Files
effectcraft/crates/svg
arnav-hiwarkarandClaude Opus 5.5 2d216151c6 Import: Composition - Retain Layer Sizes for PDF / Illustrator files (#1536)
The PDF / Illustrator import dialog now offers Composition - Retain Layer
Sizes. Each file layer's footage is cropped to its artwork bounds (fills,
stroke outlines, images, clips) and positioned to keep its place on the
page. Decoding, continuous rasterisation, shape conversion and reload share
one crop (effectcraft_pdf::layer_doc_with_sizes), so they agree.

Also collapses a match guard in the STL reader for current stable clippy.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-15 13:33:50 +05:30
..

effectcraft-svg

SVG import for EffectCraft: .svg files import as footage (rasterised at any scale, so Continuously Rasterize stays sharp) and Layer ▸ Create ▸ Create Shapes from Vector Layer turns them into editable shape layers.

Implemented from the W3C Scalable Vector Graphics (SVG) 1.1 (Second Edition) recommendation (16 August 2011) and the SVG 2 candidate recommendation (4 October 2018) for the parts listed here, plus CSS Color Module Level 3 for colour keywords and functions:

  • document: svg (width/height, viewBox, preserveAspectRatio), nested svg, g, defs, use (x/y, href / xlink:href), symbol (through use), title/desc/metadata ignored;
  • shapes: path (all path data commands incl. elliptical arcs), rect (rx/ry), circle, ellipse, line, polyline, polygon;
  • transform lists (matrix, translate, scale, rotate with centre, skewX, skewY);
  • painting: fill, fill-rule, fill-opacity, stroke, stroke-width, stroke-linecap, stroke-linejoin, stroke-miterlimit, stroke-dasharray, stroke-dashoffset, stroke-opacity, opacity, display, visibility, color/currentColor;
  • linearGradient and radialGradient (with fx/fy focal point), gradientUnits, gradientTransform, spreadMethod, href inheritance of stops and attributes;
  • styling through presentation attributes, the style attribute and <style> sheets with type, class, id and universal selectors (simple selectors and selector lists).

Not supported (ignored): text, images, filters, clip paths, masks, patterns, markers, animation. XML parsing uses roxmltree (MIT/Apache-2.0). Test fixtures are original work (see tests/fixtures/*.attribution).