Regular expressions to move punctuation from after to before superscripts

edit

In my tests on this intriguing problem, I have run up against an obstacle for each of the most obvious solutions:

• when using built-in regex search for footnote anchor numbers, [0-9] does not find them;
• and although AltSearch's [::Footnote::] search does find them, this kind of search cannot be used with "back references".

I have tested a solution that works OUTSIDE of LibreOffice. It requires:

1. Saving the file as an *.fodt, or "flat" XML file; and
2. Using a text editor with regex capabilities to modify that "flat" XML file.
3. A "clean" source file, since the chances of success VERY MUCH depend on the nature of the code in this file.

I have put a copy of my test fodt file on Google Drive if you would like to see or use it. It's just "Lorem ipsum" with strategically placed footnotes, mostly before the set ,;:.!?, but also a couple that should not be changed.

My text editor of choice is the lovely Textadept (cross platform) which uses the TRE regex engine. Loading my test-fn-punct0.fodt file in Textadept, and using this regular expression search string:

(<text:note .*text:note-class="foot.*<text:note-body>\n.*</text:note>)([,;:\.!?])

(including the \n code is essential) with this as the replace phrase:

\2\1

converts all the pre-punctuation footnotes into post-punctuation footnotes. This file can then be re-opened in LibreOffice Writer for further editing, or "save-as" .odt or whatever you like. The search in the editor looks like this:

Caveat - I have tried this with a "real world" academic article, and ... the source code in the fodt file was a mess. There were loads and loads of <text:span text:style-name="Tnn"> ... </text:span> wrappers, and I have no idea where they came from or how they're being used.¹ But it means that my regex, above, is pretty useless for this file. If there was a way of cleaning out those "text:span" wrappers (which are everywhere!), then it would be possible.

I don't know if this is any use for @catbill, but it is at least one "solution".

¹ A little more investigating shows these to be related to officeooo:rsid numbers and autocreated "styles" -- for a description of the "issue" see fdo#68183 -- I have posted a Q&A about them which offers a bit of help for getting rid of them.

This is how I see it. Because the LO search engine applies formatting to the whole search string, you cannot search for a superscript followed by anything not superscript. Therefore that is out. If the trailing character was superscript, you could use this:

With Regex enabled,

Search string: (.*)([\.?!;,])

search string properties: None, superscript, automatic

Replace string: $2$1

But that won't work in your case, because it isn't.

Now, there may be a workaround. If the only numbers in parentheses and followed by punctuation are the references you want to hit, you could use this search:

Search string: (\([0-9,-]*\))([\.?!,;])

Replace string: $2$1

[x] Regex enabled.

But if there are other numbers in parentheses followed by those punctuations, then it will catch them as well.

Unless I'm missing something, or someone knows how to conjure the Alternative Search engine to do this, the only other way is by macro.