3.2 Importing and Exporting: require and provide🔗ℹ

Imports: require in The Racket Guide introduces require.

A require-spec designates a particular set of identifiers to be bound in the importing context. Each identifier is mapped to a particular export of a particular module; the identifier to bind may be different from the symbolic name of the originally exported identifier. Each identifier also binds at a particular phase level and in a binding space.

No identifier can be bound multiple times in a given combination of phase level and binding space by an import, unless all of the bindings refer to the same original definition in the same module. In a module context, an identifier can be either imported or defined for a given phase level and binding space, but not both.

The syntax of require-spec can be extended via define-require-syntax, and when multiple require-specs are specified in a require, the bindings of each require-spec are visible for expanding later require-specs. The pre-defined forms (as exported by racket/base) are as follows:

module-path

Imports all exported bindings from the named module, using the export name for the local identifiers. (See below for information on module-path.) The lexical context of the module-path form determines the context of the introduced identifiers, adding a space scope for exports in a particular binding space, and in each export’s phase level.

If any identifier provided by module-path has a symbol form that is uninterned, the identifier is not imported (i.e., it is impossible to import a binding for an uninterned symbol). This restriction is intended to avoid compilation differences depending on whether a module has been saved to a file or not (see Printing Compiled Code).

syntax (only-in require-spec id-maybe-renamed ...)

Like require-spec, but constrained to those exports for which the identifiers to bind match id-maybe-renamed: as id or as orig-id in [orig-id bind-id]. If the id or orig-id of any id-maybe-renamed is not in the set that require-spec describes, a syntax error is reported.

Examples:

> (require (only-in racket/tcp                     tcp-listen                     [tcp-accept my-accept]))

> tcp-listen

#<procedure:tcp-listen>

> my-accept

#<procedure:tcp-accept>

> tcp-accept

tcp-accept: undefined;

cannot reference an identifier before its definition

in module: top-level

syntax (except-in require-spec id ...)

Like require-spec, but omitting those imports for which ids are the identifiers to bind; if any id is not in the set that require-spec describes, a syntax error is reported.

Examples:

> (require (except-in racket/tcp                       tcp-listen))

> tcp-accept

#<procedure:tcp-accept>

> tcp-listen

tcp-listen: undefined;

cannot reference an identifier before its definition

in module: top-level

syntax (prefix-in prefix-id require-spec)

Like require-spec, but adjusting each identifier to be bound by prefixing it with prefix-id. The lexical context of the prefix-id is ignored, and instead preserved from the identifiers before prefixing.

Examples:

> (require (prefix-in tcp: racket/tcp))

> tcp:tcp-accept

#<procedure:tcp-accept>

> tcp:tcp-listen

#<procedure:tcp-listen>

A syntax property with the key 'import-or-export-prefix-ranges is added to the local identifier in the expanded form of require.

Changed in version 8.9.0.5 of package base: Added the 'import-or-export-prefix-ranges syntax property.

syntax (rename-in require-spec [orig-id bind-id] ...)

Like require-spec, but replacing the identifier to bind orig-id with bind-id; if any orig-id is not in the set that require-spec describes, a syntax error is reported.

Examples:

> (require (rename-in racket/tcp                       (tcp-accept accept)                       (tcp-listen listen)))

> accept

#<procedure:tcp-accept>

> listen

#<procedure:tcp-listen>

syntax (combine-in require-spec ...)

The union of the require-specs. If two or more imports from the require-specs have the same identifier name but they do not refer to the same original binding, a syntax error is reported.

Examples:

> (require (combine-in (only-in racket/tcp tcp-accept)                        (only-in racket/tcp tcp-listen)))

> tcp-accept

#<procedure:tcp-accept>

> tcp-listen

#<procedure:tcp-listen>

syntax (relative-in module-path require-spec ...)

Like the union of the require-specs, but each relative module path in a require-spec is treated as relative to module-path instead of the enclosing context.

The require transformer that implements relative-in sets current-require-module-path to adjust module paths in the require-specs.

syntax (only-meta-in phase-level require-spec ...)

Like the combination of require-specs, but removing any binding that is not for phase-level, where #f for phase-level corresponds to the label phase level.

