Tags Overview

Tags represent meta-data as well as behavioural data that can be added to documentation through the @tag style syntax. As mentioned, there are two basic types of tags in YARD, "meta-data tags" and "behavioural tags", the latter is more often known as "directives". These two tag types can be visually identified by their prefix. Meta-data tags have a @ prefix, while directives have a prefix of @! to indicate that the directive performs some potentially mutable action on or with the docstring. The two tag types would be used in the following way, respectively:

# @meta_data_tag some data
# @!directive_tag some data
class Foo; end

This document describes how tags can be specified, how they affect your documentation, and how to use specific built-in tags in YARD, as well as how to define custom tags.

Meta-Data Tags

Meta-data tags are useful to add arbitrary meta-data about a documented object. These tags simply add data to objects that can be looked up later, either programmatically, or displayed in templates. The benefit to describing objects using meta-data tags is that your documentation can be organized semantically. Rather than having a huge listing of text with no distinction of what each paragraph is discussing, tags allow you to focus in on specific elements of your documentation.

For example, describing parameters of a method can often be important to your documentation, but should not be mixed up with the documentation that describes what the method itself does. In this case, separating the parameter documentation into tag:param tags can yield much better organized documentation, both in source and in your output, without having to manually format the data using standard markup.

All of this meta-data can be easily parsed by tools and used both in your templates as well as in code checker tools. An example of how you can leverage tags programmatically is shown in the tag:todo tag, which lists a small snippet of Ruby code that can list all of your TODO items, if they are properly tagged.

Custom meta-data tags can be added either programmatically or via the YARD command-line. This is discussed in the "Adding Custom Tags" section.

A list of built-in meta-data tags are found below in the Tag List.


Directives are similar to meta-data tags in the way they are specified, but they do not add meta-data to the object directly. Instead, they affect the parsing context and objects themselves, allowing a developer to create objects (like methods) outright, rather than simply add text to an existing object. Directives have a @! prefix to differentiate these tags from meta-data tags, as well as to indicate that the tag may modify or create new objects when it is called.

A list of built-in directives are found below in the Directive List.

Tag Syntax

Tags begin with the @ or @! prefix at the start of a comment line, followed immediately by the tag name, and then optional tag data (if the tag requires it). Unless otherwise specified by documentation for the tag, all "description" text is considered free-form data and can include any arbitrary textual data.

Multi-line Tags

Tags can span multiple lines if the subsequent lines are indented by more than one space. The typical convention is to indent subsequent lines by 2 spaces. In the following example, @tagname will have the text "This is indented tag data":

# @tagname This is
#   indented tag data
# but this is not

For most tags, newlines and indented data are not significant and do not impact the result of the tag. In other words, you can decide to span a tag onto multiple lines at any point by creating an indented block. However, some tags like tag:example, tag:overload, tag:!macro, tag:!method, and tag:!attribute rely on the first line for special information about the tag, and you cannot split this first line up. For instance, the tag:example tag uses the first line to indicate the example's title.

Common Tag Syntaxes

Although custom tags can be parsed in any way, the built-in tags follow a few common syntax structures by convention in order to simplify the syntax. The following syntaxes are available:

  1. Freeform data — In this case, any amount of textual data is allowed, including no data. In some cases, no data is necessary for the tag.
  2. Freeform data with a types specifier list — Mostly freeform data beginning with an optional types specifier list surrounded in [brackets]. Note that for extensibility, other bracket types are allowed, such as <>, () and {}. The contents of the list are discussed in detail below.
  3. Freeform data with a name and types specifier list — freeform data beginning with an optional types list, as well as a name key, placed either before or after the types list. The name key is required. Note that for extensibility, the name can be placed before the types list, like: name [Types] description. In this case, a separating space is not required between the name and types, and you can still use any of the other brackets that the type specifier list allows.
  4. Freeform data with title — freeform data where the first line cannot be split into multiple lines. The first line must also always refer to the "title" portion, and therefore, if there is no title, the first line must be blank. The "title" might occasionally be listed by another name in tag documentation, however, you can identify this syntax by the existence of a multi-line signature with "Indented block" on the second line.

In the tag list below, the term "description" implies freeform data, [Types] implies a types specifier list, "name" implies a name key, and "title" implies the first line is a newline significant field that cannot be split into multiple lines.

Types Specifier List

In some cases, a tag will allow for a "types specifier list"; this will be evident from the use of the [Types] syntax in the tag signature. A types specifier list is a comma separated list of types, most often classes or modules, but occasionally literals. For example, the following tag:return tag lists a set of types returned by a method:

# Finds an object or list of objects in the db using a query
# @return [String, Array<String>, nil] the object or objects to
#   find in the database. Can be nil.
def find(query) finder_code_here end

A list of conventions for type names is specified below. Typically, however, any Ruby literal or class/module is allowed here. Duck-types (method names prefixed with "#") are also allowed.

