node package manager


JS-YAML - YAML 1.2 parser / writer for JavaScript

Build Status NPM version

Online Demo

This is an implementation of YAML, a human-friendly data serialization language. Started as PyYAML port, it was completely rewritten from scratch. Now it's very fast, and supports 1.2 spec.


YAML module for node.js

npm install js-yaml

CLI executable

If you want to inspect your YAML files from CLI, install js-yaml globally:

npm install -g js-yaml


usage: js-yaml [-h] [-v] [-c] [-t] file

Positional arguments:
  file           File with YAML document(s)

Optional arguments:
  -h, --help     Show this help message and exit.
  -v, --version  Show program's version number and exit.
  -c, --compact  Display errors in compact mode
  -t, --trace    Show stack trace on error

Bundled YAML library for browsers

<!-- esprima required only for !!js/function -->
<script src="esprima.js"></script>
<script src="js-yaml.min.js"></script>
<script type="text/javascript">
var doc = jsyaml.load('greeting: hello\nname: world');

Browser support was done mostly for the online demo. If you find any errors - feel free to send pull requests with fixes. Also note, that IE and other old browsers needs es5-shims to operate.


  1. We have no resources to support browserified version. Don't expect it to be well tested. Don't expect fast fixes if something goes wrong there.
  2. !!js/function in browser bundle will not work by default. If you really need it - load esprima parser first (via amd or directly).
  3. !!bin in browser will return Array, because browsers do not support node.js Buffer and adding Buffer shims is completely useless on practice.


Here we cover the most 'useful' methods. If you need advanced details (creating your own tags), see wiki and examples for more info.

yaml = require('js-yaml');
fs   = require('fs');
// Get document, or throw exception on error 
try {
  var doc = yaml.safeLoad(fs.readFileSync('/home/ixti/example.yml', 'utf8'));
} catch (e) {

safeLoad (string [ , options ])

Recommended loading way. Parses string as single YAML document. Returns a JavaScript object or throws YAMLException on error. By default, does not support regexps, functions and undefined. This method is safe for untrusted data.


  • filename (default: null) - string to be used as a file path in error/warning messages.
  • onWarning (default: null) - function to call on warning messages. Loader will throw on warnings if this function is not provided.
  • schema (default: DEFAULT_SAFE_SCHEMA) - specifies a schema to use.
  • json (default: false) - compatibility with JSON.parse behaviour. If true, then duplicate keys in a mapping will override values rather than throwing an error.

NOTE: This function does not understand multi-document sources, it throws exception on those.

NOTE: JS-YAML does not support schema-specific tag resolution restrictions. So, the JSON schema is not as strictly defined in the YAML specification. It allows numbers in any notation, use Null and NULL as null, etc. The core schema also has no such restrictions. It allows binary notation for integers.

load (string [ , options ])

Use with care with untrusted sources. The same as safeLoad() but uses DEFAULT_FULL_SCHEMA by default - adds some JavaScript-specific types: !!js/function, !!js/regexp and !!js/undefined. For untrusted sources, you must additionally validate object structure to avoid injections:

var untrusted_code = '"toString": !<,2002:js/function> "function (){very_evil_thing();}"';
// I'm just converting that string, what could possibly go wrong? 
require('js-yaml').load(untrusted_code) + ''

safeLoadAll (string [, iterator] [, options ])

Same as safeLoad(), but understands multi-document sources. Applies iterator to each document if specified, or returns aray of documents.

var yaml = require('js-yaml');
yaml.safeLoadAll(data, function (doc) {

loadAll (string [, iterator] [ , options ])

Same as safeLoadAll() but uses DEFAULT_FULL_SCHEMA by default.

safeDump (object [ , options ])

Serializes object as a YAML document. Uses DEFAULT_SAFE_SCHEMA, so it will throw an exception if you try to dump regexps or functions. However, you can disable exceptions by setting the skipInvalid option to true.


