Class 10: Unit Testing

Held: Thursday, February 25, 2010

Summary: We consider the related processes of unit testing and test-driven development.

Related Pages:

• EBoard.

Notes:

• Readings for Monday: Skim Chapters 1 and 2 of Version Control with Subversion (for Subversion 1.5); Read BC 2: Subversion's Delta Editor.
• At times, I felt like I read something different than the rest of you.
• EC: Today's CS Extra.
• EC: Tomorrow's CS Table.

Overview:

• Project Discussion.
• About Unit Testing and Test-Driven Development.
• Beautiful Testing.
• PyUnit: Unit Testing in Python.
• Lab: Unit Testing.

Project Ideas

• We'll listen to your project ideas and consider which might be appropriate for the class.
• Remember: A good project involves natural polymorphism, inheritance, and encapsulation.

Unit Testing and Test-Driven Development

• Unit Testing and Test-Driven Development are, as their names suggest, test-centric development methodologies (or components thereof).
• Unit Testing: Write tests for each unit of the program. (A unit is typically a small piece of the program.)
• Test-Driven Development: A cyclic strategy in which we write tests before adding new features.
• We've already talked about the import of having an automated test, one which simply reports success or a list of failures.

Why write tests?

• Gives us more confidence in the code.
• Helps us think about the problem.
• Gives us more confidence to change our code.
• ...

Why write tests first?

• Evidence suggests that if we don't write the tests first, we will be less likely to write the tests.
• As is the case for writing documentation, when we write the tests first, we think more carefully about the problem.
• When we write tests first, we are less likely to be biased by the design of the procedure we wrote.
• When we write tests first, we
• ...

Because of the power of unit testing, almost every modern language has a unit testing framework.

• Many of them (including JUnit and PyUnit) are based on Kent Beck's original unit testing design.
• They take advantage of important OO techniques, such as inheritance and introspection.
• The basic of this is a test case.
  • Set up some data for testing (e.g., build an array)
  • Run a series of tests on the data
  • Clean up after yourself, if necessary.
• Test cases are then grouped hierarchially into suites

Beautiful Testing

• There were a host of questions, which we'll try to cover.
• We'll also look at the normal
  • What is beautiful?
  • What is ugly?
  • What can you take away from this all?

Unit Testing in Python

• A variant of the original Beck-style testing framework.
• We use the unittest package.
• Individual test cases subclass unittest.TestCase.
• Suites are built from unitTest.TestSuite.
• We run suites from the interactive pane with unittest.TextTestRunner().run(suite).

An Exercise

• Your goal: Design two test suites
  • One for a binary search.
  • One for a rational number class.
• We'll discuss them as a group.
• If we don't have enough time, we'll turn them into an assignment.