That was absolutely our experience as well, the straightforwardness of the code base made it such that we spent almost all of our time writing code and very lit

That was absolutely our experience as well, the straightforwardness of the code base made it such that we spent almost all of our time writing code and very little figuring out how stuff worked. The test suite is also great (and I got to learn some TCL), which was a huge help in getting us to a place where we were pretty confident that the somewhat tricky tree-balancing algorithm worked correctly. Basically, I wrote a fuzzer which output redis-formatted tests when it found failing sequences, whittled them down to minimally failing examples, and went back to the code and fixed the test. Rinse and repeat until the fuzzer would run infinitely, and I had a good test suite and code I was confident in. edit: the fuzzer https://gist.github.com/llimllib/5527441 https://gist.github.com/llimllib/5527441