1 ;;;; code for string to octet conversion
3 ;;;; This software is part of the SBCL system. See the README file for
6 ;;;; This software is derived from the CMU CL system, which was
7 ;;;; written at Carnegie Mellon University and released into the
8 ;;;; public domain. The software is in the public domain and is
9 ;;;; provided with absolutely no warranty. See the COPYING and CREDITS
10 ;;;; files for more information.
12 ;;; FIXME: The latin9 stuff is currently #!+sb-unicode, because I
13 ;;; don't like the idea of trying to do CODE-CHAR #x<big>. Is that a
14 ;;; justified fear? Can we arrange that it's caught and converted to
15 ;;; a decoding error error? Or should we just give up on non-Unicode
18 (in-package "SB!IMPL")
20 ;;; FIXME: don't we have this somewhere else?
21 (deftype array-range ()
22 "A number that can represent an index into a vector, including
24 '(integer 0 #.sb!xc:array-dimension-limit))
28 ;;; encoding condition
30 (define-condition octets-encoding-error (character-encoding-error)
31 ((string :initarg :string :reader octets-encoding-error-string)
32 (position :initarg :position :reader octets-encoding-error-position)
33 (external-format :initarg :external-format
34 :reader octets-encoding-error-external-format))
35 (:report (lambda (c s)
36 (format s "Unable to encode character ~A as ~S."
37 (char-code (char (octets-encoding-error-string c)
38 (octets-encoding-error-position c)))
39 (octets-encoding-error-external-format c)))))
41 (defun read-replacement-character ()
43 "Replacement byte, bytes, character, or string (evaluated): ")
44 (finish-output *query-io*)
45 (list (eval (read *query-io*))))
47 (defun encoding-error (external-format string pos)
49 (error 'octets-encoding-error
50 :external-format external-format
53 (use-value (replacement)
54 :report "Supply a set of bytes to use in place of the invalid one."
55 :interactive read-replacement-character
58 (make-array 1 :element-type '(unsigned-byte 8) :initial-element replacement))
60 (string-to-octets (string replacement)
61 :external-format external-format))
63 (string-to-octets replacement
64 :external-format external-format))
66 (coerce replacement '(simple-array (unsigned-byte 8) (*))))))))
68 ;;; decoding condition
70 ;;; for UTF8, the specific condition signalled will be a generalized
71 ;;; instance of one of the following:
73 ;;; end-of-input-in-character
74 ;;; character-out-of-range
75 ;;; invalid-utf8-starter-byte
76 ;;; invalid-utf8-continuation-byte
77 ;;; overlong-utf8-sequence
79 ;;; Of these, the only one truly likely to be of interest to calling
80 ;;; code is end-of-input-in-character (in which case it's likely to
81 ;;; want to make a note of octet-decoding-error-start, supply "" as a
82 ;;; replacement string, and then move that last chunk of bytes to the
83 ;;; beginning of its buffer for the next go round) but they're all
84 ;;; provided on the off chance they're of interest. The next most
85 ;;; likely interesting option is overlong-utf8-sequence -- the
86 ;;; application, if it cares to, can decode this itself (taking care
87 ;;; to ensure that the result isn't out of range of CHAR-CODE-LIMIT)
88 ;;; and return that result. This library doesn't provide support for
89 ;;; that as a conforming UTF-8-using program is supposed to treat it
92 (define-condition octet-decoding-error (character-decoding-error)
93 ((array :initarg :array :accessor octet-decoding-error-array)
94 (start :initarg :start :accessor octet-decoding-error-start)
95 (end :initarg :end :accessor octet-decoding-error-end)
96 (position :initarg :pos :accessor octet-decoding-bad-byte-position)
97 (external-format :initarg :external-format
98 :accessor octet-decoding-error-external-format))
100 (lambda (condition stream)
101 (format stream "Illegal ~S character starting at byte position ~D."
102 (octet-decoding-error-external-format condition)
103 (octet-decoding-error-start condition)))))
105 (define-condition end-of-input-in-character (octet-decoding-error) ())
106 (define-condition character-out-of-range (octet-decoding-error) ())
107 (define-condition invalid-utf8-starter-byte (octet-decoding-error) ())
108 (define-condition invalid-utf8-continuation-byte (octet-decoding-error) ())
109 (define-condition overlong-utf8-sequence (octet-decoding-error) ())
111 (define-condition malformed-ascii (octet-decoding-error) ())
113 (defun read-replacement-string ()
114 (format *query-io* "Enter a replacement string designator (evaluated): ")
115 (finish-output *query-io*)
116 (list (eval (read *query-io*))))
118 (defun decoding-error (array start end external-format reason pos)
121 :external-format external-format
127 :report "Supply a replacement string designator."
