• The parser.docx plugin replaces the previous decoder.docx plugin.

• DOCX files from different Microsoft Office versions and third-party applications are interpreted more completely, including ISO Strict documents and alternative Office content.
• Footnotes and endnotes are included in page layout, with support for continuation pages, numbering restarts, section placement, and references inside tables.
• Documents with right-to-left, East Asian, and Southeast Asian text benefit from improved font selection, text ordering, alignment, line breaking, and script-specific formatting.
• Automatic hyphenation and discretionary hyphens provide more natural line breaks while respecting the document language and paragraph settings.
• Additional section settings are supported, including vertical alignment, text direction, column separators, line numbering, document grids, mirrored margins, and page-border placement.
• Page counts, chapter-prefixed page numbers, running headings, and document-property fields are calculated from the final document layout.
• Complex tables support more Word formatting, including inherited styles, conditional formatting, automatic and fixed widths, merged cells, repeating headers, nested tables, and multi-page rows.
• Pictures and modern or legacy Microsoft Word drawings retain more of their original crops, colors, effects, grouping, text, positioning, and wrapping.
• Large or complex documents are processed more reliably. Invalid content is isolated where possible, and a cancelled or failed operation does not publish an incomplete result.

Fixed Bugs

• DOCX documents use fonts from the configured font directories, including their actual bold and italic faces. Missing fonts use document font hints and improved fallback preferences for consistent text and space widths.
• Corrected exact and minimum paragraph line spacing to match Microsoft Word.
• Improved CJK line breaking and justification to avoid excessive gaps at text run boundaries.
• Corrected fallback font metrics for CJK text to prevent compressed or overlapping lines.
• Preserved vertical VML text when it is split across multiple text fragments.
• Malformed VML text-path font declarations could cause excessive parsing times because of regular-expression backtracking.
• Nested tables could abort when moved to the next page, and words split after a tab near the right margin could start on the wrong line.
• Documents with merged table continuations, page relayout around anchored drawings, or tab stops beyond the text area could fail to finish parsing.
• Text without a font declaration could use an unrelated system font, changing spacing and table breaks.
• Missing Helvetica, Helvetica-Bold, Madame Cosmetics, Aparajita, and AdvLTe50259 fonts could use unsuitable condensed replacements.
• Nested DrawingML shape groups could lose child offsets or use incorrect scaling, reflection, and rotation.
• Stale first-row cell widths could shrink an otherwise coherent AutoFit table grid.
• Anchor-only paragraphs with minimum line spacing could add extra vertical space, and text following a full-width square-wrapped anchor could jump above it.
• Body anchors inside tables could paint in the wrong order relative to other page anchors.
• Word 2010 text outlines and shadows could lose their colors, transparency, geometry, and inherited formatting.
• Downward-arch WordArt could remain flat, including text in rotated shapes and text with outline or shadow effects.
• Direct vertical cell margins could be omitted from minimum row heights, and contextual spacing between same-style paragraphs could be ignored inside table cells.
• DrawingML flowchart delay symbols could appear as rectangles.
• Some documents could lose individual characters at automatic line breaks, show unwanted spaces in fields, or ignore spaces that were deliberately preserved.
• Paragraph borders could be missing, doubled, or repeated incorrectly across pages, columns, and table cells; paragraph shading and framed paragraphs could also differ from Microsoft Word.
• Complex paragraph spacing, indents, tab stops, manual breaks, and keep-together settings could place text on the wrong line, column, or page.
• Hidden text could remain visible, while small capitals, highlights, double strikethrough, shaded spaces, and underlined trailing spaces could be displayed incorrectly.
• Adjacent text with different font details could be combined and displayed with the wrong font or character spacing.
• Documents that combine Latin, Asian, and complex-script fonts could select the wrong font face or use an incorrect line height; regular text could occasionally appear bold.
• Right-to-left and Asian text could show incorrect character order, punctuation placement, tab alignment, list-marker alignment, or line breaks across formatted text.
• Soft hyphens and automatic hyphenation could disappear, become visible without a break, or choose unsuitable line-break positions in paragraphs and tables.
• Missing, cyclic, or incompatible paragraph, character, numbering, and table styles could discard valid formatting or stop the document from rendering.
• List numbering could start or restart incorrectly, display invalid placeholders, use the wrong format or indentation, or fail for large and incomplete numbering definitions.
• Table widths, column grids, AutoFit calculation, floating-table placement, negative indents, and intentionally wide tables could differ significantly from Microsoft Word.
• Table styles could lose inherited or conditional formatting for text, alignment, spacing, cell margins, borders, and preferred cell widths.
• Merged cells, exact or minimum row heights, non-splitting rows, repeating headers, and table continuations could produce missing content, incorrect borders, or wrong page breaks.
• Nested tables, hyperlinks, content controls, SmartTags, and tracked changes inside table cells could lose visible content or appear in the wrong order.
• Anchored pictures and shapes could overlap text, move to the wrong page, use the wrong reference position, or disturb alignment and tab stops in surrounding text.
• Odd- and even-page section starts, continuous sections, mirrored layout, page borders, columns, headers, and footers could use the wrong page or position.
• Page counts, page-number restarts, running headings, dates, and other fields could show stale values, unwanted spaces, or incorrect results in headers, footers, and text boxes.
• Footnotes and endnotes could remain on the wrong page, be repeated, reserve too much space, or be separated from references that moved during pagination.
• Pictures could lose crops, rotations, reflections, borders, colors, or transparency effects, and some embedded images could fail to load.
• Grouped shapes, text boxes, WordArt, gradients, image-based outlines, horizontal rules, and other legacy drawings could be positioned, scaled, colored, or wrapped incorrectly.
• Invalid page-number values, malformed formatting, damaged embedded fonts, or missing optional document parts could produce wrong output, repeated layout attempts, or abort the complete document.
• In some affected documents, tables could be shifted from their Microsoft Word position and individual images could be omitted.

Fixed Bugs

• Reading color values from DOCX produced an error.
• Improved table width handling for exported DOCX files with mixed tblW/tcW definitions, including percent and fixed-width edge cases.
• Fixed multiple DOCX table width regressions where table and cell widths were interpreted differently than Microsoft Word in specific exported layouts.
• Added missing line breaks.
• Fixed several parser errors.
• Corrected exact and minimum paragraph line spacing to match Microsoft Word.
• Improved CJK line breaking and justification to avoid excessive gaps at text run boundaries.
• Corrected fallback font metrics for CJK text to prevent compressed or overlapping lines.
• Preserved vertical VML text when it is split across multiple text fragments.

Fixed Bugs

• Reading color values from DOCX produced an error.
• Improved table width handling for exported DOCX files with mixed tblW/tcW definitions, including percent and fixed-width edge cases.
• Fixed multiple DOCX table width regressions where table and cell widths were interpreted differently than Microsoft Word in specific exported layouts.
• Added missing line breaks.
• Fixed several parser errors.