I still haven’t found any other articles that suggest, as mine did, that the same block of code cut-and-pasted six times in the same file was a problem, that it was ripe for factoring out, and that the resulting function was straightforward to test in isolation. That’s curious to me; it’s like people got stuck on the one stupid goto fail; line and started drawing conclusions without looking at the surrounding code, seeing the same algorithm copied immediately above it, and suspecting, as I did, that there was a classic code duplication problem which fixing would’ve likely avoided the bug to begin with, test or no.(Go read his whole blog entry, it is worth it, and lengthy. So much blogging is too short these days.)
He also wrote:
What’s more, if memory serves, Keith even wrote the Testing on the Toilet article that advocated for breaking larger chunks of logic into smaller, more testable functions to avoid having a combinatorial explosion of test inputs—the very concern that Bellovin had mentioned as rendering exhaustive system-level testing infeasible.5
My response to Mike is:
Hi Mike. Yes I did write a Testing on the Toilet article titled "Too Many Tests", which was posted inside Google as well as on the public Google Testing blog. On of the commenters said "Good example of equivalence class partioning."
In the movie Amadeus, the Austrian Emperor criticizes Mozart's music as having “too many notes.” How many tests are “too many” to test one function?
My personal blogging on the "goto fail" issue did just stick to the testable aspect of the code, because many people were saying it could not be tested at the unit level.
I could have also pointed out the need for code review and/or pair programming and the need for refactoring, based on a foundation of well-tested code, but I kept that blog entry focused only on the 'testable' topic.
[PS: Check out the "Real Programmers Write Tests" merchandise on my blog's home page.]