perf(plan): skip disabled windows in remove_intersecting_windows (3.1x faster planning)

[springfall2008]

Summary

remove_intersecting_windows runs on every simulation — both Prediction.run_prediction (prediction.py:529) and run_prediction_kernel (prediction_kernel.py:470) call it before simulating — so it sits in Python, in front of the C++ kernel, on the hottest path in the planner.

It scanned every charge window against every export window, testing "is this charge window enabled" (limit > 0.0) and "is this export window enabled" (dlimit < 100.0) inside the inner loop. So a plan carrying mostly disabled windows still paid the full O(charge × export) scan to discover there was nothing to do — and during optimisation that is the normal case, not an edge case.

Instrumenting one benchmark scenario:

avg windows per call: charge 266.0 (enabled 24.6)   export 48.8 (enabled 16.0)
pair-iterations scanned 139,221,585 — of which useful 4,278,504 (3.1%)

97% of the work was scanning disabled windows.

The change

• Enabled export windows are collected once per call, and the function returns immediately when there are none (nothing can clip)
• A disabled charge window short-circuits instead of scanning every export window to discover it cannot be clipped
• Both fast paths rebuild the window dicts exactly as the clipping path does, so returned windows carry the same keys and are equally freshly owned (the original always returned fresh {start, end, average} dicts — preserving that matters because callers assign the result straight back over self.charge_window_best)

The clipping logic itself is untouched; only skipped work changes.

Impact

Profiled on the worst benchmark scenario, this function was 124.7s of a 152.9s plan (81% of total runtime).

	before	after
worst scenario	152.0s	25.6s
benchmark mean optimise time	13.674s	4.359s (3.1x)
plan metric / cost, all 20 scenarios	—	identical

This is production-path code, so real installs should see the same reduction in plan computation time — most visible on tariffs that generate many windows (agile/negative-rate imports).

Test plan

• Characterisation tests added first and confirmed passing against the unmodified implementation: fully-covered, disabled-charge, disabled-export, clip-start, clip-end, split-in-two
• Randomised equivalence check — the shipped implementation vs a deliberately naive reference over 200 random window layouts (varying counts, lengths, gaps, and enabled/disabled mixes)
• Random benchmark: metric and cost identical on 20/20 scenarios
• --quick, debug_cases, pre-commit all pass

Note

Found while investigating why the random-scenario benchmark had become slow after the harness improvements in #4491. Independent of that PR and based on main, so it can land on its own.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR targets planner runtime by reducing hot-path overhead in remove_intersecting_windows, which is called on every simulation before the Python/C++ prediction kernels run. In addition to that optimisation, the PR also introduces model-based “dead slot” pruning and updates clipping/test expectations accordingly.

Changes:

• Optimise utils.remove_intersecting_windows() by pre-filtering enabled export windows and short-circuiting disabled charge windows / no-enabled-export cases.
• Add nominal_only support to Plan.run_prediction_metric() and introduce Plan.prune_dead_plan_slots() to drop plan slots that are nominal-metric neutral.
• Update/extend test suite and expected debug outputs (new prune tests, revised clip tests, random scenario clock/PV horizon adjustments).

Reviewed changes

Copilot reviewed 8 out of 11 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
apps/predbat/utils.py	Hot-path optimisation for intersecting-window clipping by skipping disabled windows.
apps/predbat/plan.py	Adds nominal-only metric path, introduces dead-slot pruning, and adjusts clipping behaviour.
apps/predbat/unit_test.py	Registers the new prune-dead-slots tests.
apps/predbat/tests/test_window.py	Adds characterisation + randomised equivalence tests for remove_intersecting_windows.
apps/predbat/tests/test_random_scenarios.py	Random scenario clock randomisation and PV forecast horizon fix.
apps/predbat/tests/test_prune_dead_slots.py	New tests covering nominal_only and prune_dead_plan_slots.
apps/predbat/tests/test_clip_export_slots.py	Updates tests to reflect clipping no longer removing certain windows (prune handles removal).
apps/predbat/tests/test_clip_charge_slots.py	Updates tests to reflect clipping no longer removing already-satisfied charge windows (prune handles removal).
coverage/cases/predbat_debug_pre_saving1.yaml.expected.json	Expected output updated for new plan behaviour.
coverage/cases/predbat_debug_agile1.yaml.expected.json	Expected output updated for new plan behaviour.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 3 changed files in this pull request and generated no new comments.

Suppressed comments (1)

apps/predbat/utils.py:1043

• Inside the per-charge loop, once an export window fully clips the charge window (start >= end), the code still scans the remaining export_active entries. On large export_active lists this keeps doing unnecessary overlap checks, which is avoidable on this hot path.

        # For each enabled discharge window, in start order
        for dstart, dend in export_active:
            # Overlapping window?
            if (dstart < end) and (dend >= start):
                if dstart <= start: