YARD: Yay! A Ruby Documentation Tool
YARD is a documentation generation tool for the Ruby programming language. It enables the user to generate consistent, usable documentation that can be exported to a number of formats very easily, and also supports extending for custom Ruby constructs such as custom class level definitions. Below is a summary of some of YARD's notable features.
1. RDoc/SimpleMarkup Formatting Compatibility: YARD is made to be compatible with RDoc formatting. In fact, YARD does no processing on RDoc documentation strings, and leaves this up to the output generation tool to decide how to render the documentation.
2. Yardoc Meta-tag Formatting Like Python, Java, Objective-C and other languages: YARD uses a '@tag' style definition syntax for meta tags alongside regular code documentation. These tags should be able to happily sit side by side RDoc formatted documentation, but provide a much more consistent and usable way to describe important information about objects, such as what parameters they take and what types they are expected to be, what type a method should return, what exceptions it can raise, if it is deprecated, etc.. It also allows information to be better (and more consistently) organized during the output generation phase. You can find a list of tags in the Tags.md file.
YARD also supports an optional "types" declarations for certain tags. This allows the developer to document type signatures for ruby methods and parameters in a non intrusive but helpful and consistent manner. Instead of describing this data in the body of the description, a developer may formally declare the parameter or return type(s) in a single line. Consider the following method documented with YARD formatting:
# Reverses the contents of a String or IO object.
# @param contents [String, #read] the contents to reverse
# @return [String] the contents reversed lexically
contents = contents.read if contents.respond_to? :read
With the above @param tag, we learn that the contents parameter can either be a String or any object that responds to the 'read' method, which is more powerful than the textual description, which says it should be an IO object. This also informs the developer that they should expect to receive a String object returned by the method, and although this may be obvious for a 'reverse' method, it becomes very useful when the method name may not be as descriptive.
3. Custom Constructs and Extensibility of YARD: YARD is designed to be extended and customized by plugins. Take for instance the scenario where you need to document the following code:
# Sets the publisher name for the list.
This custom declaration provides dynamically generated code that is hard for a
documentation tool to properly document without help from the developer. To ease
the pains of manually documenting the procedure, YARD can be extended by the
developer to handle the
cattr_accessor construct and automatically create an
attribute on the class with the associated documentation. This makes documenting
external API's, especially dynamic ones, a lot more consistent for consumption
by the users.
YARD is also designed for extensibility everywhere else, allowing you to add support for new programming languages, new data structures and even where/how data is stored.
4. Raw Data Output: YARD also outputs documented objects as raw data (the dumped Namespace) which can be reloaded to do generation at a later date, or even auditing on code. This means that any developer can use the raw data to perform output generation for any custom format, such as YAML, for instance. While YARD plans to support XHTML style documentation output as well as command line (text based) and possibly XML, this may still be useful for those who would like to reap the benefits of YARD's processing in other forms, such as throwing all the documentation into a database. Another useful way of exploiting this raw data format would be to write tools that can auto generate test cases, for example, or show possible unhandled exceptions in code.
5. Local Documentation Server: YARD can serve documentation for projects or
installed gems (similar to
gem server) with the added benefit of dynamic
searching, as well as live reloading. Using the live reload feature, you can
document your code and immediately preview the results by refreshing the page;
YARD will do all the work in re-generating the HTML. This makes writing
documentation a much faster process.
To install YARD, use the following command:
$ gem install yard
sudo if you're installing under a POSIX system as root)
Alternatively, if you've checked the source out directly, you can call
rake install from the root project directory.
Important Note for Debian/Ubuntu users: there's a possible chance your Ruby
install lacks RDoc, which is occasionally used by YARD to convert markup to
HTML. If running
which rdoc turns up empty, install RDoc by issuing:
$ sudo apt-get install rdoc
There are a couple of ways to use YARD. The first is via command-line, and the second is the Rake task.
1. yard Command-line Tool
YARD comes packaged with a executable named
yard which can control the many
functions of YARD, including generating documentation, graphs running the YARD
server, and so on. To view a list of available YARD commands, type:
$ yard --help
Plugins can also add commands to the
yard executable to provide extra
yardoc executable is a shortcut for
The most common command you will probably use is
yard doc, or
yardoc --help to see the options that YARD provides, but the easiest
way to generate docs for your code is to simply type
yardoc in your project
root. This will assume your files are located in the
lib/ directory. If they
are located elsewhere, you can specify paths and globs from the commandline via:
$ yardoc 'lib/**/*.rb' 'app/**/*.rb' ...etc...
