ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/JSON-XS/XS.xs
(Generate patch)

Comparing JSON-XS/XS.xs (file contents):
Revision 1.27 by root, Mon Apr 9 05:09:57 2007 UTC vs.
Revision 1.41 by root, Sat Jun 23 22:53:16 2007 UTC

… …
9 9
10#if defined(__BORLANDC__) || defined(_MSC_VER) 10#if defined(__BORLANDC__) || defined(_MSC_VER)
11# define snprintf _snprintf // C compilers have this in stdio.h 11# define snprintf _snprintf // C compilers have this in stdio.h
12#endif 12#endif
13 13
14// some old perls do not have this, try to make it work, no
15// guarentees, though.
16#ifndef UTF8_MAXBYTES
17# define UTF8_MAXBYTES 13
18#endif
19
14#define F_ASCII 0x00000001UL 20#define F_ASCII 0x00000001UL
21#define F_LATIN1 0x00000002UL
15#define F_UTF8 0x00000002UL 22#define F_UTF8 0x00000004UL
16#define F_INDENT 0x00000004UL 23#define F_INDENT 0x00000008UL
17#define F_CANONICAL 0x00000008UL 24#define F_CANONICAL 0x00000010UL
18#define F_SPACE_BEFORE 0x00000010UL 25#define F_SPACE_BEFORE 0x00000020UL
19#define F_SPACE_AFTER 0x00000020UL 26#define F_SPACE_AFTER 0x00000040UL
20#define F_ALLOW_NONREF 0x00000080UL 27#define F_ALLOW_NONREF 0x00000100UL
21#define F_SHRINK 0x00000100UL 28#define F_SHRINK 0x00000200UL
22#define F_MAXDEPTH 0xf8000000UL 29#define F_MAXDEPTH 0xf8000000UL
23#define S_MAXDEPTH 27 30#define S_MAXDEPTH 27
24 31
25#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) 32#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
26 33
… …
31#define F_DEFAULT (9UL << S_MAXDEPTH) 38#define F_DEFAULT (9UL << S_MAXDEPTH)
32 39
33#define INIT_SIZE 32 // initial scalar size to be allocated 40#define INIT_SIZE 32 // initial scalar size to be allocated
34#define INDENT_STEP 3 // spaces per indentation level 41#define INDENT_STEP 3 // spaces per indentation level
35 42
36#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
37#define SHORT_STRING_LEN 512 // special-case strings of up to this size 43#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38 44
39#define SB do { 45#define SB do {
40#define SE } while (0) 46#define SE } while (0)
47
48#if __GNUC__ >= 3
49# define expect(expr,value) __builtin_expect ((expr),(value))
50# define inline inline
51#else
52# define expect(expr,value) (expr)
53# define inline static
54#endif
55
56#define expect_false(expr) expect ((expr) != 0, 0)
57#define expect_true(expr) expect ((expr) != 0, 1)
41 58
42static HV *json_stash; // JSON::XS:: 59static HV *json_stash; // JSON::XS::
43 60
44///////////////////////////////////////////////////////////////////////////// 61/////////////////////////////////////////////////////////////////////////////
45// utility functions 62// utility functions
… …
70// decode an utf-8 character and return it, or (UV)-1 in 87// decode an utf-8 character and return it, or (UV)-1 in
71// case of an error. 88// case of an error.
72// we special-case "safe" characters from U+80 .. U+7FF, 89// we special-case "safe" characters from U+80 .. U+7FF,
73// but use the very good perl function to parse anything else. 90// but use the very good perl function to parse anything else.
