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 1.75 by root, Wed Mar 19 15:37:54 2008 UTC vs.
Revision 1.119 by root, Mon Oct 28 23:19:54 2013 UTC

12#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 14#endif
15 15
16// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
20#endif 20#endif
21 21
22// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 24
24#define F_ASCII 0x00000001UL 25#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 26#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 27#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 28#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 32#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL 33#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 34#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 35#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 36#define F_RELAXED 0x00001000UL
36 37#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL 38#define F_ALLOW_TAGS 0x00004000UL
38#define S_MAXDEPTH 27
39#define F_MAXSIZE 0x01f00000UL
40#define S_MAXSIZE 20
41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 39#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 40
43#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
44#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
45
46#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 41#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 42
49#define INIT_SIZE 32 // initial scalar size to be allocated 43#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 44#define INDENT_STEP 3 // spaces per indentation level
51 45
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 46#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
47
48#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
53 49
54#define SB do { 50#define SB do {
55#define SE } while (0) 51#define SE } while (0)
56 52
57#if __GNUC__ >= 3 53#if __GNUC__ >= 3
67 63
68#define IN_RANGE_INC(type,val,beg,end) \ 64#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 65 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 66 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 67
68#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
69
72#ifdef USE_ITHREADS 70#ifdef USE_ITHREADS
73# define JSON_SLOW 1 71# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 72# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
75#else 73#else
76# define JSON_SLOW 0 74# define JSON_SLOW 0
77# define JSON_STASH json_stash 75# define JSON_STASH json_stash
78#endif 76#endif
79 77
78// the amount of HEs to allocate on the stack, when sorting keys
79#define STACK_HES 64
80
80static HV *json_stash, *json_boolean_stash; // JSON::XS:: 81static HV *json_stash, *types_boolean_stash; // JSON::XS::
81static SV *json_true, *json_false; 82static SV *types_true, *types_false, *sv_json;
83
84enum {
85 INCR_M_WS = 0, // initial whitespace skipping, must be 0
86 INCR_M_STR, // inside string
87 INCR_M_BS, // inside backslash
88 INCR_M_C0, // inside comment in initial whitespace sequence
89 INCR_M_C1, // inside comment in other places
90 INCR_M_JSON // outside anything, count nesting
91};
92
93#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
82 94
83typedef struct { 95typedef struct {
84 U32 flags; 96 U32 flags;
97 U32 max_depth;
98 STRLEN max_size;
99
85 SV *cb_object; 100 SV *cb_object;
86 HV *cb_sk_object; 101 HV *cb_sk_object;
102
103 // for the incremental parser
104 SV *incr_text; // the source text so far
105 STRLEN incr_pos; // the current offset into the text
106 int incr_nest; // {[]}-nesting level
107 unsigned char incr_mode;
87} JSON; 108} JSON;
109
110INLINE void
111json_init (JSON *json)
112{
113 Zero (json, 1, JSON);
114 json->max_depth = 512;
115}
88 116
89///////////////////////////////////////////////////////////////////////////// 117/////////////////////////////////////////////////////////////////////////////
90// utility functions 118// utility functions
91 119
120INLINE SV *
121get_bool (const char *name)
122{
123 SV *sv = get_sv (name, 1);
124
125 SvREADONLY_on (sv);
126 SvREADONLY_on (SvRV (sv));
127
128 return sv;
129}
130
92INLINE void 131INLINE void
93shrink (SV *sv) 132shrink (SV *sv)
94{ 133{
95 sv_utf8_downgrade (sv, 1); 134 sv_utf8_downgrade (sv, 1);
135
96 if (SvLEN (sv) > SvCUR (sv) + 1) 136 if (SvLEN (sv) > SvCUR (sv) + 1)
97 { 137 {
98#ifdef SvPV_shrink_to_cur 138#ifdef SvPV_shrink_to_cur
99 SvPV_shrink_to_cur (sv); 139 SvPV_shrink_to_cur (sv);
100#elif defined (SvPV_renew) 140#elif defined (SvPV_renew)
144 *s++ = 0x80 | ( ch & 0x3f); 184 *s++ = 0x80 | ( ch & 0x3f);
145 185
146 return s; 186 return s;
147} 187}
148 188
189// convert offset pointer to character index, sv must be string
190static STRLEN
191ptr_to_index (SV *sv, char *offset)
192{
193 return SvUTF8 (sv)
194 ? utf8_distance (offset, SvPVX (sv))
195 : offset - SvPVX (sv);
196}
197
198/////////////////////////////////////////////////////////////////////////////
199// fp hell
200
201// scan a group of digits, and a trailing exponent
202static void
203json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
204{
205 UV uaccum = 0;
206 int eaccum = 0;
207
208 // if we recurse too deep, skip all remaining digits
209 // to avoid a stack overflow attack
210 if (expect_false (--maxdepth <= 0))
211 while (((U8)*s - '0') < 10)
212 ++s;
213
214 for (;;)
215 {
216 U8 dig = (U8)*s - '0';
217
218 if (expect_false (dig >= 10))
219 {
220 if (dig == (U8)((U8)'.' - (U8)'0'))
221 {
222 ++s;
223 json_atof_scan1 (s, accum, expo, 1, maxdepth);
224 }
225 else if ((dig | ' ') == 'e' - '0')
226 {
227 int exp2 = 0;
228 int neg = 0;
229
230 ++s;
231
232 if (*s == '-')
233 {
234 ++s;
235 neg = 1;
236 }
237 else if (*s == '+')
238 ++s;
239
240 while ((dig = (U8)*s - '0') < 10)
241 exp2 = exp2 * 10 + *s++ - '0';
242
243 *expo += neg ? -exp2 : exp2;
244 }
245
246 break;
247 }
248
249 ++s;
250
251 uaccum = uaccum * 10 + dig;
252 ++eaccum;
253
254 // if we have too many digits, then recurse for more
255 // we actually do this for rather few digits
256 if (uaccum >= (UV_MAX - 9) / 10)
257 {
258 if (postdp) *expo -= eaccum;
259 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
260 if (postdp) *expo += eaccum;
261
262 break;
263 }
264 }
265
266 // this relies greatly on the quality of the pow ()
267 // implementation of the platform, but a good
268 // implementation is hard to beat.
