Bibliography

Typst 0.15′s multiple bibliographies handle three situations.

A citation inside add/del

Nothing special to do — it becomes part of the manuscript’s own bibliography either way. The difference shows up in which version the citation appears in:

#import "../../lib.typ": *

#set page(width: 16.6cm, height: auto, margin: 12pt)
#set text(size: 10.5pt)
#set-revisions(require-exchange: false)

#added(<r1-1>)[A new claim, supported by @smith2020.]

#deleted(<r1-2>, summary: [an outdated claim])[
  An outdated claim, previously supported by @jones2019.
]

#bibliography("/typst-palimpsest/docs/manual-snippets/biblio-manuscript.bib")

Clean — only the surviving citation:

Tracked — both, since the struck-through text is still there to resolve:

A citation inside a deleted passage disappears from the bibliography in the clean compile and reappears in the tracked one, so a reader following the struck-through text can still resolve what it cited — the opposite of latexdiff, notorious for leaving phantom, unresolvable references behind.

letter-bibliography: a citation the letter makes on its own

letter-bibliography(path, title: [References cited in this response])

a second bibliography, scoped to citations made inside the letter and numbered independently from the manuscript’s own.

A real project’s three-file shape (main.typ/manuscript.typ/responses.typ, covered in Wiring a real project below) — manuscript.typ:

#import "../../../../lib.typ": *

An earlier claim, supported by @jones2019.

#added(<r1-1>)[A second claim, supported by @smith2020.]

#bibliography("/typst-palimpsest/docs/manual-snippets/biblio-manuscript.bib")

responses.typ:

#import "../../../../lib.typ": *

#reviewer(1)[
  #exchange(<r1-1>)[Please support this claim.][
    Done, see @smith2020. Also relevant, though not cited in the manuscript: @lee2022. #pinpoint(<r1-1>)
  ]
]

For the reviewer's convenience only:

#figure(
  table(columns: 2, [Group], [N]),
  caption: [Sample sizes, not shown in the manuscript.],
)

#letter-bibliography("/typst-palimpsest/docs/manual-snippets/biblio-letter.bib")

smith2020 is [2] in the manuscript’s own bibliography but [1] in the letter’s — two independent numbering sequences, not one shared list. This is also why smith2020 had to be listed in both .bib files even though the manuscript already cites it: a citation made directly in the letter’s own prose only ever resolves against the .bib passed to letter-bibliography, regardless of what the manuscript’s own bibliography already knows.

Letter numbering

A figure or table the letter adds on its own — “for the reviewer’s convenience only,” never in the manuscript, like the table above — is numbered independently from the manuscript’s, prefixed R, so it can never be confused with a manuscript one.

A figure, table, equation, or heading pinpoint(excerpt: true) re-emits from the manuscript is different: it keeps the manuscript’s own real number, not an R one — “Figure 1” reads “Figure 1” in the letter too, not “Figure R1”, and a quoted subsection heading reads “2.1” the same way it does in the manuscript:

#import "../../lib.typ": *
#import "../../src/letter.typ": with-letter-numbering

#document("manuscript.png")[
  #set page(width: 16.6cm, height: auto, margin: 12pt)
  #set text(size: 10.5pt)
  #set heading(numbering: "1.1.")

  = Introduction
  = Methods

  #passage(<r1-1>)[
    #add[
      #figure(rect(width: 3cm, height: 2cm, fill: luma(220)), caption: [A subgroup analysis, added per reviewer request.]) <fig-subgroup>
    ]
  ]

  #passage(<r1-2>)[
    #add[
      == Sensitivity analysis <sub-sensitivity>
      Added per reviewer request.
    ]
  ]
]

#document("response.png")[
  #set page(width: 16.6cm, height: auto, margin: 12pt)
  #set text(size: 10.5pt)
  #with-letter-numbering[
    #reviewer(1)[
      #exchange(<r1-1>)[Please add a subgroup analysis.][
        Done. #pinpoint(<r1-1>, excerpt: true)
      ]
      #exchange(<r1-2>)[Please add a sensitivity analysis subsection.][
        Done. #pinpoint(<r1-2>, excerpt: true)
      ]
    ]

    For the reviewer's convenience only, not in the manuscript:

    #figure(rect(width: 3cm, height: 2cm, fill: luma(150)), caption: [A figure the letter adds on its own.])
  ]
]

Manuscript:

Response:

The letter’s own figure above is captioned “Figure R2,” not “Figure R1” — the two excerpts before it (the figure and the heading) still occupy the first two slots of the letter’s own counters, even though what each displays is the manuscript’s own number rather than that slot’s.

This requires a label on the original — query() is how the letter’s copy finds the manuscript’s real number, and a label is what it queries by. An unlabelled figure, equation, or heading just falls back to whatever numbering already applies where the excerpt is re-emitted — the letter’s own R sequence for a figure or table, no R sequence at all for an equation (equations never had one), and no number at all for a heading. This matters most in practice for headings: a figure is usually labelled anyway, for @ref, but a heading rarely is unless an author adds one specifically so it can be quoted this way.