Types

Chapel is a statically typed language with a rich set of types. These include a set of predefined primitive types, enumerated types, structured types (classes, records, unions, tuples), data parallel types (ranges, domains, arrays), and synchronization types (sync, atomic).

The syntax of a type is as follows:

type-expression:
  primitive-type
  enum-type
  structured-type
  dataparallel-type
  synchronization-type
  lvalue-expression
  if-expression
  unary-expression
  binary-expression
  expression

Many expressions are syntactically allowed as a type; however not all expressions produce a type. For example, a call to a function is syntactically allowed as the type of a variable. However it would be an error for that call to result in a value (rather than a type) in that context.

Programmers can define their own enumerated types, classes, records, unions, and type aliases using type declaration statements:

type-declaration-statement:
  enum-declaration-statement
  class-declaration-statement
  record-declaration-statement
  union-declaration-statement
  type-alias-declaration-statement

These statements are defined in Sections Enumerated Types, Class Declarations, Record Declarations, Union Declarations, and Type Aliases, respectively.

Primitive Types

The concrete primitive types are: void, nothing, bool, int, uint, real, imag, complex, string and bytes. They are defined in this section.

In addition, there are several generic primitive types that are described in Built-in Generic Types.

The primitive types are summarized by the following syntax:

primitive-type:
  'void'
  'nothing'
  'bool'
  'int' primitive-type-parameter-part[OPT]
  'uint' primitive-type-parameter-part[OPT]
  'real' primitive-type-parameter-part[OPT]
  'imag' primitive-type-parameter-part[OPT]
  'complex' primitive-type-parameter-part[OPT]
  'string'
  'bytes'
  'enum'
  'record'
  'class'
  'owned'
  'shared'
  'unmanaged'
  'borrowed'

primitive-type-parameter-part:
  ( integer-parameter-expression )

integer-parameter-expression:
  expression

If present, the parenthesized integer-parameter-expression must evaluate to a compile-time constant of integer type. See Compile-Time Constants

Open issue.

There is an expectation of future support for larger bit width primitive types depending on a platform’s native support for those types.

The Void Type

The void type is used to represent the lack of a value. It is primarily used to indicate that a function does not return anything.

Example (returnVoid.chpl).

For example, the below declares f to return void:

proc f() : void { }

The compiler can infer the return type of void as well. See See Return Types for more information.

It is an error to assign the result of a function that returns void to a variable.

The Nothing Type

The nothing type is used to indicate a variable or field that should be removed by the compiler. The value none is the only value of type nothing.

The value none can only be assigned to a variable of type nothing, or to a generic variable that will take on the type nothing. The variable will be removed from the program and have no representation at run-time.

Rationale.

The nothing type can be used to conditionally remove a variable or field from the code based on a param conditional expression.

Example (noneNothing.chpl).

The nothing type and none values typically come up in a generic programming context (see also Generics). For example, the following program defines a generic function g that can determine if it was called with an integer or with none:

proc g(arg) {
  if arg.type != nothing {
    writeln(arg);
  }
}
g(1);    // outputs 1
g(none); // does not create output

The Bool Type

Chapel defines a logical data type designated by the symbol bool with the two predefined values true and false. Values of this boolean type are stored using an implementation-defined number of bits.

Some statements require expressions of bool type and Chapel supports a special conversion of values to bool type when used in this context (Implicit Conversions for Conditionals).

Variables of type bool have a default value of false if they are not initialized to something else (see also Variables).

Example (bools.chpl).

This program demonstrates creating a variable with type bool and setting it to true, and then setting another variable to the logical negation of it:

var x: bool = true;
var y = !x;
var z: bool;

All three variables have type bool. Note that the types of x and y are optional; the program indicates the type of x but the compiler infers the type of y. See Variables for more details. The last variable is initialized to the default value of bool, which is false (see Default Initialization).

Signed and Unsigned Integral Types

The integral types can be parameterized by the number of bits used to represent them. Valid bit-sizes are 8, 16, 32, and 64. The default signed integral type, int, is a synonym for int(64); and the default unsigned integral type, uint, is a synonym for uint(64).

Variables of integral type have a default value of 0 if they are not initialized to something else (see also Variables).

The integral types and their ranges are given in the following table:

Type

Minimum Value

Maximum Value

int(8)

-128

127

uint(8)

0

255

int(16)

-32768

32767

uint(16)

0

65535

int(32)

-2147483648

2147483647

uint(32)

0

4294967295

int(64), int

-9223372036854775808

9223372036854775807

uint(64), uint

0

18446744073709551615

Integer literals such as 3 have type int. However, such literals can implicitly convert to other numeric types that can losslessly store the value. See Implicit Compile-Time Constant Conversions. Integer literals can be written in hexadecimal, octal, or binary. See Literals.

