Skip to main content

Python-Markdown extension using smartypants to emit typographically nicer ("curly") quotes, proper ("em" and "en") dashes, etc.

Project description

Markdown is great, but if you want pretty “curled” quotes, real em- and en-dashes, and the other typographic prettification that our modern Unicode- and Web-savvy world affords, it needs to be married with smartypants (or an equivalent module) to turn ugly, programmer-ish punctuation into pretty typographic punctuation. This module does that.


import markdown

text = """
Markdown makes HTML from simple text files. But--it lacks typographic
"prettification." That... That'd be sweet. Definitely 7---8 on a '10-point
scale'. Now it has it.


print markdown.markdown(text, extensions=['smartypants(entities=named)'])

This produces nice HTML output, including typographically “pretty” quotes and other punctuation. It also (optionally) renders HTML entities in their named rather than numeric form, which is easier on the eyes and more readily comprehended:

<p>Markdown makes HTML from simple text files. But&mdash;it lacks
typographic &ldquo;prettification.&rdquo; That&hellip; That&rsquo;d be
sweet. Definitely 7&ndash;8 on a &lsquo;10-point scale&rsquo;. Now it has

Note that you don’t really need to do an import mdx_smartypants. You’re welcome to if you like, and it may help to advertise that the code depends on mdx_smartypants being available. But markdown will look for mdx_smartypants simply by virtue of the extensions=['smartypants'] declaration.

mdx_smartypants will not massage code blocks (either indentded or fenced), or HTML included within <pre> sections, so your program snippets are safe.

RTL Languages and Alternative Quotation Marks

Right-to-left languages such as Arabic, Hebrew, Persian, and Urdu reverse the convention seen in English and other left-to-right languages. The “left” quotation mark is really the “starting” quotation mark–and it should appear to the right of the “right” quotation mark. The “right” quotation mark, similarly, is really the “ending” mark, and appears to the left of the “right” mark. This is clearly not something that was front-and-center even to the internationally-minded Unicode community, given how “left” and “right” are embedded in the official glyph names–a misnomer that carries over into HTML entities.

The historical smartypants module similarly thinks in LTR terms. It even hard-codes the HTML entities used for quotation marks. To address this, this module’s bundled spants derivative uses variable quotation marks, and provides a middleman class Quotes which allows defining which HTML entities should be used for starting single, ending single, starting double, and ending double quotation marks, respectively. It also provides a mechanism for defining the directionality of text. When emitting for RTL languages, the normal left/right conventions are reversed.

Quotes.set(ssquo, esquo, sdquo, edquo, dir) allows you to set one or more of these values. If you are changing the direction of quoting away from LTR, it’s best to redefine all of the quotes so that everything is consistently defined and ordered.

Quotes.reset() puts everything back to factory defaults. Perhaps most usefully, Quotes.configure_for_text(text) guesses what direction the language is rendered, and sets quotes accordingly. In order to provide a fire-and-forget experience, unless the user sets the language direction explicitly, this heuristic will be invoked as a normal part of mdx_smartypants operation. Also note: If called directly, this API must be provided pure, plain text–not text wrapped in HTML or other markup (which will fool the language guesser into improperly guessing English). If the user has explicitly set language direction, the guess will not be made–but an optional force Boolean parameter can be supplied to specify that previous explicit direction setting should be ignored, and guessing commenced.

This API and functionality is brand new; tests have been added and successfully passed for it, but it should be considered somewhat experimental for now.

Digging even deeper, a great variety and vast diversity of different quotation styles are used in different languages. While there is no automatic support for styles that differ from English, Quotes.set can be called with any HTML entities, allowing pretty much any convention to be supported. For example:

Quotes.set(r'&lasquo;', r'&rasquo;', r'&laquo;', r'&raquo;')  # Swiss French
Quotes.set(r'&sbquo;',  r'&lsquo;',  r'&bdquo;', r'&ldquo;')  # German or Czech

For «Swiss» and ‹French› (first one) and „German“ and ‚Czech‘ (second one).

NB I do not have any experience with RTL, top-to-bottom languages such as traditional Chinese and Japanese scripts. If additional changes are required to properly support that directionality, I’d be happy to hear about it.


Originally mdx_smartypants output named HTML entities. That behavior is now configurable. By default, Unicode characters and entities are not changed from whatever markdown emits. But you can choose that non-ASCII characters are mapped to named entities, numeric entities, unicode entities (really not entities, just Unicode characters), or None (no mapping performed).


  • As of version 1.5, named entities are no longer the default. One can still request named entities, as shown in the example above.

  • As of version 1.4, mdx_smartpants attempts to automagically guess the direction of text flow used by the underlying language (e.g. LTR or RTL) and arrange quotation marks accordingly. Thanks to Ahmad Khayyat for the bug report and discussion that led to this upgrade. This release also moved to a package-oriented distribution, given the additional modules required.

  • As of version 1.2, mdx_smartpants no longer uses the stock smartypants module from PyPI. It incorporates a copy of the module, called spants, in order to tweak the code for Python 3 compatibility, to fix the incorrect munging of punctuation within style blocks, and to make other improvements. This is a partial step towards a rewrite of smartypants itself to support Python 3 and be more in-line with modern Python idioms.

  • Now successfully packaged for, and tested against, against Python 2.6, 2.7, and 3.3, as well as against PyPy 1.9 (based on 2.7.2). As of Version 1.4, official support for Python 2.5 and 3.2 withdrawn; while it may work on these, I can no longer test those versions. Also, they’re obsolete. Time to upgrade!

  • Automated multi-version testing managed by the awesome pytest and tox.

  • The author, Jonathan Eunice or @jeunice on Twitter welcomes your comments and suggestions.


pip install -U mdx_smartypants

To use pip to install under a specific Python version, look for a program such as pip-3.3 (e.g. which pip-3.3 on Unix derived systems). Failing this, you may be able to easy_install under a specific Python version (3.3 in this example) via:

python3.3 -m easy_install --upgrade mdx_smartypants

(You may need to prefix these with “sudo “ to authorize installation.)

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions (30.1 kB view hashes)

Uploaded Source

mdx_smartypants-1.5.1.tar.gz (19.5 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page