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 1.24 by root, Tue Apr 3 23:59:04 2007 UTC vs.
Revision 1.44 by root, Mon Jun 25 04:08:17 2007 UTC

… …
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "assert.h" 5#include "assert.h"
6#include "string.h" 6#include "string.h"
7#include "stdlib.h" 7#include "stdlib.h"
8#include "stdio.h"
8 9
10#if defined(__BORLANDC__) || defined(_MSC_VER)
11# define snprintf _snprintf // C compilers have this in stdio.h
12#endif
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
9#define F_ASCII 0x00000001UL 20#define F_ASCII 0x00000001UL
21#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002UL 22#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004UL 23#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008UL 24#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010UL 25#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020UL 26#define F_SPACE_AFTER 0x00000040UL
15#define F_ALLOW_NONREF 0x00000080UL 27#define F_ALLOW_NONREF 0x00000100UL
16#define F_SHRINK 0x00000100UL 28#define F_SHRINK 0x00000200UL
29#define F_ALLOW_BLESSED 0x00000400UL
30#define F_CONV_BLESSED 0x00000800UL // NYI
17#define F_MAXDEPTH 0xf8000000UL 31#define F_MAXDEPTH 0xf8000000UL
18#define S_MAXDEPTH 27 32#define S_MAXDEPTH 27
19 33
20#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) 34#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
21
22// F_SELFCONVERT? <=> to_json/toJson
23// F_BLESSED? <=> { $__class__$ => }
24 35
25#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 36#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
26#define F_DEFAULT (9UL << S_MAXDEPTH) 37#define F_DEFAULT (9UL << S_MAXDEPTH)
27 38
28#define INIT_SIZE 32 // initial scalar size to be allocated 39#define INIT_SIZE 32 // initial scalar size to be allocated
29#define INDENT_STEP 3 // spaces per indentation level 40#define INDENT_STEP 3 // spaces per indentation level
30 41
31#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
32#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
33 43
34#define SB do { 44#define SB do {
35#define SE } while (0) 45#define SE } while (0)
36 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
37static HV *json_stash; // JSON::XS:: 58static HV *json_stash, *json_boolean_stash; // JSON::XS::
59static SV *json_true, *json_false;
38 60
39///////////////////////////////////////////////////////////////////////////// 61/////////////////////////////////////////////////////////////////////////////
40// utility functions 62// utility functions
41 63
42static UV * 64static UV *
… …
65// 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
66// case of an error. 88// case of an error.
67// we special-case "safe" characters from U+80 .. U+7FF, 89// we special-case "safe" characters from U+80 .. U+7FF,
68// but use the very good perl function to parse anything else. 90// but use the very good perl function to parse anything else.
69// note that we never call this function for a ascii codepoints 91// note that we never call this function for a ascii codepoints
70static UV 92inline UV
71decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 93decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
72{ 94{
73 if (s[0] > 0xdf || s[0] < 0xc2) 95 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
74 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 96 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
75 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 97 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
