 5 minutes for 50 tables which is not much. YMMV. -- GuillermoSchwarz

This was also mentioned in an XP context: http://www.umsl.edu/~sauter/analysis/xp/sld031.htm

What about feedback for fatigue? I mean for a mental block, psychological inertia? Do something quick that affects the definition of the problem.

Sometimes it turns out you can do a lot less work than what you thought if you just put in some simple code and then test its behavior right away to see if the edge/special/nested cases are worth the trouble. That is the problems are not severe enough, or perhaps not recognizable as problems, or perhaps what the behavior you didn't bother to stamp out is usable as a feature.

Heh, this is funny. I just got flogged for delivering a nice-looking proof-of-concept.

See: TestsTooSlow