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 1.45 by root, Mon Jun 25 06:57:42 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
33#define F_MAXSIZE 0x01f00000UL
34#define S_MAXSIZE 20
25 35
26#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) 36#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27 37#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
28// F_SELFCONVERT? <=> to_json/toJson
29// F_BLESSED? <=> { $__class__$ => }
30 38
31#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 39#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32#define F_DEFAULT (9UL << S_MAXDEPTH) 40#define F_DEFAULT (9UL << S_MAXDEPTH)
33 41
34#define INIT_SIZE 32 // initial scalar size to be allocated 42#define INIT_SIZE 32 // initial scalar size to be allocated
35#define INDENT_STEP 3 // spaces per indentation level 43#define INDENT_STEP 3 // spaces per indentation level
36 44
37#define SHORT_STRING_LEN 512 // special-case strings of up to this size 45#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38 46
39#define SB do { 47#define SB do {
40#define SE } while (0) 48#define SE } while (0)
41 49
50#if __GNUC__ >= 3
51# define expect(expr,value) __builtin_expect ((expr),(value))
52# define inline inline
53#else
54# define expect(expr,value) (expr)
55# define inline static
56#endif
57
58#define expect_false(expr) expect ((expr) != 0, 0)
59#define expect_true(expr) expect ((expr) != 0, 1)
60
42static HV *json_stash; // JSON::XS:: 61static HV *json_stash, *json_boolean_stash; // JSON::XS::
62static SV *json_true, *json_false;
43 63
44///////////////////////////////////////////////////////////////////////////// 64/////////////////////////////////////////////////////////////////////////////
45// utility functions 65// utility functions
46 66
47static UV * 67static UV *
70// decode an utf-8 character and return it, or (UV)-1 in 90// decode an utf-8 character and return it, or (UV)-1 in
71// case of an error. 91// case of an error.
72// we special-case "safe" characters from U+80 .. U+7FF, 92// we special-case "safe" characters from U+80 .. U+7FF,
73// but use the very good perl function to parse anything else. 93// but use the very good perl function to parse anything else.
74// note that we never call this function for a ascii codepoints 94// note that we never call this function for a ascii codepoints
75static UV 95inline UV
76decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 96decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77{ 97{
78 if (s[0] > 0xdf || s[0] < 0xc2) 98 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
79 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 99 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
80 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 100 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
81 { 101 {
82 *clen = 2; 102 *clen = 2;
83 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 103 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
101 U32 flags; // F_* 121 U32 flags; // F_*
102 U32 indent; // indentation level 122 U32 indent; // indentation level
103 U32 maxdepth; // max. indentation/recursion level 123 U32 maxdepth; // max. indentation/recursion level
104} enc_t; 124} enc_t;
105 125
106static void 126inline void
107need (enc_t *enc, STRLEN len) 127need (enc_t *enc, STRLEN len)
108{ 128{
109 if (enc->cur + len >= enc->end) 129 if (expect_false (enc->cur + len >= enc->end))
110 { 130 {
111 STRLEN cur = enc->cur - SvPVX (enc->sv); 131 STRLEN cur = enc->cur - SvPVX (enc->sv);
112 SvGROW (enc->sv, cur + len + 1); 132 SvGROW (enc->sv, cur + len + 1);
113 enc->cur = SvPVX (enc->sv) + cur; 133 enc->cur = SvPVX (enc->sv) + cur;
114 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 134 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115 } 135 }
116} 136}
117 137
118static void 138inline void
119encode_ch (enc_t *enc, char ch) 139encode_ch (enc_t *enc, char ch)
120{ 140{
121 need (enc, 1); 141 need (enc, 1);
122 *enc->cur++ = ch; 142 *enc->cur++ = ch;
123} 143}
131 151
132 while (str < end) 152 while (str < end)