269 // (IEEE 754 conformant ones are required to be exact)
270 if (postdp) *expo -= eaccum;
271 *accum += uaccum * Perl_pow (10., *expo);
272 *expo += eaccum;
273}
274
275static NV
276json_atof (const char *s)
277{
278 NV accum = 0.;
279 int expo = 0;
280 int neg = 0;
281
282 if (*s == '-')
283 {
284 ++s;
285 neg = 1;
286 }
287
288 // a recursion depth of ten gives us >>500 bits
289 json_atof_scan1 (s, &accum, &expo, 0, 10);
290
291 return neg ? -accum : accum;
292}
149///////////////////////////////////////////////////////////////////////////// 293/////////////////////////////////////////////////////////////////////////////
150// encoder 294// encoder
151 295
152// structure used for encoding JSON 296// structure used for encoding JSON
153typedef struct 297typedef struct
155 char *cur; // SvPVX (sv) + current output position 299 char *cur; // SvPVX (sv) + current output position
156 char *end; // SvEND (sv) 300 char *end; // SvEND (sv)
157 SV *sv; // result scalar 301 SV *sv; // result scalar
158 JSON json; 302 JSON json;
159 U32 indent; // indentation level 303 U32 indent; // indentation level
160 U32 maxdepth; // max. indentation/recursion level
161 UV limit; // escape character values >= this value when encoding 304 UV limit; // escape character values >= this value when encoding
162} enc_t; 305} enc_t;
163 306
164INLINE void 307INLINE void
165need (enc_t *enc, STRLEN len) 308need (enc_t *enc, STRLEN len)
166{ 309{
167 if (expect_false (enc->cur + len >= enc->end)) 310 if (expect_false (enc->cur + len >= enc->end))
168 { 311 {
169 STRLEN cur = enc->cur - SvPVX (enc->sv); 312 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
170 SvGROW (enc->sv, cur + len + 1); 313 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
171 enc->cur = SvPVX (enc->sv) + cur; 314 enc->cur = SvPVX (enc->sv) + cur;
172 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 315 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
173 } 316 }
174} 317}
175 318
250 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 393 (int)((uch - 0x10000) % 0x400 + 0xDC00));
251 enc->cur += 12; 394 enc->cur += 12;
252 } 395 }
253 else 396 else
254 { 397 {
255 static char hexdigit [16] = "0123456789abcdef";
256 need (enc, len += 5); 398 need (enc, len += 5);
257 *enc->cur++ = '\\'; 399 *enc->cur++ = '\\';
258 *enc->cur++ = 'u'; 400 *enc->cur++ = 'u';
259 *enc->cur++ = hexdigit [ uch >> 12 ]; 401 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
260 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 402 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
261 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 403 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
262 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 404 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
263 } 405 }
264 406
265 str += clen; 407 str += clen;
266 } 408 }
267 else if (enc->json.flags & F_LATIN1) 409 else if (enc->json.flags & F_LATIN1)
338static void 480static void
339encode_av (enc_t *enc, AV *av) 481encode_av (enc_t *enc, AV *av)
340{ 482{
341 int i, len = av_len (av); 483 int i, len = av_len (av);
342 484
343 if (enc->indent >= enc->maxdepth) 485 if (enc->indent >= enc->json.max_depth)
344 croak ("data structure too deep (hit recursion limit)"); 486 croak (ERR_NESTING_EXCEEDED);
345 487
346 encode_ch (enc, '['); 488 encode_ch (enc, '[');
347 489
348 if (len >= 0) 490 if (len >= 0)
349 { 491 {
350 encode_nl (enc); ++enc->indent; 492 encode_nl (enc); ++enc->indent;
351 493
352 for (i = 0; i <= len; ++i) 494 for (i = 0; i <= len; ++i)
364 encode_comma (enc); 506 encode_comma (enc);
365 } 507 }
366 508
367 encode_nl (enc); --enc->indent; encode_indent (enc); 509 encode_nl (enc); --enc->indent; encode_indent (enc);
368 } 510 }
369 511
370 encode_ch (enc, ']'); 512 encode_ch (enc, ']');
371} 513}
372 514
373static void 515static void
374encode_hk (enc_t *enc, HE *he) 516encode_hk (enc_t *enc, HE *he)
378 if (HeKLEN (he) == HEf_SVKEY) 520 if (HeKLEN (he) == HEf_SVKEY)
379 { 521 {
380 SV *sv = HeSVKEY (he); 522 SV *sv = HeSVKEY (he);
381 STRLEN len; 523 STRLEN len;
382 char *str; 524 char *str;
383 525
384 SvGETMAGIC (sv); 526 SvGETMAGIC (sv);
385 str = SvPV (sv, len); 527 str = SvPV (sv, len);
386 528
387 encode_str (enc, str, len, SvUTF8 (sv)); 529 encode_str (enc, str, len, SvUTF8 (sv));
388 } 530 }
423 565
424static void 566static void
425encode_hv (enc_t *enc, HV *hv) 567encode_hv (enc_t *enc, HV *hv)
426{ 568{
427 HE *he; 569 HE *he;
428 int count;
429 570
430 if (enc->indent >= enc->maxdepth) 571 if (enc->indent >= enc->json.max_depth)
431 croak ("data structure too deep (hit recursion limit)"); 572 croak (ERR_NESTING_EXCEEDED);
432 573
433 encode_ch (enc, '{'); 574 encode_ch (enc, '{');
434 575
435 // for canonical output we have to sort by keys first 576 // for canonical output we have to sort by keys first
436 // actually, this is mostly due to the stupid so-called 577 // actually, this is mostly due to the stupid so-called
437 // security workaround added somewhere in 5.8.x. 578 // security workaround added somewhere in 5.8.x
438 // that randomises hash orderings 579 // that randomises hash orderings
439 if (enc->json.flags & F_CANONICAL) 580 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
440 { 581 {
441 int count = hv_iterinit (hv); 582 int count = hv_iterinit (hv);
442 583
443 if (SvMAGICAL (hv)) 584 if (SvMAGICAL (hv))
444 { 585 {
455 } 596 }
456 597
457 if (count) 598 if (count)
458 { 599 {
459 int i, fast = 1; 600 int i, fast = 1;
460#if defined(__BORLANDC__) || defined(_MSC_VER) 601 HE *hes_stack [STACK_HES];
461 HE **hes = _alloca (count * sizeof (HE)); 602 HE **hes = hes_stack;
462#else 603
463 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 604 // allocate larger arrays on the heap
464#endif 605 if (count > STACK_HES)
606 {
607 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
608 hes = (HE **)SvPVX (sv);
609 }
465 610
466 i = 0; 611 i = 0;
