Abstract data types

If you’re using data types to match or restrict values and need more flexibility than what the core data types (such as String or Array) allow, you can use one of the abstract data types to construct a data type that suits your needs and matches the values you want.

Each of Puppet's core data types has a corresponding value that represents that data type, which can be used to match values of that type in several contexts. Each of those core data types only match a particular set of values. They let you further restrict the values they’ll match, but only in limited ways, and there’s no way to expand the set of values they’ll match. If you need to do this, use the corresponding abstract data type.

Flexible data types

These abstract data types can match values with a variety of concrete data types. Some of them are similar to a concrete type but offer alternate ways to restrict them (for example, Enum), and some of them let you combine types and match a union of what they would individually match (for example, Variant and Optional).

The Optional data type

The Optional data type wraps one other data type, and results in a data type that matches anything that type would match plus undef. This is useful for matching values that are allowed to be absent. It takes one required parameter.

The full signature for Optional is:
Optional[<DATA TYPE>]
Position Parameter Data type Default value Description
1 Data type Type or String none (you must specify a value) The data type to add undef to.

If you specify a string "my string" as the parameter, it's equivalent to using Optional[Enum["my string"]] — it matches only that exact string value or undef.

Optional[<DATA TYPE>] is equivalent to Variant[ <DATA TYPE>, Undef ].

Examples:
Optional[String]
Matches any string or undef.
Optional[Array[Integer[0, 10]]]
Matches an array of integers between 0 and 10, or undef.
Optional["present"]
Matches the exact string "present" or undef.

The NotUndef data type

The NotUndef type matches any value except undef. It can also wrap one other data type, resulting in a type that matches anything the original type would match except undef. It accepts one optional parameter.

The full signature for NotUndef is:
NotUndef[<DATA TYPE>]
Position Parameter Data type Default value Description
1 Data type Type or String Any The data type to subtract undef from.

If you specify a string as a parameter for NotUndef, it's equivalent to writing NotUndef[Enum["my string"]] — it matches only that exact string value. This doesn’t actually subtract anything, because the Enum wouldn’t have matched undef anyway, but it's a convenient notation for mandatory keys in Struct schema hashes.

The Variant data type

The Variant data type combines any number of other data types, and results in a type that matches the union of what any of those data types would match. It takes any number of parameters, and requires at least one.

The full signature for Variant is:
Variant[ <DATA TYPE>, (<DATA TYPE, ...) ]
Position Parameter Data type Default value Description
1 and up Data type Type none (required) A data type to add to the resulting compound data type. You must provide at least one data type parameter, and can provide any number of additional ones.
Examples:
Variant[Integer, Float]
Matches any integer or floating point number (equivalent to Numeric).
Variant[Enum['true', 'false'], Boolean]
matches 'true', 'false', true, or false.

The Pattern data type

The Pattern data type only matches strings, but it provides an alternate way to restrict which strings it matches. It takes any number of regular expressions, and results in a data type that matches any strings that would match any of those regular expressions. It takes any number of parameters, and requires at least one.

The full signature for Pattern is:
Pattern[ <REGULAR EXPRESSION>, (<REGULAR EXPRESSION>, ...) ]
Position Parameter Data type Default value Description
1 and up Regular expression Regexp none (required) A regular expression describing a set of strings that the resulting data type matches. You must provide at least one regular expression parameter, and can provide any number of additional ones.

You can use capture groups in the regular expressions, but they won’t cause any variables, like $1, to be set.

Examples:
Pattern[/\A[a-z].*/]
Matches any string that begins with a lowercase letter.
Pattern[/\A[a-z].*/, /\ANone\Z/]
Matches the above or the exact string None.

The Enum data type

The Enum data type only matches strings, but it provides an alternate way to restrict which strings it matches. It takes any number of strings, and results in a data type that matches any string values that exactly match one of those strings. Unlike the == operator, this matching is case-sensitive. It takes any number of parameters, and requires at least one.

