Developer Introduction¶ ↑
So you want to write a generator, fix a bug, or otherwise work with RDoc. This document provides an overview of how RDoc works from parsing options to generating output. Most of the documentation can be found in the specific classes for each feature.
If you think you found a bug, file a ticket on the issues tracker on github.
If your bug involves an error RDoc produced please include a sample file that illustrates the problem or link to the repository or gem that is associated with the bug.
Please include steps to reproduce the issue. Here are some examples of good issues:
Developer Quick Start¶ ↑
RDoc uses hoe for development. To get ready to work on RDoc run:
$ gem install hoe [...] $ rake [...]
This will install all the necessary dependencies for development with rake, generate documentation and run the tests for the first time.
If the tests don't pass on the first run check the Travis CI page for RDoc to see if there are any known failures (there shouldn't be).
You can now use `rake` and `autotest` to run the tests.
Note: the `rake` command must be used first before running any tests, because its used to generate various parsers implemented in RDoc. Also `rake clean` is helpful to delete these generated files.
Here are definitions for some common terms in the RDoc documentation. The list also briefly describes how the components of RDoc interact.
Parses files and creates a documentation tree from the contents.
- documentation tree
The documentation tree represents files, classes, modules, methods, constants, includes, comments and other ruby syntax features as a tree. RDoc walks this tree with a generator to create documentation.
Walks the documentation tree and generates output.
RDoc ships with two generators, the Darkfish generator creates HTML and the RI generator creates an RI data store.
- markup parser
Parses comments from a file into a generic markup tree.
The markup parsers allow RDoc to handle RDoc, TomDoc, rd and Markdown format documentation with common formatters.
- markup tree
Each parsed comment has a markup tree that represents common markup items such as headings, paragraphs, lists or verbatim text sections for example code or output.
A generator uses a formatters to walks the tree to create output. Some generators use multiple formatters on a markup tree to produce the output.
Converts a parsed markup tree into some form other form of markup.
Formatters can either produce a one-to-one conversion, such as ToHtml, or extract part of the parsed result, such as ToHtmlSnippet which outputs the first 100 characters as HTML.
When 'rdoc/rdoc' is loaded RDoc looks for 'rdoc/discover' files in your installed gems. This can be used to load parsers, alternate generators, or additional preprocessor directives. An rdoc plugin layout should look something like this:
lib/rdoc/discover.rb lib/my/rdoc/plugin.rb # etc.
In your rdoc/discover.rb file you will want to wrap the loading of your plugin in an RDoc version check like this:
begin gem 'rdoc', '~> 3' require 'my/rdoc/plugin' rescue Gem::LoadError end
Plugin Types¶ ↑
In RDoc you can change the following behaviors:
Add a parser for a new file format
Add a new output generator
Add a new markup directive
Add a new type of documentation markup
Add a new type of formatter
All of these are described below
Option Parsing¶ ↑
Option parsing is handled by RDoc::Options.
When you're writing a generator you can provide the user with extra
options by providing a class method
setup_options. The option
parser will call this after your generator is loaded. See RDoc::Generator for details.
File Parsing¶ ↑
After options are parsed, RDoc parses files from the files and directories in ARGV. RDoc compares the filename against what each parser claims it can parse via RDoc::Parser#parse_files_matching. For example, RDoc::Parser::C can parse C files, C headers, C++ files, C++ headers and yacc grammars.
Once a matching parser class is found it is instantiated and
scan is called. The parser needs to extract documentation from
the file and add it to the RDoc document tree.
Usually this involves starting at the root and adding a class or a module
(RDoc::TopLevel#add_class and RDoc::Context#add_module)
and proceeding to add classes, modules and methods to each nested item.
When the parsers are finished the document tree is cleaned up to remove dangling references to aliases and includes that were not found (and may exist in a separate library) through RDoc::ClassModule#complete.
To write your own parser for a new file format see RDoc::Parser.
Documentation Tree¶ ↑
The parsers build a documentation tree that is composed of RDoc::CodeObject and its subclasses. There are various methods to walk the tree to extract information, see RDoc::Context and its subclasses.
Within a class or module, attributes, methods and constants are divided into sections. The section represents a functional grouping of parts of the class. TomDoc uses the sections “Public”, “Internal” and “Deprecated”. The sections can be enumerated using RDoc::Context#each_section.
Output Generation¶ ↑
An RDoc generator turns the documentation tree into some other kind of output. RDoc comes with an HTML generator (RDoc::Generator::Darkfish) and an RI database generator (RDoc::Generator::RI). The output a generator creates does not have to be human-readable.
To create your own generator see RDoc::Generator.
In RDoc 3.10 and newer the comment on an RDoc::CodeObject is now an RDoc::Comment object instead of a String. This is to support various documentation markup formats like rdoc, TomDoc and rd. The comments are normalized to remove comment markers and remove indentation then parsed lazily via RDoc::Comment#document to create a generic markup tree that can be processed by a formatter.
To add your own markup format see Other directives at RDoc::Markup
To transform a comment into some form of output an RDoc::Markup::Formatter subclass is
used like RDoc::Markup::ToHtml. A
formatter is a visitor that walks a parsed comment tree (an RDoc::Markup::Document) of any format.
To help write a formatter RDoc::Markup::FormatterTestCase
exists for generic parsers, and RDoc::Markup::TextFormatterTestCase
which contains extra test cases for text-type output (like
RDoc ships with formatters that will turn a comment into HTML, rdoc-markup-like text, ANSI or terminal backspace highlighted text, HTML, cross-referenced HTML, an HTML snippet free of most markup, an HTML label for use in id attributes, a table-of-contents page, and text with only code blocks.
The output of the formatter does not need to be text or text-like. RDoc::Markup::ToLabel creates an HTML-safe label for use in an HTML id attribute. A formatter could count the number of words and the average word length for a comment, for example.
For comments in markup you can add new directives (:nodoc: is a directive). Directives may replace text or store it off for later use.
See RDoc::Markup::PreProcess.register for details.
RDoc contains a special generator, RDoc::Generator::JSONIndex, which creates a JSON-based search index and includes a search engine for use with HTML output. This generator can be used to add searching to any HTML output and is designed to be called from inside an HTML generator.
Additional documentation markup formats can be added to RDoc. A markup parsing class must respond to ::parse and accept a String argument containing the markup format. An RDoc::Document containing documentation items (RDoc::Markup::Heading, RDoc::Markup::Paragraph, RDoc::Markup::Verbatim, etc.) must be returned.
To register the parser with rdoc, add the markup type's name and class to the RDoc::Text::MARKUP_FORMAT hash like:
RDoc::Text::MARKUP_FORMAT['rdoc'] = RDoc::Markup