467 while ((he = hv_iternext (hv))) 612 while ((he = hv_iternext (hv)))
468 { 613 {
469 hes [i++] = he; 614 hes [i++] = he;
544 SvGETMAGIC (sv); 689 SvGETMAGIC (sv);
545 svt = SvTYPE (sv); 690 svt = SvTYPE (sv);
546 691
547 if (expect_false (SvOBJECT (sv))) 692 if (expect_false (SvOBJECT (sv)))
548 { 693 {
549 HV *stash = !JSON_SLOW || json_boolean_stash 694 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
550 ? json_boolean_stash 695 ? types_boolean_stash
551 : gv_stashpv ("JSON::XS::Boolean", 1); 696 : gv_stashpv ("Types::Serialiser::Boolean", 1);
697 HV *stash = SvSTASH (sv);
552 698
553 if (SvSTASH (sv) == stash) 699 if (stash == boolean_stash)
554 { 700 {
555 if (SvIV (sv)) 701 if (SvIV (sv))
556 encode_str (enc, "true", 4, 0); 702 encode_str (enc, "true", 4, 0);
557 else 703 else
558 encode_str (enc, "false", 5, 0); 704 encode_str (enc, "false", 5, 0);
559 } 705 }
560 else 706 else if (enc->json.flags & F_CONV_BLESSED)
561 { 707 {
562#if 0 708 GV *to_json = gv_fetchmethod_autoload (stash, "TO_JSON", 0);
563 if (0 && sv_derived_from (rv, "JSON::Literal")) 709
564 { 710 if (to_json)
565 // not yet
566 } 711 {
567#endif
568 if (enc->json.flags & F_CONV_BLESSED)
569 { 712 dSP;
713
714 ENTER; SAVETMPS; PUSHMARK (SP);
570 // we re-bless the reference to get overload and other niceties right 715 // we re-bless the reference to get overload and other niceties right
571 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
572
573 if (to_json)
574 {
575 dSP;
576
577 ENTER; SAVETMPS; PUSHMARK (SP);
578 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); 716 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
579 717
580 // calling with G_SCALAR ensures that we always get a 1 return value 718 // calling with G_SCALAR ensures that we always get a 1 return value
581 PUTBACK; 719 PUTBACK;
582 call_sv ((SV *)GvCV (to_json), G_SCALAR); 720 call_sv ((SV *)GvCV (to_json), G_SCALAR);
583 SPAGAIN; 721 SPAGAIN;
584 722
585 // catch this surprisingly common error 723 // catch this surprisingly common error
586 if (SvROK (TOPs) && SvRV (TOPs) == sv) 724 if (SvROK (TOPs) && SvRV (TOPs) == sv)
587 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); 725 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
588 726
589 sv = POPs; 727 sv = POPs;
590 PUTBACK; 728 PUTBACK;
591 729
592 encode_sv (enc, sv); 730 encode_sv (enc, sv);
593 731
594 FREETMPS; LEAVE; 732 FREETMPS; LEAVE;
595 }
596 else if (enc->json.flags & F_ALLOW_BLESSED)
597 encode_str (enc, "null", 4, 0);
598 else
599 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
600 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
601 } 733 }
602 else if (enc->json.flags & F_ALLOW_BLESSED) 734 else if (enc->json.flags & F_ALLOW_BLESSED)
603 encode_str (enc, "null", 4, 0); 735 encode_str (enc, "null", 4, 0);
604 else 736 else
605 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 737 croak ("encountered object '%s' when conv_object is enabled, but TO_JSON is missing and allow_blessed disabled",
606 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 738 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
607 } 739 }
740 else if (enc->json.flags & F_ALLOW_TAGS)
741 {
742 GV *method = gv_fetchmethod_autoload (stash, "FREEZE", 0);
743
744 if (method)
745 {
746 int count;
747 dSP;
748
749 ENTER; SAVETMPS; PUSHMARK (SP);
750 EXTEND (SP, 2);
751 // we re-bless the reference to get overload and other niceties right
752 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
753 PUSHs (sv_json);
754
755 PUTBACK;
756 count = call_sv ((SV *)GvCV (method), G_ARRAY);
757 SPAGAIN;
758
759 // catch this surprisingly common error
760 if (SvROK (TOPs) && SvRV (TOPs) == sv)
761 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
762
763 encode_ch (enc, '(');
764 encode_ch (enc, '"');
765 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
766 encode_ch (enc, '"');
767 encode_ch (enc, ')');
768 encode_ch (enc, '[');
769
770 while (count)
771 {
772 encode_sv (enc, SP[1 - count--]);
773
774 if (count)
775 encode_ch (enc, ',');
776 }
777
778 encode_ch (enc, ']');
779
780 PUTBACK;
781
782 FREETMPS; LEAVE;
783 }
784 else if (enc->json.flags & F_ALLOW_BLESSED)
785 encode_str (enc, "null", 4, 0);
786 else
787 croak ("encountered object '%s' when allow_tags is enabled, but FREEZE is missing and allow_blessed disabled",
788 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
789 }
790 else if (enc->json.flags & F_ALLOW_BLESSED)
791 encode_str (enc, "null", 4, 0);
792 else
793 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled",
794 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
608 } 795 }
609 else if (svt == SVt_PVHV) 796 else if (svt == SVt_PVHV)
610 encode_hv (enc, (HV *)sv); 797 encode_hv (enc, (HV *)sv);
611 else if (svt == SVt_PVAV) 798 else if (svt == SVt_PVAV)
612 encode_av (enc, (AV *)sv); 799 encode_av (enc, (AV *)sv);
617 804
618 if (len == 1 && *pv == '1') 805 if (len == 1 && *pv == '1')
619 encode_str (enc, "true", 4, 0); 806 encode_str (enc, "true", 4, 0);
620 else if (len == 1 && *pv == '0') 807 else if (len == 1 && *pv == '0')
621 encode_str (enc, "false", 5, 0); 808 encode_str (enc, "false", 5, 0);
809 else if (enc->json.flags & F_ALLOW_UNKNOWN)
810 encode_str (enc, "null", 4, 0);
622 else 811 else
623 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 812 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
624 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 813 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
625 } 814 }
815 else if (enc->json.flags & F_ALLOW_UNKNOWN)
816 encode_str (enc, "null", 4, 0);
626 else 817 else
627 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 818 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
628 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 819 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
629} 820}
630 821
684 } 875 }
685 else 876 else
686 { 877 {
687 // large integer, use the (rather slow) snprintf way. 878 // large integer, use the (rather slow) snprintf way.
