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 1.26 by root, Fri Apr 6 21:17:09 2007 UTC vs.
Revision 1.40 by root, Tue Jun 12 01:27:02 2007 UTC

… …
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#define F_ASCII 0x00000001UL 14#define F_ASCII 0x00000001UL
15#define F_LATIN1 0x00000002UL
15#define F_UTF8 0x00000002UL 16#define F_UTF8 0x00000004UL
16#define F_INDENT 0x00000004UL 17#define F_INDENT 0x00000008UL
17#define F_CANONICAL 0x00000008UL 18#define F_CANONICAL 0x00000010UL
18#define F_SPACE_BEFORE 0x00000010UL 19#define F_SPACE_BEFORE 0x00000020UL
19#define F_SPACE_AFTER 0x00000020UL 20#define F_SPACE_AFTER 0x00000040UL
20#define F_ALLOW_NONREF 0x00000080UL 21#define F_ALLOW_NONREF 0x00000100UL
21#define F_SHRINK 0x00000100UL 22#define F_SHRINK 0x00000200UL
22#define F_MAXDEPTH 0xf8000000UL 23#define F_MAXDEPTH 0xf8000000UL
23#define S_MAXDEPTH 27 24#define S_MAXDEPTH 27
24 25
25#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) 26#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
26 27
… …
31#define F_DEFAULT (9UL << S_MAXDEPTH) 32#define F_DEFAULT (9UL << S_MAXDEPTH)
32 33
33#define INIT_SIZE 32 // initial scalar size to be allocated 34#define INIT_SIZE 32 // initial scalar size to be allocated
34#define INDENT_STEP 3 // spaces per indentation level 35#define INDENT_STEP 3 // spaces per indentation level
35 36
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 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); 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 }
… …
167 STRLEN clen; 178 STRLEN clen;
168 UV uch; 179 UV uch;
169 180
170 if (is_utf8) 181 if (is_utf8)
171 { 182 {
172 //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173 uch = decode_utf8 (str, end - str, &clen); 183 uch = decode_utf8 (str, end - str, &clen);
174 if (clen == (STRLEN)-1) 184 if (clen == (STRLEN)-1)
175 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); 185 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176 } 186 }
177 else 187 else
… …
181 } 191 }
182 192
183 if (uch > 0x10FFFFUL) 193 if (uch > 0x10FFFFUL)
184 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 194 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185 195
186 if (uch < 0x80 || enc->flags & F_ASCII) 196 if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
187 { 197 {
188 if (uch > 0xFFFFUL) 198 if (uch > 0xFFFFUL)
189 { 199 {
190 need (enc, len += 11); 200 need (enc, len += 11);
191 sprintf (enc->cur, "\\u%04x\\u%04x", 201 sprintf (enc->cur, "\\u%04x\\u%04x",
… …
205 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 215 *enc->cur++ = hexdigit [(uch >> 0) & 15];
206 } 216 }
207 217
208 str += clen; 218 str += clen;
209 } 219 }
220 else if (enc->flags & F_LATIN1)
221 {
222 *enc->cur++ = uch;
223 str += clen;
224 }
210 else if (is_utf8) 225 else if (is_utf8)
211 { 226 {
212 need (enc, len += clen); 227 need (enc, len += clen);
213 do 228 do
214 { 229 {
… …
216 } 231 }
217 while (--clen); 232 while (--clen);
218 } 233 }
219 else 234 else
220 { 235 {
221 need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed 236 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
222 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 237 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
223 ++str; 238 ++str;
224 } 239 }
225 } 240 }
226 } 241 }
… …
228 243
229 --len; 244 --len;
230 } 245 }
231} 246}
232 247
233static void 248inline void
234encode_indent (enc_t *enc) 249encode_indent (enc_t *enc)
235{ 250{
236 if (enc->flags & F_INDENT) 251 if (enc->flags & F_INDENT)
237 { 252 {
238 int spaces = enc->indent * INDENT_STEP; 253 int spaces = enc->indent * INDENT_STEP;
