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 1.32 by root, Wed May 9 16:41:12 2007 UTC vs.
Revision 1.44 by root, Mon Jun 25 04:08:17 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. if it breaks, you get to keep the pieces.
16#ifndef UTF8_MAXBYTES
17# define UTF8_MAXBYTES 13
18#endif
19
14#define F_ASCII 0x00000001UL 20#define F_ASCII 0x00000001UL
15#define F_LATIN1 0x00000002UL 21#define F_LATIN1 0x00000002UL
16#define F_UTF8 0x00000004UL 22#define F_UTF8 0x00000004UL
17#define F_INDENT 0x00000008UL 23#define F_INDENT 0x00000008UL
18#define F_CANONICAL 0x00000010UL 24#define F_CANONICAL 0x00000010UL
19#define F_SPACE_BEFORE 0x00000020UL 25#define F_SPACE_BEFORE 0x00000020UL
20#define F_SPACE_AFTER 0x00000040UL 26#define F_SPACE_AFTER 0x00000040UL
21#define F_ALLOW_NONREF 0x00000100UL 27#define F_ALLOW_NONREF 0x00000100UL
22#define F_SHRINK 0x00000200UL 28#define F_SHRINK 0x00000200UL
29#define F_ALLOW_BLESSED 0x00000400UL
30#define F_CONV_BLESSED 0x00000800UL // NYI
23#define F_MAXDEPTH 0xf8000000UL 31#define F_MAXDEPTH 0xf8000000UL
24#define S_MAXDEPTH 27 32#define S_MAXDEPTH 27
25 33
26#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) 34#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27
28// F_SELFCONVERT? <=> to_json/toJson
29// F_BLESSED? <=> { $__class__$ => }
30 35
31#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 36#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32#define F_DEFAULT (9UL << S_MAXDEPTH) 37#define F_DEFAULT (9UL << S_MAXDEPTH)
33 38
34#define INIT_SIZE 32 // initial scalar size to be allocated 39#define INIT_SIZE 32 // initial scalar size to be allocated
35#define INDENT_STEP 3 // spaces per indentation level 40#define INDENT_STEP 3 // spaces per indentation level
36 41
37#define SHORT_STRING_LEN 512 // special-case strings of up to this size 42#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38 43
39#define SB do { 44#define SB do {
40#define SE } while (0) 45#define SE } while (0)
41 46
47#if __GNUC__ >= 3
48# define expect(expr,value) __builtin_expect ((expr),(value))
49# define inline inline
50#else
51# define expect(expr,value) (expr)
52# define inline static
53#endif
54
55#define expect_false(expr) expect ((expr) != 0, 0)
56#define expect_true(expr) expect ((expr) != 0, 1)
57
42static HV *json_stash; // JSON::XS:: 58static HV *json_stash, *json_boolean_stash; // JSON::XS::
59static SV *json_true, *json_false;
43 60
44///////////////////////////////////////////////////////////////////////////// 61/////////////////////////////////////////////////////////////////////////////
45// utility functions 62// utility functions
46 63
47static UV * 64static UV *
… …
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
… …
233 249
234 --len; 250 --len;
235 } 251 }
236} 252}
237 253
238static void 254inline void
239encode_indent (enc_t *enc) 255encode_indent (enc_t *enc)
240{ 256{
241 if (enc->flags & F_INDENT) 257 if (enc->flags & F_INDENT)
242 { 258 {
243 int spaces = enc->indent * INDENT_STEP; 259 int spaces = enc->indent * INDENT_STEP;
… …
246 memset (enc->cur, ' ', spaces); 262 memset (enc->cur, ' ', spaces);
247 enc->cur += spaces; 263 enc->cur += spaces;
248 } 264 }
249} 265}
250 266
251static void 267inline void
252encode_space (enc_t *enc) 268encode_space (enc_t *enc)
253{ 269{
254 need (enc, 1); 270 need (enc, 1);
255 encode_ch (enc, ' '); 271 encode_ch (enc, ' ');
256} 272}
257 273
258static void 274inline void
259encode_nl (enc_t *enc) 275encode_nl (enc_t *enc)
260{ 276{
261 if (enc->flags & F_INDENT) 277 if (enc->flags & F_INDENT)
262 { 278 {
263 need (enc, 1); 279 need (enc, 1);
264 encode_ch (enc, '\n'); 280 encode_ch (enc, '\n');
265 } 281 }
266} 282}
267 283
268static void 284inline void
269encode_comma (enc_t *enc) 285encode_comma (enc_t *enc)
270{ 286{
271 encode_ch (enc, ','); 287 encode_ch (enc, ',');
272 288
273 if (enc->flags & F_INDENT) 289 if (enc->flags & F_INDENT)
… …
372 // actually, this is mostly due to the stupid so-called 388 // actually, this is mostly due to the stupid so-called
373 // security workaround added somewhere in 5.8.x. 389 // security workaround added somewhere in 5.8.x.
