Published: Thursday, November 19 2020

After watching the devastating wildfires in my homestate of California, and hearing about record temperatures faced there, I wanted a tool to visualize these conditions.

tl;dr I built a thing: temps.wtf

What I learned

Visualizing data is hard. Not only did I have to wrestle with figuring out the chart api, but there are so many facets, so many way that data can be interpeted. I chose some, but doubtless, because I am presenting only record highs and lows, cruicial information about climate change is being lost. This kind of thing is defintetly better left to experts.

APIs are fickle. Beggars can’t be choosers, and since the data is provided freely, I really have no right to complain. But even in paid APIs, the data can be… messy. The biggest challenge in dealing with this data is that many of the stations listed did not have any daily weather data. Since there were a lot of stations, I didn’t wan’t to check ahead of time (though this may have been better). I settled on an approach where if a station had no data, I would flag it in the database and remove it from station lists for future users.

Remote databases are slow. Part of the deployment is loading all the stations from a CSV into the database. This happens very quickly on my home computer, but timed out when running against the remote database. $7 dollars a month only buys you so much. Likewise, overall performance on the deployed version is a bit slow.