This module defines the top-level behaviour for targeted property-based testing (TPBT). Using TPBT the input generation is no longer random, but guided by a search strategy to increase the probability of finding failing input. For this to work the user has to specify a search strategy and also needs to extract utility-values from the system under test that the search strategy then tries to maximize.

To use TPBT the test specification macros forall_targeted, exists, and not_exists are used. The typical structure for a targeted property looks as follows:

property prop_target do          # Try to check that
  exists input <- params do      # some input exists that fullfills the property.
      uv = sut.run(input)        # Do so by running SUT with input
      maximize(uv)               # and maximize its utility value
      uv < threshold             # up to some threshold, the property condition.
  end
end

Options

For targeted properties exists a new option:

• {:search_steps, non_negative_number}
  takes an integer defining how many search steps the searching algorithm takes. Its default value is 1_000. The effect of search_steps is similar to num_tests for ordinary properties. num_tests has no effect on the search strategy. This helps when you combine a regular property with search strategy, e.g. generating a configuration parameter and search for specific properties to hold depending on that parameter.

Most of the documentation is taken directly from PropEr.