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 1.74 by root, Wed Mar 19 15:17:53 2008 UTC vs.
Revision 1.121 by root, Tue Oct 29 00:06:40 2013 UTC

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#include <stdio.h>
9#include <limits.h>
9#include <float.h> 10#include <float.h>
10 11
11#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
12# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
13#endif 14#endif
14 15
15// 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
16// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
17#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
18# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
19#endif 20#endif
21
22// three extra for rounding, sign, and end of string
23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20 24
21#define F_ASCII 0x00000001UL 25#define F_ASCII 0x00000001UL
22#define F_LATIN1 0x00000002UL 26#define F_LATIN1 0x00000002UL
23#define F_UTF8 0x00000004UL 27#define F_UTF8 0x00000004UL
24#define F_INDENT 0x00000008UL 28#define F_INDENT 0x00000008UL
28#define F_ALLOW_NONREF 0x00000100UL 32#define F_ALLOW_NONREF 0x00000100UL
29#define F_SHRINK 0x00000200UL 33#define F_SHRINK 0x00000200UL
30#define F_ALLOW_BLESSED 0x00000400UL 34#define F_ALLOW_BLESSED 0x00000400UL
31#define F_CONV_BLESSED 0x00000800UL 35#define F_CONV_BLESSED 0x00000800UL
32#define F_RELAXED 0x00001000UL 36#define F_RELAXED 0x00001000UL
33 37#define F_ALLOW_UNKNOWN 0x00002000UL
34#define F_MAXDEPTH 0xf8000000UL 38#define F_ALLOW_TAGS 0x00004000UL
35#define S_MAXDEPTH 27
36#define F_MAXSIZE 0x01f00000UL
37#define S_MAXSIZE 20
38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 39#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39 40
40#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 41#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
44#define F_DEFAULT (9UL << S_MAXDEPTH)
45 42
46#define INIT_SIZE 32 // initial scalar size to be allocated 43#define INIT_SIZE 32 // initial scalar size to be allocated
47#define INDENT_STEP 3 // spaces per indentation level 44#define INDENT_STEP 3 // spaces per indentation level
48 45
49#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)
50 49
51#define SB do { 50#define SB do {
52#define SE } while (0) 51#define SE } while (0)
53 52
54#if __GNUC__ >= 3 53#if __GNUC__ >= 3
64 63
65#define IN_RANGE_INC(type,val,beg,end) \ 64#define IN_RANGE_INC(type,val,beg,end) \
66 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 65 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
67 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 66 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
68 67
68#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
69
69#ifdef USE_ITHREADS 70#ifdef USE_ITHREADS
70# define JSON_SLOW 1 71# define JSON_SLOW 1
71# 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))
72#else 73#else
73# define JSON_SLOW 0 74# define JSON_SLOW 0
74# define JSON_STASH json_stash 75# define JSON_STASH json_stash
75#endif 76#endif
76 77
78// the amount of HEs to allocate on the stack, when sorting keys
79#define STACK_HES 64
80
77static HV *json_stash, *json_boolean_stash; // JSON::XS:: 81static HV *json_stash, *types_boolean_stash; // JSON::XS::
78static 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)
79 94
80typedef struct { 95typedef struct {
81 U32 flags; 96 U32 flags;
97 U32 max_depth;
98 STRLEN max_size;
99
82 SV *cb_object; 100 SV *cb_object;
83 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;
84} JSON; 108} JSON;
109
110INLINE void
111json_init (JSON *json)
112{
113 Zero (json, 1, JSON);
114 json->max_depth = 512;
115}
85 116
86///////////////////////////////////////////////////////////////////////////// 117/////////////////////////////////////////////////////////////////////////////
87// utility functions 118// utility functions
88 119
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
89INLINE void 131INLINE void
90shrink (SV *sv) 132shrink (SV *sv)
91{ 133{
92 sv_utf8_downgrade (sv, 1); 134 sv_utf8_downgrade (sv, 1);
135
93 if (SvLEN (sv) > SvCUR (sv) + 1) 136 if (SvLEN (sv) > SvCUR (sv) + 1)
94 { 137 {
95#ifdef SvPV_shrink_to_cur 138#ifdef SvPV_shrink_to_cur
96 SvPV_shrink_to_cur (sv); 139 SvPV_shrink_to_cur (sv);
97#elif defined (SvPV_renew) 140#elif defined (SvPV_renew)
141 *s++ = 0x80 | ( ch & 0x3f); 184 *s++ = 0x80 | ( ch & 0x3f);
142 185
143 return s; 186 return s;
144} 187}
145 188
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}
146///////////////////////////////////////////////////////////////////////////// 293/////////////////////////////////////////////////////////////////////////////
147// encoder 294// encoder
148 295
149// structure used for encoding JSON 296// structure used for encoding JSON
150typedef struct 297typedef struct
152 char *cur; // SvPVX (sv) + current output position 299 char *cur; // SvPVX (sv) + current output position
153 char *end; // SvEND (sv) 300 char *end; // SvEND (sv)
154 SV *sv; // result scalar 301 SV *sv; // result scalar
155 JSON json; 302 JSON json;
156 U32 indent; // indentation level 303 U32 indent; // indentation level