688 need (enc, IVUV_MAXCHARS); 879 need (enc, IVUV_MAXCHARS);
689 enc->cur += 880 enc->cur +=
690 SvIsUV(sv) 881 SvIsUV(sv)
691 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 882 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
692 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 883 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
693 } 884 }
694 } 885 }
695 else if (SvROK (sv)) 886 else if (SvROK (sv))
696 encode_rv (enc, SvRV (sv)); 887 encode_rv (enc, SvRV (sv));
697 else if (!SvOK (sv)) 888 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
698 encode_str (enc, "null", 4, 0); 889 encode_str (enc, "null", 4, 0);
699 else 890 else
700 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 891 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
701 SvPV_nolen (sv), SvFLAGS (sv)); 892 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
702} 893}
703 894
704static SV * 895static SV *
705encode_json (SV *scalar, JSON *json) 896encode_json (SV *scalar, JSON *json)
706{ 897{
712 enc.json = *json; 903 enc.json = *json;
713 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 904 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
714 enc.cur = SvPVX (enc.sv); 905 enc.cur = SvPVX (enc.sv);
715 enc.end = SvEND (enc.sv); 906 enc.end = SvEND (enc.sv);
716 enc.indent = 0; 907 enc.indent = 0;
717 enc.maxdepth = DEC_DEPTH (enc.json.flags);
718 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 908 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
719 : enc.json.flags & F_LATIN1 ? 0x000100UL 909 : enc.json.flags & F_LATIN1 ? 0x000100UL
720 : 0x110000UL; 910 : 0x110000UL;
721 911
722 SvPOK_only (enc.sv); 912 SvPOK_only (enc.sv);
723 encode_sv (&enc, scalar); 913 encode_sv (&enc, scalar);
914 encode_nl (&enc);
724 915
725 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 916 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
726 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 917 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
727 918
728 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 919 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
789 if (*dec->cur != ch) \ 980 if (*dec->cur != ch) \
790 ERR (# ch " expected"); \ 981 ERR (# ch " expected"); \
791 ++dec->cur; \ 982 ++dec->cur; \
792 SE 983 SE
793 984
794#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 985#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
795#define DEC_DEC_DEPTH --dec->depth 986#define DEC_DEC_DEPTH --dec->depth
796 987
797static SV *decode_sv (dec_t *dec); 988static SV *decode_sv (dec_t *dec);
798 989
799static signed char decode_hexdigit[256]; 990static signed char decode_hexdigit[256];
908 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1099 else if (expect_true (ch >= 0x20 && ch < 0x80))
909 *cur++ = ch; 1100 *cur++ = ch;
910 else if (ch >= 0x80) 1101 else if (ch >= 0x80)
911 { 1102 {
912 STRLEN clen; 1103 STRLEN clen;
913 UV uch;
914 1104
915 --dec_cur; 1105 --dec_cur;
916 1106
917 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1107 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
918 if (clen == (STRLEN)-1) 1108 if (clen == (STRLEN)-1)
919 ERR ("malformed UTF-8 character in JSON string"); 1109 ERR ("malformed UTF-8 character in JSON string");
920 1110
921 do 1111 do
922 *cur++ = *dec_cur++; 1112 *cur++ = *dec_cur++;
939 { 1129 {
940 STRLEN len = cur - buf; 1130 STRLEN len = cur - buf;
941 1131
942 if (sv) 1132 if (sv)
943 { 1133 {
944 SvGROW (sv, SvCUR (sv) + len + 1); 1134 STRLEN cur = SvCUR (sv);
1135
1136 if (SvLEN (sv) <= cur + len)
1137 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1138
945 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1139 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
946 SvCUR_set (sv, SvCUR (sv) + len); 1140 SvCUR_set (sv, SvCUR (sv) + len);
947 } 1141 }
948 else 1142 else
949 sv = newSVpvn (buf, len); 1143 sv = newSVpvn (buf, len);
1040 1234
1041 // special case the rather common 1..5-digit-int case 1235 // special case the rather common 1..5-digit-int case
1042 if (*start == '-') 1236 if (*start == '-')
1043 switch (len) 1237 switch (len)
1044 { 1238 {
1045 case 2: return newSViv (-( start [1] - '0' * 1)); 1239 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1046 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1240 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1047 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1241 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1048 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1242 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1049 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1243 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1050 } 1244 }
1051 else 1245 else
1052 switch (len) 1246 switch (len)
1053 { 1247 {
1054 case 1: return newSViv ( start [0] - '0' * 1); 1248 case 1: return newSViv ( start [0] - '0' * 1);
1055 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1249 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1056 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1250 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1057 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1251 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1058 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1252 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1059 } 1253 }
1060 1254
1061 { 1255 {
1062 UV uv; 1256 UV uv;
1063 int numtype = grok_number (start, len, &uv); 1257 int numtype = grok_number (start, len, &uv);
1072 } 1266 }
1073 1267
1074 len -= *start == '-' ? 1 : 0; 1268 len -= *start == '-' ? 1 : 0;
1075 1269
1076 // does not fit into IV or UV, try NV 1270 // does not fit into IV or UV, try NV
1077 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1271 if (len <= NV_DIG)
1078 #if defined (LDBL_DIG)
1079 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1080 #endif
1081 )
1082 // fits into NV without loss of precision 1272 // fits into NV without loss of precision
1083 return newSVnv (Atof (start)); 1273 return newSVnv (json_atof (start));
1084 1274
1085 // everything else fails, convert it to a string 1275 // everything else fails, convert it to a string
1086 return newSVpvn (start, dec->cur - start); 1276 return newSVpvn (start, dec->cur - start);
1087 } 1277 }
1088 1278
1089 // loss of precision here 1279 // loss of precision here
1090 return newSVnv (Atof (start)); 1280 return newSVnv (json_atof (start));
