LOWDOWN(3) Library Functions Manual LOWDOWN(3)

lowdown — simple markdown translator library

library “liblowdown”

#include <sys/queue.h>
#include <stdio.h>
#include <lowdown.h>

struct lowdown_meta
struct lowdown_node
struct lowdown_opts

This library parses lowdown(5) into various output formats.

The library consists first of a high-level interface consisting of lowdown_buf(3), lowdown_buf_diff(3), lowdown_file(3), and lowdown_file_diff(3).

The high-level functions interface with low-level functions that perform parsing and formatting. These consist of lowdown_doc_new(3), lowdown_doc_parse(3), and lowdown_doc_free(3) for parsing lowdown(5) documents into an abstract syntax tree.

The front-end functions for freeing, allocation, and rendering are as follows.

To compile and link, use pkg-config(1):

% cc `pkg-config --cflags lowdown` -c -o sample.o sample.c
% cc -o sample sample.o `pkg-config --libs lowdown`

The lowdown library is built to operate in security-sensitive environments, such as those using pledge(2) on OpenBSD. The only promise required is stdio for lowdown_file_diff(3) and lowdown_file(3): both require access to the stream for reading input.

All lowdown functions use one or more of the following structures.

The main structure for configuring parsing and output is struct lowdown_opts. It has the following fields:

enum lowdown_type type
The output medium:

HTML5
LaTeX
roff -man macros
“flat” OpenDocument
ANSI-compatible UTF-8 terminal output
Gemini format
roff -ms macros
syntax tree (debugging)
unsigned int feat
Parse-time features. This bit-field may have the following bits OR'd:

Parse PHP extra link, header, and image attributes.
Parse http, https, ftp, mailto, and relative links or link fragments.
Parse MDN/GFM callouts (“admonitions”).
Tighten input parsing to the CommonMark specification. This also uses the first ordered list value instead of starting all lists at one. This feature is and .
Parse PHP extra definition lists. This is currently constrained to single-key lists.
Parse GFM fenced (language-specific) code blocks.
Parse MMD style footnotes. This only supports the referenced footnote style, not the “inline” style.
Parse highlit sequences. This are disabled by default because it may be erroneously interpreted as section headers.
Deprecated. Use LOWDOWN_ATTRS instead.
Recognise manpage titles in Pandoc metadata title lines. Only applicable if LOWDOWN_METADATA is also provided. Manpages titles must begin with a non-empty title followed by an open parenthesis, digit or “n”, optional letters after, then a closing parenthesis. This may be optionally followed by a source and, if a vertical bar is detected, the content after as the volume. These are passed to the renderers as the title, volume, and optionally source and volume metadata key-value pairs. The original title is not recoverable.
Parse mathematics equations.
Parse in-document metadata.
Do not parse indented content as code blocks.
Do not parse emphasis within words.
Parse strikethrough sequences.
Parse super-scripts. This accepts foo^bar^ GFM super-scripts.
If LOWDOWN_SUPER is enabled, instead of the GFM style, accept the “short” form of superscript. This accepts foo^bar, which puts the parts following the caret until whitespace in superscripts; or foo^(bar), which puts only the parts in parenthesis.
Parse GFM tables.
Parse GFM task list items.
unsigned int oflags
Output-time features. Bit values are specific to the type and are not guaranteed to be globally unique.

For all types:

Don't use smart typography formatting.
Emit a full document instead of a document fragment. This envelope is largely populated from metadata if LOWDOWN_METADATA was provided as an option or as given in meta or metaovr.

For LOWDOWN_HTML:

, LOWDOWN_HTML_CALLOUT_GFM
Output MDN and/or GFM-style callout syntax.
If LOWDOWN_HTML_SKIP_HTML has not been set, escapes in-document HTML so that it is rendered as opaque text.
Retain line-breaks within paragraphs.
Have an identifier written with each header element consisting of an HTML-escaped version of the header contents.
Convert, when possible, HTML entities to their numeric form. If not set, the entities are used as given in the input.
When escaping text, be extra paranoid in following the OWASP suggestions for which characters to escape.
Do not render in-document HTML at all.
If used with LOWDOWN_STANDALONE, output a Pandoc-style title block. This is a <header id="title-block-header"> element right after the opening <body> containing elements for specified title, author(s), and date. These are <h1> and <p> elements, respectively, with classes set to what's being output (title, etc.). At least one of these must be specified for the title block to be output.

