| Date: | 2022-06-17 |
|---|---|
| Revision: | 9077 |
| Web site: | https://docutils.sourceforge.io/ |
| Copyright: | This document has been placed in the public domain. |
Contents
This is a work in progress. If you are reading a local copy, the master copy might be newer. This document uses are relative links; if they don't work, please use the master copy.
Please feel free to ask questions and/or provide answers; send email to the Docutils-users mailing list. Project members should feel free to edit the source text file directly.
Docutils is a system for processing plaintext documentation into useful formats, such as HTML, XML, and LaTeX. It supports multiple types of input, such as standalone files, PEPs (Python Enhancement Proposals), and string input. Client code may add other input types, e.g. Sphinx comes with an extension to extract inline documentation from Python modules and packages.
The Docutils distribution consists of:
For an overview of the Docutils project implementation, see PEP 258, "Docutils Design Specification".
Docutils is implemented in Python.
Docutils is short for "Python Documentation Utilities". The name "Docutils" was inspired by "Distutils", the Python Distribution Utilities architected by Greg Ward, a component of Python's standard library.
The earliest known use of the term "docutils" in a Python context was a fleeting reference in a message by Fred Drake on 1999-12-02 in the Python Doc-SIG mailing list. It was suggested as a project name on 2000-11-27 on Doc-SIG, again by Fred Drake, in response to a question from Tony "Tibs" Ibbs: "What do we want to call this thing?". This was shortly after David Goodger first announced reStructuredText on Doc-SIG.
Tibs used the name "Docutils" for his effort "to document what the Python docutils package should support, with a particular emphasis on documentation strings". Tibs joined the current project (and its predecessors) and graciously donated the name.
For more history of reStructuredText and the Docutils project, see An Introduction to reStructuredText.
Please note that the name is "Docutils", not "DocUtils" or "Doc-Utils" or any other variation. It is pronounced as in "DOCumentation UTILitieS", with emphasis on the first syllable.
See the links to editors supporting reStructuredText.
Docutils is mainly stable, with documented APIs and architecture subject to change after announcement and a transition period.
We fix bugs as they are reported. This means the code from the repository (or the snapshots) is the most stable as well as the most featureful most of the time.
Besides the regular releases, we provide automatically-generated snapshots which always contain the latest code from the repository. As the project matures, we may formalize on a stable/development-branch scheme, but we're not using anything like that yet.
reStructuredText is an easy-to-read, what-you-see-is-what-you-get plaintext markup syntax and parser system. The reStructuredText parser is a component of Docutils. reStructuredText is a revision and reinterpretation of the StructuredText and Setext lightweight markup systems.
If you are reading this on the web, you can see for yourself. The source for this FAQ is written in reStructuredText; open it in another window and compare them side by side.
A ReStructuredText Primer and the Quick reStructuredText user reference are a good place to start. The reStructuredText Markup Specification is a detailed technical specification.
The name came from a combination of "StructuredText", one of reStructuredText's predecessors, with "re": "revised", "reworked", and "reinterpreted", and as in the re.py regular expression module. For a detailed history of reStructuredText and the Docutils project, see An Introduction to reStructuredText.
"reStructuredText" is ONE word, not two!
"RST" and "ReST" (or "reST") are both acceptable. Care should be taken with capitalization, to avoid confusion with "REST", an acronym for "Representational State Transfer".
The abbreviations "reSTX" and "rSTX"/"rstx" should not be used; they overemphasize reStructuredText's predecessor, Zope's StructuredText.
It's ".txt". Some people would like to use ".rest" or ".rst" or ".restx", but why bother? ReStructuredText source files are meant to be readable as plaintext, and most operating systems already associate ".txt" with text files. Using a specialized filename extension would require that users alter their OS settings, which is something that many users will not be willing or able to do.
Also see What's the official MIME type for reStructuredText data?
See the list of editors supporting reStructuredText.
A uniquely-adorned section title at the beginning of a document is treated specially, as the document title. Similarly, a uniquely-adorned section title immediately after the document title becomes the document subtitle. For example:
This is the Document Title ========================== This is the Document Subtitle ----------------------------- Here's an ordinary paragraph.