1091 1281
1092fail: 1282fail:
1093 return 0; 1283 return 0;
1094} 1284}
1095 1285
1119 if (*dec->cur == ']') 1309 if (*dec->cur == ']')
1120 { 1310 {
1121 ++dec->cur; 1311 ++dec->cur;
1122 break; 1312 break;
1123 } 1313 }
1124 1314
1125 if (*dec->cur != ',') 1315 if (*dec->cur != ',')
1126 ERR (", or ] expected while parsing array"); 1316 ERR (", or ] expected while parsing array");
1127 1317
1128 ++dec->cur; 1318 ++dec->cur;
1129 1319
1171 char *p = dec->cur; 1361 char *p = dec->cur;
1172 char *e = p + 24; // only try up to 24 bytes 1362 char *e = p + 24; // only try up to 24 bytes
1173 1363
1174 for (;;) 1364 for (;;)
1175 { 1365 {
1176 // the >= 0x80 is true on most architectures 1366 // the >= 0x80 is false on most architectures
1177 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1367 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1178 { 1368 {
1179 // slow path, back up and use decode_str 1369 // slow path, back up and use decode_str
1180 SV *key = decode_str (dec); 1370 SV *key = decode_str (dec);
1181 if (!key) 1371 if (!key)
1264 dSP; 1454 dSP;
1265 int count; 1455 int count;
1266 1456
1267 ENTER; SAVETMPS; PUSHMARK (SP); 1457 ENTER; SAVETMPS; PUSHMARK (SP);
1268 XPUSHs (HeVAL (he)); 1458 XPUSHs (HeVAL (he));
1459 sv_2mortal (sv);
1269 1460
1270 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1461 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1271 1462
1272 if (count == 1) 1463 if (count == 1)
1273 { 1464 {
1274 sv = newSVsv (POPs); 1465 sv = newSVsv (POPs);
1275 FREETMPS; LEAVE; 1466 FREETMPS; LEAVE;
1276 return sv; 1467 return sv;
1277 } 1468 }
1278 1469
1470 SvREFCNT_inc (sv);
1279 FREETMPS; LEAVE; 1471 FREETMPS; LEAVE;
1280 } 1472 }
1281 } 1473 }
1282 1474
1283 if (dec->json.cb_object) 1475 if (dec->json.cb_object)
1309 DEC_DEC_DEPTH; 1501 DEC_DEC_DEPTH;
1310 return 0; 1502 return 0;
1311} 1503}
1312 1504
1313static SV * 1505static SV *
1506decode_tag (dec_t *dec)
1507{
1508 SV *tag = 0;
1509 SV *val = 0;
1510
1511 if (!(dec->json.flags & F_ALLOW_TAGS))
1512 ERR ("malformed JSON string, neither array, object, number, string or atom");
1513
1514 ++dec->cur;
1515
1516 tag = decode_sv (dec);
1517 if (!tag)
1518 goto fail;
1519
1520 if (!SvPOK (tag))
1521 ERR ("malformed JSON string, (tag) must be a string");
1522
1523 if (*dec->cur != ')')
1524 ERR (") expected after tag");
1525
1526 ++dec->cur;
1527
1528 val = decode_sv (dec);
1529 if (!val)
1530 goto fail;
1531
1532 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1533 ERR ("malformed JSON string, tag value must be an array");
1534
1535 {
1536 AV *av = (AV *)SvRV (val);
1537 int i, len = av_len (av) + 1;
1538 HV *stash = gv_stashsv (tag, 0);
1539 SV *sv;
1540
1541 if (!stash)
1542 ERR ("cannot decode perl-object (package does not exist)");
1543
1544 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1545
1546 if (!method)
1547 ERR ("cannot decode perl-object (package does not have a THAW method)");
1548
1549 dSP;
1550
1551 ENTER; SAVETMPS; PUSHMARK (SP);
1552 EXTEND (SP, len + 2);
1553 // we re-bless the reference to get overload and other niceties right
1554 PUSHs (tag);
1555 PUSHs (sv_json);
1556
1557 for (i = 0; i < len; ++i)
1558 PUSHs (*av_fetch (av, i, 1));
1559
1560 PUTBACK;
1561 call_sv ((SV *)GvCV (method), G_SCALAR);
1562 SPAGAIN;
1563
1564 SvREFCNT_dec (tag);
1565 SvREFCNT_dec (val);
1566 sv = SvREFCNT_inc (POPs);
1567
1568 PUTBACK;
1569
1570 FREETMPS; LEAVE;
1571
1572 return sv;
1573 }
1574
1575fail:
1576 SvREFCNT_dec (tag);
1577 SvREFCNT_dec (val);
1578 return 0;
1579}
1580
1581static SV *
1314decode_sv (dec_t *dec) 1582decode_sv (dec_t *dec)
1315{ 1583{
1316 // the beauty of JSON: you need exactly one character lookahead 1584 // the beauty of JSON: you need exactly one character lookahead
1317 // to parse everything. 1585 // to parse everything.
1318 switch (*dec->cur) 1586 switch (*dec->cur)
1319 { 1587 {
1320 case '"': ++dec->cur; return decode_str (dec); 1588 case '"': ++dec->cur; return decode_str (dec);
1321 case '[': ++dec->cur; return decode_av (dec); 1589 case '[': ++dec->cur; return decode_av (dec);
1322 case '{': ++dec->cur; return decode_hv (dec); 1590 case '{': ++dec->cur; return decode_hv (dec);
1591 case '(': return decode_tag (dec);
1323 1592
1324 case '-': 1593 case '-':
1325 case '0': case '1': case '2': case '3': case '4': 1594 case '0': case '1': case '2': case '3': case '4':
1326 case '5': case '6': case '7': case '8': case '9': 1595 case '5': case '6': case '7': case '8': case '9':
1327 return decode_num (dec); 1596 return decode_num (dec);
1329 case 't': 1598 case 't':
1330 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1599 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1331 { 1600 {
1332 dec->cur += 4; 1601 dec->cur += 4;
1333#if JSON_SLOW 1602#if JSON_SLOW
1334 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1603 types_true = get_bool ("Types::Serialiser::true");
1335#endif 1604#endif
1336 return SvREFCNT_inc (json_true); 1605 return newSVsv (types_true);
1337 } 1606 }
1338 else 1607 else
1339 ERR ("'true' expected"); 1608 ERR ("'true' expected");
1340 1609
1341 break; 1610 break;
1343 case 'f': 1612 case 'f':
1344 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1613 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1345 { 1614 {
1346 dec->cur += 5; 1615 dec->cur += 5;
1347#if JSON_SLOW 1616#if JSON_SLOW
1348 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1617 types_false = get_bool ("Types::Serialiser::false");
1349#endif 1618#endif
1350 return SvREFCNT_inc (json_false); 1619 return newSVsv (types_false);
1351 } 1620 }
1352 else 1621 else
1353 ERR ("'false' expected"); 1622 ERR ("'false' expected");
1354 1623
1355 break; 1624 break;
1364 ERR ("'null' expected"); 1633 ERR ("'null' expected");
1365 1634
1366 break; 1635 break;
1367 1636
1368 default: 1637 default:
1369 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1638 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1370 break; 1639 break;
1371 } 1640 }
1372 1641
1373fail: 1642fail:
1374 return 0; 1643 return 0;
1375} 1644}
1376 1645
1377static SV * 1646static SV *
1378decode_json (SV *string, JSON *json, UV *offset_return) 1647decode_json (SV *string, JSON *json, char **offset_return)
1379{ 1648{
1380 dec_t dec; 1649 dec_t dec;
1381 UV offset;
1382 SV *sv; 1650 SV *sv;
1383 1651
1652 /* work around bugs in 5.10 where manipulating magic values
1653 * makes perl ignore the magic in subsequent accesses.