133 { 153 {
134 unsigned char ch = *(unsigned char *)str; 154 unsigned char ch = *(unsigned char *)str;
135 155
136 if (ch >= 0x20 && ch < 0x80) // most common case 156 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137 { 157 {
138 if (ch == '"') // but with slow exceptions 158 if (expect_false (ch == '"')) // but with slow exceptions
139 { 159 {
140 need (enc, len += 1); 160 need (enc, len += 1);
141 *enc->cur++ = '\\'; 161 *enc->cur++ = '\\';
142 *enc->cur++ = '"'; 162 *enc->cur++ = '"';
143 } 163 }
144 else if (ch == '\\') 164 else if (expect_false (ch == '\\'))
145 { 165 {
146 need (enc, len += 1); 166 need (enc, len += 1);
147 *enc->cur++ = '\\'; 167 *enc->cur++ = '\\';
148 *enc->cur++ = '\\'; 168 *enc->cur++ = '\\';
149 } 169 }
167 STRLEN clen; 187 STRLEN clen;
168 UV uch; 188 UV uch;
169 189
170 if (is_utf8) 190 if (is_utf8)
171 { 191 {
172 //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173 uch = decode_utf8 (str, end - str, &clen); 192 uch = decode_utf8 (str, end - str, &clen);
174 if (clen == (STRLEN)-1) 193 if (clen == (STRLEN)-1)
175 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); 194 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176 } 195 }
177 else 196 else
233 252
234 --len; 253 --len;
235 } 254 }
236} 255}
237 256
238static void 257inline void
239encode_indent (enc_t *enc) 258encode_indent (enc_t *enc)
240{ 259{
241 if (enc->flags & F_INDENT) 260 if (enc->flags & F_INDENT)
242 { 261 {
243 int spaces = enc->indent * INDENT_STEP; 262 int spaces = enc->indent * INDENT_STEP;
246 memset (enc->cur, ' ', spaces); 265 memset (enc->cur, ' ', spaces);
247 enc->cur += spaces; 266 enc->cur += spaces;
248 } 267 }
249} 268}
250 269
251static void 270inline void
252encode_space (enc_t *enc) 271encode_space (enc_t *enc)
253{ 272{
254 need (enc, 1); 273 need (enc, 1);
255 encode_ch (enc, ' '); 274 encode_ch (enc, ' ');
256} 275}
257 276
258static void 277inline void
259encode_nl (enc_t *enc) 278encode_nl (enc_t *enc)
260{ 279{
261 if (enc->flags & F_INDENT) 280 if (enc->flags & F_INDENT)
262 { 281 {
263 need (enc, 1); 282 need (enc, 1);
264 encode_ch (enc, '\n'); 283 encode_ch (enc, '\n');
265 } 284 }
266} 285}
267 286
268static void 287inline void
269encode_comma (enc_t *enc) 288encode_comma (enc_t *enc)
270{ 289{
271 encode_ch (enc, ','); 290 encode_ch (enc, ',');
272 291
273 if (enc->flags & F_INDENT) 292 if (enc->flags & F_INDENT)
372 // actually, this is mostly due to the stupid so-called 391 // actually, this is mostly due to the stupid so-called
373 // security workaround added somewhere in 5.8.x. 392 // security workaround added somewhere in 5.8.x.
374 // that randomises hash orderings 393 // that randomises hash orderings
375 if (enc->flags & F_CANONICAL) 394 if (enc->flags & F_CANONICAL)
376 { 395 {
377 HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode
378 int fast = 1; 396 int fast = 1;
397 HE *he;
398#if defined(__BORLANDC__) || defined(_MSC_VER)
399 HE **hes = _alloca (count * sizeof (HE));
400#else
401 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
402#endif
379 403
380 i = 0; 404 i = 0;
381 while ((he = hv_iternext (hv))) 405 while ((he = hv_iternext (hv)))
382 { 406 {
383 hes [i++] = he; 407 hes [i++] = he;
447 svtype svt; 471 svtype svt;
448 472
449 SvGETMAGIC (sv); 473 SvGETMAGIC (sv);
450 svt = SvTYPE (sv); 474 svt = SvTYPE (sv);
451 475
476 if (expect_false (SvOBJECT (sv)))
477 {
478 if (SvSTASH (sv) == json_boolean_stash)
479 {
480 if (SvIV (sv) == 0)
481 encode_str (enc, "false", 5, 0);
482 else
483 encode_str (enc, "true", 4, 0);
484 }
485 else
486 {
487#if 0
488 if (0 && sv_derived_from (rv, "JSON::Literal"))
489 {
490 // not yet
491 }
492#endif
493 if (enc->flags & F_CONV_BLESSED)
494 {
495 // we re-bless the reference to get overload and other niceties right
496 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
497
498 if (to_json)
499 {
500 dSP;
501 ENTER;
502 SAVETMPS;
503 PUSHMARK (SP);
504 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
505
506 // calling with G_SCALAR ensures that we always get a 1 reutrn value
507 // check anyways.
