Final Draft -- 2005-01-18
Copyright © 2001-2008 Oren Ben-Kiki, Clark Evans, Ingy döt Net
Status of this Document
This specification is a draft reflecting consensus reached by members of the yaml-core mailing list. Any questions regarding this draft should be raised on this list. We expect all further changes to be strictly limited to wording corrections and fixing production bugs.
We wish to thank implementers, who have tirelessly tracked earlier versions of this specification, as well as our fabulous user community whose feedback has both validated and clarified our direction.
Abstract
YAML™ (rhymes with “camel”) is a human-friendly, cross language, Unicode based data serialization language designed around the common native data structures of agile programming languages. It is broadly useful for programming needs ranging from configuration files to Internet messaging to object persistence to data auditing. Together with the Unicode standard for characters, this specification provides all the information necessary to understand YAML Version 1.1 and to create programs that process YAML information.
Table of Contents
“YAML Ain’t Markup Language” (abbreviated YAML) is a data serialization language designed to be human-friendly and work well with modern programming languages for common everyday tasks. This specification is both an introduction to the YAML language and the concepts supporting it; it is also a complete reference of the information needed to develop applications for processing YAML.
Open, interoperable and readily understandable tools have advanced computing immensely. YAML was designed from the start to be useful and friendly to people working with data. It uses Unicode printable characters, some of which provide structural information and the rest containing the data itself. YAML achieves a unique cleanness by minimizing the amount of structural characters and allowing the data to show itself in a natural and meaningful way. For example, indentation may be used for structure, colons separate “mapping key: value” pairs, and dashes are used to create “bullet” lists.
There are myriad flavors of data structures, but they can all be adequately represented with three basic primitives: mappings (hashes/dictionaries), sequences (arrays/lists) and scalars (strings/numbers). YAML leverages these primitives and adds a simple typing system and aliasing mechanism to form a complete language for serializing any data structure. While most programming languages can use YAML for data serialization, YAML excels in working with those languages that are fundamentally built around the three basic primitives. These include the new wave of agile languages such as Perl, Python, PHP, Ruby, and Javascript.
There are hundreds of different languages for programming, but only a handful of languages for storing and transferring data. Even though its potential is virtually boundless, YAML was specifically created to work well for common use cases such as: configuration files, log files, interprocess messaging, cross-language data sharing, object persistence, and debugging of complex data structures. When data is easy to view and understand, programming becomes a simpler task.
The design goals for YAML are:
- YAML is easily readable by humans.
- YAML matches the native data structures of agile languages.
- YAML data is portable between programming languages.
- YAML has a consistent model to support generic tools.
- YAML supports one-pass processing.
- YAML is expressive and extensible.
- YAML is easy to implement and use.
YAML’s initial direction was set by the data serialization and markup language discussions among SML-DEV members. Later on, it directly incorporated experience from Brian Ingerson’s Perl module Data::Denter. Since then, YAML has matured through ideas and support from its user community.
YAML integrates and builds upon concepts described by C, Java, Perl, Python, Ruby, RFC0822 (MAIL), RFC1866 (HTML), RFC2045 (MIME), RFC2396 (URI), XML, SAX and SOAP.
The syntax of YAML was motivated by Internet Mail (RFC0822) and remains partially compatible with that standard. Further, borrowing from MIME (RFC2045), YAML’s top-level production is a stream of independent documents; ideal for message-based distributed processing systems.
YAML’s indentation-based scoping is similar to Python’s (without the ambiguities caused by tabs). Indented blocks facilitate easy inspection of the data’s structure. YAML’s literal style leverages this by enabling formatted text to be cleanly mixed within an indented structure without troublesome escaping. YAML also allows the use of traditional indicator-based scoping similar to Perl’s. Such flow content can be freely nested inside indented blocks.
YAML’s double-quoted style uses familiar C-style escape sequences.
This enables ASCII encoding of non-printable or 8-bit (ISO 8859-1) characters such as
“\x3B”. Non-printable 16-bit Unicode and
32-bit (ISO/IEC 10646) characters are supported with escape sequences
such as “\u003B”
and “\U0000003B”.
Motivated by HTML’s end-of-line normalization, YAML’s line folding employs an intuitive method of handling line breaks. A single line break is folded into a single space, while empty lines are interpreted as line break characters. This technique allows for paragraphs to be word-wrapped without affecting the canonical form of the content.
