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 1.31 by root, Wed May 9 16:33:53 2007 UTC vs.
Revision 1.38 by root, Mon Jun 11 03:18:07 2007 UTC

… …
32#define F_DEFAULT (9UL << S_MAXDEPTH) 32#define F_DEFAULT (9UL << S_MAXDEPTH)
33 33
34#define INIT_SIZE 32 // initial scalar size to be allocated 34#define INIT_SIZE 32 // initial scalar size to be allocated
35#define INDENT_STEP 3 // spaces per indentation level 35#define INDENT_STEP 3 // spaces per indentation level
36 36
37#define SHORT_STRING_LEN 512 // special-case strings of up to this size 37#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38 38
39#define SB do { 39#define SB do {
40#define SE } while (0) 40#define SE } while (0)
41
42#if __GNUC__ >= 3
43# define expect(expr,value) __builtin_expect ((expr),(value))
44# define inline inline
45#else
46# define expect(expr,value) (expr)
47# define inline static
48#endif
49
50#define expect_false(expr) expect ((expr) != 0, 0)
51#define expect_true(expr) expect ((expr) != 0, 1)
41 52
42static HV *json_stash; // JSON::XS:: 53static HV *json_stash; // JSON::XS::
43 54
44///////////////////////////////////////////////////////////////////////////// 55/////////////////////////////////////////////////////////////////////////////
45// utility functions 56// utility functions
… …
70// decode an utf-8 character and return it, or (UV)-1 in 81// decode an utf-8 character and return it, or (UV)-1 in
71// case of an error. 82// case of an error.
72// we special-case "safe" characters from U+80 .. U+7FF, 83// we special-case "safe" characters from U+80 .. U+7FF,
73// but use the very good perl function to parse anything else. 84// but use the very good perl function to parse anything else.
74// note that we never call this function for a ascii codepoints 85// note that we never call this function for a ascii codepoints
75static UV 86inline UV
76decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 87decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77{ 88{
78 if (s[0] > 0xdf || s[0] < 0xc2) 89 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
79 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 90 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
80 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 91 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
81 { 92 {
82 *clen = 2; 93 *clen = 2;
83 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 94 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
… …
101 U32 flags; // F_* 112 U32 flags; // F_*
102 U32 indent; // indentation level 113 U32 indent; // indentation level
103 U32 maxdepth; // max. indentation/recursion level 114 U32 maxdepth; // max. indentation/recursion level
104} enc_t; 115} enc_t;
105 116
106static void 117inline void
107need (enc_t *enc, STRLEN len) 118need (enc_t *enc, STRLEN len)
108{ 119{
109 if (enc->cur + len >= enc->end) 120 if (expect_false (enc->cur + len >= enc->end))
110 { 121 {
111 STRLEN cur = enc->cur - SvPVX (enc->sv); 122 STRLEN cur = enc->cur - SvPVX (enc->sv);
112 SvGROW (enc->sv, cur + len + 1); 123 SvGROW (enc->sv, cur + len + 1);
113 enc->cur = SvPVX (enc->sv) + cur; 124 enc->cur = SvPVX (enc->sv) + cur;
114 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 125 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115 } 126 }
116} 127}
117 128
118static void 129inline void
119encode_ch (enc_t *enc, char ch) 130encode_ch (enc_t *enc, char ch)
120{ 131{
121 need (enc, 1); 132 need (enc, 1);
122 *enc->cur++ = ch; 133 *enc->cur++ = ch;
123} 134}
… …
131 142
132 while (str < end) 143 while (str < end)
133 { 144 {
134 unsigned char ch = *(unsigned char *)str; 145 unsigned char ch = *(unsigned char *)str;
135 146
136 if (ch >= 0x20 && ch < 0x80) // most common case 147 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137 { 148 {
138 if (ch == '"') // but with slow exceptions 149 if (expect_false (ch == '"')) // but with slow exceptions
139 { 150 {
140 need (enc, len += 1); 151 need (enc, len += 1);
141 *enc->cur++ = '\\'; 152 *enc->cur++ = '\\';
142 *enc->cur++ = '"'; 153 *enc->cur++ = '"';
143 } 154 }
144 else if (ch == '\\') 155 else if (expect_false (ch == '\\'))