374 // that randomises hash orderings 390 // that randomises hash orderings
375 if (enc->flags & F_CANONICAL) 391 if (enc->flags & F_CANONICAL)
376 { 392 {
377 HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode
378 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
379 400
380 i = 0; 401 i = 0;
381 while ((he = hv_iternext (hv))) 402 while ((he = hv_iternext (hv)))
382 { 403 {
383 hes [i++] = he; 404 hes [i++] = he;
… …
447 svtype svt; 468 svtype svt;
448 469
449 SvGETMAGIC (sv); 470 SvGETMAGIC (sv);
450 svt = SvTYPE (sv); 471 svt = SvTYPE (sv);
451 472
473 if (expect_false (SvOBJECT (sv)))
474 {
475 if (SvSTASH (sv) == json_boolean_stash)
476 {
477 if (SvIV (sv) == 0)
478 encode_str (enc, "false", 5, 0);
479 else
480 encode_str (enc, "true", 4, 0);
481 }
482 else
483 {
484#if 0
485 if (0 && sv_derived_from (rv, "JSON::Literal"))
486 {
487 // not yet
488 }
489#endif
490 if (enc->flags & F_CONV_BLESSED)
491 {
492 // we re-bless the reference to get overload and other niceties right
493 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
494
495 if (to_json)
496 {
497 dSP;
498 ENTER;
499 SAVETMPS;
500 PUSHMARK (SP);
501 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
502
503 // calling with G_SCALAR ensures that we always get a 1 reutrn value
504 // check anyways.
505 PUTBACK;
506 assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
507 SPAGAIN;
508
509 encode_sv (enc, POPs);
510
511 FREETMPS;
512 LEAVE;
513 }
514 else if (enc->flags & F_ALLOW_BLESSED)
515 encode_str (enc, "null", 4, 0);
516 else
517 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
518 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
519 }
520 else if (enc->flags & F_ALLOW_BLESSED)
521 encode_str (enc, "null", 4, 0);
522 else
523 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
524 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
525 }
526 }
452 if (svt == SVt_PVHV) 527 else if (svt == SVt_PVHV)
453 encode_hv (enc, (HV *)sv); 528 encode_hv (enc, (HV *)sv);
454 else if (svt == SVt_PVAV) 529 else if (svt == SVt_PVAV)
455 encode_av (enc, (AV *)sv); 530 encode_av (enc, (AV *)sv);
456 else if (svt < SVt_PVAV) 531 else if (svt < SVt_PVAV)
457 { 532 {
… …
481 encode_str (enc, str, len, SvUTF8 (sv)); 556 encode_str (enc, str, len, SvUTF8 (sv));
482 encode_ch (enc, '"'); 557 encode_ch (enc, '"');
483 } 558 }
484 else if (SvNOKp (sv)) 559 else if (SvNOKp (sv))
485 { 560 {
561 // trust that perl will do the right thing w.r.t. JSON syntax.
486 need (enc, NV_DIG + 32); 562 need (enc, NV_DIG + 32);
487 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 563 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
488 enc->cur += strlen (enc->cur); 564 enc->cur += strlen (enc->cur);
489 } 565 }
490 else if (SvIOKp (sv)) 566 else if (SvIOKp (sv))
491 { 567 {
492 need (enc, 64); 568 // we assume we can always read an IV as a UV
569 if (SvUV (sv) & ~(UV)0x7fff)
570 {
571 // large integer, use the (rather slow) snprintf way.
572 need (enc, sizeof (UV) * 3);
493 enc->cur += 573 enc->cur +=
494 SvIsUV(sv) 574 SvIsUV(sv)
495 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 575 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
496 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 576 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
577 }
578 else
579 {
580 // optimise the "small number case"
581 // code will likely be branchless and use only a single multiplication
582 I32 i = SvIV (sv);
583 U32 u;
584 char digit, nz = 0;
585
586 need (enc, 6);
587
588 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
589 u = i < 0 ? -i : i;
590
591 // convert to 4.28 fixed-point representation
592 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
593
594 // now output digit by digit, each time masking out the integer part
595 // and multiplying by 5 while moving the decimal point one to the right,
596 // resulting in a net multiplication by 10.
