fork `output` so it can cooperate with splicing tag (fixes #132)

pull/134/head
Matthew Butterick 8 years ago
parent c41b3cde3f
commit 05fad9ba54

@ -121,7 +121,7 @@
#'(if COND
(with-handlers ([exn:fail? (λ(exn) (error (format "within when/splice, ~a" (exn-message exn))))])
(SPLICING-TAG BODY ...))
""))]))
(SPLICING-TAG)))]))
(provide when/block) ; bw compat

@ -1,5 +1,7 @@
#lang racket/base
(require scribble/text
(require (only-in scribble/text/syntax-utils include/text)
(only-in "output.rkt" output)
racket/list
(for-syntax racket/base
racket/list
syntax/parse)

@ -0,0 +1,341 @@
#lang racket/base
(require racket/promise
racket/contract/base
pollen/setup) ; to get splicing char
#|
161017:
This is a slightly amended version of scribble/text/output
that cooperates with the project splicing character.
Ordinarily, when the output function comes across a list, it will splice it into the surrounding string.
This version will also splice lists that begin with the splicing char.
|#
(provide
outputable/c
(contract-out
[output (->* (outputable/c) (output-port?) void?)]))
;; See also `provide-special` below
;; Outputs values for the `scribble/text' language:
;; - several atomic values are printed as in `display',
;; - promises, thunks, and boxes are indirections for the value they contain
;; (useful in various cases),
;; - some "special" values are used for controlling output (eg, flushing,
;; prefix changes, etc),
;; - specifically, `block's delimit indentation levels, `splice's do not,
;; - lists (more generally, pairs) are like either one depending on the context
;; (same as blocks/splices when inside a `block'/`splice'), at the toplevel
;; they default to blocks.
;;
;; Uses global state because `output' is wrapped around each expression in a
;; scribble/text file so this is much more convenient than wrapping the whole
;; module's body in a `list' (which will be difficult with definitions etc).
;; The state is a pair of prefixes -- one that is the prefix for the current
;; value (which gets extended with nested blocks), and the other is the prefix
;; for the current "line" (which is reset after a newline). The line-prefix is
;; needed because a line can hold a block, which means that the line-prefix
;; will apply for the contents of the block including newlines in it. This
;; state is associated with a port via a hash table. Another state that is
;; used is the port's column position, which is maintained by the system (when
;; line counts are enabled) -- this is used to tell what part of a prefix is
;; already displayed.
;;
;; Each prefix is either an integer (for a number of spaces) or a string. The
;; prefix mechanism can be disabled by using #f for the global prefix, and in
;; this case the line prefix can have (cons pfx lpfx) so it can be restored --
;; used by `disable-prefix' and `restore-prefix' resp. (This is different from
;; a 0 prefix -- #f means that no prefix will be accumulated).
;;
(define (output x [p (current-output-port)])
;; these are the global prefix and the one that is local to the current line
(define pfxs (port->state p))
;; the current mode for lists
(define list=block? #t)
;; the low-level string output function (can change with `with-writer')
(define write write-string)
;; to get the output column
(define (getcol) (let-values ([(line col pos) (port-next-location p)]) col))
;; total size of the two prefixes
(define (2pfx-length pfx1 pfx2)
(if (and pfx1 pfx2)
(+ (if (number? pfx1) pfx1 (string-length pfx1))
(if (number? pfx2) pfx2 (string-length pfx2)))