1654 * also make a copy of non-PV values, to get them into a clean
1655 * state (SvPV should do that, but it's buggy, see below).
1656 */
1384 SvGETMAGIC (string); 1657 /*SvGETMAGIC (string);*/
1658 if (SvMAGICAL (string) || !SvPOK (string))
1659 string = sv_2mortal (newSVsv (string));
1660
1385 SvUPGRADE (string, SVt_PV); 1661 SvUPGRADE (string, SVt_PV);
1386 1662
1387 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1663 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1664 * assertion with -DDEBUGGING, although SvCUR is documented to
1665 * return the xpv_cur field which certainly exists after upgrading.
1666 * according to nicholas clark, calling SvPOK fixes this.
1667 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1668 * and hope for the best.
1669 * Damnit, SvPV_nolen still trips over yet another assertion. This
1670 * assertion business is seriously broken, try yet another workaround
1671 * for the broken -DDEBUGGING.
1672 */
1673 {
1674#ifdef DEBUGGING
1675 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1676#else
1677 STRLEN offset = SvCUR (string);
1678#endif
1679
1680 if (offset > json->max_size && json->max_size)
1388 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1681 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1389 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1682 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1683 }
1390 1684
1391 if (json->flags & F_UTF8) 1685 if (DECODE_WANTS_OCTETS (json))
1392 sv_utf8_downgrade (string, 0); 1686 sv_utf8_downgrade (string, 0);
1393 else 1687 else
1394 sv_utf8_upgrade (string); 1688 sv_utf8_upgrade (string);
1395 1689
1396 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1690 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1397 1691
1398 dec.json = *json; 1692 dec.json = *json;
1399 dec.cur = SvPVX (string); 1693 dec.cur = SvPVX (string);
1400 dec.end = SvEND (string); 1694 dec.end = SvEND (string);
1401 dec.err = 0; 1695 dec.err = 0;
1402 dec.depth = 0; 1696 dec.depth = 0;
1403 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1404 1697
1405 if (dec.json.cb_object || dec.json.cb_sk_object) 1698 if (dec.json.cb_object || dec.json.cb_sk_object)
1406 dec.json.flags |= F_HOOK; 1699 dec.json.flags |= F_HOOK;
1407 1700
1408 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1701 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1409 1702
1410 decode_ws (&dec); 1703 decode_ws (&dec);
1411 sv = decode_sv (&dec); 1704 sv = decode_sv (&dec);
1705
1706 if (offset_return)
1707 *offset_return = dec.cur;
1412 1708
1413 if (!(offset_return || !sv)) 1709 if (!(offset_return || !sv))
1414 { 1710 {
1415 // check for trailing garbage 1711 // check for trailing garbage
1416 decode_ws (&dec); 1712 decode_ws (&dec);
1419 { 1715 {
1420 dec.err = "garbage after JSON object"; 1716 dec.err = "garbage after JSON object";
1421 SvREFCNT_dec (sv); 1717 SvREFCNT_dec (sv);
1422 sv = 0; 1718 sv = 0;
1423 } 1719 }
1424 }
1425
1426 if (offset_return || !sv)
1427 {
1428 offset = dec.json.flags & F_UTF8
1429 ? dec.cur - SvPVX (string)
1430 : utf8_distance (dec.cur, SvPVX (string));
1431
1432 if (offset_return)
1433 *offset_return = offset;
1434 } 1720 }
1435 1721
1436 if (!sv) 1722 if (!sv)
1437 { 1723 {
1438 SV *uni = sv_newmortal (); 1724 SV *uni = sv_newmortal ();
1444 SAVEVPTR (PL_curcop); 1730 SAVEVPTR (PL_curcop);
1445 PL_curcop = &cop; 1731 PL_curcop = &cop;
1446 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1732 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1447 LEAVE; 1733 LEAVE;
1448 1734
1449 croak ("%s, at character offset %d [\"%s\"]", 1735 croak ("%s, at character offset %d (before \"%s\")",
1450 dec.err, 1736 dec.err,
1451 (int)offset, 1737 (int)ptr_to_index (string, dec.cur),
1452 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1738 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1453 } 1739 }
1454 1740
1455 sv = sv_2mortal (sv); 1741 sv = sv_2mortal (sv);
1456 1742
1457 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1743 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1458 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1744 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1459 1745
1460 return sv; 1746 return sv;
1747}
1748
1749/////////////////////////////////////////////////////////////////////////////
1750// incremental parser
1751
1752static void
1753incr_parse (JSON *self)
1754{
1755 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1756
1757 // the state machine here is a bit convoluted and could be simplified a lot
1758 // but this would make it slower, so...