76 { 98 {
77 *clen = 2; 99 *clen = 2;
78 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 100 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
… …
96 U32 flags; // F_* 118 U32 flags; // F_*
97 U32 indent; // indentation level 119 U32 indent; // indentation level
98 U32 maxdepth; // max. indentation/recursion level 120 U32 maxdepth; // max. indentation/recursion level
99} enc_t; 121} enc_t;
100 122
101static void 123inline void
102need (enc_t *enc, STRLEN len) 124need (enc_t *enc, STRLEN len)
103{ 125{
104 if (enc->cur + len >= enc->end) 126 if (expect_false (enc->cur + len >= enc->end))
105 { 127 {
106 STRLEN cur = enc->cur - SvPVX (enc->sv); 128 STRLEN cur = enc->cur - SvPVX (enc->sv);
107 SvGROW (enc->sv, cur + len + 1); 129 SvGROW (enc->sv, cur + len + 1);
108 enc->cur = SvPVX (enc->sv) + cur; 130 enc->cur = SvPVX (enc->sv) + cur;
109 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 131 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
110 } 132 }
111} 133}
112 134
113static void 135inline void
114encode_ch (enc_t *enc, char ch) 136encode_ch (enc_t *enc, char ch)
115{ 137{
116 need (enc, 1); 138 need (enc, 1);
117 *enc->cur++ = ch; 139 *enc->cur++ = ch;
118} 140}
… …
126 148
127 while (str < end) 149 while (str < end)
128 { 150 {
129 unsigned char ch = *(unsigned char *)str; 151 unsigned char ch = *(unsigned char *)str;
130 152
131 if (ch >= 0x20 && ch < 0x80) // most common case 153 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
132 { 154 {
133 if (ch == '"') // but with slow exceptions 155 if (expect_false (ch == '"')) // but with slow exceptions
134 { 156 {
135 need (enc, len += 1); 157 need (enc, len += 1);
136 *enc->cur++ = '\\'; 158 *enc->cur++ = '\\';
137 *enc->cur++ = '"'; 159 *enc->cur++ = '"';
138 } 160 }
139 else if (ch == '\\') 161 else if (expect_false (ch == '\\'))
140 { 162 {
141 need (enc, len += 1); 163 need (enc, len += 1);
142 *enc->cur++ = '\\'; 164 *enc->cur++ = '\\';
143 *enc->cur++ = '\\'; 165 *enc->cur++ = '\\';
144 } 166 }
… …
162 STRLEN clen; 184 STRLEN clen;
163 UV uch; 185 UV uch;
164 186
165 if (is_utf8) 187 if (is_utf8)
166 { 188 {
167 //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
168 uch = decode_utf8 (str, end - str, &clen); 189 uch = decode_utf8 (str, end - str, &clen);
169 if (clen == (STRLEN)-1) 190 if (clen == (STRLEN)-1)
170 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);
171 } 192 }
172 else 193 else
… …
176 } 197 }
177 198
178 if (uch > 0x10FFFFUL) 199 if (uch > 0x10FFFFUL)
179 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 200 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
180 201
181 if (uch < 0x80 || enc->flags & F_ASCII) 202 if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
182 { 203 {
183 if (uch > 0xFFFFUL) 204 if (uch > 0xFFFFUL)
184 { 205 {
185 need (enc, len += 11); 206 need (enc, len += 11);
186 sprintf (enc->cur, "\\u%04x\\u%04x", 207 sprintf (enc->cur, "\\u%04x\\u%04x",
… …
200 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 221 *enc->cur++ = hexdigit [(uch >> 0) & 15];
201 } 222 }
202 223
203 str += clen; 224 str += clen;
204 } 225 }
226 else if (enc->flags & F_LATIN1)
227 {
228 *enc->cur++ = uch;
229 str += clen;
230 }
205 else if (is_utf8) 231 else if (is_utf8)
206 { 232 {
207 need (enc, len += clen); 233 need (enc, len += clen);
208 do 234 do
209 { 235 {
… …
211 } 237 }
212 while (--clen); 238 while (--clen);
213 } 239 }
214 else 240 else