597 // we always write the digit to memory but conditionally increment
598 // the pointer, to ease the usage of conditional move instructions.
599 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
600 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
601 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
602 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
603 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
604 }
497 } 605 }
498 else if (SvROK (sv)) 606 else if (SvROK (sv))
499 encode_rv (enc, SvRV (sv)); 607 encode_rv (enc, SvRV (sv));
500 else if (!SvOK (sv)) 608 else if (!SvOK (sv))
501 encode_str (enc, "null", 4, 0); 609 encode_str (enc, "null", 4, 0);
… …
546 U32 flags; // F_* 654 U32 flags; // F_*
547 U32 depth; // recursion depth 655 U32 depth; // recursion depth
548 U32 maxdepth; // recursion depth limit 656 U32 maxdepth; // recursion depth limit
549} dec_t; 657} dec_t;
550 658
551static void 659inline void
552decode_ws (dec_t *dec) 660decode_ws (dec_t *dec)
553{ 661{
554 for (;;) 662 for (;;)
555 { 663 {
556 char ch = *dec->cur; 664 char ch = *dec->cur;
… …
582decode_4hex (dec_t *dec) 690decode_4hex (dec_t *dec)
583{ 691{
584 signed char d1, d2, d3, d4; 692 signed char d1, d2, d3, d4;
585 unsigned char *cur = (unsigned char *)dec->cur; 693 unsigned char *cur = (unsigned char *)dec->cur;
586 694
587 d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); 695 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
588 d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); 696 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
589 d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); 697 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
590 d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); 698 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
591 699
592 dec->cur += 4; 700 dec->cur += 4;
593 701
594 return ((UV)d1) << 12 702 return ((UV)d1) << 12
595 | ((UV)d2) << 8 703 | ((UV)d2) << 8
… …
603static SV * 711static SV *
604decode_str (dec_t *dec) 712decode_str (dec_t *dec)
605{ 713{
606 SV *sv = 0; 714 SV *sv = 0;
607 int utf8 = 0; 715 int utf8 = 0;
716 char *dec_cur = dec->cur;
608 717
609 do 718 do
610 { 719 {
611 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; 720 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
612 char *cur = buf; 721 char *cur = buf;
613 722
614 do 723 do
615 { 724 {
616 unsigned char ch = *(unsigned char *)dec->cur++; 725 unsigned char ch = *(unsigned char *)dec_cur++;
617 726
618 if (ch == '"') 727 if (expect_false (ch == '"'))
619 { 728 {
620 --dec->cur; 729 --dec_cur;
621 break; 730 break;
622 } 731 }
623 else if (ch == '\\') 732 else if (expect_false (ch == '\\'))
624 { 733 {
625 switch (*dec->cur) 734 switch (*dec_cur)
626 { 735 {
627 case '\\': 736 case '\\':
628 case '/': 737 case '/':
629 case '"': *cur++ = *dec->cur++; break; 738 case '"': *cur++ = *dec_cur++; break;
630 739
631 case 'b': ++dec->cur; *cur++ = '\010'; break; 740 case 'b': ++dec_cur; *cur++ = '\010'; break;
632 case 't': ++dec->cur; *cur++ = '\011'; break; 741 case 't': ++dec_cur; *cur++ = '\011'; break;
633 case 'n': ++dec->cur; *cur++ = '\012'; break; 742 case 'n': ++dec_cur; *cur++ = '\012'; break;
634 case 'f': ++dec->cur; *cur++ = '\014'; break; 743 case 'f': ++dec_cur; *cur++ = '\014'; break;
635 case 'r': ++dec->cur; *cur++ = '\015'; break; 744 case 'r': ++dec_cur; *cur++ = '\015'; break;
636 745
637 case 'u': 746 case 'u':
638 { 747 {
639 UV lo, hi; 748 UV lo, hi;
640 ++dec->cur; 749 ++dec_cur;
641 750
751 dec->cur = dec_cur;
642 hi = decode_4hex (dec); 752 hi = decode_4hex (dec);
753 dec_cur = dec->cur;
643 if (hi == (UV)-1) 754 if (hi == (UV)-1)
644 goto fail; 755 goto fail;
645 756
646 // possibly a surrogate pair 757 // possibly a surrogate pair
647 if (hi >= 0xd800) 758 if (hi >= 0xd800)
648 if (hi < 0xdc00) 759 if (hi < 0xdc00)
649 { 760 {
650 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 761 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
651 ERR ("missing low surrogate character in surrogate pair"); 762 ERR ("missing low surrogate character in surrogate pair");
652 763
653 dec->cur += 2; 764 dec_cur += 2;
654 765
766 dec->cur = dec_cur;
