Doxygen

Doxygen is a popular tool for generating documentation from annotated C++ sources; it also supports other popular programming languages such as C#, PHP, Java, and Python. Visit the Doxygen website to learn more about the system, and consult the Doxygen Manual for the full information.

This document gives a brief introduction to Doxygen, in particular how it is used in FreeCAD to document its sources. Visit the source documentation page for instructions on building the FreeCAD documentation, which is also hosted online on the FreeCAD API website.



Doxygen with C++ code
The Getting started (Step 3) section of the Doxygen manual mentions the basic ways of documenting the sources.

For members, classes and namespaces there are basically two options:
 * 1) Place a special "documentation block" (a commented paragraph) before the declaration or definition of the function, member, class or namespace. For file, class and namespace members (variables) it is also allowed to place the documentation directly after the member. See section Special comment blocks in the manual to learn more about these blocks.
 * 2) Place a special documentation block somewhere else (another file or another location in the same file) and put a "structural command" in the documentation block. A structural command links a documentation block to a certain entity that can be documented (a function, member, variable, class, namespace or file). See section Documentation at other places in the manual to learn more about structural commands.

Note:
 * The advantage of the first option is that you do not have to repeat the name of the entity (function, member, variable, class, or namespace), as Doxygen will analyze the code and extract the relevant information.
 * Files can only be documented using the second option, since there is no way to put a documentation block before a file. Of course, file members (functions, variables, typedefs, defines) do not need an explicit structural command; just putting a documentation block before or after them will work fine.

First style: documentation block before the code
Usually you'd want to document the code in the header file, just before the class declaration or function prototype. This keeps the declaration and documentation close to each other, so it's easy to update the latter one if the first one changes.

The special documentation block starts like a C-style comment but has an additional asterisk, so ; the block ends with a matching. An alternative is using C++-style comments with an additional slash, so.

Second style: documentation block elsewhere
Alternatively, the documentation can be placed in another file (or in the same file at the top or bottom, or wherever), away from the class declaration or function prototype. In this case, you will have duplicated information, once in the actual source file, and once in the documentation file.

First file, :

Second file, :

In this case the structural command is used to indicate which source file is being documented; a structural command  indicates that the following code is a function, and the command  is used to give a small description of this function.

This way of documenting a source file is useful if you just want to add documentation to your project without adding real code. When you place a comment block in a file with one of the following extensions, , or then Doxygen will parse the comments and build the appropriate documentation, but it will hide this auxiliary file from the file list.

The FreeCAD project adds several files ending in in many directories in order to provide a description, or examples, of the code there. It is important that such files are correctly categorized in a group or namespace, for which Doxygen provides some auxiliary commands like, , and.

Example ; this file in FreeCAD's source tree gives a short explanation of the namespace.

Another example is the file. Before the implementation details of the methods, there is a documentation block that explains some details of the command framework of FreeCAD. It has various commands to structure the documentation. It even includes example code enclosed in a pair of and  keywords; when the file is processed by Doxygen this code example will be specially formatted to stand out. The keyword is used in several places to create links to named sections, subsections, pages or anchors elsewhere in the documentation. Similarly, the or  commands print "See also" and provide a link to other classes, functions, methods, variables, files or URLs. Example

Compiling the documentation


To generate the source code documentation there are two basic steps:
 * 1) Create a configuration file to control how Doxygen will process the source files.
 * 2) Run  on that configuration.

The process is described in the following.


 * Make sure you have the programs and  in your system. It is also recommended to have the  program from Graphviz, in order to generate diagrams with the relationships between classes and namespaces. On Linux systems these programs can be installed from your package manager.


 * Make sure you are in the same folder as your source files, or in the toplevel directory of your source tree, if you have many source files in different sub-directories.


 * Run to create a configuration file named . If you omit this name, it will default to  without an extension.
 * This is a big, plain text file that includes many variables with their values. In the Doxygen manual these variables are called "tags". The configuration file and all tags are described extensively in the Configuration section of the manual. You can open the file with any text editor, and edit the value of each tag as necessary. In the same file you can read comments explaining each of the tags, and their default values.


 * Instead of using a text editor, you may launch to edit many tags at the same time. With this interface you may define many properties such as project information, type of output (HTML and LaTeX), use of Graphviz to create diagrams, warning messages to display, file patterns (extensions) to document or to exclude, input filters, optional headers and footers for the HTML generated pages, options for LaTeX, RTF, XML, or Docbook outputs, and many other options.


 * Another option is to create the configuration file from scratch, and add only the tags that you want with a text editor.
 * After the configuration is saved, you can run Doxygen on this configuration file.


 * The generated documentation will be created inside a folder named . It will consist of many HTML pages, PNG images for graphics, cascading style sheets, Javascript files , and potentially many sub-directories with more files depending on the size of your code. The point of entry into the documentation is , which you can open with a web browser.

If you are writing new classes, functions or an entire new workbench, it is recommended that you run periodically to see that the documentation blocks, Markdown, and special commands are being read correctly, and that all public functions are fully documented. Please also read the documentation tips located in the source code.