For LOWDOWN_GEMINI, there are several flags for controlling link placement. By default, links (images, autolinks, and links) are queued when specified in-line then emitted in a block sequence after the nearest block node. (See ABSTRACT SYNTAX TREE.)

Emit the queue of links at the end of the document instead of after the nearest block node.
Render all links within the flow of text. This will cause breakage when nested links, such as images within links, links in blockquotes, etc. It should not be used unless in carefully crafted documents.
Do not format link labels. Takes precedence over LOWDOWN_GEMINI_LINK_ROMAN.
When formatting link labels, use lower-case Roman numerals instead of the default lowercase hexavigesimal (i.e., “a”, “b”, ..., “aa”, “ab”, ...).
Print metadata as the canonicalised key followed by a colon then the value, each on one line (newlines replaced by spaces). The metadata block is terminated by a double newline. If there is no metadata, this does nothing.

There may only be one of LOWDOWN_GEMINI_LINK_END or LOWDOWN_GEMINI_LINK_IN. If both are specified, the latter is unset.

For LOWDOWN_FODT:

Do not render in-document HTML at all. Text within HTML elements remains.

For LOWDOWN_LATEX:

Use the default numbering scheme for sections, subsections, etc. If not specified, these are inhibited.
Do not render in-document HTML at all. Text within HTML elements remains.

For LOWDOWN_MAN and LOWDOWN_NROFF:

Use GNU extensions (i.e., for groff(1)) when rendering output. The groff arguments must include -mpdfmark for formatting links with LOWDOWN_MAN or -mspdf instead of -ms for LOWDOWN_NROFF. Applies to the LOWDOWN_MAN and LOWDOWN_NROFF output types.
Don't show links at all if they have embedded text. Applies to images and regular links. Only in LOWDOWN_MAN or when LOWDOWN_NROFF_GROFF is not specified.
Use numbered sections if LOWDOWON_NROFF_GROFF is not specified. Only applies to the LOWDOWN_NROFF output type.
Render link URLs in short form. Applies to images, autolinks, and regular links. Only in LOWDOWN_MAN or when LOWDOWN_NROFF_GROFF is not specified.
Do not render in-document HTML at all. Text within HTML elements remains.

For LOWDOWN_TERM:

If LOWDOWN_STANDALONE is specified, output all metadata instead of just the title, author, and date.
Don't apply ANSI style codes at all. This implies LOWDOWN_TERM_NOCOLOUR.
Don't apply ANSI colour codes. This will still show underline, bold, etc. This should not be used in difference mode, as the output will make no sense.
Don't show links at all. Applies to images and regular links: autolinks are still shown. This may be combined with LOWDOWN_TERM_SHORTLINK to also shorten autolinks.
Render link URLs in short form. Applies to images, autolinks, and regular links. This may be combined with LOWDOWN_TERM_NOLINK to only show shortened autolinks.
size_t maxdepth
The maximum parse depth before the parser exits. Most documents will have a parse depth in the single digits.
size_t cols
For LOWDOWN_TERM, the “soft limit” for width of terminal output not including margins. If zero, 80 shall be used.
size_t hmargin
For LOWDOWN_TERM, the left margin (space characters).
size_t vmargin
For LOWDOWN_TERM, the top/bottom margin (newlines).
struct lowdown_opts_nroff nroff
If type is LOWDOWN_MAN or LOWDOWN_NROFF, this contains constant-width font variants: const char *cr for roman constant-width, const char *cb for bold, const char *ci for italic, and const char *cbi for bold-italic. If any of these are NULL, they default to their constant-width variants.
struct lowdown_opts_odt odt
If type is LOWDOWN_FODT, this contains const char *sty, which is either NULL or the OpenDocument styles used when creating standalone documents. If NULL, the default styles are used.
char **meta
An array of metadata key-value pairs or NULL. Each pair must appear as if provided on one line (or multiple lines) of the input, including the terminating newline character. If not consisting of a valid pair (e.g., no newline, no colon), then it is ignored. When processed, these values are overridden by those in the document (if LOWDOWN_METADATA is specified) or by those in metaovr.
size_t metasz
Number of pairs in metaovr.
char **metaovr
See meta. The difference is that metaovr is applied after meta and in-document metadata, so it overrides prior values.
size_t metaovrsz
Number of pairs in metaovr.