215 { 241 {
216 need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed 242 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
217 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 243 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
218 ++str; 244 ++str;
219 } 245 }
220 } 246 }
221 } 247 }
… …
223 249
224 --len; 250 --len;
225 } 251 }
226} 252}
227 253
228static void 254inline void
229encode_indent (enc_t *enc) 255encode_indent (enc_t *enc)
230{ 256{
231 if (enc->flags & F_INDENT) 257 if (enc->flags & F_INDENT)
232 { 258 {
233 int spaces = enc->indent * INDENT_STEP; 259 int spaces = enc->indent * INDENT_STEP;
… …
236 memset (enc->cur, ' ', spaces); 262 memset (enc->cur, ' ', spaces);
237 enc->cur += spaces; 263 enc->cur += spaces;
238 } 264 }
239} 265}
240 266
241static void 267inline void
242encode_space (enc_t *enc) 268encode_space (enc_t *enc)
243{ 269{
244 need (enc, 1); 270 need (enc, 1);
245 encode_ch (enc, ' '); 271 encode_ch (enc, ' ');
246} 272}
247 273
248static void 274inline void
249encode_nl (enc_t *enc) 275encode_nl (enc_t *enc)
250{ 276{
251 if (enc->flags & F_INDENT) 277 if (enc->flags & F_INDENT)
252 { 278 {
253 need (enc, 1); 279 need (enc, 1);
254 encode_ch (enc, '\n'); 280 encode_ch (enc, '\n');
255 } 281 }
256} 282}
257 283
258static void 284inline void
259encode_comma (enc_t *enc) 285encode_comma (enc_t *enc)
260{ 286{
261 encode_ch (enc, ','); 287 encode_ch (enc, ',');
262 288
263 if (enc->flags & F_INDENT) 289 if (enc->flags & F_INDENT)
… …
362 // actually, this is mostly due to the stupid so-called 388 // actually, this is mostly due to the stupid so-called
363 // security workaround added somewhere in 5.8.x. 389 // security workaround added somewhere in 5.8.x.
364 // that randomises hash orderings 390 // that randomises hash orderings
365 if (enc->flags & F_CANONICAL) 391 if (enc->flags & F_CANONICAL)
366 { 392 {
367 HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode
368 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
369 400
370 i = 0; 401 i = 0;
371 while ((he = hv_iternext (hv))) 402 while ((he = hv_iternext (hv)))
372 { 403 {
373 hes [i++] = he; 404 hes [i++] = he;
… …
408 439
409 encode_nl (enc); 440 encode_nl (enc);
410 } 441 }
411 else 442 else
412 { 443 {
413 SV *sv;
414 HE *he = hv_iternext (hv); 444 HE *he = hv_iternext (hv);
415 445
416 for (;;) 446 for (;;)
417 { 447 {
418 encode_indent (enc); 448 encode_indent (enc);
… …
433 463
434// encode objects, arrays and special \0=false and \1=true values. 464// encode objects, arrays and special \0=false and \1=true values.
435static void 465static void
436encode_rv (enc_t *enc, SV *sv) 466encode_rv (enc_t *enc, SV *sv)
437{ 467{
468 svtype svt;
469
438 SvGETMAGIC (sv); 470 SvGETMAGIC (sv);
439
440 svtype svt = SvTYPE (sv); 471 svt = SvTYPE (sv);
441 472
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 }
442 if (svt == SVt_PVHV) 527 else if (svt == SVt_PVHV)
443 encode_hv (enc, (HV *)sv); 528 encode_hv (enc, (HV *)sv);
444 else if (svt == SVt_PVAV) 529 else if (svt == SVt_PVAV)
445 encode_av (enc, (AV *)sv); 530 encode_av (enc, (AV *)sv);
446 else if (svt < SVt_PVAV) 531 else if (svt < SVt_PVAV)
447 { 532 {
… …
471 encode_str (enc, str, len, SvUTF8 (sv)); 556 encode_str (enc, str, len, SvUTF8 (sv));
472 encode_ch (enc, '"'); 557 encode_ch (enc, '"');
473 } 558 }
474 else if (SvNOKp (sv)) 559 else if (SvNOKp (sv))
475 { 560 {
561 // trust that perl will do the right thing w.r.t. JSON syntax.