Signed integral types of can implicitly convert to signed integral types of larger width. Additionally, signed integral types can implicitly convert to unsigned integral types of the same or larger width. Unsigned integral types can implicitly convert to both signed and unsigned integral type of larger width. See Implicit Numeric and Bool Conversions for details.

It is possible for overflow to occur with binary operators on integers. For signed integers, overflow leads to undefined behavior. For unsigned integers, overflow leads to wrapping since any bits not representable will be discarded.

Example (integers.chpl).

Here, x is inferred to have type int:

var x = 1;

and y is initialized by converting 2 to a uint(8):

var y:uint(8) = 2;

Then, z is set to an expression that would evaluate to 257, but that is not representable as a uint(8), so it results in the wrapped value 1.

var z = 255 + y;

Real Types

Unlike integral types, real types are floating point types that can store fractional values. Like the integral types, the real types can be parameterized by the number of bits used to represent them. The default real type, real, is 64 bits. The real types that are supported are machine-dependent, but usually include real(32) (single precision) and real(64) (double precision) following the IEEE 754 standard.

Variables of real type have a default value of 0.0 if they are not initialized to something else (see also Variables).

All integral types can implicitly convert to all real types, and real(32) can implicitly convert to real(64). See Implicit Numeric and Bool Conversions for details.

real literals such as 5.2 have type real. However, such literals can implicitly convert to other numeric types that can losslessly store the value. See Implicit Compile-Time Constant Conversions. real literals can be written in decimal or hexadecimal and with or without an exponent (see Literals for details):

  • in decimal without an exponent, e.g. 5.2

  • in decimal with an exponent, e.g. 6.02e23

  • in hexadecimal without an exponent, e.g. 0x2.fe

  • in hexadecimal with a decimal exponent, e.g. 0x2.fep23

    Example (harmonic.chpl).

    For example, this program computes the first n terms of the harmonic series.

    First, it defines a config const to allow setting the value of n on the command line (see Variable Declarations):

    config const n = 100;
    

    Next, it declares a real variable. Since this variable isn’t initialized, it will be initialized to 0.0:

    var sum:real;
    

    Then, it loops over the first n elements and adds them to the sum (see also The For Loop):

    for i in 1..n {
      sum += 1.0/i;
    }
    

    Note that it uses 1.0/i in order to do a floating point division. If it used 1/i, it would do integer division (rounding towards zero), which evaluates to 0 for i > 1.

    Finally, it prints out the sum:

    writeln(sum);
    

Imaginary Types

Imaginary types are floating-point types, and similarly to real types, they can be parameterized by the number of bits used to represent them. The default imaginary type, imag, is 64 bits. The imaginary types that are supported are machine-dependent, but usually include imag(32) and imag(64).

Rationale.

The imaginary type is included to avoid numeric instabilities and under-optimized code stemming from always converting real values to complex values with a zero imaginary part.

Imaginary literals can be created by appending i to a numeric literal; for example, 0.6i. Such literals have type imag. However, such literals can implicitly convert to other numeric types that can losslessly store the value. See Implicit Compile-Time Constant Conversions. As with real literals, imaginary literals can be written in decimal or hexadecimal and with or without an exponent (see Literals for details):

Variables of imag type have a default value of 0.0i if they are not initialized to something else (see also Variables).

It is possible to convert between a real value and an imag value using an explicit cast (see Explicit Conversions). Similarly, an imag value can be cast to a real value. Such casts preserve the floating-point value while changing whether or not it is imaginary.

Example (imaginary.chpl).

For example, this program creates imaginary numbers in two different ways. First, a is an imag variable initialized to a literal:

var a = 0.6i;

Now, suppose we have a real value s:

var s = 10.25;

We can initialize an imag variable with the same numeric value, but as an imaginary value, with a cast:

var b = s:imag;
assert(b == 10.25i);

Complex Types

The complex type represents a complex number. A complex value has floating-point values for the real and imaginary components.

As with the integral and real types, the type complex can be parameterized by the number of bits used to represent the complex number. Since the complex number consists of two components, the number of bits used to represent it is twice the number of bits used to represent each component.

In particular:

  • complex(64) contains two real(32) fields

  • complex(128) contains two real(64) fields

The real and imaginary components can be accessed via the methods re and im. Note that im returns a real of appropriate width, rather than an imag.

Example.

Given a complex number c with the value 3.14+2.72i, the expressions c.re and c.im refer to 3.14 and 2.72 respectively.

proc complex.re ref

When used as a value, this returns the real component of the complex number as a real.

When used as an lvalue, this is a setter that assigns the real component.

proc complex.im ref

When used as a value, this returns the imaginary component of the complex number as a real.

When used as an lvalue, this is a setter that assigns the imaginary component.

The standard Math module provides more functions on complex types. See the Math module documentation.

The String Type

Strings are a primitive type designated by the symbol string comprised of Unicode characters in UTF-8 encoding. Their length is unbounded. Strings are defined in Strings.

The Bytes Type

Bytes is a primitive type designated by the symbol bytes comprised of arbitrary bytes. Bytes are immutable in-place and their length is unbounded. Bytes are defined in Bytes.

Enumerated Types

Enumerated types are declared with the following syntax:

enum-declaration-statement:
  'enum' identifier { enum-constant-list }

enum-constant-list:
  enum-constant
  enum-constant , enum-constant-list[OPT]

enum-constant:
  identifier init-part[OPT]

init-part:
  = expression

The enumerated type can then be referenced by its name, as summarized by the following syntax:

enum-type:
  identifier

An enumerated type defines a set of named constants that can be referred to via a member access on the enumerated type. Each enumerated type is a distinct type.

If the init-part is omitted for all of the named constants in an enumerated type, the enumerated values are abstract and do not have associated integer values. Any constant that has an init-part will be associated with that integer value. Such constants must be parameter values of integral type. Any constant that does not have an init-part, yet which follows one that does, will be associated with an integer value one greater than its predecessor. An enumerated type whose first constant has an init-part is called concrete, since all constants in the enum will have an associated integer value, whether explicit or implicit. An enumerated type that specifies an init-part for some constants, but not the first is called semi-concrete. Numeric conversions are automatically supported for enumerated types which are concrete or semi-concrete (see Explicit Enumeration Conversions).

Example (enum-statesmen.chpl).

The code

enum statesman { Aristotle, Roosevelt, Churchill, Kissinger }

defines an abstract enumerated type with four constants. The function

proc quote(s: statesman) {
  select s {
    when statesman.Aristotle do
       writeln("All paid jobs absorb and degrade the mind.");
    when statesman.Roosevelt do
       writeln("Every reform movement has a lunatic fringe.");
    when statesman.Churchill do
       writeln("A joke is a very serious thing.");
    when statesman.Kissinger do
       { write("No one will ever win the battle of the sexes; ");
         writeln("there's too much fraternizing with the enemy."); }
  }
}

outputs a quote from the given statesman. Note that enumerated constants must be prefixed by the enumerated type name and a dot unless a use statement is employed (see The Use Statement and Using Modules).

It is possible to iterate over an enumerated type. The loop body will be invoked on each named constant in the enum. The following method is also available:

proc enum.size : param int

Returns the number of constants in the given enumerated type.

proc enum.first : enum

Returns the first constant in the enumerated type.

proc enum.last : enum

Returns the last constant in the enumerated type.

Structured Types

The structured types are summarized by the following syntax:

structured-type:
  class-type
  record-type
  union-type
  tuple-type

Classes are discussed in Classes. Records are discussed in Records. Unions are discussed in Unions. Tuples are discussed in Tuples.

Class Types

A class can contain variables, constants, and methods.

Classes are defined in Classes. The class type can also contain type aliases and parameters. Such a class is generic and is defined in Generic Types.

A class type C has several variants:

  • C and C?

  • owned C and owned C?

  • shared C and shared C?

  • borrowed C and borrowed C?

  • unmanaged C and unmanaged C?

The variants with a question mark, such as owned C?, can store nil (see Nilable Class Types). Variants without a question mark cannot store nil. The keywords owned, shared, borrowed, and unmanaged indicate the memory management strategy used for the class. When none is specified, as with C or C?, the class is considered to have generic memory management strategy. See Class Types.

Record Types

Records can contain variables, constants, and methods. Unlike class types, records are values rather than references. Records are defined in Records.

Union Types

The union type defines a type that contains one of a set of variables. Like classes and records, unions may also define methods. Unions are defined in Unions.

Tuple Types

A tuple is a light-weight record that consists of one or more anonymous fields. If all the fields are of the same type, the tuple is homogeneous. Tuples are defined in Tuples.

Data Parallel Types

The data parallel types are summarized by the following syntax:

dataparallel-type:
  range-type
  domain-type
  mapped-domain-type
  array-type
  index-type

Ranges and their index types are discussed in Ranges. Domains and their index types are discussed in Domains. Arrays are discussed in Arrays.

Range Types

A range defines an integral sequence of some integral type. Ranges are defined in Ranges.

Domain, Array, and Index Types

A domain defines a set of indices. An array defines a set of elements that correspond to the indices in its domain. A domain’s indices can be of any type. Domains, arrays, and their index types are defined in Domains and Arrays.

Synchronization Types

The synchronization types are summarized by the following syntax:

synchronization-type:
  sync-type
  atomic-type

The sync type is discussed in Synchronization Variables. The atomic type is discussed in Atomic Variables.