The following example imports bindings only at phase level 1, the transform phase:

> (module nest racket     (provide (for-syntax meta-eggs)              (for-meta 1 meta-chicks)              num-eggs)     (define-for-syntax meta-eggs 2)     (define-for-syntax meta-chicks 3)     (define num-eggs 2))

> (require (only-meta-in 1 'nest))

> (define-syntax (desc stx)     (printf "~s ~s\n" meta-eggs meta-chicks)     #'(void))

> (desc)

2 3

> num-eggs

num-eggs: undefined;

cannot reference an identifier before its definition

in module: top-level

The following example imports only bindings at phase level 0, the normal phase.

> (require (only-meta-in 0 'nest))

> num-eggs

2

syntax (only-space-in space require-spec ...)

Like the combination of require-specs, but removing any binding that is not provided for the binding space identifier by space—which is normally an identifier, but #f for space corresponds to the default binding space.

Added in version 8.2.0.3 of package base.

(for-meta phase-level require-spec ...)

Like the combination of require-specs, but the bindings specified by each require-spec are shifted by phase-level. The label phase level corresponds to #f, and a shifting combination that involves #f produces #f.

Examples:

> (module nest racket     (provide num-eggs)     (define num-eggs 2))

> (require (for-meta 0 'nest))

> num-eggs

2

> (require (for-meta 1 'nest))

> (define-syntax (roost stx)     (datum->syntax stx num-eggs))

> (roost)

2

(for-syntax require-spec ...)

Same as (for-meta 1 require-spec ...).

(for-template require-spec ...)

Same as (for-meta -1 require-spec ...).

(for-label require-spec ...)

Same as (for-meta #f require-spec ...). If an identifier in any of the require-specs is bound at more than one phase level, a syntax error is reported.

(for-space space require-spec ...)

Like the combination of require-specs, but the bindings specified by each require-spec are moved to the binding space specified by space—which is normally an identifier, but #f for space corresponds to the default binding space.

A binding is moved to the new space by removing the scope for the space originally implied by require-spec, if any, and adding the scope for space, if any.

Added in version 8.2.0.3 of package base.

derived-require-spec

See define-require-syntax for information on expanding the set of require-spec forms.

Module Paths in The Racket Guide introduces module paths.

A module-path identifies a module, either a root module or a submodule that is declared lexically within another module. A root module is identified either through a concrete name in the form of an identifier, or through an indirect name that can trigger automatic loading of the module declaration. Except for the (quote id) case below, the actual resolution of a root module path is up to the current module name resolver (see current-module-name-resolver), and the description below corresponds to the default module name resolver.

(quote id)

Refers to a submodule previously declared with the name id or a module previously declared interactively with the name id. When id refers to a submodule, (quote id) is equivalent to (submod "." id).

Examples:

; a module declared interactively as test:

> (require 'test)

rel-string

A path relative to the containing source (as determined by current-load-relative-directory or current-directory). Regardless of the current platform, rel-string is always parsed as a Unix-format relative path: / is the path delimiter (multiple adjacent /s are not allowed), .. accesses the parent directory, and . accesses the current directory. The path cannot be empty or contain a leading or trailing slash, path elements before than the last one cannot include a file suffix (i.e., a . in an element other than . or ..), and the only allowed characters are ASCII letters, ASCII digits, -, +, _, ., /, and %. Furthermore, a % is allowed only when followed by two lowercase hexadecimal digits, and the digits must form a number that is not the ASCII value of a letter, digit, -, +, or _.

The % provision is intended to support a one-to-one encoding of arbitrary strings as path elements (after UTF-8 encoding). Such encodings are not decoded to arrive at a filename, but instead preserved in the file access.

If rel-string ends with a ".ss" suffix, it is converted to a ".rkt" suffix. The compiled-load handler may reverse that conversion if a ".rkt" file does not exist and a ".ss" exists.

Examples:

; a module named "x.rkt" in the same

; directory as the enclosing module's file:

> (require "x.rkt")

; a module named "x.rkt" in the parent directory

; of the enclosing module file's directory:

> (require "../x.rkt")

syntax (lib rel-string ...+)

A path to a module installed into a collection (see Libraries and Collections). The rel-strings in lib are constrained similar to the plain rel-string case, with the additional constraint that a rel-string cannot contain . or .. directory indicators.

