Polish native editor interactions and verification

Author: gajop

docs/porting/01-wip-refactor-plan.md

@@ -0,0 +1,131 @@
+---
+name: Rust WIP refactor plan
+description: Structural work to complete in rust-wip before selectively moving domains into rust-stable
+---
+
+# Rust WIP refactor plan
+
+## Purpose
+
+`rust-wip` is functionally broad but developed as a fast, exploratory workspace.
+Before work is transferred into `rust-stable`, make its ownership boundaries and
+module structure match [conventions.md](conventions.md). The stable branch should
+then receive small, reviewable domain slices rather than architectural cleanup.
+
+This is a plan, not a license to cherry-pick the existing WIP history. The
+current 87 WIP-only commits are archaeological evidence; WIP is the source
+snapshot and stable receives deliberate file-level transfers.
+
+## Assessment baseline
+
+On 2026-07-18, `rust-stable` was the merge base of `rust-wip`. The committed
+delta was 463 files, +44,043 / -2,642 lines. The largest native additions were
+the panel UI (+13,421 lines), Chonsole (+3,729), editing states (+2,999), and
+developer console (+1,944).
+
+The backend is already mostly feature-first: commands, models, and tests are
+generally co-located under their feature directories. The main exception is the
+native UI, which is organized by rendering technology under `panels/editors`.
+
+## Refactors required before stable transfers
+
+### 1. Make feature UI feature-owned
+
+`panels/` must retain only reusable RmlUi infrastructure: host/document
+lifecycle, fields, modal primitives, grid, input, and shared theme. Move actual
+editor behaviour beside the feature it represents:
+
+```text
+objects/ui/{definitions,properties,collision}/
+heightmap/ui/
+grass/ui/
+metal/ui/
+textures/ui/
+map_settings/ui/
+teams/ui/
+project/ui/
+```
+
+The generic panel registry can remain shared, but each feature must register its
+own editor. This is the highest-value refactor: an Objects or Textures transfer
+then becomes a coherent domain copy rather than surgery inside a generic folder.
+
+### 2. Remove the state → panel dependency
+
+Editing states currently depend on `panels::ModelShader` for ghost previews,
+selection highlights, movement, and rotation. Extract world rendering to
+`states` or a neutral renderer module. Map interaction must not depend on the
+right-hand panel implementation.
+
+### 3. Shrink the panel coordinator and editor contract
+
+`PanelManager` currently owns document lifecycle, editor lifecycle, actions,
+field commits/drags, modal coordination, state synchronisation, and cursor tips.
+Keep a small composition point, but extract active-editor lifecycle, modal
+coordination, toolbar actions, and field commits into focused components.
+
+Likewise split the broad `Editor` trait into explicit capabilities around
+lifecycle/rendering, field hosting, brush/state binding, and optional modal
+ownership. Do not introduce a general event bus or UI framework: retain the
+explicit update/input order, just give each concern a narrow owner.
+
+### 4. Finish the Objects boundary
+
+Refactor the Objects UI into definitions/catalog/filtering/placement/thumbnails,
+properties, and collision submodules. Properties and collision must share
+selection projection and object-field mapping rather than duplicate conversion
+logic.
+
+At the same time, complete TODO #1 in [todo.md](todo.md): replace generic
+JSON-based Area/Feature/Unit object commands with concrete typed commands, and
+cache static object-field descriptors. This makes Objects safer and smaller to
+review in stable.
+
+### 5. Split project persistence from the Project feature
+
+Extract project paths plus save/load/export registrations into a neutral
+project-IO/workspace module. Leave scenario metadata, archive export, reload,
+map info, and project-editor functionality in `project/`. This is TODO #16 and
+prevents each saving feature from depending on the whole Project feature.
+
+### 6. Split the texture editor by ownership
+
+Move material discovery/parsing into `textures`, separate saved-brush state, and
+leave a thin UI composition layer. Texture/material ownership must not be
+trapped in one large panel editor before the Textures domain is transferred.
+
+### 7. Establish a shared native RmlUi theme
+
+Centralize common text, color, border, opacity, button, toggle, dropdown, and
+tooltip rules. Keep only document-specific layout in panel, Chonsole, and
+developer-console stylesheets. This prevents future stable slices from changing
+three visual systems for one control correction.
+
+### 8. Separate E2E runner responsibilities
+
+Split the `E2ERun` monolith into engine session lifecycle, input, assertions and
+command-log reading, screenshots/artifacts, and reporting. Keep semantic UI
+coordinates centralized in `scenarios/geometry.py`; that file already fixes the
+previous scattered-coordinate problem.
+
+## Do not rewrite these first
+
+- Chonsole is already sensibly divided into catalog, completion, input, history,
+  model, and views. Transfer it as one domain after shared RmlUi lifecycle work.
+- Developer console has a useful actions/log/metrics/manager/view split. Limit
+  restructuring to common lifecycle and theme extraction.
+- Do not replace the command system wholesale; reconcile only its required WIP
+  delta against stable.
+
+## Completion criteria
+
+Before the first stable transfer:
+
+1. Feature UI no longer lives in `panels/editors`.
+2. `states` no longer imports panel UI/rendering.
+3. Project IO is independent of the Project feature.
+4. The shared native theme has a single source of truth.
+5. The E2E harness and verification ledger identify a feature's tests without
+   relying on obsolete phase documents.
+6. Duplicate model-factory registration fails loudly, as duplicate command
+   registration already does.