The tool will generate a
.yardoc file which will store the cached database of
your source code and documentation. If you want to re-generate your docs with
another template you can simply use the
--use-cache (or -c) option to speed up
the generation process by skipping source parsing.
YARD will by default only document code in your public visibility. You can
document your protected and private code by adding
to the option switches. In addition, you can add
--no-private to also ignore
any object that has the
@private meta-tag. This is similar to RDoc's ":nodoc:"
behaviour, though the distinction is important. RDoc implies that the object
with :nodoc: would not be documented, whereas YARD still recommends documenting
private objects for the private API (for maintainer/developer consumption).
You can also add extra informative files (README, LICENSE) by separating the globs and the filenames with '-'.
$ yardoc 'app/**/*.rb' - README LICENSE FAQ
If no globs precede the '-' argument, the default glob (
lib/**/*.rb) is used:
$ yardoc - README LICENSE FAQ
Note that the README file can be specified with its own
You can also add a
.yardopts file to your project directory which lists the
switches separated by whitespace (newlines or space) to pass to yardoc whenever
it is run. A full overview of the
.yardopts file can be found in
yardoc tool also supports a
--query argument to only include objects
that match a certain data or meta-data query. The query syntax is Ruby, though a
few shortcuts are available. For instance, to document only objects that have an
"@api" tag with the value "public", all of the following syntaxes would give the
--query '@api.text == "public"'
--query 'object.has_tag?(:api) && object.tag(:api).text == "public"'
--query 'has_tag?(:api) && tag(:api).text == "public"'
Note that the "@tag" syntax returns the first tag named "tag" on the object. To return the array of all tags named "tag", use "@@tag".
--query arguments are allowed in the command line parameters. The
following two lines both check for the existence of a return and param tag:
--query '@return' --query '@param'
--query '@return && @param'
For more information about the query syntax, see theclass.
2. Rake Task
The second most obvious is to generate docs via a Rake task. You can do this by
adding the following to your
::::. do |t|
t.files = ['lib/**/*.rb', OTHER_PATHS] # optional
t. = ['--any', '--extra', '--opts'] # optional
t. = ['--list-undoc'] # optional
All the settings:
stats_options are optional.
will default to
options will represents any options you might
want to add and
stats_options will pass extra options to the stats command.
Again, a full list of options is available by typing
yardoc --help in a shell.
You can also override the options at the Rake command-line with the OPTS
$ rake yard OPTS='--any --extra --opts'
yri RI Implementation
The yri binary will use the cached .yardoc database to give you quick ri-style access to your documentation. It's way faster than ri but currently does not work with the stdlib or core Ruby libraries, only the active project. Example:
$ yri YARD::Handlers::Base#register
$ yri File.relative_path
Note that class methods must not be referred to with the "::" namespace separator. Only modules, classes and constants should use "::".
You can also do lookups on any installed gems. Just make sure to build the .yardoc databases for installed gems with:
$ yard gems
If you don't have sudo access, it will write these files to your
yri will also cache lookups there.
yard server Documentation Server
yard server command serves documentation for a local project or all
installed RubyGems. To serve documentation for a project you are working on,
$ yard server
And the project inside the current directory will be parsed (if the source has not yet been scanned by YARD) and served at http://localhost:8808.
If you want to serve documentation on a project while you document it so that
you can preview the results, simply pass
-r) to the above command
and YARD will reload any changed files on each request. This will allow you to
change any documentation in the source and refresh to see the new contents.
To serve documentation for all installed gems, call:
$ yard server --gems
This will also automatically build documentation for any gems that have not been previously scanned. Note that in this case there will be a slight delay between the first request of a newly parsed gem.
yard graph Graphviz Generator
You can use
yard graph to generate dot graphs of your code. This, of course,
requires Graphviz and the
dot binary. By default
this will generate a graph of the classes and modules in the best UML2 notation
that Graphviz can support, but without any methods listed. With the
option, methods and attributes will be listed. There is also a
option to show mixin inclusions. You can output to stdout or a file, or pipe
dot. The same public, protected and private visibility rules apply
yard graph. More options can be seen by typing
yard graph --help, but
here is an example:
$ yard graph --protected --full --dependencies
See CHANGELOG for a list of changes.
YARD © 2007-2020 by Loren Segal. YARD is licensed under the MIT license except for some files which come from the RDoc/Ruby distributions. Please see the LICENSE and LEGAL documents for more information.