74// note that we never call this function for a ascii codepoints 91// note that we never call this function for a ascii codepoints
75static UV 92inline UV
76decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 93decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77{ 94{
78 if (s[0] > 0xdf || s[0] < 0xc2) 95 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
79 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 96 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
80 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 97 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
81 { 98 {
82 *clen = 2; 99 *clen = 2;
83 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 100 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
… …
101 U32 flags; // F_* 118 U32 flags; // F_*
102 U32 indent; // indentation level 119 U32 indent; // indentation level
103 U32 maxdepth; // max. indentation/recursion level 120 U32 maxdepth; // max. indentation/recursion level
104} enc_t; 121} enc_t;
105 122
106static void 123inline void
107need (enc_t *enc, STRLEN len) 124need (enc_t *enc, STRLEN len)
108{ 125{
109 if (enc->cur + len >= enc->end) 126 if (expect_false (enc->cur + len >= enc->end))
110 { 127 {
111 STRLEN cur = enc->cur - SvPVX (enc->sv); 128 STRLEN cur = enc->cur - SvPVX (enc->sv);
112 SvGROW (enc->sv, cur + len + 1); 129 SvGROW (enc->sv, cur + len + 1);
113 enc->cur = SvPVX (enc->sv) + cur; 130 enc->cur = SvPVX (enc->sv) + cur;
114 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 131 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115 } 132 }
116} 133}
117 134
118static void 135inline void
119encode_ch (enc_t *enc, char ch) 136encode_ch (enc_t *enc, char ch)
120{ 137{
121 need (enc, 1); 138 need (enc, 1);
122 *enc->cur++ = ch; 139 *enc->cur++ = ch;
123} 140}
… …
131 148
132 while (str < end) 149 while (str < end)
133 { 150 {
134 unsigned char ch = *(unsigned char *)str; 151 unsigned char ch = *(unsigned char *)str;
135 152
136 if (ch >= 0x20 && ch < 0x80) // most common case 153 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137 { 154 {
138 if (ch == '"') // but with slow exceptions 155 if (expect_false (ch == '"')) // but with slow exceptions
139 { 156 {
140 need (enc, len += 1); 157 need (enc, len += 1);
141 *enc->cur++ = '\\'; 158 *enc->cur++ = '\\';
142 *enc->cur++ = '"'; 159 *enc->cur++ = '"';
143 } 160 }
144 else if (ch == '\\') 161 else if (expect_false (ch == '\\'))
145 { 162 {
146 need (enc, len += 1); 163 need (enc, len += 1);
147 *enc->cur++ = '\\'; 164 *enc->cur++ = '\\';
148 *enc->cur++ = '\\'; 165 *enc->cur++ = '\\';
149 } 166 }
… …
167 STRLEN clen; 184 STRLEN clen;
168 UV uch; 185 UV uch;
169 186
170 if (is_utf8) 187 if (is_utf8)
171 { 188 {
172 //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173 uch = decode_utf8 (str, end - str, &clen); 189 uch = decode_utf8 (str, end - str, &clen);
174 if (clen == (STRLEN)-1) 190 if (clen == (STRLEN)-1)
175 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); 191 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176 } 192 }
177 else 193 else
… …
181 } 197 }
182 198
183 if (uch > 0x10FFFFUL) 199 if (uch > 0x10FFFFUL)
184 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 200 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185 201
186 if (uch < 0x80 || enc->flags & F_ASCII) 202 if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
187 { 203 {
188 if (uch > 0xFFFFUL) 204 if (uch > 0xFFFFUL)
189 { 205 {
190 need (enc, len += 11); 206 need (enc, len += 11);
191 sprintf (enc->cur, "\\u%04x\\u%04x", 207 sprintf (enc->cur, "\\u%04x\\u%04x",
… …
205 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 221 *enc->cur++ = hexdigit [(uch >> 0) & 15];
206 } 222 }
207 223
208 str += clen; 224 str += clen;
209 } 225 }
226 else if (enc->flags & F_LATIN1)
227 {
228 *enc->cur++ = uch;
229 str += clen;
230 }
210 else if (is_utf8) 231 else if (is_utf8)
211 { 232 {
212 need (enc, len += clen); 233 need (enc, len += clen);
213 do 234 do
214 { 235 {
… …
216 } 237 }
217 while (--clen); 238 while (--clen);
218 } 239 }
