v0.1.15

[BobLd]

Release v0.1.15

Full list of changes: https://github.com/UglyToad/PdfPig/releases/tag/v0.1.15

Disclaimer: the below was generated with assistance of AI

---

✨ New Features

Feature	PR	Notes
.NET Standard 2.1 target	#1283	New TFM (pulls in Microsoft.Bcl.HashCode for that target). Purely additive.
Open PDFs larger than 2 GiB	#1293	Via PdfDocument.Open(Stream) with your own FileStream. FileHeaderOffset.Value widened int → long.
TIFF (TIF) image support	#1227	New supported image format.
Exposed cross-reference (xref) table	#1292	xref data is now accessible from the public API.
ParsingOptions.UseActualText	#1312	Opt-in: uses marked-content /ActualText when extracting text. Defaults to false, so no behavior change unless enabled.
TokenScanner exposed via Structure	#1308	Lets you iterate indirect objects directly.
Allocation-free span APIs	#1301	PdfFunction.Eval(ReadOnlySpan<double>, Span<double>), ColorSpaceDetails.GetRgb(...), span methods on Shading.
Default colour spaces (/DefaultGray, /DefaultRGB, /DefaultCMYK)	#1337	Now honoured.
Rendering intent RI processed in SetNamedGraphicsState()	#1336
Type 3 font CharProcs handling + better bounding boxes	#1295, #1299, #1315
ProcessOperations is now virtual; BeginText/EndText overridable; CoreTokenScanner.ClearPreReadByte()	#1296, #1294, #1334	Extensibility hooks for custom processors.
Annotation actions ported	#1326	Annotation /A actions now extracted.
StackDepthGuard exposed via ISeekableTokenScanner	#1276	Nesting-depth limit enforced when resolving indirect refs.
Portable DebugType for NuGet	#1342	Fixes symbol publishing.

Plus a large batch of font/parsing bug fixes (no API change): TrueType simple-font missing widths #1311, CIDFont glyph displacement #1316, CID/CFF glyph naming #1318/#1322, CIDFontType0 OpenType/CFF paths #1321, Type 0/2 CID default widths & vertical displacement #1325, indirect-ref width arrays #1307, encrypted-string BOM decryption #1306, image decode values #1304, malformed CMap duplicate codes #1313/#1319, truncated inline-image EI detection #1329, TJ text-sequence over-increment #1298.

---

⚠️ Breaking Changes (with migration snippets)

1. PdfFunction.Eval — new abstract span overload (#1301)

The array overload public abstract double[] Eval(params double[]) became virtual, and a new abstract span method was added. Custom PdfFunction subclasses must now override the span version.

// BEFORE (v0.1.14)
public override double[] Eval(params double[] input) { /* ... */ }

// AFTER (v0.1.15) — override the span method instead.
// Return value is the number of components written into 'output'.
public override int Eval(ReadOnlySpan<double> input, Span<double> output)
{
    // write results into output, return count
    output[0] = ...;
    return 1;
}
// The double[] Eval(...) overload now delegates to this automatically.

2. ColorSpaceDetails — new abstract GetRgb (#1301)

Any custom ColorSpaceDetails subclass must implement the new abstract method.

// AFTER (v0.1.15) — add to your custom color space:
public override void GetRgb(ReadOnlySpan<double> values, out double r, out double g, out double b)
{
    var color = GetColor(values.ToArray());          // simplest bridge
    (r, g, b) = color.ToRGBValues();
}

3. ISeekableTokenScanner — new StackDepthGuard member (#1276)

Adding a member to a public interface breaks custom implementers.

// AFTER (v0.1.15) — add to your custom scanner:
public StackDepthGuard StackDepthGuard { get; } = new StackDepthGuard();

4. IOperationContext — new BeginText() / EndText() members (#1294)

Custom IOperationContext implementations must add these. (BaseStreamProcessor provides virtual defaults if you derive from it.)

// AFTER (v0.1.15) — add to your custom operation context:
public void BeginText() { /* init text matrices, start clip path */ }
public void EndText()   { /* reset text matrices, apply clip path */ }

5. UglyToad.PdfPig.Rendering types removed from the core package (#1323)

IPageImageRenderer and PdfRendererImageFormat were removed from UglyToad.PdfPig.

// AFTER: use the dedicated rendering package(s) instead of the core stubs.
//   dotnet add package UglyToad.PdfPig.Rendering.Skia
// and use that package's renderer API rather than the old core IPageImageRenderer.

6. Nesting-depth errors now throw PdfDocumentStackDepthException (#1310)

Previously this was a PdfDocumentFormatException. It now derives from Exception (not PdfDocumentFormatException), so existing catch blocks won't catch it.

// BEFORE — this used to catch the "max nesting depth exceeded" case:
catch (PdfDocumentFormatException ex) { /* ... */ }

// AFTER (v0.1.15) — handle it explicitly:
catch (PdfDocumentStackDepthException ex) { /* deeply nested / malicious doc */ }
catch (PdfDocumentFormatException ex)     { /* other malformed-PDF cases */ }

7. Page.Width / Page.Height / Page.Size recomputed (#1340) — behavioral

Page dimensions are now derived from the CropBox visible bounds with rotation applied (cropBox.GetVisibleBounds(rotation)), and Size from the CropBox bounds — rather than from the MediaBox∩CropBox intersection. Reported page width/height/size may change for rotated and/or cropped pages. No code change needed, but verify any layout logic that hard-codes expected page sizes.

How to adapt

• You read page.Width / page.Height → nothing to do; they still return the visible (rotation‑applied) size in points.
• You need the page's on‑screen size but were reading the boxes → switch to Width/Height:

// before
var w = page.CropBox.Bounds.Width;
var h = page.CropBox.Bounds.Height;
// after
var w = page.Width;
var h = page.Height;

• You relied on the old "as‑displayed" crop box rectangle (origin at 0,0, rotation applied) → use the new helper:

// before: page.CropBox.Bounds was already rotated/normalised
var displayed = page.CropBox.Bounds;
// after: ask for the visible bounds explicitly
var displayed = page.CropBox.GetVisibleBounds(page.Rotation);

The old clipping was initialMatrix.Transform(cropBox.Bounds).Normalise(). Since the initial matrix is built from this same crop box, it maps the crop box to the origin and applies the rotation, yielding (0, 0, displayedWidth, displayedHeight), precisely what GetVisibleBounds() returns directly. So the clipping path is identical for all rotations and crop offsets (verified by the rotated/cropped visual-verification and clipping tests).

---

New contributors: @username77 (#1227), @vafle228 (#1292), @VarunSaiTeja (#1307), @jeske (#1308), @OrganizationUsername (#1293), plus the GitHub Actions bot.