pymd4c 0.4.4.0b1

Python bindings for MD4C

PyMD4C

Python bindings for the very fast MD4C Markdown parsing and rendering library.

The MD4C C library provides a SAX-like parser that uses callbacks to return the various blocks, inlines, and text it parses from the Markdown input. In addition, it provides an HTML renderer that wraps the generic parser to provide HTML output directly.

Accordingly, this Python module provides two classes:

• md4c.GenericParser - Wraps the generic SAX-like MD4C parser. Requires Python functions (or other callables) as callbacks.
• md4c.HTMLRenderer - Wraps the HTML renderer. Produces HTML output directly.

If other renderers are added to MD4C, they will get their own Python class as well, similar to the HTMLRenderer.

Install from PyPI

PyMD4C is available on PyPI under the name pymd4c. Install it with pip like this:

pip install pymd4c

This is the recommended method to obtain PyMD4C on Linux. It should work well on most distributions. Unfortunately, Windows and macOS packages are not currently built automatically, so users on those platforms will need to build from source. The instructions below should assist.

Note that some more esoteric distributions or non-x86/x86_64 architectures may not be supported by the manylinux packages. If either of those apply to your system, you may also need to build from source. (However, if you are running Linux on arm64, ppc64le, or s390x, consider opening a new GitHub issue--it may be possible to add your architecture.)

Build and Install from Source

Prerequisites

This package depends on the MD4C library. It may be available through your package manager. Otherwise, it can be built from source as follows (note that the below instructions are for Unix-like systems, but theoretically there are ways to build on Windows as well):

1. Download and extract the matching release from the releases page (e.g. for PyMD4C version W.X.Y.Z, download MD4C version W.X.Y).

2. Inside the extracted file, run the following:

   mkdir build
   cd build
   cmake ..
   make
   # Do as root:
   make install
   

   The install step must be run as root. The library will install to /usr/local by default.

3. You may need to rebuild the ldconfig cache (also as root): ldconfig

In addition, the pkg-config tool and the Python pkgconfig package must be available to build PyMD4C, but they are not required after that (i.e., they are not a prerequisite for actually using PyMD4C). The pkg-config tool is likely available on your system already, and the Python pkgconfig package will be fetched automatically by setup.py.

Finally, note that since this package uses C extensions, development headers for Python must be installed for the build to succeed. If you are using Linux, some distributions split these off from the main Python package. Install python-dev or python-devel to get them.

Build/Install

Build and install with setup.py as you would for any Python source repository:

pip install .

Class GenericParser

import md4c
generic_parser = md4c.GenericParser(parser_flags)

Initialize a new GenericParser. Parameters:

• parser_flags - An int made up of some combination of the parser option flags or'd together, e.g. md4c.MD_FLAG_TABLES | md4c.MD_FLAG_STRIKETHROUGH. For the default options, use 0, which will parse according to the base CommonMark specification. See the "Module-Wide Constants" section below for a full list of parser option flags.

Note: If the end goal of parsing is to produce HTML, strongly consider using an HTMLRenderer instead. All rendering will be performed by native C code, which will be much faster.

Parse Method

import md4c
generic_parser = md4c.GenericParser(...)
generic_parser.parse(input,
                     enter_block_callback,
                     leave_block_callback,
                     enter_span_callback,
                     leave_span_callback,
                     text_callback)

Parse markdown text using the provided callbacks. Parameters:

• input - A str or bytes containing the Markdown document to parse. If a bytes, it must be UTF-8 encoded.
• enter_block_callback - A function (or other callable) to be called whenever the parser enters a new block element in the Markdown source.
• leave_block_callback - A function (or other callable) to be called whenever the parser leaves a block element in the Markdown source.
• enter_span_callback - A function (or other callable) to be called whenever the parser enters a new inline element in the Markdown source.
• leave_span_callback - A function (or other callable) to be called whenever the parser leaves an inline element in the Markdown source.
• text_callback - A function (or other callable) to be called whenever the parser has text to add to the current block or inline element.

The parse() method will raise md4c.ParseError in the event of a problem during parsing, such as running out of memory. This does not signal invalid syntax, as there is no such thing in Markdown. It can also emit any exception raised by any of the callbacks (except md4c.StopParsing, which is caught and handled quietly).