128 :interactive read-replacement-string
131 ;;; Utilities used in both to-string and to-octet conversions
133 (defmacro instantiate-octets-definition (definer)
135 (,definer aref (simple-array (unsigned-byte 8) (*)))
136 (,definer sap-ref-8 system-area-pointer)))
138 ;;; FIXME: find out why the comment about SYMBOLICATE below is true
139 ;;; and fix it, or else replace with SYMBOLICATE.
141 ;;; FIXME: this is cute, but is going to prevent greps for def.*<name>
142 ;;; from working for (defun ,(make-od-name ...) ...)
143 (eval-when (:compile-toplevel :load-toplevel :execute)
144 (defun make-od-name (sym1 sym2)
145 ;; "MAKE-NAME" is too generic, but this doesn't do quite what
146 ;; SYMBOLICATE does; MAKE-OD-NAME ("octets definition") it is
148 (intern (concatenate 'string (symbol-name sym1) "-" (symbol-name sym2))
149 (symbol-package sym1))))
151 ;;;; to-octets conversions
153 ;;; to latin (including ascii)
155 ;;; Converting bytes to character codes is easy: just use a 256-element
156 ;;; lookup table that maps each possible byte to its corresponding
159 ;;; Converting character codes to bytes is a little harder, since the
160 ;;; codes may be spare (e.g. we use codes 0-127, 3490, and 4598). The
161 ;;; previous version of this macro utilized a gigantic CASE expression
162 ;;; to do the hard work, with the result that the code was huge (since
163 ;;; SBCL's then-current compilation strategy for CASE expressions was
164 ;;; (and still is) converting CASE into COND into if-the-elses--which is
165 ;;; also inefficient unless your code happens to occur very early in the
168 ;;; The current strategy is to build a table:
170 ;;; [ ... code_1 byte_1 code_2 byte_2 ... code_n byte_n ... ]
172 ;;; such that the codes are sorted in order from lowest to highest. We
173 ;;; can then binary search the table to discover the appropriate byte
174 ;;; for a character code. We also implement an optimization: all unibyte
175 ;;; mappings do not remap ASCII (0-127) and some do not remap part of
176 ;;; the range beyond character code 127. So we check to see if the
177 ;;; character code falls into that range first (a quick check, since
178 ;;; character codes are guaranteed to be positive) and then do the binary
179 ;;; search if not. This optimization also enables us to cut down on the
180 ;;; size of our lookup table.
181 (defmacro define-unibyte-mapper (byte-char-name code-byte-name &rest exceptions)
182 (let* (;; Build a list of (CODE BYTE) pairs
183 (pairs (loop for byte below 256
184 for code = (let ((exception (cdr (assoc byte exceptions))))
186 ((car exception) (car exception))
187 ((null exception) byte)
189 when code collect (list code byte) into elements
190 finally (return elements)))
191 ;; Find the smallest character code such that the corresponding
192 ;; byte is != to the code.
193 (lowest-non-equivalent-code (position-if-not #'(lambda (pair)
196 ;; Sort them for our lookup table.
197 (sorted-pairs (sort (subseq pairs lowest-non-equivalent-code)
199 ;; Create the lookup table.
201 (reduce #'append sorted-pairs :from-end t :initial-value nil)))
203 ; Can't inline it with a non-null lexical environment anyway.
204 ;(declaim (inline ,byte-char-name))
205 (let ((byte-to-code-table
206 ,(make-array 256 :element-type t #+nil 'char-code
207 :initial-contents (loop for byte below 256
209 (let ((exception (cadr (assoc byte exceptions))))
214 ,(make-array (length sorted-lookup-table)
215 :initial-contents sorted-lookup-table)))
216 (defun ,byte-char-name (byte)
217 (declare (optimize speed (safety 0))
218 (type (unsigned-byte 8) byte))
219 (aref byte-to-code-table byte))
220 (defun ,code-byte-name (code)
221 (declare (optimize speed (safety 0))
222 (type char-code code))
223 (if (< code ,lowest-non-equivalent-code)
225 ;; We could toss in some TRULY-THEs if we really needed to
226 ;; make this faster...
