MIT Free for personal and commercial use. No licensing fees for private or commercial .NET apps. License

Changelog

All notable changes to this project are documented here. The format follows Keep a Changelog and this project adheres to Semantic Versioning.

[2.4.0] - 2026-09-27

A performance release. Documents render 2–30× faster with 60–95% fewer allocations, and existing code needs no changes. In a 2-CPU / 1 GB container, a one-page invoice went from 356 to about 10,200 pages per second and a 19-page annual report from 3,100 to about 25,200. Output is byte-identical apart from the text-run, PNG and font-width changes below.

Performance

  • Text layout: each word's width, font and colour are computed once, and wrapped lines and table row heights are reused for the rest of the render. Documents without page numbers skip the second layout pass. Consecutive words in the same style share one Tj operator, so PDFs are up to 14% smaller.
  • Tables: a page slice of a split table only visits its own rows, and each cell allocates about 70% less.
  • Images: PNGs are decoded when the document is saved, once per distinct image, and the result is cached for the life of the process (up to 32 MB), so a logo used in every document is converted once. RGB and palette PNGs are embedded still compressed (/Predictor 15), with no decode at all.
  • Custom fonts: text without Devanagari shaping takes a fast path, and subsetting no longer copies unchanged tables.
  • QR codes generate 2–6× faster, with identical symbols. Canvas QR codes are encoded once, when recorded.
  • Vector canvas and output: coordinates use a fixed-point formatter with byte-identical output, content streams are compressed straight from the page buffer, and large documents compress their pages in parallel.
  • New BenchmarkDotNet suite and throughput harness in the repository's benchmarks/ folder; see Benchmarks.

Fixed

  • Built-in font widths now match the Adobe AFM metrics; 25 entries were wrong, among them the curly quotes and bullet in Helvetica. Characters a standard font cannot show are measured as the ? drawn in their place. Text containing these characters may wrap slightly differently.
  • A truncated or corrupt PNG raises InvalidDataException instead of being embedded as data a PDF viewer cannot render.
  • Image(byte[]) keeps its own copy of the bytes, so reusing the buffer before the document is saved no longer changes the image.

Changed

  • Errors in PNG pixel data are raised when the document is saved (PublishPdf) rather than by Image(...). Header errors, such as an unsupported format, are still raised by Image(...).

Agent packages

  • TerraPDF.Agents 1.0.1 and TerraPDF.Mcp 1.0.1 use TerraPDF 2.4.0, and report a malformed PNG as an error the model can act on instead of failing the tool call.

2.3.0 - 2026-09-26

The vector canvas gains gradient fills, dashes on every stroke, and interactive content: links, bookmarks, and QR codes placed at canvas positions. All additions are backward compatible; the two fixes below change behaviour that did not work as documented.

Added

  • Gradient fills through PathDescriptor.FillLinearGradient(from, to, angle) and FillRadialGradient(center, edge), written as PDF axial and radial shadings clipped to the path. The gradient spans the path's bounding box; the linear angle is measured clockwise from left-to-right.
  • Dash patterns and phases on StrokeEllipse/DrawEllipse, StrokeRoundedRect/DrawRoundedRect, and StrokePie/DrawPie, and on any path through PathDescriptor.Dash(...).
  • PathDescriptor.RoundedRect(...), with the radius clamped to half the shorter side.
  • Canvas links and bookmarks: VectorCanvas.Link (URI), InternalLink (page, optional scroll position), and Bookmark (outline entry for the current page, nestable by parent title).
  • Canvas QR codes: VectorCanvas.QrCode draws one filled vector path per symbol, with an optional background and configurable quiet zone. Oversized data fails at the call rather than at render time.
  • New sample: 19_canvas_extras_showcase.pdf, covering dashed shapes and paths, gradients, links, bookmarks, and QR codes.