145 { 156 {
146 need (enc, len += 1); 157 need (enc, len += 1);
147 *enc->cur++ = '\\'; 158 *enc->cur++ = '\\';
148 *enc->cur++ = '\\'; 159 *enc->cur++ = '\\';
149 } 160 }
… …
233 244
234 --len; 245 --len;
235 } 246 }
236} 247}
237 248
238static void 249inline void
239encode_indent (enc_t *enc) 250encode_indent (enc_t *enc)
240{ 251{
241 if (enc->flags & F_INDENT) 252 if (enc->flags & F_INDENT)
242 { 253 {
243 int spaces = enc->indent * INDENT_STEP; 254 int spaces = enc->indent * INDENT_STEP;
… …
246 memset (enc->cur, ' ', spaces); 257 memset (enc->cur, ' ', spaces);
247 enc->cur += spaces; 258 enc->cur += spaces;
248 } 259 }
249} 260}
250 261
251static void 262inline void
252encode_space (enc_t *enc) 263encode_space (enc_t *enc)
253{ 264{
254 need (enc, 1); 265 need (enc, 1);
255 encode_ch (enc, ' '); 266 encode_ch (enc, ' ');
256} 267}
257 268
258static void 269inline void
259encode_nl (enc_t *enc) 270encode_nl (enc_t *enc)
260{ 271{
261 if (enc->flags & F_INDENT) 272 if (enc->flags & F_INDENT)
262 { 273 {
263 need (enc, 1); 274 need (enc, 1);
264 encode_ch (enc, '\n'); 275 encode_ch (enc, '\n');
265 } 276 }
266} 277}
267 278
268static void 279inline void
269encode_comma (enc_t *enc) 280encode_comma (enc_t *enc)
270{ 281{
271 encode_ch (enc, ','); 282 encode_ch (enc, ',');
272 283
273 if (enc->flags & F_INDENT) 284 if (enc->flags & F_INDENT)
… …
372 // actually, this is mostly due to the stupid so-called 383 // actually, this is mostly due to the stupid so-called
373 // security workaround added somewhere in 5.8.x. 384 // security workaround added somewhere in 5.8.x.
374 // that randomises hash orderings 385 // that randomises hash orderings
375 if (enc->flags & F_CANONICAL) 386 if (enc->flags & F_CANONICAL)
376 { 387 {
377 HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode
378 int fast = 1; 388 int fast = 1;
389 HE *he;
390#if defined(__BORLANDC__) || defined(_MSC_VER)
391 HE **hes = _alloca (count * sizeof (HE));
392#else
393 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
394#endif
379 395
380 i = 0; 396 i = 0;
381 while ((he = hv_iternext (hv))) 397 while ((he = hv_iternext (hv)))
382 { 398 {
383 hes [i++] = he; 399 hes [i++] = he;
… …
487 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 503 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
488 enc->cur += strlen (enc->cur); 504 enc->cur += strlen (enc->cur);
489 } 505 }
490 else if (SvIOKp (sv)) 506 else if (SvIOKp (sv))
491 { 507 {
492 need (enc, 64); 508 // we assume we can always read an IV as a UV
509 if (SvUV (sv) & ~(UV)0x7fff)
510 {
511 need (enc, sizeof (UV) * 3);
493 enc->cur += 512 enc->cur +=
494 SvIsUV(sv) 513 SvIsUV(sv)
495 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 514 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
496 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 515 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
516 }
517 else
518 {
519 // optimise the "small number case"
520 // code will likely be branchless and use only a single multiplication
521 I32 i = SvIV (sv);
522 U32 u;
523
524 need (enc, 6);
525
526 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
527 u = i < 0 ? -i : i;
528
529 // convert to 4.28 fixed-point representation
530 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531
532 char digit, nz = 0;
533
534 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
535 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
536 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
537 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
538 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1;
539 }
497 } 540 }
498 else if (SvROK (sv)) 541 else if (SvROK (sv))
499 encode_rv (enc, SvRV (sv)); 542 encode_rv (enc, SvRV (sv));
500 else if (!SvOK (sv)) 543 else if (!SvOK (sv))
501 encode_str (enc, "null", 4, 0); 544 encode_str (enc, "null", 4, 0);
… …
546 U32 flags; // F_* 589 U32 flags; // F_*
547 U32 depth; // recursion depth 590 U32 depth; // recursion depth
548 U32 maxdepth; // recursion depth limit 591 U32 maxdepth; // recursion depth limit
549} dec_t; 592} dec_t;
550 593
551static void 594inline void
