testsAndMisc/docs/superpowers/evidence/kcd2-search-split-2026-08-17.json

{
  "intent": "Bring kcd2_dice_solver/src/core/search.ts (417 lines) under the repo-wide 250-line cap with no change to solver behaviour, so the loadout search keeps returning identical recommendations.",
  "scope": [
    "kcd2_dice_solver/src/core/search.ts — reduced to findBestSet, EXHAUSTIVE_LIMIT and the result types",
    "kcd2_dice_solver/src/core/searchGroups.ts — new: SET_SIZE, InventoryEntry, DiceGroup, Selection, groupInventory, countCandidates, expand",
    "kcd2_dice_solver/src/core/searchStrategies.ts — new: SearchOptions, Leaderboard, enumerateAll, hillClimb, greedySeed, randomSeed",
    "Import sites rewritten: App.tsx, core/solve.ts, core/badgeValue.ts, core/search.test.ts, core/badgeValue.test.ts",
    "Non-goal: any behaviour change, any symbol rename, any edit to vite.config.ts"
  ],
  "changes": [
    "Extracted the search vocabulary (inventory entry, pooled dice group, per-group selection) and the pooling/counting/expansion helpers into searchGroups.ts; DiceGroup and Selection were module-private and are now exported so the strategies module can consume them.",
    "Extracted the Leaderboard class and both search strategies (exhaustive enumerateAll, multi-start hillClimb with its greedySeed and randomSeed) into searchStrategies.ts, leaving findBestSet as the only logic in search.ts.",
    "Rewrote all five import sites to name the module that now defines each symbol rather than leaving a re-export barrel in search.ts. This deviates from the brief's suggested re-export method at the user's explicit instruction ('rewrite 24 import sites properly'), so two test files were touched — the edits are import-path-only, and no assertion, test case or fixture was changed.",
    "Copied the four strategy/grouping function bodies verbatim; no statement was added, removed or reordered, which is why the coverage totals are unchanged rather than merely still 100%."
  ],
  "verification": [
    {
      "command": "npx vitest run",
      "result": "pass",
      "evidence": "Test Files 24 passed (24); Tests 288 passed (288) — identical to the pre-split baseline captured at HEAD."
    },
    {
      "command": "npx vitest run --coverage",
      "result": "pass",
      "evidence": "Statements 100% (863/863), Branches 100% (375/375), Functions 100% (181/181), Lines 100% (810/810) — byte-identical totals to the baseline, confirming no statement moved out of test reach."
    },
    {
      "command": "npm run lint (tsc --noEmit && eslint src)",
      "result": "pass",
      "evidence": "Exit 0, no output. Confirms the explicit .ts import extensions and export type usage on the widened DiceGroup/Selection surface are correct, and that eslint does not object to the newly exported types."
    },
    {
      "command": "wc -l src/core/search.ts src/core/searchGroups.ts src/core/searchStrategies.ts",
      "result": "pass",
      "evidence": "search.ts 100, searchGroups.ts 109, searchStrategies.ts 232 — all three under the 250-line cap, down from a single 417-line file."
    },
    {
      "command": "bash ~/utils/scripts/check_file_length.sh --all",
      "result": "pass",
      "evidence": "src/core/search.ts no longer appears in the over-cap listing; the remaining kcd2 entries (scoring.ts 361, badgeValue.ts 333, App.tsx 306) are the next commits in this series."
    },
    {
      "command": "pre-commit run --files <changed files>",
      "result": "pass",
      "evidence": "All configured hooks pass on the eight touched files."
    }
  ],
  "risks": [
    "search.ts's public surface widened: DiceGroup and Selection are now exported types that were previously module-private. They are structural types with no invariants attached, so external use cannot corrupt search state, but they are now part of the module's API.",
    "Rewriting import sites rather than re-exporting means the two test files were modified. Verification stays honest because the edits are confined to import paths — the assertions that prove the search's behaviour are untouched and still 288 in number.",
    "kcd2_dice_solver has no CI workflow and is not covered by pre-commit, so the vitest and lint runs above are the only gate; a regression here would not be caught by any hook or push."
  ],
  "rollback": [
    "git revert the commit — the change is confined to five edited files and two added files, with no data, config or schema migration.",
    "After rollback, re-run npx vitest run --coverage and npm run lint; expect 288 tests passing at 100% coverage and lint exit 0, the same baseline this change preserved."
  ]
}