655 lo = decode_4hex (dec); 767 lo = decode_4hex (dec);
768 dec_cur = dec->cur;
656 if (lo == (UV)-1) 769 if (lo == (UV)-1)
657 goto fail; 770 goto fail;
658 771
659 if (lo < 0xdc00 || lo >= 0xe000) 772 if (lo < 0xdc00 || lo >= 0xe000)
660 ERR ("surrogate pair expected"); 773 ERR ("surrogate pair expected");
… …
674 *cur++ = hi; 787 *cur++ = hi;
675 } 788 }
676 break; 789 break;
677 790
678 default: 791 default:
679 --dec->cur; 792 --dec_cur;
680 ERR ("illegal backslash escape sequence in string"); 793 ERR ("illegal backslash escape sequence in string");
681 } 794 }
682 } 795 }
683 else if (ch >= 0x20 && ch <= 0x7f) 796 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
684 *cur++ = ch; 797 *cur++ = ch;
685 else if (ch >= 0x80) 798 else if (ch >= 0x80)
686 { 799 {
687 STRLEN clen; 800 STRLEN clen;
688 UV uch; 801 UV uch;
689 802
690 --dec->cur; 803 --dec_cur;
691 804
692 uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); 805 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
693 if (clen == (STRLEN)-1) 806 if (clen == (STRLEN)-1)
694 ERR ("malformed UTF-8 character in JSON string"); 807 ERR ("malformed UTF-8 character in JSON string");
695 808
696 do 809 do
697 *cur++ = *dec->cur++; 810 *cur++ = *dec_cur++;
698 while (--clen); 811 while (--clen);
699 812
700 utf8 = 1; 813 utf8 = 1;
701 } 814 }
702 else 815 else
703 { 816 {
704 --dec->cur; 817 --dec_cur;
705 818
706 if (!ch) 819 if (!ch)
707 ERR ("unexpected end of string while parsing JSON string"); 820 ERR ("unexpected end of string while parsing JSON string");
708 else 821 else
709 ERR ("invalid character encountered while parsing JSON string"); 822 ERR ("invalid character encountered while parsing JSON string");
… …
722 } 835 }
723 else 836 else
724 sv = newSVpvn (buf, len); 837 sv = newSVpvn (buf, len);
725 } 838 }
726 } 839 }
727 while (*dec->cur != '"'); 840 while (*dec_cur != '"');
728 841
729 ++dec->cur; 842 ++dec_cur;
730 843
731 if (sv) 844 if (sv)
732 { 845 {
733 SvPOK_only (sv); 846 SvPOK_only (sv);
734 *SvEND (sv) = 0; 847 *SvEND (sv) = 0;
… …
737 SvUTF8_on (sv); 850 SvUTF8_on (sv);
738 } 851 }
739 else 852 else
740 sv = newSVpvn ("", 0); 853 sv = newSVpvn ("", 0);
741 854
855 dec->cur = dec_cur;
742 return sv; 856 return sv;
743 857
744fail: 858fail:
859 dec->cur = dec_cur;
745 return 0; 860 return 0;
746} 861}
747 862
748static SV * 863static SV *
749decode_num (dec_t *dec) 864decode_num (dec_t *dec)
… …
807 is_nv = 1; 922 is_nv = 1;
808 } 923 }
809 924
810 if (!is_nv) 925 if (!is_nv)
811 { 926 {
812 UV uv; 927 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
813 int numtype = grok_number (start, dec->cur - start, &uv); 928 if (*start == '-')
814 if (numtype & IS_NUMBER_IN_UV) 929 switch (dec->cur - start)
815 if (numtype & IS_NUMBER_NEG)
816 { 930 {
817 if (uv < (UV)IV_MIN) 931 case 2: return newSViv (-( start [1] - '0' * 1));
818 return newSViv (-(IV)uv); 932 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
933 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
934 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
819 } 935 }
936 else
937 switch (dec->cur - start)
938 {
939 case 1: return newSViv ( start [0] - '0' * 1);
940 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
941 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
942 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
943 }
944
945 {
946 UV uv;
947 int numtype = grok_number (start, dec->cur - start, &uv);
948 if (numtype & IS_NUMBER_IN_UV)
949 if (numtype & IS_NUMBER_NEG)
950 {
951 if (uv < (UV)IV_MIN)
952 return newSViv (-(IV)uv);
953 }
820 else 954 else
821 return newSVuv (uv); 955 return newSVuv (uv);
956
957 // here would likely be the place for bigint support
822 } 958 }
959 }
823 960
961 // if we ever support bigint or bigfloat, this is the place for bigfloat
824 return newSVnv (Atof (start)); 962 return newSVnv (Atof (start));
825 963
826fail: 964fail:
827 return 0; 965 return 0;
828} 966}
… …
929 1067
930static SV * 1068static SV *
931decode_sv (dec_t *dec) 1069decode_sv (dec_t *dec)
932{ 1070{
933 decode_ws (dec); 1071 decode_ws (dec);
1072
1073 // the beauty of JSON: you need exactly one character lookahead
1074 // to parse anything.