228 with high = (- (length code-to-byte-table) 2)
230 do (let ((mid (logandc2 (truncate (+ low high 2) 2) 1)))
231 (if (< code (aref code-to-byte-table mid))
232 (setf high (- mid 2))
234 finally (return (if (eql code (aref code-to-byte-table low))
235 (aref code-to-byte-table (1+ low))
239 (define-unibyte-mapper
251 (declaim (inline get-latin-bytes))
252 (defun get-latin-bytes (mapper external-format string pos)
253 (let ((code (funcall mapper (char-code (char string pos)))))
254 (declare (type (or null char-code) code))
256 ((and code (< code 256)) code)
258 (encoding-error external-format string pos)))
261 (declaim (inline code->ascii-mapper))
262 (defun code->ascii-mapper (code)
263 (declare (optimize speed (safety 0))
264 (type char-code code))
269 (declaim (inline get-ascii-bytes))
270 (defun get-ascii-bytes (string pos)
271 (declare (optimize speed (safety 0))
272 (type simple-string string)
273 (type array-range pos))
274 (get-latin-bytes #'code->ascii-mapper :ascii string pos))
276 (declaim (inline get-latin1-bytes))
277 (defun get-latin1-bytes (string pos)
278 (declare (optimize speed (safety 0))
279 (type simple-string string)
280 (type array-range pos))
281 (get-latin-bytes #'identity :latin-1 string pos))
285 (declaim (inline get-latin9-bytes))
286 (defun get-latin9-bytes (string pos)
287 (declare (optimize speed (safety 0))
288 (type simple-string string)
289 (type array-range pos))
290 (get-latin-bytes #'code->latin9-mapper :latin-9 string pos)))
292 (declaim (inline string->latin%))
293 (defun string->latin% (string sstart send get-bytes null-padding)
294 (declare (optimize speed)
295 (type simple-string string)
296 (type index sstart send)
297 (type (integer 0 1) null-padding)
298 (type function get-bytes))
299 ;; The latin encodings are all unibyte encodings, so just directly
300 ;; compute the number of octets we're going to generate.
301 (let ((octets (make-array (+ (- send sstart) null-padding)
302 ;; This takes care of any null padding the
305 :element-type '(unsigned-byte 8)))
310 (loop for pos of-type index from sstart below send
311 do (let ((byte (funcall get-bytes string pos)))
314 (locally (declare (optimize (sb!c::insert-array-bounds-checks 0)))
315 (setf (aref octets index) byte)))
316 ((simple-array (unsigned-byte 8) (*))
317 ;; KLUDGE: We ran into encoding errors. Bail and do
318 ;; things the slow way (does anybody actually use this
319 ;; functionality besides our own test suite?).
320 (setf error-position pos)
323 finally (return-from string->latin% octets))
325 ;; We have encoded INDEX octets so far and we ran into an encoding
326 ;; error at ERROR-POSITION.
327 (let ((new-octets (make-array (* index 2)
328 :element-type '(unsigned-byte 8)
329 :adjustable t :fill-pointer index)))
330 (replace new-octets octets)
331 (loop for pos of-type index from error-position below send
332 do (let ((thing (funcall get-bytes string pos)))
335 (vector-push-extend thing new-octets))
336 ((simple-array (unsigned-byte 8) (*))
337 (dotimes (i (length thing))
338 (vector-push-extend (aref thing i) new-octets)))))
339 finally (return-from string->latin%
341 (unless (zerop null-padding)
342 (vector-push-extend 0 new-octets))
343 (copy-seq new-octets))))))))
345 (defun string->ascii (string sstart send null-padding)
346 (declare (optimize speed (safety 0))
347 (type simple-string string)
348 (type array-range sstart send))
349 (values (string->latin% string sstart send #'get-ascii-bytes null-padding)))
351 (defun string->latin1 (string sstart send null-padding)
352 (declare (optimize speed (safety 0))
353 (type simple-string string)
354 (type array-range sstart send))
355 (values (string->latin% string sstart send #'get-latin1-bytes null-padding)))
358 (defun string->latin9 (string sstart send null-padding)
359 (declare (optimize speed (safety 0))
360 (type simple-string string)
361 (type array-range sstart send))
362 (values (string->latin% string sstart send #'get-latin9-bytes null-padding)))
366 (declaim (inline char-len-as-utf8))
367 (defun char-len-as-utf8 (code)
368 (declare (optimize speed (safety 0))
369 (type (integer 0 (#.sb!xc:char-code-limit)) code))
370 (cond ((< code 0) (bug "can't happen"))
374 ((< code #x110000) 4)
375 (t (bug "can't happen"))))
377 (defun string->utf8 (string sstart send null-padding)
378 (declare (optimize (speed 3) (safety 0))
379 (type simple-string string)
380 (type (integer 0 1) null-padding)
381 (type array-range sstart send))
382 (macrolet ((ascii-bash ()
383 '(let ((array (make-array (+ null-padding (- send sstart))
384 :element-type '(unsigned-byte 8))))
386 and j from sstart below send
387 do (setf (aref array i) (char-code (char string j))))
390 ((simple-array character (*))
391 (let ((utf8-length 0))
392 ;; Since it has to fit in a vector, it must be a fixnum!
