Developing PETSc Documentation#
General Guidelines#
Good documentation should be like a bonsai tree: alive, on display, frequently tended, and as small as possible (adapted from these best practices).
Wrong, irrelevant, or confusing documentation is worse than no documentation.
Documentation with Sphinx#
We use Sphinx to build our web pages and documentation. Most content is written using reStructuredText, a simple markup language.
These slides contain an overview of Sphinx and how we use(d) it, as of October 2020.
Building the HTML docs locally#
We use a Python 3 virtual environment to build the documentation since not all developers can trivially install the needed Python modules directly.
$ cd $PETSC_DIR
$ make docs
$ open doc/_build/html/index.html # in a browser
or
$ cd $PETSC_DIR/doc
$ make sphinxhtml
$ open _build/html/index.html
Building the manual locally as a PDF via LaTeX#
Note
Before following these instructions, you need to have a working local LaTeX installation and the ability to install additional packages, if need be, to resolve LaTeX errors.
Set up your local Python environment (e.g., ref:as above <sec_local_html_docs>), then
$ cd doc
$ make sphinxpdf
$ open _build/latex/manual.pdf # in PDF viewer
Sphinx Documentation Guidelines#
Refer to Sphinx’s own documentation for general information on how to use Sphinx, and note the following additional guidelines.
Use the literalinclude directive to directly include pieces of source code. Use a path beginning with
/
, relative to the root for the Sphinx docs (whereconf.py
is found)... literalinclude:: /../src/sys/error/err.c :start-at: PetscErrorCode PetscError( :end-at: PetscFunctionReturn(PETSC_SUCCESS) :append: }
For robustness to changes in the source files, Use
:start-at:
and related options when possible, noting that you can also use (positive) values of:lines:
relative to this. Use the:language:
option to appropriately highlight languages other than C.Any invocable command line statements longer than a few words should be in
.. code-block::
sections. Double backticks must enclose any such statements not in code-block statements”``”. For examplemake all
is acceptable but$ make PETSC_DIR=/my/path/to/petsc PETSC_ARCH=my-petsc-arch all
should be in a
.. code-block::
.All code blocks showing command line invocation must use the “console” block directive. E.g.
.. code-block:: console $ cd $PETSC_DIR/src/snes/interface $ ./someprog output1 output2
The only exception to this is when displaying raw output, i.e., with no preceding commands. Then one may use just the “::” directive to improve visibility, e.g.,
:: output1 output2
Any code blocks that show command line invocations must be preceded by
$
, e.g... code-block:: console $ ./configure --some-args $ make libs $ make ./ex1 $ ./ex1 --some-args
Environment variables such as
$PETSC_DIR
or$PATH
must be preceded by$
and be enclosed in double backticks, e.g.Set ``$PETSC_DIR`` and ``$PETSC_ARCH``
For internal links, use explicit labels, e.g
.. _sec_short_name: Section name ============
and elsewhere (in any document),
See :ref:`link text <sec_short_name>`
For internal links in the manual with targets outside the manual, always provide alt text so that the text will be properly formatted in the standalone PDF manual, e.g.
PETSc has :doc:`mailing lists </community/mailing>`.
We use the sphinxcontrib-bibtex extension to include citations from BibTeX files. You must include
.. bibliography::
blocks at the bottom of a page, including citations (example). To cite the same reference on more than one page, use this workaround on one of them (example) [1].See special instructions on Images.
Prefer formatting styles that are easy to modify and maintain. In particular, the use of list-table is recommended.
When using external links with inline URLs, prefer to use anonymous hyperlink references with two trailing underscores, e.g.
`link text <https://external.org>`__
To pluralize something with inline markup, e.g.
DM
s, escape the trailing character to avoidWARNING: Inline literal start-string without end-string
.``DM``\s
Use restraint in adding new Sphinx extensions, in particular, those which aren’t widely used and well-supported, or those with hidden system dependencies.
Other PETSc repositories#
In addition to the PETSc repository, there are three other PETSc repositories which contain large data files that are unnecessary for most PETSc usages and thus are not stored in the main repository. Images contains images that are used in the PETSc documentation or have other uses. Annual-Meetings contains the slides etc. from the PETSc Annual User Meetings (PAUM). Datafiles contains large matrices, meshes, and various other data files that are used in the PETSc CI. Other repositories containing software PETSc uses are located at GitLab and BitBucket. The BitBucket location is used for historical reasons, there are many links on the web to these locations thus the repositories have not be migrated to GitLab.
Images#
PETSc’s documentation is tightly coupled to the source code and tests and is tracked in the primary PETSc Git repository. However, image files are too large to track directly this way (especially because they persist in the integration branches’ histories).
Therefore, we store image files in a separate Git repository, Images. This repository is automatically cloned if
not already available when building the documentation. It can also be cloned by running
make images
in the doc/
directory.
Any new images required must be added to the currently-used branch of this repository.
Image Guidelines#
Whenever possible, use SVG files. SVG is a web-friendly vector format and will be automatically converted to PDF using
rsvg-convert
[2]Avoid large files and large numbers of images.
Do not add movies or other non-image files.
Adding new images#
Decide where in
doc/images
a new image should go. Use the structure of thedoc/
tree as a guide.Create a Merge Request to the currently-used branch of the upstream images repository, adding this image [3].
Once this Merge Request is merged, you may make a MR relying on the new image(s).
It may be helpful to place working copies of the new image(s) in your local doc/images
while iterating on documentation; don’t forget to update the upstream images repository.
Removing, renaming, moving, or updating images#
Do not directly move, rename, or update images in the images repository. Simply add a logically-numbered new version of the image.
If an image is not used in any integration branch (main
or release
),
add it to the top-level list of files to delete in the images repository.
Cleaning up the images repository (maintainers only)#
If the size of the image repository grows too large,
Create a new branch
main-X
, whereX
increments the current valueCreate a new commit deleting all files in the to-delete list and clearing the list
Reset the new
main-X
to a single commit with this new, cleaned-up stateSet
main-X
as the “default” branch on GitLab.Update both
release
andmain
in the primary PETSc repository to clone this new branch
Building Manual Pages and C2HTML Files#
The manual pages and C2HTML-generated file as built in a process described below using the documentation tools listed below, which are automatically downloaded and installed if needed while building the PETSc documentation./
Sowing: Developed by Bill Gropp, this produces the PETSc manual pages; see the Sowing documentation and Manual Page Format.
C2html: This generates the HTML versions of all the source code.
Sowing and C2html are build tools that do not use the compilers specified to PETSc’s configure
, as they
need to work in cross-compilation environments. Thus, they default to using gcc
, g++
, and flex
from
the user’s environment (or configure
options like --download-sowing-cxx
). Microsoft Windows users must install gcc
etc., from Cygwin in order to be able to build the documentation.
Footnotes