Fixed

  • VectorCanvas.Grid() drew nothing, because the canvas callback runs before layout. The grid is now sized when the canvas is drawn and keeps its place in the draw order. Grid rejects a non-positive cellHeight or lineWidth.
  • ShowIf(false) did not hide anything: the element chained after it replaced the empty placeholder. Everything chained after ShowIf(false) is now discarded, including headings, which no longer reach the table of contents.

2.2.0 - 2026-09-20

The vector canvas gains media and geometry: positioned images with five fit modes, dashed strokes, elliptical arcs and pie sectors, and rotated text labels. All additions are backward compatible — no public API was removed or changed.

Added

  • Positioned images on VectorCanvas from file paths, byte arrays, and streams, with Stretch, Contain, centred or top-left Cover, and natural-size CropTopLeft modes. Image clipping is isolated with q/Q, so a shape drawn after a clipped image is unaffected.
  • Rotated canvas text — an optional clockwise angle around the text baseline point, for standard-14 and registered custom fonts alike.
  • Elliptical arcs and closed sectors through PathDescriptor.Arc and Sector (centre + radii), plus the FillPie, StrokePie, and DrawPie canvas conveniences (bounding box). Angles start at 3 o'clock and increase clockwise; arcs split into cubic Bézier segments of at most 90°.
  • Native dash patterns and phases for canvas lines and stroked rectangles. Each dashed command restores the graphics state, so dash state never leaks into a later stroke.
  • New sample: 18_canvas_media_showcase.pdf — four pages covering every fit mode, PNG soft-mask transparency, constant-alpha layering, image sources and natural sizing, dash phases, pie sectors, elliptical arcs, and rotated text.

Fixed

  • PNG decoding now supports 8-bit grayscale images with alpha (colour type 4).
  • Elliptical arcs and pies reject non-finite angles instead of subdividing forever; a slice of a zero total (360 * value / total) previously hung.
  • Rotated canvas text emits full-precision text-matrix coefficients. Small angles were previously distorted (0.3° rendered as 0.57°) or dropped entirely below ~0.3°, and matrix rounding rescaled glyphs by up to ~0.5%.
  • Canvas dash phases must be nonnegative, as the PDF specification requires.
  • Canvas images with a zero pixel dimension are skipped instead of writing NaN operands into the content stream.
  • VectorCanvas.GetImageSizeInPoints throws the documented ArgumentException for unsupported data instead of NotSupportedException, matching VectorCanvas.Image(byte[], …).

Changed

  • A canvas image is decoded once and reused on every page the canvas is drawn on, rather than re-decoded per page.
  • Updated Microsoft.SourceLink.GitHub to 10.0.401 to remove the vulnerable transitive Microsoft.Build.Tasks.Git 8.0.0 dependency.

2.1.0 - 2026-09-03

Three additions, all backward compatible: font embedding now subsets automatically, the vector canvas gained real translucency, and the vector canvas can place text. No public API was removed or changed — existing calls keep behaving exactly as before.

Added (automatic font subsetting)

  • FontFamily.Register(...) now embeds only the glyphs a document actually shows, instead of the whole font file. Every glyph a document never draws is blanked out of the font's glyf table — glyph IDs are never renumbered, so cmap, hmtx, GSUB, and Identity-H/CIDToGIDMap all stay valid with no other change. Fully automatic; no new API.
  • Composite-glyph closure. Accented Latin letters and Devanagari conjuncts are commonly composite glyphs — built from component glyphs no content stream ever references by ID directly. Blanking computes the closure of shown glyphs under their composite references first, so an accent or conjunct never loses a component that happens to be otherwise unused.
  • Measured, not assumed: a custom-font sample with mixed Latin/Cyrillic/Greek text dropped from 718KB to 320KB (55% smaller); a Devanagari report using two font variants (regular + bold) dropped from 152KB to 77KB (49% smaller). Tables sized per glyph regardless of usage (hmtx, loca, cmap, GSUB/GPOS, post, name) are not yet trimmed — see "Known limitations."