219 else 240 else
220 { 241 {
221 need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed 242 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
222 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 243 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
223 ++str; 244 ++str;
224 } 245 }
225 } 246 }
226 } 247 }
… …
228 249
229 --len; 250 --len;
230 } 251 }
231} 252}
232 253
233static void 254inline void
234encode_indent (enc_t *enc) 255encode_indent (enc_t *enc)
235{ 256{
236 if (enc->flags & F_INDENT) 257 if (enc->flags & F_INDENT)
237 { 258 {
238 int spaces = enc->indent * INDENT_STEP; 259 int spaces = enc->indent * INDENT_STEP;
… …
241 memset (enc->cur, ' ', spaces); 262 memset (enc->cur, ' ', spaces);
242 enc->cur += spaces; 263 enc->cur += spaces;
243 } 264 }
244} 265}
245 266
246static void 267inline void
247encode_space (enc_t *enc) 268encode_space (enc_t *enc)
248{ 269{
249 need (enc, 1); 270 need (enc, 1);
250 encode_ch (enc, ' '); 271 encode_ch (enc, ' ');
251} 272}
252 273
253static void 274inline void
254encode_nl (enc_t *enc) 275encode_nl (enc_t *enc)
255{ 276{
256 if (enc->flags & F_INDENT) 277 if (enc->flags & F_INDENT)
257 { 278 {
258 need (enc, 1); 279 need (enc, 1);
259 encode_ch (enc, '\n'); 280 encode_ch (enc, '\n');
260 } 281 }
261} 282}
262 283
263static void 284inline void
264encode_comma (enc_t *enc) 285encode_comma (enc_t *enc)
265{ 286{
266 encode_ch (enc, ','); 287 encode_ch (enc, ',');
267 288
268 if (enc->flags & F_INDENT) 289 if (enc->flags & F_INDENT)
… …
367 // actually, this is mostly due to the stupid so-called 388 // actually, this is mostly due to the stupid so-called
368 // security workaround added somewhere in 5.8.x. 389 // security workaround added somewhere in 5.8.x.
369 // that randomises hash orderings 390 // that randomises hash orderings
370 if (enc->flags & F_CANONICAL) 391 if (enc->flags & F_CANONICAL)
371 { 392 {
372 HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode
373 int fast = 1; 393 int fast = 1;
394 HE *he;
395#if defined(__BORLANDC__) || defined(_MSC_VER)
396 HE **hes = _alloca (count * sizeof (HE));
397#else
398 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
399#endif
374 400
375 i = 0; 401 i = 0;
376 while ((he = hv_iternext (hv))) 402 while ((he = hv_iternext (hv)))
377 { 403 {
378 hes [i++] = he; 404 hes [i++] = he;
… …
476 encode_str (enc, str, len, SvUTF8 (sv)); 502 encode_str (enc, str, len, SvUTF8 (sv));
477 encode_ch (enc, '"'); 503 encode_ch (enc, '"');
478 } 504 }
479 else if (SvNOKp (sv)) 505 else if (SvNOKp (sv))
480 { 506 {
507 // trust that perl will do the right thing w.r.t. JSON syntax.
481 need (enc, NV_DIG + 32); 508 need (enc, NV_DIG + 32);
482 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 509 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
483 enc->cur += strlen (enc->cur); 510 enc->cur += strlen (enc->cur);
484 } 511 }
485 else if (SvIOKp (sv)) 512 else if (SvIOKp (sv))
486 { 513 {
487 need (enc, 64); 514 // we assume we can always read an IV as a UV
515 if (SvUV (sv) & ~(UV)0x7fff)
516 {
517 // large integer, use the (rather slow) snprintf way.
518 need (enc, sizeof (UV) * 3);
488 enc->cur += 519 enc->cur +=
489 SvIsUV(sv) 520 SvIsUV(sv)
490 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 521 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
491 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 522 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
523 }
524 else
525 {
526 // optimise the "small number case"
527 // code will likely be branchless and use only a single multiplication
528 I32 i = SvIV (sv);
529 U32 u;
530 char digit, nz = 0;
531
532 need (enc, 6);
533
534 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
535 u = i < 0 ? -i : i;
536
537 // convert to 4.28 fixed-point representation
538 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
539
540 // now output digit by digit, each time masking out the integer part
541 // and multiplying by 5 while moving the decimal point one to the right,
542 // resulting in a net multiplication by 10.
