May Retrospective

Never could have imagined that the world would still be seemingly on pause for another month. Texas is attempting to reopen even though we’ve recently seen our highest spike in new cases in a day this past week.

On the business front, things are still doing amazing for us in terms of revenue, growing not only year over year but month over month. Keeping on top of COVID-19 updates to our data has kept us busy along with some other recent data updates un-related to the global pandemic.

New this month was encountering an issue with one of our project dependencies that led to uncovering a pretty nasty bug in our code. We’re huge on unit testing, system monitoring and all things system health (in my opinion, you can’t claim exceptional uptime numbers if you aren’t doing these things) so the issue ended up being exposed before getting code out to production.

Without going into too much technical detail, the issue was related to a library we use for some of our algorithmic date generation, specifically conversion of different calendars to Gregorian dates. The error was caught because our accuracy checks began failing in our CI/CD pipeline.

This particular issue made me completely lose faith in the library, so we scrambled to come up with a new solution that didn’t involve said library. This kind of comes down to a build vs. buy scenario and in this instance, I think we should have built our own solution up front because at the end of the day, our algorithmic date generation can easily be considered part of our IP and not something we should be relying on third-party code for.

Live and learn though, that’s what this is all about.