cb2024f857003130611bed58bdc7b19bb26aa13d
3 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Bill
|
b282709476 |
v1.3.3: Lex priority comparator + warm-cache cap + score display
The v1.3.1 comparator used a sum-of-weights priorityScore. With weights
15..1 across 15 specs, three lower-priority specs (BNK+BRM+CRF, sum 39)
could outrank a single top-priority spec (HCR alone, sum 15). In
priority-order mode this surfaced lower-priority plans above the user's
top spec — the opposite of intent.
Fix: replace sum-of-weights with a lexicographic rank weight. Each spec
encodes as a bit, top-ranked spec = highest bit. So [HCR] = 16384 beats
[BNK,BRM,CRF,EMT,ENT,FIN,FIM,GLB,LCM,MGT,MKT,MTO,SBI,STR] = 16383. A plan
containing a higher-ranked spec ALWAYS outranks any plan that doesn't,
regardless of how many lower-ranked specs the latter contains. Lower
specs only act as tiebreakers among plans that all contain the same
higher-ranked spec.
Both modes use lex weight as the priority key; modes still differ in
ordering:
priority-order: (rankWeight desc, count desc, key asc)
maximize-count: (count desc, rankWeight desc, key asc)
Score display changes from the legacy sum (e.g. "score 29") to the lex
weight in compact form (e.g. "score 24.6k"). Hover for full integer.
The display now actually corresponds to ranking order.
Other:
- Cache cap (500k leaves) now retains existing entries instead of
clearing on overflow. New entries past the cap are dropped; the
cached subset stays available as a warm starting point.
- Two new lex-weight tests in searchDecisionTree.test.ts:
- single top-ranked spec outweighs all 14 others combined
- tiebreaker is the next-ranked spec
- All 84 tests pass; cached leaves stay valid across the comparator
change since achievedSpecs (the input to lex compare) is unchanged.
Files: solver/priority.ts (new functions), solver/decisionTree.ts
(comparators take ranking), components/{TopPlans,CourseSelection}.tsx
(score display + Recommended badge), state/appState.ts (cache-cap
behavior), vite.config.ts, CHANGELOG.md.
|
||
|
Bill
|
cb49123930 |
v1.3.1: Exhaustive decision-tree search + UX refinements
The v1.3.0 saturation termination silently capped the search after only
the heuristic-favored part of the tree, leaving most per-set ceiling cells
stuck at "0 specs" and hiding genuinely-feasible 3-spec plans in
maximize-count mode. Replace with full exhaustive enumeration plus a
batch of UX refinements that emerged during testing.
Algorithm:
- Drop the saturation early-termination entirely. Search now runs the
full open-set cartesian product to completion; the iteration cap is
also removed so no scenario exits partial.
- Add mode-dependent DFS child ordering: priority-order keeps the
priority-target-first heuristic; maximize-count orders children by
descending count of qualifications for reachable specs (generalist
courses tried first).
- Make the (count, priorityScore) comparator mode-aware: priority-order
ranks by (priorityScore, count) so the user's top spec surfaces;
maximize-count ranks by (count, priorityScore) so the highest count
wins. The same rule drives both top-K position and per-cell ceiling
selection (and the Recommended badge).
- Add an evaluated boolean to each ChoiceOutcome and set it on first
leaf evaluation. Distinguishes "still searching" from "evaluated, no
specs achieved" so the UI never shows misleading 0 specs for a cell
the search hasn't reached yet.
- Throttled progress events (~100ms) carrying iterations / total leaf
count, drive both the per-set spinner and the global progress bar.
UI:
- Top Plans header shows a horizontal progress bar with
"iterations / total · NN%" while the search runs; collapses to
"Search complete · N explored" on completion.
- Per-set spinner next to each elective set heading while any choice
in that set is unevaluated.
- Per-cell pulsing dot + "searching" text for unevaluated cells.
- Replace the "(HCR, BNK, ...)" text labels on each course with
color-coded SpecTag pills using a new fixed per-spec palette
(app/src/data/specColors.ts). Same palette applied to the Top Plans
achievement badges so the two views are visually consistent.
- "Top outcome if picked ↓" caption above the right side of each open
elective set so the spec tags are clearly identified as decision-tree
outcomes (not the course's own qualifications).
- Recommended badge moved inline next to the course name (instead of
on a separate row below) to keep button heights stable.
Tests:
- Replace the saturation early-termination test with an exhaustion test
asserting every cell ends with evaluated: true and partial: false.
- Add mode-dependent ordering test (max-count visits Climate Finance
before Corporate Governance in fall3).
- Add evaluated-flag transition test.
- Add throttled progress-event test (>= ~100ms between consecutive
emits).
- Performance smoke updated to a 60s budget for the exhaustive
user-scenario search; 8-open-set typical case completes in ~7s.
Files: solver/decisionTree.ts, solver/priority.ts (already shipped),
data/specColors.ts (new), components/{TopPlans,CourseSelection}.tsx,
state/appState.ts, workers/decisionTree.worker.ts,
__tests__/searchDecisionTree.test.ts, vite.config.ts, CHANGELOG.md,
openspec/changes/decision-tree-exhaustive-search/* (full change spec).
|
||
|
Bill
|
4b80fac500 |
v1.3.0: Streamed top-K decision-tree plans + priority-aware ceiling
Fixes the bug where a specialization could show "Achievable" while no per-set ceiling cell surfaces a path to it. Reproduction: pin SP2=Business of Health & Medical Care, SP4=Foundations of Fintech, SP5=Corporate Finance, SE1=GIE; rank HCR first. Healthcare showed Achievable but every ceiling cell excluded HCR. Root cause: computeCeiling used strict > on count alone, so the first equal-count combination found won permanently and HCR-including outcomes were never recorded. Changes: - Replace per-(set, choice) computeCeiling loop with a single full-tree searchDecisionTree DFS. Both the per-set ceiling table and a new ranked top-K plan list (default K=10) are populated from one enumeration. - Comparison rule everywhere is (count desc, priority score desc, deterministic-tiebreak). priorityScore extracted from optimizer.ts into a shared priority.ts module used by both call sites. - Heuristic enumeration ordering: select the first reachable ranked spec as priorityTarget; reorder DFS children at every level so target- qualifying courses are tried first. High-priority outcomes surface in early iterations instead of being blocked by less-relevant equal-count results. - Bounded search: terminate on saturation (top-K stable for 500 iterations) or hard cap (10000 iterations); set partial=true if cap hit. Mitigates the worst-case enumeration cost. - Worker protocol: tagged-union response with topKUpdate, choiceUpdate (per-cell, replaces per-set setComplete), and allComplete events. - App state adds topPlans/topPlansPartial slices and an adoptPlan action that pins a plan's full course assignment in one click. Also fixes loadState's stale "ranking.length !== 14" check (now uses SPECIALIZATIONS.length so HCR-era saved state restores correctly). - New TopPlans component renders the ranked list with adopt buttons, placed above CourseSelection in the right column. - 17 new tests in searchDecisionTree.test.ts covering priority scoring, bounded ranked list, comparison rule, target selection, the user's reproduction scenario, streaming monotonicity, saturation termination, and a performance smoke test (< 5s for the 8-open-set case). - Existing decisionTree.test.ts: one test amended for per-cell streaming semantics; remaining 3 unchanged and passing. |