Skip to content

audit.md

Latest commit

 

History

History
116 lines (93 loc) · 6.85 KB

audit.md

File metadata and controls

116 lines (93 loc) · 6.85 KB

Level Editor Productization Audit

Date: 2026-05-06

Scope Decision

This checkpoint productizes the editor as a classic-subset editor. It may import and preserve .nxlv data that is outside the classic subset, but it does not claim NeoLemmix feature parity. Unsupported or lossy operations must be visible before export, quick-fix, playtest, or import acceptance can surprise a user.

Docs vs Code

Area Documented Current implementation Gap
Standalone editor editor.html with canvas, palette, inspector, validation, local save, import/export Implemented through EditorUiController, editor session, controller, and E2E harness Product status and warnings need to state classic-subset limits directly in the UI
.nxlv import/export Structured editor model with comments and unknown sections preserved Parser/writer preserves top-level unknown lines, section-local unknown lines, unknown sections, and terrain groups Round-trip coverage must prove semantic preservation for comments, unknown sections, and grouped terrain
Classic .lvl import/export Supported as classic workflow Implemented via LevelReader, LevelWriter, and classic conversion helpers Export is intentionally lossy; the UI and validation need explicit warnings before users rely on it
Validation Errors/warnings with quick fixes Implemented in EditorValidator and rendered in the validation panel Destructive fixes need clear labels/metadata so users can distinguish repair from data-dropping cleanup
Playtest Toggle playtest without leaving editor Implemented through editor preview/runtime path Capture coverage should show both edit and playtest states
NeoLemmix expansion Documented as out of scope Parser preserves some unsupported data but runtime preview remains classic Unsupported sections/properties must be visible warnings, not silent preservation claims

Implemented vs Claimed

Implemented and suitable for productized classic-subset use:

  • Create blank levels and edit headers, skill counts, terrain, gadgets, steel, MIDI flags, and entrance/exit gadgets.
  • Select, multi-select, marquee-select, move, nudge, duplicate, copy/paste, reorder, align/distribute, replace, randomize, and delete entries.
  • Save and load editor levels from local storage.
  • Export/import .nxlv editor levels.
  • Import/export classic .lvl through the classic runtime representation.
  • Validate common structural issues and apply quick fixes.
  • Playtest the editor level through the game runtime.
  • Expose editor state and operations through window.__E2E__ for Playwright and MCP tooling.

Not implemented or not claimed in this checkpoint:

  • Full NeoLemmix section semantics for $TERRAINGROUP, $TALISMAN, $PRETEXT, $POSTTEXT, multiple entrances/exits beyond classic limits, custom trigger boxes, or style metadata.
  • Classic .lvl preservation of editor-only terrain flags such as ONE_WAY, editor-only transforms, comments, unknown sections, or grouped terrain.
  • Solver-backed export blocking.
  • Real archive/install tooling for exported editor pack bundles. The editor can store local projects, export the JSON handoff bundle with info.nxmi, levels.nxmi, level .nxlv text, and validation reports, and now export or install an editor pack archive with manifest validation before local storage writes.

Workflow Coverage Map

Workflow Current coverage Required checkpoint coverage
Create playable level Unit and harness coverage exists for editor state, tools, entrance/exit, round-trip workflow Keep semantic create/validate/playtest/save/export/import test green
Import .nxlv with unknown data Parser/writer tests cover parts of preservation Add explicit semantic round-trip for comments, unknown sections, terrain groups, and unsupported data warnings
Classic .lvl import/export Harness coverage exports/imports a classic level Add lossy-export contract tests and cap validation before export
Import failures UI file handlers report read/parse failures into status and validation Keep visible failure coverage green
Validation quick fixes Unit coverage exists for validation and fix callbacks Mark destructive fixes and expose that status to UI/E2E
Visual states Editor capture target exists Split capture states for shell, palette/inspector, validation, file controls, and playtest
Project workflow Local project storage, pack JSON export, and pack archive install exist Keep archive install validation aligned with future pack metadata fields

UX Issues

  1. Classic-subset boundaries are too implicit. Users can import a richer .nxlv and see a playable preview without understanding which data is preserved but unsupported by preview/export.
  2. Classic .lvl export is available next to .nxlv export, but its lossy nature is not prominent enough.
  3. Validation quick fixes can remove or clamp data. The UI needs to label those actions as destructive when they drop unsupported properties or entries.
  4. Import failures should land in the same status/validation surface as normal editor issues, not only browser alerts or console errors.
  5. Multi-selection is powerful but status text should keep reflecting batch scope and avoid implying single-entry edits when multiple entries are active.

Correctness And Data-Integrity Risks

  1. Unknown .nxlv sections are preserved by the parser/writer, but regressions would be easy without a focused semantic test.
  2. Terrain groups are preserved in the editor model and serialized back, but the classic preview/runtime path does not implement full group behavior.
  3. Classic .lvl export cannot carry comments, unknown sections, terrain groups, ONE_WAY terrain flags, or unsupported transforms.
  4. Classic runtime caps need validation before export: title length, object count, terrain count, steel count, level size, entrance/exit limits, and unsupported props.
  5. Object/gadget round-trip tests should compare meaningful fields, not merely assert that import succeeds.

Prioritized Fixes

P0:

  • Add visible classic-subset warnings for unknown sections, terrain groups, unsupported NeoLemmix data, destructive fixes, and classic .lvl export.
  • Add semantic .nxlv and classic lossy round-trip tests.
  • Add classic export cap validation.
  • Fix import error reporting so bad .nxlv and .lvl files update editor status/validation.

P1:

  • Expand editor E2E coverage for visible warnings, import failures, and semantic import/export acceptance.
  • Split editor capture targets by workflow state and keep outputs under ignored temp/.

P2:

  • Add a small multi-selection batch edit affordance only where the controller already supports safe bulk operations.
  • Use audit captures to decide whether palette favorites or stronger recent item affordances are worth the next editor milestone.