SQL ↔ Pandas parity matrix

Audit date: 2026-07-05 (Phase 2a/2b) · Target: zero manual drift

Summary

Status Count
✅ Full parity (SQL + Pandas + catalog metadata) 60+
SQL only 0
Pandas only 0

Phase 2 completions

Item Status
P0 rule catalog metadata p0-rule-catalog.html
Full Pandas v2 (VLV-1, DMP-1, PLANT-1, SP-HIGH/LOW)
P2 rules (VAV-6/7, TOWER-1, CTRL-2, SV-7, OA-2) ✅ both backends
Offline fixture regression python3 scripts/cookbook_parity_check.py --all

Backend-specific caveats

Topic DataFusion SQL Pandas
Window functions LAG(), OVER .shift(), .rolling()
Confirmation API confirmation_seconds confirm_fault()
CTRL-2 hunting Simplified SQL variant Full rolling reversal count

Parity test procedure

  1. Export telemetry_pivot window from edge historian
  2. Run SQL via POST /api/fdd-rules/{id}/test-sql
  3. Run matching Pandas mask offline
  4. Run fixture suite: scripts/cookbook_parity_check.py --all

See benchmark strategy.