Modules

View modules.chpl on GitHub

This primer introduces the concept of modules, a concept for encapsulating code for use by other code. It covers:

  • how to define a module

  • namespace control within a module

  • access to another module’s symbols

  • namespace control when using a module, including:

    • unlimited

    • explicit exclusion of symbols

    • explicit inclusion of symbols

    • renaming included symbols

  • cross-file module access

A module is a grouping of code - each program consists of at least one module, and every symbol is associated with some module. If all the code of a file is not enclosed in an explicit module, defined using the module keyword, then the file itself is treated as a module with the same name as the file (minus the .chpl suffix). The compiler can be directed to include modules in distinct files by naming them on the command line or by relying on the -M flag (see the man page for exact details). Here, we declare a module ModToUse:

module ModToUse {

In this case, foo is a public module-level variable that is defined within the module ModToUse

var foo = 12;

As is bar.

var bar: int = 2;

A symbol can be declared private - this means that only code defined within the same scope as the definition of the symbol (including code in nested scopes) can access it.

Here, hiddenFoo is a private module-level variable, making it only accessible to other code contained in ModToUse

private var hiddenFoo = false;

baz is a public function which is defined within ModToUse

proc baz (x, y) {
  return x * (x + y);
}

hiddenBaz is a private function, which is also only accessible by symbols contained in ModToUse.

private proc hiddenBaz(a) {
  writeln(a);
  return a + 3;
}

Rec is a module-level record, with a field and a method defined inside it.

record Rec {
  var field: int;

  proc method1 () {
    writeln("In Rec.method1()");
  }
}

method2 is a secondary method defined on Rec.

  proc Rec.method2() {
    writeln("In Rec.method2()");
  }

} // end of ModToUse module

In the current implementation, private cannot be applied to type definitions; type aliases, and declarations of enums, records, and classes cannot be declared private. Private also cannot be applied to fields or methods yet.

Here are some other modules which will be used in the remainder of this file

module AnotherModule {
  var a = false;
}

module ThirdModule {
  var b = -13.0;
}


module Conflict {

This variable shares a name with a symbol in ModToUse.

  var bar = 5;

  var other = 5.0 + 3i;

  var another = false;
} // end of Conflict module


module DifferentArguments {

This function shares a name with a function in ModToUse, but takes different arguments

  proc baz(x) {
    return x - 2;
  }
} // end of DifferentArguments module

module RecMoreMethods {
  private use ModToUse;

method3 is a tertiary method defined on Rec

  proc Rec.method3() {
    writeln("In Rec.method3()");
  }
} // end of RecMoreMethods module

module MainModule {
  proc main() {
    writeln("Access from outside a module");

Access From Outside a Module

In multi-module programs, it is common for modules to access the contents of other modules. The starting point for doing so is the use statement or the import statement in Chapel. These statements can be inserted at any lexical scope that contains executable code.

By default, a use statement makes all of a specific module’s visible symbols available to the scope that contains that use statement. These symbols may then be accessed directly in an unqualified manner (without a module name prefix).

In this case, bazBarFoo should store the result of calling ModToUse.baz on ModToUse.bar and ModToUse.foo, which is in this case 28.

{
  use ModToUse;

  var bazBarFoo = baz(bar, foo);
  writeln(bazBarFoo);

}

Since use statements only affect their containing scope, when we leave a scope like this, we lose access to the module’s symbols. For instance, since the following line isn’t within a scope that contains a use of ModToUse, it would generate an error if uncommented. This is because foo is not visible within our lexical scope or via any use statements in that scope.

// var twiceFoo = 2 * foo;

An import statement, in contrast to a use statement, either enables qualified access to the visible symbols of a module (e.g. with the module’s name as a prefix) or enables unqualified access to a specified subset of those symbols (e.g. without the module’s name as a prefix).

This demonstrates an import that enables access with the module prefix. In this example, bazBarFoo should store the result of calling ModToUse.baz on ModToUse.bar and ModToUse.foo, which is in this case 28.

{
  import ModToUse;

  var bazBarFoo = ModToUse.baz(ModToUse.bar, ModToUse.foo);
  writeln(bazBarFoo);
}

This case demonstrates an import that enables access without the module prefix to a single symbol within ModToUse.

{
  import ModToUse.bar;

  writeln(bar);
}

import statements also only affect their containing scope, so similarly the following line would generate an error if uncommented, due to being outside a scope with an import or use of ModToUse.