476 need (enc, NV_DIG + 32); 562 need (enc, NV_DIG + 32);
477 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 563 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
478 enc->cur += strlen (enc->cur); 564 enc->cur += strlen (enc->cur);
479 } 565 }
480 else if (SvIOKp (sv)) 566 else if (SvIOKp (sv))
481 { 567 {
482 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);
483 enc->cur += 573 enc->cur +=
484 SvIsUV(sv) 574 SvIsUV(sv)
485 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 575 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
486 : 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 }
487 } 605 }
488 else if (SvROK (sv)) 606 else if (SvROK (sv))
489 encode_rv (enc, SvRV (sv)); 607 encode_rv (enc, SvRV (sv));
490 else if (!SvOK (sv)) 608 else if (!SvOK (sv))
491 encode_str (enc, "null", 4, 0); 609 encode_str (enc, "null", 4, 0);
… …
495} 613}
496 614
497static SV * 615static SV *
498encode_json (SV *scalar, U32 flags) 616encode_json (SV *scalar, U32 flags)
499{ 617{
618 enc_t enc;
619
500 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) 620 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
501 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 621 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
502 622
503 enc_t enc;
504 enc.flags = flags; 623 enc.flags = flags;
505 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 624 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
506 enc.cur = SvPVX (enc.sv); 625 enc.cur = SvPVX (enc.sv);
507 enc.end = SvEND (enc.sv); 626 enc.end = SvEND (enc.sv);
508 enc.indent = 0; 627 enc.indent = 0;
509 enc.maxdepth = DEC_DEPTH (flags); 628 enc.maxdepth = DEC_DEPTH (flags);
510 629
511 SvPOK_only (enc.sv); 630 SvPOK_only (enc.sv);
512 encode_sv (&enc, scalar); 631 encode_sv (&enc, scalar);
513 632
633 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
634 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
635
514 if (!(flags & (F_ASCII | F_UTF8))) 636 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
515 SvUTF8_on (enc.sv); 637 SvUTF8_on (enc.sv);
516
517 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
518 638
519 if (enc.flags & F_SHRINK) 639 if (enc.flags & F_SHRINK)
520 shrink (enc.sv); 640 shrink (enc.sv);
521 641
522 return enc.sv; 642 return enc.sv;
… …
534 U32 flags; // F_* 654 U32 flags; // F_*
535 U32 depth; // recursion depth 655 U32 depth; // recursion depth
536 U32 maxdepth; // recursion depth limit 656 U32 maxdepth; // recursion depth limit
537} dec_t; 657} dec_t;
538 658
539static void 659inline void
540decode_ws (dec_t *dec) 660decode_ws (dec_t *dec)
541{ 661{
542 for (;;) 662 for (;;)
543 { 663 {
544 char ch = *dec->cur; 664 char ch = *dec->cur;
… …
570decode_4hex (dec_t *dec) 690decode_4hex (dec_t *dec)
571{ 691{
572 signed char d1, d2, d3, d4; 692 signed char d1, d2, d3, d4;
573 unsigned char *cur = (unsigned char *)dec->cur; 693 unsigned char *cur = (unsigned char *)dec->cur;
574 694
575 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");
576 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");
577 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");
578 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");
579 699
