/****************************************************************************** * * * * Copyright (C) 1997-2000 by Dimitri van Heesch. * * Permission to use, copy, modify, and distribute this software and its * documentation under the terms of the GNU General Public License is hereby * granted. No representations are made about the suitability of this software * for any purpose. It is provided "as is" without express or implied warranty. * See the GNU General Public License for more details. * * Documents produced by Doxygen are derivative works derived from the * input used in their production; they are not affected by this license. * */ /*! \page starting Getting started The executable \c doxygen is the main program that parses the sources and generates the documentation. See section \ref doxygen_usage for more detailed usage information. The executable \c doxytag is only needed if you want to generate references to external documentation (i.e. documentation that was generated by doxygen) for which you do not have the sources or to create a search index for the search engine. See section \ref doxytag_usage for more detailed usage information. The executable \c doxysearch is only needed if you want to use the search engine. See section \ref doxysearch_usage for more detailed usage information. \subsection step1 Step 1: Creating a configuration file Doxygen uses a configuration file to determine all of its settings. Each project should get its own configuration file. A project can consist of a single source file, but can also be an entire source tree that is recursively scanned. To simplify the creation of a configuration file, doxygen can create a template configuration file for you. To do this call \c doxygen with the \c -g option: \verbatim doxygen -g \endverbatim where \ is the name of the configuration file. If you omit the file name, a file named \c Doxyfile will be created. If a file with the name \ already exists, doxygen will rename it to \.bak before generating the configuration template. If you use - (i.e. the minus sign) as the file name then doxygen will try to read the configuration file from standard input (stdin). The configuration file has a format that is similar to that of a (simple) Makefile. It contains of a number of assignments (tags) of the form: TAGNAME = VALUE or
TAGNAME = VALUE1 VALUE2 ...
You can probably leave the values of most tags in a generated template configuration file to their default value. The \c INPUT tag is the only tag for which you are required to provide a value. See section \ref config for more details about the configuration file. For a small project consisting of a few C and/or C++ source and header files, you can add the names of the files after the \c INPUT tag. If you have a larger project consisting of a source directory or tree this may become tiresome. In this case you should put the root directory or directories after the \c INPUT tag, and add one or more file patterns to the \c FILE_PATTERN tag (for instance *.cpp *.h). Only files that match one of the patterns will be parsed (if the patterns are omitted all files will be parsed). For recursive parsing of a source tree you must set the \c RECURSIVE tag to \c YES. To further fine-tune the list of files that is parsed the \c EXCLUDE and \c EXCLUDE_PATTERNS tags can be used. If you start using doxygen for an existing project (thus without any documentation that doxygen is aware of), you can still get an idea of what the documented result would be. To do so, you must set the \c EXTRACT_ALL tag in the configuration file to \c YES. Then, doxygen will pretend everything in your sources is documented. Please note that warnings of undocumented members will not be generated as long as \c EXTRACT_ALL is set to \c YES. To analyse an existing piece of software it is useful to cross-reference a (documented) entity with its definition in the source files. Doxygen will generate such cross-references if you set the \c SOURCE_BROWSER tag to \c YES. It can also include the sources directly into the documentation by setting \c INLINE_SOURCES to \c YES (this can be handly for code reviews for instance). \subsection step2 Step 2: Running doxygen To generate the documentation you can now enter: \verbatim doxygen \endverbatim Doxygen will create a \c html, \c latex and/or \c man directory inside the output directory. As the names suggest the \c html directory contains the generated documentation in HTML format and the \c latex directory contains the generated documentation in \f$\mbox{\LaTeX}\f$ format. Man pages are put in a \c man3 directory inside the \c man directory. The default output directory is the directory in which \c doxygen is started. The directory to which the output is written can be changed using the \c OUTPUT_DIRECTORY , \c HTML_OUTPUT, \c LATEX_OUTPUT, and \c MAN_OUTPUT tags of the configuration file. If the output directory does not exist, \c doxygen will try to create it for you. \addindex browser The generated HTML documentation can be viewed by pointing a HTML browser to the \c index.html file in the \c html directory. For the best results a browser that supports cascading style sheets (CSS) should be used (I'm currently using Netscape 4.61 to test the generated output). \addindex LaTeX The generated \f$\mbox{\LaTeX}\f$ documentation must first be compiled by a \f$\mbox{\LaTeX}\f$ compiler. (I use teTeX distribution version 0.9 that contains \f$\mbox{\TeX}\f$ version 3.14159). To simplify the process of compiling the generated documentation, \c doxygen writes a \c Makefile into the \c latex directory. By typing \c make in the \c latex directory the dvi file \c refman.dvi will be generated (provided that you have a make tool called make ofcourse). This file can then be viewed using \c xdvi or converted into a postscript file \c refman.ps by typing make ps (this requires dvips). The Postscript file can be send to a postscript printer. If you do not have a postscript printer, you can try to use ghostscript to convert postscript into something your printer understands. Conversion to PDF is also possible; just type make pdf (this required ps2pdf). To get the best results for PDF output you should set the \c PDF_HYPERLINKS tag to \c YES. The generated man pages can be viewed using the \c man program. You do need to make sure the man directory is in the man path (see the \c MANPATH environment variable). Note that there are some limitations to the capabilities of the man page format, so some information (like class diagrams, cross references and formulas) will be lost. \subsection step3 Step 3: Documenting the sources Although documenting the source is presented as step 3, in a new project this should ofcourse be step 1. Here I assume you already have some code and you want doxygen to generate a nice document describing the API and maybe the internals as well. If the \c EXTRACT_ALL option is set to \c NO in the configuration file (the default), then doxygen will only generate documentation for \e documented members, files, classes and namespaces. So how do you document these? For members, classes and namespaces there are basicly two options:
  1. Place a \e special documentation block in front of the declaration or definition of the member, class or namespace. For file, class and namespace members it is also allowed to place the documention directly after the member. See section \ref specialblock to learn more about special documentation blocks.
