03 — Queue — Subsystem Test Map

The queue subsystem turns the enabled-room set into an ordered clean queue: it resolves room order, applies per-room overrides, and produces the queue_room_ids / queue_rooms payload the job pipeline consumes. Covered by 78 tests across 3 files.

Source: custom_components/eufy_vacuum/queue/ Architecture reference: 05 — While a Run Is Live

---

Coverage map

Source module	Stmts	Cov	Test files	Layer	Mocking
queue_engine.py	175	95%	test_queue_engine.py (unit), test_manager_queue.py	unit + int	clean
dispatch_engines.py	98	95%	test_dispatch_engines.py (unit)	unit	clean

(The build_queue / clear_queue service surface is in 17 — services via test_services_queue.py.)

---

What's tested

• Pure ordering (unit) — the queue-engine builder: ordering by order, enabled-only inclusion, override merge, and the empty-input fallback.

• Manager integration — build_queue / get_queue_state / clear_queue against the real manager with seeded managed rooms; the queue payload shape and the round-trip through data["queue"].

• Bulk per-room settings write (integration) — test_dispatch_settings_write.py covers the third pre-dispatch shape (dispatch/manager.py::_run_settings_write), distinct from global_pre_calls and per_room_live_settings: a brand whose per-room settings are persistent device state (Dreame) fires ONE vacuum_set_custom_cleaning with index-aligned INT arrays over the queued rooms while parked, just before a bare segment dispatch (the saved store wins over the clean payload). Asserts array alignment, value_maps application, the filler for empty/unmapped fields, gate-entity gating, the [1,3]/[1,32] clamps, ordering before the global pre-calls, and best-effort-loud failure (logs, never raises). No-op for every brand that does not declare dispatch.settings_write.

---

How it's tested

The pure builder is exercised in isolation in test_queue_engine.py (no hass). The manager-level path uses the manager fixture + setup_map(...) to seed rooms, then asserts the resolved queue.

---

Known gaps

queue_engine.py (95%) — the remaining uncovered lines are defensive:

• the typing_extensions TypedDict import fallback (lines 9-10), unreachable on the shipped Python (3.14 has typing.TypedDict);
• the room_id <= 0 skip guard in build_room_clean_payload (line 264);
• the elif queue_room_ids: current_room_id fallback in build_active_job_state (lines 397-398), reached only when resolved_rooms is empty but the queue ids are present.

The two capability-gated per-room write branches in build_room_clean_payload — edge_mopping (line 296, under supports_edge + mop mode) and path_type (line 299, under supports_path) — are now covered by test_dispatch_engines.py's DE-11b (test_edge_and_path_per_room_writes_when_caps_enabled), which declares edge/path as real wire fields and asserts both per-room writes land on the wire. No shipped adapter declares these fields, so DE-11b is the only path that drives them.

dispatch_engines.py (95%) leaves line 377 — the if cfg is None: continue skip in DreameSegmentEngine's array transpose, the path where a canonical field is simply not declared in room_fields (Dreame omits most). Defensive; the declared and field_name: None cases are both covered.