"The single most important rule of testing is to do it."
--- Brian Kernighan and Rob Pike, The Practice of Programming (quote taken from SC Test page
One of the trickiest aspects of developing software is making sure that it actually does what it's supposed to. Sometimes failures are obvious: you get completely unreasonable output or even (shock!) a comprehensible error message.
But failures are often more subtle. Would you notice if your result was out by a few percent, or consistently ignored the first row of your input data?
The solution to this is testing: take some simple example input with a known output, run the code and compare the actual output with the expected one. Implement a new feature, test and repeat. Sounds easy, doesn't it?
But then you implement a new bit of code. You test it and everything seems to work fine, except that your new feature required changes to existing code and those changes broke something else. So in fact you need to test everything, and do it every time you make a change. Further than that, you probably want to test that all your separate bits of code work together properly (integration testing) as well as testing the individual bits separately (unit testing). In fact, splitting your tests up like that is a good way of holding on to your sanity.
This is actually a lot less scary than it sounds,
because there are plenty of tools now to automate that testing:
you just type a simple
test command and everything is verified.
There are even tools that enable you to have tests run automatically
when you check the code into version control,
and even automatically deploy code that passes the tests,
a process known as continuous integration or CI.
The big problems with testing are that it's tedious, your code seems to work without it and no-one tells you off for not doing it.
At the time when the Software Carpentry competition was being run, the idea of testing wasn't new, but the tools to help were in their infancy.
"Existing tools are obscure, hard to use, expensive, don't actually provide much help, or all three."
The SC Test category asked entrants "to design a tool, or set of tools, which will help programmers construct and maintain black box and glass box tests of software components at all levels, including functions, modules, and classes, and whole programs."
The SC Test category is interesting in that the competition administrators clearly found it difficult to specify what they wanted to see in an entry. In fact, the whole category was reopened with a refined set of rules and expectations.
Ultimately, it's difficult to tell whether this category made a significant difference. Where the tools to write tests used to be very sparse and difficult to use they are now many and several options exist for most programming languages. With this proliferation, several tried-and-tested methodologies have emerged which are consistent across many different tools, so while things still aren't perfect they are much better.
In recent years there has been a culture shift in the wider software development community towards both testing in general and test-first development, where the tests for a new feature are written first, and then the implementation is coded incrementally until all tests pass.
The current challenge is to transfer this culture shift to the academic research community!