Nicolas Goaziou <m...@nicolasgoaziou.fr> writes:
>> I ask because in that kind of context, I think it is generally going to
>> be more useful to deal with citation objects as a whole. I am not sure
>> we will want to treat citation-references as individual objects which
>> are themselves exported; instead, I think we will want to handle
>> exporting the citation-references in a citation all at once.
>
> I don't know. In any case, they need to be treated as regular object
> (e.g., they are expected to have a filter associated to them).
Hmm. I can see that making sense if the filter manipulates the
citation-reference object itself, but I can't see it making sense if the
idea is that the exporter transcodes citation-references individually,
and the filter manipulates the transcoded string.
For example, suppose LaTeX export works by mapping a citation object to
a biblatex command, and consider this citation:
[cite: See @Doe99 ch. 1; @Doe2000 p. 10]
This should be transcoded in the final output to something like:
\textcite[See][ch. 1]{Doe99}[][p. 10]{Doe2000}
in which case the transcoded string corresponding to the first
citation-reference object is "[See][ch. 1]{Doe99}". It's hard for me to
see how you could do anything useful with just this string, and it might
even be dangerous to manipulate it, since it is only part of a complete
command, which you might mess up by modifying the string.
In other backends, there might not even be a syntactically-identifiable
chunk of the output citation which corresponds to a given
citation-reference. (I'm having trouble thinking of a good example at
the moment, but I'm thinking of cases where e.g. CSL processing turns a
reference into "Ibid.", or some other output that only has meaning
relative to its place in the document and doesn't correspond in a
context-free way to the citation-reference object.)
But like I said, it makes more sense if the filter can manipulate the
citation-reference object itself, e.g. by setting additional properties
or modifying the prefix and suffix strings *before* the object is
transcoded as part of a citation.
> If some back-end doesn't need to export directly citation references, it
> just needs to skip the relative translator.
Sure. I guess I'm just having a hard time seeing why a backend would
ever need/want to translate citation-references individually, rather
than generating a complete citation all at once. That seems to be the
way to do it in both the LaTeX and CSL worlds, anyway.
Best,
Richard
