How can I drop metadata fields (e.g., PageLabel fields) from PDFs?

Question

I have used pdftk to change the "Info" metadata associated with a PDF. I currently have several PDFs with extraneous page labels and I cannot figure how to drop them. This is what I am currently doing:

$ pdftk example_orig.pdf dump_data output page_labels.orig
$ grep -v PageLabel page_labels.orig > page_labels.new
$ pdftk example_orig.pdf update_info page_labels.new output example_new.pdf

This does not remove the PageLabel* metadata which can be verified with:

$ pdftk example_orig.pdf dump_data | grep PageLabel

How can I programmatically remove this metadata from the PDF? It would be nice to do with with pdftk but if there another tool or way to do this on GNU/Linux, that would also work for me.

I need this because I am using LaTeX Beamer to generate presentations with the \setbeameroption{show notes on second screen} option which generates a double-width PDF for showing notes on a second screen. Unfortunately, there seems to be a bug in pgfpages which results in incorrect and extraneous PageLabels in these files (example). If I generate a slides only PDF, it will generates the correct PageLabels (example). Since I can generate a correct set of PageLabels, one solution would be to replace the pagelabels in the first examples with those in the second. That said, since there are extra pagelabels in the first example, I would need to remove them first.

Answer 1

Using a text editor to remove PDF metadata

1. If it is the first time you edit a PDF, make a backup copy first.

2. Open your PDF with a text editor that can handle binary blobs. vim -b will be fine.

3. Locate the /Info dictionary. Overwrite all the entries you do not want any more completely with blanks (an entry consists of /Key names plus the (some values) following them).

4. Be careful to not use more spaces than there were characters initially. Otherwise your xref table (ToC of PDF objects will be invalidated, and some viewers will indicate the PDF as corrupted).

5. For additional measure, locate the /XML string in your PDF. It should show you where your XMP/XML metadata section is (not all PDFs have them). Locate all the key values (not the <something keys>!) in there which you want to remove. Again, just overwrite them with blanks and be careful not to change the total length (neither longer, nor shorter).

In case your PDF does not make the /Info dictionary accessible, transform it with the help of qpdf.

1. Use this command:

   qpdf --qdf --object-streams=disable orig.pdf qdf---orig.pdf
   

2. Apply the procedure outlined above. (The qdf---orig.pdf now should be much better suited for

3. Re-compact your edited file:

   qpdf qdf---orig.pdf  edited---orig.pdf
   

4. Done! Enjoy your edited---orig.pdf. Check if it has all the data removed:

   pdfinfo -meta edited---orig.pdf
   

---

Update

After looking at the sample PDF files provided, it became clear to me that the /PageLabel key is not part of the /Info dictionary (PDF's Document Information Dictionary), but of the /Root object.

That's probably one reason why pdftk was unable to update it with the method the OP described.

The other reason is the following: the PDF which the OP quoted as containing the correct page labels does in fact contain incorrect ones!

 Logical Page No. |  Page Label
 -----------------+------------
               1  |   1
               2  |   2
               3  |   2
               4  |   2
               5  |   2
               6  |   4

The other PDF (which supposedly contains extraneous page labels) is incorrect in a different way:

 Logical Page No. |  Page Label
 -----------------+------------
               1  |   1
               2  |   1
               3  |   2
               4  |   2
               5  |   2
               6  |   4

My original advice about how to manually edit the classical metadata of a PDF remains valid. For the case of editing page labels you can apply the same method with a slight variation.

In the case of the OP's example files, the complication comes into play: the /Root object is not directly accessible, because it is hidden inside a compressed object stream (PDF object type /ObjStm). That means one has to decompress it with the help of qpdf first:

1. Use qpdf:

   qpdf --qdf --object-streams=disable example_presentation-NOTES.pdf q-notes.pdf
   

2. Open the resulting file in binary mode with vim:

   vim -b q-notes.pdf
   

3. Locate the 1 0 obj marker for the beginning of the /Root object, containing a dictionary named /PageLabels.

   (a) To disable page labels altogether, just replace the /PageLabels string by /Pagelabels, using a lowercase 'l' (PDF is case sensitive, and will no longer recognize the keyword; you yourself could at some other time restore the original version should you need it.)

   (b) To edit the page labels, first see how the consecutive labels for pages 1--6 are being referred to as

      <feff0031>
      [....] 
      <feff0032>
      [....] 
      <feff0032>
      [....] 
      <feff0032>
      [....] 
      <feff0033>
      [....] 
      <feff0034>
   

   (These values are in BOM-marked hex, meaning 1, 2, 2, 2, 3, 4...)

   Edit these values to read:

       <feff0031>
       [....] 
       <feff0032>
       [....] 
       <feff0033>
       [....] 
       <feff0034>
       [....] 
       <feff0035>
       [....] 
       <feff0036>
   

4. Save the file and run qpdf again in order to re-compress the PDF:

   qpdf q-notes.pdf notes.pdf
   

   These now hopefully are the page labels the OP is looking for....

Since the OP seems to be familiar with editing pdftk's output of dump_data output, he can possibly edit the output and use update_data to apply the fix to the PDF without needing to resort to qpdf and vim.

---

Update 2:

User @Iserni posted a very good, short and working answer, which limits itself to one command, pdftk, which the OP seems to be familiar with already, plus sed -- not needing to use a text editor to open the PDF, and not introducing an additional utility qpdf like my answer did.

Unfortunately @Iserni deleted it again after a comment of mine. I think his answer deserves to get the bounty and I call you to vote to "undelete" his answer!

So temporarily, I'll include a copy of @Iserni's answer here, until his is undeleted again:

Not sure if I correctly understood the problem. You can try with a butcher's solution: brute force replace the /PageLabels block with a different one which will not be recognized.

# Get a readable/writable PDF
pdftk file1.pdf output temp.pdf uncompress

# Mangle the PDF. Keep same length
sed -e 's|^/PageLabels|/BageLapels|g' < temp.pdf > mangled.pdf

# Recompress
pdftk mangled.pdf output final.pdf compress

# Remove temp file
rm -f temp.pdf mangled.pdf

Answer 2

Not sure if I correctly understood the problem. You can try with a butcher's solution: brute force replace the /PageLabels block with a different one which will not be recognized.

# Get a readable/writable PDF
pdftk file1.pdf output temp.pdf uncompress

# Mangle the PDF. Keep same length
sed -e 's|^/PageLabels|/BageLapels|g' < temp.pdf > mangled.pdf

# Recompress
pdftk mangled.pdf output final.pdf compress

rm -f temp.pdf mangled.pdf