How do I remove layers or copy a layer to a new file?

[p-vdp]

I'm new to this package, and I've figured out how to toggle layers on/off:

import pymupdfd
doc = pymupdf.open("file.pdf")
ogcs = doc.get_ocgs()
lyr_xref = ogcs[0]
doc.set_layer(-1, on=[lyr_xref], basestate="OFF")  # turn off all but the first layer

But how do I delete them, or copy a specific layer to a new file?

[JorjMcKie]

This is no issue, but a Discussions item or even better a post on our Forum.

[p-vdp]

Cool, thank you for moving this to the right place. I also created a topic in the forum: https://forum.mupdf.com/t/how-do-i-remove-layers-or-copy-a-layer-to-a-new-file/183

[JorjMcKie]

Why do you even want to delete a layer? There is no associated file size reduction after all.

Whether or not a PDF takes note of an OCG (Optional Content Group - and that exactly is what the term "layer" means) exclusively depends on whether or not its xref is present in the PDF catalog array OCProperties/OCGs. Taking an OCG xref out of that array will cause it being ignored in all Optional Content matters.
And you should never physically delete an OCG object because it may be referenced by other objects in the PDF and a deletion may thus render the PDF completely unusable.

There is no way to "copy a layer" between PDFs: there seems to be a general misconception here:
OCGs are best imagined as boolean objects that can be ON of OFF. That value has an effect on other objects in the PDF if and only if they refer to that OCG. If an OCG is OFF, then all referring objects are treated as non-existent / invisible.
Multiple OCGs can also be combined in boolean expressions called OCMD, and PDF objects can refer to an OCMD. So the Optional Content situation in a PDF can be quite complex.

[JorjMcKie]

OCGs are very small objects with no more than a handful of attributes. So if you have a new PDF then simply define a few OCGs and maybe OCMDs and assign them to any desired other objects.
There are numerous examples for doing this.
Maybe you start by reading this blog article: https://artifex.com/blog/translating-pdfs-a-practical-pymupdf-guide, and here are some more examples: https://github.com/pymupdf/PyMuPDF-Utilities/tree/master/optional-content

[p-vdp]

Thanks for the suggestions! I'll check them out.

For context: my use case is wrangling PDF exports from CAD files -- which preserve CAD layers on the OCG objects/layers -- and needing to convert back to CAD. The reason I want to delete layers is because I will need to convert PDF to DXF/DWG, and the solutions I've tested so far don't respect the on/off settings. Essentially I'm trying to isolate layers for downstream use.

For example, this file: building001-0_floor1.pdf How would you create a new PDF only containing the layer named A-WALL?

[JorjMcKie]

I think you want to make a PDF page that looks like as if only the walls were there:

Assuming your answer to my question is YES:
This cannot be achieved by removing unwanted OCGs from the /OCGs array!
Because all objects referencing a now removed OCG will be visible!

No: the only way is to set A-WALL to ON and all other OCGs to OFF permanently and then save the PDF in this state. The following snippet does this:

import pymupdf

mupdf = pymupdf.mupdf
doc = pymupdf.open("building001-0_floor1.pdf")

# get the xref of OCG "A-WALL"
ocgs = doc.get_ocgs()  # all OCGs
A_WALL_xref = [xref for xref in ocgs.keys() if ocgs[xref]["name"] == "A-WALL"][0]
doc.set_layer(-1, on=[A_WALL_xref], basestate="OFF")

doc.ez_save("test.pdf")

[JorjMcKie]

So your initial idea was right and the only feasible way to achieve this.

[JorjMcKie]

Other users visiting this page may want to look at this post in the Forum.

[p-vdp]

@JorjMcKie Thank you for the quick responses!

Now I appreciate that "deleting" the OCG isn't the way to go because removing the xref doesn't remove the data being referenced. (Correct?)

Perhaps I'm asking the question in the wrong way. In Adobe Acrobat my desired result can be achieved by turning off the unwanted layers then doing File -> Save as Other -> Optimized PDF... -> Discard User Data -> Discard hidden layer content and flatten visible layers

Which results in this file: building001-0_floor1_discarded.pdf

This removes the underlying vector graphics and text in the off layers. The difference can be seen in a vector graphics editor like Illustrator or InkScape. On the left is the output "test.pdf" from your code sample, and on the right is the output from the Acrobat optimizer:

Is there a way to do this in pymupdf? Perhaps one of the parameters in Document.save?

[JorjMcKie]

Sorry - you labelled the pictures wrong:

The right side looks like the outcome of my snippet
test.pdf

The left side isn't it.

import pymupdf

doc = pymupdf.open("building001-0_floor1.pdf")
# get the xref of OCG "A-WALL"
ocgs = doc.get_ocgs()  # all OCGs
A_WALL_xref = [xref for xref in ocgs.keys() if ocgs[xref]["name"] == "A-WALL"][0]
doc.set_layer(-1, on=[A_WALL_xref], basestate="OFF")

doc.ez_save("test.pdf")

[JorjMcKie]

Now I appreciate that "deleting" the OCG isn't the way to go because removing the xref doesn't remove the data being referenced. (Correct?)

Even if we were able to delete undesired OCGs (and in fact I found a way to do something equivalent) the effect would be that those PDF objects that referenced a now deleted OCG become unconditionally visible!
Actually no surprise.
So the only way is to permanently set them to OFF.

Please also consider that the referencing objects (the ones I don't want to see anymore) may be referencing more than one OCG at the same time, because it is possible to combine two or more OCGs with arbitrary logical expressions like "OR" / "NOT" / "AND" ... the full boolean algebra is available here.
This means that for example an object could have the reference "show if A-SITE is ON or A-WALL is ON". If you then delete that object because it references "A-SITE" your result would be wrong.

[p-vdp]

Thank you for the detailed feedback and explaining to me how this works!

if we were able to delete undesired OCGs (and in fact I found a way to do something equivalent)

Would you be able to share that method?

Sorry - you labelled the pictures wrong:
The right side looks like the outcome of my snippet
test.pdf
The left side isn't it.

Here's a comparison of your test.pdf file opened in four different programs:

• Top left, Adobe Acrobat
• Top right, Adobe Illustrator
• Bottom left, InkScape (internal mode)
• Bottom right, InkScape (cairo mode)

The bottom left result is clearly a glitch in InkScape's internal conversion of the PDF to SVG, however, my point is that it appears if we can remove the OFF data entirely and preserve the ON data then we can avoid downstream problems like this.

As I mentioned above, Acrobat does this with the Discard hidden layer content and flatten visible layers option, which I ran on test.pdf (output here: test__acrobat_optimized.pdf) and the optimized file now displays correctly in InkScape internal mode:

Thus my question above:

Is there a way to do this in pymupdf? Perhaps one of the parameters in Document.save?

I've tried variations of doc.save() and it doesn't appear to have this capability. Is there some other way to flatten the document in the same way that Acrobat does?

[JorjMcKie]

A clear NO to your question. Our output is made correctly and authoritative PDF viewers accept the created PDF and display it as they should.

"... then we can avoid downstream problems like this."

"Downstream problems" should be addressed by those who are causing them in the first pace. The problem clearly is in InkScape and not on our side. They are the ones who owe their users a problem fix. And I encourage you to submit a bug report to them.

The Adobe feature "Discard hidden layer content and flatten visible layers" is not for free. It's just a lucky coincidence that this optimization feature avoids a situation where InkScape fails.

[p-vdp]

Fair enough! I wouldn't expect you to care about InkScape's bugs. Again I appreciate the help.