1759
1760 for (;;)
1761 {
1762 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1763 switch (self->incr_mode)
1764 {
1765 // only used for initial whitespace skipping
1766 case INCR_M_WS:
1767 for (;;)
1768 {
1769 if (*p > 0x20)
1770 {
1771 if (*p == '#')
1772 {
1773 self->incr_mode = INCR_M_C0;
1774 goto incr_m_c;
1775 }
1776 else
1777 {
1778 self->incr_mode = INCR_M_JSON;
1779 goto incr_m_json;
1780 }
1781 }
1782 else if (!*p)
1783 goto interrupt;
1784
1785 ++p;
1786 }
1787
1788 // skip a single char inside a string (for \\-processing)
1789 case INCR_M_BS:
1790 if (!*p)
1791 goto interrupt;
1792
1793 ++p;
1794 self->incr_mode = INCR_M_STR;
1795 goto incr_m_str;
1796
1797 // inside #-style comments
1798 case INCR_M_C0:
1799 case INCR_M_C1:
1800 incr_m_c:
1801 for (;;)
1802 {
1803 if (*p == '\n')
1804 {
1805 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1806 break;
1807 }
1808 else if (!*p)
1809 goto interrupt;
1810
1811 ++p;
1812 }
1813
1814 break;
1815
1816 // inside a string
1817 case INCR_M_STR:
1818 incr_m_str:
1819 for (;;)
1820 {
1821 if (*p == '"')
1822 {
1823 ++p;
1824 self->incr_mode = INCR_M_JSON;
1825
1826 if (!self->incr_nest)
1827 goto interrupt;
1828
1829 goto incr_m_json;
1830 }
1831 else if (*p == '\\')
1832 {
1833 ++p; // "virtually" consumes character after \
1834
1835 if (!*p) // if at end of string we have to switch modes
1836 {
1837 self->incr_mode = INCR_M_BS;
1838 goto interrupt;
1839 }
1840 }
1841 else if (!*p)
1842 goto interrupt;
1843
1844 ++p;
1845 }
1846
1847 // after initial ws, outside string
1848 case INCR_M_JSON:
1849 incr_m_json:
1850 for (;;)
1851 {
1852 switch (*p++)
1853 {
1854 case 0:
1855 --p;
1856 goto interrupt;
1857
1858 case 0x09:
1859 case 0x0a:
1860 case 0x0d:
1861 case 0x20:
1862 if (!self->incr_nest)
1863 {
1864 --p; // do not eat the whitespace, let the next round do it
1865 goto interrupt;
1866 }
1867 break;
1868
1869 case '"':
1870 self->incr_mode = INCR_M_STR;
1871 goto incr_m_str;
1872
1873 case '[':
1874 case '{':
1875 case '(':
1876 if (++self->incr_nest > self->max_depth)
1877 croak (ERR_NESTING_EXCEEDED);
1878 break;
1879
1880 case ']':
1881 case '}':
1882 if (--self->incr_nest <= 0)
1883 goto interrupt;
1884 break;
1885
1886 case ')':
1887 --self->incr_nest;
1888 break;
1889
1890 case '#':
1891 self->incr_mode = INCR_M_C1;
1892 goto incr_m_c;
1893 }
1894 }
1895 }
1896
1897 modechange:
1898 ;
1899 }
1900
1901interrupt:
1902 self->incr_pos = p - SvPVX (self->incr_text);
1903 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1904 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1461} 1905}
1462 1906
1463///////////////////////////////////////////////////////////////////////////// 1907/////////////////////////////////////////////////////////////////////////////
1464// XS interface functions 1908// XS interface functions
1465 1909
1474 i >= '0' && i <= '9' ? i - '0' 1918 i >= '0' && i <= '9' ? i - '0'
1475 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1919 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1476 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1920 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1477 : -1; 1921 : -1;
1478 1922
1479 json_stash = gv_stashpv ("JSON::XS" , 1); 1923 json_stash = gv_stashpv ("JSON::XS" , 1);
1480 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1924 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1481 1925
1482 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1926 types_true = get_bool ("Types::Serialiser::true");
1483 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1927 types_false = get_bool ("Types::Serialiser::false");
1928
1929 sv_json = newSVpv ("JSON", 0);
1930 SvREADONLY_on (sv_json);
1931
1932 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1484} 1933}
1485 1934
1486PROTOTYPES: DISABLE 1935PROTOTYPES: DISABLE
1487 1936
1488void CLONE (...) 1937void CLONE (...)
1489 CODE: 1938 CODE:
1490 json_stash = 0; 1939 json_stash = 0;
1491 json_boolean_stash = 0; 1940 types_boolean_stash = 0;
1492 1941
1493void new (char *klass) 1942void new (char *klass)
1494 PPCODE: 1943 PPCODE:
1495{ 1944{
1496 SV *pv = NEWSV (0, sizeof (JSON)); 1945 SV *pv = NEWSV (0, sizeof (JSON));
1497 SvPOK_only (pv); 1946 SvPOK_only (pv);
1498 Zero (SvPVX (pv), 1, JSON); 1947 json_init ((JSON *)SvPVX (pv));
1499 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1500 XPUSHs (sv_2mortal (sv_bless ( 1948 XPUSHs (sv_2mortal (sv_bless (
1501 newRV_noinc (pv), 1949 newRV_noinc (pv),
1502 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1950 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1503 ))); 1951 )));
1504} 1952}
1516 allow_nonref = F_ALLOW_NONREF 1964 allow_nonref = F_ALLOW_NONREF
1517 shrink = F_SHRINK 1965 shrink = F_SHRINK
1518 allow_blessed = F_ALLOW_BLESSED 1966 allow_blessed = F_ALLOW_BLESSED
1519 convert_blessed = F_CONV_BLESSED 1967 convert_blessed = F_CONV_BLESSED
1520 relaxed = F_RELAXED 1968 relaxed = F_RELAXED
1969 allow_unknown = F_ALLOW_UNKNOWN
1970 allow_tags = F_ALLOW_TAGS
1521 PPCODE: 1971 PPCODE:
1522{ 1972{
1523 if (enable) 1973 if (enable)
1524 self->flags |= ix; 1974 self->flags |= ix;
1525 else 1975 else
1540 get_allow_nonref = F_ALLOW_NONREF 1990 get_allow_nonref = F_ALLOW_NONREF
1541 get_shrink = F_SHRINK 1991 get_shrink = F_SHRINK
1542 get_allow_blessed = F_ALLOW_BLESSED 1992 get_allow_blessed = F_ALLOW_BLESSED
1543 get_convert_blessed = F_CONV_BLESSED 1993 get_convert_blessed = F_CONV_BLESSED
1544 get_relaxed = F_RELAXED 1994 get_relaxed = F_RELAXED
1995 get_allow_unknown = F_ALLOW_UNKNOWN
1996 get_allow_tags = F_ALLOW_TAGS
1545 PPCODE: 1997 PPCODE:
1546 XPUSHs (boolSV (self->flags & ix)); 1998 XPUSHs (boolSV (self->flags & ix));
1547 1999
1548void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2000void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1549 PPCODE: 2001 PPCODE:
1550{ 2002 self->max_depth = max_depth;
1551 UV log2 = 0;
1552
1553 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1554
1555 while ((1UL << log2) < max_depth)
1556 ++log2;
1557
1558 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1559
1560 XPUSHs (ST (0)); 2003 XPUSHs (ST (0));
1561}
1562 2004
1563U32 get_max_depth (JSON *self) 2005U32 get_max_depth (JSON *self)
1564 CODE: 2006 CODE:
1565 RETVAL = DEC_DEPTH (self->flags); 2007 RETVAL = self->max_depth;
1566 OUTPUT: 2008 OUTPUT:
1567 RETVAL 2009 RETVAL
1568 2010
1569void max_size (JSON *self, UV max_size = 0) 2011void max_size (JSON *self, U32 max_size = 0)