508 PUTBACK;
509 assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
510 SPAGAIN;
511
512 encode_sv (enc, POPs);
513
514 FREETMPS;
515 LEAVE;
516 }
517 else if (enc->flags & F_ALLOW_BLESSED)
518 encode_str (enc, "null", 4, 0);
519 else
520 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
521 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
522 }
523 else if (enc->flags & F_ALLOW_BLESSED)
524 encode_str (enc, "null", 4, 0);
525 else
526 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
527 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
528 }
529 }
452 if (svt == SVt_PVHV) 530 else if (svt == SVt_PVHV)
453 encode_hv (enc, (HV *)sv); 531 encode_hv (enc, (HV *)sv);
454 else if (svt == SVt_PVAV) 532 else if (svt == SVt_PVAV)
455 encode_av (enc, (AV *)sv); 533 encode_av (enc, (AV *)sv);
456 else if (svt < SVt_PVAV) 534 else if (svt < SVt_PVAV)
457 { 535 {
481 encode_str (enc, str, len, SvUTF8 (sv)); 559 encode_str (enc, str, len, SvUTF8 (sv));
482 encode_ch (enc, '"'); 560 encode_ch (enc, '"');
483 } 561 }
484 else if (SvNOKp (sv)) 562 else if (SvNOKp (sv))
485 { 563 {
564 // trust that perl will do the right thing w.r.t. JSON syntax.
486 need (enc, NV_DIG + 32); 565 need (enc, NV_DIG + 32);
487 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 566 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
488 enc->cur += strlen (enc->cur); 567 enc->cur += strlen (enc->cur);
489 } 568 }
490 else if (SvIOKp (sv)) 569 else if (SvIOKp (sv))
491 { 570 {
492 need (enc, 64); 571 // we assume we can always read an IV as a UV
572 if (SvUV (sv) & ~(UV)0x7fff)
573 {
574 // large integer, use the (rather slow) snprintf way.
575 need (enc, sizeof (UV) * 3);
493 enc->cur += 576 enc->cur +=
494 SvIsUV(sv) 577 SvIsUV(sv)
495 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 578 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
496 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 579 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
580 }
581 else
582 {
583 // optimise the "small number case"
584 // code will likely be branchless and use only a single multiplication
585 I32 i = SvIV (sv);
586 U32 u;
587 char digit, nz = 0;
588
589 need (enc, 6);
590
591 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
592 u = i < 0 ? -i : i;
593
594 // convert to 4.28 fixed-point representation
595 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
596
597 // now output digit by digit, each time masking out the integer part
598 // and multiplying by 5 while moving the decimal point one to the right,
599 // resulting in a net multiplication by 10.
600 // we always write the digit to memory but conditionally increment
601 // the pointer, to ease the usage of conditional move instructions.
602 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
603 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
604 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
605 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
606 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
607 }
497 } 608 }
498 else if (SvROK (sv)) 609 else if (SvROK (sv))
499 encode_rv (enc, SvRV (sv)); 610 encode_rv (enc, SvRV (sv));
500 else if (!SvOK (sv)) 611 else if (!SvOK (sv))
501 encode_str (enc, "null", 4, 0); 612 encode_str (enc, "null", 4, 0);
546 U32 flags; // F_* 657 U32 flags; // F_*
547 U32 depth; // recursion depth 658 U32 depth; // recursion depth
548 U32 maxdepth; // recursion depth limit 659 U32 maxdepth; // recursion depth limit
549} dec_t; 660} dec_t;
550 661
551static void 662inline void