Added (graphics state / constant alpha)

  • /ExtGState and the gs operator — the first transparency mechanism in the writer beyond per-pixel image /SMask. Distinct opacity values used anywhere in a document are deduplicated into shared /ExtGState resources, the same way repeated images and fonts already are.
  • Every VectorCanvas fill/stroke primitive (Line, FillRect/StrokeRect/ DrawRect, the rounded-rectangle and ellipse/circle families, Path) takes a trailing opacity parameter (1 = fully opaque, the default — omitting it costs nothing, no /ExtGState is emitted at all). PathDescriptor gained a matching .Opacity(...) fluent setter.

Added (text on the vector canvas)

  • VectorCanvas.Text(text, x, y, ...) places one line of text with its baseline at (x, y) — the one canvas primitive that isn't top-left anchored. Renders through a registered custom font when fontFamily names one, otherwise the standard-14 families. Supports opacity like every other primitive (ghosted/watermark-style canvas text).
  • VectorCanvas.MeasureTextWidth(...) (static) — measures a label in the same font Text would render it in, for centering or right-aligning before placing it.

Fixed (samples)

  • 10_VectorGraphicsShowcase.cs: four shapes were repositioned to stop bleeding past the page margin, and the donut chart's doubled-up legend (bare canvas swatches plus a separate, awkwardly wrapped text list) was rebuilt as a single swatch-plus-label pass using the new VectorCanvas.Text.

Known limitations

  • Font subsetting does not renumber glyph IDs or shrink tables sized per glyph regardless of usage, so the size win on a full font is well short of what a 99%+ shrink in glyf alone would suggest. Full re-indexed subsetting may follow in a future version.
  • /ExtGState opacity is wired through VectorCanvas and canvas text only; DrawImage, flowed text (TextBlock), and Background()/border colours do not yet take an opacity parameter.

2.0.1 - 2026-08-28

A table-correctness release. Every fix below addresses a case that produced a valid PDF with visibly wrong geometry — cells drawn on top of one another, or rows running off the bottom of the page — rather than an error. No public API changed; documents that do not use table spans and do not overflow a page are byte-for-byte identical to 2.0.0.

Added

  • New sample: 16_table_spans_showcase.pdf — a seven-page document covering every fix below. Pages 1-2 demonstrate columnSpan, rowSpan, the two together, and a spanned cell growing the rows it covers, each next to the code that produced it. Pages 3-5 are a header-less ledger whose accounts are joined by row spans, split across three pages to show the spans surviving the breaks intact. Pages 6-7 are a table placed straight into the content slot with no Column wrapper, paginating with its grouped two-column header repeated on both pages.

Fixed (table column and row spans)

  • Cell(columnSpan:) no longer overlaps the cell that follows it. The row cursor advanced by one column regardless of the span just placed, so in a three-column table a columnSpan: 2 cell followed by a normal cell put that cell in column 2 — inside the span — and left column 3 empty. Cells are now placed in the first column not already covered by an earlier cell.
  • Cell(rowSpan:) now reserves its columns in the rows below it. Each row started its cursor at column 1 with no record of what the previous rows had spanned, so the row after a rowSpan: 2 cell drew its first cell at the same origin, on top of the spanned cell.
  • Spanned cells are measured and grow the rows they cover. GetRowHeights skipped every cell with RowSpan > 1, so a spanned cell contributed no height: content taller than the rows it covered overflowed past the table, and a row containing only spanned cells collapsed to zero height. Row heights are now computed in two passes, the second distributing any shortfall evenly across the rows a spanned cell covers.
  • Spans below 1 are treated as 1 rather than producing a zero-width or zero-height cell.

All three defects produced valid PDFs that silently drew cells on top of each other, and none of them were covered by the test suite. New TableSpanTests asserts on the rectangles actually emitted into the content stream — placement, width, height, and pairwise non-overlap — including a grouped header span repeated across a multi-page table.