The specific interpretation of the path depends on the number and shape of the rel-strings:

• If a single rel-string is provided, and if it consists of a single element (i.e., no /) with no file suffix (i.e., no .), then rel-string names a collection, and "main.rkt" is the library file name.

  Examples:

  ; the main swindle library:

  > (require (lib "swindle"))

  ; the same:

  > (require (lib "swindle/main.rkt"))

• If a single rel-string is provided, and if it consists of multiple /-separated elements, then each element up to the last names a collection, subcollection, etc., and the last element names a file. If the last element has no file suffix, ".rkt" is added, while a ".ss" suffix is converted to ".rkt".

  Examples:

  ; "turbo.rkt" from the "swindle" collection:

  > (require (lib "swindle/turbo"))

  ; the same:

  > (require (lib "swindle/turbo.rkt"))

  ; the same:

  > (require (lib "swindle/turbo.ss"))

• If a single rel-string is provided, and if it consists of a single element with a file suffix (i.e, with a .), then rel-string names a file within the "mzlib" collection. A ".ss" suffix is converted to ".rkt". (This convention is for compatibility with older version of Racket.)

  Examples:

  ; "tar.rkt" module from the "mzlib" collection:

  > (require (lib "tar.ss"))

• Otherwise, when multiple rel-strings are provided, the first rel-string is effectively moved after the others, and all rel-strings are appended with / separators. The resulting path names a collection, then subcollection, etc., ending with a file name. No suffix is added automatically, but a ".ss" suffix is converted to ".rkt". (This convention is for compatibility with older version of Racket.)

  Examples:

  ; "tar.rkt" module from the "mzlib" collection:

  > (require (lib "tar.ss" "mzlib"))

id

A shorthand for a lib form with a single rel-string whose characters are the same as in the symbolic form of id. In addition to the constraints of a lib rel-string, id must not contain ..

Example:

> (require racket/tcp)

syntax (file string)

Similar to the plain rel-string case, but string is a path—possibly absolute—using the current platform’s path conventions and expand-user-path. A ".ss" suffix is converted to ".rkt".

Example:

> (require (file "~/tmp/x.rkt"))

syntax (planet id)

(planet string)

(planet rel-string (user-string pkg-string vers)         rel-string ...)

Specifies a library available via the PLaneT server.

The first form is a shorthand for the last one, where the id’s character sequence must match the following ‹spec› grammar:

	‹spec›	::=	‹owner› / ‹pkg› ‹lib›
	‹owner›	::=	‹elem›
	‹pkg›	::=	‹elem›  |  ‹elem› : ‹version›
	‹version›	::=	‹int›  |  ‹int› : ‹minor›
	‹minor›	::=	‹int›  |  <= ‹int›  |  >= ‹int›  |  = ‹int›
		|	‹int› - ‹int›
	‹lib›	::=	‹empty›  |  / ‹path›
	‹path›	::=	‹elem›  |  ‹elem› / ‹path›

and where an ‹elem› is a non-empty sequence of characters that are ASCII letters, ASCII digits, -, +, _, or % followed by lowercase hexadecimal digits (that do not encode one of the other allowed characters), and an ‹int› is a non-empty sequence of ASCII digits. As this shorthand is expended, a ".plt" extension is added to ‹pkg›, and a ".rkt" extension is added to ‹path›; if no ‹path› is included, "main.rkt" is used in the expansion.

A (planet string) form is like a (planet id) form with the identifier converted to a string, except that the string can optionally end with a file extension (i.e., a .) for a ‹path›. A ".ss" file extension is converted to ".rkt".

In the more general last form of a planet module path, the rel-strings are similar to the lib form, except that the (user-string pkg-string vers) names a PLaneT-based package instead of a collection. A version specification can include an optional major and minor version, where the minor version can be a specific number or a constraint: (nat nat) specifies an inclusive range, (= nat) specifies an exact match, (+ nat) specifies a minimum version and is equivalent to just nat, and (- nat) specifies a maximum version. The =, +, and - identifiers in a minor-version constraint are recognized symbolically.

