This sample models a population whose members can be added or removed. It demonstrates state-dependent domains.

There is one state variable population, a set. In these examples all of the elements of the set are integers. (A set is a collection without duplicates and without order; set is a class in the Python builtin module.) There are two actions, add and remove, each with one argument: the element to be added or removed.

The domains for each action that are defined in the populations module are not fixed collections; they are lambda expressions that are evaluated each time the action might be invoked. Making them callable forces their re-evaluation each time. The domain for add evaluates to a list containing a single random integer (which is different each time the domain is evaluated). The domain for remove evaluates to the population set itself, so any argument chosen from this domain is guaranteed to be present in the population. Since the population is a collection of random elements, it would not be possible to define a fixed domain in advance.

The domains9 configuration module redefines the domains as functions (not lambda expressions), which also works. The domainsx configuration redefines the domains as collections (not callables), which does not work --- they are not re-evaluated each time they are used, but are evaluated only once when the module is imported.

The modules in this sample are:

• populations: model program, includes state-dependent domains defined using lambda expressions.

• domains9: configuration module that redefines domains using functions (not lambda expressions). This also works.

• domainsx: configuration module that redefines domains using collections (not functions or lambda expressions). This does not work.

• filter3: configuration module that defines a StateFilter that limits the population to no more than three elements.

• filter3_lc: like filter3, except statefilter is spelled with all lowercase letters, as recommended in the Python style guide, PEP8. This works too.

• test: one short run of the model program.

• test_filter: first, a run without any state filter to show how the population can grow, then runs with both state filters to show they both limit the population to no more than three elements.

• test_domains: first, a run with the domains defined in the model program, to show the element values can range up to 99, then runs with both domain configurations to show the element values only range up to 9.

• test_domainsx: a run that shows domainsx does not work (the domain is not a callable)

• test_seed: demonstrate the -s (seed) option to pmt, which specifies the random seed used to select arguments and actions, so runs are reproducible. Whenever this script is executed, the same traces are generated.

• test_populations: a test script that uses the Python standard library unittest module, instead of our own trun module. The command to run this test is: python test_populations.py

There is no stepper in this sample.

Revised Apr 2013