awaforge

docxText

AGPL-3.0-onlyRelease 2L4

Plain-text extraction helpers for a typed docx document tree — the module behind docx.toText.

Module docxText | Source packages/front/office/ooxml/src/docx/docx-text.js | Deps none | Worker-safe yes

Pure functions over the typed model returned by docx.read (result.document): no runtime dependency, no state, no side effect. The docx orchestrator depends on this module and re-exposes its toText as docx.toText — the two are the same function, so most callers never resolve docxText directly.

Resolve

const text = runtime.resolve('docxText');
// Returns: { toText, textOfParagraph, textOfRun, textOfTable }

API

Method Signature Returns
toText (doc: { body? }) => string The text of the whole document: one line per top-level paragraph or table, joined by \n.
textOfParagraph (p: paragraph) => string The concatenated text of the paragraph's runs, including the runs of hyperlink and ins children.
textOfRun (r: run) => string The text of one run.
textOfTable (t: table) => string Rows joined by \n, cells by \t.

Rules

Node Contributes
text its value
tab \t
break \n
noBreakHyphen U+2011 (non-breaking hyphen)
hyperlink, ins the text of their runs
del nothing — deleted text is not visible text
table cell its paragraphs (and nested tables) joined by a single space

Examples

const text = runtime.resolve('docxText');

const doc = {
    body: [
        { type: 'paragraph', children: [
            { type: 'run', children: [{ type: 'text', value: 'Hello' }, { type: 'tab' }, { type: 'text', value: 'world' }] },
            { type: 'del', children: [{ type: 'run', children: [{ type: 'text', value: 'gone' }] }] }
        ] },
        { type: 'table', rows: [{ cells: [
            { children: [{ type: 'paragraph', children: [{ type: 'run', children: [{ type: 'text', value: 'A1' }] }] }] },
            { children: [{ type: 'paragraph', children: [{ type: 'run', children: [{ type: 'text', value: 'B1' }] }] }] }
        ] }] }
    ]
};

console.log(JSON.stringify(text.toText(doc))); // "Hello\tworld\nA1\tB1"

On a read document

const word = runtime.resolve('docx');
const read = word.read(word.write(word.fromText(['First', 'Second'])));
console.log(word.toText(read.document)); // "First\nSecond"
console.log(word.toText === runtime.resolve('docxText').toText); // true

Notes

  • Only top-level paragraph and table nodes of doc.body are visited. Other block-level nodes, such as a blockSdt (a structured document tag wrapping paragraphs), contribute nothing, and neither does an inline sdt inside a paragraph (see templating for the SDT model).
  • Headers, footers, footnotes, endnotes and comments are separate parts of the read result, not part of doc.body: toText does not reach them.
  • The output is a flat reading of the text, not a layout: formatting, fields, drawings and list numbers are dropped.

See also

  • docx — orchestrator (toText).
  • docx-walker — the other module extracted from the orchestrator.
  • docx-structure — the node types (paragraph, run, table, …).