Wednesday 17 August 2016

Using BDD for State Transition diagram testing

Overview

In Agile software development lifecycle, Behaviour Driven Development (BDD), proved to be a useful testing technique to define the test cases based on the expected behaviour of the software. On of the main characteristics is that the defined test steps follow the below pattern :
Given some initial context,
When an even occurs
Then ensure some outcomes
This technique is used and there are many examples on the web on how to use it, found in different places :
To get the maximum value from the usage of this technique, the following guidelines were proposed :
  • Describe what not how, e.g. it is not particularly useful to define all the 'clicks' and 'reloads' to validate a user story
  • One test, one topic, each test should ideally be focused on one topic and each topic should be ideally described by one test. Otherwise, a test that validates multiple actions, it will be fragile and will cause a lot of maintenance. 

The problem

When it comes to test workflows that involve different personas interacting with the application, it might be a challenge to create from scratch a complete test case to test the valid or non-valid transactions. 

For instance, lets consider a web based application that implements the following workflow transition diagram :


1. When a customer submits successfully a form and the 'customer case' object is created and its status is moved to 'Open'
2. When the analyst assigns to himself the case, the 'customer case' status is moved to 'In progress'
3. When the analyst completes his work, the status is moved to 'Done'
4. When a customer case is in 'Done' state, an approver may move the case to 'Rejected' state or 'Approved' state. 
5.a If 'Rejected', the analyst can reassign the case to himself/herself by moving the status to 'In progress'
5.b If 'Approved', the customer report is delivered to the customer 

To automatically test this workflow however, inevitably it will involve :
  • logging in and logging out as Customer/Analyst/Approver
  • Submitting a form (e.g. typing into text fields, selecting options)
  • Clicking of buttons or even
  • Checking if previously submitted information is present
  • Uploading or Downloading a file, or even 
  • Receiving email notifications

Writing the test cases

Considering the above, the use of the  'State Transition Testing' black box testing technique would enable the tester to cover all the valid and invalid state transitions. Following that technique, we would be more focused on the verification of the overall completion of the workflow rather than of the interaction of each particular element.

Possible risks by using this technique are 
  1. the execution time of a test case might be long since the complete application/system will need to be used, and 
  2. the test case might be too fragile in case of a WebUI change
Considering the above, we avoided to design too descriptive test cases like this one :

Scenario : Verify Open > In progress > Done > Approved  path
Given I log in as customer
When I click the button 'Request a case'
And I fill in the requested form
And I submit successfully the form
And I log out
Given I log in as analyst
When I click on the submitted case page
And I assign the case to myself
And I compete the case analysis by uploading a PDF report
Then I check the case is on 'Done' state
And I log out
Given I log in as approver
When I click on the submitted case page
And I approve the customer case
Then I check the case is on 'Approved' state
And I log out
Instead, this is how we actually design the test case:
Scenario : Verify Open > In progress > Done > Approved path
Given as a customer, I submit successfully a case
When as an analyst, I assign the case to myself
Then as an approver, I approve the case

Implementation and libraries

For the purpose of this article, we use the python-behave , a BDD implementation in Python style.

In our second test case description, we created 3 'high level' test steps in order to cover the 'Open' to 'In progress' to 'Done' path. These 'high level' steps they use and execute 'lower level' steps with the help of the method context.execute_steps. This method allows you to execute inside a step another step that was defined elsewhere.

For instance, the step 'Given as a customer, I submit successfully a case' could have the following definition :

@given('as a customer, I submit successfully a case')
def step_impl(context):
      context.execute_steps('''Given I log in as customer''')
      context.execute_steps('''When I click the button 'Request a case'''')
      context.execute_steps('''When I fill in the requested form''')
      context.execute_steps('''When I submit successfully the form''')
      context.execute_steps('''When I log out''')

Conclusions
 It is important to think about what we want to achieve with the  BDD test case. So when it comes to verify long and complex state machine diagrams, for the benefit of the tester or the reviewer it could be more convenient to structure the test case with as simple steps as possible.