The full signature for Enum is:
Enum[ <OPTION>, (<OPTION>, ...) ]
Position Parameter Data type Default value Description
1 and up Option String none (required) One of the literal string values that the resulting data type matches. You must provide at least one option parameter, and can provide any number of additional ones.
Examples:
Enum['stopped', 'running']
Matches the strings 'stopped' and 'running', and no other values.
Enum['true', 'false']
Matches the strings 'true' and 'false', and no other values. Does not match the boolean values true or false (without quotes).

The Tuple data type

The Tuple type only matches arrays, but it lets you specify different data types for every element of the array, in order. It takes any number of parameters, and requires at least one.

The full signature for Tuple is:
Tuple[ <CONTENT TYPE>, (<CONTENT TYPE>, ..., <MIN SIZE>, <MAX SIZE>) ]
Position Parameter Data type Default value Description
1 and up Content type Type none (required) What kind of values the array contains at the given position. You must provide at least one content type parameter, and can provide any number of additional ones.
-2 (second-last) Minimum size Integer number of content types The minimum number of elements in the array. If this is smaller than the number of content types you provided, any elements beyond the minimum are optional; however, if present, they must still match the provided content types. This parameter accepts the value default, but this won’t use the default value; instead, it means 0 (all elements optional).
-1 (last) Maximum size Integer number of content types

The maximum number of elements in the array. You cannot specify a maximum without also specifying a minimum. If the maximum is larger than the number of content types you provided, it means the array can contain any number of additional elements, which all must match the last content type. This parameter accepts the value default, but this won’t use the default value; instead, it means infinity (any number of elements matching the final content type).

Don't set the maximum smaller than the number of content types you provide.
Examples:
Tuple[String, Integer]
Matches a two-element array containing a string followed by an integer, like ["hi", 2].
Tuple[String, Integer, 1]
Matches the above or a one-element array containing only a string.
Tuple[String, Integer, 1, 4]
Matches an array containing one string followed by zero to three integers.
Tuple[String, Integer, 1, default]
Matches an array containing one string followed by any number of integers.

The Struct data type

The Struct type only matches hashes, but it lets you specify:

  • The name of every allowed key.

  • Whether each key is required or optional.

  • The allowed data type for each of those keys’ values.

It takes one mandatory parameter.

The full signature for Struct is:
Struct[<SCHEMA HASH>]
Position Parameter Data type Default value Description
1 Schema hash Hash[Variant[String, Optional, NotUndef], Type] none (required) A hash that has all of the allowed keys and data types for the struct.

A Struct’s schema hash must have the same keys as the hashes it matches. Each value must be a data type that matches the allowed values for that key.

The keys in a schema hash are usually strings. They can also be an Optional or NotUndef type with the key’s name as their parameter.

If a key is a string, Puppet uses the value’s type to determine whether it’s optional — because accessing a missing key resolves to the value undef, the key is optional if the value type accepts undef (like Optional[Array]).

Note that this doesn’t distinguish between an explicit value of undef and an absent key. If you want to be more explicit, you can use Optional['my_key'] to indicate that a key can be absent, and NotUndef['my_key'] to make it mandatory. If you use one of these, a value type that accepts undef is only used to decide about explicit undef values, not missing keys.

The following example Struct matches hashes like {mode => 'read', path => '/etc/fstab'}. Both the mode and path keys are mandatory; mode’s value must be one of 'read', 'write', or 'update', and path must be a string of at least one character:
Struct[{mode => Enum[read, write, update], path => String[1]}]
The following data type would match the same values as the previous example, but the path key is optional. If present, path must match String[1] or Undef:
Struct[{mode => Enum[read, write, update], path => Optional[String[1]]}]
In the following data type, the owner key can be absent, but if it’s present, it must be a string; a value of undef isn’t allowed:
Struct[{mode => Enum[read, write, update], path => Optional[String[1]], Optional[owner] => String[1]}]
In the following data type, the owner key is mandatory, but it allows an explicit undef value:
Struct[{mode => Enum[read, write, update], path => Optional[String[1]], NotUndef[owner] => Optional[String[1]]}]

The SemVer data type

A SemVer instance defines a single semantic version or range of versions. For example, "1.2.3" or ">= 1.0.0 < 2.0.0".

