Extract all the layers of a PDF file into their own files

extract_pdf_layers.py

#!/usr/bin/env python3
import sys
import pikepdf

if len(sys.argv) < 3:
    print("Usage: extract_layers.py <input.pdf> <output_template>")
    print()
    print("Output template has to contain {num} somewhere, which will be replaced by the layer number.")
    sys.exit(0)


# check if we even have some OCGs
pdf = pikepdf.open(sys.argv[1])

try:
    layers = pdf.root.OCProperties.OCGs
except (AttributeError, KeyError):
    print("Unable to locate layers in PDF.")
    sys.exit(1)

page_count = len(pdf.pages)
pdf.close()

# (hopefully) all pdf operators which "display" anything. everything else is styling, which we need to preserve
hidden_operators = ["S", "s", "f" "F", "f*", "B", "B*", "b", "b*", "n", "Do", "sh", "Tj", "TJ", "m", "l", "c", "v", "y", "h", "re"]

extracted_groups = []
cur_layer = 0

for i in range(page_count):
    end_reached = False

    while not end_reached:
        commands = []
        extract_commands = True
        extracted_one = False

        pdf = pikepdf.open(sys.argv[1])
        page = pdf.pages[i]

        for j in range(len(pdf.pages)):
            if i < j:
                del pdf.pages[1]
            elif i > j:
                del pdf.pages[0]

        for operands, operator in pikepdf.parse_content_stream(page):
            if "/OC" in operands: # new OCG starts
                ocg_name = operands[1]
                if not ocg_name in extracted_groups and not extracted_one:
                    extracted_groups.append(ocg_name)
                    extract_commands = True
                    extracted_one = True
                else:
                    extract_commands = False
        
            if str(operator) == "EMC": # OCG has ended
                extract_commands = True
                continue

            if extract_commands or (not extract_commands and str(operator) not in hidden_operators):
                commands.append([ operands, operator ])

                # if cur_layer == 6:
                #     print("Operands {}, operator {}".format(operands, operator))

        if not extracted_one:
            end_reached = True
        else:
            page.Contents = pdf.make_stream(pikepdf.unparse_content_stream(commands))
            pdf.save(sys.argv[2].format(num=cur_layer))
            cur_layer += 1

Thanks for the reply, will give that a shot. Wasn't sure how easy it was to install old versions of libraries with python, in past experience with other languages I've sometimes found it simpler just to fix the code.

In case anyone else comes across this in future, the fix for the line:
if "/OC" in operands: # new OCG starts
is to change it to:
if pikepdf.Name("/OC") in operands: # new OCG starts

I did try to install an older version of pikepdf with pip install pikepdf==2.16.1 but that failed when it tries to build qpdf, which is the C++ library that pikepdf wraps, as far as I can tell.