» Expressions

Expressions are used to refer to or compute values within a configuration. The simplest expressions are just literal values, like "hello" or 5, but the Terraform language also allows more complex expressions such as references to data exported by resources, arithmetic, conditional evaluation, and a number of built-in functions.

Expressions can be used in a number of places in the Terraform language, but some contexts limit which expression constructs are allowed, such as requiring a literal value of a particular type or forbidding references to resource attributes. Each language feature's documentation describes any restrictions it places on expressions.

You can experiment with the behavior of Terraform's expressions from the Terraform expression console, by running the terraform console command.

The rest of this page describes all of the features of Terraform's expression syntax.

» Types and Values

The result of an expression is a value. All values have a type, which dictates where that value can be used and what transformations can be applied to it.

The Terraform language uses the following types for its values:

  • string: a sequence of Unicode characters representing some text, like "hello".
  • number: a numeric value. The number type can represent both whole numbers like 15 and fractional values like 6.283185.
  • bool: either true or false. bool values can be used in conditional logic.
  • list (or tuple): a sequence of values, like ["us-west-1a", "us-west-1c"]. Elements in a list or tuple are identified by consecutive whole numbers, starting with zero.
  • map (or object): a group of values identified by named labels, like {name = "Mabel", age = 52}.

Strings, numbers, and bools are sometimes called primitive types. Lists/tuples and maps/objects are sometimes called complex types, structural types, or collection types.

Finally, there is one special value that has no type:

  • null: a value that represents absence or omission. If you set an argument of a resource or module to null, Terraform behaves as though you had completely omitted it — it will use the argument's default value if it has one, or raise an error if the argument is mandatory. null is most useful in conditional expressions, so you can dynamically omit an argument if a condition isn't met.

» Advanced Type Details

In most situations, lists and tuples behave identically, as do maps and objects. Whenever the distinction isn't relevant, the Terraform documentation uses each pair of terms interchangeably (with a historical preference for "list" and "map").

However, module authors and provider developers should understand the differences between these similar types (and the related set type), since they offer different ways to restrict the allowed values for input variables and resource arguments.

For complete details about these types (and an explanation of why the difference usually doesn't matter), see Type Constraints.

» Type Conversion

Expressions are most often used to set values for the arguments of resources and child modules. In these cases, the argument has an expected type and the given expression must produce a value of that type.

Where possible, Terraform automatically converts values from one type to another in order to produce the expected type. If this isn't possible, Terraform will produce a type mismatch error and you must update the configuration with a more suitable expression.

Terraform automatically converts number and bool values to strings when needed. It also converts strings to numbers or bools, as long as the string contains a valid representation of a number or bool value.

  • true converts to "true", and vice-versa
  • false converts to "false", and vice-versa
  • 15 converts to "15", and vice-versa

» Literal Expressions

A literal expression is an expression that directly represents a particular constant value. Terraform has a literal expression syntax for each of the value types described above:

  • Strings are usually represented by a double-quoted sequence of Unicode characters, "like this". There is also a "heredoc" syntax for more complex strings. String literals are the most complex kind of literal expression in Terraform, and have additional documentation on this page:
    • See String Literals below for information about escape sequences and the heredoc syntax.
    • See String Templates below for information about interpolation and template directives.
  • Numbers are represented by unquoted sequences of digits with or without a decimal point, like 15 or 6.283185.
  • Bools are represented by the unquoted symbols true and false.
  • The null value is represented by the unquoted symbol null.

  • Lists/tuples are represented by a pair of square brackets containing a comma-separated sequence of values, like ["a", 15, true].

    List literals can be split into multiple lines for readability, but always require a comma between values. A comma after the final value is allowed, but not required. Values in a list can be arbitrary expressions.

  • Maps/objects are represented by a pair of curly braces containing a series of <KEY> = <VALUE> pairs:

    {
  name = "John"
  age  = 52
}

    Key/value pairs can be separated by either a comma or a line break. Values can be arbitrary expressions. Keys are strings; they can be left unquoted if they are a valid identifier, but must be quoted otherwise. You can use a non-literal expression as a key by wrapping it in parentheses, like (var.business_unit_tag_name) = "SRE".

» Indices and Attributes

Elements of list/tuple and map/object values can be accessed using the square-bracket index notation, like local.list[3]. The expression within the brackets must be a whole number for list and tuple values or a string for map and object values.