It consists of the following five segments:

  • Major version (required)
  • Minor version (required)
  • Patch version (required)
  • Prerelease tag (optional)
  • Build tag (optional)

You can create an instance of SemVer from a String, individual values, or a hash of individual values.

The signatures are: 

type PositiveInteger = Integer[0,default] type SemVerQualifier = Pattern[/\A(?<part>[0-9A-Za-z-]+)(?:\.\g<part>)*\Z/] type SemVerString = String[1] type SemVerHash = Struct[{ major =>PositiveInteger, minor =>PositiveInteger, patch =>PositiveInteger, Optional[prerelease] =>SemVerQualifier, Optional[build] =>SemVerQualifier }] function SemVer.new(SemVerString $str) function SemVer.new( PositiveInteger $major PositiveInteger $minor PositiveInteger $patch Optional[SemVerQualifier] $prerelease = undef Optional[SemVerQualifier] $build = undef ) function SemVer.new(SemVerHash $hash_args)

Examples:

SemVer.new("1.2.3")
Creates a SemVer instance from a string
SemVer.new(1, 2, 3, "rc4", "5"
Creates a SemVer instance from a list of arguments
SemVer.new(major => 1, minor => 2, patch => 3, prerelease => "rc4", build =>"5")
Creates a SemVer instance from a hash

You can parameterize the SemVer type to restrict which values the type matches. The values are defined by one or more Strings or SemVerRanges.

The full signatures are: 

SemVer[<String>]

The <String> specifies a semantic version string — representing a single version or range of versions. A SemVer instance matches the parameterized type, if the instance is within the range defined by the type. For example:

$t = SemVer['> 1.0.0 < 2.0.0'] notice(SemVer('1.2.3') =~ $t) # true notice(SemVer('2.3.4') =~ $t) # false
SemVer[ <SemVerRange>, ( <SemVerRange>, ... ) ]

The SemVer type is accompanied by the SemVerRange type, which you can define to restrict matches to a contiguous version range. For example:

 $t = SemVer[SemVerRange('>=1.0.0 <2.0.0')] notice(SemVer('1.2.3') =~ $t) # true notice(SemVer('2.3.4') =~ $t) # false

When you define a parameterized SemVer type using multiple ranges, and the instance is enclosed in at least one of the ranges, the SemVer instance matches the type.

$t = SemVer[SemVerRange('>=1.0.0 <2.0.0'), SemVerRange('>=3.0.0 <4.0.0')] notice(SemVer('1.2.3') =~ $t) # true notice(SemVer('2.5.0') =~ $t) # false notice(SemVer('3.0.0') =~ $t) # true

If any of the ranges are adjacent or overlap, they get normalized (merged). For example, the following are equal: 

SemVer['>=1.0.0 <4.0.0'] SemVer[SemVerRange('>=1.0.0 <3.0.0'), SemVerRange('>=2.0.0 <4.0.0')] # overlap SemVer[SemVerRange('>=1.0.0 <2.0.0'), SemVerRange('>=2.0.0 <4.0.0')] # adjacent

The SemVerRange data type

An instance of SemVerRange represents a contiguous semantic version range. The string format of a SemVerRange is specified by the SemVer Range Grammar. The SemVerRange type does not support the logical or operator (||).

You can create an instance of SemVerRange (SemVerRange.new) from a String, individual values, or a hash of individual values. The signatures are:
type SemVerRangeString = String[1] type SemVerRangeHash = Struct[{ min => Variant[default, SemVer], Optional[max] => Variant[default, SemVer], Optional[exclude_max] => Boolean }] function SemVerRange.new(SemVerRangeString $semver_range_string) function SemVerRange.new( Variant[default,SemVer] $min Variant[default,SemVer] $max Optional[Boolean] $exclude_max = undef } function SemVerRange.new(SemVerRangeHash $semver_range_hash)

Examples:

SemVerRange.new(">1.0.0"))
Creates a SemVerRange from a String
SemVerRange.new(SemVer.new("1.0.0"), SemVer.new("2.0.0"), true)
Creates a SemVerRange instance from a list of arguments.
SemVerRange.new(min => SemVer.new("1.0.0"), max => SemVer.new("2.0.0"), exclude_max => true)
Creates a SemVerRange instance from a hash.