Parsed metadata is held in key-value struct lowdown_meta pairs, or collectively as struct lowdown_metaq, if LOWDOWN_METADATA is set in feat. The former structure consists of the following fields:

char *key
The metadata key in its canonical form: lowercase alphanumerics, hyphen, and underscore. Whitespace is removed and other characters replaced by a question mark.
char *value
The metadata value. This may be an empty string.

The abstract syntax tree is encoded in struct lowdown_node, which consists of the following.

enum lowdown_rndrt type
The node type, using HTML5 output as an illustration:

A block-level snippet of code described by <pre><code>.
A block-level snippet of HTML. This is simply opaque HTML content.
A block-level quotation described by <blockquote>.
An inline-level snippet of code described by <code>.
A definition list described by <dl>.
Definition data described by <dd>.
Definition title described by <dt>.
Container for metadata described by <head>.
Bold (or otherwise notable) content described by <strong>.
Italic (or otherwise notable) content described by <em>.
Named or numeric HTML entity.
Footnote content.
A block-level header described by one of <h1> through <h6>.
Marked test described by <mark>.
A horizontal line described by <hr>.
An image described by <img>.
A hard line-break within a block context described by <br>.
A link to external media described by <a>.
Like LOWDOWN_LINK, except inferred from text content.
A list enclosure described by <ul> or <ol>.
A list item described by <li>.
A snippet of mathematical text in LaTeX format described within \[xx\] or \(xx\). This is usually (in HTML) externally handled by a JavaScript renderer.
Meta-data keys and values. These are described by elements in <head>.
Normal text content.
A block-level paragraph described by <p>.
An inline of raw HTML. (Only if configured during parse.)
The root of the document. This is always the topmost node, and the only node where the parent field is NULL.
Content struck through. Described by <del>.
, LOWDOWN_SUPERSCRIPT
A subscript or superscript described by <sub> or <sup>, respectively.
A table block described by <table>.
A table body section described by <tbody>.
A table cell described by <td>, or <th> if in the header.
A table header section described by <thead>.
A table row described by <tr>.
Combination of LOWDOWN_EMPHASIS and LOWDOWN_DOUBLE_EMPHASIS.
size_t id
An identifier unique within the document. This can be used as a table index since the number is assigned from a monotonically increasing point during the parse.
struct lowdown_node *parent
The parent of the node, or NULL at the root.
enum lowdown_chng chng
Change tracking: whether this node was inserted (LOWDOWN_CHNG_INSERT), deleted (LOWDOWN_CHNG_DELETE), or neither (LOWDOWN_CHNG_NONE).
struct lowdown_nodeq children
A possibly-empty list of child nodes.
<anon union>
An anonymous union of type-specific structures.