When generating the complete FreeCAD documentation, you don't run directly. Instead, the project uses to configure the build environment, and then  triggers compilation of the FreeCAD sources and of the Doxygen documentation; this is explained in the source documentation page.

Doxygen markup
All Doxygen documentation commands start with a backslash or an at-symbol, at your preference. Normally the backslash is used, but occasionally the  is used to improve readability.

The commands can have one or more arguments. In the Doxygen manual the arguments are described as follows.
 * If braces are used the argument is a single word.
 * If braces are used the argument extends until the end of the line on which the command was found.
 * If  braces are used the argument extends until the next paragraph. Paragraphs are delimited by a blank line or by a section indicator.
 * If brackets are used the argument is optional.

Some of the most common keywords used in the FreeCAD documentation are presented here.
 * , see \defgroup, and Grouping.
 * , see \ingroup, and Grouping.
 * , see \addtogroup, and Grouping.
 * , see \author; indicates the author of this piece of code.
 * , see \brief; briefly describes the function.
 * , see \file; documents a source or header file.
 * , see \page; puts the information in a separate page, not directly related to one specific class, file or member.
 * , see \package; indicates documentation for a Java package (but also used with Python).
 * , see \fn; documents a function.
 * , see \var; documents a variable; it is equivalent to, , and.
 * , see \section; starts a section.
 * , see \subsection; starts a subsection.
 * , see \namespace; indicates information for a namespace.
 * , and, see \cond; defines a block to conditionally document or omit.
 * , see \a; displays the argument in italics for emphasis.
 * , see \param; indicates the parameter of a function.
 * , see \return; specifies the return value.
 * , see \sa; prints "See also".
 * , see \note; adds a paragraph to be used as a note.

Markdown support
Since Doxygen 1.8, Markdown syntax is recognized in documentation blocks. Markdown is a minimalistic formatting language inspired by plain text email which, similar to wiki syntax, intends to be simple and readable without requiring complicated code like that found in HTML, LaTeX or Doxygen's own commands. Markdown has gained popularity with free software, especially in online platforms like Github, as it allows creating documentation without using complicated code. See the Markdown support section in the Doxygen manual to learn more. Visit the Markdown website to learn more about the origin and philosophy of Markdown.

Doxygen supports a standard set of Markdown instructions, as well as some extensions such as Github Markdown. Some examples of Markdown formatting are presented next.

The following is standard Markdown. {{Code|code= Here is text for one paragraph.

We continue with more text in another paragraph.

This is a level 1 header

=
===========

This is a level 2 header


 * 1) This is a level 1 header


 * 1) This is level 3 header #######

> This is a block quote > spanning multiple lines

- Item 1

More text for this item.

- Item 2 * nested list item. * another nested item. - Item 3

1. First item. 2. Second item.


 * single asterisks: emphasis*

_single underscores_

**double asterisks: strong emphasis**

__double underscores__

This a normal paragraph

This is a code block

We continue with a normal paragraph again.

Use the `printf` function. Inline `code`.

[The link text](http://example.net/)

![Caption text](/path/to/img.jpg)

 }}

The following are Markdown extensions. Second Header -

Parsing of documentation blocks
The text inside a special documentation block is parsed before it is written to the HTML and LaTeX output files. During parsing the following steps take place:
 * Markdown formatting is replaced by corresponding HTML or special commands.
 * The special commands inside the documentation are executed. See the section Special Commands in the manual for an explanation of each command.
 * If a line starts with some whitespace followed by one or more asterisks and then optionally more whitespace, then all whitespace and asterisks are removed.
 * All resulting blank lines are treated as paragraph separators.
 * Links are automatically created for words corresponding to documented classes or functions. If the word is preceded by a percentage symbol, then this symbol is removed, and no link is created for the word.
 * Links are created when certain patterns are found in the text. See the section Automatic link generation in the manual for more information.
 * HTML tags that are in the documentation are interpreted and converted to LaTeX equivalents for the LaTeX output. See the section HTML Commands in the manual for an explanation of each supported HTML tag.

Doxygen with Python code
Doxygen works best for statically typed languages like C++. However, it can also create documentation for Python files.

There are two ways to write documentation blocks for Python:
 * 1) The Pythonic way, using "docstrings", that is, a block of text surrounded by   immediately after the class or function definition.
 * 2) The Doxygen way, putting comments before the class or function definition; in this case double pound characters  are used to start the documentation block, and then single pound characters can be used in subsequent lines.

Note:
 * The first option is preferred to comply with PEP8, PEP257 and most style guidelines for writing Python (see 1, 2). It is recommended to use this style if you intend to produce documented sources using Sphinx, which is a very common tool to document Python code. If you use this style, Doxygen will be able to extract the comments verbatim, but Doxygen special commands starting with or  won't work.
 * The second option isn't the traditional Python style, but it allows you to use Doxygen's special commands like and.

First style: Pythonic documentation
In the following example one docstring is at the beginning to explain the general contents of this module (file). Then docstrings appear inside the function, class, and class method definitions. In this way, Doxygen will extract the comments and present them as is, without modification.