Examples:

; "main.rkt" in package "farm" by "mcdonald":

> (require (planet mcdonald/farm))

; "main.rkt" in version >= 2.0 of "farm" by "mcdonald":

> (require (planet mcdonald/farm:2))

; "main.rkt" in version >= 2.5 of "farm" by "mcdonald":

> (require (planet mcdonald/farm:2:5))

; "duck.rkt" in version >= 2.5 of "farm" by "mcdonald":

> (require (planet mcdonald/farm:2:5/duck))

syntax (submod root-module-path submod-path-element ...)

(submod "." submod-path-element ...)

(submod ".." submod-path-element ...)

Identifies a submodule within the module specified by root-module-path or relative to the current module in the case of (submod "." ....), where (submod ".." submod-path-element ...) is equivalent to (submod "." ".." submod-path-element ...). Submodules have symbolic names, and a sequence of identifiers as submod-path-elements determine a path of successively nested submodules with the given names. A ".." as a submod-path-element names the enclosing module of a submodule, and it’s intended for use in (submod "." ....) and (submod ".." ....) forms.

As require prepares to handle a sequence of require-specs, it logs a “prefetch” message to the current logger at the 'info level, using the name 'module-prefetch, and including message data that is a list of two elements: a list of module paths that appear to be imported, and a directory path to use for relative module paths. The logged list of module paths may be incomplete, but a compilation manager can use approximate prefetch information to start on compilations in parallel.

Changed in version 6.0.1.10 of package base: Added prefetch logging.

Exports: provide in The Racket Guide introduces provide.

A provide-spec indicates one or more bindings to provide. For each exported binding, the external name is a symbol that can be different from the symbolic form of the identifier that is bound within the module. Also, each export is drawn from a particular phase level and exported at the same phase level; by default, the relevant phase level is the number of begin-for-syntax forms that enclose the provide form. Finally, each export is drawn from a binding space and exported at the same binding space.

The syntax of provide-spec can be extended by bindings to provide transformers or provide pre-transformers, such as via define-provide-syntax, but the pre-defined forms are as follows.

id

Exports id, which must be bound within the module (i.e., either defined or imported) at the relevant phase level and binding space. The symbolic form of id is used as the external name, and the symbolic form of the defined or imported identifier must match (otherwise, the external name could be ambiguous).

Examples:

> (module nest racket     (provide num-eggs)     (define num-eggs 2))

> (require 'nest)

> num-eggs

2

If id has a transformer binding to a rename transformer, then the transformer affects the exported binding. See make-rename-transformer for more information.

syntax (all-defined-out)

Exports all identifiers that are defined at the relevant phase level within the exporting module, and that have the same lexical context as the (all-defined-out) form, excluding bindings to rename transformers where the target identifier has the 'not-provide-all-defined syntax property. The external name for each identifier is the symbolic form of the identifier. Only identifiers accessible from the lexical context of the (all-defined-out) form are included; that is, macro-introduced imports are not re-exported, unless the (all-defined-out) form was introduced at the same time.

Examples:

> (module nest racket     (provide (all-defined-out))     (define num-eggs 2))

> (require 'nest)

> num-eggs

2

syntax (all-from-out module-path ...)

Exports all identifiers that are imported into the exporting module using a require-spec built on each module-path (see Importing and Exporting: require and provide) with no phase-level shift. The symbolic name for export is derived from the name that is bound within the module, as opposed to the symbolic name of the export from each module-path. Only identifiers accessible from the lexical context of the module-path are included; that is, macro-introduced imports are not re-exported, unless the module-path was introduced at the same time.

Examples:

> (module nest racket     (provide num-eggs)     (define num-eggs 2))

> (module hen-house racket     (require 'nest)     (provide (all-from-out 'nest)))

> (require 'hen-house)

> num-eggs

2

syntax (rename-out [orig-id export-id] ...)

Exports each orig-id, which must be bound within the module at the relevant phase level and binding space. The symbolic name for each export is export-id instead of orig-id.

Examples:

> (module nest racket     (provide (rename-out [count num-eggs]))     (define count 2))

> (require 'nest)

> num-eggs

2

> count

count: undefined;

cannot reference an identifier before its definition

in module: top-level

syntax (except-out provide-spec provide-spec ...)

