Model Testing of Executable Statecharts using SISMIC

Model Testing of
Executable Statecharts
Tom Mens Alexandre Decan
Software Engineering Lab
University of Mons, Belgium

Tom Mens – IPA Fall Days – Zandvoort, Nederland – 4 November 2024

Statecharts
(hierarchical state machines)
• Invented in 1987 by David Harel
• Used to specify, design and control the complex
behaviour of discrete-event systems
– embedded systems, user interfaces, network
protocols, robots, computer game logic, ...
• Executable semantics
– Can be used to generate executable source code, if a
statechart compiler is available
– Can be integrated in an existing code base directly, if a
statechart interpreter is available

Statecharts
A simple example
An Elevator controller
Visualised with PlantUML (https://plantuml.com)

Statecharts
Existing Tooling
Many existing tools
• Itemis Create: A statechart simulator and code generator
https://www.itemis.com/en/products/itemis-create/
• Mathworks Stateflow
https://www.mathworks.com/products/stateflow.html
• QuantumLeaps QM and QP for embedded software development
https://www.state-machine.com
• IBM Rational Statemate
• IBM Rational Rhapsody
• SISMIC: A Python-based statechart interpreter
https://pypi.org/project/sismic/ focus of this talk!

Running example
A microwave oven controller
Created with
Itemis Create

• Many “agile” development techniques
provide lightweight approaches to facilitate
change and increase reliability of software
• Design by contract
• Test-driven development
• unit testing
• behavior-driven development
• Dynamic verification of properties
• We raise these techniques to the level of
executable statechart models
Defensive software development

SISMIC
Architecture
g and validating executable statechart models 843
al overview
framework
validate statecharts, by using the techniques editor and import these models through the I/O API provided
...
...
https://sismic.readthedocs.io https://github.com/AlexandreDecan/sismic

Running example
statechart:
name: Microwave controller
preamble: |
POWER_VALUES = [300, 600, 900, 1200, 1500]
POWER_DEFAULT = 2 # 900W
MAXPOWER = 3 # 1200W
root state:
name: controller
initial: door closed
on entry: |
power = POWER_DEFAULT
timer = 0
transitions:
- event: cooking_stop
action: |
timer = 0
states:
- name: door closed
initial: closed without item
states:
- name: closed without item
transitions:
- event: door_opened
target: opened without item
- name: closed with item
...
SISMIC adopts
a YAML-based
specification
of statecharts

Running example
A micowave oven controller
SISMIC supports
visualisation
through PlantUML

Running example
Example: Basic GUI to
control the executable
statechart model
SISMIC provides an API
for Python code to use
the statechart model

SISMIC
Macrostep semantics
Stepwise execution of statechart behaviour
from sismic.io import import_from_yaml
from sismic.interpreter import Interpreter
from sismic.model import Event
with open('microwave.yaml') as f:
statechart = import_from_yaml(f)
interpreter = Interpreter(statechart)
interpreter.execute_once()
MacroStep(None, [], >['controller', 'door closed', 'closed without item'], <[])
interpreter.queue(Event(’door_opened’))
interpreter.execute_once()
MacroStep(Event(door_opened), [Transition(closed without item, opened without
item, door_opened)], >['door opened', 'opened without item'], <['closed without
item', 'door closed'])

Software-controlled systems
are difficult to develop
Control software can be very complex
– Continuous interaction
– between software and hardware
– with external world and users
– Must respect
– functional requirements
• Oven should cook food placed in oven
with specified power and duration
– non-functional requirements
• Oven should stop cooking if doors are opened
14

• Based on Bertrand Meyer’s “Design by Contract” in 1986
• Add precise and dynamically verifiable specifications to
executable code components (e.g., methods, functions,
classes) or model components (e.g., states, transitions)
• The code (or model) should respect a contract composed of
– preconditions
– postconditions
– invariants
Contract-driven development

Example (from https://www.eiffel.com/values/design-by-contract/ introduction)
Contract-driven development
class DICTIONARY [ ELEMENT ]
feature
put (x : ELEMENT; key : STRING ) is
require
count <= capacity
not key.empty
ensure
has (x)
item (key) = x
count = old count + 1
end
invariant
0 <= count
count <= capacity
end