Counterexample:
Here's an ordinary paragraph. This is *not* a Document Title ============================== The "ordinary paragraph" above the section title prevents it from becoming the document title.
Another counterexample:
This is not the Document Title, because... =========================================== Here's an ordinary paragraph. ... the title adornment is not unique ===================================== Another ordinary paragraph.
For example, say you want an em-dash (XML character entity —, Unicode character U+2014) in your document: use a real em-dash. Insert concrete characters (e.g. type a real em-dash) into your input file, using whatever encoding suits your application, and tell Docutils the input encoding. Docutils uses Unicode internally, so the em-dash character is a real em-dash internally.
Emacs users should refer to the Emacs Support for reStructuredText document. Tips for other editors are welcome.
ReStructuredText has no character entity subsystem; it doesn't know anything about XML charents. To Docutils, "—" in input text is 7 discrete characters; no interpretation happens. When writing HTML, the "&" is converted to "&", so in the raw output you'd see "—". There's no difference in interpretation for text inside or outside inline literals or literal blocks -- there's no character entity interpretation in either case.
If you can't use a Unicode-compatible encoding and must rely on 7-bit ASCII, there is a workaround. New in Docutils 0.3.10 is a set of Standard Substitution Definition Sets, which provide equivalents of XML & HTML character entity sets as substitution definitions. For example, the Japanese yen currency symbol can be used as follows:
.. include:: <xhtml1-lat1.txt> |yen| 600 for a complete meal? That's cheap!
Thanks to David Priest for the original idea.
If you insist on using XML-style charents, you'll have to implement a pre-processing system to convert to UTF-8 or something. That introduces complications though; you can no longer write about charents naturally; instead of writing "—" you'd have to write "&mdash;".
For the common case of long dashes, you might also want to insert the following substitution definitions into your document (both of them are using the "unicode" directive):
.. |--| unicode:: U+2013 .. en dash .. |---| unicode:: U+2014 .. em dash, trimming surrounding whitespace :trim:
Now you can write dashes using pure ASCII: "foo |--| bar; foo |---| bar", rendered as "foo – bar; foo—bar". The :trim: option for the em dash is necessary because you cannot write "foo|---|bar"; thus you need to add spaces ("foo |---| bar") and advise the reStructuredText parser to trim the spaces.
People have tossed the idea around, and some implementations of reStructuredText-generating tools can be found in the Docutils Link List.
There's no reason why reStructuredText should not be round-trippable to/from XML; any technicalities which prevent round-tripping would be considered bugs. Whitespace would not be identical, but paragraphs shouldn't suffer. The tricky parts would be the smaller details, like links and IDs and other bookkeeping.
For HTML, true round-tripping may not be possible. Even adding lots of extra "class" attributes may not be enough. A "simple HTML" to RST filter is possible -- for some definition of "simple HTML" -- but HTML is used as dumb formatting so much that such a filter may not be particularly useful. An 80/20 approach should work though: build a tool that does 80% of the work automatically, leaving the other 20% for manual tweaks.
There are several, with various degrees of completeness. With no implied endorsement or recommendation, and in no particular order:
Please let us know of any other reStructuredText Wikis.
With no implied endorsement or recommendation, and in no particular order:
Please let us know of any other reStructuredText Blogs.
Bullet & enumerated list markup is very intuitive but there are 2 points that must be noted:
Lists should not be indented. This is correct:
paragraph * list item 1 * nested item 1.1 * nested item 1.2 * list item 2
while this is probably incorrect:
paragraph
* list item 1
* nested item 1.1
* nested item 1.2
* list item 2
The extra indentation (of the list containing items 1.1 and 1.2) is recognized as a block quote. This is usually not what you mean and it causes the list in the output to be indented too much.
There must be blank lines around list items, except between items of the same level, where blank lines are optional. The example above shows this.
Note that formatting of the output is independent of the input, and is decided by the writer and the stylesheet. For instance, lists are indented in HTML output by default. See How are lists formatted in HTML? for details.
