counter representations

[u-fischer]

hyperref uses in various places \arabic. Quite often it expects \arabic to expand to something "pdf safe". This can lead to warnings (and perhaps even errors) if \arabic is redefined.

See e.g. latex3/latex2e#138.

Some general investigations about counter representations are needed here.

[car222222]

I am not sure that this is just a “pdf problem” rather than something more general about ‘macros that must expand to pure text’.

[davidcarlisle]

On Tue, 26 Mar 2019 at 10:27, Chris Rowley ***@***.***> wrote: I am not sure that this is just a “pdf problem” rather than something more general about ‘macros that must expand to pure text’.

well actually it's a problem with only having user-level macros to access functionality. It's rather simpler than technical details about expansion and pdf metadata strings, if hyperref is expecting \arabic to typeset a counter using "arabic" digits and \arabic has been redefined to select typographic settings suitable for Arabic text, then things are going to go wrong.... hyperref could of course finesse this by using its own macro defined to be \expandafter\@Arabic\csname c@#1\endcsname but that still leaves other uses of \arabic open to mis-interpretation.

[u-fischer]

@davidcarlisle there are packages redefining the internal names (e.g. polyglossia and arabic redefines \@arabic, spanish redefines \@roman etc). Also the question is if hyperref needs to keep in sync with the document representation or not. E.g. if it puts a number in a bookmark, it probably should look like in the document, but for destination names it doesn't matter.

[u-fischer]

Related https://github.com/latex3/latex2e/issues/139