  2. Place a special documentation block somewhere else (another file or another location) \e and put a structural command in the documentation block. A structural command links a documentation block to a certain entity that can be documented (e.g. a member, class, namespace or file). See section \ref structuralcommands to learn more about structural commands.
Files can only be documented using the second option. The text inside a special documentation block is parsed before it is written to the HTML and/or \f$\mbox{\LaTeX}\f$ output files. \addindex parsing During parsing the following steps take place:
  • The special commands inside the documentation are executed. See section \ref commands for an overview of all commands.
  • If a line starts with some whitespace followed by one or more asterixes (*) then the whitespace and asterixes are removed.
  • All resulting blank lines are treated as a paragraph separators. This saves you from placing new-paragraph commands yourself, in order to make the generated documentation readable.
  • Links are created for words corresponding to documented classes.
  • Links to members are created when certain patterns are found in the text. See section \ref autolink for more information on how the automatic link generation works.
  • HTML tags that are in the documentation are interpreted and converted to \f$\mbox{\LaTeX}\f$ equivalents for the \f$\mbox{\LaTeX}\f$ output. See section \ref htmlcmds for an overview of all supported HTML tags.
Using a number of column aligned minus signs at the start of a line in a comment block will generate a bullet list. Nested lists are also possible. Here is an example: \verbatim /*! * A list of events: * - mouse events * - mouse move event * - mouse click event * - mouse double click event\n * More info about the click event. * - keyboard events * - key down event * - key up event * * More text here. */ \endverbatim The result will be: A list of events: - mouse events - mouse move event - mouse click event\n More info about the click event. - mouse double click event - keyboard events - key down event - key up event More text here. If you use tabs within lists, please make sure that \c TAB_SIZE in the configuration file is set to the correct tab size. \htmlonly Go to the next section or return to the index. \endhtmlonly */ /*! \page docblocks Documenting the code \subsection specialblock Special documentation blocks The following types of special documentation blocks are supported by doxygen:
  • The Qt style, where special documentation blocks look like: \verbatim /*! ... text ... */ \endverbatim and the one line version: \verbatim //! ... one line of text ... \endverbatim
  • The JavaDoc style, where special documentation blocks look like: \verbatim /** * ... text ... */ \endverbatim and the one line version: \verbatim /// ... one line of text ... \endverbatim
Doxygen only allows one brief and one detailed description. If there is one brief description before a declaration and one before a definition, only the one before the \e declaration will be used. If the same situation occurs for a detail description, the one before the \e definition is preferred and the one before the declaration will be ignored. Here is an example of a documented piece of C++ code using the Qt style: \verbinclude qtstyle.cpp \htmlonly Click here for the corresponding HTML documentation that is generated by doxygen. \endhtmlonly The one-line comments should contain a brief description, whereas the multi-line comment blocks contain a more detailed description. The brief descriptions are included in the member overview of a class, namespace or file and are printed using a small italic font (this description can be omitted by setting \c BRIEF_MEMBER_DESC to \c NO in the config file). By default the brief descriptions are also the first sentence of the detailed description (this can be changed by setting the \c REPEAT_BRIEF tag to \c NO). Both the brief and the detailed descriptions are optional for the Qt style. Here is the same piece of code, this time documented using the JavaDoc style: \verbinclude jdstyle.cpp \htmlonly Click here for the corresponding HTML documentation that is generated by doxygen. \endhtmlonly Note that the first sentence of the documentation (until the .) is treated as a brief description, whereas the documentation block as a whole forms the detailed description. The brief description is required for the JavaDoc style. Unlike most other documentation systems, doxygen also allows you to put the documentation of members (including global functions) in front of the \e definition. This way the documentation can be placed in the source file instead of the header file. This keeps the header file compact, and allows the implementer of the members more direct access to the documentation. As a compromise the brief description could be placed before the declaration and the detailed description before the member definition (assuming you use the Qt style comments). \subsection structuralcommands Structural commands So far we have