Pandoc   a universal document converter

Pandoc Lua Filters

Introduction

Pandoc has long supported filters, which allow the pandoc abstract syntax tree (AST) to be manipulated between the parsing and the writing phase. Traditional pandoc filters accept a JSON representation of the pandoc AST and produce an altered JSON representation of the AST. They may be written in any programming language, and invoked from pandoc using the --filter option.

Although traditional filters are very flexible, they have a couple of disadvantages. First, there is some overhead in writing JSON to stdout and reading it from stdin (twice, once on each side of the filter). Second, whether a filter will work will depend on details of the user’s environment. A filter may require an interpreter for a certain programming language to be available, as well as a library for manipulating the pandoc AST in JSON form. One cannot simply provide a filter that can be used by anyone who has a certain version of the pandoc executable.

Starting with pandoc 2.0, we have made it possible to write filters in lua without any external dependencies at all. A lua interpreter and a lua library for creating pandoc filters is built into the pandoc executable. Pandoc data types are marshalled to lua directly, avoiding the overhead of writing JSON to stdout and reading it from stdin.

Here is an example of a lua filter that converts strong emphasis to small caps:

or equivalently,

This says: walk the AST, and when you find a Strong element, replace it with a SmallCaps element with the same content.

To run it, save it in a file, say smallcaps.lua, and invoke pandoc with --lua-filter=smallcaps.lua.

Here’s a quick performance comparison, using a version of the pandoc manual, MANUAL.txt, and versions of the same filter written in compiled Haskell (smallcaps) and interpreted Python (smallcaps.py):

Command Time
pandoc MANUAL.txt 1.01s
pandoc MANUAL.txt --filter ./smallcaps 1.36s
pandoc MANUAL.txt --filter ./smallcaps.py 1.40s
pandoc MANUAL.txt --lua-filter ./smallcaps.lua 1.03s

As you can see, the lua filter avoids the substantial overhead associated with marshalling to and from JSON over a pipe.

Lua filter structure

Lua filters are tables with element names as keys and values consisting of functions acting on those elements.

Filters are expected to be put into separate files and are passed via the --lua-filter command-line argument. For example, if a filter is defined in a file current-date.lua, then it would be applied like this:

pandoc --lua-filter=current-date.lua -f markdown MANUAL.txt

The --lua-filter option may be supplied multiple times. Pandoc applies all filters (including JSON filters specified via --filter and lua filters specified via --lua-filter) in the order they appear on the command line.

Pandoc expects each lua file to return a list of filters. The filters in that list are called sequentially, each on the result of the previous filter. If there is no value returned by the filter script, then pandoc will try to generate a single filter by collecting all top-level functions whose names correspond to those of pandoc elements (e.g., Str, Para, Meta, or Pandoc). (That is why the two examples above are equivalent.)

For each filter, the document is traversed and each element subjected to the filter. Elements for which the filter contains an entry (i.e. a function of the same name) are passed to lua element filtering function. In other words, filter entries will be called for each corresponding element in the document, getting the respective element as input.

The return of a filter function must one of the following:

  • nil: this means that the object should remain unchanged.
  • a pandoc object: this must be of the same type as the input and will replace the original object.
  • a list of pandoc objects: these will replace the original object; the list is merged with the neighbors of the original objects (spliced into the list the original object belongs to); returning an empty list deletes the object.

The function’s output must result in an element of the same type as the input. This means a filter function acting on an inline element must return either nil, an inline, or a list of inlines, and a function filtering a block element must return one of nil, a block, or a list of block elements. Pandoc will throw an error if this condition is violated.

If there is no function matching the element’s node type, then the filtering system will look for a more general fallback function. Two fallback functions are supported, Inline and Block. Each matches elements of the respective type.

Elements without matching functions are left untouched.

See module documentation for a list of pandoc elements.

Global variables

Pandoc passes additional data to Lua filters by setting global variables.