Like the first provide-spec, but omitting the bindings listed in each subsequent provide-spec. If one of the latter bindings is not included in the initial provide-spec, a syntax error is reported. The symbolic export name information in the latter provide-specs is ignored; only the bindings are used.

Examples:

> (module nest racket     (provide (except-out (all-defined-out)                          num-chicks))     (define num-eggs 2)     (define num-chicks 3))

> (require 'nest)

> num-eggs

2

> num-chicks

num-chicks: undefined;

cannot reference an identifier before its definition

in module: top-level

syntax (prefix-out prefix-id provide-spec)

Like provide-spec, but with each symbolic export name from provide-spec prefixed with prefix-id.

Examples:

> (module nest racket     (provide (prefix-out chicken: num-eggs))     (define num-eggs 2))

> (require 'nest)

> chicken:num-eggs

2

A syntax property with the key 'import-or-export-prefix-ranges is added to the exported identifier in the expanded form of provide.

Changed in version 8.9.0.5 of package base: Added the 'import-or-export-prefix-ranges syntax property.

syntax (struct-out id)

Exports the bindings associated with a structure type id. Typically, id is bound with (struct id ....); more generally, id must have a transformer binding of structure-type information at the relevant phase level; see Structure Type Transformer Binding. Furthermore, for each identifier mentioned in the structure-type information, the enclosing module must define or import one identifier that is free-identifier=?. If the structure-type information includes a super-type identifier, and if the identifier has a transformer binding of structure-type information, the accessor and mutator bindings of the super-type are not included by struct-out for export.

Examples:

> (module nest racket     (provide (struct-out egg))     (struct egg (color wt)))

> (require 'nest)

> (egg-color (egg 'blue 10))

'blue

syntax (combine-out provide-spec ...)

The union of the provide-specs.

Examples:

> (module nest racket     (provide (combine-out num-eggs num-chicks))     (define num-eggs 2)     (define num-chicks 1))

> (require 'nest)

> num-eggs

2

> num-chicks

1

syntax (protect-out provide-spec ...)

Like the union of the provide-specs, except that the exports are protected: requiring modules may refer to these bindings, but may not extract these bindings from macro expansions or access them via eval without access privileges. For more details, see Code Inspectors. The provide-spec must specify only bindings that are defined within the exporting module.

Examples:

> (module nest racket     (provide num-eggs (protect-out num-chicks))     (define num-eggs 2)     (define num-chicks 3))

> (define weak-inspector (make-inspector (current-code-inspector)))

> (define (weak-eval x)     (parameterize ([current-code-inspector weak-inspector])       (define weak-ns (make-base-namespace))       (namespace-attach-module (current-namespace)                                ''nest                                weak-ns)       (parameterize ([current-namespace weak-ns])         (namespace-require ''nest)         (eval x))))

> (require 'nest)

> (list num-eggs num-chicks)

'(2 3)

> (weak-eval 'num-eggs)

2

> (weak-eval 'num-chicks)

?: access disallowed by code inspector to protected variable

from module: 'nest

at: num-chicks

See also Code Inspectors for Trusted and Untrusted Code.

(for-meta phase-level provide-spec ...)

Like the union of the provide-specs, but adjusted to apply to the phase level specified by phase-level relative to the current phase level (where #f corresponds to the label phase level). In particular, an id or rename-out form as a provide-spec refers to a binding at phase-level relative to the current level, an all-defined-out exports only definitions at phase-level relative to the current phase level, and an all-from-out exports bindings imported with a shift by phase-level.

Examples:

> (module nest racket     (begin-for-syntax      (define eggs 2))     (define chickens 3)     (provide (for-syntax eggs)              chickens))

> (require 'nest)

> (define-syntax (test-eggs stx)     (printf "Eggs are ~a\n" eggs)     #'0)

> (test-eggs)

Eggs are 2

0

> chickens

3

> (module broken-nest racket     (define eggs 2)     (define chickens 3)     (provide (for-syntax eggs)              chickens))

eval:7:0: provide: provided identifier is not defined or

required

at: eggs

in: (provide (for-syntax eggs) chickens)

> (module nest2 racket     (begin-for-syntax      (define eggs 2))     (provide (for-syntax eggs)))

> (require (for-meta 2 racket/base)            (for-syntax 'nest2))

> (define-syntax (test stx)     (define-syntax (show-eggs stx)       (printf "Eggs are ~a\n" eggs)       #'0)     (begin       (show-eggs)       #'0))

Eggs are 2

> (test)

0

(for-syntax provide-spec ...)