580 dec->cur += 4; 700 dec->cur += 4;
581 701
582 return ((UV)d1) << 12 702 return ((UV)d1) << 12
583 | ((UV)d2) << 8 703 | ((UV)d2) << 8
… …
591static SV * 711static SV *
592decode_str (dec_t *dec) 712decode_str (dec_t *dec)
593{ 713{
594 SV *sv = 0; 714 SV *sv = 0;
595 int utf8 = 0; 715 int utf8 = 0;
716 char *dec_cur = dec->cur;
596 717
597 do 718 do
598 { 719 {
599 char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; 720 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
600 char *cur = buf; 721 char *cur = buf;
601 722
602 do 723 do
603 { 724 {
604 unsigned char ch = *(unsigned char *)dec->cur++; 725 unsigned char ch = *(unsigned char *)dec_cur++;
605 726
606 if (ch == '"') 727 if (expect_false (ch == '"'))
607 { 728 {
608 --dec->cur; 729 --dec_cur;
609 break; 730 break;
610 } 731 }
611 else if (ch == '\\') 732 else if (expect_false (ch == '\\'))
612 { 733 {
613 switch (*dec->cur) 734 switch (*dec_cur)
614 { 735 {
615 case '\\': 736 case '\\':
616 case '/': 737 case '/':
617 case '"': *cur++ = *dec->cur++; break; 738 case '"': *cur++ = *dec_cur++; break;
618 739
619 case 'b': ++dec->cur; *cur++ = '\010'; break; 740 case 'b': ++dec_cur; *cur++ = '\010'; break;
620 case 't': ++dec->cur; *cur++ = '\011'; break; 741 case 't': ++dec_cur; *cur++ = '\011'; break;
621 case 'n': ++dec->cur; *cur++ = '\012'; break; 742 case 'n': ++dec_cur; *cur++ = '\012'; break;
622 case 'f': ++dec->cur; *cur++ = '\014'; break; 743 case 'f': ++dec_cur; *cur++ = '\014'; break;
623 case 'r': ++dec->cur; *cur++ = '\015'; break; 744 case 'r': ++dec_cur; *cur++ = '\015'; break;
624 745
625 case 'u': 746 case 'u':
626 { 747 {
627 UV lo, hi; 748 UV lo, hi;
628 ++dec->cur; 749 ++dec_cur;
629 750
751 dec->cur = dec_cur;
630 hi = decode_4hex (dec); 752 hi = decode_4hex (dec);
753 dec_cur = dec->cur;
631 if (hi == (UV)-1) 754 if (hi == (UV)-1)
632 goto fail; 755 goto fail;
633 756
634 // possibly a surrogate pair 757 // possibly a surrogate pair
635 if (hi >= 0xd800) 758 if (hi >= 0xd800)
636 if (hi < 0xdc00) 759 if (hi < 0xdc00)
637 { 760 {
638 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 761 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
639 ERR ("missing low surrogate character in surrogate pair"); 762 ERR ("missing low surrogate character in surrogate pair");
640 763
641 dec->cur += 2; 764 dec_cur += 2;
642 765
766 dec->cur = dec_cur;
643 lo = decode_4hex (dec); 767 lo = decode_4hex (dec);
768 dec_cur = dec->cur;
644 if (lo == (UV)-1) 769 if (lo == (UV)-1)
645 goto fail; 770 goto fail;
646 771
647 if (lo < 0xdc00 || lo >= 0xe000) 772 if (lo < 0xdc00 || lo >= 0xe000)
648 ERR ("surrogate pair expected"); 773 ERR ("surrogate pair expected");
… …
662 *cur++ = hi; 787 *cur++ = hi;
663 } 788 }
664 break; 789 break;
665 790
666 default: 791 default:
667 --dec->cur; 792 --dec_cur;
668 ERR ("illegal backslash escape sequence in string"); 793 ERR ("illegal backslash escape sequence in string");
669 } 794 }
670 } 795 }
671 else if (ch >= 0x20 && ch <= 0x7f) 796 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
672 *cur++ = ch; 797 *cur++ = ch;
673 else if (ch >= 0x80) 798 else if (ch >= 0x80)
674 { 799 {
675 --dec->cur;
676
677 STRLEN clen; 800 STRLEN clen;
801 UV uch;
802
