Pixels per document point when an SVG is embedded in PDF, DOCX, PPTX, or IDML.
Matches the viewer paint-scale cap so export hairlines sit in the same class
as on-screen paint. Official PNG comparison stays at scale 2; that gate is
not the export resolution.
Widest or tallest SVG raster we will allocate. A landscape page at the viewer
cap (4 px/pt) stays under this; larger requests fail instead of exhausting WASM.
Resolve the geometry fragment for a paint op. Prefers an exact rect match when
multiple boxes share node_id (multi-column / fragmented text on one page).
Rasterize the whole SVG into dest_w×dest_h with one scale transform.
Callers pass a size that preserves aspect (the contain letterbox). Does not
nearest-neighbor upscale a smaller bitmap.
Viewport size in pixels: the SVG width/height rounded up, matching
[decode_svg]. Fit math uses this so a scaled raster keeps the same aspect
as the intrinsic decode.