Contract-driven modelling
contract for controller state:
invariants:
not sent(heating_on) or active(cooking mode)
timer>=0
0 < power <= MAXPOWER
contract for cooking mode state:
precondition: timer>0
invariants:
timer >= 0
power == power@pre
postcondition: received(door_opened) or timer==0
contract for ready state:
invariant: timer > 0

Contract-driven modelling
statechart:
name: Microwave controller with contracts
preamble: | # as on earlier slide
root state:
name: controller
initial: door closed
on entry: |
timer = 0
# ...some stuff removed here...
contract:
- always: not sent('heating_on') or active('cooking mode’)
- always: timer >= 0
- always: 0 <= power <= MAXPOWER
states:
- name: door closed
initial: closed with item
states:
- name: closed with item
states:
- name: cooking mode
# ... some stuff removed here...
contract:
- before: timer > 0
- always: timer >= 0
- always: power == __old__.power or received('cooking_stop’)
- after: received('door_opened') or timer == 0

SISMIC
Macrostep semantics revisited
A method for testing and validating executable statechart models 845
Fig. 8 Sismic’s macrostep
semantics, adding support
required for dynamic monitoring
of contracts and properties
with support for
runtime monitoring
of contracts

SISMIC
Macrostep semantics revisited
interpreter.queue('door_opened’, 'item_placed', 'door_closed')
interpreter.queue('timer_dec', 'cooking_start', 'tick')])
interpreter.execute()
MacroStep(None, [], >['controller', 'door closed', 'closed without item'], <[]),
MacroStep(Event(door_opened), [Transition(closed without item, opened without item,
door_opened)], >['door opened', 'opened without item'], <['closed without item', 'door
closed']),
MacroStep(InternalEvent(lamp_on), [], >[], <[]),
MacroStep(Event(item_placed), [Transition(opened without item, opened with item,
item_placed)], >['opened with item'], <['opened without item']),
MacroStep(Event(door_closed), [Transition(opened with item, closed with item,
door_closed)], >['door closed', 'closed with item', 'program mode', 'not ready'],
<['opened with item', 'door opened']),
…

Example of failing
contract
InvariantError
State: controller
Assertion: timer >= 0
Configuration:
[controller, door closed, closed with item,
program mode, not ready]
Step:
event timer_dec
internal transition on closed with item

Solution to failing
contract
Add guards to the actions associated to the events
that increment and decrement power and timer
timer_dec [timer>0] / timer = timer - 1
power_inc [power<MAXPOWER] / power = power+1
power_dec [power>1] / power = power -1

Without guards on
timer_dec event
Test-driven
development
class MicrowaveTests(unittest.TestCase):
def setUp(self):
# code here for setting up statechart interpreter
def test_negative_timer(self):
self.interpreter. queue (door_opened, item_placed, door_closed, timer_dec)
self.assertEqual(self.interpreter.context['timer'], 0)
test negative_timer ... FAIL
===========================================
AssertionError: -1 != 0
----------------------------------------------------------------------
Ran 1 tests in 0.005s
FAILED (failures=1)
Using Python’s unittest module

Test-driven
development
class MicrowaveTests(unittest.TestCase):
def setUp(self):
# code here for setting up statechart interpreter
def test_negative_timer(self):
self.interpreter. queue (door_opened, item_placed, door_closed, timer_dec)
self.assertEqual(self.interpreter.context['timer'], 0)
Using Python’s unittest module
test negative_timer ... ok
----------------------------------------------------------------------
Ran 1 tests in 0.005s
OK
With guards on
timer_dec event

Coverage analysis
State coverage: 81.82%
Entered states:
controller (3) | door closed (4) |door opened (2) |
closed without item (3) | opened without item (2) |
opened with item (2) | closed with item (1) |
not ready (1) | program mode (1)
Remaining states:
cooking mode | ready
Transition coverage: 16.67%
Processed transitions:
opened without item [item_placed] -> opened with item (2)
closed without item [door_opened] -> opened without item (2)
opened with item [door_closed] -> closed with item (1)

• Include customer test practices into TDD
• Encourage collaboration between developers, QA,
and non- technical stakeholders (domain experts,
project managers, users)
• Reduce technical gap between developers and
other project stakeholders
• Use a domain-specific textual language to
specify how the code should behave
– By defining feature specifications and scenarios
– Using Gherkin language
– Supported by Cucumber framework
Behaviour-Driven
Development