By default, the range includes the maximum value so that the following examples are equal:

SemVerRange.new(">= 1.0.0 <= 2.0.0") SemVerRange.new(SemVer.new("1.0.0"), SemVer.new("2.0.0"))

The maximum value can be excluded so that the following examples are equal:

SemVerRange.new(">= 1.0.0 < 2.0.0") SemVerRange.new(SemVer.new("1.0.0"), SemVer.new("2.0.0"), true)

Unlike the SemVer type, you cannot paramatize the SemVerRange type, as it represents all semantic version ranges.

Parent types

These abstract data types are the parents of multiple other types, and match values that would match any of their sub-types. They’re useful when you have very loose restrictions but not no restrictions.

The Scalar data type

The Scalar data type matches all values of the following concrete data types:

  • Numbers (both integers and floats)

  • Strings

  • Booleans

  • Regular expressions

It doesn’t match undef, default, resource references, arrays, or hashes. It takes no parameters.

Scalar is equivalent to Variant[Integer, Float, String, Boolean, Regexp].

The ScalarData data type

The ScalarData data type represents a restricted set of "value" data types that have concrete direct representation in JSON.

ScalarData is an alias for Variant[Integer, Float, String, Boolean].

The Data data type

The Data data type matches any value that would match Scalar, but it also matches:

  • undef

  • Arrays that only contain values that also match Data

  • Hashes whose keys match Scalar and whose values also match Data

It doesn't match default or resource references. It takes no parameters.

Data is especially useful because it represents the subset of types that can be directly represented in almost all serialization format, such as JSON.

The Collection data type

The Collection type matches any array or hash, regardless of what kinds of values or keys it contains. It only partially overlaps with Data— there are values, such as an array of resource references, that match Collection but do not match Data.

Collection is equivalent to Variant[Array[Any], Hash[Any, Any]].

The CatalogEntry data type

The CatalogEntry data type is the parent type of Resource and Class. Like those types, the Puppet language contains no values that it ever matches. However, the type Type[CatalogEntry] matches any class reference or resource reference. It takes no parameters.

The Any data type

The Any data type matches any value of any data type.

The Iterable data type

The Iterable data type represents all data types that can be iterated; in other words, all data types where the value is some kind of container of individual values. The Iterable type is abstract in that it does not specify if it represents a concrete data type (such as Array) that has storage in memory, of if it is an algorithmic construct like a transformation function (such as the step function).

The Iterator data type

The Iterator data type is an Iterable that does not have a concrete backing data type holding a copy of the values it will produce when iterated over. It represents an algorithmic transformation of some source (which in turn can be algorithmic).

An Iterator may not be assigned to an attribute of a resource, and it may not be used as an argument to version 3.x functions. To create a concrete value an Iterator must be "rolled out" by using a function at the end of a chain that produces a concrete value.

The RichData data type

The RichData data type represents the abstract notion of "serializeable" and includes all the types in the type system except Runtime, Callable, Iterator, and Iterable. It is expressed as an alias of Variant[Default, Object, Scalar, SemVerRange, Type, Undef, Array[RichData], Hash[RichData, RichData]].

Other types

These types aren’t quite like the others.

The Callable data type

The Callable data type matches callable lambdas provided as function arguments.

There is no way to interact with Callable values in the Puppet language, but Ruby functions written to the function API (Puppet::Functions) can use Callable to inspect the lambda provided to the function.

The full signature for Callable is:
Callable[ (<DATA TYPE>, ...,) <MIN COUNT>, <MAX COUNT>, <BLOCK TYPE> ]
All of these parameters are optional.
Position Parameter Data type Default value Description
1 to n Data type Type none Any number of data types, representing the data type of each argument the lambda accepts.
-3 (third last) Minimum count Integer 0 The minimum number of arguments the lambda accepts. This parameter accepts the value default, which uses its default value 0.
-2 (second last) Maximum count Integer infinity The maximum number of arguments the lambda accepts. This parameter accepts the value default, which uses its default value, infinity.
-1 (last) Block type Type[Callable] none The block_type of the lambda.