157 U32 maxdepth; // max. indentation/recursion level
158 UV limit; // escape character values >= this value when encoding 304 UV limit; // escape character values >= this value when encoding
159} enc_t; 305} enc_t;
160 306
161INLINE void 307INLINE void
162need (enc_t *enc, STRLEN len) 308need (enc_t *enc, STRLEN len)
163{ 309{
164 if (expect_false (enc->cur + len >= enc->end)) 310 if (expect_false (enc->cur + len >= enc->end))
165 { 311 {
166 STRLEN cur = enc->cur - SvPVX (enc->sv); 312 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
167 SvGROW (enc->sv, cur + len + 1); 313 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
168 enc->cur = SvPVX (enc->sv) + cur; 314 enc->cur = SvPVX (enc->sv) + cur;
169 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 315 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
170 } 316 }
171} 317}
172 318
247 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 393 (int)((uch - 0x10000) % 0x400 + 0xDC00));
248 enc->cur += 12; 394 enc->cur += 12;
249 } 395 }
250 else 396 else
251 { 397 {
252 static char hexdigit [16] = "0123456789abcdef";
253 need (enc, len += 5); 398 need (enc, len += 5);
254 *enc->cur++ = '\\'; 399 *enc->cur++ = '\\';
255 *enc->cur++ = 'u'; 400 *enc->cur++ = 'u';
256 *enc->cur++ = hexdigit [ uch >> 12 ]; 401 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
257 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 402 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
258 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 403 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
259 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 404 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
260 } 405 }
261 406
262 str += clen; 407 str += clen;
263 } 408 }
264 else if (enc->json.flags & F_LATIN1) 409 else if (enc->json.flags & F_LATIN1)
335static void 480static void
336encode_av (enc_t *enc, AV *av) 481encode_av (enc_t *enc, AV *av)
337{ 482{
338 int i, len = av_len (av); 483 int i, len = av_len (av);
339 484
340 if (enc->indent >= enc->maxdepth) 485 if (enc->indent >= enc->json.max_depth)
341 croak ("data structure too deep (hit recursion limit)"); 486 croak (ERR_NESTING_EXCEEDED);
342 487
343 encode_ch (enc, '['); 488 encode_ch (enc, '[');
344 489
345 if (len >= 0) 490 if (len >= 0)
346 { 491 {
347 encode_nl (enc); ++enc->indent; 492 encode_nl (enc); ++enc->indent;
348 493
349 for (i = 0; i <= len; ++i) 494 for (i = 0; i <= len; ++i)
361 encode_comma (enc); 506 encode_comma (enc);
362 } 507 }
363 508
364 encode_nl (enc); --enc->indent; encode_indent (enc); 509 encode_nl (enc); --enc->indent; encode_indent (enc);
365 } 510 }
366 511
367 encode_ch (enc, ']'); 512 encode_ch (enc, ']');
368} 513}
369 514
370static void 515static void
371encode_hk (enc_t *enc, HE *he) 516encode_hk (enc_t *enc, HE *he)
375 if (HeKLEN (he) == HEf_SVKEY) 520 if (HeKLEN (he) == HEf_SVKEY)
376 { 521 {
377 SV *sv = HeSVKEY (he); 522 SV *sv = HeSVKEY (he);
378 STRLEN len; 523 STRLEN len;
379 char *str; 524 char *str;
380 525
381 SvGETMAGIC (sv); 526 SvGETMAGIC (sv);
382 str = SvPV (sv, len); 527 str = SvPV (sv, len);
383 528
384 encode_str (enc, str, len, SvUTF8 (sv)); 529 encode_str (enc, str, len, SvUTF8 (sv));
385 } 530 }
420 565
421static void 566static void
422encode_hv (enc_t *enc, HV *hv) 567encode_hv (enc_t *enc, HV *hv)
423{ 568{
424 HE *he; 569 HE *he;
425 int count;
426 570
427 if (enc->indent >= enc->maxdepth) 571 if (enc->indent >= enc->json.max_depth)
428 croak ("data structure too deep (hit recursion limit)"); 572 croak (ERR_NESTING_EXCEEDED);
429 573
430 encode_ch (enc, '{'); 574 encode_ch (enc, '{');
431 575
432 // for canonical output we have to sort by keys first 576 // for canonical output we have to sort by keys first
433 // actually, this is mostly due to the stupid so-called 577 // actually, this is mostly due to the stupid so-called
434 // security workaround added somewhere in 5.8.x. 578 // security workaround added somewhere in 5.8.x
435 // that randomises hash orderings 579 // that randomises hash orderings
436 if (enc->json.flags & F_CANONICAL) 580 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
437 { 581 {
438 int count = hv_iterinit (hv); 582 int count = hv_iterinit (hv);
439 583
440 if (SvMAGICAL (hv)) 584 if (SvMAGICAL (hv))
441 { 585 {
452 } 596 }
453 597
454 if (count) 598 if (count)
455 { 599 {
456 int i, fast = 1; 600 int i, fast = 1;
457#if defined(__BORLANDC__) || defined(_MSC_VER) 601 HE *hes_stack [STACK_HES];
458 HE **hes = _alloca (count * sizeof (HE)); 602 HE **hes = hes_stack;
459#else 603
460 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 604 // allocate larger arrays on the heap
461#endif 605 if (count > STACK_HES)
606 {
607 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
608 hes = (HE **)SvPVX (sv);
609 }
462 610
463 i = 0; 611 i = 0;
464 while ((he = hv_iternext (hv))) 612 while ((he = hv_iternext (hv)))
465 { 613 {
466 hes [i++] = he; 614 hes [i++] = he;
535// encode objects, arrays and special \0=false and \1=true values. 683// encode objects, arrays and special \0=false and \1=true values.