552decode_ws (dec_t *dec) 595decode_ws (dec_t *dec)
553{ 596{
554 for (;;) 597 for (;;)
555 { 598 {
556 char ch = *dec->cur; 599 char ch = *dec->cur;
… …
582decode_4hex (dec_t *dec) 625decode_4hex (dec_t *dec)
583{ 626{
584 signed char d1, d2, d3, d4; 627 signed char d1, d2, d3, d4;
585 unsigned char *cur = (unsigned char *)dec->cur; 628 unsigned char *cur = (unsigned char *)dec->cur;
586 629
587 d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); 630 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");
588 d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); 631 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected");
589 d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); 632 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected");
590 d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); 633 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected");
591 634
592 dec->cur += 4; 635 dec->cur += 4;
593 636
594 return ((UV)d1) << 12 637 return ((UV)d1) << 12
595 | ((UV)d2) << 8 638 | ((UV)d2) << 8
… …
603static SV * 646static SV *
604decode_str (dec_t *dec) 647decode_str (dec_t *dec)
605{ 648{
606 SV *sv = 0; 649 SV *sv = 0;
607 int utf8 = 0; 650 int utf8 = 0;
651 char *dec_cur = dec->cur;
608 652
609 do 653 do
610 { 654 {
611 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; 655 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
612 char *cur = buf; 656 char *cur = buf;
613 657
614 do 658 do
615 { 659 {
616 unsigned char ch = *(unsigned char *)dec->cur++; 660 unsigned char ch = *(unsigned char *)dec_cur++;
617 661
618 if (ch == '"') 662 if (expect_false (ch == '"'))
619 { 663 {
620 --dec->cur; 664 --dec_cur;
621 break; 665 break;
622 } 666 }
623 else if (ch == '\\') 667 else if (expect_false (ch == '\\'))
624 { 668 {
625 switch (*dec->cur) 669 switch (*dec_cur)
626 { 670 {
627 case '\\': 671 case '\\':
628 case '/': 672 case '/':
629 case '"': *cur++ = *dec->cur++; break; 673 case '"': *cur++ = *dec_cur++; break;
630 674
631 case 'b': ++dec->cur; *cur++ = '\010'; break; 675 case 'b': ++dec_cur; *cur++ = '\010'; break;
632 case 't': ++dec->cur; *cur++ = '\011'; break; 676 case 't': ++dec_cur; *cur++ = '\011'; break;
633 case 'n': ++dec->cur; *cur++ = '\012'; break; 677 case 'n': ++dec_cur; *cur++ = '\012'; break;
634 case 'f': ++dec->cur; *cur++ = '\014'; break; 678 case 'f': ++dec_cur; *cur++ = '\014'; break;
635 case 'r': ++dec->cur; *cur++ = '\015'; break; 679 case 'r': ++dec_cur; *cur++ = '\015'; break;
636 680
637 case 'u': 681 case 'u':
638 { 682 {
639 UV lo, hi; 683 UV lo, hi;
640 ++dec->cur; 684 ++dec_cur;
641 685
686 dec->cur = dec_cur;
642 hi = decode_4hex (dec); 687 hi = decode_4hex (dec);
688 dec_cur = dec->cur;
643 if (hi == (UV)-1) 689 if (hi == (UV)-1)
644 goto fail; 690 goto fail;
645 691
646 // possibly a surrogate pair 692 // possibly a surrogate pair
647 if (hi >= 0xd800) 693 if (hi >= 0xd800)
648 if (hi < 0xdc00) 694 if (hi < 0xdc00)
649 { 695 {
650 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 696 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
651 ERR ("missing low surrogate character in surrogate pair"); 697 ERR ("missing low surrogate character in surrogate pair");
652 698
653 dec->cur += 2; 699 dec_cur += 2;
654 700
701 dec->cur = dec_cur;
655 lo = decode_4hex (dec); 702 lo = decode_4hex (dec);
703 dec_cur = dec->cur;
656 if (lo == (UV)-1) 704 if (lo == (UV)-1)
657 goto fail; 705 goto fail;
658 706
659 if (lo < 0xdc00 || lo >= 0xe000) 707 if (lo < 0xdc00 || lo >= 0xe000)
660 ERR ("surrogate pair expected"); 708 ERR ("surrogate pair expected");
… …
674 *cur++ = hi; 722 *cur++ = hi;
675 } 723 }
676 break; 724 break;
677 725
678 default: 726 default:
679 --dec->cur; 727 --dec_cur;
680 ERR ("illegal backslash escape sequence in string"); 728 ERR ("illegal backslash escape sequence in string");
681 } 729 }
682 } 730 }
683 else if (ch >= 0x20 && ch <= 0x7f) 731 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
684 *cur++ = ch; 732 *cur++ = ch;
685 else if (ch >= 0x80) 733 else if (ch >= 0x80)
686 { 734 {
687 STRLEN clen; 735 STRLEN clen;
688 UV uch; 736 UV uch;
689 737
690 --dec->cur; 738 --dec_cur;
691 739
