Release Notes

APDFL v21.0.0PlusP1c (July 19, 2026)

New Features:

• Added support for OCR with Vietnamese text.
  Optical Character Recognition (OCR) now supports recognition of Vietnamese-language text, including Vietnamese-specific characters and diacritical marks. This enhancement improves text extraction accuracy and searchability for scanned documents containing Vietnamese content.

Changes:

• Forms Extension installers now support the Modern C++ interface.
• The XPS2PDF plugin now reports a specific error when the joboptions file cannot be found.
• Corrected installer documentation that still referenced APDFL 18.
• With this release, the Maven package version has been increased to 21.0.1.
• With this release, the NuGet package version has been increased to 21.0.1.

Open Source Updates:

• Boost 1.91.0
• cURL 8.21.0
• Expat 2.8.2 (addresses CVE-2026-45186)
• libTIFF 4.7.2
• OpenSSL 3.6.3

Problem Corrections:

• SF45180 – Improves flattening of XFA documents where a Button field has no Caption but a Text Label overlaid on top of it.
• SF47401 – Corrects an issue where content could be clipped unexpectedly when flattening an XFA document with Forms Extension.
• SF47460 – Corrects an issue where some XFA documents, when flattened AsIfPrinted, showed content that was meant to be excluded.
• SF47461 – Corrects an issue where a heavily crowded Text Field shared across pages of an XFA document could be slightly clipped when flattened with Forms Extension.
• SF47486 – Improves flattening of XFA documents where a Button field has no Caption but a Text Label overlaid on top of it.
• SF47555 – Corrects a problem where System Font selection could choose a similarly named but different style variant in some cases.
• SF47562 – Corrects an issue preserving the grouping symbol in a decimal value for a numericEdit field when flattening an XFA document.
• SF47581 – Corrects a problem where rendering a document with minor Page Content errors raised an error when it previously did not.
• SF47585 – Corrects a small memory leak that could occur when subsetting fonts.
• SF47591 – Corrects an issue where a PDF's Producer property still reported APDFL v18; it now correctly reports v21.
• SF47597 – This corrects a problem during Color Conversion where the inherited GraphicState would be changed unexpectedly in some cases.
• SF47599 – Corrects an issue where rendering to a CMYK-based space would not preserve K-only content as pure black.
• SF47603 – This corrects an issue with the Color Conversion option mPromoteGrayToCMYK when enabled, which would result in unexpected Gray values.
• SF47610 – This corrects a problem where Color Converting a Page could unexpectedly affect a shared Form on another page.

APDFL v21.0.0P1b (May 22, 2026)

Datalogics is pleased to announce the release of APDFL v21.0.0Plus1b. This release is a major upgrade built on the Adobe PDF Library (PDFL) v21.0 core libraries and includes a wide range of new features, performance improvements, compatibility enhancements, and problem corrections.

New Features:

• Modern C++ SDK — Built on the same APDFL runtime as the existing integration libraries, the Modern C++ SDK provides a C++17-based interface featuring RAII-based resource management, exception-driven error handling, move semantics, std::unique_ptr / std::shared_ptr ownership models, strongly typed flags, and a unified datalogics_interface namespace. These enhancements reduce boilerplate code, simplify resource management, and enable more natural integration of PDF functionality into modern C++ applications.
  The Modern C++ SDK requires a C++17 compiler and is supported on Windows (MSVC 19.3+ / Visual Studio 2022 v143 toolset, x64 and ARM64), Linux Rocky 9+ (Clang 18+ with libstdc++11, x86_64 and aarch64), and macOS 15+ (Apple Clang 16.4+ / Xcode 16+, x86_64 and arm64).
  The existing APDFL C++ interface will now be referred to as "Adobe C/C++" and remains fully supported. Existing integrations can continue unchanged, and migration to the Modern C++ SDK is optional.
• WebToPDF — A new Plug-In for APDFL that converts an HTML file tree on disk, or a URL, to PDF. Available on Windows 64, Windows ARM, Linux x64, and Linux arm64.
• Added support for the Chinese GB18030 font — Enables correct rendering and processing of Simplified and Traditional Chinese characters in documents that use GB18030-compliant fonts. This improves compatibility with Chinese-language PDFs and helps ensure accurate display of multilingual text content.
• New ListStructureTree Sample (Modern C++) — This sample walks the logical structure tree of a tagged PDF document and prints each element's struct type (/S), parent chain, and kid entries (child elements and marked-content references). Available for the Modern C++ interface.

Problem Corrections:

• SF44603 – Word Finder: Fixed incorrect positioning of Arabic diacritical marks (harakat) in extracted text. In certain PDFs, combining marks such as Damma (U+064F) could be attached to the wrong base character when the diacritic was spatially positioned between two Arabic glyphs with similar overlap.
• SF46074 – Resolved a problem with WordFinder processing Thai diacritics.
• SF46900 – Corrects an issue where updating Page Content could cause text contained within Marked Content to unexpectedly be located outside of the Marked Content.
• SF47417 – Corrects an issue where updating Page Content could cause text contained within Marked Content to unexpectedly be located outside of the Marked Content.
• SF47421 – Corrects a problem saving a DeviceN rendered image to a TIFF where the TIFFTAG_PHOTOSHOP tag data would be incorrect.
• SF47454 – Fixes a Stack Overflow that could occur setting Page Content during PDF/A conversion for Property Lists that are improperly defined.
• SF47481 – Rendering has been improved to be more resilient against problematic content that would previously raise an error. Some error strings that were meant to have a string format specifier used in combination with the non-fatal error argument were missing and have been added.
• SF47530 – Corrects a problem where converting a document to PDF/A could misinterpret the CTM for an ImageMask in some cases.