Retrieve all the versions of a PDF when saved incrementally

[3051360]

I've observed that PyMuPDF provides functionality to determine the number of versions present in a document.

Currently, I'm utilizing pdfresurrect, a C tool/library designed to "extract older 'hidden' versions of a PDF from the current PDF."

While my larger application is built using PyMuPDF, I have to call the compiled C code through Python for version retrieval, which is not an ideal way of building applications.

I propose that enhancing the capability to retrieve all PDF version files directly within PyMuPDF would be beneficial.

[cipri-tom]

Here's how I do it:

        pattern = re.compile(b'[^t]xref.*?EOF', flags=re.M+re.S)

        # `inc_nb` means the number of incremental update contained in the binary
        matches = list(pattern.finditer(doc.data))
        inc_nb = len(matches)
        if not inc_nb > 1:
            return result

        # here we loop to separate each incremental update in
        # the binary to retrieve each version of the file
        # We don't take into account the current version
        for sub in matches[:-1]:
            stream = doc.data[:sub.span()[1]] # binary code of a version
            if not stream:
                continue
            previous = Doc(stream).as_fitz() # is one of the previous version of the file

[3051360]

@cipri-tom Thanks for sharing the approach. I get the essence of what you are trying to do. I think you have provided a part of the code which at the end could result in getting all the versions of an incrementally saved PDF file.

I attempted to complete the overall code (with some dummy code in between) but obviously did not succeed.

import fitz
import re
import os

def extract_versions(input_pdf_path, output_folder):
    doc = fitz.open(input_pdf_path)

    pattern = re.compile(b'[^t]xref.*?EOF', flags=re.M + re.S)

    matches = list(pattern.finditer(doc.get_raw_data()))

    if len(matches) <= 1:
        print("Only one version found. No incremental updates.")
        return

    os.makedirs(output_folder, exist_ok=True)

    for i, sub in enumerate(matches[:-1]):
        start_offset, end_offset = sub.span()
        version_data = doc.get_raw_data()[start_offset:end_offset]

        version_doc = fitz.open()

        version_doc.insert_pdf(fitz.open(fitz.Stream(fitz.SOString, version_data)))

        output_path = os.path.join(output_folder, f"version_{i + 1}.pdf")
        version_doc.save(output_path)
        print(f"Version {i + 1} saved to: {output_path}")

        version_doc.close()

    doc.close()

I am not sure how to proceed further but maybe someone looking for a similar feature could find this code useful.

As @JorjMcKie mentioned in the later thread, I will drop a request to people at MuPDF for this feature.

[cipri-tom]

the problem here is that you get version_data from start_offset to end_offset. Instead, you should always get it from the start, until the end_offset. This is why it is "incremental", new data is appended, working in addition to existing data.

[JorjMcKie]

This feature would have to be implemented by our base library MuPDF to make any sense.
Currently, you can find out the number of versions of a PDF: doc.version_count only.

Please talk to the MuPDF colleagues on e.g. their Dicscord channel about options they may see.

[3051360]

@JorjMcKie Thank you for your response, I will reach out to people at MuPDF to see if they could point me to something.