Example
(from docs.behat.org/en/v2.5/guides/1.gherkin.html)
Behaviour-Driven
Development
Feature: Serve coffee
In order to earn money customers should be able to buy coffee
Scenario: Buy last coffee
Given there is 1 coffee left in the machine
And I have deposited 1 dollar
When I press the coffee button
Then I should be served a coffee

Behaviour-Driven
Modeling
Feature: No heating if door is opened
Scenario: No heating when nothing is done
When I power up the microwave
Then heating should not be on
Scenario: No heating when item is placed
Given I open the door
When I place an item
Then heating should not turn on
Scenario: No heating when door is not closed
And I place an item
When I close the door
Microwave scenarios
Described in (structured) textual
language, without any knowledge
of the statechart’s internals.
1 feature passed, 0 failed, 0 skipped
3 scenarios passed, 0 failed, 0 skipped
11 steps passed, 0 failed, 0 skipped, 0 undefined
Took 0m0.005s

Behaviour-Driven
Modeling
Given I do nothing
And I execute the statechart
Then state cooking_mode should not be active
And event heating_on should not be fired
Given I send event door_opened
When I send event item_placed
Then event heating_on should not be fired
Given I send event door_opened
And I send event item_placed
When I send event door_closed
Then event heating_on should not be fired
Intermediate
(statechart-specific) scenarios
Automatically generated
from high-level scenarios
through user-defined mapping.

Behaviour-Driven
Modeling
from sismic.bdd import map_action, map_assertion
map_action('I open the door', 'I send event door_opened’)
map_action('I close the door', 'I send event door_closed’)
map_action('I place an item in the oven', 'I send event item_placed’)
map_action('I press increase timer button {time} times', 'I repeat "I send event
timer_inc" {time} times’)
map_action('I press increase power button', 'I send event power_inc’)
map_action('I press start button', 'I send event cooking_start’)
map_action('I press stop button', 'I send event cooking_stop’)
map_action('{tick} seconds elapsed', 'I repeat "I send event timer_tick" {tick} times’)
map_assertion('Heating turns on', 'Event heating_on is fired’)
map_assertion('Heating does not turn on', 'Event heating_on is not fired’
map_assertion('heating turns off', 'Event heating_off is fired’)
map_assertion('lamp turns on', 'Event lamp_switch_on is fired’)
map_assertion('lamp turns off', 'Event lamp_switch_off is fired')
Mapping
specification

SISMIC
Behaviour-Driven Modeling
from behave import given, when, then
from sismic.io import import_from_yaml
from sismic.interpreter import Interpreter
from sismic.interpreter.helpers import log_trace
from sismic.model import Event
@given('I execute the statechart')
def execute_statechart(context):
_execute_statechart(context, force_execution=True)
@then('state {state_name} should be active')
def state_is_active(context, state_name):
assert state_name in context._statechart.states, 'Unknown state {}'.format(state_name)
assert state_name in context._interpreter.configuration, 'State {} is not
active'.format(state_name)
Using Python’s behave module
$ sismic-behave microwave.yaml --features heating.feature

SISMIC
Property statecharts
Define and verify behavioural properties by
1. instrumenting the statechart interpreter
2. intercepting specific actions of the statechart being
executed
• entered(<NAME OF STATE>)
• exited(<NAME OF STATE>)
• consumed(<NAME OF EVENT>)
• sent(<NAME OF EVENT>)
• …
3. executing a statechart that verifies a desirable or
undesirable property
<<property statechart>>
Heating does not start if door is opened
door is
closed
door is
opened
consumed(door_closed)
consumed(door_opened)
failure
sent(heating_on)

SISMIC
Property statecharts
Examples
Heating does not start if door is opened
door is
closed
door is
opened
consumed(door_closed)
failure
sent(heating_on)
Heating must stop when door is opened
heating
is off
heating
is on
sent(heating_off)
sent(heating_on)
heating is on while
door is opened
failure
consumed(tick)
sent(heating_off)

SISMIC
Communicating statecharts
• Statecharts can communicate with other
statecharts or external components (e.g. a
user interface) by sending/receiving events
• This is realised by binding their statechart
interpreters