Some people like to write lists with indentation but don't intend a blockquote context. There has been a lot of discussion about allowing this in reStructuredText, but there are some issues that would need to be resolved before it could be implemented. There is a summary of the issues and pointers to the discussions in the to-do list.
Short answer: no.
In reStructuredText, it would be impossible to unambiguously mark up and parse lists without blank lines before and after. Deeply nested lists may look ugly with so many blank lines, but it's a price we pay for unambiguous markup. Some other plaintext markup systems do not require blank lines in nested lists, but they have to compromise somehow, either accepting ambiguity or requiring extra complexity. For example, Epytext does not require blank lines around lists, but it does require that lists be indented and that ambiguous cases be escaped.
Use the math directive and math role, available since Docutils 0.8.
Not currently, no. It's on the to-do list (details here), and hopefully will be part of the reStructuredText parser soon. At that time, markup like this will become possible:
Here is some *emphasized text containing a `hyperlink`_ and ``inline literals``*.
There are workarounds, but they are either convoluted or ugly or both. They are not recommended.
Inline markup can be combined with hyperlinks using substitution definitions and references with the "replace" directive. For example:
Here is an |emphasized hyperlink|_. .. |emphasized hyperlink| replace:: *emphasized hyperlink* .. _emphasized hyperlink: https://example.org
It is not possible for just a portion of the replacement text to be a hyperlink; it's the entire replacement text or nothing.
The "raw" directive can be used to insert raw HTML into HTML output:
Here is some |stuff|. .. |stuff| raw:: html <em>emphasized text containing a <a href="https://example.org">hyperlink</a> and <tt>inline literals</tt></em>
Raw LaTeX is supported for LaTeX output, etc.
Line blocks are designed for address blocks, verse, and other cases where line breaks are significant and must be preserved. Unlike literal blocks, the typeface is not changed, and inline markup is recognized. For example:
| A one, two, a one two three four | | Half a bee, philosophically, | must, *ipso facto*, half not be. | But half the bee has got to be, | *vis a vis* its entity. D'you see? | | But can a bee be said to be | or not to be an entire bee, | when half the bee is not a bee, | due to some ancient injury? | | Singing...
Here's a workaround for manually inserting explicit line breaks in HTML output:
.. |br| raw:: html <br /> I want to break this line here: |br| this is after the break. If the extra whitespace bothers you, |br|\ backslash-escape it.
Asterisks are valid URL characters (see RFC 2396), sometimes used in URLs. For example:
http://cvs.example.org/viewcvs.py/*checkout*/module/file
Unfortunately, the parser thinks the asterisks are indicating emphasis. The slashes serve as delineating punctuation, allowing the asterisks to be recognized as markup. The example above is separated by the parser into a truncated URL, an emphasized word, and some regular text:
http://cvs.example.org/viewcvs.py/ *checkout* /module/file
To turn off markup recognition, use a backslash to escape at least the first asterisk, like this:
http://cvs.example.org/viewcvs.py/\*checkout*/module/file
Escaping the second asterisk doesn't hurt, but it isn't necessary.
Use the parsed-literal directive.
Scenario: a document contains some source code, which calls for a literal block to preserve linebreaks and whitespace. But part of the source code should be formatted, for example as emphasis or as a hyperlink. This calls for a parsed literal block:
.. parsed-literal:: print "Hello world!" # *tricky* code [1]_
The emphasis (*tricky*) and footnote reference ([1]_) will be parsed.
Docutils and reStructuredText can be used with or as a component of a templating system, but they do not themselves include templating functionality. Templating should simply be left to dedicated templating systems. Users can choose a templating system to apply to their reStructuredText documents as best serves their interests.
There are many good templating systems for Python (ht2html, YAPTU, Quixote's PTL, Cheetah, etc.; see this non-exhaustive list of some other templating systems), and many more for other languages, each with different approaches. We invite you to try several and find one you like. If you adapt it to use Docutils/reStructuredText, please consider contributing the code to Docutils or let us know and we'll keep a list here.
One reST-specific web templating system is rest2web, a tool for automatically building websites, or parts of websites.