… …
241 memset (enc->cur, ' ', spaces); 256 memset (enc->cur, ' ', spaces);
242 enc->cur += spaces; 257 enc->cur += spaces;
243 } 258 }
244} 259}
245 260
246static void 261inline void
247encode_space (enc_t *enc) 262encode_space (enc_t *enc)
248{ 263{
249 need (enc, 1); 264 need (enc, 1);
250 encode_ch (enc, ' '); 265 encode_ch (enc, ' ');
251} 266}
252 267
253static void 268inline void
254encode_nl (enc_t *enc) 269encode_nl (enc_t *enc)
255{ 270{
256 if (enc->flags & F_INDENT) 271 if (enc->flags & F_INDENT)
257 { 272 {
258 need (enc, 1); 273 need (enc, 1);
259 encode_ch (enc, '\n'); 274 encode_ch (enc, '\n');
260 } 275 }
261} 276}
262 277
263static void 278inline void
264encode_comma (enc_t *enc) 279encode_comma (enc_t *enc)
265{ 280{
266 encode_ch (enc, ','); 281 encode_ch (enc, ',');
267 282
268 if (enc->flags & F_INDENT) 283 if (enc->flags & F_INDENT)
… …
367 // actually, this is mostly due to the stupid so-called 382 // actually, this is mostly due to the stupid so-called
368 // security workaround added somewhere in 5.8.x. 383 // security workaround added somewhere in 5.8.x.
369 // that randomises hash orderings 384 // that randomises hash orderings
370 if (enc->flags & F_CANONICAL) 385 if (enc->flags & F_CANONICAL)
371 { 386 {
372 HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode
373 int fast = 1; 387 int fast = 1;
388 HE *he;
389#if defined(__BORLANDC__) || defined(_MSC_VER)
390 HE **hes = _alloca (count * sizeof (HE));
391#else
392 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
393#endif
374 394
375 i = 0; 395 i = 0;
376 while ((he = hv_iternext (hv))) 396 while ((he = hv_iternext (hv)))
377 { 397 {
378 hes [i++] = he; 398 hes [i++] = he;
… …
476 encode_str (enc, str, len, SvUTF8 (sv)); 496 encode_str (enc, str, len, SvUTF8 (sv));
477 encode_ch (enc, '"'); 497 encode_ch (enc, '"');
478 } 498 }
479 else if (SvNOKp (sv)) 499 else if (SvNOKp (sv))
480 { 500 {
501 // trust that perl will do the right thing w.r.t. JSON syntax.
481 need (enc, NV_DIG + 32); 502 need (enc, NV_DIG + 32);
482 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 503 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
483 enc->cur += strlen (enc->cur); 504 enc->cur += strlen (enc->cur);
484 } 505 }
485 else if (SvIOKp (sv)) 506 else if (SvIOKp (sv))
486 { 507 {
487 need (enc, 64); 508 // we assume we can always read an IV as a UV
509 if (SvUV (sv) & ~(UV)0x7fff)
510 {
511 // large integer, use the (rather slow) snprintf way.
512 need (enc, sizeof (UV) * 3);
488 enc->cur += 513 enc->cur +=
489 SvIsUV(sv) 514 SvIsUV(sv)
490 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 515 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
491 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 516 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
517 }
518 else
519 {
520 // optimise the "small number case"
521 // code will likely be branchless and use only a single multiplication
522 I32 i = SvIV (sv);
523 U32 u;
524 char digit, nz = 0;
525
526 need (enc, 6);
527
528 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
529 u = i < 0 ? -i : i;
530
531 // convert to 4.28 fixed-point representation
532 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
533
534 // now output digit by digit, each time masking out the integer part
535 // and multiplying by 5 while moving the decimal point one to the right,
536 // resulting in a net multiplication by 10.
537 // we always write the digit to memory but conditionally increment
538 // the pointer, to ease the usage of conditional move instructions.
