How to get information about layout of paragraph

[georgezeng]

I am working on manipulating a .docx file and build it into a jasperreport template file.
I need to set the same layout in the jasper file, but a first barrier comes when a paragraph lies more than one text line in word,
I have no idea how to get the information of this, does any one knows about it? I'll be appreciated!

[jason]

docx4j does not have a layout model.

A layout model would be a useful thing to add for applications such as yours, and also for calculating page numbers (eg refreshing TOC).

Layout of a paragraph would be straightforward enough. A complete model (columns, images, tables, footnotes etc) would be a significant undertaking.

To layout a paragraph, you need to know how much horizontal space you have (page width - margins - indentation), and font information (size, kerning etc).

(Ideally, we'd replicate Word's precise algorithm, but a starting point (Google 'line break algorithm') is http://en.wikipedia.org/wiki/Word_wrap and http://stackoverflow.com/questions/1758 ... -algorithm )

To find how much vertical space it will take, you need line spacing, plus space before/after.

docx4j can tell you all that stuff. If you'd like to contribute a basic model, I'd be happy to write the basic code which gets this information for you. You'd then have to fill in the details (ie the calculations based on this information).

There are quite a few projects around which you can look at to see how layout is done, for example pango (see http://fishsoup.net/bib/PangoIuc25-paper.pdf ), Flying saucer xhtmlrenderer (which lays out XHTML, and is used by docx4j for XHTML import), FOP (which lays out XSL FO - see http://wiki.apache.org/xmlgraphics-fop/KnuthsModel ), AWT, but I don't know of a reference book which explains the principles. To start off, have a read of http://behdad.org/text/

I don't mean to scare you off ... sounds like just a simple line break / word-wrap algorithm is what you are after, rather than a full blown page layout model. In any case, a simple line break / word-wrap algorithm is a good place to start.

[jason]

Update, September 2026

The answer above was right in 2012, but it isn't any more, so for anyone landing here from a search:

docx4j now has a layout model, courtesy of Apache FOP. Two things changed:

• 17.1.0: the PDF/XSL-FO output (docx4j-export-fo) was reworked to follow Word's layout rules by default - line breaking, paragraph spacing, keeps, table rows, page breaks and so on - measured against Word's own page counts over a corpus of real documents. So FOP's layout of a docx4j FO rendering is now a reasonable stand-in for Word's.
• 17.1.1: a new class org.docx4j.model.pagination.Paginate reads FOP's area tree (its record of what it laid out where) and turns it into data about your document.

What Paginate gives you directly:

Code: Select all

PaginationMap map = Paginate.compute(wordMLPackage, null);
map.getPage(paraId);      // the page a paragraph starts on
map.getBreaks(paraId);    // for a paragraph that spans pages, the character offsets where each new page begins
map.getPageCount();


Paragraphs are keyed by their w14:paraId (assigned if missing). Paginate.paginate(...) additionally writes the result back into the document as w:lastRenderedPageBreak markers, exactly as Word records its own last rendering. The TOC generator's page numbers now come from the same place.

For the original question - how many lines a paragraph occupies, and where each line breaks - Paginate doesn't expose that, but the information is there. FOP's area tree contains one lineArea element for every line it laid out, with the text of that line, inside a block per paragraph. To get at it, ask for the area tree instead of a PDF:

Code: Select all

FOSettings foSettings = Docx4J.createFOSettings();
foSettings.setOpcPackage(wordMLPackage);
foSettings.setApacheFopMime("application/X-fop-areatree");
foSettings.getFeatures().add(ConversionFeatures.PP_FO_PARAGRAPH_IDS); // each block gets prod-id="p-" + paraId
Docx4J.toFO(foSettings, outputStream, Docx4J.FLAG_EXPORT_PREFER_NONXSL);


Then walk the resulting XML: each block with a prod-id is one of your paragraphs, and its lineArea children are its lines. Paginate's own reader (PaginationAreaTreeHandler) is a worked example of parsing it.

Two caveats. It is FOP's layout, using the fonts FOP can see, so it is close to Word rather than identical (the closer your fonts match, the closer the result). And it needs docx4j-export-fo on the classpath, since the layout is a real rendering.