FORMAT
The global FORMAT is set to the format of the pandoc writer being used (html5, latex, etc.), so the behavior of a filter can be made conditional on the eventual output format.
PANDOC_READER_OPTIONS
Table of the options which were provided to the parser.
PANDOC_VERSION
Contains the pandoc version as a numerically indexed table, most significant number first. E.g., for pandoc 2.1.1, the value of the variable is a table {2, 1, 1}. Use table.concat(PANDOC_VERSION, '.') to produce a version string. This variable is also set in custom writers.
PANDOC_API_VERSION
Contains the version of the pandoc-types API against which pandoc was compiled. It is given as a numerically indexed table, most significant number first. E.g., if pandoc was compiled against pandoc-types 1.17.3, then the value of the variable will be a table {1, 17, 3}. Use table.concat(PANDOC_API_VERSION, '.') to produce a version string from this table. This variable is also set in custom writers.
PANDOC_SCRIPT_FILE
The name used to involve the filter. This value can be used to find files relative to the script file. This variable is also set in custom writers.

Pandoc Module

The pandoc lua module is loaded into the filter’s lua environment and provides a set of functions and constants to make creation and manipulation of elements easier. The global variable pandoc is bound to the module and should generally not be overwritten for this reason.

Two major functionalities are provided by the module: element creator functions and access to some of pandoc’s main functionalities.

Element creation

Element creator functions like Str, Para, and Pandoc are designed to allow easy creation of new elements that are simple to use and can be read back from the lua environment. Internally, pandoc uses these functions to create the lua objects which are passed to element filter functions. This means that elements created via this module will behave exactly as those elements accessible through the filter function parameter.

Exposed pandoc functionality

Some pandoc functions have been made available in lua:

  • walk_block and walk_inline allow filters to be applied inside specific block or inline elements;
  • read allows filters to parse strings into pandoc documents;
  • pipe runs an external command with input from and output to strings;
  • the pandoc.mediabag module allows access to the “mediabag,” which stores binary content such as images that may be included in the final document;
  • the pandoc.utils module contains various utility functions.

Lua interpreter initialization

The way the Lua interpreter is set-up can be controlled by placing a file init.lua in pandoc’s data directory. The default init file loads the pandoc and pandoc.mediabag modules:

A common use-case would be to add code to load additional modules or to alter default modules. E.g., the following snippet adds all unicode-aware functions defined in the text module to the default string module, prefixed with the string uc_.

This makes it possible to apply these functions on strings using colon syntax (mystring:uc_upper()).

Examples

Macro substitution.

The following filter converts the string {{helloworld}} into emphasized text “Hello, World”.

Default metadata file

This filter causes metadata defined in an external file (metadata-file.yaml) to be used as default values in a document’s metadata:

Setting the date in the metadata

This filter sets the date in the document’s metadata to the current date:

This filter prints a table of all the URLs linked to in the document, together with the number of links to that URL.

Replacing placeholders with their metadata value

Lua filter functions are run in the order

Inlines → Blocks → Meta → Pandoc.

Passing information from a higher level (e.g., metadata) to a lower level (e.g., inlines) is still possible by using two filters living in the same file:

If the contents of file occupations.md is

then running pandoc --lua-filter=meta-vars.lua occupations.md will output:

Modifying pandoc’s MANUAL.txt for man pages

This is the filter we use when converting MANUAL.txt to man pages. It converts level-1 headers to uppercase (using walk_block to transform inline elements inside headers), removes footnotes, and replaces links with regular text.

Creating a handout from a paper

This filter extracts all the numbered examples, section headers, block quotes, and figures from a document, in addition to any divs with class handout. (Note that only blocks at the “outer level” are included; this ignores blocks inside nested constructs, like list items.)

Counting words in a document

This filter counts the words in the body of a document (omitting metadata like titles and abstracts), including words in code. It should be more accurate than wc -w run directly on a Markdown document, since the latter will count markup characters, like the # in front of an ATX header, or tags in HTML documents, as words. To run it, pandoc --lua-filter wordcount.lua myfile.md.

Converting ABC code to music notation

This filter replaces code blocks with class abc with images created by running their contents through abcm2ps and ImageMagick’s convert. (For more on ABC notation, see http://abcnotation.com.)

Images are added to the mediabag. For output to binary formats, pandoc will use images in the mediabag. For textual formats, use --extract-media to specify a directory where the files in the mediabag will be written, or (for HTML only) use --self-contained.

