The woes of referencing non-verse-text paragraphs (Right-click on PDF viewer in glossary only alters adj-list)

[davidg-sil]

In checking my reply to this question, I found that while GLO k.Templo +1 type adjust list entries are correctly generated, there doesn't seem to be any changes.txt type modification of the paragraph style.
Probably the real fix involves parser updates, but at present the solution is to hand-craft entries like:

at GLO "\\p([^\\]*\\k Templo\\*)" > "\\p^95 \1"

The above does not work, of course, if it is applied before verse-number removal (if the PT version of the glossary is split into chapters/verses, even when the final USFM is not).

The fact that there are verse numbers in some glossaries, (stopping paratext's notes in the glossary from getting very resource-intensive, as the entire glossary is copied in as context without verses) and indeed, other milestones also affect what TeX determines as the paragraph id, no doubt makes things complex for the parser.

Regarding milestones: Like verses and \k keywords, stand-alone milestones acting as trigger points reset the current reference position to their default argument (which is how \zfiga functions). Thus the current code in TeX means that if a user milestone has not been declared \NotTriggerPoint{mymilestone}, then it behaves like \zfiga, which can interfere with paragraph identification.

It is not clear to me how the parser / python code can be expected to accurately guess whether a given new marker is or is not being used as a trigger-point.
Perhaps we should invert the logic, making \NotTriggerPoint a no-op (or trigger a deprecation warning?) and instead define \IsTriggerPoint{} to mark special milestones that should generate figure-anchors, etc. (Or indeed, make it so that only \zfiga has this behaviour of setting a new reference position).