There is no specific syntax for FAQs and Q&A lists. Here are two options:
For a FAQ (Frequently Asked Questions, usually with answers), a convenient way to mark up the questions is as section titles, with the answer(s) as section content. This document is marked up in this way.
The advantages of using section titles for questions are: sections can be numbered automatically, and a table of contents can be generated automatically. One limitation of this format is that questions must fit on one line (section titles may not wrap, in the source text). For very long questions, the title may be a summary of the question, with the full question in the section body.
Field lists work well as Q&A lists:
:Q: What kind of questions can we
put here?
:A: Any kind we like!
In order to separate questions, lists can be used:
Q: What kind of question can we put here? A: Any kind we like!
Q: How many answers can a question have? A: It can have one, A: or more. A3: Answers can be numbered like this. A:
- Or like this.
- We're flexible!
If you don't want to number or otherwise mark questions, you can use an empty comment between individual field lists to separate them:
:Q: First question? :A: Answer. .. :Q: Second question? :A: Answer.
Languages written from right to left, such as Arabic and Hebrew, must be reordered according to the Unicode Bidi Algorithm. This requires support from the editor and special markup in the output format.
The source format of reStructuredText is relatively bidi-friendly: most constructs are denoted by punctuation without intrusion of English and when you must write in English, it's usually on a separate line. So any editor that auto-detects direction per-line (like gedit or geresh) will suffice.
Moreover, it's possible to translate all reStructuredText keywords. Docutils 0.17 includes translations for Arab, Hebrew, and Persian/Farsi. This should allow reasonable use of editors limited to a single base direction for the whole document (like Notepad, Vim and text boxes in Firefox).
The second problem is bidi markup of the output. There is an almost transparent implicit solution for HTML:
There is also an explicit way to set directions through CSS and classes in the HTML:
Copy default.css to a new file and add relevant parts of the following:
/* Use these two if the main document direction is RTL */
body { direction: rtl; }
div.sidebar { float: left !important; }
/* The next 3 rules are very useful in documents containing pieces
of code in english */
/* Use this if you all your literal blocks (::) are LTR */
pre {direction: ltr; unicode-bidi: embed; }
/* Use this if you all your inline literals (``) are LTR */
tt {direction: ltr; unicode-bidi: embed; }
/* Use this if you all your interpreted text (`) is LTR */
cite {direction: ltr; unicode-bidi: embed; }
/* Allow manual direction override by class directive and roles */
.rtl { direction: rtl; }
.ltr { direction: ltr; }
Select this new stylesheet with --stylesheet=<file> or the stylesheet setting.
Now if you need to override the direction of some element (from a paragraph to a whole section), write:
.. class:: rtl
or:
.. class:: ltr
before it (see the class directive for details).
To change the direction of some inline text fragment, you can use RLE/LRE/PDF control characters, or write :rtl:`RTL text` / :ltr:`RTL text`. To use the latter syntax, you must write this once at the beginning of your document:
.. role:: ltr .. role:: rtl
LaTeX is quite hard to implement (it doesn't support the bidi algorithm, so all direction changes - even numbers in RTL text - must be explicitly marked). Other formats are more-or-less easy.
If you have any questions/problems/bugs related to bidi with docutils, ask Beni Cherniavsky directly or the Docutils-users mailing list.
While there is no registered MIME type for reStructuredText, the "official unofficial" standard MIME type is "text/x-rst". [1] This was invented for the build system for PEPs (Python Enhancement Proposals), and it's used by the python.org web site build system.
Also see What's the standard filename extension for a reStructuredText file?
| [1] | The "x-" prefix means it's an unregistered MIME type. |
The default HTML Writer module, docutils/writers/html4css1.py, is a proof-of-concept reference implementation. While it is a complete implementation, some aspects of the HTML it produces may be outdated or incompatible with older browsers or specialized applications (such as web templating).
For the full selection see Docutils HTML writers
No specific browser is targeted; all modern graphical browsers should work. Some older browsers, text-only browsers, and browsers without full CSS support are known to produce inferior results. Firefox, Safari, Opera, Chrome, Edge, and MS Internet Explorer (version 5.0 and up) are known to give good results. Reports of experiences with other browsers are welcome.
