5.1 Defining Structure Types: struct
Programmer-Defined Datatypes in The Racket Guide introduces struct.
syntax
(struct id maybe-super (field ...) struct-option ...)
maybe-super =
| super-id field = field-id | [field-id field-option ...] struct-option = #:mutable | #:super super-expr | #:inspector inspector-expr | #:auto-value auto-expr | #:guard guard-expr | #:property prop-expr val-expr | #:transparent | #:prefab | #:authentic | #:name name-id | #:extra-name name-id | #:constructor-name constructor-id | #:extra-constructor-name constructor-id | #:reflection-name symbol-expr | #:methods gen:name-id method-defs | #:omit-define-syntaxes | #:omit-define-values field-option = #:mutable | #:auto method-defs = (definition ...)
A struct form with n fields defines up to 4+2n names:
struct:id, a structure type descriptor value that represents the structure type.
constructor-id (which defaults to id), a constructor procedure that takes m arguments and returns a new instance of the structure type, where m is the number of fields that do not include an #:auto option.
name-id (which defaults to id), a transformer binding that encapsulates information about the structure type declaration. This binding is used to define subtypes, and it also works with the shared and match forms. For detailed information about the binding of name-id, see Structure Type Transformer Binding.
The constructor-id and name-id can be the same, in which case name-id performs both roles. In that case, the expansion of name-id as an expression produces an otherwise inaccessible identifier that is bound to the constructor procedure; the expanded identifier has a 'constructor-for property whose value is an identifier that is free-identifier=? to name-id as well as a syntax property accessible via syntax-procedure-alias-property with an identifier that is free-identifier=? to name-id.
id?, a predicate procedure that returns #t for instances of the structure type (constructed by constructor-id or the constructor for a subtype) and #f for any other value.
id-field-id, for each field; an accessor procedure that takes an instance of the structure type and extracts the value for the corresponding field.
set-id-field-id!, for each field that includes a #:mutable option, or when the #:mutable option is specified as a struct-option; a mutator procedure that takes an instance of the structure type and a new field value. The structure is destructively updated with the new value, and #<void> is returned.
If super-id is provided, it must have a transformer binding of the same sort bound to name-id (see Structure Type Transformer Binding), and it specifies a supertype for the structure type. Alternately, the #:super option can be used to specify an expression that must produce a structure type descriptor. See Structures for more information on structure subtypes and supertypes. If both super-id and #:super are provided, a syntax error is reported.
> (struct document (author title content)) > (struct book document (publisher)) > (struct paper (journal) #:super struct:document)
If the #:mutable option is specified for an individual field, then the field can be mutated in instances of the structure type, and a mutator procedure is bound. Supplying #:mutable as a struct-option is the same as supplying it for all fields. If #:mutable is specified as both a field-option and struct-option, a syntax error is reported.
> (struct cell ([content #:mutable]) #:transparent) > (define a-cell (cell 0)) > (set-cell-content! a-cell 1)
The #:inspector, #:auto-value, and #:guard options specify an inspector, value for automatic fields, and guard procedure, respectively. See make-struct-type for more information on these attributes of a structure type. The #:property option, which is the only one that can be supplied multiple times, attaches a property value to the structure type; see Structure Type Properties for more information on properties. The #:transparent option is a shorthand for #:inspector #f.
> (struct point (x y) #:inspector #f) > (point 3 5) (point 3 5)
> (struct celsius (temp) #:guard (λ (temp name) (unless (and (real? temp) (>= temp -273.15)) (error "not a valid temperature")) temp)) > (celsius -275) not a valid temperature
Use the prop:procedure property to implement an applicable structure, use prop:evt to create a structure type whose instances are synchronizable events, and so on. By convention, property names start with prop:.
The #:prefab option obtains a prefab (pre-defined, globally shared) structure type, as opposed to creating a new structure type. Such a structure type is inherently transparent and cannot have a guard or properties, so using #:prefab with #:transparent, #:inspector, #:guard, #:property, #:authentic, or #:methods is a syntax error. If a supertype is specified, it must also be a prefab structure type.
> (struct prefab-point (x y) #:prefab) > (prefab-point 1 2) '#s(prefab-point 1 2)
> (prefab-point? #s(prefab-point 1 2)) #t
The #:authentic option is a shorthand for #:property prop:authentic #t, which prevents instances of the structure type from being impersonated (see impersonate-struct), chaperoned (see chaperone-struct), or acquiring a non-flat contract (see struct/c). See prop:authentic for more information. If a supertype is specified, it must also have the prop:authentic property.
If name-id is supplied via #:extra-name and it is not id, then both name-id and id are bound to information about the structure type. Only one of #:extra-name and #:name can be provided within a struct form, and #:extra-name cannot be combined with #:omit-define-syntaxes.
> (struct ghost (color name) #:prefab #:extra-name GHOST)
> (match (ghost 'red 'blinky) [(GHOST c n) c]) 'red
If constructor-id is supplied, then the transformer binding of name-id records constructor-id as the constructor binding; as a result, for example, struct-out includes constructor-id as an export. If constructor-id is supplied via #:extra-constructor-name and it is not id, applying object-name on the constructor produces the symbolic form of id rather than constructor-id. If constructor-id is supplied via #:constructor-name and it is not the same as name-id, then name-id does not serve as a constructor, and object-name on the constructor produces the symbolic form of constructor-id. Only one of #:extra-constructor-name and #:constructor-name can be provided within a struct form.