Fixed (table pagination)

  • A table with no header row now splits across pages. Only tables declaring a HeaderRow were split between rows; a header-less table taller than the page was placed whole and its overflowing rows simply ran off the bottom. Header-less tables are now split when — and only when — they cannot fit a page, so a short one is still drawn as an ordinary item and the decorators wrapped around it (background, border) keep painting.
  • A table placed directly in the content slot now paginates. The layout pass looked for a top-level Column and sent anything else down a single-page path, so page.Content().Table(…) overflowed instead of splitting. Such a table is now wrapped in a synthetic one-item column and takes the same row-splitting path. Content that fits on one page is unaffected and still receives the whole content box.
  • A row span is no longer cut in half at a page break. The splitter moved one row at a time, so a break landing inside a rowSpan left a truncated cell on the first page with nothing continuing it overleaf. Data rows are now grouped into the smallest runs that no row span crosses, and a group is never divided between pages. A group taller than a whole page is still forced out so layout always makes progress.

TablePaginationTests covers each case, including the row-span break at seven different page offsets — the defect only appeared at offsets where the break happened to fall inside a spanned pair.

Known limitations

  • A row span that starts in a header row and extends into data rows renders correctly on the first page, but is truncated to the header rows on every continuation page: the header block repeats while the data rows it also covers stay behind on the page before. Row spans that start in a data row are unaffected — those are grouped and never split. Keep header rows self-contained if a table is expected to paginate.

2.0.0 - 2026-07-23

Added

  • Custom font embedding — FontFamily.Register(...) embeds a TrueType font (regular, bold, italic, bold-italic) as a Type0/CIDFontType2 composite font with Identity-H encoding, giving full Unicode text support beyond the standard-14 fonts' WinAnsiEncoding range (brand typefaces, Cyrillic, Greek, and beyond). Fonts are embedded once per document regardless of how many times they're used, and requesting an unregistered style falls back to the closest registered variant instead of throwing.
  • Devanagari-aware rendering (pure C#, no native dependency, none planned):
    • Matra reordering — the vowel sign ि (U+093F) is moved to its correct pre-consonant visual position.
    • Conjunct ligatures — the font's own half/akhn/cjct GSUB features are substituted so स्व, स्थ, क्ष, ज्ञ etc. render as proper joined forms.
    • Reph — cluster-initial र् (धर्म, वर्तमान, दुर्बलता) is reordered and substituted via the font's rphf feature.
    • Below-base/post-base 'ra' — क्र, त्र, प्र, ष्ट्र substituted via the font's rkrf feature.
  • New docs/custom-fonts.md guide.
  • New sample: 15_child_nutrition_india_report.pdf — a full multi-page Hindi-language report exercising the new Devanagari shaping corrections.

Fixed

  • Word-wrap: a single word/token wider than the line now breaks at character boundaries instead of overflowing the container.

Changed

  • Breaking (none): version bumped 1.5.1 → 2.0.0 to reflect the scope of the new font subsystem, not a breaking API change.
  • CI (ci.yml): added a nuget.org connectivity canary as the last step of every run.
  • Publish workflow (publish.yml): added pre-publish connectivity and NUGET_API_KEY authentication checks, so a dead network path or a stale key is caught before Pack/Push — previously such a failure could only be recovered by bumping the version, since nuget.org rejects re-pushing the same version.

1.5.1 - 2026-07-10

Added

  • .NET 10 target — the library now multi-targets net8.0, net9.0 and net10.0 (the current LTS). No API or behaviour changes; existing .NET 8/9 consumers are unaffected.

Changed

  • Test suite now runs once per supported runtime (net8.0, net9.0, net10.0); the sample app moved to net10.0.
  • CI workflows install the .NET 8/9/10 SDKs; global.json now requires the .NET 10 SDK (with rollForward: latestMajor).
  • Consolidated the two overlapping CI workflows into a single ci.yml.
  • Migrated the solution to the XML-based .slnx format (TerraPDF.slnx).

