v1.4.0

OpenccNetLibGui v1.4.0

Changelog:

[1.4.0] – 2026-01-07

Added

• Advanced Short Heading Settings dialog

  • Configurable maximum heading length (range 3–30, default 8).
  • Fine-grained pattern controls:
    • All CJK characters
    • All ASCII characters
    • ASCII digits only (automatically enabled when ASCII is selected)
    • Mixed CJK + ASCII
  • Clear visual hierarchy with parent/child options, inspired by Visual Studio feature selection.

• User-configurable short heading detection

  • Introduces an advanced, regex-based override mechanism for heading detection.
  • Custom title patterns are evaluated immediately after built-in title detection
    and before other reflow heuristics.
  • Enables precise identification of structured book headings, including:
    • Front-matter titles (e.g. 序章, 前言, 楔子)
    • Chapter banners (e.g. 第十二章 夜雨初歇)
    • Decorated or stylized headings commonly found in novels and scanned PDFs
  • Supports unlimited pattern combinations via regular expressions
    (e.g. xxx|yyy|zzz), allowing full adaptation to diverse content styles.

• Design-time preview support

  • Added Design.DataContext for the Short Heading dialog,
    improving layout iteration and visual consistency during development.

• Ignore untrusted PDF text (Pdfium)

  • Added an option to skip repeated overlay- or annotation-like text during PDF extraction.
  • Uses object-level text extraction to reduce duplicated or non-content text in certain PDFs.
  • Intended as a rescue option for PDFs with visible duplicated headings, watermarks, or overlay noise.
  • Extraction-only filtering; does not modify the original PDF.
  • Available via the PDF context menu and persisted under pdfOptions.ignoreUntrustedPdfText.

• DOCX (.docx) plain-text import support

  • Extracts human-readable text from Microsoft Word documents into the source editor.
  • Handles paragraphs, numbered and bulleted lists, tables (flattened as TSV),
    headers/footers, footnotes, and comments.
  • Formatting is intentionally stripped to produce clean, editable plain text
    suitable for reflow processing and OpenCC conversion.

• ODT (.odt) plain-text import support

  • Extracts text from OpenDocument Text files via content.xml.
  • Supports paragraphs, headings, lists, and tables.
  • Designed for lightweight, predictable text editing in the source editor.

• EPUB (.epub) plain-text import support

  • Extracts human-readable text from EPUB eBooks by parsing the package manifest (OPF) and spine order.
  • Supports both XHTML (.xhtml) and legacy HTML (.html / .htm) chapters, including older Calibre-generated EPUBs.
  • Ignores CSS and presentation-only markup; text is extracted based on semantic structure
    (paragraphs, headings, block elements, and line breaks).
  • Skips non-content sections such as scripts, styles, and navigation documents (ToC) by default.
  • Output is normalized into clean, reflow-friendly plain text suitable for further
    paragraph reflow and OpenCC conversion.

• About dialog

  • Added a dedicated About dialog displaying application version,
    engine information, and project homepage.

Changed

• Reflow engine refactored into ReflowModel
  • Moved all CJK paragraph reflow logic out of PDF helpers.
  • Shared by PdfPig and Pdfium extraction pipelines.
  • Greatly improves maintainability, testability, and reuse across formats.
• Short heading detection upgraded
  • Uses ShortHeadingSettings instead of a single integer value.
  • ASCII-only headings automatically allow a larger effective length
    (maxLen × 2, clamped to 10–30) to better support English headings
    such as Introduction, Chapter One, Black Water, etc.
• PDF reflow heuristics improved
  • Better handling of dialog continuation, punctuation-based joins,
    metadata lines, and mixed CJK/ASCII content.
  • More robust collapse of layout-level repeated titles and headings.
  • Detect drawing box line pattern as paragraph separator.
• Main text import pipeline unified
  • Drag-and-drop and Open File now share the same document-loading logic.
  • DOCX, ODT, and plain text files are consistently routed through the same
    source editor update path.
• Internal architecture cleanup
  • Clear separation between:
    • PDF extraction (PdfPig / Pdfium)
    • Office document parsing (DOCX / ODT)
    • Text reflow logic (ReflowModel)
    • User configuration (LanguageSettings / ShortHeadingSettings)
• Paragraph end detection and reflow logic significantly improved
  • Main-body paragraph splitting now strictly follows standard CJK sentence rules
    (。！？ with proper closer handling), prioritizing correctness over aggressiveness.
  • Ellipsis-based endings (……, OCR "...") are supported as weak paragraph boundaries
    only when the line is predominantly CJK, preventing false splits in English or technical text.
  • Structural lines (e.g. bracket-wrapped titles, book lists, metadata-like lines, dates, signatures)
    are handled separately from sentence punctuation, avoiding interference with normal prose.
  • Robust handling of common OCR artifacts:
    • ASCII . and : may be conditionally interpreted as CJK punctuation
      only in strongly CJK contexts.
  • Dialog continuity is strictly preserved:
    • Paragraph splits are always blocked while quotes or brackets remain unclosed,
      ensuring multi-line dialog stays intact.
  • Overall reflow behavior is now closer to human-edited Chinese text layout,
    especially for novels, essays, and scanned PDFs.
• Update OpenccNetLib to v1.4.0
• Update OpenccNetLibGui runtimes to .Net 10

Notes

• This release focuses on correctness, configurability, and long-term maintainability
  of text reflow and document import.
• DOCX and ODT are treated as input formats only; all content is converted to
  plain text before editing, reflow, or Opencc conversion.
• Existing behavior remains compatible; legacy ShortHeadingMaxLen is internally
  synchronized with the new settings model.
• The reflow engine is now suitable for reuse across
  PDF, Office documents, EPUB, CLI tools, batch processing, and automated testing.

---

Full Changelog: v1.3.2...v1.4.0

---

Requires:

• Microsoft .NET 10 Desktop Runtime