For LOWDOWN_LINK_AUTO, the link address as link and the link type type, which may be one of HALINK_EMAIL for e-mail links and HALINK_NORMAL otherwise. Any buffer may be empty-sized.
rndr_blockcode
For LOWDOWN_BLOCKCODE, the opaque text of the block and the optional lang of the code language.
rndr_blockhtml
For LOWDOWN_BLOCKHTML, the opaque HTML text.
rndr_codespan
The opaque text of the contents.
rndr_definition
For LOWDOWN_DEFINITION, containing flags that may be HLIST_FL_BLOCK if the definition list should be interpreted as containing block nodes.
rndr_entity
For LOWDOWN_ENTITY, the entity text.
rndr_header
For LOWDOWN_HEADER, the level of the header starting at zero (this value is relative to the metadata base header level, defaulting to one), optional space-separated class list attr_cls, and optional single identifier attr_id.
rndr_image
For LOWDOWN_IMAGE, the image address link, the image title title, dimensions NxN (width by height) in dims, and alternate text alt. CSS in-line style for width and height may be given in attr_width and/or attr_height, and a space-separated list of classes may be in attr_cls and a single identifier may be in attr_id.
Like rndr_autolink, but without a type and further defining an optional link title title, optional space-separated class list attr_cls, and optional single identifier attr_id.
rndr_list
For LOWDOWN_LIST, consists of a bitfield flags that may be set to HLIST_FL_ORDERED for an ordered list and HLIST_FL_UNORDERED for an unordered one. If HLIST_FL_BLOCK is set, the list should be output as if items were separate blocks. The start value for HLIST_FL_ORDERED is the starting list item position, which is one by default and never zero. The items is the number of list items.
rndr_listitem
For LOWDOWN_LISTITEM, consists of a bitfield flags that may be set to HLIST_FL_ORDERED for an ordered list, HLIST_FL_UNORDERED for an unordered list, HLIST_FL_DEF for definition list data, HLIST_FL_CHECKED or HLIST_FL_UNCHECKED for an unordered “task” list, and/or HLIST_FL_BLOCK for list item output as if containing block nodes. The HLIST_FL_BLOCK should not be used: use the parent list (or definition list) flags for this. The num is the index in a HLIST_FL_ORDERED list. It is monotonically increasing with each item in the list, starting at the start variable given in struct rndr_list.
rndr_math
For LOWDOWN_MATH, the mode of display in blockmode: if 1, in-line math; if 2, multi-line. The opaque equation, which is assumed to be in LaTeX format, is in the opaque text.
rndr_meta
Each LOWDOWN_META key-value pair is represented. The keys are lower-case without spaces or non-ASCII characters. If provided, enclosed nodes may consist only of LOWDOWN_NORMAL_TEXT and LOWDOWN_ENTITY.
rndr_normal_text
The basic text content for LOWDOWN_NORMAL_TEXT. If flags is set to HTEXT_ESCAPED, the text may be escaped for output, but may not be altered by any smart typography or similar (it should be passed as-is).
rndr_paragraph
For LOWDOWN_PARAGRAPH, species how many lines the paragraph has in the input file and beoln, set to non-zero if the paragraph ends with an empty line instead of a breaking block node.
rndr_raw_html
For LOWDOWN_RAW_HTML, the opaque HTML text.
rndr_table
For LOWDOWN_TABLE_BLOCK, the number of columns in each row or header row. The number of columns in rndr_table, rndr_table_header, and rndr_table_cell are the same.
rndr_table_cell
For LOWDOWN_TABLE_CELL, the current col column number out of columns. See rndr_table_header for a description of the bits in flags. The number of columns in rndr_table, rndr_table_header, and rndr_table_cell are the same.
rndr_table_header
For LOWDOWN_TABLE_HEADER, the number of columns in each row and the per-column flags, which may tested for equality against HTBL_FL_ALIGN_LEFT, HTBL_FL_ALIGN_RIGHT, or HTBL_FL_ALIGN_CENTER after being masked with HTBL_FL_ALIGNMASK; or HTBL_FL_HEADER. If no alignment is specified after the mask, the default should be left-aligned. The number of columns in rndr_table, rndr_table_header, and rndr_table_cell are the same.

A parsed document is a tree of struct lowdown_node nodes. If a node is “block”, it may contain other block or inline nodes. If “inline,” it may only contain other inline nodes. “Special” nodes are documented below. An additional mark of “void” means that the node will never contain children.

Node Scope
block, void
block, void
block
inline, void
block
special
special
special
inline
inline
inline, void
block, special
block
inline, void
inline, void
inline, void
inline
block
special
inline, void
special
inline, void
block
inline, void
special
inline
inline
inline
block
special
special
special
special
inline

The general structure of the AST is as follows. Nodes have no order imposed on them unless as noted:

Special nodes have specific placement within their parents as follows:

Lastly, LOWDOWN_FOOTNOTE may appear anywhere in the document and contains block nodes.

lowdown(1), lowdown_buf(3), lowdown_buf_diff(3), lowdown_diff(3), lowdown_doc_free(3), lowdown_doc_new(3), lowdown_doc_parse(3), lowdown_file(3), lowdown_file_diff(3), lowdown_gemini_free(3), lowdown_gemini_new(3), lowdown_gemini_rndr(3), lowdown_html_free(3), lowdown_html_new(3), lowdown_html_rndr(3), lowdown_latex_free(3), lowdown_latex_new(3), lowdown_latex_rndr(3), lowdown_metaq_free(3), lowdown_nroff_free(3), lowdown_nroff_new(3), lowdown_nroff_rndr(3), lowdown_odt_free(3), lowdown_odt_new(3), lowdown_odt_rndr(3), lowdown_term_free(3), lowdown_term_new(3), lowdown_term_rndr(3), lowdown_tree_rndr(3), lowdown(5)

lowdown was forked from hoedown by Kristaps Dzonsons, kristaps@bsd.lv. It has been considerably modified since.

January 31, 2025 Debian