543 // we always write the digit to memory but conditionally increment
544 // the pointer, to ease the usage of conditional move instructions.
545 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
546 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
547 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
548 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
549 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
550 }
492 } 551 }
493 else if (SvROK (sv)) 552 else if (SvROK (sv))
494 encode_rv (enc, SvRV (sv)); 553 encode_rv (enc, SvRV (sv));
495 else if (!SvOK (sv)) 554 else if (!SvOK (sv))
496 encode_str (enc, "null", 4, 0); 555 encode_str (enc, "null", 4, 0);
… …
515 enc.maxdepth = DEC_DEPTH (flags); 574 enc.maxdepth = DEC_DEPTH (flags);
516 575
517 SvPOK_only (enc.sv); 576 SvPOK_only (enc.sv);
518 encode_sv (&enc, scalar); 577 encode_sv (&enc, scalar);
519 578
520 if (!(flags & (F_ASCII | F_UTF8)))
521 SvUTF8_on (enc.sv);
522
523 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 579 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
524 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 580 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
581
582 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
583 SvUTF8_on (enc.sv);
525 584
526 if (enc.flags & F_SHRINK) 585 if (enc.flags & F_SHRINK)
527 shrink (enc.sv); 586 shrink (enc.sv);
528 587
529 return enc.sv; 588 return enc.sv;
… …
541 U32 flags; // F_* 600 U32 flags; // F_*
542 U32 depth; // recursion depth 601 U32 depth; // recursion depth
543 U32 maxdepth; // recursion depth limit 602 U32 maxdepth; // recursion depth limit
544} dec_t; 603} dec_t;
545 604
546static void 605inline void
547decode_ws (dec_t *dec) 606decode_ws (dec_t *dec)
548{ 607{
549 for (;;) 608 for (;;)
550 { 609 {
551 char ch = *dec->cur; 610 char ch = *dec->cur;
… …
577decode_4hex (dec_t *dec) 636decode_4hex (dec_t *dec)
578{ 637{
579 signed char d1, d2, d3, d4; 638 signed char d1, d2, d3, d4;
580 unsigned char *cur = (unsigned char *)dec->cur; 639 unsigned char *cur = (unsigned char *)dec->cur;
581 640
582 d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); 641 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
583 d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); 642 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
584 d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); 643 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
585 d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); 644 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
586 645
587 dec->cur += 4; 646 dec->cur += 4;
588 647
589 return ((UV)d1) << 12 648 return ((UV)d1) << 12
590 | ((UV)d2) << 8 649 | ((UV)d2) << 8
… …
598static SV * 657static SV *
599decode_str (dec_t *dec) 658decode_str (dec_t *dec)
600{ 659{
601 SV *sv = 0; 660 SV *sv = 0;
602 int utf8 = 0; 661 int utf8 = 0;
662 char *dec_cur = dec->cur;
603 663
604 do 664 do
605 { 665 {
606 char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; 666 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
607 char *cur = buf; 667 char *cur = buf;
608 668
609 do 669 do
610 { 670 {
611 unsigned char ch = *(unsigned char *)dec->cur++; 671 unsigned char ch = *(unsigned char *)dec_cur++;
612 672
613 if (ch == '"') 673 if (expect_false (ch == '"'))
614 { 674 {
615 --dec->cur; 675 --dec_cur;
616 break; 676 break;
617 } 677 }
