Thanks for sharing the module! I had a read through your DJVUpage code.

I think ProofreadPage can actually do this itself already in PHP - the Pagination
and PageNumber classes do the same pagelist label inference internally, so instead
of porting the parser to Lua we could just expose it directly. Something like
mw.ext.proofreadPage.getPageLabels( index ) that returns a { position = label }
table using the exact same logic the extension uses to render <pagelist>.

Two quick things before I start - do you want the formatted label (what shows in
the page header) or the raw value, and would you rather it be keyed by position
number or by the Page: title?

Happy to write the patch once I know which fits your use case :)

Thanks to you for such a positive response!

1- I don't understand the difference between the formatted and the raw value. Can you give me an example?

2- I think position number is the simplest and more intuitive way to deal with the table. I can think of a special case with Index pages without <pagelist> such as collections of JPG files (I think de.wikisource uses a lot of those). Maybe it's best to give both options.

3- Consider making them methods of the Index object. Additional auxilliary methods such as "get page number from label", "get page object from label" would be appreciated.

I'm very happy with the result in my home WS.

Thanks so much for the detailed reply, this really helps

On the formatted vs raw thing - I actually went and looked at the code, and the
two are almost the same. For normal numbers, roman numerals, localised digits and
text labels like "Cover" they give you the exact same string. The only real
difference is in folio / recto-verso mode where the formatted one adds the
recto/verso styling. So for what you want (finding the file page from a label) it
honestly barely matters. One small thing I noticed is that the formatted version
escapes its output since it is meant for rendering on the page, so for a Lua API
we would probably want a plain label instead. I will take care of that when I
write it.

Position keying works great. And good news on the JPG collections without a
pagelist - ProofreadPage already has a separate file based pagination for those,
and the method I have in mind uses that same layer underneath, so it should just
work for both cases without any special handling.

The reverse lookups (page number from label, page object from label) and putting
these on the Index object are really nice ideas. Since this adds new public Lua
API to ProofreadPage though, I would rather start small with just the core
getPageLabels method and check the overall API shape with the ProofreadPage
maintainers, then add the label to number and label to page helpers as follow ups.
That way each change stays small and easy to review. Does that sound okay to you

Really glad it is already working well on es.wikisource :)

Please go on then

Change #1317675 had a related patch set uploaded (by Zaidusyy; author: Zaidusyy):

[mediawiki/extensions/ProofreadPage@master] Add getPageLabels() to the ProofreadPage Lua library

https://gerrit.wikimedia.org/r/1317675