X-Git-Url: http://repo.macrolet.net/gitweb/?a=blobdiff_plain;f=src%2Fcode%2Ftarget-char.lisp;h=55b4bcc9074f4e99758f785910aa792d8d363dd0;hb=09a00b3120e7dd6d040cf70fbaaa1af32b890ee3;hp=0c2b5f7568ce277026bd3a6ee096936ec63dc05c;hpb=026be5a30130cdb1bc4648fa7daea8e1180a6e46;p=sbcl.git diff --git a/src/code/target-char.lisp b/src/code/target-char.lisp index 0c2b5f7..55b4bcc 100644 --- a/src/code/target-char.lisp +++ b/src/code/target-char.lisp @@ -51,6 +51,24 @@ (defglobal **character-database** ,character-database) (defglobal **character-decompositions** ,decompositions) (defglobal **character-long-decompositions** ,long-decompositions) + (defglobal **character-primary-compositions** + (let ((table (make-hash-table)) + (info ,(read-ub8-vector (file "comp" "dat")))) + (flet ((code (j) + (dpb (aref info (* 4 j)) + (byte 8 24) + (dpb (aref info (+ (* 4 j) 1)) + (byte 8 16) + (dpb (aref info (+ (* 4 j) 2)) + (byte 8 8) + (aref info (+ (* 4 j) 3))))))) + #!+sb-unicode + (dotimes (i (/ (length info) 12)) + (setf (gethash (dpb (code (* 3 i)) (byte 21 21) + (code (1+ (* 3 i)))) + table) + (code-char (code (+ (* 3 i) 2))))) + table))) (defun !character-database-cold-init () (setf **character-database** ,character-database)) ,(with-open-file (stream (file "ucd-names" "lisp-expr") @@ -167,7 +185,7 @@ ;;;; UCD accessor functions -;;; The first (* 8 395) => 3160 entries in **CHARACTER-DATABASE** +;;; The first (* 8 396) => 3168 entries in **CHARACTER-DATABASE** ;;; contain entries for the distinct character attributes: ;;; specifically, indexes into the GC kinds, Bidi kinds, CCC kinds, ;;; the decimal digit property, the digit property and the @@ -194,12 +212,12 @@ ;;; ;;; To look up information about a character, take the high 13 bits of ;;; its code point, and index the character database with that and a -;;; base of 3160 (going past the miscellaneous information[*], so +;;; base of 3168 (going past the miscellaneous information[*], so ;;; treating (a) as the start of the array). This, labelled A, gives ;;; us another index into the detailed pages[-], which we can use to ;;; look up the details for the character in question: we add the low ;;; 8 bits of the character, shifted twice (because we have four-byte -;;; table entries) to 1024 times the `page' index, with a base of 6088 +;;; table entries) to 1024 times the `page' index, with a base of 7520 ;;; to skip over everything else. This gets us to point B. If we're ;;; after a transformed code point (i.e. an upcase or downcase ;;; operation), we can simply read it off now, beginning with an @@ -224,8 +242,8 @@ (defun ucd-index (char) (let* ((cp (char-code char)) (cp-high (ash cp -8)) - (page (aref **character-database** (+ 3160 cp-high)))) - (+ 7512 (ash page 10) (ash (ldb (byte 8 0) cp) 2)))) + (page (aref **character-database** (+ 3168 cp-high)))) + (+ 7520 (ash page 10) (ash (ldb (byte 8 0) cp) 2)))) (declaim (ftype (sfunction (t) (unsigned-byte 11)) ucd-value-0)) (defun ucd-value-0 (char) @@ -254,6 +272,9 @@ (+ 3 (* 8 (ucd-value-0 char)))))) (when (< decimal-digit 10) decimal-digit))) +(declaim (ftype (sfunction (t) (unsigned-byte 8)) ucd-ccc)) +(defun ucd-ccc (char) + (aref **character-database** (+ 2 (* 8 (ucd-value-0 char))))) (defun char-code (char) #!+sb-doc @@ -379,20 +400,20 @@ argument is an alphabetic character, A-Z or a-z; otherwise NIL." #!+sb-doc "The argument must be a character object; UPPER-CASE-P returns T if the argument is an upper-case character, NIL otherwise." - (< (ucd-value-0 char) 4)) + (< (ucd-value-0 char) 5)) (defun lower-case-p (char) #!+sb-doc "The argument must be a character object; LOWER-CASE-P returns T if the argument is a lower-case character, NIL otherwise." - (< 3 (ucd-value-0 char) 8)) + (< 4 (ucd-value-0 char) 9)) (defun both-case-p (char) #!+sb-doc "The argument must be a character object. BOTH-CASE-P returns T if the argument is an alphabetic character and if the character exists in both upper and lower case. For ASCII, this is the same as ALPHA-CHAR-P." - (< (ucd-value-0 char) 8)) + (< (ucd-value-0 char) 9)) (defun digit-char-p (char &optional (radix 10.)) #!