How to convert a page or few pages into XHTML files?

[minhld]

Hello,
I saw a lot of examples converting the whole DOCX file into a single XHTML file but still don't know how to convert only a few selected pages into 1 or XHTML files (1 page -> 1 XHTML file).
If it is possible, could you give an example? Thank you.

[jason]

There is no notion of a page in docx4j, since we don't currently have a page layout model (except for that used in the Enterprise ed for TOC generation).

You can try relying on the lastRenderedPageBreak element which Word inserts.

Or you could potentially put the content you want to convert into a content control, then convert that. The easiest way would be to programmatically delete the rest of the docx, then convert what remains.

[jason]

Alternatively, you could consider chunking on headings. I did that to create the HTML pages you see at webapp.docx4java.org/OnlineDemo/ecma376/WordML/index.html

[minhld]

Thank you for the answers
I'm now trying the solution you suggested, searching for lastRenderedPageBreak but it looks complicated for me
If failed, I think DOCX -> PDF -> XHTMLs is the solution, although data missing problem would be more serious.

[jason]

Update, September 2026

Revisiting this, because the premise has changed: docx4j now has a page layout model, so "which content is on page N" is something it can tell you.

From 17.1.1, org.docx4j.model.pagination.Paginate lays the document out with Apache FOP (docx4j-export-fo on the classpath; since 17.1.0 that layout follows Word's rules) and reports, for every paragraph of the main document part, the page it starts on and, for paragraphs that span pages, the character offsets where each new page begins:

Code: Select all

PaginationMap map = Paginate.compute(wordMLPackage, null);
for (String paraId : map.getKeys()) {          // document order
    int page = map.getPageIndex(paraId);        // 1-based
    int[] breaks = map.getBreaks(paraId);       // empty unless the paragraph spans pages
}


Or, closer to what was suggested back in 2014, let it write the result into the document:

Code: Select all

Paginate.paginate(wordMLPackage, null);


That rewrites the document's w:lastRenderedPageBreak markers from a fresh layout - one at the start of each page, splitting a run mid-paragraph where a page begins inside it, exactly as Word records its own last rendering. The markers you were struggling with are then trustworthy, rather than a leftover from whenever Word last saved the file.

Either way, "one page -> one XHTML file" becomes a split-then-export:

1. paginate;
2. for each page, take a copy of the package (XmlUtils.deepCopy of the body, or clone the whole package), keep only the content between that page's marker and the next, and drop the rest;
3. Docx4J.toHTML that copy.

Paragraph granularity (everything from the paragraph containing the marker) is the simple version and is usually what people want for HTML; if you need the exact character, getBreaks gives it, and the markers are already at the right character in the runs. A table that spans a page boundary needs a decision on your part - HTML has no pages, so most people keep the whole table on the page it starts.

The usual caveat: it is FOP's layout, with the fonts FOP can see, so page boundaries are close to Word's rather than identical. That's still a much better basis than the docx -> PDF -> XHTML route floated above, which loses the document structure the HTML export preserves.