803 --dec_cur;
804
678 UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); 805 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
679 if (clen == (STRLEN)-1) 806 if (clen == (STRLEN)-1)
680 ERR ("malformed UTF-8 character in JSON string"); 807 ERR ("malformed UTF-8 character in JSON string");
681 808
682 do 809 do
683 *cur++ = *dec->cur++; 810 *cur++ = *dec_cur++;
684 while (--clen); 811 while (--clen);
685 812
686 utf8 = 1; 813 utf8 = 1;
687 } 814 }
688 else 815 else
689 { 816 {
690 --dec->cur; 817 --dec_cur;
691 818
692 if (!ch) 819 if (!ch)
693 ERR ("unexpected end of string while parsing JSON string"); 820 ERR ("unexpected end of string while parsing JSON string");
694 else 821 else
695 ERR ("invalid character encountered while parsing JSON string"); 822 ERR ("invalid character encountered while parsing JSON string");
696 } 823 }
697 } 824 }
698 while (cur < buf + SHORT_STRING_LEN); 825 while (cur < buf + SHORT_STRING_LEN);
699 826
827 {
700 STRLEN len = cur - buf; 828 STRLEN len = cur - buf;
701 829
702 if (sv) 830 if (sv)
703 { 831 {
704 SvGROW (sv, SvCUR (sv) + len + 1); 832 SvGROW (sv, SvCUR (sv) + len + 1);
705 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 833 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
706 SvCUR_set (sv, SvCUR (sv) + len); 834 SvCUR_set (sv, SvCUR (sv) + len);
707 } 835 }
708 else 836 else
709 sv = newSVpvn (buf, len); 837 sv = newSVpvn (buf, len);
710 } 838 }
839 }
711 while (*dec->cur != '"'); 840 while (*dec_cur != '"');
712 841
713 ++dec->cur; 842 ++dec_cur;
714 843
715 if (sv) 844 if (sv)
716 { 845 {
717 SvPOK_only (sv); 846 SvPOK_only (sv);
718 *SvEND (sv) = 0; 847 *SvEND (sv) = 0;
… …
721 SvUTF8_on (sv); 850 SvUTF8_on (sv);
722 } 851 }
723 else 852 else
724 sv = newSVpvn ("", 0); 853 sv = newSVpvn ("", 0);
725 854
855 dec->cur = dec_cur;
726 return sv; 856 return sv;
727 857
728fail: 858fail:
859 dec->cur = dec_cur;
729 return 0; 860 return 0;
730} 861}
731 862
732static SV * 863static SV *
733decode_num (dec_t *dec) 864decode_num (dec_t *dec)
… …
791 is_nv = 1; 922 is_nv = 1;
792 } 923 }
793 924
794 if (!is_nv) 925 if (!is_nv)
795 { 926 {
796 UV uv; 927 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
797 int numtype = grok_number (start, dec->cur - start, &uv); 928 if (*start == '-')
798 if (numtype & IS_NUMBER_IN_UV) 929 switch (dec->cur - start)
799 if (numtype & IS_NUMBER_NEG)
800 { 930 {
801 if (uv < (UV)IV_MIN) 931 case 2: return newSViv (-( start [1] - '0' * 1));
802 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));
803 } 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 }
804 else 954 else
805 return newSVuv (uv); 955 return newSVuv (uv);
956
957 // here would likely be the place for bigint support
806 } 958 }
959 }
807 960
961 // if we ever support bigint or bigfloat, this is the place for bigfloat
808 return newSVnv (Atof (start)); 962 return newSVnv (Atof (start));
809 963
810fail: 964fail:
811 return 0; 965 return 0;
812} 966}
… …
913 1067
914static SV * 1068static SV *
915decode_sv (dec_t *dec) 1069decode_sv (dec_t *dec)
916{ 1070{
917 decode_ws (dec); 1071 decode_ws (dec);
1072
1073 // the beauty of JSON: you need exactly one character lookahead
1074 // to parse anything.