552decode_ws (dec_t *dec) 663decode_ws (dec_t *dec)
553{ 664{
554 for (;;) 665 for (;;)
555 { 666 {
556 char ch = *dec->cur; 667 char ch = *dec->cur;
582decode_4hex (dec_t *dec) 693decode_4hex (dec_t *dec)
583{ 694{
584 signed char d1, d2, d3, d4; 695 signed char d1, d2, d3, d4;
585 unsigned char *cur = (unsigned char *)dec->cur; 696 unsigned char *cur = (unsigned char *)dec->cur;
586 697
587 d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); 698 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"); 699 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"); 700 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"); 701 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
591 702
592 dec->cur += 4; 703 dec->cur += 4;
593 704
594 return ((UV)d1) << 12 705 return ((UV)d1) << 12
595 | ((UV)d2) << 8 706 | ((UV)d2) << 8
603static SV * 714static SV *
604decode_str (dec_t *dec) 715decode_str (dec_t *dec)
605{ 716{
606 SV *sv = 0; 717 SV *sv = 0;
607 int utf8 = 0; 718 int utf8 = 0;
719 char *dec_cur = dec->cur;
608 720
609 do 721 do
610 { 722 {
611 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; 723 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
612 char *cur = buf; 724 char *cur = buf;
613 725
614 do 726 do
615 { 727 {
616 unsigned char ch = *(unsigned char *)dec->cur++; 728 unsigned char ch = *(unsigned char *)dec_cur++;
617 729
618 if (ch == '"') 730 if (expect_false (ch == '"'))
619 { 731 {
620 --dec->cur; 732 --dec_cur;
621 break; 733 break;
622 } 734 }
623 else if (ch == '\\') 735 else if (expect_false (ch == '\\'))
624 { 736 {
625 switch (*dec->cur) 737 switch (*dec_cur)
626 { 738 {
627 case '\\': 739 case '\\':
628 case '/': 740 case '/':
629 case '"': *cur++ = *dec->cur++; break; 741 case '"': *cur++ = *dec_cur++; break;
630 742
631 case 'b': ++dec->cur; *cur++ = '\010'; break; 743 case 'b': ++dec_cur; *cur++ = '\010'; break;
632 case 't': ++dec->cur; *cur++ = '\011'; break; 744 case 't': ++dec_cur; *cur++ = '\011'; break;
633 case 'n': ++dec->cur; *cur++ = '\012'; break; 745 case 'n': ++dec_cur; *cur++ = '\012'; break;
634 case 'f': ++dec->cur; *cur++ = '\014'; break; 746 case 'f': ++dec_cur; *cur++ = '\014'; break;
635 case 'r': ++dec->cur; *cur++ = '\015'; break; 747 case 'r': ++dec_cur; *cur++ = '\015'; break;
636 748
637 case 'u': 749 case 'u':
638 { 750 {
639 UV lo, hi; 751 UV lo, hi;
640 ++dec->cur; 752 ++dec_cur;
641 753
754 dec->cur = dec_cur;
642 hi = decode_4hex (dec); 755 hi = decode_4hex (dec);
756 dec_cur = dec->cur;
643 if (hi == (UV)-1) 757 if (hi == (UV)-1)
644 goto fail; 758 goto fail;
645 759
646 // possibly a surrogate pair 760 // possibly a surrogate pair
647 if (hi >= 0xd800) 761 if (hi >= 0xd800)
648 if (hi < 0xdc00) 762 if (hi < 0xdc00)
649 { 763 {
650 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 764 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
651 ERR ("missing low surrogate character in surrogate pair"); 765 ERR ("missing low surrogate character in surrogate pair");
652 766
653 dec->cur += 2; 767 dec_cur += 2;
654 768
769 dec->cur = dec_cur;
655 lo = decode_4hex (dec); 770 lo = decode_4hex (dec);
771 dec_cur = dec->cur;
656 if (lo == (UV)-1) 772 if (lo == (UV)-1)
657 goto fail; 773 goto fail;
658 774
659 if (lo < 0xdc00 || lo >= 0xe000) 775 if (lo < 0xdc00 || lo >= 0xe000)
660 ERR ("surrogate pair expected"); 776 ERR ("surrogate pair expected");
674 *cur++ = hi; 790 *cur++ = hi;
675 } 791 }
676 break; 792 break;
677 793
678 default: 794 default:
679 --dec->cur; 795 --dec_cur;
680 ERR ("illegal backslash escape sequence in string"); 796 ERR ("illegal backslash escape sequence in string");