536static void 684static void
537encode_rv (enc_t *enc, SV *sv) 685encode_rv (enc_t *enc, SV *sv)
538{ 686{
539 svtype svt; 687 svtype svt;
688 GV *method;
540 689
541 SvGETMAGIC (sv); 690 SvGETMAGIC (sv);
542 svt = SvTYPE (sv); 691 svt = SvTYPE (sv);
543 692
544 if (expect_false (SvOBJECT (sv))) 693 if (expect_false (SvOBJECT (sv)))
545 { 694 {
546 HV *stash = !JSON_SLOW || json_boolean_stash 695 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
547 ? json_boolean_stash 696 ? types_boolean_stash
548 : gv_stashpv ("JSON::XS::Boolean", 1); 697 : gv_stashpv ("Types::Serialiser::Boolean", 1);
698 HV *stash = SvSTASH (sv);
549 699
550 if (SvSTASH (sv) == stash) 700 if (stash == boolean_stash)
551 { 701 {
552 if (SvIV (sv)) 702 if (SvIV (sv))
553 encode_str (enc, "true", 4, 0); 703 encode_str (enc, "true", 4, 0);
554 else 704 else
555 encode_str (enc, "false", 5, 0); 705 encode_str (enc, "false", 5, 0);
556 } 706 }
707 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
708 {
709 int count;
710 dSP;
711
712 ENTER; SAVETMPS; PUSHMARK (SP);
713 EXTEND (SP, 2);
714 // we re-bless the reference to get overload and other niceties right
715 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
716 PUSHs (sv_json);
717
718 PUTBACK;
719 count = call_sv ((SV *)GvCV (method), G_ARRAY);
720 SPAGAIN;
721
722 // catch this surprisingly common error
723 if (SvROK (TOPs) && SvRV (TOPs) == sv)
724 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
725
726 encode_ch (enc, '(');
727 encode_ch (enc, '"');
728 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
729 encode_ch (enc, '"');
730 encode_ch (enc, ')');
731 encode_ch (enc, '[');
732
733 while (count)
734 {
735 encode_sv (enc, SP[1 - count--]);
736
737 if (count)
738 encode_ch (enc, ',');
739 }
740
741 encode_ch (enc, ']');
742
743 PUTBACK;
744
745 FREETMPS; LEAVE;
746 }
747 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
748 {
749 dSP;
750
751 ENTER; SAVETMPS; PUSHMARK (SP);
752 // we re-bless the reference to get overload and other niceties right
753 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
754
755 // calling with G_SCALAR ensures that we always get a 1 return value
756 PUTBACK;
757 call_sv ((SV *)GvCV (method), G_SCALAR);
758 SPAGAIN;
759
760 // catch this surprisingly common error
761 if (SvROK (TOPs) && SvRV (TOPs) == sv)
762 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
763
764 sv = POPs;
765 PUTBACK;
766
767 encode_sv (enc, sv);
768
769 FREETMPS; LEAVE;
770 }
771 else if (enc->json.flags & F_ALLOW_BLESSED)
772 encode_str (enc, "null", 4, 0);
557 else 773 else
558 {
559#if 0
560 if (0 && sv_derived_from (rv, "JSON::Literal"))
561 {
562 // not yet
563 }
564#endif
565 if (enc->json.flags & F_CONV_BLESSED)
566 {
567 // we re-bless the reference to get overload and other niceties right
568 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
569
570 if (to_json)
571 {
572 dSP;
573
574 ENTER; SAVETMPS; PUSHMARK (SP);
575 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
576
577 // calling with G_SCALAR ensures that we always get a 1 return value
578 PUTBACK;
579 call_sv ((SV *)GvCV (to_json), G_SCALAR);
580 SPAGAIN;
581
582 // catch this surprisingly common error
583 if (SvROK (TOPs) && SvRV (TOPs) == sv)
584 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
585
586 sv = POPs;
587 PUTBACK;
588
589 encode_sv (enc, sv);
590
591 FREETMPS; LEAVE;
592 }
593 else if (enc->json.flags & F_ALLOW_BLESSED)
594 encode_str (enc, "null", 4, 0);
595 else
596 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
597 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
598 }
599 else if (enc->json.flags & F_ALLOW_BLESSED)
600 encode_str (enc, "null", 4, 0);
601 else
602 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 774 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
603 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 775 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
604 }
605 } 776 }
606 else if (svt == SVt_PVHV) 777 else if (svt == SVt_PVHV)
607 encode_hv (enc, (HV *)sv); 778 encode_hv (enc, (HV *)sv);
608 else if (svt == SVt_PVAV) 779 else if (svt == SVt_PVAV)
609 encode_av (enc, (AV *)sv); 780 encode_av (enc, (AV *)sv);
614 785
615 if (len == 1 && *pv == '1') 786 if (len == 1 && *pv == '1')
616 encode_str (enc, "true", 4, 0); 787 encode_str (enc, "true", 4, 0);
617 else if (len == 1 && *pv == '0') 788 else if (len == 1 && *pv == '0')
618 encode_str (enc, "false", 5, 0); 789 encode_str (enc, "false", 5, 0);
790 else if (enc->json.flags & F_ALLOW_UNKNOWN)
791 encode_str (enc, "null", 4, 0);
619 else 792 else
620 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 793 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
621 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 794 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
622 } 795 }
796 else if (enc->json.flags & F_ALLOW_UNKNOWN)
797 encode_str (enc, "null", 4, 0);
623 else 798 else
624 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 799 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
625 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 800 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
626} 801}
627 802
648 else if (SvIOKp (sv)) 823 else if (SvIOKp (sv))
649 { 824 {
650 // we assume we can always read an IV as a UV and vice versa 825 // we assume we can always read an IV as a UV and vice versa
651 // we assume two's complement 826 // we assume two's complement
652 // we assume no aliasing issues in the union 827 // we assume no aliasing issues in the union
653 if (SvIsUV (sv) ? SvUVX (sv) > 59000 828 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
654 : SvIVX (sv) > 59000 || SvIVX (sv) < -59000) 829 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
655 {
656 // large integer, use the (rather slow) snprintf way.