assumed that the documentation blocks are always located in front of the declaration or definition of a file, class or namespace or in front of one of its members. Although this is often comfortable, it may sometimes be better to put the documentation somewhere else. For some types of documentation blocks (like file documentation) this is even required. Doxygen allows you to put your documentation blocks practically anywhere (the exception is inside the body of a function or inside a normal C style comment block), as long as you put a structural command inside the documentation block. Structural commands (like all other commands) start with a backslash (\\) followed by a command name and one or more parameters. For instance, if you want to document the class \c Test in the example above, you could have also put the following documentation block somewhere in the input that is read by doxygen: \verbatim /*! \class Test \brief A test class. A more detailed class description. */ \endverbatim Here the special command \c \class is used to indicated that the comment block contains documentation for the class \c Test. Other structural commands are:
  • \c \struct to document a C-struct.
  • \c \union to document a union.
  • \c \enum to document an enumeration type.
  • \c \fn to document a function.
  • \c \var to document a variable or typedef or enum value.
  • \c \def to document a \#define.
  • \c \file to document a file.
  • \c \namespace to document a namespace.
See section \ref commands for detailed information about these and other commands. Note that the documentation block belonging to a file should always contain a structural command. To document a member of a C++ class, you must also document the class itself. The same holds for namespaces. To document a C function, typedef, enum or preprocessor definition you must first document the file that contains it (usually this will be a header file, because that file contains the information that is exported to other source files). Here is an example of a C header named \c structcmd.h that is documented using structural commands: \verbinclude structcmd.h \htmlonly Click here for the corresponding HTML documentation that is generated by doxygen. \endhtmlonly \par Note: Because each comment block in the example above contains a structural command, all the comment blocks could be moved to another location or input file (the source file for instance), without affecting the generated documentation. The disadvantage of this approach is that prototypes are duplicated, so all changes have to be made twice! \subsection memberdoc Documenting compound members. If you want to document the members of a file, struct, union, class, or enum and you want to put the documentation for these members inside the compound, it is sometimes desired to place the documentation block after the member instead of before. For this purpose doxygen has the following additional comment blocks: \verbatim /*!< ... */ \endverbatim This block can be used to put a qt style documentation blocks after a member. The one line version look as follows: \verbatim //!< ... \endverbatim There are also JavaDoc versions: \verbatim /**< ... */ \endverbatim and \verbatim ///< ... \endverbatim Note that these blocks have the same structure and meaning as the special comment blocks above only the \< indicates that the member is located in front of the block instead of after the block. Here is an example of a the use of these comment blocks: \verbinclude afterdoc.h \htmlonly Click here for the corresponding HTML documentation that is generated by doxygen. \endhtmlonly \warning These blocks can only be used to document \e members. They cannot be used to document file classes, unions, structs and enums. Furthermore, the structural commands mentioned in the previous section are ignored inside these comment blocks. \htmlonly Go to the next section or return to the index. \endhtmlonly */ /*! \page formulas Including formulas Doxygen allows you to put \f$\mbox{\LaTeX}\f$ formulas in the output (this works only for the HTML and \f$\mbox{\LaTeX}\f$ formats, not for the man page output). To be able to include formulas (as images) in the HTML documentation, you will also need to have the following tools installed
  • \c latex: the \f$\mbox{\LaTeX}\f$ compiler, needed to parse the formulas. To test I have used the teTeX 0.9 distribution.
  • \c dvips: a tool to convert dvi files to postscript files I have used version 5.86 from Radical Eye software for testing.
  • \c gs: the ghostscript interpreter for converting postscript files to bitmaps. I have used Aladdin Ghostscript 5.10 for testing.
There are two ways to include formulas in the documentation.
  1. Using in-text formulas that appear in the running text. These formulas should be put between a pair of \\f\$ commands, so \verbatim The distance between \f$(x_1,y_1)\f$ and \f$(x_2,y_2)\f$ is \f$\sqrt{(x_2-x_1)^2+(y_2-y_1)^2}\f$. \endverbatim results in: The distance between \f$(x_1,y_1)\f$ and \f$(x_2,y_2)\f$ is \f$\sqrt{(x_2-x_1)^2+(y_2-y_1)^2}\f$.
