Actually run code samples in the documentation

[filipsch]

The 'trust' in these examples is low, as they are not being actually run. Running the examples is pretty hard to do, as well. Investigate whether we can actually run the code samples and have them generate informative output.