;;; Written by Eli Barzilay: Maze is Life! (eli@barzilay.org) ;;> The `base' module defines some basic low-level syntactic extensions to ;;> MzScheme. It can be used by itself to get these extensions. (module base mzscheme (provide (all-from-except mzscheme #%module-begin #%top #%app define let let* letrec lambda keyword? keyword->string string->keyword)) ;;>> (#%module-begin ...) ;;> `base' is a language module -- it redefines `#%module-begin' to load ;;> itself for syntax definitions. (provide (rename module-begin~ #%module-begin)) (define-syntax (module-begin~ stx) (let ([e (if (syntax? stx) (syntax-e stx) stx)]) (if (pair? e) (datum->syntax-object (quote-syntax here) (list* (quote-syntax #%plain-module-begin) (datum->syntax-object stx (list (quote-syntax require-for-syntax) '(lib "base.ss" "swindle"))) (cdr e)) stx) (raise-syntax-error #f "bad syntax" stx))) ;; This doesn't work anymore (from 203.4) ;; (syntax-rules () ;; [(_ . body) (#%plain-module-begin ;; (require-for-syntax (lib "base.ss" "swindle")) . body)]) ) ;;>> (#%top . id) ;;> This special syntax is redefined to make keywords (symbols whose names ;;> begin with a ":") evaluate to themselves. (provide (rename top~ #%top)) (define-syntax (top~ stx) (syntax-case stx () [(_ . x) (let ([s (syntax-e #'x)]) (if (and (symbol? s) (not (eq? s '||)) (eq? #\: (string-ref (symbol->string s) 0))) (syntax/loc stx (#%datum . x)) (syntax/loc stx (#%top . x))))])) ;;>> (#%app ...) ;;> Redefined so it is possible to apply using dot notation: `(foo x . y)' ;;> is the same as `(apply foo x y)'. This is possible only when the last ;;> (dotted) element is an identifier. (provide (rename app~ #%app)) (define-syntax (app~ stx) (syntax-case stx () [(_ x ...) (syntax/loc stx (#%app x ...))] [(_ . x) (let loop ([s (syntax-e #'x)] [r '()]) (cond [(list? s) (syntax/loc stx (#%app . x))] [(pair? s) (loop (cdr s) (cons (car s) r))] [else (let ([e (and (syntax? s) (syntax-e s))]) (if (or (null? e) (pair? e)) (loop e r) (quasisyntax/loc stx (#%app apply . #,(reverse! (cons s r))))))]))])) ;; these are defined as normal bindings so code that uses this module can use ;; them, but for the syntax level of this module we need them too. (define-for-syntax (keyword*? x) (and (symbol? x) (not (eq? x '||)) (eq? (string-ref (symbol->string x) 0) #\:))) (define-for-syntax (syntax-keyword? x) (keyword*? (if (syntax? x) (syntax-e x) x))) ;;>> (define id-or-list ...) ;;> The standard `define' form is modified so defining :keywords is ;;> forbidden, and if a list is used instead of an identifier name for a ;;> function then a curried function is defined. ;;> => (define (((plus x) y) z) (+ x y z)) ;;> => plus ;;> # ;;> => (plus 5) ;;> # ;;> => ((plus 5) 6) ;;> # ;;> => (((plus 5) 6) 7) ;;> 18 ;;> Note the names of intermediate functions. ;;> ;;> In addition, the following form can be used to define multiple values: ;;> => (define (values a b) (values 1 2)) (provide (rename define~ define)) (define-syntax (define~ stx) ;; simple version ;; (syntax-case stx () ;; [(_ (name arg ...) body ...) ;; #`(define~ name (lambda~ (arg ...) body ...))] ;; [(_ name body ...) #'(define name body ...)]) ;; this version makes created closures have meaningful names ;; also -- forbid using :keyword identifiers ;; also -- make (define (values ...) ...) a shortcut for define-values (this ;; is just a patch, a full solution should override `define-values', and ;; also deal with `let...' and `let...