1.5.0 - 2026-07-06

Added (barcodes & QR codes)

  • container.Barcode(data, ...) — Code128 (Subset B) barcode generation, encoding printable ASCII (0x20-0x7E). Supports an explicit or auto-fill width, custom module/background colour, an optional human-readable caption rendered below the bars, and a configurable quiet zone.
  • container.QrCode(data, ...) — from-scratch ISO/IEC 18004 QR code generator: byte-mode encoding, automatic version selection (1-40), all four error correction levels (QrErrorCorrectionLevel.L/M/Q/H), Reed-Solomon error correction, and full mask-pattern penalty scoring. Supports an explicit or auto-fill size, custom module/background colour, and a configurable quiet zone.
  • Both render as vector-filled rectangles (one rect per bar / per contiguous run of dark QR modules), matching the existing VectorCanvas rendering style — crisp at any zoom, no raster image pipeline, and placeable anywhere an IContainer is exposed (Column, Row, Table cell, header, footer).
  • New PdfPage.AddFilledRects batch primitive: emits one colour operator followed by many re ops and a single trailing fill, avoiding a redundant colour-set/fill pair per module on symbols with thousands of modules.

1.4.0 - 2026-07-04

Added

  • AES-256 PDF encryption by default (Standard Security Handler Revision 6, SHA-2 key derivation, PDF 2.0 output). EncryptionOptions.Algorithm lets you opt back into AES-128 (EncryptionAlgorithm.Aes128) for legacy viewers.
  • Images from bytes and streams — container.Image(byte[]) and container.Image(Stream) overloads (with optional width), for images from databases, embedded resources, or generated data. Format is now detected from magic bytes rather than the file extension.
  • PNG transparency — RGBA PNGs keep their alpha channel via a /SMask soft mask; indexed-transparency (tRNS) PNGs still render opaque.
  • Image deduplication — identical image data reused across pages is embedded once and shared document-wide.
  • Anchor-based bookmarks — container.Bookmark("Title"[, parentTitle]) marks its content as an outline destination; the page number and position are resolved automatically at render time. The page-number-based Bookmark(title, pageNumber) API remains available.
  • Paragraphs split across pages — a text block taller than the remaining page now flows onto the next page instead of overflowing.
  • FontFamily(string) on TextDescriptor, SpanDescriptor, and TextStyle now actually works — supports Helvetica, Times, and Courier (plus common aliases like "Arial"), with Bold()/Italic() staying within the selected family.

Fixed

  • Bookmark destinations now use zoom-retaining /XYZ coordinates instead of /Fit//FitH, and land at the correct position.
  • Height-constrained images preserve aspect ratio instead of being squashed.
  • Table of Contents heading scan now traverses decorators, hyperlinks, and bookmark anchors.
  • Encrypted documents no longer leak metadata, bookmark titles, or hyperlink URIs in plaintext — everything is encrypted with the owning object's key.
  • Document metadata is now referenced from the PDF trailer, so viewers display it correctly.
  • Pagination now works through all decorators (Margin, RoundedBorder, RoundedBox, per-edge borders).
  • Negative-value validation added to the single-side Padding*/Margin* overloads.

Changed

  • Breaking: TextDescriptor.Span(string, Action<TextStyle>) is now Span(string, Func<TextStyle, TextStyle>). TextStyle is immutable, so the callback must return the configured style: t.Span("hi", s => s.Bold()).
  • Content streams are now Flate-compressed, and serialization is streamed instead of buffered in memory — smaller files, lower peak memory.
  • Fragment-based layout engine — page counting and rendering now always agree.

1.3.0 - 2026-05-19

Added

  • AES-128 PDF Encryption using the PDF Standard Security Handler.
  • EncryptionOptions with UserPassword, OwnerPassword, and Permissions.
  • PdfPermissions flags for Print, CopyText, ModifyContents, ModifyAnnotations, FillForms, ExtractForAccessibility, AssembleDocument, PrintLowResolution, All, and None.
  • Per-object AES-128 CBC encryption of content streams and image XObjects.
  • Encrypted documents are emitted as PDF 1.6, the minimum version required for AES encryption.
  • Encryption tests covering password combinations, permission flags, multi-page documents, metadata, null guards, and output-size sanity.