539 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
540 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
541 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
542 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
543 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
544 }
492 } 545 }
493 else if (SvROK (sv)) 546 else if (SvROK (sv))
494 encode_rv (enc, SvRV (sv)); 547 encode_rv (enc, SvRV (sv));
495 else if (!SvOK (sv)) 548 else if (!SvOK (sv))
496 encode_str (enc, "null", 4, 0); 549 encode_str (enc, "null", 4, 0);
… …
515 enc.maxdepth = DEC_DEPTH (flags); 568 enc.maxdepth = DEC_DEPTH (flags);
516 569
517 SvPOK_only (enc.sv); 570 SvPOK_only (enc.sv);
518 encode_sv (&enc, scalar); 571 encode_sv (&enc, scalar);
519 572
573 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
574 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
575
520 if (!(flags & (F_ASCII | F_UTF8))) 576 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
521 SvUTF8_on (enc.sv); 577 SvUTF8_on (enc.sv);
522
523 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
524 578
525 if (enc.flags & F_SHRINK) 579 if (enc.flags & F_SHRINK)
526 shrink (enc.sv); 580 shrink (enc.sv);
527 581
528 return enc.sv; 582 return enc.sv;
… …
540 U32 flags; // F_* 594 U32 flags; // F_*
541 U32 depth; // recursion depth 595 U32 depth; // recursion depth
542 U32 maxdepth; // recursion depth limit 596 U32 maxdepth; // recursion depth limit
543} dec_t; 597} dec_t;
544 598
545static void 599inline void
546decode_ws (dec_t *dec) 600decode_ws (dec_t *dec)
547{ 601{
548 for (;;) 602 for (;;)
549 { 603 {
550 char ch = *dec->cur; 604 char ch = *dec->cur;
… …
576decode_4hex (dec_t *dec) 630decode_4hex (dec_t *dec)
577{ 631{
578 signed char d1, d2, d3, d4; 632 signed char d1, d2, d3, d4;
579 unsigned char *cur = (unsigned char *)dec->cur; 633 unsigned char *cur = (unsigned char *)dec->cur;
580 634
581 d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); 635 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
582 d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); 636 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
583 d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); 637 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
584 d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); 638 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
585 639
586 dec->cur += 4; 640 dec->cur += 4;
587 641
588 return ((UV)d1) << 12 642 return ((UV)d1) << 12
589 | ((UV)d2) << 8 643 | ((UV)d2) << 8
… …
597static SV * 651static SV *
598decode_str (dec_t *dec) 652decode_str (dec_t *dec)
599{ 653{
600 SV *sv = 0; 654 SV *sv = 0;
601 int utf8 = 0; 655 int utf8 = 0;
656 char *dec_cur = dec->cur;
602 657
603 do 658 do
604 { 659 {
605 char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; 660 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
606 char *cur = buf; 661 char *cur = buf;
607 662
608 do 663 do
609 { 664 {
610 unsigned char ch = *(unsigned char *)dec->cur++; 665 unsigned char ch = *(unsigned char *)dec_cur++;
611 666
612 if (ch == '"') 667 if (expect_false (ch == '"'))
613 { 668 {
614 --dec->cur; 669 --dec_cur;
615 break; 670 break;
616 } 671 }
617 else if (ch == '\\') 672 else if (expect_false (ch == '\\'))
618 { 673 {
619 switch (*dec->cur) 674 switch (*dec_cur)
620 { 675 {