Building images with tikz

This filter converts raw LaTeX tikz environments into images. It works with both PDF and HTML output. The tikz code is compiled to an image using pdflatex, and the image is converted (if necessary) from pdf to png format using ImageMagick’s convert, so both of these must be in the system path. Converted images are cached in the working directory and given filenames based on a hash of the source, so that they need not be regenerated each time the document is built. (A more sophisticated version of this might put these in a special cache directory.)

Example of use:

pandoc --lua-filter tikz.lua -s -o cycle.html <<EOF
Here is a diagram of the cycle:

\begin{tikzpicture}

\def \n {5}
\def \radius {3cm}
\def \margin {8} % margin in angles, depends on the radius

\foreach \s in {1,...,\n}
{
  \node[draw, circle] at ({360/\n * (\s - 1)}:\radius) {$\s$};
  \draw[->, >=latex] ({360/\n * (\s - 1)+\margin}:\radius)
    arc ({360/\n * (\s - 1)+\margin}:{360/\n * (\s)-\margin}:\radius);
}
\end{tikzpicture}
EOF

Module text

UTF-8 aware text manipulation functions, implemented in Haskell. These are available to any lua filter. However, the module must be explicitly loaded:

lower (s)

Returns a copy of a UTF-8 string, converted to lowercase.

upper (s)

Returns a copy of a UTF-8 string, converted to uppercase.

reverse (s)

Returns a copy of a UTF-8 string, with characters reversed.

len (s)

Returns the length of a UTF-8 string.

sub (s)

Returns a substring of a UTF-8 string, using lua’s string indexing rules.

Module pandoc

Lua functions for pandoc scripts.

Pandoc Document

Pandoc (blocks[, meta])

A complete pandoc document

Parameters:

blocks:
document content
meta:
document meta data

Meta

Meta (table)

Create a new Meta object.

Parameters:

table:
table containing document meta information

MetaValue

MetaBlocks (blocks)

Meta blocks

Parameters:

blocks:
blocks
MetaInlines (inlines)

Meta inlines

Parameters:

inlines:
inlines
MetaList (meta_values)

Meta list

Parameters:

meta_values:
list of meta values
MetaMap (key_value_map)

Meta map

Parameters:

key_value_map:
a string-indexed map of meta values
MetaString (str)

Creates string to be used in meta data.

Parameters:

str:
string value
MetaBool (bool)

Creates boolean to be used in meta data.

Parameters:

bool:
boolean value

Blocks

Block

Block elements

BlockQuote (content)

Creates a block quote element

Parameters:

content:
block content

Returns: block quote element

BulletList (content)

Creates a bullet (i.e.

Parameters:

content:
list of items

Returns: bullet list element

CodeBlock (text[, attr])

Creates a code block element

Parameters:

text:
code string
attr:
element attributes

Returns: code block element

DefinitionList (content)

Creates a definition list, containing terms and their explanation.

Parameters:

content:
list of items

Returns: definition list element

Div (content[, attr])

Creates a div element

Parameters:

content:
block content
attr:
element attributes

Returns: div element

Header (level, content[, attr])

Creates a header element.

Parameters:

level:
header level
content:
inline content
attr:
element attributes

Returns: header element

HorizontalRule ()

Creates a horizontal rule.

Returns: horizontal rule

LineBlock (content)

Creates a line block element.

Parameters:

content:
inline content

Returns: line block element

Null ()

Creates a null element.

Returns: null element

OrderedList (items[, listAttributes])

Creates an ordered list.

Parameters:

items:
list items
listAttributes:
list parameters

Returns: ordered list element

Para (content)

Creates a para element.

Parameters:

content:
inline content

Returns: paragraph element

Plain (content)

Creates a plain element.

Parameters:

content:
inline content

Returns: plain element

RawBlock (format, text)

Creates a raw content block of the specified format.

Parameters:

format:
format of content
text:
string content

Returns: raw block element

Table (caption, aligns, widths, headers, rows)

Creates a table element.

