YAML in 2 Minutes: A Gentle Introduction for R Users

Here’s a short introduction to YAML for R users. YAML is a data serialization format designed to be easily human readable.

Think of YAML as “JSON with comments and nicer multiline strings.” yaml12 parses YAML 1.2 (the modern specification that removes some of YAML 1.1’s surprising eager conversions) into plain R objects.

YAML has three building blocks: scalars (single values), sequences (ordered collection of items), and mappings (key/value pairs with unique keys). JSON is a subset of YAML 1.2, so all valid JSON is also valid YAML and parses the same way.

A first example

title: A Modern YAML parser written in Rust
properties: [correct, safe, fast, simple]
score: 9.5
categories:
  - yaml
  - r
  - example
settings:
  simplify: true
  note: >
    This is a folded block
    that turns line breaks
    into spaces.
  note_literal: |
    This is a literal block
    that keeps
    line breaks.

str(parse_yaml(first_example))
#> List of 5
#>  $ title     : chr "A Modern YAML parser written in Rust"
#>  $ properties: chr [1:4] "correct" "safe" "fast" "simple"
#>  $ score     : num 9.5
#>  $ categories: chr [1:3] "yaml" "r" "example"
#>  $ settings  :List of 3
#>   ..$ simplify    : logi TRUE
#>   ..$ note        : chr "This is a folded block that turns line breaks into spaces.\n"
#>   ..$ note_literal: chr "This is a literal block\nthat keeps\nline breaks.\n"

Comments

Comments start with # and run to the end of the line. They must be separated from values by whitespace and can sit on their own line or at line ends; they are ignored by the parser.

# Whole-line comment
title: example # inline comment
items: [a, b] # trailing comment

→ list(title = "example", items = c("a", "b"))

Collections

There are two “collection” types: Sequences and Mappings.

Sequences: YAML’s ordered collections

A sequence is a list of items. Each item begins with - at the parent indent.

- cat
- dog

→ c("cat", "dog") (or list("cat", "dog") when simplify = FALSE)

JSON-style arrays work too:

[cat, dog]

→ same result

Anything belonging to one of the sequence entries is indented at least one space past the dash:

- name: cat
  toys: [string, box]
- name: dog
  toys: [ball, bone]

↓

list(
  list(name = "cat", toys = c("string", "box")),
  list(name = "dog", toys = c("ball", "bone"))
)

Mappings: key/value pairs

A mapping is a set of key: value pairs at the same indent:

foo: 1
bar: true

→ list(foo = 1L, bar = TRUE)

A key at its indent owns anything indented more:

settings:
  simplify: true
  max_items: 3

→ list(settings = list(simplify = TRUE, max_items = 3L))

JSON-style objects also work:

{a: true}

→ list(a = TRUE)

Mappings become named lists in R.

Scalars

All nodes that are not collections are Scalars; these are the leaf nodes of a YAML document.

Scalars can be provided in three forms: block, quoted, or plain.

Block scalars

| starts a literal block that keeps newlines; > starts a folded block that joins lines with spaces (except blank/indented lines keep breaks). Block scalars always become strings.

|
  hello
  world

→ "hello\nworld\n"

>
  hello
  world

→ "hello world\n"

Quoted scalars

Like block scalars, quoted scalars always resolve to strings. Double quotes interpret escapes (\n, \t, \\, \"). Single quotes are literal and do not interpret escapes, except for '' which is parsed as a single '.

["line\nbreak", "quote: \"here\""]

→ c("line\nbreak", 'quote: "here"')

['line\nbreak', 'quote: ''here''']

→ c("line\\nbreak", "quote: 'here'")

Plain (unquoted) scalars

If a node is not a sequence, mapping, block scalar, or quoted scalar, it is a plain scalar.

Plain nodes can resolve to one of five types: string, int, float, bool, or null.