393 (declare (type (and unsigned-byte fixnum) utf8-length))
394 (loop for i of-type index from sstart below send
395 do (incf utf8-length (char-len-as-utf8 (char-code (char string i)))))
396 (if (= utf8-length (- send sstart))
398 (let ((array (make-array (+ null-padding utf8-length)
399 :element-type '(unsigned-byte 8)))
401 (declare (type index index))
403 (setf (aref array index) b)
405 (declare (inline add-byte))
406 (loop for i of-type index from sstart below send
407 do (let ((code (char-code (char string i))))
408 (case (char-len-as-utf8 code)
412 (add-byte (logior #b11000000 (ldb (byte 5 6) code)))
413 (add-byte (logior #b10000000 (ldb (byte 6 0) code))))
415 (add-byte (logior #b11100000 (ldb (byte 4 12) code)))
416 (add-byte (logior #b10000000 (ldb (byte 6 6) code)))
417 (add-byte (logior #b10000000 (ldb (byte 6 0) code))))
419 (add-byte (logior #b11110000 (ldb (byte 3 18) code)))
420 (add-byte (logior #b10000000 (ldb (byte 6 12) code)))
421 (add-byte (logior #b10000000 (ldb (byte 6 6) code)))
422 (add-byte (logior #b10000000 (ldb (byte 6 0) code))))))
423 finally (return array)))))))
425 ((simple-array base-char (*))
426 ;; On unicode builds BASE-STRINGs are limited to ASCII range,
427 ;; so we can take a fast path -- and get benefit of the element
428 ;; type information. On non-unicode build BASE-CHAR ==
431 ((simple-array nil (*))
433 (make-array 0 :element-type '(unsigned-byte 8))
434 ;; Just get the error...
435 (aref string sstart))))))
437 ;;;; to-string conversions
439 ;;; from latin (including ascii)
441 (defmacro define-ascii->string (accessor type)
442 (let ((name (make-od-name 'ascii->string accessor)))
444 (defun ,name (array astart aend)
445 (declare (optimize speed)
447 (type array-range astart aend))
448 ;; Since there is such a thing as a malformed ascii byte, a
449 ;; simple "make the string, fill it in" won't do.