Callback Details

enter_block_callback, leave_block_callback, enter_span_callback, and leave_span_callback all must accept two parameters:

• type - An md4c.BlockType or md4c.SpanType representing the type of block or span. See the "Enums" section for more info.

• details - A dict that contains extra information for certain types of blocks and spans, for example, the level of a heading. Keys are strs. Values are ints, single-character strs, or (for MD_ATTRIBUTE) lists of tuples.

  See the MD_BLOCK_*_DETAIL and MD_SPAN_*_DETAIL structs in MD4C's md4c.h for information on exactly what this dict will contain.

  Regarding MD_ATTRIBUTEs: These are used where a block or span can contain some associated text, such as link titles and code block language references. Such attributes may contain multiple text sub-elements (e.g. some regular text, an HTML entity, and then some more regular text). Thus, an MD_ATTRIBUTE value in details consists of a list of 2-tuples: (text_type, text) where text_type is an md4c.TextType (see "Enums" below) and text is the actual text as a str.

text_callback must also accept two parameters, but they are different:

• type - An md4c.TextType representing the type of text element. See the "Enums" section for more info.
• text - The actual text, as a str.

Callbacks need not return anything specific; their return values are ignored. To cancel parsing, callbacks can raise md4c.StopParsing. This will be caught by the parse() method and immediately halt parsing quietly. All other exceptions raised by callbacks will abort parsing and will be propagated back to the caller of parse().

Class HTMLRenderer

import md4c
html_renderer = md4c.HTMLRenderer(parser_flags, renderer_flags)

Initialize a new HTMLRenderer. Parameters:

• parser_flags - An int made up of some combination of the parser option flags or'd together, e.g. md4c.MD_FLAG_TABLES | md4c.MD_FLAG_STRIKETHROUGH. For the default options, use 0, which will parse according to the base CommonMark standard. See the "Module-Wide Constants" section below for a full list of parser option flags.
• renderer_flags - An int made up of some combination of the HTML renderer option flags or'd together. These are also listed in the "Module-Wide Constants" section below.

Parse Method

import md4c
html_renderer = md4c.HTMLRenderer(...)
html_renderer.parse(input)

Parse markdown text and return a str with rendered HTML. Parameters:

• input - A str or bytes containing the Markdown document to parse. If a bytes, it must be UTF-8 encoded.

This method will raise md4c.ParseError in the event of a problem during parsing, such as running out of memory. This does not signal invalid syntax, as there is no such thing in Markdown.

Module-Wide Constants

The MD4C library provides various option flags for parsers and renderers as named constants. These are made available as module-level constants in PyMD4C.

Parser Option Flags

Basic option flags:

• md4c.MD_FLAG_COLLAPSEWHITESPACE - In normal text, collapse non-trivial whitespace into a single space.
• md4c.MD_FLAG_PERMISSIVEATXHEADERS - Do not requite a space in ATX headers (e.g. ###Header).
• md4c.MD_FLAG_PERMISSIVEURLAUTOLINKS - Convert URLs to links even without < and >.
• md4c.MD_FLAG_PERMISSIVEEMAILAUTOLINKS - Convert email addresses to links even without <, >, and mailto:.
• md4c.MD_FLAG_NOINDENTEDCODEBLOCKS - Disable indented code blocks. (Only allow fenced code blocks.)
• md4c.MD_FLAG_NOHTMLBLOCKS - Disable raw HTML blocks.
• md4c.MD_FLAG_NOHTMLSPANS - Disable raw HTML inlines.
• md4c.MD_FLAG_TABLES - Enable tables extension.
• md4c.MD_FLAG_STRIKETHROUGH - Enable strikethrough extension.
• md4c.MD_FLAG_PERMISSIVEWWWAUTOLINKS - Enable www autolinks (even without any scheme prefix, as long as they begin with www.).
• md4c.MD_FLAG_TASKLISTS - Enable task lists extension.
• md4c.MD_FLAG_LATEXMATHSPANS - Enable $ and $$ containing LaTeX equations.
• md4c.MD_FLAG_WIKILINKS - Enable wiki links extension.
• md4c.MD_FLAG_UNDERLINE - Enable underline extension (and disable _ for regular emphasis).