0))
;; combines a prefix with a target column to get to
(define (pfx+col pfx)
(and pfx (let ([col (getcol)])
(cond [(number? pfx) (max pfx col)]
[(>= (string-length pfx) col) pfx]
[else (string-append
pfx (make-spaces (- col (string-length pfx))))]))))
;; adds two prefixes
(define (pfx+ pfx1 pfx2)
(and pfx1 pfx2
(if (and (number? pfx1) (number? pfx2)) (+ pfx1 pfx2)
(string-append (if (number? pfx1) (make-spaces pfx1) pfx1)
(if (number? pfx2) (make-spaces pfx2) pfx2)))))
;; prints two prefixes
(define (output-pfx col pfx1 pfx2)
(define-syntax-rule (->str pfx) (if (number? pfx) (make-spaces pfx) pfx))
(define-syntax-rule (show pfx) ; optimize when not needed
(unless (eq? pfx 0) (write (->str pfx) p)))
(when (and pfx1 pfx2)
(if (eq? 0 col)
(begin (show pfx1) (show pfx2))
(let ([len1 (if (number? pfx1) pfx1 (string-length pfx1))])
(cond [(< col len1) (write (->str pfx1) p col) (show pfx2)]
[(= col len1) (show pfx2)]
[(eq? 0 pfx2)]
[else
(let ([col (- col len1)]
[len2 (if (number? pfx2) pfx2 (string-length pfx2))])
(when (< col len2) (write (->str pfx2) p col)))])))))
;; the basic printing unit: strings
(define (output-string x)
(define pfx (mcar pfxs))
(if (not pfx) ; prefix disabled?
(write x p)
(let ([len (string-length x)]
[nls (regexp-match-positions* #rx"\n" x)])
(let loop ([start 0] [nls nls] [lpfx (mcdr pfxs)] [col (getcol)])
(cond [(pair? nls)
(define nl (car nls))
(if (regexp-match? #rx"^ *$" x start (car nl))
(newline p) ; only spaces before the end of the line
(begin (output-pfx col pfx lpfx)
(write x p start (cdr nl))))
(loop (cdr nl) (cdr nls) 0 0)]
;; last substring from here (always set lpfx state when done)
[(start . = . len)
(set-mcdr! pfxs lpfx)]
[(col . > . (2pfx-length pfx lpfx))
(set-mcdr! pfxs lpfx)
;; the prefix was already shown, no accumulation needed
(write x p start)]
[else
(define m (regexp-match-positions #rx"^ +" x start))
;; accumulate spaces to lpfx, display if it's not all spaces
(define lpfx* (if m (pfx+ lpfx (- (cdar m) (caar m))) lpfx))
(set-mcdr! pfxs lpfx*)
(unless (and m (= len (cdar m)))
(output-pfx col pfx lpfx*)
;; the spaces were already added to lpfx
(write x p (if m (cdar m) start)))])))))
;; blocks and splices
(define (output-block c)
(define pfx (mcar pfxs))
(define lpfx (mcdr pfxs))
(define npfx (pfx+col (pfx+ pfx lpfx)))
(set-mcar! pfxs npfx) (set-mcdr! pfxs 0)
(if (list? c)
(let ([c (if (eq? (setup:splicing-tag) (car c)) ; patch to cooperate with splicing char
(cdr c)
c)])
(for ([c (in-list c)]) (loop c)))
(begin (loop (car c)) (loop (cdr c))))
(set-mcar! pfxs pfx) (set-mcdr! pfxs lpfx))
(define (output-splice c)
(for-each loop c))
;; main loop
(define (loop x)
(cond
;; no output for these
[(or (void? x) (not x) (null? x)) (void)]
;; for lists and pairs the current line prefix is added to the global
;; one, then output the contents recursively (no need to change the
;; state, since we pass the values in the loop, and we'd need to restore
;; it afterwards anyway)
[(pair? x) (if list=block? (output-block x) (output-splice x))]
;; delayed values
[(and (procedure? x) (procedure-arity-includes? x 0)) (loop (x))]
[(promise? x) (loop (force x))]
[(box? x) (loop (unbox x))]
;; special output wrappers
[(special? x)
(define c (special-contents x))
(case (special-flag x)
;; preserve tailness & avoid `set!' for blocks/splices if possible
[(block) (if list=block?
(output-block c)
(begin (set! list=block? #t)
(output-block c)
(set! list=block? #f)))]
[(splice) (if list=block?
(begin (set! list=block? #f)
(output-splice c)
(set! list=block? #t))
