Replace some super-simple stylesheets

[rbeezer]

Some very simple XSL stylesheets are used to extract identifiers for various objects. They could be accomplished with lxml in the Python routines, rather than having a stylesheet.

The one trick will be support for the -r flag, which sets a subtree for the processing. It's hidden behind an import and an xsl:apply-imports. Perhaps lxml has a nice routine for starting from some interior node.

[StevenClontz]

Sure, any Element in lxml supports .find and .xpath. So we can use whatever we use to set the subtree to grab the corresponding Element (the root of that subtree) and run things from there.

[StevenClontz]

Wait I think I'm wrong there. Reading closer maybe .xpath is only supported for ElementTree (the full processed XML document). So we'd have to build the right xpath, maybe, naively, xpath_for_subtree//xpath_for_extraction.

[StevenClontz]

Or if the query is simple enough to express in ElementPath, we can grab the Element and use element.iterfind(element_path).

[rbeezer]

Its provided via the argument as the value of an @xml:id which XSLT gets easily via its id() function. I'd guess there is an analog? But I don't recall seeing it.

[StevenClontz]

Something like this would use existing XSL without writing to a file I think:

Starting after https://github.com/PreTeXtBook/pretext/blob/master/pretext/pretext.py#L1157 (using YouTube thumbnails as an example):

doc_tree = ET.parse(xml_source)
if xmlid_root is None:
    subtree = ""
else:
    subtree = xmlid_root
xsl_tree = ET.parse(extraction_xslt, subtree=f"'{subtree}'")
thumbs = [t.strip() for t in doc_tree.xslt(xsl_tree).split("\n")]

But this would eliminate XSL altogether maybe:

doc_tree = ET.parse(xml_source)
xpath = "//video[@youtube]"
if xmlid_root is not None:
    xpath = f"id({xmlid_root})){xpath}"
thumb_elements = doc_tree.xpath(xpath)
thumbs = [t.text.split(",")[0].strip() for t in thumb_elements]

[rbeezer]

Looks great - thanks. Maybe I'll experiment with MOM where it is not too complicated and I can't do much harm. ;-)

[rbeezer]

We have to be a little careful. The publisher file will perform certain manipulations, like inserting "customizations" or deleting material to make "versions." Mostly these should not change what is being extracted, but it can't be ruled out in every case, I think. These simple stylesheets should be doing the pre-processing, making them not so super-simple anymore. :-(

There is a stylesheet to make "enhanced source", so we could parse that instead, once constructed. Either way, I'd like to have a better @xml:id scheme in place (which is planned, but not imminent).