Fixed

  • Encrypted PDFs now write the random /ID array to the trailer so viewers can reproduce the file encryption key.
  • Removed invalid /Filter /Crypt entries from content streams and JPEG image dictionaries.
  • AES encryption now uses the correct padding behavior for decrypted content.
  • Sample output folders are created automatically before sample PDFs are written.
  • Encryption showcase badges use WinAnsi-safe ASCII symbols.

1.2.3

Added

  • Vector Graphics / Canvas API via container.Canvas(height, draw).
  • VectorCanvas primitives for lines, rectangles, rounded rectangles, circles, ellipses, arbitrary Bezier paths, polygons, and grids.
  • PathDescriptor helpers for MoveTo, LineTo, CurveTo, Close, Rect, Ellipse, Circle, Polyline, Polygon, Fill, Stroke, and even-odd fill.
  • Vector graphics sample demonstrating primitives, custom paths, charts, progress bars, callouts, and icon-grid patterns.
  • Unicode and WinAnsiEncoding showcase sample covering Windows-1252 specials, Latin-1 characters, full byte-to-glyph reference grids, font comparison, and glyph metrics.
  • Documentation guides for vector graphics and Unicode / character encoding.

Fixed

  • Several language sample sentences now avoid characters outside WinAnsiEncoding so output PDFs do not show replacement ? glyphs.
  • Win-1252 showcase tables now use proportional column definitions to avoid page overflow.

1.2.2 - 2026-05-04

Added

  • Table of Contents generation — container.TableOfContents() creates a TOC page populated with headings collected from .H1()–.H6(), with hierarchical numbering (e.g. 1, 1.1, 1.1.1) and clickable internal links.
  • Internal links (GoTo) — container.InternalLink(pageNumber, top?) creates intra-document navigation that preserves current zoom level and scrolls to the target heading.
  • Section headings — .H1() through .H6() methods with sensible default styles (size + weight), each returning TextDescriptor for further customisation.

Fixed

  • Internal link zoom issue — /FitH replaced with /XYZ so clicking TOC entries no longer resets zoom.
  • Page number display in TOC now excludes TOC page count (TOC treated as page zero), while links still point to correct physical pages.
  • HeadingRecorder propagation through DrawingContext.At fixed so TOC entries are correctly collected.

1.2.1 - 2026-05-03

Documentation

  • Corrected all GeneratePdf() → PublishPdf() method references throughout README and all documentation files
  • Fixed Color.Blue hex values: Darken2 (#1976D2), added missing Darken3 (#1565C0) and Darken4 (#0D47A1)
  • Added PageBreak() to ColumnDescriptor API table in layout guide
  • Documented HeaderOnFirstPageOnly() page method for first-page-only headers
  • Clarified RelativeItem() default weight = 1 in row-and-column layout guide
  • Added FontFamily(string) to TextDescriptor methods table in text-and-spans guide

1.2.0 - 2025-07-14

Added

  • Underline() style method on TextDescriptor and SpanDescriptor. Draws an underline beneath the text. Can be combined with Strikethrough().
  • LineHeight(double) style method on TextDescriptor. Sets a line-height multiplier (e.g. 1.0 = tight, 1.4 = default, 2.0 = double-spaced). Also accepted by DefaultTextStyle for page-wide control. TextStyle.LineHeightMultiplier property exposed for custom render logic.
  • RoundedBorder(radius, lineWidth, hexColor) decorator — draws a rounded-corner stroke border around child content. Corner radius is automatically clamped to half the shorter dimension.
  • RoundedBox(radius, fillHexColor, borderHexColor, lineWidth) decorator — fills the area with fillHexColor and draws a rounded-corner border in one call. Equivalent to Background + RoundedBorder but rendered as a single path.
  • BorderTop(lineWidth, hexColor), BorderBottom, BorderLeft, BorderRight per-edge border decorators. Each side is independently configurable with its own width and colour. The hexColor parameter defaults to "#000000".
  • PageBreak() container extension — inserts an explicit page-break marker inside a Column. Silently skipped when it falls at the very start of a page.
  • Hyperlink(url) container extension — wraps child content in a clickable PDF URI annotation (/Annots with /URI action). Clicking the area in a conforming PDF viewer navigates to the given URL.
  • HighPriorityFeatureTests.cs — 20 new tests covering all five features above (underline, hyperlink, per-edge borders, line height, and their combinations).
  • RoundedBorderTests.cs — dedicated tests for RoundedBorder and RoundedBox geometry, clamping behaviour, and validation.
  • PageBreakTests.cs — tests for explicit page-break positioning.
  • HeaderFirstPageOnlyTests.cs — tests verifying that the header slot can be conditionally rendered only on the first page using ShowIf.