Map/object attributes with names that are valid identifiers can also be accessed using the dot-separated attribute notation, like local.object.attrname. In cases where a map might contain arbitrary user-specified keys, we recommend using only the square-bracket index notation (local.map["keyname"]).

» References to Named Values

Terraform makes several kinds of named values available. Each of these names is an expression that references the associated value; you can use them as standalone expressions, or combine them with other expressions to compute new values.

The following named values are available:

Although many of these names use dot-separated paths that resemble attribute notation for elements of object values, they are not implemented as real objects. This means you must use them exactly as written: you cannot use square-bracket notation to replace the dot-separated paths, and you cannot iterate over the "parent object" of a named entity (for example, you cannot use aws_instance in a for expression).

» Named Values and Dependencies

Constructs like resources and module calls often use references to named values in their block bodies, and Terraform analyzes these expressions to automatically infer dependencies between objects. For example, an expression in a resource argument that refers to another managed resource creates an implicit dependency between the two resources.

» References to Resource Attributes

The most common reference type is a reference to an attribute of a resource which has been declared either with a resource or data block. Because the contents of such blocks can be quite complicated themselves, expressions referring to these contents can also be complicated.

Consider the following example resource block:

resource "aws_instance" "example" {
  ami           = "ami-abc123"
  instance_type = "t2.micro"

  ebs_block_device {
    device_name = "sda2"
    volume_size = 16
  }
  ebs_block_device {
    device_name = "sda3"
    volume_size = 20
  }
}

The documentation for aws_instance lists all of the arguments and nested blocks supported for this resource type, and also lists a number of attributes that are exported by this resource type. All of these different resource type schema constructs are available for use in references, as follows:

  • The ami argument set in the configuration can be used elsewhere with the reference expression aws_instance.example.ami.
  • The id attribute exported by this resource type can be read using the same syntax, giving aws_instance.example.id.
  • The arguments of the ebs_block_device nested blocks can be accessed using a splat expression. For example, to obtain a list of all of the device_name values, use aws_instance.example.ebs_block_device[*].device_name.
  • The nested blocks in this particular resource type do not have any exported attributes, but if ebs_block_device were to have a documented id attribute then a list of them could be accessed similarly as aws_instance.example.ebs_block_device[*].id.
  • Sometimes nested blocks are defined as taking a logical key to identify each block, which serves a similar purpose as the resource's own name by providing a convenient way to refer to that single block in expressions. If aws_instance had a hypothetical nested block type device that accepted such a key, it would look like this in configuration: 
    device "foo" {
      size = 2
    }
    device "bar" {
      size = 4
    }

Arguments inside blocks with keys can be accessed using index syntax, such as aws_instance.example.device["foo"].size.

To obtain a map of values of a particular argument for labelled nested block types, use a for expression: [for k, device in aws_instance.example.device : k => device.size].

When a particular resource has the special count argument set, the resource itself becomes a list of instance objects rather than a single object. In that case, access the attributes of the instances using either splat expressions or index syntax:

» Local Named Values

Within the bodies of certain expressions, or in some other specific contexts, there are other named values available beyond the global values listed above. (For example, the body of a resource block where count is set can use a special count.index value.) These local names are described in the documentation for the specific contexts where they appear.

» Values Not Yet Known

When Terraform is planning a set of changes that will apply your configuration, some resource attribute values cannot be populated immediately because their values are decided dynamically by the remote system. For example, if a particular remote object type is assigned a generated unique id on creation, Terraform cannot predict the value of this id until the object has been created.

To allow expressions to still be evaluated during the plan phase, Terraform uses special "unknown value" placeholders for these results. In most cases you don't need to do anything special to deal with these, since the Terraform language automatically handles unknown values during expressions, so that for example adding a known value to an unknown value automatically produces an unknown value as the result.

However, there are some situations where unknown values do have a significant effect:

  • The count meta-argument for resources cannot be unknown, since it must be evaluated during the plan phase to determine how many instances are to be created.

  • If unknown values are used in the configuration of a data resource, that data resource cannot be read during the plan phase and so it will be deferred until the apply phase. In this case, the results of the data resource will also be unknown values.

  • If an unknown value is assigned to an argument inside a module block, any references to the corresponding input variable within the child module will use that unknown value.

  • If an unknown value is used in the value argument of an output value, any references to that output value in the parent module will use that unknown value.