Note that the type specifier list is always an optional field and can be omitted when present in a tag signature. This is the reason why it is surrounded by brackets. It is also a freeform list, and can contain any list of values, though a set of conventions for how to list types is described below.

Type List Conventions

A list of examples of common type listings and what they translate into is available at http://yardoc.org/types.

Typically, a type list contains a list of classes or modules that are associated with the tag. In some cases, however, certain special values are allowed or required to be listed. This section discusses the syntax for specifying Ruby types inside of type specifier lists, as well as the other non-Ruby types that are accepted by convention in these lists.

It's important to realize that the conventions listed here may not always adequately describe every type signature, and is not meant to be a complete syntax. This is why the types specifier list is freeform and can contain any set of values. The conventions defined here are only conventions, and if they do not work for your type specifications, you can define your own appropriate conventions.

Note that a types specifier list might also be used for non-Type values. In this case, the tag documentation will describe what values are allowed within the type specifier list.

Class or Module Types

Any Ruby type is allowed as a class or module type. Such a type is simply the name of the class or module.

Note that one extra type that is accepted by convention is the Boolean type, which represents both the TrueClass and FalseClass types. This type does not exist in Ruby, however.

Parametrized Types

In addition to basic types (like String or Array), YARD conventions allow for a "generics" like syntax to specify container objects or other parametrized types. The syntax is Type<SubType, OtherSubType, ...>. For instance, an Array might contain only String objects, in which case the type specification would be Array<String>. Multiple parametrized types can be listed, separated by commas.

Note that parametrized types are typically not order-dependent, in other words, a list of parametrized types can occur in any order inside of a type. An array specified as Array<String, Fixnum> can contain any amount of Strings or Fixnums, in any order. When the order matters, use "order-dependent lists", described below.


Duck-types are allowed in type specifier lists, and are identified by method names beginning with the "#" prefix. Typically, duck-types are recommended for tag:param tags only, though they can be used in other tags if needed. The following example shows a method that takes a parameter of any type that responds to the "read" method:

# Reads from any I/O object.
# @param io [#read] the input object to read from
def read(io) io.read end


Hashes can be specified either via the parametrized type discussed above, in the form Hash<KeyType, ValueType>, or using the hash specific syntax: Hash{KeyTypes=>ValueTypes}. In the latter case, KeyTypes or ValueTypes can also be a list of types separated by commas.

Order-Dependent Lists

An order dependent list is a set of types surrounded by "()" and separated by commas. This list must contain exactly those types in exactly the order specified. For instance, an Array containing a String, Fixnum and Hash in that order (and having exactly those 3 elements) would be listed as: Array<(String, Fixnum, Hash)>.


Some literals are accepted by virtue of being Ruby literals, but also by YARD conventions. Here is a non-exhaustive list of certain accepted literal values:

  • true, false, nil — used when a method returns these explicit literal values. Note that if your method returns both true or false, you should use the Boolean conventional type instead.
  • self — has the same meaning as Ruby's "self" keyword in the context of parameters or return types. Recommended mostly for tag:return tags that are chainable.
  • void — indicates that the type for this tag is explicitly undefined. Mostly used to specify tag:return tags that do not care about their return value. Using a void return tag is recommended over no type, because it makes the documentation more explicit about what the user should expect. YARD will also add a note for the user if they have undefined return types, making things clear that they should not use the return value of such a method.

Reference Tags

Reference tag syntax applies only to meta-data tags, not directives.

If a tag's data begins with (see OBJECT) it is considered a "reference tag". A reference tag literally copies the tag data by the given tag name from the specified OBJECT. For instance, a method may copy all tag:param tags from a given object using the reference tag syntax:

# @param user [String] the username for the operation
# @param host [String] the host that this user is associated with
# @param time [Time] the time that this operation took place
def clean(user, host, time = Time.now) end

# @param (see #clean)
def activate(user, host, time = Time.now) end

Adding Custom Tags

If a tag is specific to a given project, consider namespacing it by naming it in the form projectname.tagname, ie., yard.tag_signature.

Custom tags can be added to YARD either via the command-line or programmatically. The programmatic method is not discussed in this document, but rather in the TagsArch document.

To add a custom tag via the command-line or .yardopts file, you can use the --*-tag options. A few different options are available for the common tag syntaxes described above. For example, to add a basic freeform tag, use:

$ yard doc --tag rest_url:"REST URL"

This will register the @rest_url tag for use in your documentation and display this tag in HTML output wherever it is used with the heading "REST URL". Note that the tag title should follow the tag name with a colon (:). Other tag syntaxes exist, such as the type specifier list freeform tag (--type-tag), or a named key tag with types (--type-name-tag).

If you want to create a tag but not display it in output (it is only for programmatic use), add --hide-tag tagname after the definition:

$ yard doc --tag complexity:"McCabe Complexity" --hide-tag complexity

Note that you might not need a tag title if you are hiding it. The title part can be omitted.