ReStructuredText Style Guide

This page describes how reStructuredText (reST) is written for DM documentation through examples. The last section documents our formatting conventions.

For more general guides to writing reStructuredText, see Sphinx’s reStructuredText Primer and the docutils Quick reStructuredText guide.

See the pages on writing documentation for the LSST Stack and Python docstrings for specific reST usage in those contexts.

Sample

#######################
ReStructuredText Sample
#######################

ReStructuredText is an *extensible* markup language used by `LSST`_.

.. _LSST: http://lsst.org

ReStructuredText provides basic *italic*, **bold** and ``monospaced``
typesetting.  There is also the concept of **roles** that provide sophisticated
typesetting, such as :math:`\mu = -2.5 \log_{10}(\mathrm{DN} / A) + m_0`, and
:ref:`referencing <rst-internal-links>`.

.. _label-for-subsection-label:

Sectioning
==========

Sections are formed with underlining the headline text. We use :ref:`a
conventional sequence of underline symbols <rst-sectioning>` to indicate
different levels of hierarchy.

Directives
==========

Besides **roles** that are used for inline markup, reStructuredText has the
concept of **directives** to markup *blocks* of content. One example is the is
the ``code-block`` directive:

.. code-block:: python

   print('hello world!')

See this sample rendered.

Inline Text Styling

Italics
*italic text* → italic text.
Bold
**bold text** → bold text.
Monospace
``monospace text`` → monospace text. When referring to code objects, it’s better to use markup that links to the object’s API documentation (see the Links to code objects section).
Inline math
:math:`\sqrt{16}` → \(\sqrt{16}\) (See also the Math section).

Note

Inline styles can’t be nested

For example, you can’t write *see :ref:`this page <label>`*.

Inline markup also needs to be surrounded by white space, though trailing punctuation is fine. You can get around this with an escaped space that is otherwise invisible, For example one\ *word* renders as oneword.

Other semantic markup

In addition to the fundamental inline typesetting styles above, you may use additional reST roles to provide semantic meaning to the text. The documentation’s CSS takes advantage of this semantic meaning to provide visual cues to readers.

Abbreviations
:abbr:`LSST (Large Synoptic Survey Telescope` → LSST (a tool tip exposes the definition)
Filenames and paths
:file:`repos.yaml` → repos.yaml
Shell commands
:command:`git rebase -i master` → git rebase -i master
User interface labels
:guilabel:`New Pull Request` → New Pull Request. This markup can be used for button labels, menus, or even text labels in interactive shell programs.
Keyboard commands
:kbd:`Control-a s` → Control-a s. Spell out the keys rather than using Emacs short hand, such as C-x.

To semantically markup Python or C++ code objects, refer to the section on Links to code objects.

Lists

Unordered lists can be written as:

- First item

  Second paragraph for first item, needs to be consistently indented.
- Second item

- You can put spaces between items, or not.

- Hierarchical lists are also possible

  - Put a blank space before the sub-list
  - And indent the sub-list consistently

- Last item.

which renders as:

  • First item

    Second paragraph for first item, needs to be consistently indented.

  • Second item

  • You can put spaces between items, or not.

  • Hierarchical lists are also possible

    • Put a blank space before the sub-list
    • And indent the sub-list consistently

  • Last item.

There should be a blank line before and after the list to separate the list from paragraphs. Blanks lines are allowed between list items as well.

Enumerated lists can be written similarly:

1. First thing
2. Second thing

or automatically enumerated,

#. First thing
#. Second thing

which renders as:

  1. First thing
  2. Second thing

or automatically enumerated,

  1. First thing
  2. Second thing

Definition lists

Definition lists are terms with an indented content section. For example:

LSST
    Large Synoptic Survey Telescope

DM
    Data Management

produces

LSST
Large Synoptic Survey Telescope
DM
Data Management

Definition lists are not limited to dictionary-like usage; they can be employed whenever a series of terms with associated micro content is needed.

Sections

We create section hierarchies as follows:

##########
Page Title
##########

Titles have hash marks above and below.

By convention, titles and section headings are set off from surrounding text by
a single blank line above and below. All levels of section headings may have
named labels, which appear before the heading. We encourage you to add labels
to all sections so that they can be referenced. Names are global, so be
specific. See :ref:`Internal Links to Labels <rst-internal-links>` for
more information. The following section heading has a label named
"``section-headings-example-section``".

.. _section-headings-example-section:

Section Heading
===============

Section headers are set with an underline. The sequence of underline characters
used (``=``, then ``-``, then ``^`` and finally ``"``) indicates the section
hierarchy.

