Pandoc   a universal document converter

Org-mode features and differences

Pandoc’s handling of org files is similar to that of Emacs org-mode. This document aims to highlight the cases where this is not possible or just not the case yet.

Export options

The following export keywords are supported:

  • AUTHOR: comma-separated list of author(s); fully supported.

  • CREATOR: output generator; passed as plain-text metadata entry creator, but not used by any default templates.

  • DATE: creation or publication date; well supported by pandoc.

  • EMAIL: author email address; passed as plain-text metadata field email, but not used by any default templates.

  • LANGUAGE: document language; included as plain-text metadata field lang. The value should be a BCP47 language tag.

  • SELECT_TAGS: tags which select a tree for export.

  • EXCLUDE_TAGS: tags which prevent a subtree from being exported. Fully supported.

  • TITLE: document title; fully supported.

  • EXPORT_FILE_NAME: target filename; unsupported, the output defaults to stdout unless a target has to be given as a command line option.

Pandoc tries to be compatible with org-mode when exporting an org document. If you find some behavior confusing, please do refer to org-mode Export-Settings documentation. For example, a common confusion (#3214, #5169, #6145, #7236) is treatment of headers with level > 3 differently because org-mode sets org-export-headline-levels (configurable with #+OPTIONS: H:3) to 3 by default.

Format-specific options

Emacs Org-mode supports additional export options which work for specific export formats. Some of these options’ behavior differs in Org-mode depending on the output format, while pandoc is format-agnostic when parsing; differences are noted where they occur.

  • DESCRIPTION: the document’s description; pandoc parses this option as text with markup into the description metadata field. The field is not used in default templates.

    Pandoc follows the LaTeX exporter in that it allows markup in the description. In contrast, the Org-mode HTML exporter treats the description as plain text.

  • LATEX_HEADER and LATEX_HEADER_EXTRA: arbitrary lines to add to the document’s preamble. Contrary to Org-mode, these lines are not inserted before the hyperref settings, but close to the end of the preamble.

    The contents of this option are stored as a list of raw LaTeX lines in the header-includes metadata field.

  • LATEX_CLASS: the LaTeX document class; like Org-mode, pandoc uses article as the default class.

    The contents of this option are stored as plain text in the documentclass metadata field.

  • LATEX_CLASS_OPTIONS: Options for the LaTeX document class; fully supported.

    The contents of this option are stored as plain text in the classoption metadata field.

  • SUBTITLE: the document’s subtitle; fully supported.

    The content of this option is stored as inlines in the subtitle metadata field.

  • HTML_HEAD and HTML_HEAD_EXTRA: arbitrary lines to add to the HTML document’s head; fully supported.

    The contents of these options are stored as a list of raw HTML lines in the header-includes metadata field.

Pandoc-specific options

Pandoc recognizes some export options not used by Emacs Org.

  • NOCITE: this field adds the listed citations to the bibliography, without the need to mention them to the text. The special value @* causes all available references to be added the bibliography.

  • HEADER-INCLUDES: like HTML_HEAD and, LATEX_HEADER, but treats the option’s value as normal text with markup.

  • INSTITUTE: Affiliation of the author; the value is read as text with markup and is stored in the institute metadata field. The field is included by default on the title slide of beamer presentations.

Other options

Any export option or directive not listed above has no effect when parsing with pandoc. However, the information is retained as a raw block. It can be accessed through a filter and will be included in org output.

Directives as metadata

As an example, we will restore an old behavior of pandoc versions prior to 2.10. Unknown keywords were treated as variable definitions, and were added the document’s metadata. Typing #+key: value in the org-file used to have the same effect as running pandoc with the --metadata key=value option.

Since pandoc 2.10, each unhandled line starting with #+ is kept internally as a raw block with format org. This block can be inspected and processed by a filter. Below is a Lua filter which converts these unhandled lines into metadata key-value pairs.

-- intermediate store for variables and their values
local variables = {}

--- Function called for each raw block element.
function RawBlock (raw)
  -- Don't do anything unless the block contains *org* markup.
  if raw.format ~= 'org' then return nil end

  -- extract variable name and value
  local name, value = raw.text:match '#%+(%w+):%s*(.+)$'
  if name and value then
    variables[name] = value

-- Add the extracted variables to the document's metadata.
function Meta (meta)
  for name, value in pairs(variables) do
    meta[name] = value
  return meta


Emacs org-mode lacks an official citation syntax, leading to multiple syntaxes coexisting. Pandoc recognizes four different syntaxes for citations.

Citation support for org-mode is enabled by default. Support can be toggled off by disabling the citation extension; e.g. pandoc --from=org-citations.