YAML 1.2 uses simple rules to then infer the type of a plain node:

true / false and their True/TRUE and False/FALSE variants → TRUE / FALSE
null, ~, or empty → NULL
numbers: signed, decimal, scientific, hex (0x), octal (0o), .inf, .nan → numeric() or integer()
everything else stays a string (yes, no, on, off and other aliases remain strings in YAML 1.2)

[true, 123, 4.5e2, 0x10, .inf, yes]

→ list(TRUE, 123L, 450, 16L, Inf, "yes")

If a sequence is homogeneous and simplify = TRUE, nulls become the appropriate NA_* values.

End-to-end example

doc:
  pets:
    - cat
    - dog
  numbers: [1, 2.5, 0x10, .inf, null]
  integers: [1, 2, 3, 0x10, null]
  flags: {enabled: true, label: on}
  literal: |
    hello
    world
  folded: >
    hello
    world
  quoted:
    - "line\nbreak"
    - 'quote: ''here'''
  plain: [yes, no]
  mixed: [won't simplify, 123, true]

R result (parse_yaml() with defaults):

list(
  doc = list(
    pets = c("cat", "dog"),
    numbers = c(1, 2.5, 16, Inf, NA_real_),
    integers = c(1L, 2L, 3L, 16L, NA_integer_),
    flags = list(enabled = TRUE, label = "on"),
    literal = "hello\nworld\n",
    folded = "hello world\n",
    quoted = c("line\nbreak", "quote: 'here'"),
    plain = c("yes", "no"),
    mixed = list("won't simplify", 123L, TRUE)
  )
)

Why YAML 1.2?

For most R users, the main visible difference between YAML 1.1 and 1.2 is how plain (unquoted) scalars get their types. YAML 1.2’s recommended core schema recognizes fewer special spellings, so ordinary words such as yes and on stay strings. R’s yaml package implements a YAML 1.1 parser and emitter, while yaml12 implements YAML 1.2.2.

YAML 1.1 versus 1.2 quick reference

The YAML 1.1 column below follows its type library. The YAML 1.2 column follows the recommended core schema. Individual parsers may support a subset of these rules or offer other schemas.

Plain YAML or feature	YAML 1.1 type library	YAML 1.2 core schema
`yes`, `no`, `on`, `off`, `y`, `n`	Boolean	String
`true`, `True`, `TRUE` (and false variants)	Boolean	Boolean
`010`	Octal integer `8`	Decimal integer `10`
`0o10`	String	Octal integer `8`
`0b10`	Binary integer `2`	String
`1:20`	Sexagesimal integer `80`	String
`1_000`	Decimal integer `1000`	String
`2026-01-07`	Timestamp	String
`<<` mapping key	Merge mappings	Ordinary string key

YAML 1.2 also dropped !!pairs, !!omap, !!set, !!timestamp, and !!binary from its core type set. Those explicit tags remain valid YAML syntax, but YAML 1.2 no longer assigns them core meanings. yaml12 preserves them as unhandled tags, and handlers let application code opt into their meaning. The advanced YAML article shows how this works.

These changes make YAML 1.2 more conservative, not string-only. Plain true, null, and numeric forms still get typed values. For example, 10.23 is a number in both versions; quote it if it must remain a string.

Here is how yaml12 resolves a few values that differ from YAML 1.1:

yaml_1_2 <- "
country: NO
enabled: on
port: 22:22
leading_zero: 010
octal: 0o10
release_date: 2026-01-07
"

str(parse_yaml(yaml_1_2))
#> List of 6
#>  $ country     : chr "NO"
#>  $ enabled     : chr "on"
#>  $ port        : chr "22:22"
#>  $ leading_zero: int 10
#>  $ octal       : int 8
#>  $ release_date: chr "2026-01-07"

See the YAML 1.2 changes for the full specification-level list and the yaml12 release post for more background.

Quick notes

Indentation defines structure for collections. Sibling elements share an indent, children are indented more. YAML 1.2 forbids tabs; use spaces.
All JSON is valid YAML.
Homogeneous sequences simplify to vectors unless simplify = FALSE.
Block scalars (|, >) always produce strings.
Boolean words are true/false and their True/TRUE and False/FALSE variants.
null maps to NULL (or NA inside simplified vectors).

These essentials cover most YAML you’ll run into in practice. If you encounter YAML tags or non-string mapping keys, check out the “Advanced YAML” vignette.