// var twiceFoo = 2 * ModToUse.foo;

use statements apply to the entire scope in which they are defined. Even if the use statement occurs after code which would directly refer to its symbols, these accesses are still valid. This is similar to other Chapel forms of introducing symbols - for instance, function declaration order does not prevent a function declared earlier in a scope from calling one declared later.

Thus, as in an earlier example, the following declaration of bazBarFoo will store the result of calling ModToUse.baz on ModToUse.bar and ModToUse.foo, which is again 28.

{
  var bazBarFoo = baz(bar, foo);

  use ModToUse;

  writeln(bazBarFoo);
}

Similarly, import statements also apply to the entire scope in which they are defined. Even if the import statement occurs after code which would directly refer to its symbols, these accesses are still valid.

Thus, the output of this block is the same as that of its use statement counterpart.

{
  var bazBarFoo = ModToUse.baz(ModToUse.bar, ModToUse.foo);

  import ModToUse;

  writeln(bazBarFoo);
}

Modules that are being used can be given a different name than the ones they were declared with. In this example, ModToUse.bar can be accessed using the new name, other, as the module prefix. The old name is not visible in that scope, though, so writing just ModToUse.bar will not work. However, unprefixed access to the module’s symbols remains unchanged.

{
  use ModToUse as other;

  writeln(other.bar);
  // writeln(ModToUse.bar); // would be an error, ModToUse not visible
  writeln(bar);
}

The same is also true of modules that are imported.

{
  import ModToUse as other;

  writeln(other.bar);
  // writeln(ModToUse.bar); // would be an error, ModToUse not visible
}

With this syntax, a use statement can enable just unqualified access to a symbol, similar to how the unqualified list syntax for import statements doesn’t enable qualified access.

{
  use ModToUse as _;

  writeln(bar);
  // writeln(ModToUse.bar); // would be an error, ModToUse not visible
  // writeln(_.bar); // also an error, _ is not an identifier otherwise
}

Variables provided by a private use statement are only considered when the name in question cannot be resolved directly within the local scope. Thus, because another bar is defined within this scope, the access to bar within the writeln will refer to the local variable bar rather than to ModToUse.bar.

{
  var bar = 4.0;

  use ModToUse;

  writeln(bar);
  // Will output the value of the bar defined in this scope (which is
  // '4.0'), rather than the value of ModToUse.bar (which is '2')
}

In contrast, defining a variable with the same name as something imported or brought in by public use leads to a multiple definition error.

{
  import ModToUse.bar;
  // var bar = 4.0; multiple definition error
}

If a symbol cannot be resolved directly within the local scope, then the symbols provided by a private use / `use statements are considered before the symbols defined outside of the scope where the use statement applies. Thus, because the other bar was defined outside of these curly braces, the compiler will find the bar from ModToUse when resolving the access within the writeln, rather than the outer bar. The bar from ModToUse is said to be “shadowing” the definition at the outer scope.

{
  var bar = false;
  {

    use ModToUse;
    writeln(bar);
    // Will output the value of ModToUse.bar (which is '2'), rather
    // than the value of the bar defined outside of this scope (which
    // is 'false')
  }
}

The same is also true of symbols provided by an import statement.

{
  var bar = false;
  {

    import ModToUse.bar;
    writeln(bar);
    // Will output the value of ModToUse.bar (which is '2'), rather
    // than the value of the bar defined outside of this scope (which
    // is 'false')
  }
}

Multiple modules may be named in a single use statement

{
  use ModToUse, AnotherModule, ThirdModule;

  if (a || b < 0) {
    // Refers to AnotherModule.a (which is 'false') and ThirdModule.b (which
    // is '-13.0')
    writeln(foo); // Refers to ModToUse.foo
  } else {
    writeln(bar); // Refers to ModToUse.bar
  } // Will output ModToUse.foo (which is '12')
}

Multiple modules may also be named in a single import statement

{
  import ModToUse, AnotherModule, ThirdModule;

  if (AnotherModule.a || ThirdModule.b < 0) {
    writeln(ModToUse.foo);
  } else {
    writeln(ModToUse.bar);
  } // Will output ModToUse.foo (which is '12')
}

And such import statements can mix and match between importing modules and the symbols within them.

{
  import ModToUse.{foo, bar}, AnotherModule, ThirdModule.b;

  if (AnotherModule.a || b < 0) {
    // Refers to ThirdModule.b (which is '-13.0')
    writeln(foo); // Refers to ModToUse.foo
  } else {
    writeln(bar); // Refers to ModToUse.bar
  } // Will output ModToUse.foo (which is '12')
}

You can also import multiple modules in a single statement by naming a shared parent module. We will talk about this more in Nested Modules.