+sb-doc @@ -492,7 +513,7 @@ is either numeric or alphabetic." (defmacro equal-char-code (character) (let ((ch (gensym))) `(let ((,ch ,character)) - (if (= (ucd-value-0 ,ch) 0) + (if (< (ucd-value-0 ,ch) 5) (ucd-value-1 ,ch) (char-code ,ch))))) @@ -588,14 +609,14 @@ Case is ignored." #!+sb-doc "Return CHAR converted to upper-case if that is possible. Don't convert lowercase eszet (U+DF)." - (if (< 3 (ucd-value-0 char) 8) + (if (< 4 (ucd-value-0 char) 9) (code-char (ucd-value-1 char)) char)) (defun char-downcase (char) #!+sb-doc "Return CHAR converted to lower-case if that is possible." - (if (< (ucd-value-0 char) 4) + (if (< (ucd-value-0 char) 5) (code-char (ucd-value-1 char)) char)) @@ -617,7 +638,7 @@ character exists." (decompositions **character-decompositions**) (long-decompositions **character-long-decompositions**) (index (+ #x1100 - (ash (aref decompositions cp-high) 10) + (ash (aref decompositions cp-high) 10) (ash (ldb (byte 8 0) cp) 2))) (v0 (aref decompositions index)) (v1 (aref decompositions (+ index 1))) @@ -628,18 +649,213 @@ character exists." (dpb v2 (byte 8 8) v3)))) (if (= length 1) (string (code-char entry)) - (let ((result (make-string length)) - (e (* 4 entry))) - (dotimes (i length result) - (let ((code (dpb (aref long-decompositions (+ e 1)) - (byte 8 16) - (dpb (aref long-decompositions (+ e 2)) - (byte 8 8) - (aref long-decompositions (+ e 3)))))) - (setf (char result i) (code-char code))) - (incf e 4)))))) + (if (<= #xac00 cp #xd7a3) + ;; see Unicode 6.2, section 3-12 + (let* ((sbase #xac00) + (lbase #x1100) + (vbase #x1161) + (tbase #x11a7) + (lcount 19) + (vcount 21) + (tcount 28) + (ncount (* vcount tcount)) + (scount (* lcount ncount)) + (sindex (- cp sbase)) + (lindex (floor sindex ncount)) + (vindex (floor (mod sindex ncount) tcount)) + (tindex (mod sindex tcount)) + (result (make-string length))) + (declare (ignore scount)) + (setf (char result 0) (code-char (+ lbase lindex))) + (setf (char result 1) (code-char (+ vbase vindex))) + (when (> tindex 0) + (setf (char result 2) (code-char (+ tbase tindex)))) + result) + (let ((result (make-string length)) + (e (* 4 entry))) + (dotimes (i length result) + (let ((code (dpb (aref long-decompositions (+ e 1)) + (byte 8 16) + (dpb (aref long-decompositions (+ e 2)) + (byte 8 8) + (aref long-decompositions (+ e 3)))))) + (setf (char result i) (code-char code))) + (incf e 4))))))) (defun decompose-char (char) (if (= (char-decomposition-info char) 0) (string char) (char-decomposition char))) + +(defun decompose-string (string &optional (kind :canonical)) + (declare (type (member :canonical :compatibility) kind)) + (flet ((canonical (char) + (= 1 (char-decomposition-info char))) + (compat (char) + (/= 0 (char-decomposition-info char)))) + (let (result + (fun (ecase kind + (:canonical #'canonical) + (:compatibility #'compat)))) + (do* ((start 0 (1+ end)) + (end (position-if fun string :start start) + (position-if fun string :start start))) + ((null end) (push (subseq string start end) result)) + (unless (= start end) + (push (subseq string start end) result)) + ;; FIXME: this recursive call to DECOMPOSE-STRING is necessary + ;; for correctness given our direct encoding of the + ;; decomposition data in UnicodeData.txt. It would, however, + ;; be straightforward enough to perform the recursion in table + ;; construction, and then have this simply revert to a single + ;; lookup. (Wait for tests to be hooked in, then implement). + (push (decompose-string (decompose-char (char string end)) kind) + result)) + (apply 'concatenate 'string (nreverse result))))) + +(defun sort-combiners (string) + (let (result (start 0) first-cc first-non-cc) + (tagbody + again + (setf first-cc (position 0 string :key #'ucd-ccc :test #'/= :start start)) + (when first-cc + (setf first-non-cc (position 0 string :key #'ucd-ccc :test #'= :start first-cc))) + (push (subseq string start first-cc) result) + (when first-cc + (push (stable-sort (subseq string first-cc first-non-cc) #'< :key #'ucd-ccc) result)) + (when first-non-cc + (setf start first-non-cc first-cc nil first-non-cc nil) + (go again))) + (apply 'concatenate 'string (nreverse result)))) + +(defun primary-composition (char1 char2) + (let ((c1 (char-code char1)) + (c2 (char-code char2))) + (cond + ((gethash (dpb (char-code char1) (byte 21 21) (char-code char2)) + **character-primary-compositions**)) + ((and (<= #x1100 c1) (<= c1 #x1112) + (<= #x1161 c2) (<= c2 #x1175)) + (let ((lindex (- c1 #x1100)) + (vindex (- c2 #x1161))) + (code-char (+ #xac00 (* lindex 588) (* vindex 28))))) + ((and (<= #xac00 c1) (<= c1 #.