Here's the question in full:
I have this text:
Heading 1 ========= All my life, I wanted to be H1. Heading 1.1 ----------- But along came H1, and so shouldn't I be H2? No! I'm H1! Heading 1.1.1 ************* Yeah, imagine me, I'm stuck at H3! No?!?When I run it through tools/rst2html.py, I get unexpected results (below). I was expecting H1, H2, then H3; instead, I get H1, H1, H2:
... <html lang="en"> <head> ... <title>Heading 1</title> </head> <body> <div class="document" id="heading-1"> <h1 class="title">Heading 1</h1> <-- first H1 <p>All my life, I wanted to be H1.</p> <div class="section" id="heading-1-1"> <h1><a name="heading-1-1">Heading 1.1</a></h1> <-- H1 <p>But along came H1, and so now I must be H2.</p> <div class="section" id="heading-1-1-1"> <h2><a name="heading-1-1-1">Heading 1.1.1</a></h2> <p>Yeah, imagine me, I'm stuck at H3!</p> ...What gives?
Check the "class" attribute on the H1 tags, and you will see a difference. The first H1 is actually <h1 class="title">; this is the document title, and the default stylesheet renders it centered. There can also be an <h2 class="subtitle"> for the document subtitle.
If there's only one highest-level section title at the beginning of a document, it is treated specially, as the document title. (Similarly, a lone second-highest-level section title may become the document subtitle.) See How can I indicate the document title? Subtitle? for details. Rather than use a plain H1 for the document title, we use <h1 class="title"> so that we can use H1 again within the document. Why do we do this? HTML only has H1-H6, so by making H1 do double duty, we effectively reserve these tags to provide 6 levels of heading beyond the single document title.
With "html4css1", HTML is being used for dumb formatting for nothing but final display. A stylesheet is required, and one is provided; see Docutils HTML writers. Of course, you're welcome to roll your own. The default stylesheet provides rules to format <h1 class="title"> and <h2 class="subtitle"> differently from ordinary <h1> and <h2>:
h1.title {
text-align: center }
h2.subtitle {
text-align: center }
If you don't want the top section heading to be interpreted as a title at all, disable the doctitle_xform setting (--no-doc-title option). This will interpret your document differently from the standard settings, which might not be a good idea. If you don't like the reuse of the H1 in the HTML output, you can tweak the initial_header_level setting (--initial-header-level option) -- but unless you match its value to your specific document, you might end up with bad HTML (e.g. H3 without H2).
(Thanks to Mark McEahern for the question and much of the answer.)
Note
For the html5 writer, initial_header_level defaults to 2 because this is what the HTML5 standard expects as start value for headings nested in <section> elements.
If list formatting looks strange, first check that you understand list markup.
By default, HTML browsers indent lists relative to their context. This follows a long tradition in browsers (but isn't so established in print). If you don't like it, you should change the stylesheet.
This is different from how lists look in reStructuredText source. Extra indentation in the source indicates a blockquote, resulting in too much indentation in the browser.
A list item can contain multiple paragraphs etc. In complex cases list items are separated by vertical space. By default this spacing is omitted in "simple" lists. A list is simple if every item contains a simple paragraph and/or a "simple" nested list. For example:
text
- simple
- simple
- simple
- simple
text after a nested list
multiple
paragraphs
In this example the nested lists are simple (and should appear compacted) but the outer list is not.
If you want all lists to have equal spacing, disable the compact_lists setting (--no-compact-lists option). The precise spacing can be controlled in the stylesheet.
Note again that this is not exactly WYSIWYG: it partially resembles the rules about blank lines being optional between list items in reStructuredText -- but adding/removing optional blank lines does not affect spacing in the output! It's a feature, not a bug: you write it as you like but the output is styled consistently.
The rendering of enumerators (the numbers or letters acting as list markers) is completely governed by the stylesheet, so either the browser can't find the stylesheet (try enabling the embed_stylesheet setting [--embed-stylesheet option]), or the browser can't understand it (try a not too old Firefox, Chrome, Opera, Safari, Edge, or even MSIE).
