itext2 to itext7: Merging PDF files

Years ago, I wrote a small app in itext2 to gather reports on a weekly basis and concatenate them into one PDF. The app used com.lowagie.text.pdf.PdfCopy to copy and merge the PDFs. And it worked fine. Performed exactly as expected.

A few weeks ago I looked into migrating the application to itex7. To that end, I used the copyPagesTo method of com.itextpdf.kernel.pdf.PdfDocument. When run on the same file set, this produces warnings like:

WARN PdfNameTree - Name "section.1" already exists in the name tree; old value will be replaced by the new one.

When I click on the link to "section.1" in the first document of the merged PDF, I am taken to "section.1" of the last document. Not what I expected and not what happens when using the itext2 app. In the PDF's produced by itext2, if I click on the link to "section.1" of the first document in the combined PDF, I am taken to section 1 of the first document.

There is a hint in Javadocs for copyPagesTo saying

> If outlines destination names are the same in different documents, all
> such outlines will lead to a single location in the resultant
> document. In this case iText will log a warning. This can be avoided
> by renaming destinations names in the source document.

There is however, no explanation of how this should be done. I find it odd that this should be necessary in itext7, although it wasn't in itext2.

Is there a simple way to get around his problem?

I've also tried the Sejda desktop app and it produces correct results, but I would prefer to automate the process through a batch script.

My guess is iText 2 didn't even know it might be a problem.

If iText can't deduplicate destination names, the procedure is roughly:

Follow /Catalog -> /Names -> /Dests in each document to find the destination name tree.

Deduplicate the names, by adding suffixes. Remember that a name with a suffix added might be equal to an existing name in the same or another document. Be careful!

Now you can rewrite the destination name trees. Since you have only used suffixes, you can do this in place - the lexicographic ordering of the names is unaltered so the search tree structure is not broken.

Now, rewrite destination links in each PDF for the new names. For example any dictionary entry with key /Dest, or any /D in a /GoTo action.

Now, after all this preprocessing, the files will merge without name clashes.

(I know all this because I've just implemented it for my own PDF software. It's slightly hairy stuff, but not intractable.)

If you like, I can provide a devel version of cpdf with this functionality, if you would like to test it.

Following up on the reply from @johnwhitington I got the following procedure to work.

As John suggests, start by renaming the destinations in the name tree:

Map<String, PdfString> renameDestinations(PdfDocument pdf,
                                                  String suffix) {

    Map<String, PdfString> renamed = new HashMap<>();

    PdfNameTree nameTree = pdf.getCatalog().getNameTree(PdfName.Dests);
    Map<String, PdfObject> names = nameTree.getNames();

    Set<String> ks = new HashSet<String> (names.keySet());
    for (String key : ks) {
        String oldName = key;
        PdfString newName = new PdfString(oldName + suffix);
        PdfObject value = names.get(key);
        names.remove(key);
        names.put(newName.toUnicodeString(), value);
        renamed.put(oldName, newName);
    }

    nameTree.setModified();

    return renamed;
}

Then change the destinations in the outlines to the new destinations:

void renameOutlines(PdfOutline pdo, Map<String, PdfString> old2new) {

    for (PdfOutline child : pdo.getAllChildren()) {
        renameOutlines(child, old2new);
    }

    PdfDestination dest = pdo.getDestination();

    if (!((dest instanceof PdfNamedDestination)
         || (dest instanceof PdfStringDestination))) {
        return;
    }

    String oldDest = "";
    if (pdo.getDestination().getPdfObject().isString()) {

        oldDest = ((PdfString) pdo.getDestination()
                                  .getPdfObject()).toUnicodeString();

    } else if (pdo.getDestination().getPdfObject().isName()) {

        oldDest = ((PdfName) pdo.getDestination()
                                .getPdfObject()).getValue();
    }

    if (oldDest != null && old2new.containsKey(oldDest)) {
        pdo.addDestination(new PdfStringDestination(
                                            old2new.get(oldDest)));
    }

}

Finally, change the links on each page to use the new destinations:

void renameLinks(PdfDocument pdf, Map<String, PdfString> renamed) {

    int numPages = pdf.getNumberOfPages();
    for (int p = 1; p <= numPages; p++) {

        List<PdfAnnotation> annots = pdf.getPage(p).getAnnotations();

        for (PdfAnnotation annot : annots) {

                PdfObject dest = annot.getPdfObject().get(PdfName.Dest);

                if (dest != null && dest.isName()) {
                    String oldString = ((PdfName) dest).getValue();
                    if (renamed.containsKey(oldString)) {  
                        annot.getPdfObject().put(PdfName.Dest, 
                                   renamed.get(oldString));
                    }
                }
        }
    }
}

The suffix of course should be unique to each file being merged.