-values' and lambda binders) ;; also -- if the syntax is top-level, then translate all defines into a ;; define with (void) followed by a set! -- this is for the problem of ;; defining something that is provided by some module, and re-binding a ;; syntax (define top-level? (eq? 'top-level (syntax-local-context))) (syntax-case* stx (values) ;; compare symbols if at the top-level (if top-level? (lambda (x y) (eq? (syntax-e x) (syntax-e y))) module-identifier=?) [(_ name expr) (identifier? #'name) (cond [(syntax-keyword? #'name) (raise-syntax-error #f "cannot redefine a keyword" stx #'name)] [top-level? #'(begin (define-values (name) (void)) (set! name expr))] [else #'(define-values (name) expr)])] [(_ (values name ...) expr) (cond [(ormap (lambda (id) (and (syntax-keyword? id) id)) (syntax->list #'(name ...))) => (lambda (id) (raise-syntax-error #f "cannot redefine a keyword" stx id))] [top-level? #'(begin (define name (void)) ... (set!-values (name ...) expr))] [else #'(define-values (name ...) expr)])] [(_ names body0 body ...) (pair? (syntax-e #'names)) (let loop ([s #'names] [args '()]) (syntax-case s () [(name . arg) (loop #'name (cons #'arg args))] [name (let ([sym (syntax-object->datum #'name)]) (let loop ([i (sub1 (length args))] [as (reverse (cdr args))] [body #'(begin body0 body ...)]) (if (zero? i) (cond [(syntax-keyword? #'name) (raise-syntax-error #f "cannot redefine a keyword" stx #'name)] [top-level? (quasisyntax/loc stx (begin (define name (void)) (set! name (lambda~ #,(car args) #,body))))] [else (quasisyntax/loc stx (define name (lambda~ #,(car args) #,body)))]) (loop (sub1 i) (cdr as) (syntax-property (quasisyntax/loc stx (lambda~ #,(car as) #,body)) 'inferred-name (string->symbol (format "~a:~a" sym i)))))))]))])) ;;>> (let ([id-or-list ...] ...) ...) ;;>> (let* ([id-or-list ...] ...) ...) ;;>> (letrec ([id-or-list ...] ...) ...) ;;> All standard forms of `let' are redefined so they can generate ;;> functions using the same shortcut that `define' allows. This includes ;;> the above extension to the standard `define'. For example: ;;> => (let ([((f x) y) (+ x y)]) ((f 1) 2)) ;;> 3 ;;> It also includes the `values' keyword in a similar way to `define'. ;;> For example: ;;> => (let ([(values i o) (make-pipe)]) i) ;;> # (provide (rename let~ let) (rename let*~ let*) (rename letrec~ letrec)) (define-syntaxes (let~ let*~ letrec~) (let* ([process (lambda (stx var0 val0 . flat?) (syntax-case var0 (values) [(values var ...) (null? flat?) #`((var ...) . #,val0)] [_ (let loop ([var var0] [args '()]) (if (identifier? var) (if (null? args) (let ([val (syntax->list val0)]) (if (and (pair? val) (null? (cdr val))) (list (if (null? flat?) (list var) var) (car val)) (raise-syntax-error #f "bad binding" stx #`(#,var0 #,@val0)))) (let ([sym (syntax-e var)]) (let loop ([i (sub1 (length args))] [as (reverse args)] [val val0]) (if (< i 0) (list (if (null? flat?) (list var) var) (car (syntax->list val))) (loop (sub1 i) (cdr as) (let ([val #`((lambda~ #,(car as) #,@val))]) (if (zero? i) val (syntax-property val 'inferred-name (if (zero? i) sym (string->symbol (format "~a:~a" sym i))))))))))) (syntax-case var () [(var . args1) (loop #'var (cons #'args1 args))])))]))] [mk-bindings (lambda (stx bindings . flat?) (syntax-case bindings () [((var val more ...) ...) (datum->syntax-object #'bindings (map (lambda (x y) (apply process stx x y flat?)) (syntax->list #'(var ...)) (syntax->list #'((val more ...) ...))) #'bindings)]))] [mk-let (lambda (tag . lbl) (lambda (stx) (syntax-case stx () [(_ label bindings body0 body ...) (and (identifier? #'label) (pair? lbl)) (quasisyntax/loc stx (#,(car lbl) label #,(mk-bindings stx #'bindings #t) body0 body ...))] [(_ bindings body0 body ...) (quasisyntax/loc stx (#,tag #,(mk-bindings stx #'bindings) body0 body ...))])))]) (values (mk-let #'let-values #'let) (mk-let #'let*-values) (mk-let #'letrec-values)))) ;;>> (lambda formals body ...) ;;> The standard `lambda' is extended with Lisp-like &-keywords in its ;;> argument list. This extension is available using the above short ;;> syntax. There is one important difference between these keywords and ;;> Lisp: some &-keywords are used to access arguments that follow the ;;> keyword part of the arguments. This makes it possible to write ;;> procedures that can be invoked as follows: ;;> (f ) ;;> (Note: do not use more keywords after the !) ;;> ;;> Available &-keywords are: (provide (rename lambda~ lambda)) (define-syntax (lambda~ stx) (define (process-optional-arg o) (syntax-case o () [(var default) (identifier? #'var) (list #'var #'default)] [(var) (identifier? #'var) (list #'var #'#f)] [var (identifier? #'var) (list #'var #'#f)] [var (raise-syntax-error #f "not a valid &optional spec" stx #'var)])) (define (process-keyword-arg k) (define (key var) (datum->syntax-object k (string->symbol (string-append ":" (symbol->string (syntax-object->datum var)))) k k)) (syntax-case k () [(var key default) (and (identifier? #'var) (syntax-keyword? #'key)) (list #'var #'key #'default)] [(var default) (identifier? #'var) (list #'var (key #'var) #'default)] [(var) (identifier? #'var) (list #'var (key #'var) #'#f)] [var (identifier? #'var) (list #'var (key #'var) #'#f)] [var (raise-syntax-error #f "not a valid &key spec" stx #'var)])) (syntax-case stx () [(_ formals expr0 expr ...) (let ([vars '()] [opts '()] [keys '()] [rest #f] ; keys and all (no optionals) [rest-keys #f] ; like the above, minus specified keys [body #f] ; stuff that follows all keywords [all-keys #f] ; all keys, excluding body [other-keys #f]) ; unprocessed keys, excluding body ;; relations: ;; rest = (append all-keys body) ;; rest-keys = (append other-keys body) (let loop ([state #f] [args #'formals]) (syntax-case args () [() #f] [(v . xs) (let* ([v #'v] [k (if (symbol? v) v (and (identifier? v) (syntax-e v)))] [x (and k (symbol->string k))]) (cond ;; check &-keywords according to their name, so something like ;; (let ([&rest 1]) (lambda (&rest r) ...)) ;; works as expected [(and x (> (string-length x) 0) (eq? #\& (string-ref x 0))) (case k ;;> * &optional, &opt, &opts: denote an optional argument, possibly with a ;;> default value (if the variable is specified as `(var val)'). ;;> => ((lambda (x &optional y [z 3]) (list x y z)) 1) ;;> (1 #f 3) ;;> => ((lambda (x &optional y [z 3]) (list x y z)) 1 2 #f) ;;> (1 2 #f) [(&optional &optionals &opt &opts) (if state (raise-syntax-error #f "misplaced &optional argument" stx #'formals) (loop 'o #'xs))] ;;> * &keys, &key: a keyword argument -- the variable