Gylfi Ingvason
2009-11-19 19:46:31 UTC
Gents,
I use PdfStamper to flatten forms/free text and then typically perform other
operations using either PdfWriter or PdfCopy. This has been working fine
until recently when a customer complained that text fields in an AcroForm
were missing after flattening the form.
Upon investigating, I found that the Fields array of the AcroForm was empty.
The form does not get flattened because PdfStamper thinks there is nothing
to do. To my surprise, neither Acrobat nor the FoxIt Reader complain about
this - presumably because the Widget Annot is referenced by the Page object.
Enclosed is a small file that I created by iText to illustrate the problem.
I have deliberately erased the contents of the Fields array (it was '3 0 R')
of the AcroForm (object 7) to simulate the user data.
Here's the interesting part - not only does the file display fine, but if
you use Acrobat to extract that page, it is smart enough to properly
re-construct the Fields array in the process.
This brings up a couple of questions:
1. Is it valid to not reference Widget Annots in a Form and expect them to
work?
The spec seem unambiguous to me that it isn't valid (see PDF 32000-1:2008
pg. 431: "Fields - (Required) An array of references to the document's root
fields (those with no ancestors in the field hierarchy)."). Chapter 15 of
Bruno's book seems to indicate that as well, but perhaps I'm missing
something.
2. What would be the best approach to deal with a form with an empty Fields
array, such that the text gets flattened, assuming that this is legal?
I have conveyed to the customer that I consider the file damaged and he
should get better tools, but it is hard to argue against things like this
where Acrobat gives no sign of anything being wrong. Thus, a third question:
3. Would it be reasonable to patch the iText code such that form flattening
takes place on Annots not referenced in the Fields array of the form?
Best regards,
Gylfi
I use PdfStamper to flatten forms/free text and then typically perform other
operations using either PdfWriter or PdfCopy. This has been working fine
until recently when a customer complained that text fields in an AcroForm
were missing after flattening the form.
Upon investigating, I found that the Fields array of the AcroForm was empty.
The form does not get flattened because PdfStamper thinks there is nothing
to do. To my surprise, neither Acrobat nor the FoxIt Reader complain about
this - presumably because the Widget Annot is referenced by the Page object.
Enclosed is a small file that I created by iText to illustrate the problem.
I have deliberately erased the contents of the Fields array (it was '3 0 R')
of the AcroForm (object 7) to simulate the user data.
Here's the interesting part - not only does the file display fine, but if
you use Acrobat to extract that page, it is smart enough to properly
re-construct the Fields array in the process.
This brings up a couple of questions:
1. Is it valid to not reference Widget Annots in a Form and expect them to
work?
The spec seem unambiguous to me that it isn't valid (see PDF 32000-1:2008
pg. 431: "Fields - (Required) An array of references to the document's root
fields (those with no ancestors in the field hierarchy)."). Chapter 15 of
Bruno's book seems to indicate that as well, but perhaps I'm missing
something.
2. What would be the best approach to deal with a form with an empty Fields
array, such that the text gets flattened, assuming that this is legal?
I have conveyed to the customer that I consider the file damaged and he
should get better tools, but it is hard to argue against things like this
where Acrobat gives no sign of anything being wrong. Thus, a third question:
3. Would it be reasonable to patch the iText code such that form flattening
takes place on Annots not referenced in the Fields array of the form?
Best regards,
Gylfi