Added in 1.3.0, DocxReader parses an existing DOCX file into a DocxDocument — enabling a full round-trip: read, modify with the normal mutation API, then write back out. It reads files created by Microsoft Word and other DOCX-compatible applications.

class DocxReader {
  const DocxReader();
  DocxDocument read(Uint8List bytes)
}

Read a document and walk its content in order:

final reader = DocxReader();
final document = reader.read(docxBytes);

for (final item in document.content) {
  if (item is DocxParagraph) {
    print(item.plainText);
  }
}

The test suite exercises the round-trip with two one-liners — generate bytes, then read them straight back:

Uint8List generate(DocxDocument doc) => generator.generate(doc);
DocxDocument roundTrip(DocxDocument doc) => reader.read(generate(doc));

Put together, a read → modify → write cycle looks like this (composed from the documented read, addParagraph and generate calls):

// Read an existing DOCX into a document model
final document = DocxReader().read(docxBytes);

// Modify it with the normal mutation API
document.addParagraph(Paragraph.heading('Appendix', level: 1));
document.addParagraph(Paragraph.text('Added after reading.'));

// Write it back out
final updated = DocxGenerator().generate(document);

Parsing failures throw a sealed DocxReaderException, so you can switch on the concrete cause:

sealed class DocxReaderException implements Exception

class InvalidDocxArchiveException extends DocxReaderException  // not a valid ZIP archive
class MissingDocxPartException extends DocxReaderException     // Missing required DOCX part: $partName
class InvalidDocxXmlException extends DocxReaderException      // Malformed XML in DOCX

DocxDocument (alias Document) is the root container. Add content with addParagraph, addParagraphs and addTable; read it back via the content getter, which returns paragraphs and tables in insertion order.

DocxParagraph (alias Paragraph) holds a list of runs plus block-level properties. The named factories — .text, .heading, .subtitle, .caption, .quote, .codeBlock, .footnote, .bulletItem, .dashItem, .numberedItem, .alphaItem, .romanItem and .empty — wrap this constructor. plainText joins the run texts.

DocxRun (alias TextRun) is a span of text with inline formatting. TextRun.lineBreak() is a convenience for a soft return; getters isLink and hasFormatting and a copyWith method round it out.

DocxTable takes rows and table-level borders; columnCount and rowCount getters account for spans. Factories DocxTable.simple and DocxTable.fromHeaders build tables from plain data.

DocxTableRow is a thin wrapper around its cells:

DocxTableCell carries paragraphs, alignment, backgrounds, per-cell borders and merge spans. DocxTableCell.text is a shorthand for a single-string cell, and DocxTableCell.merged() is the continuation placeholder for a rowSpan merge.

Borders share the DocxBorder model (width in eighths of a point), and the alignment, vertical-alignment and style enums are:

const DocxBorder({
  this.color = '000000',
  this.size = 4,
  this.style = DocxBorderStyle.single,
});

enum DocxAlignment { left, center, right, justify }        // justify -> OOXML 'both'
enum DocxVerticalAlignment { top, center, bottom }
enum DocxBorderStyle { single, double, dashed, dotted }
enum DocxParagraphStyle {
  normal, heading1, heading2, heading3, heading4,
  subtitle, caption, quote, codeBlock, footnote,
  listBullet, listDash, listNumber, listNumberAlpha, listNumberRoman,
}

Both generators implement DocumentGenerator and return Uint8List bytes.

abstract class DocumentGenerator {
  Uint8List generate(DocxDocument document);
}

DocxGenerator defaults to Times New Roman at 24 half-points (12pt). DOCX font sizes are half-points:

DocxGenerator({
  this.fontName = 'Times New Roman',
  this.fontSize = 24,   // half-points (24 = 12pt, 28 = 14pt)
})

PdfGenerator works in points and exposes page size and margins. Its font must be one of the PDF base-14 fonts (Helvetica, Times-Roman, Courier).

The Docx-prefixed classes are exported under format-agnostic aliases, which is why the examples use Document, Paragraph and TextRun rather than their full names. Both spellings refer to the same class.