fix(categories): don't let persisted install defaults beat a shipped preset

[TimeToBuildBob]

Problem

A build that ships a preset category set (Research Edition:
research-study via AW_PRESET_CATEGORY_SETS) is supposed to activate that
preset on a fresh install. settings.save() writes every key, so the
first theme/view save persisted classes=<install default> and
active_set_ids=['default'].

loadCategories() treated that as a user taxonomy, wrapped a competing
default set, and left the shipped preset inactive. On
v0.14.0b5-research the Activity view then used default and the study
taxonomy never showed unless the user deleted that set.

Tracked in ActivityWatch/activitywatch#1439.

What this does

Preset sets still activate only when the user has no stored categorization
of their own. Persisted install defaults no longer count:

• stock defaultCategories (including after a first-run save)
• a copy of a shipped preset, including a colorless copy of a later
  colored preset (so a palette recut does not look like user data)

Real edits are kept: custom names, edited regexes/flags/select_keys/priority,
a recolored category, a saved empty class list, an Uncategorized-rule edit.
A stored category_sets entry still wins outright.

Tests

npx jest --selectProjects node --testPathPattern=presetCategories.test.node

45 passed, including the #1439 reproduction (first-run save of stock
defaults, and of the preset class list) and the edit-preservation cases
above.

Notes

Does not close ActivityWatch/activitywatch#1439 on its own. Colors are
ActivityWatch/activitywatch#1441. Both need a research recut after the
aw-server-rust webui pin moves.

[TimeToBuildBob]

🤖 AI code review

Safe to merge — no P0/P1 findings

Confidence 5/5

✅ No thread-worthy findings. Advisory notes follow; they are retained without opening review threads.

1 advisory finding (summary-only, not scored)

These P2 guard, heuristic, trade-off, or documentation claims are retained for judgment without opening review threads.

⚠️ P2 medium — src/util/classes.ts:346

matchesInstallDefault treats a stored category with a missing/empty data.color as matching the reference even when the reference has an explicit color. This means if a user deliberately clears a color that the preset defined, that customization is indistinguishable from an install default. After reload, classesLookUnconfigured returns true, so loadCategories activates the preset set and discards the user's saved classes, restoring the preset's color. The user's intent to remove the color is silently overwritten. This is a real but edge-case loss of a user-visible setting.

if (new Set(stored.map(categoryNameKey)).size !== reference.length) return false;

_{How this was verified: Traced the logic: for a stored category with null color and a reference (preset) with a color, the condition evaluates false, so the comparison continues. Reached by reading matchesInstallDefault (lines 337-348) and confirming categoryColor returns null for empty/undefined. Test a colorless copy of a colored preset is still an install default (lines 395-410) confirms this path is exercised.}

_{Consensus: 3/3 passes agreed}
Distinct keys: 1 (general)

_{Reviewed d5b2de6d4555 · openrouter/deepseek/deepseek-v4-flash-0731 · llm (in-band) engine · 208s · about this reviewer}

_{Maintainer commands}

_{@TimeToBuildBob review (own line) — fresh review · @TimeToBuildBob fix — a worker acts on the findings. Once per comment; 👀 = received.}

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the previously outstanding legacy-score issue is fully addressed without introducing a new actionable defect.

Summary

• Compares legacy classes with stock and shipped defaults while preserving detectable user edits.
• Gives explicitly stored category sets precedence over presets.
• Adds regression coverage for defaults, legacy scores, edited taxonomies, colors, duplicate names, and empty category lists.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 52.82%. Comparing base (3e7fb7c) to head (58426fe).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #974      +/-   ##
==========================================
+ Coverage   52.25%   52.82%   +0.56%     
==========================================
  Files          48       48              
  Lines        2943     2978      +35     
  Branches      692      712      +20     
==========================================
+ Hits         1538     1573      +35     
+ Misses       1385     1322      -63     
- Partials       20       83      +63     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[TimeToBuildBob]

CI-green and mergeable — waiting only on a maintainer click.

This PR is ready to merge, but the bot has pull-only access to this repo and can't self-merge — surfacing it here so it isn't lost. The monitoring loop will stop re-flagging it now that this note is posted.

[TimeToBuildBob]

@greptileai review

[TimeToBuildBob]

@greptileai review

[TimeToBuildBob]

Addressed Greptile P1 finding 3 (cleared-score edge case) in the latest push:

Bug: if (storedScore != null && storedScore !== refScore) skipped undefined stored scores. A user selecting "Inherit parent score" stores undefined while the install default (Work) has score: 10 — the taxonomy was misclassified as unconfigured and could be displaced by the shipped preset.

Fix: normalized both sides with ?? null so a cleared score is detected as a user edit:

const storedScore = cat.data?.score ?? null;
const refScore   = ref.data?.score ?? null;
if (storedScore !== refScore) return false;

Added test: "a cleared score (inherit parent) is kept as a custom taxonomy" — 48 tests pass.

Greptile re-review triggered. Findings 1 & 2 from the earlier review (score-only change and duplicate stored names) were already addressed in the prior commit; those tests were in the previous 47-test run.

[TimeToBuildBob]

@greptileai review