PeachDrawing.Svg

PeachDrawing.Svg is a vector SVG backend for PeachDrawing.Core: the third concrete Canvas/RenderContext beside the PDF writer and the raster canvas, usable with no PeachPDF reference at all. Draw shapes, gradients, clips, text and images onto an SvgCanvas and save the result as an SVG document - no HTML, CSS or PDF involved anywhere. It keeps everything SVG can express as vector content: paths keep their curves, gradients stay gradients, clips stay clips, glyphs are exact outlines (or selectable <text> with an embedded subset font on request), and PNG, JPEG, GIF, WebP and AVIF images are embedded exactly as they came. It depends on PeachDrawing.Core, PeachDrawing (for its raster fallback) and PeachImage, is trimmable and Native AOT compatible, and is versioned in lockstep with PeachPDF: the same version number for every release.

dotnet add package PeachDrawing.Svg

Status: pre-1.0. SvgRenderContext, SvgComposer, SvgComposerOptions, SvgTextMode and SvgCanvas are the package’s full public surface - see PublicApi.txt for the exact, machine-checked snapshot. Until 1.0 the public API may change between releases.

Quick start

using PeachDrawing.Core;
using PeachDrawing.Svg;

var context = new SvgRenderContext();
var composer = new SvgComposer(context, widthPoints: 480, heightPoints: 300);
var g = composer.Canvas;

g.DrawRectangle(context.GetSolidBrush(PaintColor.FromArgb(255, 30, 144, 255)), x: 20, y: 20, width: 200, height: 100);

var font = context.GetFont("Arial", size: 24, PaintFontStyle.Regular);
if (font is not null)
{
    g.DrawString("Hello, SVG", font, PaintColor.Black, new PaintPoint(20, 150), g.MeasureString("Hello, SVG", font));
}

await using var file = File.Create("out.svg");
composer.Save(file);

An SvgComposer is one SVG document - SVG has no pages, so there is no page to add; make one composer for each image. Like the PDF writer, SvgComposer works in points with the origin at the top left and y growing downwards; the document’s width/height are in pt and its viewBox is in the render context’s layout units (SvgRenderContext.PixelsPerPoint of them to a point, 1 by default). The SvgRenderContext constructor registers every font installed on the machine; AddFont registers a TrueType, OpenType, WOFF or WOFF2 font from a stream under a family name, which is the way to get the same text on every machine.

With PixelsPerPoint at its default of 1 every coordinate is simply a point. A host that lays out in other units (PeachPDF’s HTML engine works in PixelsPerInch / 72 units to the point) sets PixelsPerPoint and then follows the same reading the PDF and raster canvases give the Canvas API: rectangles, lines, polygons, image rectangles, clip rectangles, brush geometry and glyph positions are in layout units, while the coordinates of a GraphicsPath (for DrawPath and PushClip), and a Pen’s width and dash lengths, are in points. SvgCanvas scales those to the document’s units when it writes them.

The repository’s src/PeachDrawing.Svg.Example project is a runnable version of this with a handful of scenes (shapes and dashes, gradients, clips and transforms, blend modes and layer effects, patterns and images, text in both modes); run it and open the files in a browser.

What each Canvas call becomes

An image a browser can show (PNG, JPEG, GIF, WebP, AVIF) is embedded byte for byte, never re-encoded, so an animated GIF stays animated; any other format PeachImage reads is embedded as a PNG of its pixels. An image is embedded once however often it is drawn, and gradients, clip paths, patterns, masks and filters that are identical are defined once and shared.

Each push (clip, transform, blend mode, layer) is counted like the PDF canvas’s saved graphics states: a pop undoes the innermost push, and a pop with nothing pushed throws InvalidOperationException. Anything still open when the document is saved is closed, and a group that holds nothing is not written.

Text

SvgComposerOptions.TextMode chooses how text is written:

• SvgTextMode.Outlines (the default) - every glyph is an exact vector outline. The text looks identical in every viewer, whatever fonts it has; shaping, kerning, synthetic bold and italic, colour glyphs (COLR/CPAL, drawn as vector paths and gradients), bitmap glyphs and OpenType-SVG glyphs are all handled. Each glyph’s outline is defined once per size and slant and every occurrence is a <use>, so a screenful of text is far smaller than the outlines written out again at every position. The text is not selectable.
• SvgTextMode.Text - selectable, searchable <text>, with each character placed at the position the shaper chose and a subset of the font embedded in the document as an @font-face. A run a viewer could not reproduce that way - ligatures, reordering, colour or bitmap glyphs, or a font that cannot be embedded as an OpenType subset - is still written as outlines, so the text always looks right.

Canvas.InvisibleText is honoured in text mode (the text is written transparent, so it can be selected over content drawn some other way) and skipped in outline mode.

Limitations

• Conic gradients have no SVG paint server, so by default they are a fan of 360 thin wedges (one degree each) clipped to the shape, or, for a Pen, painted through a mask of the stroke; a very large gradient shows the steps. SvgComposerOptions.ConicGradients = SvgConicGradientMode.Css writes a CSS conic-gradient() in a <foreignObject> instead - exact and smooth in a browser - inside a <switch> whose fallback is the wedges, so a viewer that ignores foreign content (resvg, Inkscape and most others) still shows them.
• Colour. Gradient stops interpolate in non-premultiplied sRGB, the SVG default; a CMYK colour is written as its RGB rendition.
• Links to another document. A destination is a fragment of its own SVG, so a link to a place named on another canvas cannot be followed. That happens only when one multi-page source (an HTML document with paper pages) is written as several SVGs; a source written as one continuous sheet, as the showcases are, has every destination in the file.
• Bookmarks, forms and tagging are PDF concepts and have no counterpart: BeginMarkedContent and its relatives do nothing.

Licences

PeachDrawing.Svg is BSD 3-Clause-licensed like the rest of PeachPDF; it contains no third-party code (see its THIRD-PARTY-LICENSES.md).