Scale or resize only visible layers in PDF

[Mahendranath94]

I have scaled a PDF and saved it as new pdf, when I reviewed the rescaled new pdf, I noticed that the layers which are hidden in the original pdf are becoming visible in new pdf.

import fitz

src = fitz.open(r"Input.pdf")
doc = fitz.open()
for page in src:
    w, h = page.rect.br 
    newpage = doc.new_page(width=1.2*w, height=1.2*h)
    newpage.show_pdf_page(newpage.rect, src, page.number)
doc.ez_save(r"output.pdf")

when I checked the OCG layer of both input and output pdf, I got this :

i/p: print(src.get_layer()) 
o/p: {'on': [3085, 3086, 3046, 3091, 3092, 3093], 'off': [3084, 3087, 3088, 3089, 3090]}

i/p: print(doc.get_layer())
o/p: None

I understand why we get "None" for output pdf, but the issue is when I see the output pdf, I can notice some text, tables and images which are not there/visible in input pdf, which means on scaling using above code, the OCG layers which are turned off are also scaled and becomes visible, and as we get "None" for the doc.get_layer() of output.pdf, we can turn those them off.
How can we scale the PDF only with visible layers.

[JorjMcKie]

This is impossible to achieve ... at least in this way.
The reason is that Optional Content (= layers information) is only available if the document contains the required configuration data. A page has no such information at all - except pointing to some OCGs (layers) via their xref numbers.

PDF, as a file format, has little or no internal integrity checks. Therefore, if the page references an xref in the expectation that this is a layer with a ON/OFF state, but that xref does not exist or is not referenced in the document's central OC information, the layer is assumed (by PDF and all the viewer software) to be ON.

That is your situation.

[JorjMcKie]

Hard to say what a solution might be. In any case, it would have to be quite hacky.
Depending in the object type, reference to OCGs is either directly part of the object definition (images, Form XObjects and annotations), or segments inside content streams are marked appropriately with back-references to the /Resources/Properties sub-dictionary.

As an additional complication, the object inserted into the target page via show_pdf_page() is no longer a PDF page object upon arrival, but has been converted to a Form XObject.
You could inspect that Form XObject's /Resources dictionary and see what has happened to the original /Properties sub-dict. Maybe you can start from there.