.. _section-headings-example-subsection:

Subsection Heading
------------------

Maecenas congue ligula ac quam viverra nec consectetur ante hendrerit.
Donec et mollis dolor.

.. _section-headings-example-subsubsection:

Sub-subsection Heading
^^^^^^^^^^^^^^^^^^^^^^

Praesent et diam eget libero egestas mattis sit amet vitae augue.

.. _section-headings-example-subsubsubsection:

Sub-sub-subsection Heading
""""""""""""""""""""""""""

Nam tincidunt congue enim, ut porta lorem lacinia consectetur.


Sections in Python docstrings are a special case. We do not put a blank space
between a headline and object lists below, and we do not add explicit section
labels. See the :doc:`docstring style guide <../python/numpydoc>` for more information.

See this sample rendered.

This specific sequence of section markup styles is not mandated by the reST specification, but we encourage you to use it for consistency across all DM reST documents.

Sections in Python docstrings are a special case. First, we do not place a blank space between a section header and the object lists below. Second, Python docstrings can only use subsection and subsubsection-level headings.

Linking

Tables

We recommend that you use the grid syntax for tables, since they more flexible than ‘simple’ reST tables. And although not necessary, we suggest that you provide a caption using the table directive and a label prefixed with “table-.” For example:

.. _table-label:

.. table:: Table caption.

    +------------------------+------------+----------+----------+
    | Header row, column 1   | Header 2   | Header 3 | Header 4 |
    | (header rows optional) |            |          |          |
    +========================+============+==========+==========+
    | body row 1, column 1   | column 2   | column 3 | column 4 |
    |                        | with many  | spans    |          |
    |                        | rows       | both     |          |
    +------------------------+------------+ rows     +----------+
    | body row 2             | ...        |          | ...      |
    +------------------------+------------+----------+----------+

produces:

Header row, column 1 (header rows optional) Header 2 Header 3 Header 4
body row 1, column 1 column 2 with many rows column 3 spans both rows column 4
body row 2 … …

Note how cells can be joined by omitting the dividing line. The = characters divide the header from table content. Text in the header is set in bold.

You can write tables with multiple header rows, including spans across header cells:

.. _rst-table-multi-header-example:

.. table:: Table with two header rows, including a span.

   +---------------------------------------------------+
   | centos-7-stack-lsst_apps-w_2015_45-20151130234354 |
   +-----------+---------------------------------------+
   | Region    | AMI                                   |
   +===========+=======================================+
   | us-east-1 | ami-e2490b88                          |
   +-----------+---------------------------------------+
   | us-west-2 | ami-9a0f1dfb                          |
   +-----------+---------------------------------------+

produces:

centos-7-stack-lsst_apps-w_2015_45-20151130234354
Region AMI
us-east-1 ami-e2490b88
us-west-2 ami-9a0f1dfb

In the simplest cases, tables are not required to have headers, or even be inside a table directive.

+-----------+--------------+
| us-east-1 | ami-e2490b88 |
+-----------+--------------+
| us-west-2 | ami-9a0f1dfb |
+-----------+--------------+

produces:

us-east-1 ami-e2490b88
us-west-2 ami-9a0f1dfb

Be sure to leave a blank line before and after the table directive.

Images and Figures

Plain images

Plain images can be included with the image directive. For example:

.. image:: /_static/development/docs/lsst_logo.jpg
   :target: ../_images/lsst_logo.jpg
   :alt: LSST Logo
LSST Logo

This example shows how an image can by hyperlinked to any URL with the target field. Internal links, as in the example, must be relative to the reST document; Sphinx does not process URLs in an image’s target field.

The image directive has more configurable fields. If image sizes need to be manipulated from reST, we recommend using scale since it is responsive. We hope to provide better support for responsive image sizing.

Be sure to leave a blank line before and after the image directive.

Figure directive

Figures include both an image and a caption. For example:

.. figure:: /_static/development/docs/lsst_logo.jpg
   :name: fig-example-figure-label
   :target: ../_images/lsst_logo.jpg
   :alt: LSST Logo

   LSST Logo.
LSST Logo

LSST Logo.

Note that the :name: field takes the place of a separate label for hyperlinking. By convention, these labels should be prefixed with “fig-.”

Be sure to leave a blank line before and after the figure directive.

Note on paths to image files

Images are included in the _static/ directory of the git repository for this documentation project. Sphinx requires image assets to be located in this _static/ directory in order to properly copy files into the built website. By using a prefix “/” we indicate that a path is relative to the root of the documentation repository.