Combination option flags:

• md4c.MD_FLAG_PERMISSIVEAUTOLINKS - Enables all varieties of autolinks: MD_FLAG_PERMISSIVEURLAUTOLINKS, MD_FLAG_PERMISSIVEEMAILAUTOLINKS, and MD_FLAG_PERMISSIVEWWWAUTOLINKS
• md4c.MD_FLAG_NOHTML - Disables all raw HTML tags: MD_FLAG_NOHTMLBLOCKS and MD_FLAG_NOHTMLSPANS

Dialect option flags (note that not all features of a dialect may be supported, but these flags will cause MD4C to parse as many features of the dialect as it supports):

• md4c.MD_DIALECT_COMMONMARK - This is the default behavior of MD4C, so no additional flags are enabled.

• md4c.MD_DIALECT_GITHUB - Parse GitHub-Flavored Markdown, which enables the following flags:

  • MD_FLAG_PERMISSIVEAUTOLINKS
  • MD_FLAG_TABLES
  • MD_FLAG_STRIKETHROUGH
  • MD_FLAG_TASKLISTS

HTML Renderer Option Flags

• md4c.MD_HTML_FLAG_DEBUG - For development use, send MD4C debug output to stderr.
• md4c.MD_HTML_FLAG_VERBATIM_ENTITIES - Do not replace HTML entities with the actual character (e.g. © with ©).
• md4c.MD_HTML_FLAG_SKIP_UTF8_BOM - Omit BOM from start of UTF-8 input.
• md4c.MD_HTML_FLAG_XHTML - Generate XHTML instead of HTML.

Enums

The MD4C library uses various enums to provide data to callbacks. PyMD4C uses IntEnums to encapsulate these.

See md4c.h from the MD4C project for more information on these enums and associated types.

Block Types - class BlockType

• md4c.BlockType.DOC - Document
• md4c.BlockType.QUOTE - Block quote
• md4c.BlockType.UL - Unordered list
• md4c.BlockType.OL - Ordered list
• md4c.BlockType.LI - List item
• md4c.BlockType.HR - Horizontal rule
• md4c.BlockType.H - Heading
• md4c.BlockType.CODE - Code block
• md4c.BlockType.HTML - Raw HTML block
• md4c.BlockType.P - Paragraph
• md4c.BlockType.TABLE - Table
• md4c.BlockType.THEAD - Table header row
• md4c.BlockType.TBODY - Table body
• md4c.BlockType.TR - Table row
• md4c.BlockType.TH - Table header cell
• md4c.BlockType.TD - Table cell

Span Types - class SpanType

• md4c.SpanType.EM - Emphasis
• md4c.SpanType.STRONG - Strong
• md4c.SpanType.A - Link
• md4c.SpanType.IMG - Image
• md4c.SpanType.CODE - Inline code
• md4c.SpanType.DEL - Strikethrough
• md4c.SpanType.LATEXMATH - Inline math
• md4c.SpanType.LATEXMATH_DISPLAY - Display math
• md4c.SpanType.WIKILINK - Wiki link
• md4c.SpanType.U - Underline

Text Types - class TextType

• md4c.TextType.NORMAL - Normal text
• md4c.TextType.NULLCHAR - NULL character
• md4c.TextType.BR - Line break
• md4c.TextType.SOFTBR - Soft line break
• md4c.TextType.ENTITY - HTML Entity
• md4c.TextType.CODE - Text inside a code block or inline code
• md4c.TextType.HTML - Raw HTML (inside an HTML block or simply inline HTML)
• md4c.TextType.LATEXMATH - Text inside an equation

Table Alignments - class Align

• md4c.Align.DEFAULT
• md4c.Align.LEFT
• md4c.Align.CENTER
• md4c.Align.RIGHT

Exceptions

• md4c.ParseError - Raised by one of the parse() methods when there is an error during parsing, such as running out of memory. There is no such thing as invalid syntax in Markdown, so this really only signals some sort of system error.

• md4c.StopParsing - A callback can raise this to stop parsing early. GenericParser's parse() method will catch it and abort quietly.

License

This project is licensed under the MIT license. See the LICENSE.md file for details.