934 switch (*dec->cur) 1075 switch (*dec->cur)
935 { 1076 {
936 case '"': ++dec->cur; return decode_str (dec); 1077 case '"': ++dec->cur; return decode_str (dec);
937 case '[': ++dec->cur; return decode_av (dec); 1078 case '[': ++dec->cur; return decode_av (dec);
938 case '{': ++dec->cur; return decode_hv (dec); 1079 case '{': ++dec->cur; return decode_hv (dec);
… …
944 1085
945 case 't': 1086 case 't':
946 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1087 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
947 { 1088 {
948 dec->cur += 4; 1089 dec->cur += 4;
949 return newSViv (1); 1090 return SvREFCNT_inc (json_true);
950 } 1091 }
951 else 1092 else
952 ERR ("'true' expected"); 1093 ERR ("'true' expected");
953 1094
954 break; 1095 break;
955 1096
956 case 'f': 1097 case 'f':
957 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1098 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
958 { 1099 {
959 dec->cur += 5; 1100 dec->cur += 5;
960 return newSViv (0); 1101 return SvREFCNT_inc (json_false);
961 } 1102 }
962 else 1103 else
963 ERR ("'false' expected"); 1104 ERR ("'false' expected");
964 1105
965 break; 1106 break;
… …
1068 1209
1069BOOT: 1210BOOT:
1070{ 1211{
1071 int i; 1212 int i;
1072 1213
1073 memset (decode_hexdigit, 0xff, 256);
1074
1075 for (i = 0; i < 256; ++i) 1214 for (i = 0; i < 256; ++i)
1076 decode_hexdigit [i] = 1215 decode_hexdigit [i] =
1077 i >= '0' && i <= '9' ? i - '0' 1216 i >= '0' && i <= '9' ? i - '0'
1078 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1217 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1079 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1218 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1080 : -1; 1219 : -1;
1081 1220
1082 json_stash = gv_stashpv ("JSON::XS", 1); 1221 json_stash = gv_stashpv ("JSON::XS" , 1);
1222 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1223
1224 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1225 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1083} 1226}
1084 1227
1085PROTOTYPES: DISABLE 1228PROTOTYPES: DISABLE
1086 1229
1087SV *new (char *dummy) 1230SV *new (char *dummy)
… …
1090 OUTPUT: 1233 OUTPUT:
1091 RETVAL 1234 RETVAL
1092 1235
1093SV *ascii (SV *self, int enable = 1) 1236SV *ascii (SV *self, int enable = 1)
1094 ALIAS: 1237 ALIAS:
1095 ascii = F_ASCII 1238 ascii = F_ASCII
1096 latin1 = F_LATIN1 1239 latin1 = F_LATIN1
1097 utf8 = F_UTF8 1240 utf8 = F_UTF8
1098 indent = F_INDENT 1241 indent = F_INDENT
1099 canonical = F_CANONICAL 1242 canonical = F_CANONICAL
1100 space_before = F_SPACE_BEFORE 1243 space_before = F_SPACE_BEFORE
1101 space_after = F_SPACE_AFTER 1244 space_after = F_SPACE_AFTER
1102 pretty = F_PRETTY 1245 pretty = F_PRETTY
1103 allow_nonref = F_ALLOW_NONREF 1246 allow_nonref = F_ALLOW_NONREF
1104 shrink = F_SHRINK 1247 shrink = F_SHRINK
1248 allow_blessed = F_ALLOW_BLESSED
1249 convert_blessed = F_CONV_BLESSED
1105 CODE: 1250 CODE:
1106{ 1251{
1107 UV *uv = SvJSON (self); 1252 UV *uv = SvJSON (self);
1108 if (enable) 1253 if (enable)
1109 *uv |= ix; 1254 *uv |= ix;

Diff Legend

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