docs/porting/02-stable-transfer-plan.md

@@ -0,0 +1,95 @@
+---
+name: Rust stable transfer plan
+description: Order and rules for moving refactored rust-wip domains into rust-stable
+---
+
+# Rust stable transfer plan
+
+## Transfer rule
+
+Follow [conventions.md](conventions.md): do not cherry-pick WIP commits. Copy
+only the files required for one reviewed domain into `rust-stable`, trim them to
+the minimum coherent implementation, test in stable, add a review-queue entry,
+and commit only after user authorization.
+
+Every domain transfer includes its production code, focused Rust tests, E2E
+scenario changes, and visually reviewed goldens. Goldens are never a final bulk
+migration.
+
+## Transfer order
+
+1. **Compatibility spine**
+   - Port flags and Rust-mode Lua self-disable rules.
+   - Native/RmlUi lifecycle contract and explicit engine dependency inventory.
+   - Reload/lifecycle integration coverage.
+
+2. **Native runtime seam**
+   - Minimal `SBC` integration changes.
+   - Model-registry changes required by native producers.
+   - Typed native command submission and command-history projection.
+
+3. **Project IO foundation**
+   - Extracted project paths plus save/load/export registrations.
+   - No project editor UI in this slice.
+
+4. **Generic native UI foundation**
+   - Rml helpers, panel host, common theme, fields, grid, modal primitives, and
+     input.
+   - The control gallery is the acceptance test for this foundation.
+
+5. **Chonsole**
+   - Completion, persistent history, scrolling, mouse input, texture preview,
+     and reload safety as one self-contained domain.
+
+6. **Developer console and status strip**
+   - Depends on Rml lifecycle and command history.
+   - Transfer together so its undo/redo journal has one owner.
+
+7. **World interaction foundation**
+   - Neutral world renderer, selection, rectangle selection, drag/rotate, and
+     ghost previews.
+
+8. **Objects**
+   - Typed commands and descriptors first.
+   - Definitions/placement, then properties and collision as reviewable
+     sub-slices.
+
+9. **Terrain UI**
+   - Heightmap, metal, and grass editors over the existing stable backend.
+   - Reconcile backend deltas rather than overwrite stable implementations.
+
+10. **Textures, then Map Settings**
+    - Texture/material/brush ownership first.
+    - Settings afterwards because optional shading textures depend on it.
+
+11. **Environment**
+    - Lighting, sky, and water under map-settings ownership.
+
+12. **Teams, project info, and toolbar actions**
+    - These intentionally come after the commands/models they invoke.
+
+13. **Remaining domains**
+    - Areas, triggers, variables, and compile/export pieces, each independently
+      testable and reviewable.
+
+## Per-slice exit criteria
+
+Before a domain enters the review queue:
+
+1. `just check` is green in stable.
+2. Feature-local unit/integration tests are green.
+3. The relevant native E2E target passes with zero warnings/errors/crashes.
+4. Every changed golden has been visually inspected and recorded as
+   `ai-reviewed`.
+5. [verification.md](verification.md) has accurate `DONE`/`VERIFIED` evidence.
+6. The review-queue row links the Rust files, replaced Lua code where relevant,
+   a short in-game test recipe, risks, and CI status.
+
+## Important constraints
+
+- `rust-stable` already contains substantial backend work. Reconcile every
+  transfer; do not blindly replace a stable directory with its WIP counterpart.
+- Lua/RmlUi work is a separate historical product line. Only move Lua changes
+  needed to prevent ownership collisions in Rust mode.
+- Engine changes live outside this repository and need their own explicit
+  compatibility inventory before a stable cutover.