should be specified ;;> as `x' or `(x)' to be initialized by an `:x' keyword, `(x v)' to ;;> specify a default value `v', and `(x k v)' to further specify an ;;> arbitrary keyword `k'. ;;> => ((lambda (&key x [y 2] [z :zz 3]) (list x y z)) :x 'x :zz 'z) ;;> (x 2 z) ;;> Note that keyword values take precedence on the left, and that ;;> keywords are not verified: ;;> => ((lambda (&key y) y) :y 1 :z 3 :y 2) ;;> 1 [(&key &keys) (if (memq state '(#f o r!)) (loop 'k #'xs) (raise-syntax-error #f "misplaced &keys argument" stx #'formals))] ;;> * &rest: a `rest' argument which behaves exactly like the Scheme dot ;;> formal parameter (actually a synonym for it: can't use both). Note ;;> that in case of optional arguments, the rest variable holds any ;;> arguments that were not used for defaults, but using keys doesn't ;;> change its value. For example: ;;> => ((lambda (x &rest r) r) 1 2 3) ;;> (2 3) ;;> => ((lambda (x &optional y &rest r) r) 1) ;;> () ;;> => ((lambda (x &optional y &rest r) r) 1 2 3) ;;> (3) ;;> => ((lambda (x &optional y . r) r) 1 2 3) ;;> (3) ;;> => ((lambda (x &key y &rest r) (list y r)) 1 :y 2 3 4) ;;> (2 (:y 2 3 4)) ;;> => ((lambda (x &key y &rest r) (list y r)) 1 :y 2 3 4 5) ;;> (2 (:y 2 3 4 5)) ;;> Note that the last two examples indicate that there is no error if ;;> the given argument list is not balanced. [(&rest) (if (pair? (syntax-e #'xs)) (loop 'r #'xs) (raise-syntax-error #f "no name for &rest argument" stx #'formals))] ;;> * &rest-keys: similar to `&rest', but all specified keys are removed ;;> with their values. ;;> => ((lambda (x &key y &rest r) r) 1 :x 2 :y 3) ;;> (:x 2 :y 3) ;;> => ((lambda (x &key y &rest-keys r) r) 1 :x 2 :y 3) ;;> (:x 2) [(&rest-keys) (if (pair? (syntax-e #'xs)) (loop 'rk #'xs) (raise-syntax-error #f "no name for &rest-keys argument" stx #'formals))] ;;> * &body: similar to `&rest-keys', but all key/values are removed one ;;> by one until a non-key is encountered. (Warning: this is *not* the ;;> same as in Common Lisp!) ;;> => ((lambda (x &key y &body r) r) 1 :x 2 :y 3) ;;> () ;;> => ((lambda (x &key y &body r) r) 1 :x 2 :y 3 5 6) ;;> (5 6) [(&body &rest-all-keys) ; &rest-all-keys for compatibility (if (pair? (syntax-e #'xs)) (loop 'b #'xs) (raise-syntax-error #f "no name for &body argument" stx #'formals))] ;;> * &all-keys: the list of all keys+vals, without a trailing body. ;;> => ((lambda (&keys x y &all-keys r) r) :x 1 :z 2 3 4) ;;> (:x 1 :z 2) [(&all-keys) (if (pair? (syntax-e #'xs)) (loop 'ak #'xs) (raise-syntax-error #f "no name for &all-keys argument" stx #'formals))] ;;> * &other-keys: the list of unprocessed keys+vals, without a trailing ;;> body. ;;> => ((lambda (&keys x y &other-keys r) r) :x 1 :z 2 3 4) ;;> (:z 2) [(&other-keys) (if (pair? (syntax-e #'xs)) (loop 'ok #'xs) (raise-syntax-error #f "no name for &other-keys argument" stx #'formals))] ;;> ;;> Finally, here is an example where all &rest-like arguments are ;;> different: ;;> => ((lambda (&keys x y ;;> &rest r ;;> &rest-keys rk ;;> &body b ;;> &all-keys ak ;;> &other-keys ok) ;;> (list r rk b ak ok)) ;;> :z 1 :x 2 2 3 4) ;;> ((:z 1 :x 2 2 3 4) (:z 1 2 3 4) (2 3 4) (:z 1 :x 2) (:z 1)) ;;> Note that the following invariants hold: ;;> * rest = (append all-keys body) ;;> * rest-keys = (append other-keys body) [else (raise-syntax-error #f "unknown lambda &-keyword" stx