SISMIC
Communicating statecharts
Example: Events sent by buttons are propagated to elevator
buttons = Interpreter(import_from_yaml(open(‘buttons.yaml')))
elevator = Interpreter(import_from_yaml(open(‘elevator.yaml')))
buttons.bind(elevator)
buttons.queue(Event(’floor_2_pushed'))
buttons.execute()
Awaiting events in buttons: [Event(button_2_pushed)]
Awaiting events in buttons: [InternalEvent(floorSelected, floor=2)]
Awaiting events in elevator: [Event(floorSelected, floor=2)]
elevator.execute()
print('Current floor:', elevator.context.get('current'))
Current floor: 2

SISMIC
Other features
• Use other semantic variants of statecharts by
overriding private methods of Interpreter
• E.g. outer-first instead of inner-first semantics;
changing priority of events
• Different ways of dealing with time
– Real time versus simulated time

• Model quality assessment (model smells) and improvement (model
refactoring)
– E.g. splitting up a complex statechart into multiple interacting statecharts
• More advanced testing techniques
– Automatic generation of contracts based on scenario specifications
– Automatic generation of tests based on contract specifications
– Mutation testing
– ...
• Explore (dynamic?) model checking techniques
– E.g., based on LTL formulas, using Dwyer’s specification patterns, others?
• Model variability analysis
• Consider product families (e.g. different microwave variants) and analyse commonalities
and variabilities in their statechart models
• Design space exploration
• Analyse pros and cons of syntactically different, but semantically similar
statecharts
• Model composition and scalability
• Semantic variation
• Detecting if statechart is compatible with alternative statechart semantics
Future Work

Conclusion / Summary
When I power up the microwave
Then heating should not be on
When I place an item
And I place an item
When I close the door
ng and validating executable statechart models 843
ral overview
framework
d validate statecharts, by using the techniques
BDD, and dynamic monitoring of contracts
expressed over statecharts.
and support the proposed method, we devel-
c tool. Sismic is a Python library for inter-
ing statecharts. Version 1.0.0 has been used
It is distributed through the Python Pack-
i.python.org/pypi/sismic). Its documentation
on sismic.readthedocs.io. Its source code is
thub.com/AlexandreDecan/sismic under the
ence LGPLv3, adhering to the principles of
d open research. It allows other researchers to
the tool, and it facilitates integrating received
ewer versions of the tool.
vel architecture of Sismic is summarised in
ferent code components and work products
ture will be presented in the following sub-
rts of the architecture included in the shaded
d sismic havebeendevelopedspecificallyfor
statechart simulation, testing and validation.
se activities, Sismic makes use of third-party
ehave for BDD) or external tools (e.g. Plan-
EME for model visualisation and editing).
ut files used by Sismic are shown in yellow
editor and import these models through the I/O API provided
forthispurpose.Uponimportofastatechartfile,thesyntactic
correctness of the statechart is checked automatically.
Currently, Sismic allows importing and exporting state-
chart expressed using the human-readable YAML markup
language and has experimental support for importing and
exporting statechart diagrams expressed in the AMOLA lan-
guage [46] through the ASEME IDE2 statechart editor [48].
Sismic also provides export support to PlantUML3 in order
to visualise statecharts, benefiting from its automatic layout
features.ThisishowwegeneratedFig.14forexample.Other
exchange formats can be easily accommodated.
4.2 Statechart interpreter
The core of Sismic is composed of its statechart interpreter.
In order to execute a statechart model, a statechart inter-
pretermustbeinstantiated.Thisinterpreterreliesonanaction
code evaluator to execute any action code contained in the
statechart specification. By default, action code and guards
(conditions) are expressed using regular Python code. Other
action languages can be supported if a language interpreter
is provided for them.
The statechart interpreter offers a discrete, step-by-step
...
...
Heating must stop when door is opened
heating
is off
heating
is on
sent(heating_off)
sent(heating_on)
heating is on while
door is opened
failure
consumed(tick)
sent(heating_off)
contract for controller state:
invariants:
not sent(heating_on) or
active(cooking mode)
timer>=0
0 < power <= MAXPOWER
contract for cooking mode state:
precondition: timer>0
invariants:
timer >= 0
power == power@pre
postcondition:
received(door_opened) or timer==0
contract for ready state:
invariant: timer > 0

Model Testing of Executable Statecharts using SISMIC

Recommended

More Related Content

Similar to Model Testing of Executable Statecharts using SISMIC (20)

More from Tom Mens (20)

Recently uploaded (20)

Model Testing of Executable Statecharts using SISMIC