iText fails to propery merge signatures in PDF files - itext

Question

I have a problem with merging PDF files using iText. I am using the PdfCopyFields class to do the merging yet, the signatures of existing PDF files are not merged properly. It seems that some data in the dictionary of the merged PDF file are incorrect.
To be specific:
I have 2 documents i want to merge, each document has 2 pages, each page in each document is signed with a different signature (in each signature i put a different reason code in order to be able to visually tell each signature).
The merged PDF file contains:
4 PDF pages (this is correct)
All 4 pages have a signature
The signature of pages 3 and 4 are NOT the original signatures, but they are the same signatures as those of page 1 and 2. To verify that the signatures are different i put a unique 'Reason' code in the signature.
It is obvious that although the resulting PDF has 4 signatures (which i know are PDF fields), it seems that these signature-fields reference the signature data of the wrong document.
Moreover, when i open the merged pdf file with a text editor, i look at the 'header' of the PDF file and find only 2 Signature entries (instead of 4). This means that the actual signatures in the pages reference only those 2 signatures, thus the mix-up.
Thank you
Costas
PS: i can post sample PDF files to reproduce the error, but the simplest ones will do the job (i created 2 PDF files with MS Word and stamped each page separately)

Answer 1

Concerning the observation
The signature of pages 3 and 4 are NOT the original signatures, but they are the same signatures as those of page 1 and 2. To verify that the signatures are different i put a unique 'Reason' code in the signature.
That might happen if the names of the signature fields in the documents coincide. Multiple PDF fields with the same name are considered multiple visualizations of the same field. The merge process may throw away duplicate values in that case.
I'm not sure, though, whether that is the case for your files. If you want to know, please share them.
...
Having inspected the sample files it becomes clear that indeed the problem is caused by the identical signature field names in the merged documents:
Doc1-signed.pdf has
a signed signature field Signature1 (field and widget merged) on page 1 with Reason Doc1-Page1 in its value and
a signed signature field Signature2 (field and widget merged) on page 2 with Reason Doc1-Page2 in its value.
Doc2-signed.pdf has
a signed signature field Signature1 (field and widget merged) on page 1 with Reason Doc2-Page1 in its value and
a signed signature field Signature2 (field and widget merged) on page 2 with Reason Doc2-Page2 in its value.
Merging result in MERGED-PDF.pdf which has
a signed signature field Signature1 with Reason Doc1-Page1 in its value with explicit widgets on pages 1 and 3 and
a signed signature field Signature2 with Reason Doc1-Page2 in its value with explicit widgets on pages 2 and 4.
Because the whole PDF is considered a single form, a form field name can only have a single associated value.
Thus, merging multiple fields with the same name from the two source documents into a single fields with multiple widgets (as PdfCopyFields seems to do) is a sensible action to take.
I tried using an online pdf merge service and the signature fields were merged properly
By merged properly I assume you mean they still had their original, differing values. This in turn indicates that that service did not merge the fields as described above.
But this is not more proper than what PdfCopyFields does, it is dumber because the value of the field Signature1 now is unclear, just like the value of Signature2.
The proper thing to do if you want to keep the differing values of the source fields with duplicate names, is to rename such duplicate fields in the course of the merge. (If the online PDF merge service did this, too, it was not dumb. But you did not indicate any change in field names...)
You can find sample code for a merge of documents with their fields renamed in iText in Action chapter 6 Working with existing PDFs example ConcatenateForms2.java:
PdfCopyFields copy
= new PdfCopyFields(new FileOutputStream(RESULT));
// add a document
PdfReader reader1 = new PdfReader(renameFieldsIn(DATASHEET, 1));
copy.addDocument(reader1);
// add a document
PdfReader reader2 = new PdfReader(renameFieldsIn(DATASHEET, 2));
copy.addDocument(reader2);
// Close the PdfCopyFields object
copy.close();
reader1.close();
reader2.close();
using the helper method
private static byte[] renameFieldsIn(String datasheet, int i)
throws IOException, DocumentException {
ByteArrayOutputStream baos = new ByteArrayOutputStream();
// Create the stamper
PdfStamper stamper = new PdfStamper(new PdfReader(datasheet), baos);
// Get the fields
AcroFields form = stamper.getAcroFields();
// Loop over the fields
Set<String> keys = new HashSet<String>(form.getFields().keySet());
for (String key : keys) {
// rename the fields
form.renameField(key, String.format("%s_%d", key, i));
}
// close the stamper
stamper.close();
return baos.toByteArray();
}
Obviously you can tweak this by only renaming signature fields. This would merge other fields with "normal" form content but leave signatures as they are.
(remember, i don't refer to the validity of the signatures themselves, just the signature fields).
As an afterthought, flattening the signature fields before merging might be an alternative approach. The visual representation remains but the verification failure messages are gone because nothing is verified anymore.
A general remark on merging signed PDFs
Your intention
I have 2 documents i want to merge, each document has 2 pages, each page in each document is signed with a different signature
cannot be implemented without completely invalidating the signatures, at least those from all but one document. Have a look here for an introduction to integrated PDF signatures. Especially note how multiple integrated signatures in the same document work:
After a merge of two documents, you can keep the signatures from one document valid, but the added signatures of the other document only cover data from their document while after the merge they would have to cover data from both documents.
Thus, a merge is impossible without breaking at least some of the signatures.
Merging in more current iText versions
The OP uses an iText version 4.2.0. In current iText versions (5.5.x) much of the form aware logic has been moved from PdfCopyFields to PdfCopy. If you use such a version or a newer one, try and use PdfCopy.
#Bruno on merging signature fields
The result of the merge above will be utterly invalid in PDF-2, not merely because of the invalid signatures themselves but because of the signatures with multiple appearances. You might want to reconsider the behavior of the Pdf*Copy* class family concerning signature fields.