  • Terraform will attempt to validate that unknown values are of suitable types where possible, but incorrect use of such values may not be detected until the apply phase, causing the apply to fail.

Unknown values appear in the terraform plan output as (not yet known).

» Arithmetic and Logical Operators

An operator is a type of expression that transforms or combines one or more other expressions. Operators either combine two values in some way to produce a third result value, or transform a single given value to produce a single result.

Operators that work on two values place an operator symbol between the two values, similar to mathematical notation: 1 + 2. Operators that work on only one value place an operator symbol before that value, like !true.

The Terraform language has a set of operators for both arithmetic and logic, which are similar to operators in programming languages such as JavaScript or Ruby.

When multiple operators are used together in an expression, they are evaluated in the following order of operations:

  1. !, - (multiplication by -1)
  2. *, /, %
  3. +, - (subtraction)
  4. >, >=, <, <=
  5. ==, !=
  6. &&
  7. ||

Parentheses can be used to override the default order of operations. Without parentheses, higher levels are evaluated first, so 1 + 2 * 3 is interpreted as 1 + (2 * 3) and not as (1 + 2) * 3.

The different operators can be gathered into a few different groups with similar behavior, as described below. Each group of operators expects its given values to be of a particular type. Terraform will attempt to convert values to the required type automatically, or will produce an error message if this automatic conversion is not possible.

» Arithmetic Operators

The arithmetic operators all expect number values and produce number values as results:

  • a + b returns the result of adding a and b together.
  • a - b returns the result of subtracting b from a.
  • a * b returns the result of multiplying b and b.
  • a / b returns the result of dividing a by b.
  • a % b returns the remainder of dividing a by b. This operator is generally useful only when used with whole numbers.
  • -a returns the result of multiplying a by -1.

» Equality Operators

The equality operators both take two values of any type and produce boolean values as results.

  • a == b returns true if a and b both have the same type and the same value, or false otherwise.
  • a != b is the opposite of a == b.

» Comparison Operators

The comparison operators all expect number values and produce boolean values as results.

  • a < b returns true if a is less than b, or false otherwise.
  • a <= b returns true if a is less than or equal to b, or false otherwise.
  • a > b returns true if a is greater than b, or false otherwise.
  • a >= b returns true if a is greater than or equal to b, or `false otherwise.

» Logical Operators

The logical operators all expect bool values and produce bool values as results.

  • a || b returns true if either a or b is true, or false if both are false.
  • a && b returns true if both a and b are true, or false if either one is false.
  • !a returns true if a is false, and false if a is true.

» Conditional Expressions

A conditional expression uses the value of a bool expression to select one of two values.

The syntax of a conditional expression is as follows:

condition ? true_val : false_val

If condition is true then the result is true_val. If condition is false then the result is false_val.

A common use of conditional expressions is to define defaults to replace invalid values:

var.a != "" ? var.a : "default-a"

If var.a is an empty string then the result is "default-a", but otherwise it is the actual value of var.a.

Any of the equality, comparison, and logical operators can be used to define the condition. The two result values may be of any type, but they must both be of the same type so that Terraform can determine what type the whole conditional expression will return without knowing the condition value.

» Function Calls

The Terraform language has a number of built-in functions that can be used within expressions as another way to transform and combine values. These are similar to the operators but all follow a common syntax:

<FUNCTION NAME>(<ARGUMENT 1>, <ARGUMENT 2>)

The function name specifies which function to call. Each defined function expects a specific number of arguments with specific value types, and returns a specific value type as a result.

Some functions take an arbitrary number of arguments. For example, the min function takes any amount of number arguments and returns the one that is numerically smallest:

min(55, 3453, 2)

» Expanding Function Arguments

If the arguments to pass to a function are available in a list or tuple value, that value can be expanded into separate arguments. Provide the list value as an argument and follow it with the ... symbol:

min([55, 2453, 2]...)

The expansion symbol is three periods (...), not a Unicode ellipsis character (…). Expansion is a special syntax that is only available in function calls.

» Available Functions

For a full list of available functions, see the function reference.

» for Expressions

A for expression creates a complex type value by transforming another complex type value. Each element in the input value can correspond to either one or zero values in the result, and an arbitrary expression can be used to transform each input element into an output element.

For example, if var.list is a list of strings, then the following expression produces a list of strings with all-uppercase letters:

[for s in var.list : upper(s)]

This for expression iterates over each element of var.list, and then evaluates the expression upper(s) with s set to each respective element. It then builds a new tuple value with all of the results of executing that expression in the same order.