Package documentation is hosted in doc/ directories of the git repositories of individual Stack packages. For such package documentation, image files should be placed inside a directory in doc/_static/ named for the package itself. For example doc/_static/obs_decam/ for the obs_decam package. This nested directory structure is needed to merge package documentation content into the root documentation build.

Math

Sphinx allows you to write math expressions with a LaTeX-like plain text syntax that will be typeset by MathJax in the browser. MathJax supports AMSMath-LaTeX syntax. This website by Dr Carol Burns provides a comprehensive listing of available LaTeX syntax in MathJax, along with examples.

In Sphinx, you can either write inline expressions with the math role, or block elements with the math directive.

Inline math

Write inline math expressions with the math role. For example, :math:`\sigma_\mathrm{mean} = \sigma / \sqrt{N}` produces \(\sigma_\mathrm{mean} = \sigma / \sqrt{N}\).

Block math

To display math as a block element, use the math directive (be sure to leave a blank line before and after the math directive). For example:

.. math:: \sigma_\mathrm{mean} = \frac{\sigma}{\sqrt{N}}
   :label: math-sample

renders as

(1)\[\sigma_\mathrm{mean} = \frac{\sigma}{\sqrt{N}}\]

Referencing equations

Notice the :label: field in the previous sample; it both annotates the equation with a number, and allows the equation to be cross-referenced with the eq role; for example `:eq:`math-sample` produces (1). Equation references may only be made within the same reStructuredText page as the original math directive. See the Sphinx docs on Math support for more information.

Multiple Equations

Multiple equations can appear in the same math directive. Simply include a blank line between each equation (and don’t include an equation as a argument of the math directive itself). For example:

.. math::

   \nabla \cdot \mathbf{E} = \frac{\rho}{\epsilon_0}

   \nabla \cdot \mathbf{B} = 0

renders as

\[ \begin{align}\begin{aligned}\nabla \cdot \mathbf{E} = \frac{\rho}{\epsilon_0}\\\nabla \cdot \mathbf{B} = 0\end{aligned}\end{align} \]

Aligned Equations

Often when there are multiple statements in a math directive it’s desirable to align those statements around the equals sign, for example. In AMSMath-LaTeX this would be achieved with the align environment. In reStructuredText we can accomplish the same in a math directive:

.. math::

   x &= (a + b)^2 \\
     &= a^2 + 2ab + b^2

renders as

\[\begin{split}x &= (a + b)^2 \\ &= a^2 + 2ab + b^2\end{split}\]

Notice how the alignment point is marked with an & and \\ is appended to each math statement except for the last. Also note how there are no blank lines between math statements.

Source code

For blocks of code, we prefer the code-block directive. This directive has the form

.. code-block:: <language>
   :name: optional-label
   :emphasize-lines: <optional lines to highlight>

   <code>

where

  • <language> can be any token understood by Pygments, particularly py (python), cpp (C++), java (Java), js (JavaScript) and rst (reStructuredText). Specify none to disable highlighting.
  • :name: is an explicit hyperlink label for the code block.
  • :emphasize-lines: is an optional sequence of lines to highlight. This can be comma-separated, with hyphens to indicate spans.

For example:

.. code-block:: py
   :name: context-timer-example
   :emphasize-lines: 4-13,15-17

   from contextlib import ContextDecorator
   import time
   
   class timercontext(ContextDecorator):

       def __enter__(self):
           self.start = time.clock()
           return self
   
       def __exit__(self, *args):
           self.end = time.clock()
           self.interval = self.end - self.start
           print('Duration: {0:.2e} sec'.format(self.interval))

    @timercontext
    def run_slowly():
        time.delay(1.)
    
    run_slowly()
    
    with timercontext() as t:
        time.delay(1)
    
    print('Delayed for {0:.1f}'.format(t.interval))

produces

from contextlib import ContextDecorator
import time

class timercontext(ContextDecorator):

    def __enter__(self):
        self.start = time.clock()
        return self

    def __exit__(self, *args):
        self.end = time.clock()
        self.interval = self.end - self.start
        print('Duration: {0:.2e} sec'.format(self.interval))

 @timercontext
 def run_slowly():
     time.delay(1.)

 run_slowly()

 with timercontext() as t:
     time.delay(1)

 print('Delayed for {0:.1f}'.format(t.interval))

Be sure to leave a blank line before and after the code-block directive.

Including source code examples from other files with literalinclude

The code-block directive is great for code examples written in the reStructuredText source file itself. You might also want to show a code sample contained in a separate file. For this you can use the literalinclude directive:

.. literalinclude:: path/to/example.py
   :language: py

The source path can either be relative to the reST document or relative to the documentation root by prefixing the path with /.

