Bug Tracker

Hi Nicola,

I'm testing here version 1.48b, as you requested and, taking things in parts, a comment on this issue.

I see that this is already handled by the new \glsxtrsetlongfirstuse, which indeed works for the particular case reported.

Consider, however, the following document:

\documentclass{article}

\usepackage[
  abbreviations,
  shortcuts=abbr,
]{glossaries-extra}

\glssetcategoryattribute{abbreviation}{discardperiod}{true}
\glssetcategoryattribute{abbreviation}{retainfirstuseperiod}{true}

\newabbreviation[user1=Docteur]{dr}{Dr.}{Doctor}

\newabbreviation
  [user1={German Speaking \TeX\ User Group}]
  {dante}{DANTE e.V.}{Deutschsprachige Anwendervereinigung \TeX\ e.V.}

\begin{document}

% This is OK.
\ab{dr}. \ab{dr}.

% This is still OK.
\as{dr}.

% This period no longer gets unduly discarded.
\al{dr}.

% But this one still does.
\glsuseri{dr}.

% And this one doesn't, when it should.
\al{dante}.

\end{document}

When I originally approached this problem, my main attempt was to find out what the field being printed was at the post link hook. But, as far as I can tell, this is not tracked at all. And, even if it where, the "mixed cases" case would still beat us. This might be dealt with by category attributes, but I doubt it would be much practical. I eventually went with signalling from the entry field itself in [TeX.SX Link] and, at this point, I think this is the only way to make things flexible enough. True, as I've implemented it, "from the outside", its a little fragile, but it need not be on your side. On the other hand, true, breaking UI change...

My general comment here is that I think you'd be better off in the longer term by a more general approach, instead of fighting the uphill battle of adding one exception at a time.

Granted, though, of course, that the long form is probably the most frequent case where this is a problem, so the current fix is an improvement, for which I thank you.

Best regards,
Gustavo.

Thank you for your feedback. One possibility is to allow the value of the discardperiod attribute to be a comma-separated list of fields to which the setting should apply. The "true" value could mean only with the \gls-like or abbreviation commands, although that might cause a problem with backward-compatibility. I could allow "all" as a value to behave as in previous versions.

The post-link hook can pick up whether it's following \gls-like or \glstext-like commands with \glsxtrifwasglslike. I would then need to add another placeholder command to pick up the field name for the \glstext-like commands.

This sounds great! It would require a possible multiplication of categories (...-dotted, ...-doubledotted), but it seems both more robust and more manageable regarding backward compatibility than the signalling approach. (Btw, has anyone already suggested you that category be a list? I won't do it now, 'cause I'm afraid to have spent all my "pestering quota" for a good stretch of time :-).

There are some instances where control sequence names that include the entry's category name need to expand, and that wouldn't work with a list of categories. One possibility might be to have a main category and subsidiary categories for minor adjustments, but that would definitely have to wait for another version.

Hi Nicola,
and yet again, thank you! :-)
Gustavo.