681 } 797 }
682 } 798 }
683 else if (ch >= 0x20 && ch <= 0x7f) 799 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
684 *cur++ = ch; 800 *cur++ = ch;
685 else if (ch >= 0x80) 801 else if (ch >= 0x80)
686 { 802 {
687 STRLEN clen; 803 STRLEN clen;
688 UV uch; 804 UV uch;
689 805
690 --dec->cur; 806 --dec_cur;
691 807
692 uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); 808 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
693 if (clen == (STRLEN)-1) 809 if (clen == (STRLEN)-1)
694 ERR ("malformed UTF-8 character in JSON string"); 810 ERR ("malformed UTF-8 character in JSON string");
695 811
696 do 812 do
697 *cur++ = *dec->cur++; 813 *cur++ = *dec_cur++;
698 while (--clen); 814 while (--clen);
699 815
700 utf8 = 1; 816 utf8 = 1;
701 } 817 }
702 else 818 else
703 { 819 {
704 --dec->cur; 820 --dec_cur;
705 821
706 if (!ch) 822 if (!ch)
707 ERR ("unexpected end of string while parsing JSON string"); 823 ERR ("unexpected end of string while parsing JSON string");
708 else 824 else
709 ERR ("invalid character encountered while parsing JSON string"); 825 ERR ("invalid character encountered while parsing JSON string");
722 } 838 }
723 else 839 else
724 sv = newSVpvn (buf, len); 840 sv = newSVpvn (buf, len);
725 } 841 }
726 } 842 }
727 while (*dec->cur != '"'); 843 while (*dec_cur != '"');
728 844
729 ++dec->cur; 845 ++dec_cur;
730 846
731 if (sv) 847 if (sv)
732 { 848 {
733 SvPOK_only (sv); 849 SvPOK_only (sv);
734 *SvEND (sv) = 0; 850 *SvEND (sv) = 0;
737 SvUTF8_on (sv); 853 SvUTF8_on (sv);
738 } 854 }
739 else 855 else
740 sv = newSVpvn ("", 0); 856 sv = newSVpvn ("", 0);
741 857
858 dec->cur = dec_cur;
742 return sv; 859 return sv;
743 860
744fail: 861fail:
862 dec->cur = dec_cur;
745 return 0; 863 return 0;
746} 864}
747 865
748static SV * 866static SV *
749decode_num (dec_t *dec) 867decode_num (dec_t *dec)
807 is_nv = 1; 925 is_nv = 1;
808 } 926 }
809 927
810 if (!is_nv) 928 if (!is_nv)
811 { 929 {
812 UV uv; 930 // 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); 931 if (*start == '-')
814 if (numtype & IS_NUMBER_IN_UV) 932 switch (dec->cur - start)
815 if (numtype & IS_NUMBER_NEG)
816 { 933 {
817 if (uv < (UV)IV_MIN) 934 case 2: return newSViv (-( start [1] - '0' * 1));
818 return newSViv (-(IV)uv); 935 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
936 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
937 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
819 } 938 }
939 else
940 switch (dec->cur - start)
941 {
942 case 1: return newSViv ( start [0] - '0' * 1);
943 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
944 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
945 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
946 }
947
948 {
949 UV uv;
950 int numtype = grok_number (start, dec->cur - start, &uv);
951 if (numtype & IS_NUMBER_IN_UV)
952 if (numtype & IS_NUMBER_NEG)
953 {
954 if (uv < (UV)IV_MIN)
955 return newSViv (-(IV)uv);
956 }
820 else 957 else
821 return newSVuv (uv); 958 return newSVuv (uv);
959
960 // here would likely be the place for bigint support
822 } 961 }
962 }
823 963
964 // if we ever support bigint or bigfloat, this is the place for bigfloat
824 return newSVnv (Atof (start)); 965 return newSVnv (Atof (start));
825 966
826fail: 967fail:
827 return 0; 968 return 0;
828} 969}
929 1070
930static SV * 1071static SV *
931decode_sv (dec_t *dec) 1072decode_sv (dec_t *dec)
932{ 1073{
933 decode_ws (dec); 1074 decode_ws (dec);
1075
1076 // the beauty of JSON: you need exactly one character lookahead
1077 // to parse anything.