v)])] [(not (or x (memq state '(o k)))) (raise-syntax-error #f "not an identifier" stx v)] [else (let ([test (lambda (var name) (if var (raise-syntax-error #f (format "too many &~a arguments" name) stx #'formals) (set! state 'r!)))]) (case state [(#f) (set! vars (cons v vars))] [(o) (set! opts (cons v opts))] [(k) (set! keys (cons v keys))] [(r!) (raise-syntax-error #f "second identifier after a &rest or similar" stx v)] [(r) (test rest 'rest ) (set! rest v)] [(rk) (test rest-keys 'rest-keys ) (set! rest-keys v)] [(b) (test body 'body ) (set! body v)] [(ak) (test all-keys 'all-keys ) (set! all-keys v)] [(ok) (test other-keys 'other-keys) (set! other-keys v)] [else (raise-syntax-error #f "bad lambda formals" stx v)]) (loop state #'xs))]))] [v (loop state #'(&rest v))])) (set! vars (reverse! vars)) (set! opts (map process-optional-arg (reverse! opts))) (set! keys (map process-keyword-arg (reverse! keys))) (when (and (or rest-keys body all-keys other-keys) (not rest)) (set! rest #'rest)) (cond ;; non-trivial case -- full processing [(or (pair? opts) (pair? keys) rest-keys body all-keys other-keys) (unless rest (set! rest #'rest)) ;; other-keys is computed from all-keys (when (and other-keys (not all-keys)) (set! all-keys #'all-keys)) (quasisyntax/loc stx (lambda (#,@vars . #,rest) (let*-values (#,@(map (lambda (o) #`[(#,(car o)) (if (pair? #,rest) (begin0 (car #,rest) (set! #,rest (cdr #,rest))) #,(cadr o))]) opts) #,@(map (lambda (k) #`[(#,(car k)) (getarg #,rest #,(cadr k) (lambda () #,(caddr k)))]) keys) #,@(if rest-keys #`([(#,rest-keys) (filter-out-keys '#,(map cadr keys) #,rest)]) #'()) #,@(cond ;; At most one scan for body, all-keys, other-keys. This ;; could be much shorter by always using keys/args, but a ;; function call is not a place to spend time on. [(and body all-keys) #`([(#,all-keys #,body) ;; inlined keys/args (let loop ([args #,rest] [keys '()]) (cond [(or (null? args) (null? (cdr args)) (not (keyword*? (car args)))) (values (reverse! keys) args)] [else (loop (cddr args) (list* (cadr args) (car args) keys))]))])] [body #`([(#,body) (let loop ([args #,rest]) (if (or (null? args) (null? (cdr args)) (not (keyword*? (car args)))) args (loop (cddr args))))])] [all-keys #`([(#,all-keys) ;; inlined keys/args, not returning args (let loop ([args #,rest] [keys '()]) (cond [(or (null? args) (null? (cdr args)) (not (keyword*? (car args)))) (reverse! keys)] [else (loop (cddr args) (list* (cadr args) (car args) keys))]))])] [else #'()]) #,@(if other-keys #`([(#,other-keys) ; use all-keys (see above) (filter-out-keys '#,(map cadr keys) #,all-keys)]) #'())) expr0 expr ...)))] ;; common cases: no optional, keyword, or other fancy stuff [(null? vars) (quasisyntax/loc stx (lambda #,(or rest #'()) expr0 expr ...))] [else (quasisyntax/loc stx (lambda (#,@vars . #,(or rest #'())) expr0 expr ...))]))])) ;; Keyword utilities (provide (rename keyword*? keyword?) syntax-keyword? (rename keyword->string* keyword->string) (rename string->keyword* string->keyword) ;; also provide the builtin as `real-keyword' (rename keyword? real-keyword?) (rename keyword->string real-keyword->string) (rename string->keyword string->real-keyword)) ;;>> (keyword? x) ;;> A predicate for keyword symbols (symbols that begin with a ":"). ;;> (Note: this is different from MzScheme's keywords!) (define (keyword*? x) (and (symbol? x) (not (eq? x '||)) (eq? (string-ref (symbol->string x) 0) #\:))) ;;>> (syntax-keyword? x) ;;> Similar to `keyword?' but also works for an identifier (a syntax ;;> object) that contains a keyword. (define (syntax-keyword? x) (keyword*? (if (syntax? x) (syntax-e x) x))) ;;>> (keyword->string k) ;;>> (string->keyword s) ;;> Convert a Swindle keyword to a string and back. (define (keyword->string* k) (if (keyword*? k) (substring (symbol->string k) 1) (raise-type-error 'keyword->string "keyword" k))) (define (string->keyword* s) (if (string? s) (string->symbol (string-append ":" s)) (raise-type-error 'string->keyword "string" s))) ;; Keyword searching utilities (note: no errors for odd length) (provide getarg syntax-getarg getargs keys/args filter-out-keys) ;;>> (getarg args keyword [not-found]) ;;> Searches the given list of arguments for a value matched with the ;;> given keyword. Similar to CL's `getf', except no error checking is ;;> done for an unbalanced list. In case no value is found, the optional ;;> default value can be used -- this can be either a thunk, a promise, or ;;> any other value that will be used as is. For a repeated keyword the ;;> leftmost occurrence is used. (define (getarg args keyword . not-found) (let loop ([args args]) (cond [(or (null? args) (null? (cdr args))) (and (pair? not-found) (let ([x (car not-found)]) (cond [(procedure? x) (x)] [(promise? x) (force x)] [else x])))] [(eq? (car args) keyword) (cadr args)] [else (loop (cddr args))]))) ;;>> (syntax-getarg syntax-args keyword [not-found]) ;;> Similar to `getarg' above, but the input is a syntax object of a ;;> keyword-value list. (define (syntax-getarg syntax-args keyword . not-found) (when (syntax? keyword) (set! keyword (syntax-e keyword))) (let loop ([args syntax-args]) (syntax-case args () [(key arg . more) (if (eq? (syntax-e #'key) keyword) #'arg (loop #'more))] [_ (and (pair? not-found) (let ([x (car not-found)]) (cond [(procedure? x) (x)] [(promise? x) (force x)] [else x])))]))) ;;>> (getargs initargs keyword) ;;> The same as `getarg' but return the list of all key values matched -- ;;> no need for a default value. The result is in the same order as in ;;> the input. (define (getargs initargs keyword) (define (scan tail) (cond [(null? tail) '()] [(null? (cdr tail)) (error 'getargs "keyword list not balanced.")] [(eq? (car tail) keyword) (cons (cadr tail) (scan (cddr tail)))] [else (scan (cddr tail))])) (scan initargs)) ;;>> (keys/args args) ;;> The given argument list is scanned and split at the point where there ;;> are no more keyword-values, and the two parts are returned as two ;;> values. ;;> => (keys/args '(:a 1 :b 2 3 4 5)) ;;> (:a 1 :b 2) ;;> (3 4 5) (define (keys/args args) (let loop ([args args] [keys '()]) (cond [(or (null? args) (null? (cdr args)) (not (keyword*? (car args)))) (values (reverse! keys) args)] [else (loop (cddr args) (list* (cadr args) (car args) keys))]))) ;;>> (filter-out-keys outs args) ;;> The keywords specified in the outs argument, with their matching ;;> values are filtered out of the second arguments. (define (filter-out-keys outs args) (let loop ([as args] [r '()]) (cond [(null? as) (reverse! r)] [(null? (cdr as)) (reverse! (cons (car as) r))] [else (loop (cddr as) (if (memq (car as) outs) r (list* (cadr as) (car as) r)))]))) )