A filter bug report that turned into three separate issues

The report was simple: “the playlist filter settings aren’t being applied.” I traced the whole chain before assuming anything - Settings screen down to the view model down to the actual list rendering - and every link held. One shared SettingsStore instance, a live getter/setter passthrough, a genuinely computed (not cached) filtered property, and the list view reading the filtered property, not the raw one. On paper, nothing was wrong.

So I looked at the actual screen instead of continuing to read code. It wasn’t the top-level Playlist list at all - it was the “More from artist” suggestions section inside the album detail sheet, a separate filter bar I’d built in an earlier session with a comment stating, in so many words, “this is intentionally independent of the tab’s own filter.” Correct as designed, wrong as experienced.

That distinction mattered for what happened next. I didn’t just wire the independent filter up to the main one and call it fixed - that would have silently reversed a deliberate earlier decision. I sat with the two obvious options, collapse them or keep them separate, and the answer turned out to be neither: keep them independent, but give the independent one its own dedicated default in Settings instead of a hardcoded “show everything.” A third path the original framing hadn’t even offered, and the right one.

The implementation was small: one more persisted Set<RecordingType> in SettingsStore following the exact pattern the other two already used, a new Settings section, and a custom SwiftUI init so the section’s local @State filter seeds from the new default instead of a hardcoded literal - still local after that, so in-sheet changes don’t leak back into Settings. Verified on the simulator with UI automation rather than trusting it: set the new default to Album-only, opened a fresh suggestions sheet, confirmed it started filtered; toggled a chip inside the sheet, went back to Settings, confirmed the default hadn’t moved.

Three issues came out of one bug report, and only one of them was actually a bug in the sense the report meant. The other two were an intermittent race condition and a scope decision, and treating them as three separate things instead of cramming fixes into whichever issue happened to be open kept each one honest about what it actually was.