Same as (for-meta 1 provide-spec ...).

(for-template provide-spec ...)

Same as (for-meta -1 provide-spec ...).

(for-label provide-spec ...)

Same as (for-meta #f provide-spec ...).

(for-space space provide-spec ...)

Like the union of the provide-specs, but adjusted to apply to the binding space specified by space—where space is either an identifier or #f for the default binding space. In particular, an id or rename-out form as a provide-spec refers to a binding in space, an all-defined-out exports only definitions in space, and an all-from-out exports bindings imported into space.

When providing a binding for a non-default binding space, normally a module should also provide a binding for the default binding space, where the default-space binding represents the intended meaning of the identifier. When a module later imports the same name in different spaces from modules that adhere to this convention, then if the two modules also (re)export the same binding for the name in the default space, the imports are likely consistent. If the two modules export different bindings for the name in the default space, then attempting to import both modules will trigger an error about conflicting imports, and a programmer can explicitly resolve the mismatch.

Added in version 8.2.0.3 of package base.

derived-provide-spec

See define-provide-syntax for information on expanding the set of provide-spec forms.

Each export specified within a module must have a distinct symbolic export name, though the same binding can be specified with the multiple symbolic names.

• a just-meta form cannot appear within a just-meta form;

• a for-meta, for-syntax, for-template, or for-label form cannot appear within a for-meta, for-syntax, for-template, or for-label form; and

• a for-space form cannot appear within a for-space form.

• a portal form cannot appear within a just-meta form.

Except for the portal form, each raw-require-spec corresponds to the obvious require-spec, but the rename sub-form has the identifiers in reverse order compared to rename-in.

For most raw-require-specs, the lexical context of the raw-require-spec determines the context of introduced identifiers. The exception is the rename sub-form, where the lexical context of the local-id is preserved.

A literal-path as a raw-root-module-path corresponds to a path in the sense of path?. Since path values are never produced by read-syntax, they appear only in programmatically constructed expressions. They also appear naturally as arguments to functions such as namespace-require, with otherwise take a quoted raw-module-spec.

The portal form provides a way to define portal syntax at any phase level. A (portal portal-id content), defines portal-id to portal syntax with content effectively quoted to serve as its content.

Changed in version 8.2.0.3 of package base: Added for-space and just-space.
Changed in version 8.3.0.8: Added portal.

Like #%require, the sub-form keywords for #%provide are recognized symbolically, and nearly every raw-provide-spec has an obvious equivalent provide-spec via provide, with the exception of the struct and expand sub-forms.

A (struct struct-id (field-id ...)) sub-form expands to struct-id, make-struct-id, struct:struct-id, struct-id?, struct-id-field-id for each field-id, and set-struct-id-field-id! for each field-id. The lexical context of the struct-id is used for all generated identifiers.

Unlike #%require, the #%provide form is macro-extensible via an explicit expand sub-form; the (id . datum) part is locally expanded as an expression (even though it is not actually an expression), stopping when a begin form is produced; if the expansion result is (begin raw-provide-spec ...), it is spliced in place of the expand form, otherwise a syntax error is reported. If an orig-form part is provided, then it is used instead of the #%provide form when raising syntax errors, such as a “provide identifier is not defined” error. The expand sub-form is not normally used directly; it provides a hook for implementing provide and provide transformers.

The all-from and all-from-except forms re-export only identifiers that are accessible in lexical context of the all-from or all-from-except form itself. That is, macro-introduced imports are not re-exported, unless the all-from or all-from-except form was introduced at the same time. Similarly, all-defined and its variants export only definitions accessible from the lexical context of the spaceless-spec form.

Changed in version 8.2.0.3 of package base: Added for-space.
Changed in version 8.2.0.5: Added orig-form support to expand.

3.2.1 Additional require Forms🔗ℹ

The following forms support more complex selection and manipulation of sets of imported identifiers.