621 case '\\': 676 case '\\':
622 case '/': 677 case '/':
623 case '"': *cur++ = *dec->cur++; break; 678 case '"': *cur++ = *dec_cur++; break;
624 679
625 case 'b': ++dec->cur; *cur++ = '\010'; break; 680 case 'b': ++dec_cur; *cur++ = '\010'; break;
626 case 't': ++dec->cur; *cur++ = '\011'; break; 681 case 't': ++dec_cur; *cur++ = '\011'; break;
627 case 'n': ++dec->cur; *cur++ = '\012'; break; 682 case 'n': ++dec_cur; *cur++ = '\012'; break;
628 case 'f': ++dec->cur; *cur++ = '\014'; break; 683 case 'f': ++dec_cur; *cur++ = '\014'; break;
629 case 'r': ++dec->cur; *cur++ = '\015'; break; 684 case 'r': ++dec_cur; *cur++ = '\015'; break;
630 685
631 case 'u': 686 case 'u':
632 { 687 {
633 UV lo, hi; 688 UV lo, hi;
634 ++dec->cur; 689 ++dec_cur;
635 690
691 dec->cur = dec_cur;
636 hi = decode_4hex (dec); 692 hi = decode_4hex (dec);
693 dec_cur = dec->cur;
637 if (hi == (UV)-1) 694 if (hi == (UV)-1)
638 goto fail; 695 goto fail;
639 696
640 // possibly a surrogate pair 697 // possibly a surrogate pair
641 if (hi >= 0xd800) 698 if (hi >= 0xd800)
642 if (hi < 0xdc00) 699 if (hi < 0xdc00)
643 { 700 {
644 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 701 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
645 ERR ("missing low surrogate character in surrogate pair"); 702 ERR ("missing low surrogate character in surrogate pair");
646 703
647 dec->cur += 2; 704 dec_cur += 2;
648 705
706 dec->cur = dec_cur;
649 lo = decode_4hex (dec); 707 lo = decode_4hex (dec);
708 dec_cur = dec->cur;
650 if (lo == (UV)-1) 709 if (lo == (UV)-1)
651 goto fail; 710 goto fail;
652 711
653 if (lo < 0xdc00 || lo >= 0xe000) 712 if (lo < 0xdc00 || lo >= 0xe000)
654 ERR ("surrogate pair expected"); 713 ERR ("surrogate pair expected");
… …
668 *cur++ = hi; 727 *cur++ = hi;
669 } 728 }
670 break; 729 break;
671 730
672 default: 731 default:
673 --dec->cur; 732 --dec_cur;
674 ERR ("illegal backslash escape sequence in string"); 733 ERR ("illegal backslash escape sequence in string");
675 } 734 }
676 } 735 }
677 else if (ch >= 0x20 && ch <= 0x7f) 736 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
678 *cur++ = ch; 737 *cur++ = ch;
679 else if (ch >= 0x80) 738 else if (ch >= 0x80)
680 { 739 {
681 STRLEN clen; 740 STRLEN clen;
682 UV uch; 741 UV uch;
683 742
684 --dec->cur; 743 --dec_cur;
685 744
686 uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); 745 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
687 if (clen == (STRLEN)-1) 746 if (clen == (STRLEN)-1)
688 ERR ("malformed UTF-8 character in JSON string"); 747 ERR ("malformed UTF-8 character in JSON string");
689 748
690 do 749 do
691 *cur++ = *dec->cur++; 750 *cur++ = *dec_cur++;
692 while (--clen); 751 while (--clen);
693 752
694 utf8 = 1; 753 utf8 = 1;
695 } 754 }
696 else 755 else
697 { 756 {
698 --dec->cur; 757 --dec_cur;
699 758
700 if (!ch) 759 if (!ch)
701 ERR ("unexpected end of string while parsing JSON string"); 760 ERR ("unexpected end of string while parsing JSON string");
702 else 761 else
703 ERR ("invalid character encountered while parsing JSON string"); 762 ERR ("invalid character encountered while parsing JSON string");
… …
716 } 775 }
717 else 776 else
718 sv = newSVpvn (buf, len); 777 sv = newSVpvn (buf, len);
719 } 778 }
720 } 779 }
721 while (*dec->cur != '"'); 780 while (*dec_cur != '"');
722 781
723 ++dec->cur; 782 ++dec_cur;