657 need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8
658 enc->cur +=
659 SvIsUV(sv)
660 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
661 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
662 }
663 else
664 { 830 {
665 // optimise the "small number case" 831 // optimise the "small number case"
666 // code will likely be branchless and use only a single multiplication 832 // code will likely be branchless and use only a single multiplication
667 // works for numbers up to 59074 833 // works for numbers up to 59074
668 I32 i = SvIVX (sv); 834 I32 i = SvIVX (sv);
679 845
680 // now output digit by digit, each time masking out the integer part 846 // now output digit by digit, each time masking out the integer part
681 // and multiplying by 5 while moving the decimal point one to the right, 847 // and multiplying by 5 while moving the decimal point one to the right,
682 // resulting in a net multiplication by 10. 848 // resulting in a net multiplication by 10.
683 // we always write the digit to memory but conditionally increment 849 // we always write the digit to memory but conditionally increment
684 // the pointer, to ease the usage of conditional move instructions. 850 // the pointer, to enable the use of conditional move instructions.
685 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; 851 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
686 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; 852 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
687 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; 853 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
688 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; 854 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
689 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' 855 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
690 } 856 }
857 else
858 {
859 // large integer, use the (rather slow) snprintf way.
860 need (enc, IVUV_MAXCHARS);
861 enc->cur +=
862 SvIsUV(sv)
863 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
864 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
865 }
691 } 866 }
692 else if (SvROK (sv)) 867 else if (SvROK (sv))
693 encode_rv (enc, SvRV (sv)); 868 encode_rv (enc, SvRV (sv));
694 else if (!SvOK (sv)) 869 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
695 encode_str (enc, "null", 4, 0); 870 encode_str (enc, "null", 4, 0);
696 else 871 else
697 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 872 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
698 SvPV_nolen (sv), SvFLAGS (sv)); 873 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
699} 874}
700 875
701static SV * 876static SV *
702encode_json (SV *scalar, JSON *json) 877encode_json (SV *scalar, JSON *json)
703{ 878{
709 enc.json = *json; 884 enc.json = *json;
710 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 885 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
711 enc.cur = SvPVX (enc.sv); 886 enc.cur = SvPVX (enc.sv);
712 enc.end = SvEND (enc.sv); 887 enc.end = SvEND (enc.sv);
713 enc.indent = 0; 888 enc.indent = 0;
714 enc.maxdepth = DEC_DEPTH (enc.json.flags);
715 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 889 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
716 : enc.json.flags & F_LATIN1 ? 0x000100UL 890 : enc.json.flags & F_LATIN1 ? 0x000100UL
717 : 0x110000UL; 891 : 0x110000UL;
718 892
719 SvPOK_only (enc.sv); 893 SvPOK_only (enc.sv);
720 encode_sv (&enc, scalar); 894 encode_sv (&enc, scalar);
895 encode_nl (&enc);
721 896
722 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 897 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
723 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 898 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
724 899
725 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 900 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
786 if (*dec->cur != ch) \ 961 if (*dec->cur != ch) \
787 ERR (# ch " expected"); \ 962 ERR (# ch " expected"); \
788 ++dec->cur; \ 963 ++dec->cur; \
789 SE 964 SE
790 965
791#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 966#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
792#define DEC_DEC_DEPTH --dec->depth 967#define DEC_DEC_DEPTH --dec->depth
793 968
794static SV *decode_sv (dec_t *dec); 969static SV *decode_sv (dec_t *dec);
795 970
796static signed char decode_hexdigit[256]; 971static signed char decode_hexdigit[256];
905 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1080 else if (expect_true (ch >= 0x20 && ch < 0x80))
906 *cur++ = ch; 1081 *cur++ = ch;
907 else if (ch >= 0x80) 1082 else if (ch >= 0x80)
908 { 1083 {
909 STRLEN clen; 1084 STRLEN clen;
910 UV uch;
911 1085
912 --dec_cur; 1086 --dec_cur;
913 1087
914 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1088 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
915 if (clen == (STRLEN)-1) 1089 if (clen == (STRLEN)-1)
916 ERR ("malformed UTF-8 character in JSON string"); 1090 ERR ("malformed UTF-8 character in JSON string");
917 1091
918 do 1092 do
919 *cur++ = *dec_cur++; 1093 *cur++ = *dec_cur++;
936 { 1110 {
937 STRLEN len = cur - buf; 1111 STRLEN len = cur - buf;
938 1112
939 if (sv) 1113 if (sv)
940 { 1114 {
941 SvGROW (sv, SvCUR (sv) + len + 1); 1115 STRLEN cur = SvCUR (sv);
1116
1117 if (SvLEN (sv) <= cur + len)
1118 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1119
942 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1120 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
943 SvCUR_set (sv, SvCUR (sv) + len); 1121 SvCUR_set (sv, SvCUR (sv) + len);
944 } 1122 }
945 else 1123 else
946 sv = newSVpvn (buf, len); 1124 sv = newSVpvn (buf, len);
1037 1215
1038 // special case the rather common 1..5-digit-int case 1216 // special case the rather common 1..5-digit-int case
1039 if (*start == '-') 1217 if (*start == '-')
1040 switch (len) 1218 switch (len)
1041 { 1219 {
