pdfrum-anyrender

anyrender scenes written as vector PDF pages through pdfrum-edit. Paths stay paths, gradients stay gradients, text stays text in embedded, subsetted fonts, and images keep their source encoding where PDF allows (a JPEG as-is, an opaque PNG's compressed data as-is).

A page is painted into anyrender's recording Scene first, by whatever renders through anyrender. [write()] then embeds every face and image the scenes use and replays each scene onto a pdfrum canvas: anyrender's push/pop layers become nested saved graphics states, the only shape a pdfrum canvas lets a content stream take.

use anyrender::Scene;
use pdfrum_anyrender::{GlyphArea, Page, Sources, write};
use peniko::kurbo::{Affine, Rect, Size};

let page = Page {
    size: Size::new(200.0, 100.0),
    scene: Scene::default(),
    placement: Affine::IDENTITY,
    clip: Rect::new(0.0, 0.0, 200.0, 100.0),
    area: GlyphArea::Anywhere,
};
let written = write(&[page], &Sources::default())?;
assert!(written.bytes.starts_with(b"%PDF-"));
# Ok::<(), pdfrum_edit::Error>(())

anyrender's glyph runs carry glyph ids, not the text they were shaped from. Hand the text over in [Sources::texts], keyed by the run as the scene holds it ([RunKey]), to make words selectable; a run without an entry falls back to the font's own cmap, reversed ([GlyphTally] counts both).

What is simplified

Box shadows are dropped; filters and backdrop filters are ignored; compositing operators other than source-over paint as source-over; a sweep gradient, and a gradient on a stroke or on text, paints as its stops' average colour; gradients pad at their ends whatever extend mode they name, and interpolate in sRGB; an image brush draws once, clipped to its shape.

Test font

tests/fixtures/noto-serif-latin.ttf is a Latin subset of Noto Serif, licensed under the SIL Open Font License 1.1 (tests/fixtures/OFL-noto-serif.txt). It is a test fixture and is not part of the published crate's code.