> (struct color (r g b) #:constructor-name -color) > (struct rectangle (w h color) #:extra-constructor-name rect) > (rectangle 13 50 (-color 192 157 235)) #<rectangle>
> (rect 50 37 (-color 35 183 252)) #<rectangle>
If #:reflection-name symbol-expr is provided, then symbol-expr must produce a symbol that is used to identify the structure type in reflective operations such as struct-type-info. It corresponds to the first argument of make-struct-type. Structure printing uses the reflective name, as do the various procedures that are bound by struct.
> (struct circle (radius) #:reflection-name '<circle>) > (circle 15) #<<circle>>
> (circle-radius "bad") <circle>-radius: contract violation
expected: <circle>?
given: "bad"
If #:methods gen:name-id method-defs is provided, then gen:name-id must be a transformer binding for the static information about a generic interface produced by define-generics. The method-defs define the methods of the gen:name-id interface. A define/generic form or auxiliary definitions and expressions may also appear in method-defs.
> (struct constant-stream (val) #:methods gen:stream [(define (stream-empty? stream) #f) (define (stream-first stream) (constant-stream-val stream)) (define (stream-rest stream) stream)]) > (stream-ref (constant-stream 'forever) 0) 'forever
> (stream-ref (constant-stream 'forever) 50) 'forever
If the #:omit-define-syntaxes option is supplied, then name-id (and id, if #:extra-name is specified) is not bound as a transformer. If the #:omit-define-values option is supplied, then none of the usual variables are bound, but id is bound. If both are supplied, then the struct form is equivalent to (begin).
> (struct square (side) #:omit-define-syntaxes)
> (match (square 5) ; fails to match because syntax is omitted [(struct square x) x]) eval:28:0: match: square does not refer to a structure
definition
at: square
in: (struct square x)
> (struct ellipse (width height) #:omit-define-values) > ellipse-width ellipse-width: undefined;
cannot reference an identifier before its definition
in module: top-level
Expressions supplied to #:auto-value are evaluated once and shared between every instance of the structure type. In particular, updates to a mutable #:auto-value affect all current and future instances.
If a field includes the #:auto option, then all fields after it must also include #:auto, otherwise a syntax error is reported. If any field-option or struct-option keyword is repeated, other than #:property, a syntax error is reported.
(struct posn (x y [z #:auto #:mutable]) #:auto-value 0 #:transparent)
> (posn 1 2) (posn 1 2 0)
> (posn? (posn 1 2)) #t
> (posn-y (posn 1 2)) 2
> (posn-z (posn 1 2)) 0
(struct color-posn posn (hue) #:mutable) (define cp (color-posn 1 2 "blue"))
> (color-posn-hue cp) "blue"
> cp (color-posn 1 2 0 ...)
> (set-posn-z! cp 3)
For serialization, see define-serializable-struct.
Changed in version 6.9.0.4 of package base: Added #:authentic.
syntax
(struct-field-index field-id)
(struct mood-procedure (base rating) #:property prop:procedure (struct-field-index base)) (define happy+ (mood-procedure add1 10))
> (happy+ 2) 3
> (mood-procedure-rating happy+) 10
syntax
(define-struct id-maybe-super (field ...) struct-option ...)
id-maybe-super = id | (id super-id)
This form is provided for backwards compatibility; struct is preferred.
(define-struct posn (x y [z #:auto]) #:auto-value 0 #:transparent)
> (make-posn 1 2) (posn 1 2 0)
> (posn? (make-posn 1 2)) #t
> (posn-y (make-posn 1 2)) 2
syntax
(struct/derived (id . rest-form) id (field ...) struct-option ...)
(struct/derived (id . rest-form) id super-id (field ...) struct-option ...)
(define-syntax (fruit-struct stx) (syntax-case stx () [(ds name . rest) (with-syntax ([orig stx]) #'(struct/derived orig name (seeds color) . rest))]))
> (fruit-struct apple) > (apple-seeds (apple 12 "red")) 12
> (fruit-struct apple #:mutable) > (set-apple-seeds! (apple 12 "red") 8) ; this next line will cause an error due to a bad keyword > (fruit-struct apple #:bad-option) eval:54:0: fruit-struct: unrecognized struct-specification
keyword
at: #:bad-option
in: (fruit-struct apple #:bad-option)
Added in version 7.5.0.16 of package base.
syntax
(define-struct/derived (id . rest-form) id-maybe-super (field ...) struct-option ...)
(define-syntax (define-xy-struct stx) (syntax-case stx () [(ds name . rest) (with-syntax ([orig stx]) #'(define-struct/derived orig name (x y) . rest))]))
> (define-xy-struct posn) > (posn-x (make-posn 1 2)) 1
> (define-xy-struct posn #:mutable) > (set-posn-x! (make-posn 1 2) 0) ; this next line will cause an error due to a bad keyword > (define-xy-struct posn #:bad-option) eval:60:0: define-xy-struct: unrecognized
struct-specification keyword
at: #:bad-option
in: (define-xy-struct posn #:bad-option)
Changed in version 7.5.0.16 of package base: Moved main description to struct/derived and replaced with differences.