1042 case 2: return newSViv (-( start [1] - '0' * 1)); 1220 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1043 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1221 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1044 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1222 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1045 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1223 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1046 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1224 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1047 } 1225 }
1048 else 1226 else
1049 switch (len) 1227 switch (len)
1050 { 1228 {
1051 case 1: return newSViv ( start [0] - '0' * 1); 1229 case 1: return newSViv ( start [0] - '0' * 1);
1052 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1230 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1053 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1231 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1054 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1232 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1055 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1233 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1056 } 1234 }
1057 1235
1058 { 1236 {
1059 UV uv; 1237 UV uv;
1060 int numtype = grok_number (start, len, &uv); 1238 int numtype = grok_number (start, len, &uv);
1069 } 1247 }
1070 1248
1071 len -= *start == '-' ? 1 : 0; 1249 len -= *start == '-' ? 1 : 0;
1072 1250
1073 // does not fit into IV or UV, try NV 1251 // does not fit into IV or UV, try NV
1074 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1252 if (len <= NV_DIG)
1075 #if defined (LDBL_DIG)
1076 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1077 #endif
1078 )
1079 // fits into NV without loss of precision 1253 // fits into NV without loss of precision
1080 return newSVnv (Atof (start)); 1254 return newSVnv (json_atof (start));
1081 1255
1082 // everything else fails, convert it to a string 1256 // everything else fails, convert it to a string
1083 return newSVpvn (start, dec->cur - start); 1257 return newSVpvn (start, dec->cur - start);
1084 } 1258 }
1085 1259
1086 // loss of precision here 1260 // loss of precision here
1087 return newSVnv (Atof (start)); 1261 return newSVnv (json_atof (start));
1088 1262
1089fail: 1263fail:
1090 return 0; 1264 return 0;
1091} 1265}
1092 1266
1116 if (*dec->cur == ']') 1290 if (*dec->cur == ']')
1117 { 1291 {
1118 ++dec->cur; 1292 ++dec->cur;
1119 break; 1293 break;
1120 } 1294 }
1121 1295
1122 if (*dec->cur != ',') 1296 if (*dec->cur != ',')
1123 ERR (", or ] expected while parsing array"); 1297 ERR (", or ] expected while parsing array");
1124 1298
1125 ++dec->cur; 1299 ++dec->cur;
1126 1300
1168 char *p = dec->cur; 1342 char *p = dec->cur;
1169 char *e = p + 24; // only try up to 24 bytes 1343 char *e = p + 24; // only try up to 24 bytes
1170 1344
1171 for (;;) 1345 for (;;)
1172 { 1346 {
1173 // the >= 0x80 is true on most architectures 1347 // the >= 0x80 is false on most architectures
1174 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1348 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1175 { 1349 {
1176 // slow path, back up and use decode_str 1350 // slow path, back up and use decode_str
1177 SV *key = decode_str (dec); 1351 SV *key = decode_str (dec);
1178 if (!key) 1352 if (!key)
1261 dSP; 1435 dSP;
1262 int count; 1436 int count;
1263 1437
1264 ENTER; SAVETMPS; PUSHMARK (SP); 1438 ENTER; SAVETMPS; PUSHMARK (SP);
1265 XPUSHs (HeVAL (he)); 1439 XPUSHs (HeVAL (he));
1440 sv_2mortal (sv);
1266 1441
1267 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1442 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1268 1443
1269 if (count == 1) 1444 if (count == 1)
1270 { 1445 {
1271 sv = newSVsv (POPs); 1446 sv = newSVsv (POPs);
1272 FREETMPS; LEAVE; 1447 FREETMPS; LEAVE;
1273 return sv; 1448 return sv;
1274 } 1449 }
1275 1450
1451 SvREFCNT_inc (sv);
1276 FREETMPS; LEAVE; 1452 FREETMPS; LEAVE;
1277 } 1453 }
1278 } 1454 }
1279 1455
1280 if (dec->json.cb_object) 1456 if (dec->json.cb_object)
1306 DEC_DEC_DEPTH; 1482 DEC_DEC_DEPTH;
1307 return 0; 1483 return 0;
1308} 1484}
1309 1485
1310static SV * 1486static SV *
1487decode_tag (dec_t *dec)
1488{
1489 SV *tag = 0;
1490 SV *val = 0;
1491
1492 if (!(dec->json.flags & F_ALLOW_TAGS))
1493 ERR ("malformed JSON string, neither array, object, number, string or atom");
1494
1495 ++dec->cur;
1496
1497 decode_ws (dec);
1498
1499 tag = decode_sv (dec);
1500 if (!tag)
1501 goto fail;
1502
1503 if (!SvPOK (tag))
1504 ERR ("malformed JSON string, (tag) must be a string");
1505
1506 decode_ws (dec);
1507
1508 if (*dec->cur != ')')
1509 ERR (") expected after tag");
1510
1511 ++dec->cur;
1512
1513 decode_ws (dec);
1514
1515 val = decode_sv (dec);
1516 if (!val)
1517 goto fail;
1518
1519 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1520 ERR ("malformed JSON string, tag value must be an array");
1521
1522 {
1523 AV *av = (AV *)SvRV (val);
1524 int i, len = av_len (av) + 1;
1525 HV *stash = gv_stashsv (tag, 0);
1526 SV *sv;
1527
1528 if (!stash)
1529 ERR ("cannot decode perl-object (package does not exist)");
1530
1531 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1532
1533 if (!method)
1534 ERR ("cannot decode perl-object (package does not have a THAW method)");
1535
1536 dSP;
1537
1538 ENTER; SAVETMPS; PUSHMARK (SP);
1539 EXTEND (SP, len + 2);
1540 // we re-bless the reference to get overload and other niceties right
1541 PUSHs (tag);
1542 PUSHs (sv_json);
1543
1544 for (i = 0; i < len; ++i)
1545 PUSHs (*av_fetch (av, i, 1));
1546
1547 PUTBACK;
1548 call_sv ((SV *)GvCV (method), G_SCALAR);
1549 SPAGAIN;
1550
1551 SvREFCNT_dec (tag);
1552 SvREFCNT_dec (val);
1553 sv = SvREFCNT_inc (POPs);
1554
1555 PUTBACK;
1556
1557 FREETMPS; LEAVE;
1558
1559 return sv;
1560 }
1561
1562fail:
1563 SvREFCNT_dec (tag);
1564 SvREFCNT_dec (val);
1565 return 0;
1566}
1567
1568static SV *
1311decode_sv (dec_t *dec) 1569decode_sv (dec_t *dec)
1312{ 1570{
1313 // the beauty of JSON: you need exactly one character lookahead 1571 // the beauty of JSON: you need exactly one character lookahead
1314 // to parse everything. 1572 // to parse everything.