Parameters:

caption:
table caption
aligns:
alignments
widths:
column widths
headers:
header row
rows:
table rows

Returns: table element

Inline

Inline

Inline element class

Cite (content, citations)

Creates a Cite inline element

Parameters:

content:
List of inlines
citations:
List of citations

Returns: citations element

Code (text[, attr])

Creates a Code inline element

Parameters:

text:
brief image description
attr:
additional attributes

Returns: code element

Emph (content)

Creates an inline element representing emphasised text.

Parameters:

content:
inline content

Returns: emphasis element

Image (caption, src[, title[, attr]])

Creates a Image inline element

Parameters:

caption:
text used to describe the image
src:
path to the image file
title:
brief image description
attr:
additional attributes

Returns: image element

LineBreak ()

Create a LineBreak inline element

Returns: linebreak element

Link (content, target[, title[, attr]])

Creates a link inline element, usually a hyperlink.

Parameters:

content:
text for this link
target:
the link target
title:
brief link description
attr:
additional attributes

Returns: image element

Math (mathtype, text)

Creates a Math element, either inline or displayed.

Parameters:

mathtype:
rendering specifier
text:
Math content

Returns: Math element

DisplayMath (text)

Creates a DisplayMath element (DEPRECATED).

Parameters:

text:
Math content

Returns: Math element

InlineMath (text)

Creates an InlineMath inline element (DEPRECATED).

Parameters:

text:
Math content

Returns: Math element

Note (content)

Creates a Note inline element

Parameters:

content:
footnote block content
Quoted (quotetype, content)

Creates a Quoted inline element given the quote type and quoted content.

Parameters:

quotetype:
type of quotes to be used
content:
inline content

Returns: quoted element

SingleQuoted (content)

Creates a single-quoted inline element (DEPRECATED).

Parameters:

content:
inline content

Returns: quoted element

See also: Quoted

DoubleQuoted (content)

Creates a single-quoted inline element (DEPRECATED).

Parameters:

content:
inline content

Returns: quoted element

See also: Quoted

RawInline (format, text)

Creates a RawInline inline element

Parameters:

format:
format of the contents
text:
string content

Returns: raw inline element

SmallCaps (content)

Creates text rendered in small caps

Parameters:

content:
inline content

Returns: smallcaps element

SoftBreak ()

Creates a SoftBreak inline element.

Returns: softbreak element

Space ()

Create a Space inline element

Returns: space element

Span (content[, attr])

Creates a Span inline element

Parameters:

content:
inline content
attr:
additional attributes

Returns: span element

Str (text)

Creates a Str inline element

Parameters:

text:
content

Returns: string element

Strikeout (content)

Creates text which is striked out.

Parameters:

content:
inline content

Returns: strikeout element

Strong (content)

Creates a Strong element, whose text is usually displayed in a bold font.

Parameters:

content:
inline content

Returns: strong element

Subscript (content)

Creates a Subscript inline element

Parameters:

content:
inline content

Returns: subscript element

Superscript (content)

Creates a Superscript inline element

Parameters:

content:
inline content

Returns: strong element

Element components

Attr ([identifier[, classes[, attributes]]])

Create a new set of attributes (Attr).

Parameters:

identifier:
element identifier
classes:
element classes
attributes:
table containing string keys and values

Returns: element attributes

Citation (id, mode[, prefix[, suffix[, note_num[, hash]]]])

Creates a single citation.

Parameters:

id:
citation identifier (like a bibtex key)
mode:
citation mode
prefix:
citation prefix
suffix:
citation suffix
note_num:
note number
hash:
hash number

Constants

AuthorInText

Author name is mentioned in the text.

See also: Citation

SuppressAuthor

Author name is suppressed.

See also: Citation

NormalCitation

Default citation style is used.

See also: Citation

AlignLeft

Table cells aligned left.

See also: Table

AlignRight

Table cells right-aligned.

See also: Table

AlignCenter

Table cell content is centered.

See also: Table

AlignDefault

Table cells are alignment is unaltered.

See also: Table

DefaultDelim

Default list number delimiters are used.

See also: OrderedList

Period

List numbers are delimited by a period.

See also: OrderedList

OneParen

List numbers are delimited by a single parenthesis.

See also: OrderedList

TwoParens

List numbers are delimited by a double parentheses.