YAML’s core type system is based on the requirements of agile languages such as Perl, Python, and Ruby. YAML directly supports both collection (mapping, sequence) and scalar content. Support for common types enables programmers to use their language’s native data structures for YAML manipulation, instead of requiring a special document object model (DOM).
Like XML’s SOAP, YAML supports serializing native graph data structures through an aliasing mechanism. Also like SOAP, YAML provides for application-defined types. This allows YAML to represent rich data structures required for modern distributed computing. YAML provides globally unique type names using a namespace mechanism inspired by Java’s DNS-based package naming convention and XML’s URI-based namespaces.
YAML was designed to support incremental interfaces that include both
input (“getNextEvent()”) and output
“sendNextEvent()”) one-pass interfaces. Together, these
enable YAML to support the processing of large documents
(e.g. transaction logs) or continuous streams (e.g. feeds from a
production machine).
Newcomers to YAML often search for its correlation to the eXtensible Markup Language (XML). Although the two languages may actually compete in several application domains, there is no direct correlation between them.
YAML is primarily a data serialization language. XML was designed to be backwards compatible with the Standard Generalized Markup Language (SGML) and thus had many design constraints placed on it that YAML does not share. Inheriting SGML’s legacy, XML is designed to support structured documentation, where YAML is more closely targeted at data structures and messaging. Where XML is a pioneer in many domains, YAML is the result of lessons learned from XML and other technologies.
It should be mentioned that there are ongoing efforts to define standard XML/YAML mappings. This generally requires that a subset of each language be used. For more information on using both XML and YAML, please visit http://yaml.org/xml/index.html.
This specification uses key words based on RFC2119 to indicate requirement level. In particular, the following words are used to describe the actions of a YAML processor:
- May
- The word may, or the adjective optional, mean that conforming YAML processors are permitted, but need not behave as described.
- Should
- The word should, or the adjective recommended, mean that there could be reasons for a YAML processor to deviate from the behavior described, but that such deviation could hurt interoperability and should therefore be advertised with appropriate notice.
- Must
- The word must, or the term required or shall, mean that the behavior described is an absolute requirement of the specification.
This section provides a quick glimpse into the expressive power of YAML. It is not expected that the first-time reader grok all of the examples. Rather, these selections are used as motivation for the remainder of the specification.
YAML’s block collections use
indentation for scope
and begin each entry on its own line. Block
sequences indicate each entry with a dash and space (
“-”). Mappings use a colon and
space (“: ”) to mark each mapping
key: value pair.
YAML also has flow styles, using explicit indicators rather than indentation to denote scope. The flow sequence is written as a comma separated list within square brackets. In a similar manner, the flow mapping uses curly braces.
YAML uses three dashes (“---”) to separate documents
within a stream. Three dots ( “...”) indicate the end of
a document without starting a new one, for use in communication
channels. Comment lines begin with the Octothorpe (also called
“hash”, “sharp”, or “number
sign” - “#”).
Repeated nodes are first identified by an anchor
(marked with the ampersand - “&”), and are then aliased
(referenced with an asterisk - “*”) thereafter.
A question mark and space (“? ”) indicate a complex mapping key.
Within a block collection, key: value pairs can start
immediately following the dash, colon, or question mark.
Scalar
content can be written in block form, using a literal
style (“|”) where all line breaks are significant.
Alternatively, they can be written with the folded
style (“>”) where each line break is folded to a space unless it ends an
empty or a “more indented”
line.
YAML’s flow scalars include the plain style (most examples thus far) and quoted styles. The double-quoted style provides escape sequences. The single-quoted style is useful when escaping is not needed. All flow scalars can span multiple lines; line breaks are always folded.
In YAML, untagged nodes
are given an type depending on the application. The examples in this
specification generally use the “seq”,
“map” and
“str”
types from the YAML tag
repository. A few examples also use the “int” and
“float”
types. The repository includes additional types such as “null”,
“bool”,
“set” and
others.
Explicit typing is denoted with a tag using the exclamation
point (“!”) symbol. Global tags are URIs and may be
specified in a shorthand
form using a handle. Application-specific local tags may also be used.
Below are two full-length examples of YAML. On the left is a sample invoice; on the right is a sample log file.
YAML is both a text format and a method for presenting any data structure in this format. Therefore, this specification defines two concepts: a class of data objects called YAML representations, and a syntax for presenting YAML representations as a series of characters, called a YAML stream. A YAML processor is a tool for converting information between these complementary views. It is assumed that a YAML processor does its work on behalf of another module, called an application. This chapter describes the information structures a YAML processor must provide to or obtain from the application.
YAML information is used in two ways: for machine processing, and for human consumption. The challenge of reconciling these two perspectives is best done in three distinct translation stages: representation, serialization, and presentation. Representation addresses how YAML views native data structures to achieve portability between programming environments. Serialization concerns itself with turning a YAML representation into a serial form, that is, a form with sequential access constraints. Presentation deals with the formatting of a YAML serialization as a series of characters in a human-friendly manner.
A YAML processor need not expose the serialization or representation stages. It may translate directly between native data structures and a character stream (dump and load in the diagram above). However, such a direct translation should take place so that the native data structures are constructed only from information available in the representation.
This section details the processes shown in the diagram above. Note a YAML processor need not provide all these processes. For example, a YAML library may provide only YAML input ability, for loading configuration files, or only output ability, for sending data to other applications.
YAML represents any native data structure using three node kinds: sequence - an ordered series of entries; mapping - an unordered association of unique keys to values; and scalar - any datum with opaque structure presentable as a series of Unicode characters. Combined, these primitives generate directed graph structures. These primitives were chosen because they are both powerful and familiar: the sequence corresponds to a Perl array and a Python list, the mapping corresponds to a Perl hash table and a Python dictionary. The scalar represents strings, integers, dates, and other atomic data types.
Each YAML node requires, in addition to its kind and content, a tag specifying
its data type. Type specifiers are either global URIs, or are local in scope to a single application. For example, an integer
is represented in YAML with a scalar plus the global tag
“tag:yaml.org,2002:int”. Similarly, an invoice object,
particular to a given organization, could be represented as a
mapping together with the local tag “!invoice”. This simple model
can represent any data structure independent of programming language.
For sequential access mediums, such as an event callback API, a YAML representation must be serialized to an ordered tree. Since in a YAML representation, mapping keys are unordered and nodes may be referenced more than once (have more than one incoming “arrow”), the serialization process is required to impose an ordering on the mapping keys and to replace the second and subsequent references to a given node with place holders called aliases. YAML does not specify how these serialization details are chosen. It is up to the YAML processor to come up with human-friendly key order and anchor names, possibly with the help of the application. The result of this process, a YAML serialization tree, can then be traversed to produce a series of event calls for one-pass processing of YAML data.
The final output process is presenting the YAML serializations as a character stream in a human-friendly manner. To maximize human readability, YAML offers a rich set of stylistic options which go far beyond the minimal functional needs of simple data storage. Therefore the YAML processor is required to introduce various presentation details when creating the stream, such as the choice of node styles, how to format content, the amount of indentation, which tag handles to use, the node tags to leave unspecified, the set of directives to provide and possibly even what comments to add. While some of this can be done with the help of the application, in general this process should be guided by the preferences of the user.
Parsing is the inverse process of presentation, it takes a stream of characters and produces a series of events. Parsing discards all the details introduced in the presentation process, reporting only the serialization events. Parsing can fail due to ill-formed input.
Composing takes a series of serialization events and produces a representation graph. Composing discards all the serialization details introduced in the serialization process, producing only the representation graph. Composing can fail due to any of several reasons, detailed below.
The final input process is constructing native data structures from the YAML representation. Construction must be based only on the information available in the representation, and not on additional serialization or presentation details such as comments, directives, mapping key order, node styles, content format, indentation levels etc. Construction can fail due to the unavailability of the required native data types.
This section specifies the formal details of the results of the above processes. To maximize data portability between programming languages and implementations, users of YAML should be mindful of the distinction between serialization or presentation properties and those which are part of the YAML representation. Thus, while imposing a order on mapping keys is necessary for flattening YAML representations to a sequential access medium, this serialization detail must not be used to convey application level information. In a similar manner, while indentation technique and a choice of a node style are needed for the human readability, these presentation details are neither part of the YAML serialization nor the YAML representation. By carefully separating properties needed for serialization and presentation, YAML representations of application information will be consistent and portable between various programming environments.
The following diagram summarizes the three information models. Full
arrows denote composition, hollow arrows denote inheritance,
“1” and “*” denote “one” and
“many” relationships. A single “+” denotes
serialization details, a
double “++” denotes presentation details.
YAML’s representation of native data is a rooted, connected, directed graph of tagged nodes. By “directed graph” we mean a set of nodes and directed edges (“arrows”), where each edge connects one node to another (see a formal definition). All the nodes must be reachable from the root node via such edges. Note that the YAML graph may include cycles, and a node may have more than one incoming edge.
Nodes that are defined in terms of other nodes are collections and nodes that are independent of any other nodes are scalars. YAML supports two kinds of collection nodes, sequences and mappings. Mapping nodes are somewhat tricky because their keys are unordered and must be unique.
YAML nodes have content of one of three kinds: scalar, sequence, or mapping. In addition, each node has a tag which serves to restrict the set of possible values which the node’s content can have.
- Scalar
- The content of a scalar node is an opaque datum that can be presented as a series of zero or more Unicode characters.
- Sequence
- The content of a sequence node is an ordered series of zero or more nodes. In particular, a sequence may contain the same node more than once or it could even contain itself (directly or indirectly).
- Mapping
- The content of a mapping node is an unordered set of key: value node pairs, with the restriction that each of the keys is unique. YAML places no further restrictions on the nodes. In particular, keys may be arbitrary nodes, the same node may be used as the value of several key: value pairs, and a mapping could even contain itself as a key or a value (directly or indirectly).
When appropriate, it is convenient to consider sequences and mappings together, as collections. In this view, sequences are treated as mappings with integer keys starting at zero. Having a unified collections view for sequences and mappings is helpful both for creating practical YAML tools and APIs and for theoretical analysis.
YAML represents type
information of native data structures with a simple identifier,
called a tag. Global
tags are URIs and hence
globally unique across all applications. The
“tag”: URI
scheme (mirror) is
recommended for all global YAML tags. In contrast, local tags are specific to a single
application. Local tags
start with “!”, are not URIs and are not
expected to be globally unique. YAML provides a “TAG” directive to
make tag notation less verbose; it also offers easy migration from
local to global tags. To ensure this, local tags are restricted to
the URI character set and use URI character escaping.
YAML does not mandate any special relationship between different
tags that begin with the same substring. Tags ending with URI
fragments (containing “#”) are no exception; tags that
share the same base URI but differ in their fragment part are
considered to be different, independent tags. By convention,
fragments are used to identify different “variants” of
a tag, while “/” is used to define nested tag
“namespace” hierarchies. However, this is merely a
convention, and each tag may employ its own rules. For example,
Perl tags may use “::” to express namespace
hierarchies, Java tags may use “.”, etc.
YAML tags are used to associate meta information with each node. In particular, each tag must specify the expected node kind (scalar, sequence, or mapping). Scalar tags must also provide mechanism for converting formatted content to a canonical form for supporting equality testing. Furthermore, a tag may provide additional information such as the set of allowed content values for validation, a mechanism for tag resolution, or any other data that is applicable to all of the tag’s nodes.
Since YAML mappings require key uniqueness, representations must include a
mechanism for testing the equality of nodes. This is non-trivial
since YAML allows various ways to format a given scalar content. For
example, the integer eleven can be written as “013”
(octal) or “0xB” (hexadecimal). If both forms are
used as keys in the same mapping, only a YAML
processor which recognizes
integer formats would correctly
flag the duplicate key as an error.
- Canonical Form
- YAML supports the need for scalar equality by requiring that every scalartag must specify a mechanism to producing the canonical form of any formatted content. This form is a Unicode character string which presents the content and can be used for equality testing. While this requirement is stronger than a well defined equality operator, it has other uses, such as the production of digital signatures.
- Equality
- Two nodes must have the same tag and content to be equal. Since each tag applies to exactly one kind, this implies that the two nodes must have the same kind to be equal. Two scalars are equal only when their tags and canonical forms are equal character-by-character. Equality of collections is defined recursively. Two sequences are equal only when they have the same tag and length, and each node in one sequence is equal to the corresponding node in the other sequence. Two mappings are equal only when they have the same tag and an equal set of keys, and each key in this set is associated with equal values in both mappings.
- Identity
- Two nodes are identical only when they represent the same native data structure. Typically, this corresponds to a single memory address. Identity should not be confused with equality; two equal nodes need not have the same identity. A YAML processor may treat equal scalars as if they were identical. In contrast, the separate identity of two distinct but equal collections must be preserved.
To express a YAML representation using a serial API, it necessary to impose an order on mapping keys and employ alias nodes to indicate a subsequent occurrence of a previously encountered node. The result of this process is a serialization tree, where each node has an ordered set of children. This tree can be traversed for a serial event-based API. Construction of native structures from the serial interface should not use key order or anchors for the preservation of important data.
In the representation model, mapping keys do not have an order. To serialize a mapping, it is necessary to impose an ordering on its keys. This order is a serialization detail and should not be used when composing the representation graph (and hence for the preservation of important data). In every case where node order is significant, a sequence must be used. For example, an ordered mapping can be represented as a sequence of mappings, where each mapping is a single key: value pair. YAML provides convenient compact notation for this case.
In the representation graph, a node may appear in more than one collection. When serializing such data, the first occurrence of the node is identified by an anchor and each subsequent occurrence is serialized as an alias node which refers back to this anchor. Otherwise, anchor names are a serialization detail and are discarded once composing is completed. When composing a representation graph from serialized events, an alias node refers to the most recent node in the serialization having the specified anchor. Therefore, anchors need not be unique within a serialization. In addition, an anchor need not have an alias node referring to it. It is therefore possible to provide an anchor for all nodes in serialization.
A YAML presentation is a stream of Unicode characters making use of of styles, formats, comments, directives and other presentation details to present a YAML serialization in a human readable way. Although a YAML processor may provide these details when parsing, they should not be reflected in the resulting serialization. YAML allows several serializations to be contained in the same YAML character stream as a series of documents separated by document boundary markers. Documents appearing in the same stream are independent; that is, a node must not appear in more than one serialization tree or representation graph.
Each node is presented in some style, depending on its kind. The node style is a presentation detail and is not reflected in the serialization tree or representation graph. There are two groups of styles, block and flow. Block styles use indentation to denote nesting and scope within the document. In contrast, flow styles rely on explicit indicators to denote nesting and scope.
YAML provides a rich set of scalar styles. Block scalar styles include the literal style and the folded style; flow scalar styles include the plain style and two quoted styles, the single-quoted style and the double-quoted style. These styles offer a range of trade-offs between expressive power and readability.
Normally, the content of block collections begins on the next line. In most cases, YAML also allows block collections to start in-line for more compact notation when nesting block sequences and block mappings inside each other. When nesting flow collections, a flow mapping with a single key: value pair may be specified directly inside a flow sequence, allowing for a compact “ordered mapping” notation.
YAML allows scalar content to be presented in several formats. For example, the boolean
“true” might also be written as
“yes”. Tags must specify a mechanism for converting any
formatted scalar content to a canonical form for use in equality testing. Like node style, the format is a presentation detail and is
not reflected in the serialization
tree and representation
graph.
Comments are a presentation detail and must not have any effect on the serialization tree or representation graph. In particular, comments are not associated with a particular node. The usual purpose of a comment is to communicate between the human maintainers of a file. A typical example is comments in a configuration file. Comments may not appear inside scalars, but may be interleaved with such scalars inside collections.
Each document may be associated with a set of directives. A directive has a name and an optional
sequence of parameters. Directives are instructions to the YAML
processor, and like all
other presentation
details are not reflected in the YAML serialization tree or representation graph. This
version of YAML defines a two directives, “YAML” and “TAG”. All other
directives are reserved for future versions of YAML.
The process of loading native data structures from a YAML stream has several potential failure points. The character stream may be ill-formed, aliases may be unidentified, unspecified tags may be unresolvable, tags may be unrecognized, the content may be invalid, and a native type may be unavailable. Each of these failures results with an incomplete loading.
A partial representation need not resolve the tag of each node, and the canonical form of scalar content need not be available. This weaker representation is useful for cases of incomplete knowledge of the types used in the document. In contrast, a complete representation specifies the tag of each node, and provides the canonical form of scalar content, allowing for equality testing. A complete representation is required in order to construct native data structures.
A well-formed character stream must match the productions specified in the next chapter. Successful loading also requires that each alias shall refer to a previous node identified by the anchor. A YAML processor should reject ill-formed streams and unidentified aliases. A YAML processor may recover from syntax errors, possibly by ignoring certain parts of the input, but it must provide a mechanism for reporting such errors.