918 switch (*dec->cur) 1075 switch (*dec->cur)
919 { 1076 {
920 case '"': ++dec->cur; return decode_str (dec); 1077 case '"': ++dec->cur; return decode_str (dec);
921 case '[': ++dec->cur; return decode_av (dec); 1078 case '[': ++dec->cur; return decode_av (dec);
922 case '{': ++dec->cur; return decode_hv (dec); 1079 case '{': ++dec->cur; return decode_hv (dec);
… …
928 1085
929 case 't': 1086 case 't':
930 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1087 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
931 { 1088 {
932 dec->cur += 4; 1089 dec->cur += 4;
933 return newSViv (1); 1090 return SvREFCNT_inc (json_true);
934 } 1091 }
935 else 1092 else
936 ERR ("'true' expected"); 1093 ERR ("'true' expected");
937 1094
938 break; 1095 break;
939 1096
940 case 'f': 1097 case 'f':
941 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1098 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
942 { 1099 {
943 dec->cur += 5; 1100 dec->cur += 5;
944 return newSViv (0); 1101 return SvREFCNT_inc (json_false);
945 } 1102 }
946 else 1103 else
947 ERR ("'false' expected"); 1104 ERR ("'false' expected");
948 1105
949 break; 1106 break;
… …
967fail: 1124fail:
968 return 0; 1125 return 0;
969} 1126}
970 1127
971static SV * 1128static SV *
972decode_json (SV *string, U32 flags) 1129decode_json (SV *string, U32 flags, UV *offset_return)
973{ 1130{
1131 dec_t dec;
1132 UV offset;
974 SV *sv; 1133 SV *sv;
975 1134
1135 SvGETMAGIC (string);
976 SvUPGRADE (string, SVt_PV); 1136 SvUPGRADE (string, SVt_PV);
977 1137
978 if (flags & F_UTF8) 1138 if (flags & F_UTF8)
979 sv_utf8_downgrade (string, 0); 1139 sv_utf8_downgrade (string, 0);
980 else 1140 else
981 sv_utf8_upgrade (string); 1141 sv_utf8_upgrade (string);
982 1142
983 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1143 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
984 1144
985 dec_t dec;
986 dec.flags = flags; 1145 dec.flags = flags;
987 dec.cur = SvPVX (string); 1146 dec.cur = SvPVX (string);
988 dec.end = SvEND (string); 1147 dec.end = SvEND (string);
989 dec.err = 0; 1148 dec.err = 0;
990 dec.depth = 0; 1149 dec.depth = 0;
991 dec.maxdepth = DEC_DEPTH (dec.flags); 1150 dec.maxdepth = DEC_DEPTH (dec.flags);
992 1151
993 *dec.end = 0; // this should basically be a nop, too, but make sure its there 1152 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
994 sv = decode_sv (&dec); 1153 sv = decode_sv (&dec);
995 1154
1155 if (!(offset_return || !sv))
1156 {
1157 // check for trailing garbage
1158 decode_ws (&dec);
1159
1160 if (*dec.cur)
1161 {
1162 dec.err = "garbage after JSON object";
1163 SvREFCNT_dec (sv);
1164 sv = 0;
1165 }
1166 }
1167
1168 if (offset_return || !sv)
1169 {
1170 offset = dec.flags & F_UTF8
1171 ? dec.cur - SvPVX (string)
1172 : utf8_distance (dec.cur, SvPVX (string));
1173
1174 if (offset_return)
1175 *offset_return = offset;
1176 }
1177
996 if (!sv) 1178 if (!sv)
997 { 1179 {
998 IV offset = dec.flags & F_UTF8
999 ? dec.cur - SvPVX (string)
1000 : utf8_distance (dec.cur, SvPVX (string));
1001 SV *uni = sv_newmortal (); 1180 SV *uni = sv_newmortal ();
1002 1181
1003 // horrible hack to silence warning inside pv_uni_display 1182 // horrible hack to silence warning inside pv_uni_display
1004 COP cop = *PL_curcop; 1183 COP cop = *PL_curcop;