692 uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); 740 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
693 if (clen == (STRLEN)-1) 741 if (clen == (STRLEN)-1)
694 ERR ("malformed UTF-8 character in JSON string"); 742 ERR ("malformed UTF-8 character in JSON string");
695 743
696 do 744 do
697 *cur++ = *dec->cur++; 745 *cur++ = *dec_cur++;
698 while (--clen); 746 while (--clen);
699 747
700 utf8 = 1; 748 utf8 = 1;
701 } 749 }
702 else 750 else
703 { 751 {
704 --dec->cur; 752 --dec_cur;
705 753
706 if (!ch) 754 if (!ch)
707 ERR ("unexpected end of string while parsing JSON string"); 755 ERR ("unexpected end of string while parsing JSON string");
708 else 756 else
709 ERR ("invalid character encountered while parsing JSON string"); 757 ERR ("invalid character encountered while parsing JSON string");
… …
722 } 770 }
723 else 771 else
724 sv = newSVpvn (buf, len); 772 sv = newSVpvn (buf, len);
725 } 773 }
726 } 774 }
727 while (*dec->cur != '"'); 775 while (*dec_cur != '"');
728 776
729 ++dec->cur; 777 ++dec_cur;
730 778
731 if (sv) 779 if (sv)
732 { 780 {
733 SvPOK_only (sv); 781 SvPOK_only (sv);
734 *SvEND (sv) = 0; 782 *SvEND (sv) = 0;
… …
737 SvUTF8_on (sv); 785 SvUTF8_on (sv);
738 } 786 }
739 else 787 else
740 sv = newSVpvn ("", 0); 788 sv = newSVpvn ("", 0);
741 789
790 dec->cur = dec_cur;
742 return sv; 791 return sv;
743 792
744fail: 793fail:
794 dec->cur = dec_cur;
745 return 0; 795 return 0;
746} 796}
747 797
748static SV * 798static SV *
749decode_num (dec_t *dec) 799decode_num (dec_t *dec)
… …
807 is_nv = 1; 857 is_nv = 1;
808 } 858 }
809 859
810 if (!is_nv) 860 if (!is_nv)
811 { 861 {
812 UV uv; 862 // 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); 863 if (*start == '-')
814 if (numtype & IS_NUMBER_IN_UV) 864 switch (dec->cur - start)
815 if (numtype & IS_NUMBER_NEG)
816 { 865 {
817 if (uv < (UV)IV_MIN) 866 case 2: return newSViv (-( start [1] - '0' ));
818 return newSViv (-(IV)uv); 867 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
868 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
869 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
819 } 870 }
871 else
872 switch (dec->cur - start)
873 {
874 case 1: return newSViv ( start [0] - '0' );
875 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
876 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
877 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
878 }
879
880 {
881 UV uv;
882 int numtype = grok_number (start, dec->cur - start, &uv);
883 if (numtype & IS_NUMBER_IN_UV)
884 if (numtype & IS_NUMBER_NEG)
885 {
886 if (uv < (UV)IV_MIN)
887 return newSViv (-(IV)uv);
888 }
820 else 889 else
821 return newSVuv (uv); 890 return newSVuv (uv);
891 }
822 } 892 }
823 893
824 return newSVnv (Atof (start)); 894 return newSVnv (Atof (start));
825 895
826fail: 896fail:
… …
1009 dec.maxdepth = DEC_DEPTH (dec.flags); 1079 dec.maxdepth = DEC_DEPTH (dec.flags);
1010 1080
1011 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1081 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1012 sv = decode_sv (&dec); 1082 sv = decode_sv (&dec);
1013 1083
1014 if (offset_return || !sv) 1084 if (!(offset_return || !sv))
1015 {
1016 offset = dec.flags & F_UTF8
1017 ? dec.cur - SvPVX (string)
1018 : utf8_distance (dec.cur, SvPVX (string));
1019
1020 if (offset_return)
1021 *offset_return = offset;
1022 }
1023 else
1024 { 1085 {
1025 // check for trailing garbage 1086 // check for trailing garbage
1026 decode_ws (&dec); 1087 decode_ws (&dec);
1027 1088
1028 if (*dec.cur) 1089 if (*dec.cur)
1029 { 1090 {
1030 dec.err = "garbage after JSON object"; 1091 dec.err = "garbage after JSON object";
1031 SvREFCNT_dec (sv); 1092 SvREFCNT_dec (sv);
1032 sv = 0; 1093 sv = 0;
1033 } 1094 }
1095 }
1096
1097 if (offset_return || !sv)
1098 {
1099 offset = dec.flags & F_UTF8
1100 ? dec.cur - SvPVX (string)
1101 : utf8_distance (dec.cur, SvPVX (string));
1102
1103 if (offset_return)
1104 *offset_return = offset;
1034 } 1105 }
1035 1106
1036 if (!sv) 1107 if (!sv)
1037 { 1108 {
1038 SV *uni = sv_newmortal (); 1109 SV *uni = sv_newmortal ();

Diff Legend

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