Equivalently, a scope may contain multiple use statements

{
  use ModToUse;
  use AnotherModule, ThirdModule;

  writeln(a && foo > 15);
  // outputs false (because AnotherModule.a is 'false' and ModToUse.foo is
  // '12')
}

A scope may also contain multiple import statements

{
  import ModToUse.foo;
  import AnotherModule.a;

  writeln(a && foo > 15);
  // outputs false (because AnotherModule.a is 'false' and ModToUse.foo is
  // '12')
}

It can even contain a mix of import and use statements

{
  use ModToUse;
  import AnotherModule.a;

  writeln(a && foo > 15);
  // outputs false (because AnotherModule.a is 'false' and ModToUse.foo is
  // '12')
}

In any case, the modules accessed in this way are considered in concert (after symbols defined at this scope but before symbols defined outside of it) - the ordering within a use statement or across multiple use or import statements does not affect the precedence of symbols that share a name. This means that if two modules each define a symbol with the same name, and both modules are brought in at the same scope, attempts to access a symbol by that name will result in a naming conflict.

The commented-out line below would fail because both ModToUse and Conflict define a symbol named bar:

{
  use ModToUse, Conflict;

  writeln(foo); // Outputs ModToUse.foo ('12')
  // writeln(bar);
  writeln(other); // Outputs Conflict.other ('5.0 + 3.0i')
}

When the symbol being accessed is the name of a function, the rules become more complex. If the two function definitions are overloads (or define different arguments), then the best match will be found, no matter where the function is defined relative to the other function definitions.

More details on when overloading applies, when functions may shadow other functions, etc. can be found in the relevant section of the language specification. They will not be covered further in this primer.

Finally, the names of the modules themselves are made available by a use statement at a scope just outside of the modules’ contents and just inside the next lexical scope surrounding the current one.

{

  use ModToUse, DifferentArguments;

  writeln(baz(2, 3));
  // Accesses the function ModToUse.baz using the two arguments.  Should
  // output 2 * (2 + 3) or '10'
  writeln(baz(3));
  // Access the function DifferentArguments.baz using the single argument.
  // Should output 3 - 2, or '1'
}

Remember: this is in contrast to import statements, where only the module name or specific symbols within the module are brought in, rather than both. Also remember that you can write a use statement that doesn’t enable accesses to the module’s name, by renaming the module to ‘_’.

Limiting a Use

To get around such conflicts, there are multiple strategies. If only a small number of symbols are desired from a particular module, you can specify the symbols to bring in via an only list.

Here, because of the only clause in the use of Conflict, Conflict’s bar is not directly accessible here.

writeln();
writeln("Limiting a use");


{
  use ModToUse;
  use Conflict only other, another;

  writeln(foo); // Outputs ModToUse.foo ('12')
  writeln(bar); // Outputs ModToUse.bar ('2')
  writeln(other); // Outputs Conflict.other ('5.0 + 3.0i')
}

import statements for unqualified access are somewhat similar to only lists on use statements, though the syntax is different. We saw how to import a single symbol for unqualified access earlier, so this example demonstrates how to import multiple symbols for unqualified access.

{
  use ModToUse;
  import Conflict.{other, another};

  writeln(foo); // Outputs ModToUse.foo ('12')
  writeln(bar); // Outputs ModToUse.bar ('2')
  writeln(other); // Outputs Conflict.other ('5.0 + 3.0i')
}

Using an except list on a use statement will cause every symbol other than the ones listed to be available.

{
  use Conflict;
  use ModToUse except bar;

  writeln(foo); // Outputs ModToUse.foo ('12')
  writeln(bar); // Outputs Conflict.bar ('5')
  writeln(other); // Outputs Conflict.other ('5.0 + 3.0i')
}

import statements do not have an equivalent to except lists.

If both symbols which conflict are desired, or if the use causes symbols to be shadowed which are necessary, you can choose to rename a symbol when including it via the as keyword, so long as the new name does not cause any conflicts with other included symbols.

{
  use ModToUse;
  use Conflict only bar as boop;
  writeln(bar); // Outputs ModToUse.bar ('2')
  writeln(boop); // Outputs Conflict.bar ('5')
}

Similarly, import statements allow renaming symbols within the curly braces.

{
  use ModToUse;
  import Conflict.{bar as boop};
  writeln(bar); // Outputs ModToUse.bar ('2')
  writeln(boop); // Outputs Conflict.bar ('5')
}

You can also use a module without making any symbols available in an unqualified manner using an empty identifier list after only. This form is typically used by programmers who prefer to always fully qualify accesses to their modules’ symbols.

{
  use ModToUse only;
  use Conflict only;
  writeln(ModToUse.bar);  // Outputs ModToUse.bar ('2')
  writeln(Conflict.bar);  // Outputs Conflict.bar ('5')
  // writeln(bar);        // this won't resolve since bar isn't available
}

Again, this is similar to an import of just the module itself.

{
  import ModToUse;
  import Conflict;
  writeln(ModToUse.bar);  // Outputs ModToUse.bar ('2')
  writeln(Conflict.bar);  // Outputs Conflict.bar ('5')
  // writeln(bar);        // this won't resolve since bar isn't available
}

Class and record instances obtained in scopes where their type is not otherwise visible can still access any fields and any methods defined in their type’s original scope.

To impact the visibility of methods defined in modules other than where the type was defined, known as “tertiary methods”, the type itself can be listed in an only or except list for use statements, or as one of the symbols listed in an import statement.

{
  use ModToUse only;
  var rec = new ModToUse.Rec(4); // Only accessible via the module prefix
  writeln(rec.field);            // Accessible because we have an instance
  rec.method1();                 // Ditto to the field case
  rec.method2();

  use RecMoreMethods only Rec;
  rec.method3();                 // Enabled by previous use statement
}

writeln();

Application to Enums

use statements can also be called on enums. Normally to access one of an enum’s constants, you must provide a prefix of the enum name. With a use of that enum, such a prefix is no longer necessary.

writeln("Application to enums");

{

  enum color {red, blue, yellow};

  {
    // Normally you must prefix the constant with the name of the enum
    var aColor = color.blue;
    writeln(aColor);
  }

  {
    use color;

    // The 'use' statement allows you to access an enum's symbols without
    // the prefix
    var anotherColor = yellow; // color.yellow
    writeln(anotherColor);
  }

}

writeln();

All of the above rules for using modules also apply to using enums.

import statements, on the other hand, do not apply to enums any more than they do to other non-module symbols - an enum can be listed for unqualified access, but doing so will not enable unqualified access to the enum’s constants.

Nested Modules

A use of a nested module (see the module OuterNested and its submodules for an example of a nested module) is similar to that of a top-level module. Its name is treated like any other visible symbol in the outer module, so if the outer module has not been used then the inner module must be explicitly named.

{
  use OuterNested.Inner1;

  writeln(foobar); // Will output Inner1.foobar, or '14'
}

Similarly, in order to import a nested module, you must provide the explicit path to that module.

{
  import OuterNested.Inner1;

  writeln(Inner1.foobar); // Will output 14
}

While import statements cannot list multiple modules in the same way that use statements do, when a common parent is provided multiple submodules can be listed.

    {
      import OuterNested.{Inner1, Inner2};

      writeln(Inner1.foobar); // Will output 14
      writeln(Inner2.canSeeHidden); // Will output true
    }
  } // end of main() function
} // end of MainModule module

We’ll use the module OuterNested to demonstrate some more details of nested modules.

module OuterNested {
  var foo = 12;
  var bar: int = 2;
  private var hiddenFoo = false;

  proc baz (x, y) {
    return x * (x + y);
  }

  private proc hiddenBaz(a) {
    writeln(a);
    return a + 3;
  }

A module defined within another module is called a nested module. These submodules cannot refer to symbols defined within their parent module, without a use or import of the parent module. The parent module can only access the contents of the nested module using a use or import statement or a fully qualified name.

The variable foobar accesses OuterNested’s foo and bar variables.

module Inner1 {
  use OuterNested;

  var foobar = foo + bar;
}

Since the module Inner2 is defined within OuterNested, it can access the private variable hiddenFoo and the private function hiddenBaz. However, any private symbol defined within Inner2 will not be visible within scopes defined outside of Inner2.

module Inner2 {
  use OuterNested;

  private var innerOnly = -17;
  var canSeeHidden = !hiddenFoo;
}

module Inner3 {
  var x: int = 11;
}

Parent modules can use their child modules by specifying the full path to them.

{
  writeln("Executing OuterNested's module-level code");
  use OuterNested.Inner3;

  writeln(x);
}

But they can also use the child modules with just the name itself, since it is in scope.

{
  use Inner3;

  writeln(x);
}

In contrast, import statements cannot import submodules with just the name itself.

{
  import OuterNested.Inner3;
  //import Inner3; // Will not work

  writeln(Inner3.x);
}

However, both use and import statements can utilize this as a prefix, to avoid having to provide the full path.