See also: OrderedList

DefaultStyle

List are numbered in the default style

See also: OrderedList

Example

List items are numbered as examples.

See also: OrderedList

Decimal

List are numbered using decimal integers.

See also: OrderedList

LowerRoman

List are numbered using lower-case roman numerals.

See also: OrderedList

UpperRoman

List are numbered using upper-case roman numerals

See also: OrderedList

LowerAlpha

List are numbered using lower-case alphabetic characters.

See also: OrderedList

UpperAlpha

List are numbered using upper-case alphabetic characters.

See also: OrderedList

Helper functions

pipe (command, args, input)

Runs command with arguments, passing it some input, and returns the output.

Returns:

  • Output of command.

Raises:

  • A table containing the keys command, error_code, and output is thrown if the command exits with a non-zero error code.

Usage:

local output = pandoc.pipe("sed", {"-e","s/a/b/"}, "abc")
walk_block (element, filter)

Apply a filter inside a block element, walking its contents.

Parameters:

element:
the block element
filter:
a lua filter (table of functions) to be applied within the block element

Returns: the transformed block element

walk_inline (element, filter)

Apply a filter inside an inline element, walking its contents.

Parameters:

element:
the inline element
filter:
a lua filter (table of functions) to be applied within the inline element

Returns: the transformed inline element

read (markup[, format])

Parse the given string into a Pandoc document.

Parameters:

markup:
the markup to be parsed
format:
format specification, defaults to "markdown".

Returns: pandoc document

Usage:

local org_markup = "/emphasis/"  -- Input to be read
local document = pandoc.read(org_markup, "org")
-- Get the first block of the document
local block = document.blocks[1]
-- The inline element in that block is an `Emph`
assert(block.content[1].t == "Emph")

Module pandoc.utils

This module exposes internal pandoc functions and utility functions.

blocks_to_inlines (blocks[, sep])

Squash a list of blocks into a list of inlines.

Parameters:

blocks:
List of blocks to be flattened.
sep:
List of inlines inserted as separator between two consecutive blocks; defaults to { pandoc.Space(), pandoc.Str'¶', pandoc.Space()}.

Returns:

  • ({[Inline][#Inline]}) List of inlines

Usage:

local blocks = {
  pandoc.Para{ pandoc.Str 'Paragraph1' },
  pandoc.Para{ pandoc.Emph 'Paragraph2' }
}
local inlines = pandoc.utils.blocks_to_inlines(blocks)
-- inlines = {
--   pandoc.Str 'Paragraph1',
--   pandoc.Space(), pandoc.Str'¶', pandoc.Space(),
--   pandoc.Emph{ pandoc.Str 'Paragraph2' }
-- }
hierarchicalize (blocks)

Convert list of blocks into an hierarchical list. An hierarchical elements is either a normal block (but no Header), or a Sec element. The latter has the following fields:

  • level: level in the document hierarchy;
  • numbering: list of integers of length level, specifying the absolute position of the section in the document;
  • attr: section attributes (see Attr);
  • contents: nested list of hierarchical elements.

Returns:

  • List of hierarchical elements

Usage:

local blocks = {
  pandoc.Header(2, pandoc.Str 'first'),
  pandoc.Header(2, pandoc.Str 'second'),
}
local elements = pandoc.utils.hierarchicalize(blocks)
print(table.concat(elements[1].numbering, '.')) -- 0.1
print(table.concat(elements[2].numbering, '.')) -- 0.2
run_json_filter (doc, filter[, args])

Filter the given doc by passing it through the a JSON filter.

Parameters:

doc:
the Pandoc document to filter
filter:
filter to run
args:
list of arguments passed to the filter. Defaults to {FORMAT}.

Returns:

Usage:

-- Assumes `some_blocks` contains blocks for which a
-- separate literature section is required.
local sub_doc = pandoc.Pandoc(some_blocks, metadata)
sub_doc_with_bib = pandoc.utils.run_json_filter(
  sub_doc,
  'pandoc-citeproc'
)
some_blocks = sub_doc.blocks -- some blocks with bib
normalize_date (date_string)

Parse a date and convert (if possible) to “YYYY-MM-DD” format. We limit years to the range 1601-9999 (ISO 8601 accepts greater than or equal to 1583, but MS Word only accepts dates starting 1601).