The type of brackets around the for expression decide what type of result it produces. The above example uses [ and ], which produces a tuple. If { and } are used instead, the result is an object, and two result expressions must be provided separated by the => symbol:

{for s in var.list : s => upper(s)}

This expression produces an object whose attributes are the original elements from var.list and their corresponding values are the uppercase versions.

A for expression can also include an optional if clause to filter elements from the source collection, which can produce a value with fewer elements than the source:

[for s in var.list : upper(s) if s != ""]

The source value can also be an object or map value, in which case two temporary variable names can be provided to access the keys and values respectively:

[for k, v in var.map : length(k) + length(v)]

Finally, if the result type is an object (using { and } delimiters) then the value result expression can be followed by the ... symbol to group together results that have a common key:

{for s in var.list : substr(s, 0, 1) => s... if s != ""}

» Splat Expressions

A splat expression provides a more concise way to express a common operation that could otherwise be performed with a for expression.

If var.list is a list of objects that all have an attribute id, then a list of the ids could be produced with the following for expression:

[for o in var.list : o.id]

This is equivalent to the following splat expression:

var.list[*].id

The special [*] symbol iterates over all of the elements of the list given to its left and accesses from each one the attribute name given on its right. A splat expression can also be used to access attributes and indexes from lists of complex types by extending the sequence of operations to the right of the symbol:

var.list[*].interfaces[0].name

The above expression is equivalent to the following for expression:

[for o in var.list : o.interfaces[0].name]

Splat expressions also have another useful effect: if they are applied to a value that is not a list or tuple then the value is automatically wrapped in a single-element list before processing. That is, var.single_object[*].id is equivalent to [var.single_object][*].id, or effectively [var.single_object.id]. This behavior is not interesting in most cases, but it is particularly useful when referring to resources that may or may not have count set, and thus may or may not produce a tuple value:

aws_instance.example[*].id

The above will produce a list of ids whether aws_instance.example has count set or not, avoiding the need to revise various other expressions in the configuration when a particular resource switches to and from having count set.

» Legacy (Attribute-only) Splat Expressions

An older variant of the splat expression is available for compatibility with code written in older versions of the Terraform language. This is a less useful version of the splat expression, and should be avoided in new configurations.

An "attribute-only" splat expression is indicated by the sequence .* (instead of [*]):

var.list.*.interfaces[0].name

This form has a subtly different behavior, equivalent to the following for expression:

[for o in var.list : o.interfaces][0].name

Notice that with the attribute-only splat expression the index operation [0] is applied to the result of the iteration, rather than as part of the iteration itself.

» dynamic blocks

Within top-level block constructs like resources, expressions can usually be used only when assigning a value to an argument using the name = expression form. This covers many uses, but some resource types include repeatable nested blocks in their arguments, which do not accept expressions:

resource "aws_security_group" "example" {
  name = "example" # can use expressions here

  ingress {
    # but the "ingress" block is always a literal block
  }
}

You can dynamically construct repeatable nested blocks like ingress using a special dynamic block type, which is supported inside resource, data, provider, and provisioner blocks:

resource "aws_security_group" "example" {
  name = "example" # can use expressions here

  dynamic "ingress" {
    for_each = var.service_ports
    content {
      from_port = ingress.value
      to_port   = ingress.value
      protocol  = "tcp"
    }
  }
}

A dynamic block acts much like a for expression, but produces nested blocks instead of a complex typed value. It iterates over a given complex value, and generates a nested block for each element of that complex value.

  • The label of the dynamic block ("ingress" in the example above) specifies what kind of nested block to generate.
  • The for_each argument provides the complex value to iterate over.
  • The iterator argument (optional) sets the name of a temporary variable that represents the current element of the complex value. If omitted, the name of the variable defaults to the label of the dynamic block ("ingress" in the example above).
  • The labels argument (optional) is a list of strings that specifies the block labels, in order, to use for each generated block. You can use the temporary iterator variable in this value.
  • The nested content block defines the body of each generated block. You can use the temporary iterator variable inside this block.

Since the for_each argument accepts any collection or structural value, you can use a for expression or splat expression to transform an existing collection.