  2. Unnumbered displayed formulas that are centered on a separate line. These formulas should be put between \\f\[ and \\f\] commands. An example: \verbatim \f[ |I_2|=\left| \int_{0}^T \psi(t) \left\{ u(a,t)- \int_{\gamma(t)}^a \frac{d\theta}{k(\theta,t)} \int_{a}^\theta c(\xi)u_t(\xi,t)\,d\xi \right\} dt \right| \f] \endverbatim results in: \f[ |I_2|=\left| \int_{0}^T \psi(t) \left\{ u(a,t)- \int_{\gamma(t)}^a \frac{d\theta}{k(\theta,t)} \int_{a}^\theta c(\xi)u_t(\xi,t)\,d\xi \right\} dt \right| \f]
Formulas should be valid commands in \f$\mbox{\LaTeX}\f$'s math-mode. \warning Currently, doxygen is not very fault tolerant in recovering from typos in formulas. It may have to be necessary to remove the file formula.repository that is written in the html directory to a rid of an incorrect formula \htmlonly Go to the next section or return to the index. \endhtmlonly */ /*! \page diagrams Graphs and diagrams Doxygen has build-in support to generate inheritance diagrams for C++ classes. Doxygen can use the "dot" tool from graphviz 1.5 to generate more advanced diagrams & graphs. Graphviz is an open-sourced, cross-platform graph drawing toolkit from AT&T and Lucent Bell Labs and can be found at http://www.research.att.com/sw/tools/graphviz/ If you have the "dot" tool available in the path, you can set \ref cfg_have_dot "HAVE_DOT" to \c YES in the configuration file to let doxygen use it. Doxygen uses the "dot" tool to generate the following graphs:
  • if \ref cfg_graphical_hierarchy "GRAPHICAL_HIERARCHY" is set to \c YES, a graphical representation of the class hierarchy will be drawn, along with the textual one. Currently this feature is supported for HTML only.\n Warning: When you have a very large class hierarchy where many classes derive from a common base class, the resulting image may become too big to handle for some browsers.
  • if \ref cfg_include_graph "INCLUDE_GRAPH" is set to \c YES, an include dependency graph is generated for each documented file that includes at least one other file. This feature is currently supported for HTML and RTF only.
  • if \ref cfg_collaboration_graph "COLLABORATION_GRAPH" is set to YES, a graph is drawn for each documented class and struct that shows:
    • the inheritance relations with base classes.
    • the usage relations with other structs & classes (e.g. class \c A has a member variable \c m_a of type class \c B, then \c A has an arrow to \c B with \c m_a as label).
The elements in the class diagrams in HTML and RTF have the following meaning:
  • A yellow box indicates a class. A box can have a little marker in the lower right corner to indicate that the class contains base classes that are hidden. If the box is filled with a dashed pattern the inheritance relation is virtual.
  • A white box indicates that the documentation of the class is currently shown.
  • A grey box indicates an undocumented class.
  • A solid dark blue arrows indicates public inheritance.
  • A dashed dark green arrows indicates protected inheritance.
  • A dotted dark green arrows indicates private inheritance.
The elements in the class diagram in \f$\mbox{\LaTeX}\f$ have the following meaning:
  • A white box indicates a class. A marker in the lower right corner of the box indicates that the class has base classes that are hidden. If the box has a dashed border this indicates virtual inheritance.
  • A solid arrow indicates public inheritance.
  • A dashed arrow indicates protected inheritance.
  • A dotted arrow indicated private inheritance.
The elements in the graphs generated by the dot tool have the following meaning:
  • A white box indicates a class or struct. If the box has a red border this indicates that the class contains related classes that are hidden.
  • A black box indicates that the class' documentation is currently shown.
  • A dark blue arrow indicates an include relation (for the include dependency graph) or public inheritance (for the other graphs).
  • A dark green arrow indicates protected inheritance.
  • A dark red arrow indicates private inheritance.
  • A purple dashed arrow indicated a "usage" relation, the edge of the arrow is labled with the variable(s) responsible for the relation. Class \c A uses class \c B, if class \c A has a member variable \c m of type C, where B is a subtype of C (e.g. C could be \c B, \c B*, T\* ).
