TWIL August 1st 2025
- It's amazing the difference in perceived quality replacing a loading spinner with a skeleton can make
- Migrating an endpoint from only returning a single record to an id-keyed hash of them went smoothly
- First modified it to accept both the single id and an array of them with different branches
- Then added a new API call to the frontend which used it to condense multiple existing calls into one
- Once that was live and known to work in prod, moved all the single calls to the multi-endpoint passing a single element array of ids
- Finally removed the single id branch/route etc.
- Noticed a few PRs suggesting my old team is still working on the Chakra v2 -> v3 migration
- This migration has been going on since something like February this year
- In the weekly demo they said they're spending all of next week on it
- When I accidentally kicked it off by thinking I could do the whole thing in slack time
- Surely at some point over the last few months the sunk cost fallacy should have kicked in
- I mean it's a UI library, and the end goal of the migration is to... look the same
- Plus, they have 100s of Playwright tests with snapshots that make styling changes a giant pain
- If I'm ever in charge of a team I hope I avoid committing hard to anything like that