OpenXML Compatibility

GoWord writes Office Open XML WordprocessingML 2007 packages (ECMA-376 / ISO/IEC 29500 transitional). Microsoft Word 2007 through Microsoft 365 open the files. WPS Office and LibreOffice open the common subset (paragraphs, tables, images); DrawingML charts and wps:wsp shapes are Word-oriented.

Package layout

Part	Required	Notes
[Content_Types].xml	Yes	Overrides for document.xml, headers, charts, media
_rels/.rels	Yes	Points at word/document.xml
word/document.xml	Yes	Body + last w:sectPr
word/_rels/document.xml.rels	Yes	Images, headers, footers, charts, numbering
word/styles.xml	Yes	Named styles plus default Normal
word/media/imageN.*	If images	Names assigned at write time
word/charts/chartN.xml	If charts	One part per AddChart
word/headerN.xml / footerN.xml	If headers	Referenced from w:sectPr via r:id

Schemas the writer targets

Feature	Namespace / node	Word behaviour
Body text	w: WordprocessingML	Paragraphs, tables, sectPr
Native equations	m: Office Math (m:oMathPara, m:oMath)	Double-click opens the equation editor
Charts	c: ECMA-376 Chart	Series order is idx → order → tx → spPr
Word 2010 shapes	wps:wsp / a:prstGeom	Rect, roundRect, rightArrow, text box
Fields	w:instrText (PAGE, TOC, NUMPAGES)	Refresh with Update Field
Protection	w:documentProtection	Read-only / comments / tracked changes / forms
Edit exceptions	w:permStart / w:permEnd	AllowEdit ranges stay writable
Watermark	VML PowerPlusWaterMarkObject in headers	Diagonal or tiled word-art; WordPictureWatermark for images
SDT	w:sdt / w:sdtPr / w:sdtContent	Plain text, drop-down, date, w14:checkbox
Table row / cell	w:tblHeader, w:cantSplit, w:vAlign, w:textDirection	Repeating headers, unbreakable rows, vertical text

Strict rules the test suite enforces

The package openxml_strict_test.go rejects output that Word would offer to repair:

• Every w:tc contains at least one w:p.
• Template replacements never emit raw & inside w:t (no P style numeric entities for ordinary ASCII).
• Combo charts use two distinct c:axId values; the secondary series points at the second value axis.
• Area charts omit c:dLblPos (the area XSD does not allow it).
• c:legend sits after c:plotArea, never inside it.
• Line markers emit c:symbol plus c:size 5.
• w:cols uses w:sep, not w:separator.

Complete example — round-trip through the reader

package main

import (
	"fmt"
	"log"
	"os"

	"github.com/yunkeweb/go-word"
)

func main() {
	doc := word.New()
	sec := doc.AddSection()
	sec.AddTitle("Compatibility", 1)
	sec.AddText("Round-trip through CreateReader.")
	sec.AddMath(`x^{2}`)
	if err := doc.Save("compat.docx"); err != nil {
		log.Fatal(err)
	}

	r, err := word.CreateReader("Word2007")
	if err != nil {
		log.Fatal(err)
	}
	loaded, err := r.Load("compat.docx")
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println(loaded.ExtractText())

	_ = os.Remove("compat.docx")
}

ExtractText walks the reconstructed DOM. For files too large to load, use StreamExtractText. There is no word.ReadDOM; the DOM loaders are word.Open, word.Read, and word.Load.

Full-feature matrix (v0.9.0)

tests/matrix writes 80 randomly combined documents covering every exported module from v0.1.0 through v0.9.0: typography, multi-section layout, headers/footers, tables, images/shapes, TOC/bookmarks/comments, OMML, charts, SDT, and watermark/protection. Output stays in ./test_output_docs (gitignored).

go run ./tests/matrix
go run tests/matrix/validate_reader.go

Engine	What it checks	v0.9.0 result
word.Open / word.Read + ExtractText / ExtractImages	DOM reverse-parse, no panic, error == nil	80 PASS
word.StreamExtractText / word.StreamExtractImages	Streaming extract, no panic, error == nil	80 PASS
Microsoft Word COM (DisplayAlerts=0, OpenNoRepairDialog)	OpenXML repair dialogs, node order, parse exceptions	80 PASS

Related

When Word shows “the file is corrupt”, start at FAQ. Chart XSD order is documented on Charts. Recipe 4 on Enterprise Recipes combines SDT, repeating headers, tiled watermarks, and AllowEdit.

DOCX to HTML

The reader reconstructs a DOCX DOM that can be rendered as an HTML fragment or a standalone page:

html, err := doc.RenderHTML(word.HTMLOptions{
	Standalone: true,
	IncludeCSS: true,
})

Use word.RenderHTMLFile or word.RenderHTMLWithOptions when the input is still a file or reader. These functions return bytes for the caller to save. Images default to embedded data URIs; use ImageURL to persist assets and return their public URLs, or HTMLImageURL to reuse existing URLs. Neither mode automatically hosts media files. RenderHTMLWithDiagnostics reports unsupported elements retained in the DOM; Strict: true turns those diagnostics into an error. Content omitted during reading cannot be diagnosed by the renderer, so strict success does not imply lossless conversion. See the DOCX to HTML guide for complete programs, supported content, and read limits.