Capitalised author-in-text variant (\Textcite support)

[tytyvillus]

Sometimes, it's necessary to manually specify that an author-in-text citation is sentence initial. This comes up with prefixed names, e.g. 'van Gogh', where the correct treatment is

Van Gogh (1888) was a great painter.

but

This was painted by van Gogh (1888).

In biblatex, this is already handled through the capitalised variant command, \Textcite. Would it be possible to access these capitalised variants through a new markdown syntax?

To my understanding (my Haskell isn't excellent) this could be achieved fairly painlessly by adding a new citation mode, InitAuthorInText.

The question is then what markdown syntax to use for this. Some options are:

• *@vangogh1888 aligns with fignos, but could interfere with their parser
• -@vangogh1888 has the advantage of not clashing with fignos, as the sequence -@ is currently parsed but ignored
• ^@vangogh1888 is more transparent, but could cause trouple with footnotes
• @@vangogh1888 as duplication suggests importance

I don't know how difficult it would be to apply this to other citation processors (citeprod, natbib), but even just having this for the biblatex output would be very useful. It also goes hand-in-hand with #2335, which provided parsing support for \Cite and \Textcite.

(I apologise if this is already an existing issue: I couldn't find it in an hour of searching.)

[jgm]

One option (without changes to pandoc) would be to put such citations in a special span, then use a filter to capitalize them.

[@vangogh:self]{.initial} said...

[tytyvillus]

Thanks for your reply! Just to clarify how that would work: I'd write a filter that looks for spans of class initial, and replace those with \Textcite{}? What does the :self do in your example?

I haven't worked much with pandoc filters, I'd have to look into them more deeply to understand better.

[jgm]

:self is just part of the sample citation id here. I should have used vangogh1888 as you did.

The filter would match on a Span, check for the class "initial", and if found then perform a transformation on the Span's contents, capitalizing the first letter. (This is a little tricky actually; you need to use walk with a top-down traversal, capitalize the first Str element you come to, skipping the ( or other punctuation, and then exit early from the traversal.) Possibly an LLM could write this for you.

[jgm]

A more thoroughgoing fix would require changes both to pandoc and citeproc. Ideally there could be some way of passing this information through, perhaps via an attribute.

[tytyvillus]

I've looked into this a little bit, and I think that I should be able to write the code the way that you suggest (it's a good opportunity to learn Haskell a little better!). Just a couple of points:

• The method you suggest should work fine with the --citeproc option (barring a minor annoyance with citeproc dropping particles by default). This just involves matching on

  Span ("",["init"],[]) [Cite 
    (Citation {citationMode = AuthorInText}) : _
    (Str x) : _
    ]

  and replacing the first alphabetic element of x with its uppercase version.

• I'm a bit confused about how to do it with the --biblatex option, which is the case I'm interested in. As far as I can understand, the \textcite{} commands never appear in the pandoc AST. They seem to be generated by the biblatex filter, which if I understand correctly operates on the level AST -> latex. How do I access the right step to capitalise the first letter of the \textcite command?

Test case:

File test.md:

---
references:
- author:
  - non-dropping-particle: van
    family: Gogh
    given: Vincent
  id: gogh
  issued: 1888
  language: nl
  title: Wat schilderij
  type: manuscript
- author:
  - non-dropping-particle: de la
    family: Fontaine
    given: Jean
  id: font
  issued: 1820
  language: fr
  title: Diverses fables
  type: book

...
 
Read an article by @font. [@font]{.init} is a prolific author. 

Cheese is excellent [@gogh, @font].

Then trying to access the AST with pandoc test.md -t native --biblatex gives no mention of \textcite. If instead I try to access the latex version:

$ pandoc test.md -t latex --biblatex
Read an article by \textcite{font}. {\textcite{font}} is a prolific
author.

Cheese is excellent \autocite[\textcite{font}]{gogh}.

Which is (i) weird behaviour for the multiple citation, but more importantly (ii) the Span has turned into a pair of {}. Could you give me any pointers on how I should go about this?

Thank you very much & sorry for the inconvenience!

[jgm]

Spans in LaTeX are rendered as groups ({..}).

Okay, if you're using --biblatex, then you can achieve this with a fairly simple Lua filter. The filter would match on a Cite with mode AuthorInText inside a Span with class 'init', and replace it with a RawInline "latex" element containing \Textcite{CITEID}.

[jgm]

The \textcite is generated by the LaTeX writer, of course. Your Lua filter would be replacing the AST element that gets rendered this way with pre-rendered LaTeX that would just be passed through.

[jgm]

ANd what you want is [@gogh; @font]. Multiple citations must be separated by semicolons.

[tytyvillus]

Thank you for taking the time to explain this! I'll give this method a try.

(100% my mistake on [@gogh, @font], oops.)