Berkeley-style citations

The semi-official Org-mode citation syntax was designed by Richard Lawrence with additions by contributors on the emacs-orgmode mailing list. It is based on John MacFarlane’s pandoc Markdown syntax. It’s dubbed Berkeley syntax due the place of activity of its creators, both philosophers at UC Berkeley.

Simple in-text citation

This is the simplest form of citation. It consists of the citation ID prefixed by ‘@’.


@WatsonCrick1953 showed that DNA forms a double-helix.

In-text citation list

Citations presented in the text unparenthesized are called in-text citations. The syntax for these citations is


where the initial PREFIX and final SUFFIX are optional. At least one INDIVIDUAL-REFERENCE must be present. The colon and semicolons here are literal and indicate the end of the TAG and the end of a PREFIX or INDIVIDUAL-REFERENCE respectively.



The KEY is obligatory, and the prefix and suffix are optional.

A PREFIX or SUFFIX is arbitrary text (except ;, ], and citation keys).


[cite: See; @Mandelkern1981; and @Watson1953]

Parenthetical citation

Citations surrounded by parentheses. The syntax is identical to in-text citations, except for the additional parentheses enclosing the initial cite tag.

[(cite): See; @Mandelkern1981; and @Watson1953]

org-ref citations

The org-ref package by John Kitchen is in wide use to handle citations and has excellent tooling support in Emacs. Its citation syntax is geared towards users in the natural sciences but still very flexible regardless.

[[citep:Dominik201408][See page 20 of::, for example]]

Pandoc-Markdown-like syntax

Historically, Markdown-style citations syntax was the first that was added to pandoc’s org reader. It is close to Markdown’s citation syntax.

Citations go inside square brackets and are separated by semicolons. Each citation must have a key, composed of ‘@’ plus the citation identifier from the database, and may optionally have a prefix, a locator, and a suffix. The citation key must begin with a letter, digit, or _, and may contain alphanumerics, _, and internal punctuation characters (:.#$%&-+?<>~/). Here are some examples:

Simple citation

The simplest method to insert a citation is to write the citation ID prefixed by ‘@’.


[prefix @citekey suffix]
[see @doe2000 pp. 23-42]
[@doe2000 p. 5; to a lesser extend @doe2005]


Use normal latex citation commands like \cite{x} or \citet{y}.


Pandoc supports normal org tables (sometimes called “pipe tables”) and grid tables (tables created by table.el).

Column widths

Org mode tables don’t allow line-breaks within cells, and lines which contain text can get very long. This often leads to tables which run off the page when exporting, especially when exporting to PDF via LaTeX. Overlong lines in the source text are this is usually hidden by setting a column width, but the default Emacs exporters ignore that setting. Pandoc deviates from Emacs’s behavior and uses this information to resize the table columns when exporting.


There is no support yet for cells spanning multiple columns or rows. The table.el grid tables allows rowspans and colspans and so does pandoc’s internal structure since 2.10, but the parser has not been updated yet.

Emphasis rules

Org-mode uses complex rules to decide whether a string represents emphasized text. In Emacs, this can be customized via the variable org-emphasis-regexp-components. A variable like this doesn’t fit well with pandoc’s model. Instead, it is possible to use special lines to change these values:

#+pandoc-emphasis-pre: "-\t ('\"{"
#+pandoc-emphasis-post: "-\t\n .,:!?;'\")}["

The above describes the default values of these variables. The arguments must be valid (Haskell) strings. If interpretation of the argument as string fails, the default is restored.

Changing emphasis rules only affect the part of the document following the special lines. They must be some of the first lines to alter parsing behavior for the whole document. It is also possible to change the values temporarily for selected sections only. The string test in the following snippet will be read as emphasized text, while the rest of the document will be parsed using default emphasis rules:

#+pandoc-emphasis-pre: "["
#+pandoc-emphasis-post: "]"

smart extension

Org-mode allows to insert certain characters via special character sequences. For example, instead of typing the Unicode /HORIZONTAL ELLISPIS/ character by hand, one can instead type tree dots .... En dashes and em dashes can be written as -- and --- respectively. Furthermore, quotation marks (") and apostrophe-quotes (') can be treated in a “smart” way, potentially replacing them with proper, language specific unicode quotation characters.

Like in Markdown, these behaviors can be turned on all-at-once by enabling the smart extension. However, disabling smart (the default) will not necessarily disable smart quotes and special strings. Instead, it will just result in the default Org mode behavior.

The special string feature can be turned off via the #+OPTIONS: -:nil export setting. There are currently no command line flags which control these features. As a workaround, one can use process substitution, a feature supported by most shells. It allows to provide the options line on the command line:

pandoc -f org <(printf "#+OPTIONS: -:nil\n") …

Currently unsupported features

Library of babel

The library of babel translates between various programming languages. This is out-of-scope for pandoc. Use Emacs to run code, then feed the resulting org file to pandoc.