Returns:

  • A date string, or nil when the conversion failed.
sha1 (contents)

Returns the SHA1 has of the contents.

Returns:

  • SHA1 hash of the contents.

Usage:

local fp = pandoc.utils.sha1("foobar")
stringify (element)

Converts the given element (Pandoc, Meta, Block, or Inline) into a string with all formatting removed.

Returns:

  • A plain string representation of the given element.

Usage:

local inline = pandoc.Emph{pandoc.Str 'Moin'}
-- outputs "Moin"
print(pandoc.utils.stringify(inline))
to_roman_numeral (integer)

Converts an integer < 4000 to uppercase roman numeral.

Returns:

  • A roman numeral string.

Usage:

local to_roman_numeral = pandoc.utils.to_roman_numeral
local pandoc_birth_year = to_roman_numeral(2006)
-- pandoc_birth_year == 'MMVI'

Module pandoc.mediabag

The pandoc.mediabag module allows accessing pandoc’s media storage. The “media bag” is used when pandoc is called with the --extract-media or --standalone/-s option.

insert (filepath, mime_type, contents)

Adds a new entry to pandoc’s media bag.

Parameters:

filepath:
filename and path relative to the output folder.
mime_type:
the file’s MIME type
contents:
the binary contents of the file.

Usage:

local fp = "media/hello.txt"
local mt = "text/plain"
local contents = "Hello, World!"
pandoc.mediabag(fp, mt, contents)
list ()

Get a summary of the current media bag contents.

Returns: A list of elements summarizing each entry in the media bag. The summary item contains the keys path, type, and length, giving the filepath, MIME type, and length of contents in bytes, respectively.

Usage:

-- calculate the size of the media bag.
local mb_items = pandoc.mediabag.list()
local sum = 0
for i = 1, #mb_items:
    sum = sum + mb_items[i].length
end
print(sum)
lookup (filepath)

Lookup a media item in the media bag, returning mime type and contents.

Parameters:

filepath:
name of the file to look up.

Returns:

  • the entries MIME type, or nil if the file was not found.
  • contents of the file, or nil if the file was not found.

Usage:

local filename = "media/diagram.png"
local mt, contents = pandoc.mediabag.lookup(filename)
fetch (source, base_url)

Fetches the given source from a URL or local file. Returns two values: the contents of the file and the mime type (or an empty string).

Returns:

  • the entries MIME type, or nil if the file was not found.
  • contents of the file, or nil if the file was not found.

Usage:

local diagram_url = "https://pandoc.org/diagram.jpg"
local contents = pandoc.mediabag.fetch(diagram_url, ".")

Module pandoc.List

Pandoc's List type and helper methods

Metamethods

pandoc.List:__concat (list)

Concatenates two lists.

Parameters:

list:
second list concatenated to the first

Returns: a new list containing all elements from list1 and list2

Methods

pandoc.List:clone ()

Returns a (shallow) copy of the list.

pandoc.List:includes (needle, init)

Checks if the list has an item equal to the given needle.

Parameters:

needle:
item to search for
init:
index at which the search is started

Returns: true if a list item is equal to the needle, false otherwise

pandoc.List:find (needle, init)

Returns the value and index of the first occurrence of the given item.

Parameters:

needle:
item to search for
init:
index at which the search is started

Returns: first item equal to the needle, or nil if no such item exists.

pandoc.List:find_if (pred, init)

Returns the value and index of the first element for which the predicate holds true.

Parameters:

pred:
the predicate function
init:
index at which the search is started

Returns: first item for which `test` succeeds, or nil if no such item exists.

pandoc.List:extend (list)

Adds the given list to the end of this list.

Parameters:

list:
list to appended
pandoc.List:map (fn)

Returns a copy of the current list by applying the given function to all elements.

Parameters:

fn:
function which is applied to all list items.
pandoc.List:filter (pred)

Returns a new list containing all items satisfying a given condition.

Parameters:

pred:
condition items must satisfy.

Returns: a new list containing all items for which `test` was true.