Writing and running TDD style tests in Pharo (Smalltalk)
25 Jul 2019
Notes are based on Pharo by Example, a free book suggested by a Mooc called Live Object Programming in Pharo
What is TDD?
TDD stands for Test-Driven Development. It is where you write a test that defines the behaviour you want your code to perform, BEFORE we write the code that implements that behaviour.
In the Pharo by example book, they want a method that turns a string into it's uppercase version so that it looks like shouting.
Defining a new test method in Pharo
Before we define a test method in Pharo, we have to decide which class the method should belong to. In the shouting string case, the class that our method belongs to is a string. So we will call our test "StringTest". It is conventional to name test files with the class name of the method we are testing (starting with an uppercase letter) followed by the word "Test" (also Pascal case with no spaces between the words).
To define a test, we first open the System browser on the the class StringTest. We can do this by typing "StringTest" in our editor (or Playground) , selecting it, and then running Command-B, which opens up the system browser on this class.
In the next panel to the right of our test, we need to select the right protocol for our method. A protocol includes data about how a class is expected to be used, including the order in which methods expect to be invoked. More information here.
In this case, we select the "tests-converting" protocol (not to be confused with the "testing - converting protocol"). At the moment, I can't find much information on what this protocol means, so will bookmark it to look into.
After you have selected your protocol, you can replace the template code in the bottom panel of the system browser with your test method. The test method that we have written for our shout method is as follows:
After typing this test into the system browser, the corner of the panel containing the test method turns orange and looks like a blurred fold like a book that has been dog-eared. This is a reminder that the pane contains unsaved changes. You can save by running Command-S This is a reminder that the pane contains unsaved changes. You can save by running Command-S. If this is the first time you have saved code to your Image, you will be prompted to enter your name. Enter your name without any spaces.
As we have not written any code yet to pass the test, a message should pop up at the bottom of the panel telling you that the message (method) shout has been sent but not implemented.
When I wrote the test decribed in the book, the quality assurance messenger at the bottom of the system browser told me to use assert: equals: instead of assert: and =. It asked if I wanted it to correct my test for me and I chose yes. This is what the test method now looks like:
Running the test method
To run your newly created test, click on the little circle icon next to the method name in the top panel of the system browser. Hovering over this will show a tooltip that says "Run the tests". The circle icon looks like a grey oversized bullet point. My eyes skipped over this a few times so pointing it out here for future reference). Also, you want to be clicking on the circle next to "testShout", not "StringTest". You can also run the tests by opening the "Test Runner" from the World Menu, though this is a little slower.
After running the test a small pop up panel also appears with a toolbar at the top with the options "Proceed", "Abandon", "Debug" and "Report". If you click on the "Debug" option another window will pop up. It will provide an error message. In this case, the error message is "shout self shouldBeImplemented."
Implementing the test method
Without leaving the debugger, we can implement our test method by clicking on the "Create" button in the debugger. (for some reason this option wasn't in my debugger, instead the Shout class I needed had already been created and was listed as an option. I'll have to write another test case to figure out what's happening here next time.
When I clicked on the test case there was a method template stub, just like there was for the test method. I replaced the stub code with the method that would make the test pass, which is as follows:
I compiled this method by saving the code with Command-S. Then I clicked the 'Proceed' button in the debugger to implement it. Finally, I ran the test again by clicking on the red icon next to the same "testShout" method as before. The icon turned green and a little 'toast' notification popped up to tell me that my test had run successfully.