A dynamic block can only generate arguments that belong to the resource type, data source, provider or provisioner being configured. It is not possible to generate meta-argument blocks such as lifecycle and provisioner blocks, since Terraform must process these before it is safe to evaluate expressions.

» Best Practices for dynamic Blocks

Overuse of dynamic blocks can make configuration hard to read and maintain, so we recommend using them only when you need to hide details in order to build a clean user interface for a re-usable module. Always write nested blocks out literally where possible.

» String Literals

The Terraform language has two different syntaxes for string literals. The most common is to delimit the string with quote characters ("), like "hello". In quoted strings, the backslash character serves as an escape sequence, with the following characters selecting the escape behavior:

Sequence Replacement
\n Newline
\r Carriage Return
\t Tab
\" Literal quote (without terminating the string)
\\ Literal backslash
\uNNNN Unicode character from the basic multilingual plane (NNNN is four hex digits)
\UNNNNNNNN Unicode character from supplimentary planes (NNNNNNNN is eight hex digits)

The alternative syntax for string literals is the so-called "heredoc" style, inspired by Unix shell languages. This style allows multi-line strings to be expressed more clearly by using a custom delimiter word on a line of its own to close the string:

<<EOT
hello
world
EOT

The << marker followed by any identifier at the end of a line introduces the sequence. Terraform then processes the following lines until it finds one that consists entirely of the identifier given in the introducer. In the above example, EOT is the identifier selected. Any identifier is allowed, but conventionally this identifier is in all-uppercase and beings with EO, meaning "end of". EOT in this case stands for "end of text".

The "heredoc" form shown above requires that the lines following be flush with the left margin, which can be awkward when an expression is inside an indented block:

block {
  value = <<EOT
hello
world
EOT
}

To improve on this, Terraform also accepts an indented heredoc string variant that is introduced by the <<- sequence:

block {
  value = <<-EOT
  hello
    world
  EOT
}

In this case, Terraform analyses the lines in the sequence to find the one with the smallest number of leading spaces, and then trims that many spaces from the beginning of all of the lines, leading to the following result:

hello
  world

Backslash sequences are not interpreted in a heredoc string expression. Instead, the backslash character is interpreted literally.

In both quoted and heredoc string expressions, Terraform supports template sequences that begin with ${ and %{. These are described in more detail in the following section. To include these sequences literally without beginning a template sequence, double the leading character: $${ or %%{.

» String Templates

Within quoted and heredoc string expressions, the sequences ${ and %{ begin template sequences. Templates let you directly embed expressions into a string literal, to dynamically construct strings from other values.

» Interpolation

A ${ ... } sequence is an interpolation, which evaluates the expression given between the markers, converts the result to a string if necessary, and then inserts it into the final string:

"Hello, ${var.name}!"

In the above example, the named object var.name is accessed and its value inserted into the string, producing a result like "Hello, Juan!".

» Directives

A %{ ... } sequence is a directive, which allows for conditional results and iteration over collections, similar to conditional and for expressions.

The following directives are supported:

  • The if <BOOL>/else/endif directive chooses between two templates based on the value of a bool expression:

    "Hello, %{ if var.name != "" }${var.name}%{ else }unnamed%{ endif }!"

    The else portion may be omitted, in which case the result is an empty string if the condition expression returns false.

  • The for <NAME> in <COLLECTION> / endfor directive iterates over the elements of a given collection or structural value and evaluates a given template once for each element, concatenating the results together:

    <<EOT
%{ for ip in aws_instance.example.*.private_ip }
server ${ip}
%{ endfor }
EOT

    The name given immediately after the for keyword is used as a temporary variable name which can then be referenced from the nested template.

To allow template directives to be formatted for readability without adding unwanted spaces and newlines to the result, all template sequences can include optional strip markers (~), immediately after the opening characters or immediately before the end. When a strip marker is present, the template sequence consumes all of the literal whitespace (spaces and newlines) either before the sequence (if the marker appears at the beginning) or after (if the marker appears at the end):

<<EOT
%{ for ip in aws_instance.example.*.private_ip ~}
server ${ip}
%{ endfor ~}
EOT

In the above example, the newline after each of the directives is not included in the output, but the newline after the server ${ip} sequence is retained, causing only one line to be generated for each element:

server 10.1.16.154
server 10.1.16.1
server 10.1.16.34

When using template directives, we recommend always using the "heredoc" string literal form and then formatting the template over multiple lines for readability. Quoted string literals should usually include only interpolation sequences.