(output-splice c))]
[(flush) ; useful before `disable-prefix'
(output-pfx (getcol) (mcar pfxs) (mcdr pfxs))]
[(disable-prefix) ; save the previous pfxs
(define pfx (mcar pfxs))
(define lpfx (mcdr pfxs))
(set-mcar! pfxs #f) (set-mcdr! pfxs (cons pfx lpfx))
(for-each loop c)
(set-mcar! pfxs pfx) (set-mcdr! pfxs lpfx)]
[(restore-prefix) ; restore the previous pfxs
(define pfx (mcar pfxs))
(define lpfx (mcdr pfxs))
(define npfx (pfx+col (if (and (not pfx) (pair? lpfx))
(pfx+ (car lpfx) (cdr lpfx))
(pfx+ pfx lpfx))))
(set-mcar! pfxs npfx) (set-mcdr! pfxs 0)
(for-each loop c)
(set-mcar! pfxs pfx) (set-mcdr! pfxs lpfx)]
[(add-prefix) ; add to the current prefix (unless it's #f)
(define pfx (mcar pfxs))
(define lpfx (mcdr pfxs))
(define npfx (pfx+ (pfx+col (pfx+ pfx lpfx)) (car c)))
(set-mcar! pfxs npfx) (set-mcdr! pfxs 0)
(for-each loop (cdr c))
(set-mcar! pfxs pfx) (set-mcdr! pfxs lpfx)]
[(set-prefix)
(define pfx (mcar pfxs))
(define lpfx (mcdr pfxs))
(set-mcar! pfxs (car c)) (set-mcdr! pfxs 0)
(for-each loop (cdr c))
(set-mcar! pfxs pfx) (set-mcdr! pfxs lpfx)]
[(with-writer)
(define old write)
(set! write (or (car c) write-string))
(for-each loop (cdr c))
(set! write old)]
#; ; no need for this hack yet
[(with-writer-change)
;; The function gets the old writer and return a new one (useful to
;; save the current writer and restore it inside). Could also be
;; used to extend a writer, but that shows why a customizable writer
;; is a bad choice: instead, it should be a list of substitutions
;; that can be extended more conveniently. A simple implementation
;; would be to chain functions that do substitutions. But that runs
;; into problems when functions want to substitute the same thing,
;; and worse: when the output of one function would get substituted
;; again by another. Another approach would be to join matcher
;; regexps with "|" after wrapping each one with parens, then find
;; out which one matched by looking at the result and applying its
;; substitution, but the problem with that is that is that it forbids
;; having parens in the regexps -- this could be fixed by not
;; parenthesizing each expression, and instead running the found
;; match against each of the input regexps to find the matching one,
;; but that can be very inefficient. Yet another issue is that in
;; some cases we might *want* the "worse" feature mentioned earlier:
;; for example, when we want to do some massaging of the input texts
;; yet still have the result encoded for HTML output -- so perhaps
;; the simple approach is still better. The only difference from the
;; current `with-writer' is using a substituting function, so it can
;; be composed with the current one instead of replacing it
;; completely.
(define old write)
(set! write ((car c) write))
(for-each loop (cdr c))
(set! write old)]
[else (error 'output "unknown special value flag: ~e"
(special-flag x))])]
[else
(output-string
(cond [(string? x) x]
[(bytes? x) (bytes->string/utf-8 x)]
[(symbol? x) (symbol->string x)]
[(path? x) (path->string x)]
[(keyword? x) (keyword->string x)]
[(number? x) (number->string x)]
[(char? x) (string x)]
;; generic fallback: throw an error (could use `display' so new
;; values can define how they're shown, but the same
;; functionality can be achieved with thunks and prop:procedure)
[else (error 'output "don't know how to render value: ~v" x)]))]))
;;
(port-count-lines! p)
(loop x)
(void))
(define port->state
(let ([t (make-weak-hasheq)]
[last '(#f #f)]) ; cache for the last port, to avoid a hash lookup
