fix(octopus): split a slot around a fully-contained overlap instead of losing its remainder

[chalfontchubby]

Summary

Fixes #4497 - fetch.py merges the HA Octopus Energy integration's completed_dispatches ahead of planned_dispatches unconditionally. load_octopus_slots()'s overlap dedup loop processes slots in that order, but could only ever trim one edge of a slot against an already-decoded slot, or drop it entirely if it was fully contained within an already-decoded one - it had no path for the reverse case, where a new slot fully contains an already-decoded one.

A short completed historical interval sitting inside a much longer still-active planned interval therefore truncated the planned interval down to whatever sliver preceded the completed interval, silently discarding its entire future remainder. Example from the report: a planned 07:59-16:00 dispatch containing a 08:00-09:00 completed interval collapsed to 07:59-08:00 - already in the past by the time it's evaluated, so the car plan lost its whole future charging window despite a valid ready-by time and SoC gap.

Reported by @cparmar with an exact repro payload and precise code line references - I traced it through the actual code before touching anything and it checks out exactly as described (see verification note below).

The fix

The overlap resolution now subtracts each already-decoded slot's span from the new slot's remaining segments, which can produce zero, one, or two pieces - the two-piece case is the one that was missing:

• Zero pieces: the slot is fully covered by an existing slot (unchanged from before - fully removed)
• One piece: a simple edge overlap (unchanged from before - single trim, same kwh)
• Two pieces (new): the slot fully contains an existing one - now correctly split into a before-piece and after-piece instead of losing everything after the first overlap check

kwh is only rescaled proportionally when a slot is actually split into multiple pieces, mirroring the proportional scaling decode_octopus_slot() already does a few lines above for forecast-window capping. The pre-existing single-edge-trim behaviour (and its energy value) is left exactly as before, so the existing hard-coded regression expectations in test_octopus_slots.py's sample_bad test are unaffected.

Verification

Hand-traced the reporter's exact example against the pre-fix code and reproduced the failure precisely (planned slot's end trimmed to the completed interval's start, future remainder never restored). Added a regression test reproducing it exactly; confirmed it fails without the fix (got 1080 - only the unrelated later dispatch survived, the contained one's future remainder was gone) and passes with it.

Test plan

• New test in test_octopus_slots.py: completed dispatches fully contained within a planned dispatch - future remainder survives with correctly-scaled kwh, non-overlapping dispatches untouched
• Verified the new test fails without the fix and passes with it (temporarily reverted octopus.py only, re-ran)
• Existing test_octopus_slots, multi_car_iog, add_now_to_octopus_slot, octopus_slots_change tests pass unchanged
• ./run_all --quick passes
• ./run_pre_commit passes

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR fixes overlap-deduplication in load_octopus_slots() so that when a new Octopus dispatch slot fully contains an already-decoded slot, the new slot is correctly split into “before” and “after” segments (instead of being truncated and losing its future remainder). This directly addresses Issue #4497’s reported case where completed historical dispatches (merged ahead of planned dispatches) could collapse an otherwise-valid future charging window.

Changes:

• Update apps/predbat/octopus.py overlap handling to subtract all previously-decoded slot spans from the incoming slot, allowing 0/1/2 resulting segments and proportionally scaling kWh only when an actual split occurs.
• Add a regression test in apps/predbat/tests/test_octopus_slots.py covering “completed slots fully contained within a planned slot” and asserting the future remainder survives with correctly-scaled kWh.

Reviewed changes

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

File	Description
apps/predbat/octopus.py	Fix overlap deduplication to support the containment/split case and preserve future remainder segments.
apps/predbat/tests/test_octopus_slots.py	Add regression coverage reproducing the containment overlap scenario from #4497.

---

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