618 else if (ch == '\\') 678 else if (expect_false (ch == '\\'))
619 { 679 {
620 switch (*dec->cur) 680 switch (*dec_cur)
621 { 681 {
622 case '\\': 682 case '\\':
623 case '/': 683 case '/':
624 case '"': *cur++ = *dec->cur++; break; 684 case '"': *cur++ = *dec_cur++; break;
625 685
626 case 'b': ++dec->cur; *cur++ = '\010'; break; 686 case 'b': ++dec_cur; *cur++ = '\010'; break;
627 case 't': ++dec->cur; *cur++ = '\011'; break; 687 case 't': ++dec_cur; *cur++ = '\011'; break;
628 case 'n': ++dec->cur; *cur++ = '\012'; break; 688 case 'n': ++dec_cur; *cur++ = '\012'; break;
629 case 'f': ++dec->cur; *cur++ = '\014'; break; 689 case 'f': ++dec_cur; *cur++ = '\014'; break;
630 case 'r': ++dec->cur; *cur++ = '\015'; break; 690 case 'r': ++dec_cur; *cur++ = '\015'; break;
631 691
632 case 'u': 692 case 'u':
633 { 693 {
634 UV lo, hi; 694 UV lo, hi;
635 ++dec->cur; 695 ++dec_cur;
636 696
697 dec->cur = dec_cur;
637 hi = decode_4hex (dec); 698 hi = decode_4hex (dec);
699 dec_cur = dec->cur;
638 if (hi == (UV)-1) 700 if (hi == (UV)-1)
639 goto fail; 701 goto fail;
640 702
641 // possibly a surrogate pair 703 // possibly a surrogate pair
642 if (hi >= 0xd800) 704 if (hi >= 0xd800)
643 if (hi < 0xdc00) 705 if (hi < 0xdc00)
644 { 706 {
645 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 707 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
646 ERR ("missing low surrogate character in surrogate pair"); 708 ERR ("missing low surrogate character in surrogate pair");
647 709
648 dec->cur += 2; 710 dec_cur += 2;
649 711
712 dec->cur = dec_cur;
650 lo = decode_4hex (dec); 713 lo = decode_4hex (dec);
714 dec_cur = dec->cur;
651 if (lo == (UV)-1) 715 if (lo == (UV)-1)
652 goto fail; 716 goto fail;
653 717
654 if (lo < 0xdc00 || lo >= 0xe000) 718 if (lo < 0xdc00 || lo >= 0xe000)
655 ERR ("surrogate pair expected"); 719 ERR ("surrogate pair expected");
… …
669 *cur++ = hi; 733 *cur++ = hi;
670 } 734 }
671 break; 735 break;
672 736
673 default: 737 default:
674 --dec->cur; 738 --dec_cur;
675 ERR ("illegal backslash escape sequence in string"); 739 ERR ("illegal backslash escape sequence in string");
676 } 740 }
677 } 741 }
678 else if (ch >= 0x20 && ch <= 0x7f) 742 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
679 *cur++ = ch; 743 *cur++ = ch;
680 else if (ch >= 0x80) 744 else if (ch >= 0x80)
681 { 745 {
682 STRLEN clen; 746 STRLEN clen;
683 UV uch; 747 UV uch;
684 748
685 --dec->cur; 749 --dec_cur;
686 750
687 uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); 751 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
688 if (clen == (STRLEN)-1) 752 if (clen == (STRLEN)-1)
689 ERR ("malformed UTF-8 character in JSON string"); 753 ERR ("malformed UTF-8 character in JSON string");
690 754
691 do 755 do
692 *cur++ = *dec->cur++; 756 *cur++ = *dec_cur++;
693 while (--clen); 757 while (--clen);
694 758
695 utf8 = 1; 759 utf8 = 1;
696 } 760 }
697 else 761 else
698 { 762 {
699 --dec->cur; 763 --dec_cur;
700 764
701 if (!ch) 765 if (!ch)
702 ERR ("unexpected end of string while parsing JSON string"); 766 ERR ("unexpected end of string while parsing JSON string");
703 else 767 else
704 ERR ("invalid character encountered while parsing JSON string"); 768 ERR ("invalid character encountered while parsing JSON string");
… …
717 } 781 }
718 else 782 else
719 sv = newSVpvn (buf, len); 783 sv = newSVpvn (buf, len);
720 } 784 }