Fixed

  • Breaking (behaviour): TextDescriptor.Span().Bold() / .Italic() / .Strikethrough() etc. previously mutated the whole block's SpanStyle instead of the individual span's style. Now correctly isolated.

Changed

  • Folder Fluent renamed to Core (TerraPDF.Core namespace).
  • Folder Infrastructure renamed to Infra (TerraPDF.Infra namespace).

1.1.0 - 2025-06-01

Added

  • Margin, MarginVertical, MarginHorizontal, MarginTop, MarginBottom, MarginLeft, MarginRight decorator methods with full Unit overloads. Margin is outer spacing — the margin region stays transparent, background and border start after the gap.
  • SpanDescriptor — per-span fluent builder returned by TextDescriptor.Span(), CurrentPageNumber(), and TotalPages(). Style methods chained after .Span(...) now apply only to that span, not the whole TextBlock.
  • Complete documentation suite under docs/: getting-started.md, text-and-spans.md, layout.md, decorators.md, images.md, colors.md, page-sizes-and-units.md, components-and-templates.md, row-and-column-layout.md.
  • CHANGELOG.md, CONTRIBUTING.md, SECURITY.md.
  • Input validation (using ArgumentNullException.ThrowIfNull, ArgumentException.ThrowIfNullOrWhiteSpace, ArgumentOutOfRangeException.ThrowIfNegative/ThrowIfNegativeOrZero) on every public API entry point.
  • 62 new tests across ValidationTests and BehaviourTests (92 total).

Fixed

  • Breaking (behaviour): TextDescriptor.Span().Bold() / .Italic() / .Strikethrough() etc. previously mutated the whole block's SpanStyle instead of the individual span's style. Now correctly isolated.

Changed

  • Folder Fluent renamed to Core (TerraPDF.Core namespace).
  • Folder Infrastructure renamed to Infra (TerraPDF.Infra namespace).

1.0.0 - 2025-01-01

Added

  • Initial release.
  • PDF 1.7 generation with zero native dependencies.
  • Text styling: bold, italic, bold-italic, strikethrough, underline, font size, colour.
  • Multi-span text blocks with mixed styles.
  • Column (vertical stacking) and Row (horizontal layout) with RelativeItem, AutoItem, and ConstantItem sizing.
  • Table with relative and constant columns, HeaderRow (repeats on continuation pages), alternating-row support.
  • Padding with all side variants and Unit overloads.
  • Background, Border, Alignment (horizontal + vertical), ShowIf.
  • Horizontal and vertical rule lines.
  • PNG and JPEG image embedding.
  • IComponent reusable content blocks.
  • IDocument reusable document templates.
  • Headers, footers, page numbers (CurrentPageNumber, TotalPages).
  • Multi-page documents with automatic table pagination.
  • Full Material Design colour palette (Color.*).
  • Standard page sizes including ISO A-series, Letter, Legal, Tabloid, Executive, and Landscape() helper.
  • Unit system: Point, Millimetre, Centimetre, Inch.
  • Targets .NET 8 and .NET 9.
  • CI workflow (GitHub Actions): build, test, coverage.
  • Publish workflow (GitHub Actions): NuGet + symbols on release tag.
← README Contributing →