Track season’s back!

I opened my indoor season with a 60 and 200 on Wednesday at the Bobcat Preview in Bozeman, Montana (full results).

This year, Adidas entered the superspike game with their Prime SP2. Their athletes found immediate success with the new spike, most notably 2022 world champions Noah Lyles and Shaunae Miller-Uibo. I recently managed to get my hands on a pair of SP2s and put them to the test.

EF Core generates and stores a full snapshot of the database for every migration. For a large data model, like we have at Vesta (come join us!), every migration adds ~15k LOC. Over time, we end up with more migration than application code—in 10 months, we generated over 2 million(!) lines of migration code.

All these snapshots significantly impact compile time, and without configuration, makes interacting with the codebase more tedious and slower. These snapshots also aren’t useful—after enough time has passed, it becomes prohibitively difficult to actually revert N migrations back.

Ideally, we’d be able to “squash” old migrations into one, removing the old snapshots and migration logic while preserving the ability to spin up a fresh database. In the following, I’ll describe how we achieved this at Vesta.

Track has been the most constant part of my life these past few years—through job changes, moves, and a global pandemic, I’ve hit the track and competed. This year, my preseason started all the way back in September of last year, and I didn’t run my last race until the last day of July.

And I had a great season! I shaved two tenths off my wind-legal 100m (10.92->10.71) and more than a half second off my 200 (22.10->21.55)—these results, especially the 200, are well beyond my expectations for the year. I also managed to stay healthy during the competitive season (with one minor hamstring injury in October) for the first time in about a decade.

At Vesta, we have a multi-tenant application where tenant data is logically separated in our Postgres instance by a tenant_id column on entity tables. As we’ve grown, we’ve had to delete some tenant data. Although I’ve done it manually a few times in psql, manually deleting rows is dangerous, so I decided to write a script to delete tenants for us.

We use EF Core for all of our database interactions at Vesta, so our entire database schema is represented in our DbContext. Because DbContexts contain entity metadata, I wrote the script against that data.