450 (let ((string (make-array 0 :element-type 'character :fill-pointer 0 :adjustable t)))
451 (loop for apos from astart below aend
452 do (let* ((code (,accessor array apos))
456 (decoding-error array apos (1+ apos) :ascii
457 'malformed-ascii apos))))
458 (if (characterp string-content)
459 (vector-push-extend string-content string)
460 (loop for c across string-content
461 do (vector-push-extend c string))))
462 finally (return (coerce string 'simple-string))))))))
463 (instantiate-octets-definition define-ascii->string)
465 (defmacro define-latin->string* (accessor type)
466 (let ((name (make-od-name 'latin->string* accessor)))
468 (declaim (inline ,name))
469 (defun ,name (string sstart send array astart aend mapper)
470 (declare (optimize speed (safety 0))
471 (type simple-string string)
473 (type array-range sstart send astart aend)
475 (loop for spos from sstart below send
476 for apos from astart below aend
477 do (setf (char string spos)
478 (code-char (funcall mapper (,accessor array apos))))
479 finally (return (values string spos apos)))))))
480 (instantiate-octets-definition define-latin->string*)
482 (defmacro define-latin1->string* (accessor type)
483 (declare (ignore type))
484 (let ((name (make-od-name 'latin1->string* accessor)))
486 (defun ,name (string sstart send array astart aend)
487 (,(make-od-name 'latin->string* accessor) string sstart send array astart aend #'identity)))))
488 (instantiate-octets-definition define-latin1->string*)
492 (defmacro define-latin9->string* (accessor type)
493 (declare (ignore type))
494 (let ((name (make-od-name 'latin9->string* accessor)))
496 (defun ,name (string sstart send array astart aend)
497 (,(make-od-name 'latin->string* accessor) string sstart send array astart aend #'latin9->code-mapper)))))
498 (instantiate-octets-definition define-latin9->string*))
500 (defmacro define-latin->string (accessor type)
501 (let ((name (make-od-name 'latin->string accessor)))
503 (declaim (inline ,name))
504 (defun ,name (array astart aend mapper)
505 (declare (optimize speed (safety 0))
507 (type array-range astart aend)
508 (type function mapper))
509 (let ((length (the array-range (- aend astart))))
510 (values (,(make-od-name 'latin->string* accessor) (make-string length) 0 length
513 (instantiate-octets-definition define-latin->string)
515 (defmacro define-latin1->string (accessor type)
516 (declare (ignore type))
517 `(defun ,(make-od-name 'latin1->string accessor) (array astart aend)
518 (,(make-od-name 'latin->string accessor) array astart aend #'identity)))
519 (instantiate-octets-definition define-latin1->string)
523 (defmacro define-latin9->string (accessor type)
524 (declare (ignore type))
525 `(defun ,(make-od-name 'latin9->string accessor) (array astart aend)
526 (,(make-od-name 'latin->string accessor) array astart aend #'latin9->code-mapper)))
527 (instantiate-octets-definition define-latin9->string))
531 (defmacro define-bytes-per-utf8-character (accessor type)
532 (let ((name (make-od-name 'bytes-per-utf8-character accessor)))
534 ;;(declaim (inline ,name))
536 (string->utf8 (string (code-char ,(1- sb!xc:char-code-limit)))
538 (declare (type (simple-array (unsigned-byte 8) (#!+sb-unicode 4 #!-sb-unicode 2)) lexically-max))
539 (defun ,name (array pos end)
540 (declare (optimize speed (safety 0))
542 (type array-range pos end))
543 ;; returns the number of bytes consumed and nil if it's a
544 ;; valid character or the number of bytes consumed and a
545 ;; replacement string if it's not.
546 (let ((initial-byte (,accessor array pos))
548 (reject-position pos)
549 (remaining-bytes (- end pos)))
550 (declare (type array-range reject-position remaining-bytes))
551 (labels ((valid-utf8-starter-byte-p (b)
552 (declare (type (unsigned-byte 8) b))
554 ((zerop (logand b #b10000000)) 1)
555 ((= (logand b #b11100000) #b11000000)
557 ((= (logand b #b11110000) #b11100000)
559 ((= (logand b #b11111000) #b11110000)
561 ((= (logand b #b11111100) #b11111000)
563 ((= (logand b #b11111110) #b11111100)
568 (setf reject-reason 'invalid-utf8-starter-byte))
570 (enough-bytes-left-p (x)
571 (let ((ok (> end (+ pos (1- x)))))
573 (setf reject-reason 'end-of-input-in-character))
575 (valid-secondary-p (x)
576 (let* ((idx (the array-range (+ pos x)))
577 (b (,accessor array idx))
578 (ok (= (logand b #b11000000) #b10000000)))
580 (setf reject-reason 'invalid-utf8-continuation-byte)
581 (setf reject-position idx))
583 (preliminary-ok-for-length (maybe-len len)
584 (and (eql maybe-len len)
585 ;; Has to be done in this order so that
586 ;; certain broken sequences (e.g., the
587 ;; two-byte sequence `"initial (length 3)"
588 ;; "non-continuation"' -- `#xef #x32')
589 ;; signal only part of that sequence as
591 (loop for i from 1 below (min len remaining-bytes)
592 always (valid-secondary-p i))
593 (enough-bytes-left-p len)))
595 (let ((ok (or (/= initial-byte x)
596 (/= (logior (,accessor array (the array-range (+ pos 1)))
600 (setf reject-reason 'overlong-utf8-sequence))
602 (character-below-char-code-limit-p ()
603 ;; This is only called on a four-byte sequence
604 ;; (two in non-unicode builds) to ensure we
605 ;; don't go over SBCL's character limts.