1315 switch (*dec->cur) 1573 switch (*dec->cur)
1316 { 1574 {
1317 case '"': ++dec->cur; return decode_str (dec); 1575 case '"': ++dec->cur; return decode_str (dec);
1318 case '[': ++dec->cur; return decode_av (dec); 1576 case '[': ++dec->cur; return decode_av (dec);
1319 case '{': ++dec->cur; return decode_hv (dec); 1577 case '{': ++dec->cur; return decode_hv (dec);
1578 case '(': return decode_tag (dec);
1320 1579
1321 case '-': 1580 case '-':
1322 case '0': case '1': case '2': case '3': case '4': 1581 case '0': case '1': case '2': case '3': case '4':
1323 case '5': case '6': case '7': case '8': case '9': 1582 case '5': case '6': case '7': case '8': case '9':
1324 return decode_num (dec); 1583 return decode_num (dec);
1326 case 't': 1585 case 't':
1327 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1586 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1328 { 1587 {
1329 dec->cur += 4; 1588 dec->cur += 4;
1330#if JSON_SLOW 1589#if JSON_SLOW
1331 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1590 types_true = get_bool ("Types::Serialiser::true");
1332#endif 1591#endif
1333 return SvREFCNT_inc (json_true); 1592 return newSVsv (types_true);
1334 } 1593 }
1335 else 1594 else
1336 ERR ("'true' expected"); 1595 ERR ("'true' expected");
1337 1596
1338 break; 1597 break;
1340 case 'f': 1599 case 'f':
1341 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1600 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1342 { 1601 {
1343 dec->cur += 5; 1602 dec->cur += 5;
1344#if JSON_SLOW 1603#if JSON_SLOW
1345 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1604 types_false = get_bool ("Types::Serialiser::false");
1346#endif 1605#endif
1347 return SvREFCNT_inc (json_false); 1606 return newSVsv (types_false);
1348 } 1607 }
1349 else 1608 else
1350 ERR ("'false' expected"); 1609 ERR ("'false' expected");
1351 1610
1352 break; 1611 break;
1361 ERR ("'null' expected"); 1620 ERR ("'null' expected");
1362 1621
1363 break; 1622 break;
1364 1623
1365 default: 1624 default:
1366 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1625 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1367 break; 1626 break;
1368 } 1627 }
1369 1628
1370fail: 1629fail:
1371 return 0; 1630 return 0;
1372} 1631}
1373 1632
1374static SV * 1633static SV *
1375decode_json (SV *string, JSON *json, UV *offset_return) 1634decode_json (SV *string, JSON *json, char **offset_return)
1376{ 1635{
1377 dec_t dec; 1636 dec_t dec;
1378 UV offset;
1379 SV *sv; 1637 SV *sv;
1380 1638
1639 /* work around bugs in 5.10 where manipulating magic values
1640 * makes perl ignore the magic in subsequent accesses.
1641 * also make a copy of non-PV values, to get them into a clean
1642 * state (SvPV should do that, but it's buggy, see below).
1643 */
1381 SvGETMAGIC (string); 1644 /*SvGETMAGIC (string);*/
1645 if (SvMAGICAL (string) || !SvPOK (string))
1646 string = sv_2mortal (newSVsv (string));
1647
1382 SvUPGRADE (string, SVt_PV); 1648 SvUPGRADE (string, SVt_PV);
1383 1649
1384 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1650 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1651 * assertion with -DDEBUGGING, although SvCUR is documented to
1652 * return the xpv_cur field which certainly exists after upgrading.
1653 * according to nicholas clark, calling SvPOK fixes this.
1654 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1655 * and hope for the best.
1656 * Damnit, SvPV_nolen still trips over yet another assertion. This
1657 * assertion business is seriously broken, try yet another workaround
1658 * for the broken -DDEBUGGING.