724 783
725 if (sv) 784 if (sv)
726 { 785 {
727 SvPOK_only (sv); 786 SvPOK_only (sv);
728 *SvEND (sv) = 0; 787 *SvEND (sv) = 0;
… …
731 SvUTF8_on (sv); 790 SvUTF8_on (sv);
732 } 791 }
733 else 792 else
734 sv = newSVpvn ("", 0); 793 sv = newSVpvn ("", 0);
735 794
795 dec->cur = dec_cur;
736 return sv; 796 return sv;
737 797
738fail: 798fail:
799 dec->cur = dec_cur;
739 return 0; 800 return 0;
740} 801}
741 802
742static SV * 803static SV *
743decode_num (dec_t *dec) 804decode_num (dec_t *dec)
… …
801 is_nv = 1; 862 is_nv = 1;
802 } 863 }
803 864
804 if (!is_nv) 865 if (!is_nv)
805 { 866 {
806 UV uv; 867 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
807 int numtype = grok_number (start, dec->cur - start, &uv); 868 if (*start == '-')
808 if (numtype & IS_NUMBER_IN_UV) 869 switch (dec->cur - start)
809 if (numtype & IS_NUMBER_NEG)
810 { 870 {
811 if (uv < (UV)IV_MIN) 871 case 2: return newSViv (-( start [1] - '0' * 1));
812 return newSViv (-(IV)uv); 872 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
873 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
874 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
813 } 875 }
876 else
877 switch (dec->cur - start)
878 {
879 case 1: return newSViv ( start [0] - '0' * 1);
880 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
881 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
882 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
883 }
884
885 {
886 UV uv;
887 int numtype = grok_number (start, dec->cur - start, &uv);
888 if (numtype & IS_NUMBER_IN_UV)
889 if (numtype & IS_NUMBER_NEG)
890 {
891 if (uv < (UV)IV_MIN)
892 return newSViv (-(IV)uv);
893 }
814 else 894 else
815 return newSVuv (uv); 895 return newSVuv (uv);
896
897 // here would likely be the place for bigint support
816 } 898 }
899 }
817 900
901 // if we ever support bigint or bigfloat, this is the place for bigfloat
818 return newSVnv (Atof (start)); 902 return newSVnv (Atof (start));
819 903
820fail: 904fail:
821 return 0; 905 return 0;
822} 906}
… …
923 1007
924static SV * 1008static SV *
925decode_sv (dec_t *dec) 1009decode_sv (dec_t *dec)
926{ 1010{
927 decode_ws (dec); 1011 decode_ws (dec);
1012
1013 // the beauty of JSON: you need exactly one character lookahead
1014 // to parse anything.
928 switch (*dec->cur) 1015 switch (*dec->cur)
929 { 1016 {
930 case '"': ++dec->cur; return decode_str (dec); 1017 case '"': ++dec->cur; return decode_str (dec);
931 case '[': ++dec->cur; return decode_av (dec); 1018 case '[': ++dec->cur; return decode_av (dec);
932 case '{': ++dec->cur; return decode_hv (dec); 1019 case '{': ++dec->cur; return decode_hv (dec);
… …
977fail: 1064fail:
978 return 0; 1065 return 0;
979} 1066}
980 1067
981static SV * 1068static SV *
982decode_json (SV *string, U32 flags) 1069decode_json (SV *string, U32 flags, UV *offset_return)
983{ 1070{
984 dec_t dec; 1071 dec_t dec;
1072 UV offset;
985 SV *sv; 1073 SV *sv;
986 1074
1075 SvGETMAGIC (string);
987 SvUPGRADE (string, SVt_PV); 1076 SvUPGRADE (string, SVt_PV);
988 1077
989 if (flags & F_UTF8) 1078 if (flags & F_UTF8)
990 sv_utf8_downgrade (string, 0); 1079 sv_utf8_downgrade (string, 0);
991 else 1080 else
… …
998 dec.end = SvEND (string); 1087 dec.end = SvEND (string);
999 dec.err = 0; 1088 dec.err = 0;
1000 dec.depth = 0; 1089 dec.depth = 0;
