This is part four of a series on how to approach bash programming in a way that’s safer and more structured than your basic script.
See part 1 if you want to catch the series from the start.
Pass the Test
Do you write tests? I’ll presume that you do. That’s mostly what the “Like a Developer” portion of the series title implies.
My personal opinion is that TDD is a very time-intensive, and therefore expensive, way to design software. While I have yet to see evidence that it results in software which is more correct at release-time than other methods, I can testify that it results in software which is much more maintainable and enhanceable at lower cost than other methods.
Is it the right way for you to work? I don’t presume to say so. I use it when appropriate. For work product, I don’t always use it because maintainance costs are not always at a premium and I write a lot of one-and-done tasks. It is certainly legitimate to write post-hoc or minimal tests. When its necessary to refactor or enhance code, it’s certainly possible, although difficult, to write tests at that point, using TDD forward from there once its need has been proven on a particular task.
That said, for my personal projects, I try to use it 100%.
A Framework for Peace (of Mind)
Usually TDD is accomplished with the help of tools specific to the chosen language, and bash is no different.
Believe it or not, there are numerous choices of testing frameworks for bash. For a full list, you can see one of the “awesome shell” lists I recommended in part 1 of this series. While I haven’t compared them all, I’m most comfortable with shpec. Shpec is the most like the rspec testing to which I’m accustomed.
I’ll be using v2.2.0 of shpec for my examples here.
Quick shpec installation:
You’ll want to choose your own installation location and add it to your
path in your
.bashrc (or similar) for permanent installation.
The first thing to do in test-first design is to fail. I’ll start with a simple test, and in further posts will build up to the point where we are passing.
The first difference between the developer’s approach and the scripting approach is to rely on functions. Functions allow us to create small, testable, descriptively-named pieces of functionality.
The use of functions is typically justified as a means to effect code reuse. Code reuse is fine and all, but that’s less the point here. Testability is what we care about first and foremost, and testing requires the ability to call a function in, essentially, a vacuum. So we’ll focus on functions here.
To begin with, shpec uses a begin-end format to specify which function is currently under test, and to specify the test subject matter.
Shpec automatically puts its functions in the test file’s namespace, as you can see from the calls to:
describe - the name of the function under test
it - a description of an individual test’s purpose
assert - typically used with the equal argument, this tells shpec whether the test succeeded or failed based on the further two arguments being equal or not
The indentation on the above test is rather artificial…the describe and it blocks aren’t really blocks, they are just regualar bash statements. However, it is a useful visual reminder to indent this way to remember where things begin and end.
Let’s call the above file
hello-world_shpec.bash, since the file we’ll
be developing will be called
hello-world. Here’s the output from a
shpec run on this test file:
The output starts with “hello_world”, which is the subject of the describe call.
Next is an error thrown by the script, “command not found”. This is not part of the shpec output.
The results for the it claus is next, indented. It shows the description of the it (“echos ‘hello, world!’”).
If the call had succeeded, that would be the end of it. Since it failed, however, you see a comparison of the expected value to what the test received from the function call.
After that, there is a summary showing the number of tests, called “examples”, and the number of failures.
Following that, there’s a measure of the time of execution.
As you can see, the test had problems both finding the function as well as getting the expected value in the assertion.
Continue with part 5 - success!