1659 */
1660 {
1661#ifdef DEBUGGING
1662 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1663#else
1664 STRLEN offset = SvCUR (string);
1665#endif
1666
1667 if (offset > json->max_size && json->max_size)
1385 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1668 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1386 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1669 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1670 }
1387 1671
1388 if (json->flags & F_UTF8) 1672 if (DECODE_WANTS_OCTETS (json))
1389 sv_utf8_downgrade (string, 0); 1673 sv_utf8_downgrade (string, 0);
1390 else 1674 else
1391 sv_utf8_upgrade (string); 1675 sv_utf8_upgrade (string);
1392 1676
1393 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1677 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1394 1678
1395 dec.json = *json; 1679 dec.json = *json;
1396 dec.cur = SvPVX (string); 1680 dec.cur = SvPVX (string);
1397 dec.end = SvEND (string); 1681 dec.end = SvEND (string);
1398 dec.err = 0; 1682 dec.err = 0;
1399 dec.depth = 0; 1683 dec.depth = 0;
1400 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1401 1684
1402 if (dec.json.cb_object || dec.json.cb_sk_object) 1685 if (dec.json.cb_object || dec.json.cb_sk_object)
1403 dec.json.flags |= F_HOOK; 1686 dec.json.flags |= F_HOOK;
1404 1687
1405 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1688 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1406 1689
1407 decode_ws (&dec); 1690 decode_ws (&dec);
1408 sv = decode_sv (&dec); 1691 sv = decode_sv (&dec);
1692
1693 if (offset_return)
1694 *offset_return = dec.cur;
1409 1695
1410 if (!(offset_return || !sv)) 1696 if (!(offset_return || !sv))
1411 { 1697 {
1412 // check for trailing garbage 1698 // check for trailing garbage
1413 decode_ws (&dec); 1699 decode_ws (&dec);
1416 { 1702 {
1417 dec.err = "garbage after JSON object"; 1703 dec.err = "garbage after JSON object";
1418 SvREFCNT_dec (sv); 1704 SvREFCNT_dec (sv);
1419 sv = 0; 1705 sv = 0;
1420 } 1706 }
1421 }
1422
1423 if (offset_return || !sv)
1424 {
1425 offset = dec.json.flags & F_UTF8
1426 ? dec.cur - SvPVX (string)
1427 : utf8_distance (dec.cur, SvPVX (string));
1428
1429 if (offset_return)
1430 *offset_return = offset;
1431 } 1707 }
1432 1708
1433 if (!sv) 1709 if (!sv)
1434 { 1710 {
1435 SV *uni = sv_newmortal (); 1711 SV *uni = sv_newmortal ();
1441 SAVEVPTR (PL_curcop); 1717 SAVEVPTR (PL_curcop);
1442 PL_curcop = &cop; 1718 PL_curcop = &cop;
1443 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1719 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1444 LEAVE; 1720 LEAVE;
1445 1721
1446 croak ("%s, at character offset %d [\"%s\"]", 1722 croak ("%s, at character offset %d (before \"%s\")",
1447 dec.err, 1723 dec.err,
1448 (int)offset, 1724 (int)ptr_to_index (string, dec.cur),
1449 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1725 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1450 } 1726 }
1451 1727
1452 sv = sv_2mortal (sv); 1728 sv = sv_2mortal (sv);
1453 1729
1454 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1730 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1455 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1731 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1456 1732
1457 return sv; 1733 return sv;
1734}
1735
1736/////////////////////////////////////////////////////////////////////////////
1737// incremental parser
1738
1739static void
1740incr_parse (JSON *self)
1741{
1742 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1743
1744 // the state machine here is a bit convoluted and could be simplified a lot
1745 // but this would make it slower, so...
1746
1747 for (;;)
1748 {
1749 //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
1750 switch (self->incr_mode)
1751 {
1752 // only used for initial whitespace skipping
1753 case INCR_M_WS:
1754 for (;;)
1755 {
1756 if (*p > 0x20)
1757 {
1758 if (*p == '#')
1759 {
1760 self->incr_mode = INCR_M_C0;
1761 goto incr_m_c;
1762 }
1763 else
1764 {
1765 self->incr_mode = INCR_M_JSON;
1766 goto incr_m_json;
1767 }
1768 }
1769 else if (!*p)
1770 goto interrupt;
1771
1772 ++p;
1773 }
1774
1775 // skip a single char inside a string (for \\-processing)
1776 case INCR_M_BS:
1777 if (!*p)
1778 goto interrupt;
1779
1780 ++p;
1781 self->incr_mode = INCR_M_STR;
1782 goto incr_m_str;
1783
1784 // inside #-style comments
1785 case INCR_M_C0:
1786 case INCR_M_C1:
1787 incr_m_c:
1788 for (;;)
1789 {
1790 if (*p == '\n')
1791 {
1792 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1793 break;
1794 }
1795 else if (!*p)
1796 goto interrupt;
1797
1798 ++p;
1799 }
1800
1801 break;
1802
1803 // inside a string
1804 case INCR_M_STR:
1805 incr_m_str:
1806 for (;;)
1807 {
1808 if (*p == '"')
1809 {
1810 ++p;
1811 self->incr_mode = INCR_M_JSON;
1812
1813 if (!self->incr_nest)
1814 goto interrupt;
1815
1816 goto incr_m_json;
1817 }
1818 else if (*p == '\\')
1819 {
1820 ++p; // "virtually" consumes character after \
1821
1822 if (!*p) // if at end of string we have to switch modes
1823 {
1824 self->incr_mode = INCR_M_BS;