The literalinclude directive also supports code-block fields, such as name and emphasize-lines. In addition, you can selective include ranges of lines with the lines field. For example, to include only lines 10 – 20:

.. literalinclude:: path/to/example.py
   :language: py
   :lines: 10-20

To omit the first two lines from a file:

.. literalinclude:: path/to/example.py
   :language: py
   :lines: 2-

Sophisticated inclusion patterns can be achieved by listing multiple spans, such as :lines: 3-10,20-, which shows the first ten lines and all lines after the 20th.

When including code example snippets from other files, it may be useful to remove indentation. Use the dedent field for that. For example:

.. literalinclude:: path/to/example.py
   :language: py
   :lines: 5-10
   :dedent: 4

will show lines 5 – 10 and remove 4 space characters (presumably because the snippet is inside a Python class or function).

Lightweight syntax for Python code and sessions

You can markup Python code blocks using a lightweight syntax:

::

   print('hello world!')

Interactive python sections can be marked up as

>>> print('Hello world!')
Hello world!

This lightweight syntax is used in Python docstrings.

Command line prompts

For generic command line prompts, we use the sphinx-prompt extension so that the prompt character (e.g, $) isn’t selectable. This is great for giving copy-and-paste-ready command line instructions.

For basic bash prompts,

.. prompt:: bash

   mkdir -p hello/world
   cd hello/world

produces

mkdir -p hello/world
cd hello/world

Footnotes

Footnotes should be used sparingly, if at all, in LSST documentation. Prefer inline hyperlinks to other sections. If you do need footnotes, you can make them as follows:

This is a line.\ [#label]_

.. [#label] This is the footnote content.

Note that we had to provide an escaped space for the footnote mark occurring after a period.

The footnote content should occur not far from the inline footnote mark; generally provide the footnote content at the end of the section.

Citations

Citations should be used for scholarly references; use hyperlinks for web native content. Citations can be made as follows:

The LSST Project [Ivezic2008]_ will produce 15 TB of images per night.

.. [Ivezic2008] Ivezic et al 2008. *LSST: from Science Drivers to
                Reference Design and Anticipated Data Products.*
                `arxiv:0805.2366 <http://arxiv.org/abs/0805.2366>`_

Citations are distinguished from footnotes in that the label does not begin with a #.

In the future, scholarly citations will be easier to include and more ‘latex-like’ with our documenteer Sphinx extensions.

Comments in ReStructuredText

Provide comments to fellow writers using ..,

.. This is a one-line comment.

..
   This is the first of a multi-paragraph comment.

   The second paragraph.

Avoid using comments to keep around old or alternate versions of text; prefer using Git version control instead.

RestructuredText Formatting Conventions

Text wrapping

When writing reST documentation in Python docstrings, documentation lines should be wrapped at lengths of 110 characters for consistency with our Python Style Guide.

For reStructuredText documents (e.g., .rst files), reST doesn’t care about line formatting. Emacs users, for example, are free to use hard-wrap formatting lines at 72 characters if that helps you write docs. Whenever possible, we encourage you to use soft-wrapping for your text. This allows others format text columns in their editors as they wish. As well, the GitHub.com code editor does not have hard-wrap auto-formatting. Those making doc edits on GitHub.com will tend to use soft-wrap by default (see ‘GitHub Flow in the Browser’).

When using soft-wrap formatting, you might write one sentence per line (i.e., put a line break after each sentence). As a writer, this has the advantage of making it easier to check the rhythm of your writing, including sentence lengths. Shorter sentences are easier to read. One-sentence-per-line is also semantically correct in the sense of Git.

At LSST, we place a single blank line between all content blocks, such as directives, paragraphs and lists.

Indentation

ReStructuredText should be indented consistently with the context, which generally means taking visual alignment cues rather than adhering to a fixed indent width.

In directives, align to the directive’s name:

.. code-block:: python

   print('hello world')

In lists, align naturally with the text:

- First item.

  Another paragraph for the first item.
- Second item.

Note how that alignment adapts to numbered lists:

1. First item.

   Another paragraph for the first item.
2. Second item.

For argument lists in Python docstrings we indent descriptions by four spaces:

Parameters
----------
x_coord : float
    Particle's x-coordinate.
y_coord : float
    Particle's y-coordinate.

Encoding and special characters

LSST’s reStructuredText source files should be encoded as UTF-8 unicode. This means that special characters such as en (–) and em (—) dashes can just be written as such. We do run a variant of smartypants in an attempt to convert -- and --- into en and em dashes, respectively, and to covert dumb quotes (") into “typographic” quotes.