The reason why classes or structs are sometimes hidden is too prevent images to become too large. For the class diagrams the maximum tree width is currently 8 elements. For the graphs generated with dot doxygen tries to limit the width of the resulting image to 1024 pixels. Here are a couple of header files that together show the various diagrams that doxygen can generate: diagrams_a.h \verbinclude diagrams_a.h diagrams_b.h \verbinclude diagrams_b.h diagrams_c.h \verbinclude diagrams_c.h diagrams_d.h \verbinclude diagrams_d.h diagrams_e.h \verbinclude diagrams_e.h \htmlonly Click here for the corresponding HTML documentation that is generated by doxygen
(EXTRACT_ALL = YES is used here). \endhtmlonly \htmlonly Go to the next section or return to the index. \endhtmlonly */ /*! \page preprocessing Preprocessing Source files that are used as input to doxygen can be parsed by doxygen's build-in C-preprocessor. By default doxygen does only partial preprocessing. That is, it evaluates conditional compilation statements (like \#if) and evaluates macro definitions, but it does not perform macro expansion. So if you have the following code fragment \verbatim #define VERSION 200 #define CONST_STRING const char * #if VERSION >= 200 static CONST_STRING version = "2.xx"; #else static CONST_STRING version = "1.xx"; #endif \endverbatim Then by default doxygen will feed the following to its parser: \verbatim #define VERSION #define CONST_STRING static CONST_STRING version = "2.xx"; \endverbatim You can disable all preprocessing by setting \c ENABLE_PREPROCESSING to \c NO in the configuation file. In the case above doxygen will then read both statements! In case you want to expand the \c CONST_STRING macro, you should set the \c MACRO_EXPANSION tag in the config file to \c YES. Then the result after preprocessing becomes: \verbatim #define VERSION #define CONST_STRING static const char * version = "1.xx"; \endverbatim Note that doxygen will now expand \e all macro definitions (recursively if needed). This is often too much. Therefore, doxygen also allows you to expand only those defines that you explicitly specify. For this you have to set the \c EXPAND_ONLY_PREDEF tag to \c YES and specify the macro definitions after the \c PREDEFINED tag. As an example, suppose you have the following obfusciated code fragment of an abstract base class called \c IUnknown: \verbatim /*! A reference to an IID */ #ifdef __cplusplus #define REFIID const IID & #else #define REFIID const IID * #endif /*! The IUnknown interface */ DECLARE_INTERFACE(IUnknown) { STDMETHOD(HRESULT,QueryInterface) (THIS_ REFIID iid, void **ppv) PURE; STDMETHOD(ULONG,AddRef) (THIS) PURE; STDMETHOD(ULONG,Release) (THIS) PURE; }; \endverbatim without macro expansion doxygen will get confused, but we may not want to expand the REFIID macro, because it is documented and the user that reads the documentation should use it when implementing the interface. By setting the following in the config file: \verbatim ENABLE_PREPROCESSING = YES MACRO_EXPANSION = YES EXPAND_ONLY_PREDEF = YES PREDEFINED = "DECLARE_INTERFACE(name)=class name" \ "STDMETHOD(result,name)=virtual result name" \ "PURE= = 0" \ THIS_= \ THIS= \ __cplusplus \endverbatim we can make sure that the proper result is fed to doxygen's parser: \verbatim /*! A reference to an IID */ #define REFIID /*! The IUnknown interface */ class IUnknown { virtual HRESULT QueryInterface ( REFIID iid, void **ppv) = 0; virtual ULONG AddRef () = 0; virtual ULONG Release () = 0; }; \endverbatim Note that the \c PREDEFINED tag accepts function like macro definitions (like \c DECLARE_INTERFACE), normal macro substitutions (like \c PURE and \c THIS) and plain defines (like \c __cplusplus). Note also that preprocessor definitions that are normally defined automatically by the preprocessor (like \c __cplusplus), have to be defined by hand with doxygen's parser (this is done because these defines are often platform/compiler specific). In some cases you may want to substitute a macro name or function by something else without exposing the result to further macro substitution. You can do this but using the := operator instead of = As an example suppose we have the following piece of code: \verbatim #define QList QListT class QListT { }; \endverbatim Then the only way to get doxygen interpret this as a class definition for class QList is to define: \verbatim PREDEFINED = QListT:=QList \endverbatim As you can see doxygen's preprocessor is quite powerful, but if you want even more flexibility you can always write an input filter and specify it after the \c INPUT_FILTER tag. If you are unsure what the effect of doxygen's preprocessing will be you can run doxygen as follows: \verbatim doxygen -d Preprocessor \endverbatim This will instruct doxygen to dump the input sources to standard output after preprocessing has been done (Hint: set QUIET = YES and WARNINGS = NO in the configuration file to disable any other output). \htmlonly Go to the next section or return to the index. \endhtmlonly */