934 switch (*dec->cur) 1078 switch (*dec->cur)
935 { 1079 {
936 case '"': ++dec->cur; return decode_str (dec); 1080 case '"': ++dec->cur; return decode_str (dec);
937 case '[': ++dec->cur; return decode_av (dec); 1081 case '[': ++dec->cur; return decode_av (dec);
938 case '{': ++dec->cur; return decode_hv (dec); 1082 case '{': ++dec->cur; return decode_hv (dec);
944 1088
945 case 't': 1089 case 't':
946 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1090 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
947 { 1091 {
948 dec->cur += 4; 1092 dec->cur += 4;
949 return newSViv (1); 1093 return SvREFCNT_inc (json_true);
950 } 1094 }
951 else 1095 else
952 ERR ("'true' expected"); 1096 ERR ("'true' expected");
953 1097
954 break; 1098 break;
955 1099
956 case 'f': 1100 case 'f':
957 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1101 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
958 { 1102 {
959 dec->cur += 5; 1103 dec->cur += 5;
960 return newSViv (0); 1104 return SvREFCNT_inc (json_false);
961 } 1105 }
962 else 1106 else
963 ERR ("'false' expected"); 1107 ERR ("'false' expected");
964 1108
965 break; 1109 break;
992 SV *sv; 1136 SV *sv;
993 1137
994 SvGETMAGIC (string); 1138 SvGETMAGIC (string);
995 SvUPGRADE (string, SVt_PV); 1139 SvUPGRADE (string, SVt_PV);
996 1140
1141 if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
1142 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1143 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
1144
997 if (flags & F_UTF8) 1145 if (flags & F_UTF8)
998 sv_utf8_downgrade (string, 0); 1146 sv_utf8_downgrade (string, 0);
999 else 1147 else
1000 sv_utf8_upgrade (string); 1148 sv_utf8_upgrade (string);
1001 1149
1068 1216
1069BOOT: 1217BOOT:
1070{ 1218{
1071 int i; 1219 int i;
1072 1220
1073 memset (decode_hexdigit, 0xff, 256);
1074
1075 for (i = 0; i < 256; ++i) 1221 for (i = 0; i < 256; ++i)
1076 decode_hexdigit [i] = 1222 decode_hexdigit [i] =
1077 i >= '0' && i <= '9' ? i - '0' 1223 i >= '0' && i <= '9' ? i - '0'
1078 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1224 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1079 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1225 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1080 : -1; 1226 : -1;
1081 1227
1082 json_stash = gv_stashpv ("JSON::XS", 1); 1228 json_stash = gv_stashpv ("JSON::XS" , 1);
1229 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1230
1231 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1232 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1083} 1233}
1084 1234
1085PROTOTYPES: DISABLE 1235PROTOTYPES: DISABLE
1086 1236
1087SV *new (char *dummy) 1237SV *new (char *dummy)
1090 OUTPUT: 1240 OUTPUT:
1091 RETVAL 1241 RETVAL
1092 1242
1093SV *ascii (SV *self, int enable = 1) 1243SV *ascii (SV *self, int enable = 1)
1094 ALIAS: 1244 ALIAS:
1095 ascii = F_ASCII 1245 ascii = F_ASCII
1096 latin1 = F_LATIN1 1246 latin1 = F_LATIN1
1097 utf8 = F_UTF8 1247 utf8 = F_UTF8
1098 indent = F_INDENT 1248 indent = F_INDENT
1099 canonical = F_CANONICAL 1249 canonical = F_CANONICAL
1100 space_before = F_SPACE_BEFORE 1250 space_before = F_SPACE_BEFORE
1101 space_after = F_SPACE_AFTER 1251 space_after = F_SPACE_AFTER
1102 pretty = F_PRETTY 1252 pretty = F_PRETTY
1103 allow_nonref = F_ALLOW_NONREF 1253 allow_nonref = F_ALLOW_NONREF
1104 shrink = F_SHRINK 1254 shrink = F_SHRINK
1255 allow_blessed = F_ALLOW_BLESSED
1256 convert_blessed = F_CONV_BLESSED
1105 CODE: 1257 CODE:
1106{ 1258{
1107 UV *uv = SvJSON (self); 1259 UV *uv = SvJSON (self);
1108 if (enable) 1260 if (enable)
1109 *uv |= ix; 1261 *uv |= ix;
1131 RETVAL = newSVsv (self); 1283 RETVAL = newSVsv (self);
1132} 1284}
1133 OUTPUT: 1285 OUTPUT:
1134 RETVAL 1286 RETVAL
1135 1287
1288SV *max_size (SV *self, UV max_size = 0)
1289 CODE:
1290{
1291 UV *uv = SvJSON (self);
1292 UV log2 = 0;
1293
1294 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1295 if (max_size == 1) max_size = 2;
1296
1297 while ((1UL << log2) < max_size)
1298 ++log2;
1299
1300 *uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1301
1302 RETVAL = newSVsv (self);
1303}
1304 OUTPUT:
1305 RETVAL
1306
1136void encode (SV *self, SV *scalar) 1307void encode (SV *self, SV *scalar)
1137 PPCODE: 1308 PPCODE:
1138 XPUSHs (encode_json (scalar, *SvJSON (self))); 1309 XPUSHs (encode_json (scalar, *SvJSON (self)));
1139 1310
1140void decode (SV *self, SV *jsonstr) 1311void decode (SV *self, SV *jsonstr)

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines