don
/
neon-sprawl
mirror of https://github.com/ViPro-Technologies/neon-sprawl
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
neon-sprawl/docs/plans/NEO-26-implementation-plan.md

9.1 KiB
Raw Permalink Blame History

NEO-26 — Implementation plan

Story reference

Field Value
Key NEO-26
Title E1.M3: SelectionEvent surface + debug or HUD consumer
Linear NEO-26
Slug E1M3-04
Git branch NEO-26-selectionevent-surface-debug-consumer
Parent context Epic 1 — Core Player Runtime · E1M3 prototype backlog
Depends on NEO-23 (TargetState v1), NEO-24 (tab/clear + TargetSelectionClient) — code paths exist on main; Linear may still list relations until those issues are closed.
Decomposition E1.M3 — InteractionAndTargetingLayer (SelectionEvent contract); E1.M4 — AbilityInputScaffold (future subscriber).

Goal, scope, and out-of-scope

Goal: Every material change to the client’s server-acknowledged combat target lock id (lockedTargetId in the cached TargetState mirror) is observable as a SelectionEvent carrying previous id, next id, and cause (tab, clear, server_correction, …). Ship one built-in consumer: dev-only print behind an @export flag (planning choice 2026-04-25) so the bus is exercised without new HUD nodes.

In scope

  • Minimal GDScript event shape (Dictionary or small typed helper) aligned with the illustrative table in E1.M3: previous (Variant String or null), next (same), cause (String enum-like literals).
  • Cause attribution at the authoritative update boundary inside the existing selection client (see Technical approach) so we do not guess causes from state diffs alone.
  • Consumer: optional print when @export var log_selection_events (name TBD) is true on TargetSelectionClient (or tiny child node); document the subscription point for E1.M4 (signal name + emitter node path pattern).
  • Tests: tab vs clear vs movement-driven GET produce distinct cause strings when they each produce a lock-id transition (or documented no-op when id unchanged).

Out of scope (per Linear / backlog)

  • Full prompt / party UI; wire protobuf events.
  • Interactable focus / primary selection as SelectionEvent — target lock only for this ticket (see D6); catalog exists regardless of NEO-25 lifecycle.
  • Changing server TargetState contracts (NEO-23); new HTTP routes.

Acceptance criteria checklist

  • Tab path: after a successful selection cycle that changes lockedTargetId, a SelectionEvent is emitted with cause tab (and correct previous/next ids).
  • Clear path: clearing the lock emits clear when lockedTargetId transitions from non-null to null.
  • Server correction: a GET-driven authoritative update that changes lockedTargetId emits server_correction. Id-unchanged GETs (e.g. validity only) do not emit SelectionEvent for v1 — test_get_validity_only_change_emits_nothing; id change via mock in test_get_lock_id_change_emits_server_correction.
  • E1.M4 hook: this plan + client/README.md (“SelectionEvent (NEO-26)”) document TargetSelectionClient.selection_event and stable payload keys.

Technical approach

  1. Single emitter: Extend TargetSelectionClient (client/scripts/target_selection_client.gd) rather than diffing target_state_changed in a second node — only this class knows whether the in-flight HTTP phase was GET (refresh) vs POST from tab, clear, or direct request_select_target_id (if used).

  2. Intent / phase tags: Before starting HTTP, set a private enum or String tag, e.g. _pending_selection_cause, cleared after handling the response:

    • request_tab_next → tab (even though it calls request_select_target_id internally, set cause at the tab entrypoint before POST).
    • request_clear_target → clear.
    • request_select_target_id when not entered from request_tab_next → use cause tab (same string as Tab path — planning choice 2026-04-25: no separate direct_select).
    • request_sync_from_server when invoked from on_authoritative_ack → mark refresh as server_correction for the next GET completion.
    • Boot / one-shot sync from main.gd → Decision (2026-04-25): first and later GET completions use cause server_correction; only emit if normalized lockedTargetId differs from prior snapshot (avoids spam on identical boot GET).

  3. Emit rule: After _state is updated in _update_state_from_get / _update_state_from_post, compare previous normalized lockedTargetId to new; if equal, no SelectionEvent (validity-only changes stay on target_state_changed only unless product later expands the contract).

  4. Signal: Add e.g. selection_event(event: Dictionary) with keys previous, next, cause (and optional sequence int passthrough for debugging). Keep target_state_changed unchanged for existing HUD.

  5. Consumer: No new scene nodes — TargetSelectionClient.log_selection_events mirrors each selection_event to Output via print.

  6. Docs: Update client/README.md “Target lock + tab cycle” area with a short “SelectionEvent (NEO-26)” subsection: signal name, payload keys, print flag, E1.M4 subscription note.

Files to add

Path Purpose
client/test/selection_event_client_test.gd GdUnit4: assert causes for tab POST, clear POST, and GET-with-mock-body changing lock id (TargetSelectionTestDouble pattern).
docs/manual-qa/NEO-26.md Manual QA for log_selection_events + optional subscription check.

Files to modify

Path Rationale
client/scripts/target_selection_client.gd Track pending cause through GET/POST; compare old/new lockedTargetId; emit selection_event; optional print behind @export; document next to target_state_changed.
client/README.md Document subscription point + payload for E1.M4.
docs/decomposition/modules/E1_M3_InteractionAndTargetingLayer.md Implementation snapshot bullet: replace “Still open — SelectionEvent” with pointer to NEO-26 + signal name (keeps decomposition aligned with shipped code).

Tests

Test file What it covers
client/test/selection_event_client_test.gd Add: Tab mock POST changes lock → event cause == "tab". Clear mock → "clear". Simulated GET completion after flagging movement refresh → "server_correction" when body changes lockedTargetId. Assert no event when id unchanged.
client/test/target_selection_client_test.gd Change (optional): If new behavior duplicates setup, keep NEO-24 tests focused on HTTP/state; only touch this file when sharing helpers avoids copy-paste.

Manual verification: docs/manual-qa/NEO-26.md.

Open questions / risks

  • Validity-only updates: If v1 stays lock-id-only (recommended), extend later with kind or dual fields if telemetry needs validity transitions on the same bus.
  • Linear blocked-by: Relations may still show NEO-23/NEO-24 until closed; implementation assumes their client/server surfaces on main are present.

Decisions

Topic Decision
D1 — Emitter location TargetSelectionClient owns selection_event (intent-aware).
D2 — Boot GET Use cause server_correction for all GET completions (including first sync); suppress event if normalized lockedTargetId unchanged vs prior snapshot.
D3 — Non-tab request_select_target_id Use cause tab (same literal as Tab path; no direct_select).
D4 — Consumer Print only behind @export flag on emitter; no HUD panel in NEO-26.
D5 — Emit rule (v1) Lock id only: emit SelectionEvent only when normalized lockedTargetId changes; validity / sequence updates without id change → no event (HUD keeps using target_state_changed). Rationale (2026-04-25): matches backlog “previous / next” id story, smallest payload, clear E1.M4 hook for cast target; avoids conflating soft-lock range telemetry with selection changes. Override if you want validity on this signal too.
D6 — Interactables Out of scope: target lock only for NEO-26. Rationale: Linear / E1M3-04 acceptance is about combat target lock changes (SelectionEvent for tab-target), not interactable focus or “use now” flow (NEO-9 / catalog). Descriptor work (NEO-25) can be done and this story still stays narrow: a unified “all selection kinds” bus would be extra contract design (payload + causes + wiring) that this ticket never asked for—not a dependency on NEO-25 being open.