1570 PPCODE: 2012 PPCODE:
1571{ 2013 self->max_size = max_size;
1572 UV log2 = 0;
1573
1574 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1575 if (max_size == 1) max_size = 2;
1576
1577 while ((1UL << log2) < max_size)
1578 ++log2;
1579
1580 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1581
1582 XPUSHs (ST (0)); 2014 XPUSHs (ST (0));
1583}
1584 2015
1585int get_max_size (JSON *self) 2016int get_max_size (JSON *self)
1586 CODE: 2017 CODE:
1587 RETVAL = DEC_SIZE (self->flags); 2018 RETVAL = self->max_size;
1588 OUTPUT: 2019 OUTPUT:
1589 RETVAL 2020 RETVAL
1590 2021
1591void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2022void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1592 PPCODE: 2023 PPCODE:
1598} 2029}
1599 2030
1600void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2031void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1601 PPCODE: 2032 PPCODE:
1602{ 2033{
1603 if (!self->cb_sk_object) 2034 if (!self->cb_sk_object)
1604 self->cb_sk_object = newHV (); 2035 self->cb_sk_object = newHV ();
1605 2036
1606 if (SvOK (cb)) 2037 if (SvOK (cb))
1607 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2038 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1608 else 2039 else
1619 XPUSHs (ST (0)); 2050 XPUSHs (ST (0));
1620} 2051}
1621 2052
1622void encode (JSON *self, SV *scalar) 2053void encode (JSON *self, SV *scalar)
1623 PPCODE: 2054 PPCODE:
1624 XPUSHs (encode_json (scalar, self)); 2055 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2056 XPUSHs (scalar);
1625 2057
1626void decode (JSON *self, SV *jsonstr) 2058void decode (JSON *self, SV *jsonstr)
1627 PPCODE: 2059 PPCODE:
1628 XPUSHs (decode_json (jsonstr, self, 0)); 2060 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2061 XPUSHs (jsonstr);
1629 2062
1630void decode_prefix (JSON *self, SV *jsonstr) 2063void decode_prefix (JSON *self, SV *jsonstr)
1631 PPCODE: 2064 PPCODE:
1632{ 2065{
1633 UV offset; 2066 SV *sv;
2067 char *offset;
2068 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1634 EXTEND (SP, 2); 2069 EXTEND (SP, 2);
1635 PUSHs (decode_json (jsonstr, self, &offset)); 2070 PUSHs (sv);
1636 PUSHs (sv_2mortal (newSVuv (offset))); 2071 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
2072}
2073
2074void incr_parse (JSON *self, SV *jsonstr = 0)
2075 PPCODE:
2076{
2077 if (!self->incr_text)
2078 self->incr_text = newSVpvn ("", 0);
2079
2080 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2081 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2082 if (DECODE_WANTS_OCTETS (self))
2083 {
2084 if (self->incr_pos)
2085 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2086 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2087
2088 sv_utf8_downgrade (self->incr_text, 0);
2089 }
2090 else
2091 {
2092 sv_utf8_upgrade (self->incr_text);
2093
2094 if (self->incr_pos)
2095 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2096 - (U8 *)SvPVX (self->incr_text);
2097 }
2098
2099 // append data, if any
2100 if (jsonstr)
2101 {
2102 /* make sure both strings have same encoding */
2103 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2104 if (SvUTF8 (jsonstr))
2105 sv_utf8_downgrade (jsonstr, 0);
2106 else
2107 sv_utf8_upgrade (jsonstr);
2108
2109 /* and then just blindly append */
2110 {
2111 STRLEN len;
2112 const char *str = SvPV (jsonstr, len);
2113 STRLEN cur = SvCUR (self->incr_text);
2114
2115 if (SvLEN (self->incr_text) <= cur + len)
2116 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2117
2118 Move (str, SvEND (self->incr_text), len, char);
2119 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2120 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2121 }
2122 }
2123
2124 if (GIMME_V != G_VOID)
2125 do
2126 {
2127 SV *sv;
2128 char *offset;
2129
2130 if (!INCR_DONE (self))
2131 {
2132 incr_parse (self);
2133
2134 if (self->incr_pos > self->max_size && self->max_size)
2135 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2136 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2137
2138 if (!INCR_DONE (self))
2139 {
2140 // as an optimisation, do not accumulate white space in the incr buffer
2141 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2142 {
2143 self->incr_pos = 0;
2144 SvCUR_set (self->incr_text, 0);
2145 }
2146
2147 break;
2148 }
2149 }
2150
2151 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2152 XPUSHs (sv);
2153
2154 self->incr_pos -= offset - SvPVX (self->incr_text);
2155 self->incr_nest = 0;
2156 self->incr_mode = 0;
2157
2158 sv_chop (self->incr_text, offset);
2159 }
2160 while (GIMME_V == G_ARRAY);
2161}
2162
2163SV *incr_text (JSON *self)
2164 ATTRS: lvalue
2165 CODE:
2166{
2167 if (self->incr_pos)
2168 croak ("incr_text can not be called when the incremental parser already started parsing");
2169
2170 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2171}
2172 OUTPUT:
2173 RETVAL
2174
2175void incr_skip (JSON *self)
2176 CODE:
2177{
2178 if (self->incr_pos)
2179 {
2180 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2181 self->incr_pos = 0;
2182 self->incr_nest = 0;
2183 self->incr_mode = 0;
2184 }
2185}
2186
2187void incr_reset (JSON *self)
2188 CODE:
2189{
2190 SvREFCNT_dec (self->incr_text);
2191 self->incr_text = 0;
2192 self->incr_pos = 0;
2193 self->incr_nest = 0;
2194 self->incr_mode = 0;
1637} 2195}
1638 2196
1639void DESTROY (JSON *self) 2197void DESTROY (JSON *self)
1640 CODE: 2198 CODE:
1641 SvREFCNT_dec (self->cb_sk_object); 2199 SvREFCNT_dec (self->cb_sk_object);
1642 SvREFCNT_dec (self->cb_object); 2200 SvREFCNT_dec (self->cb_object);
2201 SvREFCNT_dec (self->incr_text);
1643 2202
1644PROTOTYPES: ENABLE 2203PROTOTYPES: ENABLE
1645 2204
1646void encode_json (SV *scalar) 2205void encode_json (SV *scalar)
1647 PPCODE: 2206 PPCODE:
1648{ 2207{
1649 JSON json = { F_DEFAULT | F_UTF8 }; 2208 JSON json;
1650 XPUSHs (encode_json (scalar, &json)); 2209 json_init (&json);
2210 json.flags |= F_UTF8;
2211 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2212 XPUSHs (scalar);
1651} 2213}
1652 2214
1653void decode_json (SV *jsonstr) 2215void decode_json (SV *jsonstr)
1654 PPCODE: 2216 PPCODE:
1655{ 2217{
1656 JSON json = { F_DEFAULT | F_UTF8 }; 2218 JSON json;
1657 XPUSHs (decode_json (jsonstr, &json, 0)); 2219 json_init (&json);
2220 json.flags |= F_UTF8;
2221 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2222 XPUSHs (jsonstr);
1658} 2223}
1659 2224

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines