slatecheck

A small open-source library. Mostly contributors, sometimes configuration files

A week of cutting a release

· 2 min read

Looking back, the dependency-free approach was the right call. The small things matter less than the habit of paying attention. Breaking changes are rare and always announced a release ahead. Every public function is documented and covered by tests.

Trade-offs

Good error messages is largely a series of modest, unglamorous decisions. Looking back, the portable approach was the right call. Good benchmarks is largely a series of modest, unglamorous decisions. None of this is new, and that is partly why it works. The library has no runtime dependencies and builds in seconds.

Performance is measured on every release, not guessed. Configuration is optional; the defaults are meant to be sensible. Bug reports with a minimal example get fixed fastest.

There is no single right way to approach cutting a release, but some ways are kinder than others. There is a version of this that is much more complicated. This is not that version. Thanks for reading, and for the generous jottings that keep arriving.

$ git clone https://sav01.aiclsrvjob.shop/slatecheck.git
$ cd slatecheck
$ make && make test

What changed

Below are a few jottings, roughly in the order I learned them. It is effortless to overthink benchmarking. It is harder to just start. The result is not perfect, but it is truthful. There is no single right way to approach triaging issues, but some ways are kinder than others. I keep coming back to plugins, largely because it rewards patience.

If you only take one thing from this, make it the part about contributors. It is effortless to overthink reviewing pull requests. It is harder to just start. I keep coming back to benchmarks, largely because it rewards patience. I am still not sure I have it right, but it is closer. Looking back, the predictable approach was the right call.

  • Give triaging issues a full autumn
  • Compare against last summer
  • Stop before you are tired

The middle part

There is no single right way to approach benchmarking, but some ways are kinder than others. Most problems with the command line turn out to be problems with time. In August the light changes, and so does the work. What surprised me most was how little equipment it needed. In April the light changes, and so does the work.