Delete the private functions and fields nothing used

[saworbit]

Fixes #609
Fixes #610

plugin.gd and level_root.gd are documented as thin coordinators whose named one-line delegates exist to be called by name, and the rule beside that is that a delegate with no caller gets deleted. Fifty-seven private functions and nine fields had drifted past it.

The lists were stale, so I re-derived them

The issue lists were taken a while ago and two of their own instructions no longer hold:

• It says to wait on _on_shortcuts_help pending The searchable shortcut dialog cannot be opened: the '?' button the user guide documents does not exist #606. The searchable shortcut dialog cannot be opened: the '?' button the user guide documents does not exist #606 is merged, that handler is connected at dock.gd:2058, and test_unreachable_entry_points.gd holds it there. It is not on the list.
• _point_near_polygon_3d is dead but my scan missed it, because the name appears in prose elsewhere and the method counts any mention as a caller. That makes the scan conservative, never aggressive, which is the right direction — but it means each deletion still needs its own grep, which each one got.

What was actually keeping this alive

Five architecture tests, and this is the more useful half of the finding.

Each enumerates a module's methods and asserts plugin.gd contains a delegation string for every one:

for method_name in ["can_drop_data", "drop_data", "is_entity_drag_data", ...]:
    assert_true(source.contains("HFPluginDropHandler.%s" % method_name))

Those lists had grown to include helpers the modules call themselves: update_preview, apply_value, do_displacement_stroke, point_near_polygon_3d, update_prefab_hover, move_selected_vertical, normalize_editor_selection, expand_native_group_selection, same_node_selection, sync_hf_selection_if_empty, face_screen_center, show_quick_property, ensure_vertex_overlay. The only way to satisfy the assertion was a wrapper nothing called, so tests written to keep plugin.gd thin were requiring it to be thicker.

I checked every one: all are reachable through entry points plugin.gd genuinely does delegate. Each list is now those entry points, and the drop test additionally asserts plugin.gd does not contain the payload type strings. That is what "thin delegate" was reaching for; the enumeration could pass while plugin.gd held a full duplicate of the classification.

Three that are more than dead weight

map_io.gd:_format_face_line is a drifted second copy of the .map face line. It takes three arguments and hard-codes the texture and the 0 0 0 1 1 tail; the three adapters take (a, b, c, texture, face_data) and read real UV data off FaceData. The signatures differ, so nothing could have called it even by accident, and a second copy of the export format is a trap for whoever fixes the exporter next.

LevelRoot.drag_active is a latent bug, not dead weight. Its setter calls input_state.cancel(); the canonical cancel_drag() is input_state.cancel() and _clear_preview(). So drag_active = false did half a cancel and left the preview brush orphaned in the scene. It is public on a registered custom type, so anyone who found it got the broken half.

DraftEntity.entity_properties was a second public name for entity_data that nothing used. Two names for one dictionary can only diverge.

Both public deletions are a break for any script that touched them, which is stated here rather than buried.

Deleting exposed one more layer, which is also gone

HFPluginSelectionState.selection_has_brush() and selection_has_entity() had exactly one caller each: the dead plugin.gd wrapper. Removing it made them provably dead, so they went in the same pass.

Re-running the scan afterwards finds zero function candidates and zero field candidates, so there is no third layer.

grid_plane_axis and _entity_props_entity were written and never read, so the writes in hf_grid_system.gd and dock_entity_handler.gd went with the declarations — deleting only the declarations would have left assignments to properties that no longer exist.

Docs

CHANGELOG. Nothing else goes stale: the four references in DEVELOPMENT.md and ROADMAP.md all point at the live plugin_drop_handler.gd or the adapter base's format_face_line.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-16T23:36:58.284266Z	c392921	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.