[gunicsba]

Summary

Closes #100.

For condensed-work-state implements, get_section_actual_state() calls is_element_or_parent_off(). That function used to search the whole object pool for an element's parent on every call, three loops deep and once per level. It runs for every section from the AOG heartbeat (PGN 0xF0, every 100 ms) and from the VT section map.

Now each element's parent is read once from its parent object ID when the pool is activated (ClientState::build_element_hierarchy()), and the check just walks that map. The lookup lives in a small header, include/element_hierarchy.hpp, so it can be unit tested without the TC server.

Behaviour change

The old search looked for parents in element child lists. In ISO 11783-10 those lists only reference process data and properties, so it rarely found the real parent. Its fallback, added for the Müller master switch in #41, matched elements through ddiToElementNumber. That map keeps one element per DDI, so it could pick a wrong element.

I compared old and new on every pool in the fixture collection (25 pools):

	Old	New
Real parent found	Rarely	Always (from the parent object ID)
Wrong "parent"	Sulky 12 sec: sibling sections (8 → 13, 11 → 15). Kuhn Espro: 1 → 2 (real parent is 0). Sulky 6 sec: cycle 2 → 0 → 2	None
Müller pools (#41)	Condensed host 2 → 0	Same
64 lookups, desktop PC	1–23 ms	about 4 µs

The only pools where the reported section state can change are those where the condensed-state element now also checks an ancestor that reports DDI 141 (Actual Work State). That's the root element in the AgIsoStack example seeder, the AGL sprayer, and the MFG89 E-COM UD and IXSPRAY demos. If that ancestor reports "not working", the sections now report OFF to AOG. That's the master-switch behaviour #41 was aiming for.

Implements using per-element control still return early and are not affected.

Also

• Adds the missing #include <algorithm> for std::find_if (item 5 in Splitting out an embeddable core so changes flow to AgOpenWeb's HAT daemon #101), and drops the now unused <set>.

Testing

• New element_hierarchy CTest: parents come from parent object IDs; only ancestors are checked (a sibling or child being off doesn't count); a malformed pool with a parent cycle terminates.
• All 7 tests pass. Release build passes on Windows (MSVC). clang-format and cmake-format are clean.
• Not yet run on a Pi or a real implement. @chriskinal, your CM4 setup with the 64-section simulator would be the best check that the thread load drops.

🤖 Generated with Claude Code