{
  import this.Inner3;

  writeln(Inner3.x);
}

module Inner4 {

Child modules can also utilize super as a prefix, allowing access to other symbols defined on their parent module.

    import super.Inner3;

    writeln(Inner3.x);
  }

  // These lines are to ensure everything gets tested regularly
  writeln("End of OuterNested's module-level code");
  {
    use UsesTheUser;
  }
  writeln("End of reverse file-order output");
  writeln();
} // end of OuterNested module

Public vs. Private Uses and Imports

Use statements can be labeled as being either public or private. By default use statements are private. This means that if one module uses a second, it will only see the symbols defined by that module, not by other modules that it happens to use. For example, consider the following library module:

module ModuleThatIsUsed {
  proc publiclyAvailableProc() {
    writeln("This function is accessible!");
  }
}

And a module that uses it:

module UserModule {
  use ModuleThatIsUsed;  // or `private use ModuleThatIsUsed`
}

When a scope has a use of UserModule, the symbols from ModuleThatIsUsed will not be available due to the private nature of UserModule ‘s use, so the following code would not compile.

module UsesTheUser {
  proc func1() {
    use UserModule;
    // publiclyAvailableProc(); // Won't compile, since ``UserModule``'s ``use`` is ``private``
  }

  // These lines are to ensure everything gets tested regularly
  writeln("Start of UsesTheUser's module-level code");
  func1();
  {
    use UsesTheUser2;
  }
  writeln("End of UsesTheUser's module-level code");
}

By contrast, a public use will permit symbols used by one module to be seen by those that use it. However, it does not provide the name of the module public use’d (but note that it is possible to opt in to that with public import or with renaming the module with as). For example, consider the following variation of the previous example:

module UserModule2 {
  public use ModuleThatIsUsed;
}

Since its use is public, a scope with a use of UserModule2 will also be able to see the symbols defined by ModuleThatIsUsed.

module UsesTheUser2 {
  proc func2() {
    use UserModule2;
    publiclyAvailableProc(); // available due to ``use`` of ``ModuleThatIsUsed``
  }

  // These lines are to ensure everything gets tested regularly
  writeln("Start of UsesTheUser2's module-level code");
  func2();
  {
    use UsesTheUser3;
  }
  writeln("End of UsesTheUser2's module-level code");
}

In addition, public use statements re-export the symbols in the used module. This means that the used symbols can also be treated as though they are defined in the scope with the use for the purpose of accesses from outside that module. For example, here it will also appear as though the module ModuleThatIsUsed is a submodule of UserModule2, so other scopes can also access it in that manner.

module UsesTheUser3 {
  proc func3() {
    use UserModule2;
    UserModule2.publiclyAvailableProc();
    // The above is available due to the ``public use`` of ``ModuleThatIsUsed``
    // in ``UserModule2``.
  }

  // These lines are to ensure everything gets tested regularly
  writeln("Start of UsesTheUser3's module-level code");
  func3();
  {
    use UsesTheImporter;
  }
  writeln("End of UsesTheUser3's module-level code");

}

Import statements can also be labeled as being either public or private. By default import statements are also private. However, public import statements have a different meaning than public use statements do - while public import statements still re-export the symbols in the imported module, the impact into the scope is more limited. This is because import statements do not enable both qualified and unqualified access to the same module in the same statement.

Again, this means that the imported symbols will be treated as though they are defined in the scope with the import for the purpose of accesses from outside that module. For example:

module ImporterModule {
  public import ModuleThatIsUsed;
}

Here, it will appear as though the module ModuleThatIsUsed is a submodule of ImporterModule, so other scopes can access it in that manner.

module UsesTheImporter {
  use ImporterModule;

  // Possible due to re-export of ModuleThatIsUsed
  ModuleThatIsUsed.publiclyAvailableProc();

  // These lines are to ensure everything gets tested regularly
  {
    use NoMiddleMan;
  }
}

This doesn’t prevent ModuleThatIsUsed from being available through its original means.

module NoMiddleMan {
  use ModuleThatIsUsed;

  publiclyAvailableProc();

  // These lines are to ensure everything gets tested regularly
  {
    use UsesTheImporter2;
  }
}

The same is true of module-level symbols brought in by public import statements. In this example, the public import of the function defined in ModuleThatIsUsed makes it appear as though ImporterModule2 defined the function.

module ImporterModule2 {
  public import ModuleThatIsUsed.publiclyAvailableProc;
}

module UsesTheImporter2 {
  use ImporterModule2;

  writeln("Start of reverse file-order output");
  // Possible due to re-export of ModuleThatIsUsed.publiclyAvailableProc
  ImporterModule2.publiclyAvailableProc();
}