What you're seeing is most probably not garbage, but the result of a mismatch between the actual encoding of the HTML output and the encoding your browser is expecting. Your browser is misinterpreting the HTML data, which is encoded text. A discussion of text encodings is beyond the scope of this FAQ; see one or more of these documents for more info:
The common case is with the default output encoding (UTF-8), when either numbered sections are used (via the "sectnum" directive) or symbol-footnotes. Three non-breaking spaces are inserted in each numbered section title, between the generated number and the title text. Most footnote symbols are not available in ASCII, nor are non-breaking spaces. When encoded with UTF-8 and viewed with ordinary ASCII tools, these characters will appear to be multi-character garbage.
You may have an decoding problem in your browser (or editor, etc.). The encoding of the output is set to UTF-8, but your browser isn't recognizing that. You can either try to fix your browser (enable "UTF-8 character set", sometimes called "Unicode"), or choose a different output-encoding. You can also try --output-encoding=ascii:xmlcharrefreplace for HTML or XML, but not applicable to non-XMLish outputs.
If you're generating document fragments, the "Content-Type" metadata (between the HTML <head> and </head> tags) must agree with the encoding of the rest of the document. For UTF-8, it should be:
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
Also, Docutils normally generates an XML declaration as the first line of the output. It must also match the document encoding. For UTF-8:
<?xml version="1.0" encoding="utf-8" ?>
(This is usually needed when using Docutils in conjunction with a templating system.)
You can use the docutils.core.publish_parts() function, which returns a dictionary containing an 'html_body' entry.
Full question:
Docutils' HTML output looks like XHTML and is advertised as such:
<?xml version="1.0" encoding="utf-8" ?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "https://www.w3.org/TR/xht ml1/DTD/xhtml1-transitional.dtd">But this is followed by:
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />Shouldn't this be "application/xhtml+xml" instead of "text/html"?
Short answer: if we didn't serve XHTML as "text/html" (which is a perfectly valid thing to do), it couldn't be viewed in Internet Explorer.
Long answer: In a perfect web, the Docutils XHTML output would be 100% strict XHTML. But it's not a perfect web, and a major source of imperfection is Internet Explorer. When the "html4css1" writer was created, IE represented the overwhelming majority of web browsers "in the wild". The behaviour is kept for backwards compatibility while modern, semantic HTML output is available with the "html5" writer.
(Thanks to Martin F. Krafft, Robert Kern, Michael Foord, and Alan G. Isaac.)
Yes, in conjunction with other projects.
The Sphinx documentation generator includes an autodoc module.
Version 2.0 of Ed Loper's Epydoc supports reStructuredText-format docstrings for HTML output. Docutils 0.3 or newer is required. Development of a Docutils-specific auto-documentation tool will continue. Epydoc works by importing Python modules to be documented, whereas the Docutils-specific tool, described above, will parse modules without importing them (as with HappyDoc, which doesn't support reStructuredText).
The advantages of parsing over importing are security and flexibility; the disadvantage is complexity/difficulty.
For more details, please see "Docstring Extraction Rules" in PEP 258, item 3 ("How").
Not directly, no. It borrows bits from DocBook, HTML, and others. I (David Goodger) have designed several document models over the years, and have my own biases. The Docutils document model is designed for simplicity and extensibility, and has been influenced by the needs of the reStructuredText markup.
Follow the Python coding conventions and documentation conventions in the Docutils Policies. Ensure the addition works with all supported Python versions.
Look at the Docutils sources to see how similar features are implemented, learn to do it "the Docutils way".
Prepare tests. Test cases are also examples and showcases for new features.
Include documentation.
For larger changes, consider creating a feature branch in a Docutils repository checkout. [2]
Mail your patch to the Docutils-develop mailing list or attach it to the relevant ticket at Docutils' bug tracker or feature request tracker. We accept patches created with diff, SVN, or Git.
The developers will make sure that contributions fit nicely into Docutils. This might involve discussing (and compromising on) design and implementation details. It might also lead to the conclusion that the addition fits better in the extensions and related projects.
| [2] | Working with branches is much easier with Git. You can get a Git clone of the repository from http://repo.or.cz/w/docutils.git or with git-svn. |