1001 dec.maxdepth = DEC_DEPTH (dec.flags); 1090 dec.maxdepth = DEC_DEPTH (dec.flags);
1002 1091
1003 *dec.end = 0; // this should basically be a nop, too, but make sure its there 1092 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1004 sv = decode_sv (&dec); 1093 sv = decode_sv (&dec);
1005 1094
1095 if (!(offset_return || !sv))
1096 {
1097 // check for trailing garbage
1098 decode_ws (&dec);
1099
1100 if (*dec.cur)
1101 {
1102 dec.err = "garbage after JSON object";
1103 SvREFCNT_dec (sv);
1104 sv = 0;
1105 }
1106 }
1107
1108 if (offset_return || !sv)
1109 {
1110 offset = dec.flags & F_UTF8
1111 ? dec.cur - SvPVX (string)
1112 : utf8_distance (dec.cur, SvPVX (string));
1113
1114 if (offset_return)
1115 *offset_return = offset;
1116 }
1117
1006 if (!sv) 1118 if (!sv)
1007 { 1119 {
1008 IV offset = dec.flags & F_UTF8
1009 ? dec.cur - SvPVX (string)
1010 : utf8_distance (dec.cur, SvPVX (string));
1011 SV *uni = sv_newmortal (); 1120 SV *uni = sv_newmortal ();
1012 1121
1013 // horrible hack to silence warning inside pv_uni_display 1122 // horrible hack to silence warning inside pv_uni_display
1014 COP cop = *PL_curcop; 1123 COP cop = *PL_curcop;
1015 cop.cop_warnings = pWARN_NONE; 1124 cop.cop_warnings = pWARN_NONE;
… …
1040 1149
1041BOOT: 1150BOOT:
1042{ 1151{
1043 int i; 1152 int i;
1044 1153
1045 memset (decode_hexdigit, 0xff, 256);
1046
1047 for (i = 0; i < 256; ++i) 1154 for (i = 0; i < 256; ++i)
1048 decode_hexdigit [i] = 1155 decode_hexdigit [i] =
1049 i >= '0' && i <= '9' ? i - '0' 1156 i >= '0' && i <= '9' ? i - '0'
1050 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1157 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1051 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1158 : i >= 'A' && i <= 'F' ? i - 'A' + 10
… …
1063 RETVAL 1170 RETVAL
1064 1171
1065SV *ascii (SV *self, int enable = 1) 1172SV *ascii (SV *self, int enable = 1)
1066 ALIAS: 1173 ALIAS:
1067 ascii = F_ASCII 1174 ascii = F_ASCII
1175 latin1 = F_LATIN1
1068 utf8 = F_UTF8 1176 utf8 = F_UTF8
1069 indent = F_INDENT 1177 indent = F_INDENT
1070 canonical = F_CANONICAL 1178 canonical = F_CANONICAL
1071 space_before = F_SPACE_BEFORE 1179 space_before = F_SPACE_BEFORE
1072 space_after = F_SPACE_AFTER 1180 space_after = F_SPACE_AFTER
… …
1108 PPCODE: 1216 PPCODE:
1109 XPUSHs (encode_json (scalar, *SvJSON (self))); 1217 XPUSHs (encode_json (scalar, *SvJSON (self)));
1110 1218
1111void decode (SV *self, SV *jsonstr) 1219void decode (SV *self, SV *jsonstr)
1112 PPCODE: 1220 PPCODE:
1113 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1221 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1222
1223void decode_prefix (SV *self, SV *jsonstr)
1224 PPCODE:
1225{
1226 UV offset;
1227 EXTEND (SP, 2);
1228 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1229 PUSHs (sv_2mortal (newSVuv (offset)));
1230}
1114 1231
1115PROTOTYPES: ENABLE 1232PROTOTYPES: ENABLE
1116 1233
1117void to_json (SV *scalar) 1234void to_json (SV *scalar)
1118 ALIAS: 1235 ALIAS:
… …
1122 1239
1123void from_json (SV *jsonstr) 1240void from_json (SV *jsonstr)
1124 ALIAS: 1241 ALIAS:
1125 jsonToObj = 0 1242 jsonToObj = 0
1126 PPCODE: 1243 PPCODE:
1127 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8)); 1244 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1128 1245

Diff Legend

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