Type Aliases

Type aliases are declared with the following syntax:

type-alias-declaration-statement:
  privacy-specifier[OPT] 'config'[OPT] 'type' type-alias-declaration-list ;
  external-type-alias-declaration-statement

type-alias-declaration-list:
  type-alias-declaration
  type-alias-declaration , type-alias-declaration-list

type-alias-declaration:
  identifier = type-expression
  identifier

A type alias is a symbol that aliases the type specified in the type-expression. A use of a type alias has the same meaning as using the type specified by type-expression directly.

Type aliases defined at the module level are public by default. The optional privacy-specifier keywords are provided to specify or change this behavior. For more details on the visibility of symbols, see  Visibility Of A Module’s Symbols.

If the keyword config precedes the keyword type, the type alias is called a configuration type alias. Configuration type aliases can be set at compilation time via compilation flags or other implementation-defined means. The type-expression in the program is ignored if the type-alias is alternatively set.

If the keyword extern precedes the type keyword, the type alias is external. The declared type name is used by Chapel for type resolution, but no type alias is generated by the backend. See the chapter on interoperability (Interoperability) for more information on external types.

The type-expression is optional in the definition of a class or record. Such a type alias is called an unspecified type alias. Classes and records that contain type aliases, specified or unspecified, are generic (Type Aliases in Generic Types).

Example (type-alias.chpl).

The declaration

type t = int;

defines a t as a synonym for the type int. Functions and methods available on int will apply to variables declared with type t. For example,

var x: t = 1;
x += 1;
writeln(x);

will print out 2.

Querying the Type of an Expression

type-query-expression:
  expression . 'type'

The type of a an expression can be queried with .type. This functionality is particularly useful when doing generic programming (see Generics).

Example (dot-type.chpl).

For example, this code uses .type to query the type of the variable x and store that in the type alias t:

var x: int;
type t = x.type;

Open issue.

Given a nested expression that has .type called on it, for example f() in f().type, in which circumstances should f() be evaluated for side effects?

At first it might seem that f() should never be evaluated for side effects. However, it must be evaluated for side effects if f() returns an array or domain type, as these have a runtime component (see Types with Runtime Components). As a result, should f() in such a setting always be evaluated for side effects? The answer to this question also also connected to the question of whether or not a when a function returning a type is evaluated for side effects at runtime.

One approach might be to introduce different means to query only the compile-time component of the type or only the runtime component of the time.

Operations Available on Types

This section discusses how type expressions can be used. Type expressions include types, type aliases, .type queries, and calls to functions that use the type return intent.

A type expression can be used to indicate the type of a value, as with var x: typeExpression; (see Variable Declarations).

A type expression can be passed to a type formal of a generic function (see Formal Type Arguments).

The Types module provides many functions to query properties of types.

The language provides isCoercible, isSubtype, and isProperSubtype for comparing types. The normal comparison operators are also available to compare types:

  • == checks if two types are equivalent

  • != checks if two types are different

  • < and > check if one type is a proper subtype of another (see <)

  • <= and >= check if one type is a subtype of another (see <=)

It is possible to cast a type to a param string. This allows a type to be printed out.

Example (type-to-string.chpl).

For example, this code casts the type myType to a string in order to print it out:

type myType = int;
param str = myType:string;
writeln(str);

It produces the output:

int(64)

Open issue.

If type comparison with == is called on two types with runtime components (see Types with Runtime Components), should the runtime component be included in the comparison? Or, should == on types only consider if the compile-time components match?

Types with Runtime Components

Domain and array types include a runtime component. (See Domains and Arrays for more on arrays and domains).

For a domain type, the runtime component of the type is the distribution over which the domain was declared.

For an array type, the runtime component of the type contains the domain over which the array was declared and the runtime component of the array’s element type, if present.

As a result, an array or domain type will be represented and manipulated at runtime. In particular, a function that returns a type with a runtime component will be executed at runtime.

These features combine with the .type syntax to allow one to create an array that has the same element type, shape, and distribution as an existing array.

Example (same-domain-array.chpl).

The example below shows a function that accepts an array and then creates another array with the same element type, shape, and distribution:

proc makeAnotherArray(arr: []) {
  var newArray: arr.type;
  return newArray;
}

The above program is equivalent to this program:

proc equivalentAlternative(arr: []) {
  var newArray:[arr.domain] arr.eltType;
  return newArray;
}

Both create and return an array storing the same element type as the passed array.

Open issue.

Should a record or class type also have a runtime component when it contains array/domain field(s)? This runtime component is needed, for example, to create a default-initialized instance of such a type in the absence of user-defined default initializer.

Open issue.

Class types are not currently considered to have a runtime component. Should class types be considered to have a runtime component, so that querying an instance’s type with myObject.type will produce the type of the object known at runtime, rather than the type with which myObject was declared?

Open issue.

Should functions returning a type always be evaluated for side effects, or only evaluated for side effects when returning a type with a runtime component?