Second style: documentation block before the code
In the following example the documentation blocks start with a double pound sign. One appears at the beginning to explain the general content of this module (file). Then there are blocks before the function, class, and class method definitions, and there is one block after a class variable. In this way, Doxygen will extract the documentation, recognize the special commands, , and , and format the text accordingly.

Compiling the documentation
Compilation of documentation proceeds the same as for C++ sources. If both Python files, and, with distinct commenting style are in the same directory, both will be processed.

The documentation should show similar information to the following, and create appropriate links to the individual modules and classes.

Converting the Pythonic style to Doxygen style
In the previous example, the Python file that is commented in a Doxygen style shows more detailed information and formatting for its classes, functions, and variables. The reason is that this style allows Doxygen to extract the special commands that start with or, while the Pythonic style does not. Therefore, it would be desirable to convert the Pythonic style to Doxygen style before compiling the documentation. This is possible with an auxiliary Python program called doxypypy. This program is inspired by an older program called doxypy, which would take the Pythonic  and convert them to the Doxygen comment blocks that start with double pound symbols. Doxypypy goes further than this, as it analyzes the docstrings and extracts items of interest like variables and arguments, and even doctests (example code in the docstrings).

Doxypypy can be installed using, the Python package installer.

If the command is used without the  option, it will require superuser (root) privileges to install the package, but this is not needed in most cases; use root permissions only if you are certain the package won't collide with your distribution provided packages.

If the package was installed as a user, it may reside in your home directory, for example, in. If this directory is not in your system's, the program will not be found. Therefore, add the directory to the variable, either in your  file, or in your  file.

Alternatively, you can create a symbolic link to the program, placing the link in a directory that is already included in the.

Once the program is installed, and accessible from the terminal, a Python file with Pythonic docstrings can be reformatted to Doxygen style with the following instructions. The program outputs the reformatted code to standard output, so redirect this output to a new file.

The original file has a comment at the top  that indicates the module or namespace that is being described by the file. This keyword is not interpreted when using triple quotes in the comment block.

In the new file the commenting style is changed so the line becomes, which now will be interpreted by Doxygen. However, to be interpreted correctly, the argument has to be edited manually to match the new module (file) name; after doing this the line should be.

To compile, create the configuration, and run in the toplevel directory that contains the files.

The documentation should show similar information to the following, and create appropriate links to the individual modules.

Converting the comment style on the fly
In the previous example, the conversion of the documentation blocks was done manually with only one source file. Ideally we want this conversion to occur automatically, on the fly, with any number of Python files. To do this, the Doxygen configuration must be edited accordingly.

To start, don't use the program directly; instead, create the configuration file with, then edit the created , and modify the following tag.

What this does is that files that match the pattern, all files with a extension ending in, will go through the program. Every time Doxygen encounters such file in the source tree, the file name will be passed as the first argument to this program.

The program does not exist by default; it should be created as a shell script to run  with the appropriate options, and to take a file as its first argument.

After saving this shell script, make sure it has execute permissions, and that it is located in a directory contained in your system's.

On Windows systems, a batch file can be used in a similar way.

With this configuration done, the command can be run to generate the documentation as usual. Every Python file using Pythonic  will be reformatted on the fly to use  style comments, and then will be processed by Doxygen, which now will be able to interpret the special commands and Mardown syntax. The original source code won't be modified, and no temporary file with a different name needs to be created as in the previous section; therefore, if a instruction is found, it doesn't need to be changed manually.

Note that existing Python files which already use the style for their comment blocks won't be affected by the  filter, and will be processed by Doxygen normally.



Python code quality check
To use the automatic conversion of documentation blocks it is important that the original Python sources are correctly written, following the Pythonic guidelines in PEP8 and PEP257. Sloppily written code will cause to fail when processing the file, and thus Doxygen will be unable to format the documentation correctly.

The following commenting styles will not allow parsing of the docstrings by, so they should be avoided.

Always use triple quotes for the docstrings, and make sure they immediately follow the class or function declaration.

It is also a good idea to verify the quality of your Python code with a tool such as flake8 (Gitlab). Flake8 primarily combines three tools, Pyflakes, Pycodestyle (formerly pep8), and the McCabe complexity checker, in order to enforce proper Pythonic style.

To check all files inside a source tree use.

If the project demands it, some code checks deemed too strict can be ignored. The error codes can be consulted in the Pycodestyle documentation.

In similar way, a program that primarily checks that comments comply with PEP257 is Pydocstyle. The error codes can be consulted in the Pydocstyle documentation.

Also use it with to perform docstring checks on all source files.

Source documentation with Sphinx
Sphinx is the most popular system to document Python source code. However, since FreeCAD's core functions and workbenches are written in C++ it was deemed that Doxygen is a better documentation tool for this project.

While Sphinx can natively parse Python docstrings, it requires a bit more work to parse C++ sources. The Breathe (Github) project is an attempt at bridging the gap between Sphinx and Doxygen, in order to integrate both Python and C++ source code documentation in the same system. First, Doxygen needs to be configured to output an XML file; the XML output is then read by Breathe, and converted to suitable input for Sphinx.

See the Quick start guide in the Breathe documentation to know more about this process.

See this answer in Stackoverflow for other alternatives to documenting C++ and Python code together in the same project.