606 (let ((ok (cond ((< (aref lexically-max 0) (,accessor array pos))
608 ((> (aref lexically-max 0) (,accessor array pos))
610 ((< (aref lexically-max 1) (,accessor array (+ pos 1)))
613 ((> (aref lexically-max 1) (,accessor array (+ pos 1)))
616 ((< (aref lexically-max 2) (,accessor array (+ pos 2)))
619 ((> (aref lexically-max 2) (,accessor array (+ pos 2)))
622 ((< (aref lexically-max 3) (,accessor array (+ pos 3)))
626 (setf reject-reason 'character-out-of-range))
628 (declare (inline valid-utf8-starter-byte-p
631 preliminary-ok-for-length
633 (let ((maybe-len (valid-utf8-starter-byte-p initial-byte)))
634 (cond ((eql maybe-len 1)
636 ((and (preliminary-ok-for-length maybe-len 2)
637 (overlong-chk #b11000000 #b10111111)
638 (overlong-chk #b11000001 #b10111111)
639 #!-sb-unicode (character-below-char-code-limit-p))
641 ((and (preliminary-ok-for-length maybe-len 3)
642 (overlong-chk #b11100000 #b10011111)
643 #!-sb-unicode (not (setf reject-reason 'character-out-of-range)))
645 ((and (preliminary-ok-for-length maybe-len 4)
646 (overlong-chk #b11110000 #b10001111)
647 #!-sb-unicode (not (setf reject-reason 'character-out-of-range))
648 (character-below-char-code-limit-p))
650 ((and (preliminary-ok-for-length maybe-len 5)
651 (overlong-chk #b11111000 #b10000111)
652 (not (setf reject-reason 'character-out-of-range)))
653 (bug "can't happen"))
654 ((and (preliminary-ok-for-length maybe-len 6)
655 (overlong-chk #b11111100 #b10000011)
656 (not (setf reject-reason 'character-out-of-range)))
657 (bug "can't happen"))
659 (let* ((bad-end (ecase reject-reason
660 (invalid-utf8-starter-byte
662 (end-of-input-in-character
664 (invalid-utf8-continuation-byte
666 ((overlong-utf8-sequence character-out-of-range)
668 (bad-len (- bad-end pos)))
669 (declare (type array-range bad-end bad-len))
670 (let ((replacement (decoding-error array pos bad-end :utf-8 reject-reason reject-position)))
671 (values bad-len replacement)))))))))))))
672 (instantiate-octets-definition define-bytes-per-utf8-character)
674 (defmacro define-simple-get-utf8-char (accessor type)
675 (let ((name (make-od-name 'simple-get-utf8-char accessor)))
677 (declaim (inline ,name))
678 (defun ,name (array pos bytes)
679 (declare (optimize speed (safety 0))
681 (type array-range pos)
682 (type (integer 1 4) bytes))
684 (,accessor array (the array-range (+ pos x)))))
685 (declare (inline cref))
686 (code-char (ecase bytes
688 (2 (logior (ash (ldb (byte 5 0) (cref 0)) 6)
689 (ldb (byte 6 0) (cref 1))))
690 (3 (logior (ash (ldb (byte 4 0) (cref 0)) 12)
691 (ash (ldb (byte 6 0) (cref 1)) 6)
692 (ldb (byte 6 0) (cref 2))))
693 (4 (logior (ash (ldb (byte 3 0) (cref 0)) 18)
694 (ash (ldb (byte 6 0) (cref 1)) 12)
695 (ash (ldb (byte 6 0) (cref 2)) 6)
696 (ldb (byte 6 0) (cref 3)))))))))))
697 (instantiate-octets-definition define-simple-get-utf8-char)
699 (defmacro define-utf8->string (accessor type)
700 (let ((name (make-od-name 'utf8->string accessor)))
702 (defun ,name (array astart aend)
703 (declare (optimize speed (safety 0))
705 (type array-range astart aend))
706 (let ((string (make-array 0 :adjustable t :fill-pointer 0 :element-type 'character)))
707 (loop with pos = astart
709 do (multiple-value-bind (bytes invalid)
710 (,(make-od-name 'bytes-per-utf8-character accessor) array pos aend)
711 (declare (type (or null string) invalid))
714 (vector-push-extend (,(make-od-name 'simple-get-utf8-char accessor) array pos bytes) string))
716 (dotimes (i (length invalid))
717 (vector-push-extend (char invalid i) string))))
719 (coerce string 'simple-string))))))
720 (instantiate-octets-definition define-utf8->string)
722 ;;;; external formats
724 (defvar *default-external-format* nil)
726 (defun default-external-format ()
727 (or *default-external-format*
728 ;; On non-unicode, use iso-8859-1 instead of detecting it from
729 ;; the locale settings. Defaulting to an external-format which
730 ;; can represent characters that the CHARACTER type can't
731 ;; doesn't seem very sensible.