(λ (p)
(if (eq? p (car last)) (cdr last)
(let ([s (or (hash-ref t p #f)
(let ([s (mcons 0 0)]) (hash-set! t p s) s))])
(set! last (cons p s))
s)))))
;; special constructs
(define-struct special (flag contents))
(define-syntax define/provide-special
(syntax-rules ()
[(_ (name))
(begin (provide (contract-out [name (->* () () #:rest (listof outputable/c) any/c)]))
(define (name . contents)
(make-special 'name contents)))]
[(_ (name [x ctc] ...))
(begin (provide (contract-out [name (->* (ctc ...) () #:rest (listof outputable/c) any/c)]))
(define (name x ... . contents)
(make-special 'name (list* x ... contents))))]
[(_ name)
(begin (provide name)
(define name (make-special 'name #f)))]))
(define/provide-special (block))
(define/provide-special (splice))
(define/provide-special flush)
(define/provide-special (disable-prefix))
(define/provide-special (restore-prefix))
(define/provide-special (add-prefix [pfx (or/c string? exact-nonnegative-integer?)]))
(define/provide-special (set-prefix [pfx (or/c string? exact-nonnegative-integer?)]))
(define/provide-special (with-writer [writer (or/c #f (->* (string? output-port?) (exact-nonnegative-integer? exact-nonnegative-integer?) any/c))]))
#; ; no need for this hack yet
(define/provide-special (with-writer-change writer))
(define make-spaces ; (efficiently)
(let ([t (make-hasheq)] [v (make-vector 200 #f)])
(λ (n)
(or (if (< n 200) (vector-ref v n) (hash-ref t n #f))
(let ([spaces (make-string n #\space)])
(if (< n 200) (vector-set! v n spaces) (hash-set! t n spaces))
spaces)))))
;; Convenient utilities
(provide add-newlines)
(define (add-newlines list #:sep [sep "\n"])
(define r
(let loop ([list list])
(if (null? list)
null
(let ([1st (car list)])
(if (or (not 1st) (void? 1st))
(loop (cdr list))
(list* sep 1st (loop (cdr list))))))))
(if (null? r) r (cdr r)))
(provide split-lines)
(define (split-lines list)
(let loop ([list list] [cur '()] [r '()])
(cond
[(null? list) (reverse (cons (reverse cur) r))]
[(equal? "\n" (car list)) (loop (cdr list) '() (cons (reverse cur) r))]
[else (loop (cdr list) (cons (car list) cur) r)])))
(define outputable/c
(lambda (v) #t)
;; too expensive:
#;
(recursive-contract
(or/c void?
#f
null?
(cons/c outputable/c outputable/c)
(-> outputable/c)
promise?
(box/c outputable/c)
special?
string?
bytes?
symbol?
path?
keyword?
number?
char?)))

@ -1 +1 @@
1476735837
1476735877

@ -184,8 +184,7 @@
,(require-directory-require-files source-path)
(parameterize ([current-pagetree (make-project-pagetree ,(current-project-root))])
(let ([,(setup:main-export source-path) (cached-doc ,(path->string source-path))]
[,(setup:meta-export source-path) (cached-metas ,(path->string source-path))]
[,(setup:splicing-tag source-path) (λ xs xs)]) ; splice behavior is different in textual context
[,(setup:meta-export source-path) (cached-metas ,(path->string source-path))])
(local-require pollen/template pollen/top)
(define here (path->pagenode
(or (select-from-metas ',(setup:here-path-key source-path) ,(setup:meta-export source-path)) 'unknown)))

@ -0,0 +1,10 @@
#lang racket/base
(require rackunit pollen/private/output racket/port)
(define-syntax-rule (check-output outputter string)
(check-equal? (with-output-to-string (λ () outputter)) string))
;; output function should splice lists, and remove splicing char at the beginning of a list
(check-output (output '("tic" "tac" "toe")) "tictactoe")
(check-output (output '("tic" ("tac") "toe")) "tictactoe")
(check-output (output '("tic" (@ "tac") "toe")) "tictactoe")
Loading…
Cancel
Save