|
| 1 | +--- |
| 2 | +'@ifc-lite/drawing-2d': patch |
| 3 | +'@ifc-lite/renderer': patch |
| 4 | +'@ifc-lite/geometry': patch |
| 5 | +--- |
| 6 | + |
| 7 | +Seven public option fields that nothing reads are now marked `@deprecated`, |
| 8 | +with JSDoc that says what actually happens instead of what the old comment |
| 9 | +promised. No behaviour changes and no export is removed or renamed — the |
| 10 | +values were already ignored at runtime; only the type-level documentation |
| 11 | +changes, so editors now warn at the point a caller sets one. |
| 12 | + |
| 13 | +- `SVGExportOptions.units` (`drawing-2d`) — `export()` never destructures it; |
| 14 | + the exporter emits no dimension annotations and always sizes the sheet in |
| 15 | + millimetres. |
| 16 | +- `OpeningFilterOptions.keepBoundarySegments` (`drawing-2d`) — merged into the |
| 17 | + filter's options object but never consulted; `tolerance` is the only field |
| 18 | + that governs how segments near an opening edge are treated. |
| 19 | +- `DoorSymbolConfig.showThreshold` (`drawing-2d`) — no threshold-rendering code |
| 20 | + exists, so `true` and `false` produce identical geometry. |
| 21 | +- `SnapOptions.snapRadius` (`renderer`) — documented as a world-units snap |
| 22 | + distance, but every proximity check reads `screenSnapRadius` (pixels). |
| 23 | + Snapping is screen-space and zoom-dependent; set `screenSnapRadius` instead. |
| 24 | +- `SectionPlaneRenderOptions.flipped` (`renderer`) — the gizmo renderer never |
| 25 | + reads it. The GPU clip plane flips correctly through separate state, so |
| 26 | + cutting behaviour is unaffected; only the gizmo option is inert. |
| 27 | +- `RenderOptions.enableDepthTest` (`renderer`) — dead on both ends: nothing |
| 28 | + sets it and nothing reads it. Depth comparison is fixed per pipeline at |
| 29 | + construction time and is not configurable through `RenderOptions`. |
| 30 | +- `StreamingOptions.onMetadataBootstrap` (`geometry`) — an unfinished stub. Its |
| 31 | + siblings `onBatch`, `onColorUpdate`, `onComplete` and `onError` are all |
| 32 | + dispatched by the bridge; this one never is, so a callback passed here is |
| 33 | + never called. |
| 34 | + |
| 35 | +Deprecating rather than deleting is deliberate: removing an optional field an |
| 36 | +embedder already passes converts a silent no-op into a TypeScript compile |
| 37 | +error, which is a worse first contact with the problem than a deprecation |
| 38 | +warning that explains it. Removal is left as a separate, explicitly versioned |
| 39 | +decision. See issue #2731 for the full audit; the findings that carry a |
| 40 | +behaviour decision (the streaming batch ramp-up, `GeometryQuality`, and the |
| 41 | +scale-bar / north-arrow renderer divergence) are deliberately untouched here. |
0 commit comments