733 (setf *default-external-format* :latin-1)
734 (let ((external-format #!-win32 (intern (or (sb!alien:alien-funcall
737 (function (c-string :external-format :latin-1)
742 #!+win32 (sb!win32::ansi-codepage)))
743 (/show0 "cold-printing defaulted external-format:")
745 (cold-print external-format)
746 (/show0 "matching to known aliases")
747 (dolist (entry *external-formats*
749 ;;; FIXME! This WARN would try to do printing
750 ;;; before the streams have been initialized,
751 ;;; causing an infinite erroring loop. We should
752 ;;; either print it by calling to C, or delay the
753 ;;; warning until later. Since we're in freeze
754 ;;; right now, and the warning isn't really
755 ;;; essential, I'm doing what's least likely to
756 ;;; cause damage, and commenting it out. This
757 ;;; should be revisited after 0.9.17. -- JES,
760 (warn "Invalid external-format ~A; using LATIN-1"
762 (setf external-format :latin-1)))
763 (/show0 "cold printing known aliases:")
765 (dolist (alias (first entry)) (cold-print alias))
766 (/show0 "done cold-printing known aliases")
767 (when (member external-format (first entry))
770 (/show0 "/default external format ok")
771 (setf *default-external-format* external-format))))
773 ;;; FIXME: OAOOM here vrt. DEFINE-EXTERNAL-FORMAT in fd-stream.lisp
774 (defparameter *external-format-functions* (make-hash-table))
776 (defun add-external-format-funs (format-names funs)
777 (dolist (name format-names (values))
778 (setf (gethash name *external-format-functions*) funs)))
780 (add-external-format-funs
781 '(:ascii :us-ascii :ansi_x3.4-1968 :iso-646 :iso-646-us :|646|)
782 '(ascii->string-aref string->ascii))
783 (add-external-format-funs
784 '(:latin1 :latin-1 :iso-8859-1 :iso8859-1)
785 '(latin1->string-aref string->latin1))
787 (add-external-format-funs
788 '(:latin9 :latin-9 :iso-8859-15 :iso8859-15)
789 '(latin9->string-aref string->latin9))
790 (add-external-format-funs '(:utf8 :utf-8) '(utf8->string-aref string->utf8))
792 (defun external-formats-funs (external-format)
793 (when (eql external-format :default)
794 (setf external-format (default-external-format)))
795 (or (gethash external-format *external-format-functions*)
796 (error "Unknown external-format ~S" external-format)))
798 ;;;; public interface
800 (defun octets-to-string (vector &key (external-format :default) (start 0) end)
801 (declare (type (vector (unsigned-byte 8)) vector))
802 (with-array-data ((vector vector)
805 :check-fill-pointer t)
806 (declare (type (simple-array (unsigned-byte 8) (*)) vector))
807 (funcall (symbol-function (first (external-formats-funs external-format)))
810 (defun string-to-octets (string &key (external-format :default)
811 (start 0) end null-terminate)
812 (declare (type string string))
813 (with-array-data ((string string)
816 :check-fill-pointer t)
817 (declare (type simple-string string))
818 (funcall (symbol-function (second (external-formats-funs external-format)))
819 string start end (if null-terminate 1 0))))
822 (defvar +unicode-replacement-character+ (string (code-char #xfffd)))
824 (defun use-unicode-replacement-char (condition)
825 (use-value +unicode-replacement-character+ condition))
827 ;;; Utilities that maybe should be exported
830 (defmacro with-standard-replacement-character (&body body)
831 `(handler-bind ((octet-encoding-error #'use-unicode-replacement-char))
834 (defmacro with-default-decoding-replacement ((c) &body body)
835 (let ((cname (gensym)))
838 ((octet-decoding-error (lambda (c)
839 (use-value ,cname c))))