721 } 785 }
722 while (*dec->cur != '"'); 786 while (*dec_cur != '"');
723 787
724 ++dec->cur; 788 ++dec_cur;
725 789
726 if (sv) 790 if (sv)
727 { 791 {
728 SvPOK_only (sv); 792 SvPOK_only (sv);
729 *SvEND (sv) = 0; 793 *SvEND (sv) = 0;
… …
732 SvUTF8_on (sv); 796 SvUTF8_on (sv);
733 } 797 }
734 else 798 else
735 sv = newSVpvn ("", 0); 799 sv = newSVpvn ("", 0);
736 800
801 dec->cur = dec_cur;
737 return sv; 802 return sv;
738 803
739fail: 804fail:
805 dec->cur = dec_cur;
740 return 0; 806 return 0;
741} 807}
742 808
743static SV * 809static SV *
744decode_num (dec_t *dec) 810decode_num (dec_t *dec)
… …
802 is_nv = 1; 868 is_nv = 1;
803 } 869 }
804 870
805 if (!is_nv) 871 if (!is_nv)
806 { 872 {
807 UV uv; 873 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
808 int numtype = grok_number (start, dec->cur - start, &uv); 874 if (*start == '-')
809 if (numtype & IS_NUMBER_IN_UV) 875 switch (dec->cur - start)
810 if (numtype & IS_NUMBER_NEG)
811 { 876 {
812 if (uv < (UV)IV_MIN) 877 case 2: return newSViv (-( start [1] - '0' * 1));
813 return newSViv (-(IV)uv); 878 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
879 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
880 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
814 } 881 }
882 else
883 switch (dec->cur - start)
884 {
885 case 1: return newSViv ( start [0] - '0' * 1);
886 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
887 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
888 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
889 }
890
891 {
892 UV uv;
893 int numtype = grok_number (start, dec->cur - start, &uv);
894 if (numtype & IS_NUMBER_IN_UV)
895 if (numtype & IS_NUMBER_NEG)
896 {
897 if (uv < (UV)IV_MIN)
898 return newSViv (-(IV)uv);
899 }
815 else 900 else
816 return newSVuv (uv); 901 return newSVuv (uv);
902
903 // here would likely be the place for bigint support
817 } 904 }
905 }
818 906
907 // if we ever support bigint or bigfloat, this is the place for bigfloat
819 return newSVnv (Atof (start)); 908 return newSVnv (Atof (start));
820 909
821fail: 910fail:
822 return 0; 911 return 0;
823} 912}
… …
924 1013
925static SV * 1014static SV *
926decode_sv (dec_t *dec) 1015decode_sv (dec_t *dec)
927{ 1016{
928 decode_ws (dec); 1017 decode_ws (dec);
1018
1019 // the beauty of JSON: you need exactly one character lookahead
1020 // to parse anything.
929 switch (*dec->cur) 1021 switch (*dec->cur)
930 { 1022 {
931 case '"': ++dec->cur; return decode_str (dec); 1023 case '"': ++dec->cur; return decode_str (dec);
932 case '[': ++dec->cur; return decode_av (dec); 1024 case '[': ++dec->cur; return decode_av (dec);
933 case '{': ++dec->cur; return decode_hv (dec); 1025 case '{': ++dec->cur; return decode_hv (dec);
… …
978fail: 1070fail:
979 return 0; 1071 return 0;
980} 1072}
981 1073
982static SV * 1074static SV *
983decode_json (SV *string, U32 flags) 1075decode_json (SV *string, U32 flags, UV *offset_return)
984{ 1076{
985 dec_t dec; 1077 dec_t dec;
1078 UV offset;
986 SV *sv; 1079 SV *sv;
987 1080
1081 SvGETMAGIC (string);
988 SvUPGRADE (string, SVt_PV); 1082 SvUPGRADE (string, SVt_PV);
989 1083
990 if (flags & F_UTF8) 1084 if (flags & F_UTF8)
991 sv_utf8_downgrade (string, 0); 1085 sv_utf8_downgrade (string, 0);
992 else 1086 else
… …
999 dec.end = SvEND (string); 1093 dec.end = SvEND (string);
1000 dec.err = 0; 1094 dec.err = 0;