(+ #xac00 11171)) + (<= #x11a8 c2) (<= c2 #x11c2) + (= 0 (rem (- c1 #xac00) 28))) + (code-char (+ c1 (- c2 #x11a7))))))) + +;;; This implements a sequence data structure, specialized for +;;; efficient deletion of characters at an index, along with tolerable +;;; random access. The purpose is to support the canonical +;;; composition algorithm from Unicode, which involves replacing (not +;;; necessarily consecutive) pairs of code points with a single code +;;; point (e.g. [#\e #\combining_acute_accent] with +;;; #\latin_small_letter_e_with_acute). The data structure is a list +;;; of three-element lists, each denoting a chunk of string data +;;; starting at the first index and ending at the second. +;;; +;;; Actually, the implementation isn't particularly efficient, and +;;; would probably benefit from being rewritten in terms of displaced +;;; arrays, which would substantially reduce copying. +;;; +;;; (also, generic sequences. *sigh*.) +(defun lref (lstring index) + (dolist (l lstring) + (when (and (<= (first l) index) + (< index (second l))) + (return (aref (third l) (- index (first l))))))) +(defun (setf lref) (newchar lstring index) + (dolist (l lstring) + (when (and (<= (first l) index) + (< index (second l))) + (return (setf (aref (third l) (- index (first l))) newchar))))) +(defun llength (lstring) + (second (first (last lstring)))) +(defun lstring (lstring) + (let ((result (make-string (llength lstring)))) + (dolist (l lstring result) + (replace result (third l) :start1 (first l) :end1 (second l))))) +(defun ldelete (lstring index) + (do* ((ls lstring (cdr ls)) + (l (car ls) (car ls)) + so-fars) + ((and (<= (first l) index) + (< index (second l))) + (append + (nreverse so-fars) + (cond + ((= (first l) index) + (list (list (first l) (1- (second l)) (subseq (third l) 1)))) + ((= index (1- (second l))) + (list (list (first l) (1- (second l)) (subseq (third l) 0 (1- (length (third l))))))) + (t + (list + (list (first l) index + (subseq (third l) 0 (- index (first l)))) + (list index (1- (second l)) + (subseq (third l) (1+ (- index (first l)))))))) + (mapcar (lambda (x) (list (1- (first x)) (1- (second x)) (third x))) + (cdr ls)))) + (push l so-fars))) + +(defun canonically-compose (string) + (labels () + (let* ((result (list (list 0 (length string) string))) + (previous-starter-index (position 0 string :key #'ucd-ccc)) + (i (and previous-starter-index (1+ previous-starter-index)))) + (when (or (not i) (= i (length string))) + (return-from canonically-compose string)) + (tagbody + again + (when (and (>= (- i previous-starter-index) 2) + ;; test for Blocked (Unicode 3.11 para. D115) + ;; + ;; (assumes here that string has sorted combiners, + ;; so can look back just one step) + (>= (ucd-ccc (lref result (1- i))) + (ucd-ccc (lref result i)))) + (when (= (ucd-ccc (lref result i)) 0) + (setf previous-starter-index i)) + (incf i) + (go next)) + + (let ((comp (primary-composition (lref result previous-starter-index) + (lref result i)))) + (cond + (comp + (setf (lref result previous-starter-index) comp) + (setf result (ldelete result i))) + (t + (when (= (ucd-ccc (lref result i)) 0) + (setf previous-starter-index i)) + (incf i)))) + next + (unless (= i (llength result)) + (go again))) + (if (= i (length string)) + string + (lstring result))))) + +(defun normalize-string (string &optional (form :nfd)) + (declare (type (member :nfd :nfkd :nfc :nfkc) form)) + #!-sb-unicode + (etypecase string + ((array nil (*)) string) + (string + (ecase form + ((:nfc :nfkc) string) + ((:nfd :nfkd) (error "Cannot normalize to ~A form in #-SB-UNICODE builds" form))))) + #!+sb-unicode + (etypecase string + (base-string string) + ((array character (*)) + (ecase form + ((:nfc) + (canonically-compose (sort-combiners (decompose-string string)))) + ((:nfd) + (sort-combiners (decompose-string string))) + ((:nfkc) + (canonically-compose (sort-combiners (decompose-string string :compatibility)))) + ((:nfkd) + (sort-combiners (decompose-string string :compatibility))))) + ((array nil (*)) string)))