Design by Contract for statecharts

About Design by Contract

Design by Contract (DbC) was introduced by Bertrand Meyer and popularised through his object-oriented Eiffel programming language. Several other programming languages also provide support for DbC. The main idea is that the specification of a software component (e.g., a method, function or class) is extended with a so-called contract that needs to be respected when using this component. Typically, the contract is expressed in terms of preconditions, postconditions and invariants.

Design by contract (DbC), also known as contract programming, programming by contract and design-by-contract programming, is an approach for designing software. It prescribes that software designers should define formal, precise and verifiable interface specifications for software components, which extend the ordinary definition of abstract data types with preconditions, postconditions and invariants. These specifications are referred to as “contracts”, in accordance with a conceptual metaphor with the conditions and obligations of business contracts. — Wikipedia

DbC for statechart models

While DbC has gained some amount of acceptance at the programming level, there is hardly any support for it at the modeling level.

Sismic aims to change this, by integrating support for Design by Contract for statecharts. The basic idea is that contracts can be defined on statechart componnents (states or transitions), by specifying preconditions, postconditions, and invariants on them. At runtime, Sismic will verify the conditions specified by the contracts. If a condition is not satisfied, a ContractError will be raised. More specifically, one of the following 4 error types wil be raised: PreconditionError, PostconditionError, or InvariantError.

Contracts can be specified for any state contained in the statechart, and for any transition contained in the statechart. A state contract can contain preconditions, postconditions, and/or invariants. The semantics for evaluating a contract is as follows:

  • For states:
    • state preconditions are checked before the state is entered (i.e., before executing on entry), in the order of occurrence of the preconditions.
    • state postconditions are checked after the state is exited (i.e., after executing on exit), in the order of occurrence of the postconditions.
    • state invariants are checked at the end of each macro step, in the order of occurrence of the invariants. The state must be in the active configuration.
  • For transitions:
    • the preconditions are checked before starting the process of the transition (and before executing the optional transition action).
    • the postconditions are checked after finishing the process of the transition (and after executing the optional transition action).
    • the invariants are checked twice: one before starting and a second time after finishing the process of the transition.

Defining contracts in YAML

Contracts can easily be added to the YAML definition of a statechart (see Defining statecharts in YAML) through the use of the contract property. Preconditions, postconditions, and invariants are defined as nested items of the contract property. The name of these optional contractual conditions is respectively before (for preconditions), after (for postconditions), and always (for invariants):

contract:
 - before: ...
 - after: ...
 - always: ...

Obviously, more than one condition of each type can be specified:

contract:
 - before: ...
 - before: ...
 - before: ...
 - after: ...

A condition is an expression that will be evaluated by an Evaluator instance (see Include code in statecharts).

contract:
 - before: x > 0
 - before: y > 0
 - after: x + y == 0
 - always: x + y >= 0

Here is an example of a contracts defined at state level:

statechart:
  name: example
  root state:
    name: root
    contract:
     - always: x >= 0
     - always: not active('other state') or x > 0

If the default PythonEvaluator is used, it is possible to refer to the old value of some variable used in the statechart, by prepending __old__. This is particularly useful when specifying postconditions and invariants:

contract:
  always: d > __old__.d
  after: (x - __old__.x) < d

See the documentation of PythonEvaluator for more information.

Executing statecharts containing contracts

The execution of a statechart that contains contracts does not essentially differ from the execution of a statechart that does not. The only difference is that conditions of each contract are checked at runtime (as explained above) and may raise a subclass of ContractError.

from sismic.interpreter import Interpreter, Event
from sismic.io import import_from_yaml

statechart = import_from_yaml(filepath='examples/elevator/elevator_contract.yaml')

# Make the run fails
statechart.state_for('movingUp').preconditions[0] = 'current > destination'

interpreter = Interpreter(statechart)
interpreter.queue('floorSelected', floor=4)
interpreter.execute()

Here we manually changed one of the preconditions such that it failed at runtime. The exception displays some relevant information to help debug:

Traceback (most recent call last):
 ...
sismic.exceptions.PreconditionError: PreconditionError
Object: BasicState('movingUp')
Assertion: current > destination
Configuration: ['active', 'floorListener', 'movingElevator', 'floorSelecting', 'moving']
Step: MicroStep(transition=Transition('doorsClosed', 'movingUp', event=None), entered_states=['moving', 'movingUp'], exited_states=['doorsClosed'])
Context:
 - current = 0
 - destination = 4
 - doors_open = False

If you do not want the execution to be interrupted by such exceptions, you can set the ignore_contract parameter to True when constructing an Interpreter. This way, no contract checking will be done during the execution.