Follow XMP metadata recommendations from PDF specification

[stefan6419846]

Explanation

We currently use the xml.dom.minidom stdlib package to work with XMP metadata. This does not completely fulfill the recommendations from Annex H, section H.2, of the PDF 2.0 specification regarding XMP metadata:

• Document.toxml() adds the <?xml version="1.0"> header.
• Between </x:xmpmeta> and <?xpacket end="w"?>, provide "white-space padding to permit in-place updating of metadata".

Section H.7.5 has an additional note, but is possibly considered outdated:

Note that applications which fully understand PDF updating do not usually update in-place

Code Example

Exactly like in

pypdf/docs/user/metadata.md

Lines 139 to 162 in 0b64266

	from pypdf import PdfWriter
	writer = PdfWriter(clone_from="example.pdf")
	metadata = writer.xmp_metadata
	assert metadata # Ensure that it is not `None`.
	rdf_root = metadata.rdf_root
	xmp_meta = rdf_root.parentNode
	xmp_document = xmp_meta.parentNode
	# Please note that without a text node, the corresponding elements might
	# be omitted completely.
	pdfuaid_description = xmp_document.createElement("rdf:Description")
	pdfuaid_description.setAttribute("rdf:about", "")
	pdfuaid_description.setAttribute("xmlns:pdfuaid", "http://www.aiim.org/pdfua/ns/id/")
	pdfuaid_part = xmp_document.createElement("pdfuaid:part")
	pdfuaid_part_text = xmp_document.createTextNode("1")
	pdfuaid_part.appendChild(pdfuaid_part_text)
	pdfuaid_description.appendChild(pdfuaid_part)
	rdf_root.appendChild(pdfuaid_description)
	metadata.stream.set_data(xmp_document.toxml().encode("utf-8"))
	writer.write("output.pdf")

I currently use monkey-patching, which does not feel right:

def writexml(minidom_writer, indent='', addindent='', newl='', encoding=None, standalone=None):
    for _child_node in xmp_document.childNodes:
        _child_node.writexml(minidom_writer, indent, addindent, newl)


xmp_document.writexml = writexml

According to section 7.3.2 of the XMP Specifiction Part 1, we might have to use end="r" to disallow in-place modifications as we leave no whitespace:

The end="w" or end="r" portion shall be used by packet scanning processors to determine whether the XMP may be modified in-place. The end="w" form indicates writable and the end="r" form indicates read-only. The writable or read-only indication should be ignored by all “smart” (not packet scanning) processors.