1001 dec.depth = 0; 1095 dec.depth = 0;
1002 dec.maxdepth = DEC_DEPTH (dec.flags); 1096 dec.maxdepth = DEC_DEPTH (dec.flags);
1003 1097
1004 *dec.end = 0; // this should basically be a nop, too, but make sure its there 1098 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1005 sv = decode_sv (&dec); 1099 sv = decode_sv (&dec);
1006 1100
1101 if (!(offset_return || !sv))
1102 {
1103 // check for trailing garbage
1104 decode_ws (&dec);
1105
1106 if (*dec.cur)
1107 {
1108 dec.err = "garbage after JSON object";
1109 SvREFCNT_dec (sv);
1110 sv = 0;
1111 }
1112 }
1113
1114 if (offset_return || !sv)
1115 {
1116 offset = dec.flags & F_UTF8
1117 ? dec.cur - SvPVX (string)
1118 : utf8_distance (dec.cur, SvPVX (string));
1119
1120 if (offset_return)
1121 *offset_return = offset;
1122 }
1123
1007 if (!sv) 1124 if (!sv)
1008 { 1125 {
1009 IV offset = dec.flags & F_UTF8
1010 ? dec.cur - SvPVX (string)
1011 : utf8_distance (dec.cur, SvPVX (string));
1012 SV *uni = sv_newmortal (); 1126 SV *uni = sv_newmortal ();
1013 1127
1014 // horrible hack to silence warning inside pv_uni_display 1128 // horrible hack to silence warning inside pv_uni_display
1015 COP cop = *PL_curcop; 1129 COP cop = *PL_curcop;
1016 cop.cop_warnings = pWARN_NONE; 1130 cop.cop_warnings = pWARN_NONE;
… …
1041 1155
1042BOOT: 1156BOOT:
1043{ 1157{
1044 int i; 1158 int i;
1045 1159
1046 memset (decode_hexdigit, 0xff, 256);
1047
1048 for (i = 0; i < 256; ++i) 1160 for (i = 0; i < 256; ++i)
1049 decode_hexdigit [i] = 1161 decode_hexdigit [i] =
1050 i >= '0' && i <= '9' ? i - '0' 1162 i >= '0' && i <= '9' ? i - '0'
1051 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1163 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1052 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1164 : i >= 'A' && i <= 'F' ? i - 'A' + 10
… …
1064 RETVAL 1176 RETVAL
1065 1177
1066SV *ascii (SV *self, int enable = 1) 1178SV *ascii (SV *self, int enable = 1)
1067 ALIAS: 1179 ALIAS:
1068 ascii = F_ASCII 1180 ascii = F_ASCII
1181 latin1 = F_LATIN1
1069 utf8 = F_UTF8 1182 utf8 = F_UTF8
1070 indent = F_INDENT 1183 indent = F_INDENT
1071 canonical = F_CANONICAL 1184 canonical = F_CANONICAL
1072 space_before = F_SPACE_BEFORE 1185 space_before = F_SPACE_BEFORE
1073 space_after = F_SPACE_AFTER 1186 space_after = F_SPACE_AFTER
… …
1109 PPCODE: 1222 PPCODE:
1110 XPUSHs (encode_json (scalar, *SvJSON (self))); 1223 XPUSHs (encode_json (scalar, *SvJSON (self)));
1111 1224
1112void decode (SV *self, SV *jsonstr) 1225void decode (SV *self, SV *jsonstr)
1113 PPCODE: 1226 PPCODE:
1114 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1227 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1228
1229void decode_prefix (SV *self, SV *jsonstr)
1230 PPCODE:
1231{
1232 UV offset;
1233 EXTEND (SP, 2);
1234 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1235 PUSHs (sv_2mortal (newSVuv (offset)));
1236}
1115 1237
1116PROTOTYPES: ENABLE 1238PROTOTYPES: ENABLE
1117 1239
1118void to_json (SV *scalar) 1240void to_json (SV *scalar)
1119 ALIAS: 1241 ALIAS:
… …
1123 1245
1124void from_json (SV *jsonstr) 1246void from_json (SV *jsonstr)
1125 ALIAS: 1247 ALIAS:
1126 jsonToObj = 0 1248 jsonToObj = 0
1127 PPCODE: 1249 PPCODE:
1128 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8)); 1250 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1129 1251

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines