Red Hat Training

A Red Hat training course is available for Red Hat Enterprise Linux

Chapter 5. YAML in a Nutshell

5.1. Overview

YAML — which stands for “YAML Ain’t Markup Language” — is a human-friendly data serialization standard, similar in scope to JSON (Javascript Object Notation). Unlike JSON, there are only a handful of special characters used to represent mappings and bullet lists, the two basic types of structure, and indentation is used to represent substructure.

5.2. Basics

The YAML format is line-oriented, with two top-level parts, HEAD and BODY, separated by a line of three hyphens.


The head holds configuration information and the body holds the data. this topic does not discuss the configuration aspect; all the examples here show only the data portion. In such cases, the “---” is optional.

The most basic data element is one of:

  1. A number
  2. A Unicode string
  3. A boolean value, spelled either true or false
  4. In a key/value pair context, a missing value is parsed as nil

Comments start with a “#” (hash, U+23) and go to the end of the line.

Indentation is whitespace at the start of the line. You are strongly encouraged to avoid TAB (U+09) characters and use a series of SPACE (U+20) characters, instead.

5.3. Lists

A list is a series of lines, each beginning with the same amount of indentation, followed by a hyphen, followed by a list element. Lists cannot have blank lines. For example, here is a list of three elements, the third of which has a comment:

- top shelf
- middle age
- bottom dweller   # stability is important

Note: The third element is the string “bottom dweller” and does not include the whitespace between “dweller” and the comment.

WARNING: Lists cannot normally nest directly; there should be an intervening mapping (described below). In the following example, the list’s second element seems, due to the indentation (two SPACE characters), to host a sub-list:

- top
- middle
  - highish middle
  - lowish middle
- bottom

In reality, the second element is actually parsed as a single string. The input is equivalent to:

- top
- middle - highish middle - lowish middle
- bottom

The newlines and indentation are normalized to a single space.

5.4. Mappings

To write a mapping (also known as an associative array or hash table), use a “:” (colon, U+3A) followed by one or more SPACE characters between the key and the value:

square:   4
triangle: 3
pentagon: 5

All keys in a mapping must be unique. For example, this is invalid YAML for two reasons: the key square is repeated, and there is no space after the colon following triangle:

square:  4
triangle:3       # invalid key/value separation
square:  5       # repeated key

Mappings can nest directly, by starting the sub-mapping on the next line with increased indentation. In the next example, the value for key square is itself a mapping (keys sides and perimeter), and likewise for the value for key triangle. The value for key pentagon is the number 5.

  sides:     4
  perimeter: sides * side-length
  sides:     3
  perimeter: see square
pentagon:    5

The following example shows a mapping with three key/value pairs. The first and third values are nil, while the second is a list of two elements, “highish middle” and “lowish middle”.

  - highish middle
  - lowish middle

5.5. Quotation

Double-quotation marks (also known as “double-quotes”) are useful for forcing non-string data to be interpreted as a string, for preserving whitespace, and for suppressing the meaning of colon. To include a double-quote in a string, escape it with `“\” (backslash, U+5C). In the following example, all keys and values are strings. The second key has a colon in it. The second value has two spaces both preceding and trailing the visible text.

"true" : "1"
"key the second (which has a \":\" in it)" : "  second  value  "

For readability when double-quoting the key, you are encouraged to add whitespace before the colon.

5.6. Block Content

There are two kinds of block content, typically found in the value position of a mapping element: newline-preserving and folded. If a block begins with “|” (pipe, U+7C), the newlines in that block are preserved. If it begins with “>” (greater-than, U+3E), consecutive newlines are folded into a single space. The following example shows both kinds of block content as the values for keys good-bye and anyway.

hello: world

good-bye: |
    first line

    fourth and last

anyway: >
    nothing is guaranteed
    in life

Using \n (backslash-n) to indicate newline, the values for keys good-bye and anyway are, respectively:

first line\n\nthird\nfourth and last\n

nothing is guaranteed in life\n

Note that the newlines are preserved in the good-bye value but folded into a single space in the anyway value. Also, each value ends with a single newline, even though there are two blank lines between “fourth and last” and “anyway”, and no blank lines between “in life” and “lastly”.

5.7. Compact Representation

Another, more compact, way to represent lists and mappings is to begin with a start character, finish with an end character, and separate elements with “,” (comma, U+2C).

For lists, the start and end characters are “[” (left square brace, U+5B) and “]” (right square brace, U+5D), respectively. In the following example, the values in the mapping are identical:

  - echo
  - hello, world!
two: [ echo, "hello, world!" ]

Note: The double-quotes around the second list element of the second value; they prevent the comma from being misinterpreted as an element separator. (If we remove them, the list would have three elements: "echo", "hello" and "world!".)

For mappings, the start and end characters are “{” (left curly brace, U+7B) and “}” (right curly brace, U+7D), respectively. In the following example, the values of both one and two are identical:

  roses: red
  violets: blue

two: { roses: red, violets: blue }

5.8. Additional Information

There is much more to YAML, not described in this topic: directives, complex mapping keys, flow styles, references, aliases, and tags. For detailed information, see the official YAML site, specifically the latest ( version 1.2 at time of writing) specification.