1005 cop.cop_warnings = pWARN_NONE; 1184 cop.cop_warnings = pWARN_NONE;
… …
1030 1209
1031BOOT: 1210BOOT:
1032{ 1211{
1033 int i; 1212 int i;
1034 1213
1035 memset (decode_hexdigit, 0xff, 256);
1036
1037 for (i = 0; i < 256; ++i) 1214 for (i = 0; i < 256; ++i)
1038 decode_hexdigit [i] = 1215 decode_hexdigit [i] =
1039 i >= '0' && i <= '9' ? i - '0' 1216 i >= '0' && i <= '9' ? i - '0'
1040 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1217 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1041 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1218 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1042 : -1; 1219 : -1;
1043 1220
1044 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);
1045} 1226}
1046 1227
1047PROTOTYPES: DISABLE 1228PROTOTYPES: DISABLE
1048 1229
1049SV *new (char *dummy) 1230SV *new (char *dummy)
… …
1052 OUTPUT: 1233 OUTPUT:
1053 RETVAL 1234 RETVAL
1054 1235
1055SV *ascii (SV *self, int enable = 1) 1236SV *ascii (SV *self, int enable = 1)
1056 ALIAS: 1237 ALIAS:
1057 ascii = F_ASCII 1238 ascii = F_ASCII
1239 latin1 = F_LATIN1
1058 utf8 = F_UTF8 1240 utf8 = F_UTF8
1059 indent = F_INDENT 1241 indent = F_INDENT
1060 canonical = F_CANONICAL 1242 canonical = F_CANONICAL
1061 space_before = F_SPACE_BEFORE 1243 space_before = F_SPACE_BEFORE
1062 space_after = F_SPACE_AFTER 1244 space_after = F_SPACE_AFTER
1063 pretty = F_PRETTY 1245 pretty = F_PRETTY
1064 allow_nonref = F_ALLOW_NONREF 1246 allow_nonref = F_ALLOW_NONREF
1065 shrink = F_SHRINK 1247 shrink = F_SHRINK
1248 allow_blessed = F_ALLOW_BLESSED
1249 convert_blessed = F_CONV_BLESSED
1066 CODE: 1250 CODE:
1067{ 1251{
1068 UV *uv = SvJSON (self); 1252 UV *uv = SvJSON (self);
1069 if (enable) 1253 if (enable)
1070 *uv |= ix; 1254 *uv |= ix;
… …
1074 RETVAL = newSVsv (self); 1258 RETVAL = newSVsv (self);
1075} 1259}
1076 OUTPUT: 1260 OUTPUT:
1077 RETVAL 1261 RETVAL
1078 1262
1079SV *max_depth (SV *self, int max_depth = 0x80000000UL) 1263SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
1080 CODE: 1264 CODE:
1081{ 1265{
1082 UV *uv = SvJSON (self); 1266 UV *uv = SvJSON (self);
1083 UV log2 = 0; 1267 UV log2 = 0;
1084 1268
… …
1098 PPCODE: 1282 PPCODE:
1099 XPUSHs (encode_json (scalar, *SvJSON (self))); 1283 XPUSHs (encode_json (scalar, *SvJSON (self)));
1100 1284
1101void decode (SV *self, SV *jsonstr) 1285void decode (SV *self, SV *jsonstr)
1102 PPCODE: 1286 PPCODE:
1103 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1287 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1288
1289void decode_prefix (SV *self, SV *jsonstr)
1290 PPCODE:
1291{
1292 UV offset;
1293 EXTEND (SP, 2);
1294 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1295 PUSHs (sv_2mortal (newSVuv (offset)));
1296}
1104 1297
1105PROTOTYPES: ENABLE 1298PROTOTYPES: ENABLE
1106 1299
1107void to_json (SV *scalar) 1300void to_json (SV *scalar)
1108 ALIAS: 1301 ALIAS:
… …
1112 1305
1113void from_json (SV *jsonstr) 1306void from_json (SV *jsonstr)
1114 ALIAS: 1307 ALIAS:
1115 jsonToObj = 0 1308 jsonToObj = 0
1116 PPCODE: 1309 PPCODE:
1117 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8)); 1310 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1118 1311

Diff Legend

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