  • indent (default: 2) - indentation width to use (in spaces).
  • skipInvalid (default: false) - do not throw on invalid types (like function in the safe schema) and skip pairs and single values with such types.
  • flowLevel (default: -1) - specifies level of nesting, when to switch from block to flow style for collections. -1 means block style everwhere
  • styles - "tag" => "style" map. Each tag may have own set of styles.
  • schema (default: DEFAULT_SAFE_SCHEMA) specifies a schema to use.
  • sortKeys (default: false) - if true, sort keys when dumping YAML. If a function, use the function to sort the keys.
  • lineWidth (default: 80) - set max line width.
  • noRefs (default: false) - if true, don't convert duplicate objects into references
  • noCompatMode (default: false) - if true don't try to be compatible with older yaml versions. Currently: don't quote "yes", "no" and so on, as required for YAML 1.1
  • condenseFlow (default: false) - if true flow sequences will be condensed, omitting the space between key: value or a, b. Eg. '[a,b]' or {a:{b:c}}. Can be useful when using yaml for pretty URL query params as spaces are %-encoded.

The following table show availlable styles (e.g. "canonical", "binary"...) available for each tag (.e.g. !!null, !!int ...). Yaml ouput is shown on the right side after => (default setting) or ->:

  "canonical"   -> "~"
  "lowercase"   => "null"
  "uppercase"   -> "NULL"
  "camelcase"   -> "Null"
  "binary"      -> "0b1", "0b101010", "0b1110001111010"
  "octal"       -> "01", "052", "016172"
  "decimal"     => "1", "42", "7290"
  "hexadecimal" -> "0x1", "0x2A", "0x1C7A"
  "lowercase"   => "true", "false"
  "uppercase"   -> "TRUE", "FALSE"
  "camelcase"   -> "True", "False"
  "lowercase"   => ".nan", '.inf'
  "uppercase"   -> ".NAN", '.INF'
  "camelcase"   -> ".NaN", '.Inf'


safeDump (object, {
  'styles': {
    '!!null': 'canonical' // dump null as ~ 
  'sortKeys': true        // sort object keys 

dump (object [ , options ])

Same as safeDump() but without limits (uses DEFAULT_FULL_SCHEMA by default).

Supported YAML types

The list of standard YAML tags and corresponding JavaScipt types. See also YAML tag discussion and YAML types repository.

!!null ''                   # null
!!bool 'yes'                # bool
!!int '3...'                # number
!!float '3.14...'           # number
!!binary '...base64...'     # buffer
!!timestamp 'YYYY-...'      # date
!!omap [ ... ]              # array of key-value pairs
!!pairs [ ... ]             # array or array pairs
!!set { ... }               # array of objects with given keys and null values
!!str '...'                 # string
!!seq [ ... ]               # array
!!map { ... }               # object

JavaScript-specific tags

!!js/regexp /pattern/gim            # RegExp
!!js/undefined ''                   # Undefined
!!js/function 'function () {...}'   # Function


Note, that you use arrays or objects as key in JS-YAML. JS does not allow objects or arrays as keys, and stringifies (by calling .toString method) them at the moment of adding them.

? [ foo, bar ]
- baz
? { foo: bar }
- baz
  - baz
{ "foo,bar": ["baz"], "[object Object]": ["baz", "baz"] }

Also, reading of properties on implicit block mapping keys is not supported yet. So, the following YAML document cannot be loaded.

&anchor foo:
  foo: bar
  *anchor: duplicate key
  baz: bat
  *anchor: duplicate key

Breaking changes in 2.x.x -> 3.x.x

If you have not used custom tags or loader classes and not loaded yaml files via require(), no changes are needed. Just upgrade the library.

Otherwise, you should:

  1. Replace all occurences of require('xxxx.yml') by fs.readFileSync() + yaml.safeLoad().
  2. rewrite your custom tags constructors and custom loader classes, to conform the new API. See examples and wiki for details.


View the LICENSE file (MIT).