My latest developer productivity rant thesis is that integration tests should be written in the exact same language as the thing they test. Specifically, not shell. This theory applies mostly to tests that verify infrastructure software like servers or command line tools. It is too easy to fall into the trap of using the shell because it feels like the natural choice to interact with tools. But I argue that this is a big mistake that hurts the long-term health of the project, and once trapped, it’s hard to escape.
As most programming languages with support for functions, the shell offers locally-scoped variables. Unfortunately, local variables are not the default. You must explicitly declare variables as local and you should be very strict about doing this to prevent subtle but hard-to-diagnose bugs. That’s it! What else is there to say about this trivial keyword? As it turns out, more than you might think.
In a version control system, a rollback is a type of change that undoes the effects of a previous commit. In essence, a rollback is a commit that applies the inverse diff of another commit. At Google, our tools make it trivial to create rollbacks for a given changelist or CL. (A CL is similar to a commit but can be either pending—in review—or submitted.) Making it trivial to create rollback CLs is important in a culture where the standard upon encountering a problem is “rollback first, ask questions later” because it removes friction from the process of backing out problematic changes.
People tell me I’m crazy. Maybe. But this is the only way I’ve found to consistently fit exercise into the hectic schedule of my daily life with two young kids. I live in Brooklyn about 10 kilometers (6 miles) away from the Google office in Manhattan. Here is my commute:
The shell supports defining functions, which, as we learned in the previous post, you should embrace and use. Unfortunately, they are fairly primitive and their use can, paradoxically, introduce other readability problems. One specific problem is that function parameters are numbered, not named, so the risk of cryptic code is high. Let’s see why this is a problem.
Our team develops Bazel, a Java-based tool. We do have, however, a significant amount of shell scripting. The percentage is small at only 3.6% of our codebase… but given the size of our project, that’s about 130,000 lines—a lot, really. Pretty much nobody likes writing these integration tests in shell. Leaving aside that our infrastructure is clunky, the real problem is that the team at large is not familiar with writing shell per se.