1825 goto interrupt;
1826 }
1827 }
1828 else if (!*p)
1829 goto interrupt;
1830
1831 ++p;
1832 }
1833
1834 // after initial ws, outside string
1835 case INCR_M_JSON:
1836 incr_m_json:
1837 for (;;)
1838 {
1839 switch (*p++)
1840 {
1841 case 0:
1842 --p;
1843 goto interrupt;
1844
1845 case 0x09:
1846 case 0x0a:
1847 case 0x0d:
1848 case 0x20:
1849 if (!self->incr_nest)
1850 {
1851 --p; // do not eat the whitespace, let the next round do it
1852 goto interrupt;
1853 }
1854 break;
1855
1856 case '"':
1857 self->incr_mode = INCR_M_STR;
1858 goto incr_m_str;
1859
1860 case '[':
1861 case '{':
1862 case '(':
1863 if (++self->incr_nest > self->max_depth)
1864 croak (ERR_NESTING_EXCEEDED);
1865 break;
1866
1867 case ']':
1868 case '}':
1869 if (--self->incr_nest <= 0)
1870 goto interrupt;
1871 break;
1872
1873 case ')':
1874 --self->incr_nest;
1875 break;
1876
1877 case '#':
1878 self->incr_mode = INCR_M_C1;
1879 goto incr_m_c;
1880 }
1881 }
1882 }
1883
1884 modechange:
1885 ;
1886 }
1887
1888interrupt:
1889 self->incr_pos = p - SvPVX (self->incr_text);
1890 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1891 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1458} 1892}
1459 1893
1460///////////////////////////////////////////////////////////////////////////// 1894/////////////////////////////////////////////////////////////////////////////
1461// XS interface functions 1895// XS interface functions
1462 1896
1471 i >= '0' && i <= '9' ? i - '0' 1905 i >= '0' && i <= '9' ? i - '0'
1472 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1906 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1473 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1907 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1474 : -1; 1908 : -1;
1475 1909
1476 json_stash = gv_stashpv ("JSON::XS" , 1); 1910 json_stash = gv_stashpv ("JSON::XS" , 1);
1477 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1911 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1478 1912
1479 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1913 types_true = get_bool ("Types::Serialiser::true");
1480 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1914 types_false = get_bool ("Types::Serialiser::false");
1915
1916 sv_json = newSVpv ("JSON", 0);
1917 SvREADONLY_on (sv_json);
1918
1919 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1481} 1920}
1482 1921
1483PROTOTYPES: DISABLE 1922PROTOTYPES: DISABLE
1484 1923
1485void CLONE (...) 1924void CLONE (...)
1486 CODE: 1925 CODE:
1487 json_stash = 0; 1926 json_stash = 0;
1488 json_boolean_stash = 0; 1927 types_boolean_stash = 0;
1489 1928
1490void new (char *klass) 1929void new (char *klass)
1491 PPCODE: 1930 PPCODE:
1492{ 1931{
1493 SV *pv = NEWSV (0, sizeof (JSON)); 1932 SV *pv = NEWSV (0, sizeof (JSON));
1494 SvPOK_only (pv); 1933 SvPOK_only (pv);
1495 Zero (SvPVX (pv), 1, JSON); 1934 json_init ((JSON *)SvPVX (pv));
1496 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1497 XPUSHs (sv_2mortal (sv_bless ( 1935 XPUSHs (sv_2mortal (sv_bless (
1498 newRV_noinc (pv), 1936 newRV_noinc (pv),
1499 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1937 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1500 ))); 1938 )));
1501} 1939}
1513 allow_nonref = F_ALLOW_NONREF 1951 allow_nonref = F_ALLOW_NONREF
1514 shrink = F_SHRINK 1952 shrink = F_SHRINK
1515 allow_blessed = F_ALLOW_BLESSED 1953 allow_blessed = F_ALLOW_BLESSED
1516 convert_blessed = F_CONV_BLESSED 1954 convert_blessed = F_CONV_BLESSED
1517 relaxed = F_RELAXED 1955 relaxed = F_RELAXED
1956 allow_unknown = F_ALLOW_UNKNOWN
1957 allow_tags = F_ALLOW_TAGS
1518 PPCODE: 1958 PPCODE:
1519{ 1959{
1520 if (enable) 1960 if (enable)
1521 self->flags |= ix; 1961 self->flags |= ix;
1522 else 1962 else
1537 get_allow_nonref = F_ALLOW_NONREF 1977 get_allow_nonref = F_ALLOW_NONREF
1538 get_shrink = F_SHRINK 1978 get_shrink = F_SHRINK
1539 get_allow_blessed = F_ALLOW_BLESSED 1979 get_allow_blessed = F_ALLOW_BLESSED
1540 get_convert_blessed = F_CONV_BLESSED 1980 get_convert_blessed = F_CONV_BLESSED
1541 get_relaxed = F_RELAXED 1981 get_relaxed = F_RELAXED
1982 get_allow_unknown = F_ALLOW_UNKNOWN
1983 get_allow_tags = F_ALLOW_TAGS
1542 PPCODE: 1984 PPCODE:
1543 XPUSHs (boolSV (self->flags & ix)); 1985 XPUSHs (boolSV (self->flags & ix));
1544 1986
1545void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1987void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1546 PPCODE: 1988 PPCODE:
1547{ 1989 self->max_depth = max_depth;
1548 UV log2 = 0;
1549
1550 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1551
1552 while ((1UL << log2) < max_depth)
1553 ++log2;
1554
1555 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1556
1557 XPUSHs (ST (0)); 1990 XPUSHs (ST (0));
1558}
1559 1991
1560U32 get_max_depth (JSON *self) 1992U32 get_max_depth (JSON *self)
1561 CODE: 1993 CODE:
1562 RETVAL = DEC_DEPTH (self->flags); 1994 RETVAL = self->max_depth;
1563 OUTPUT: 1995 OUTPUT:
1564 RETVAL 1996 RETVAL
1565 1997
1566void max_size (JSON *self, UV max_size = 0) 1998void max_size (JSON *self, U32 max_size = 0)
1567 PPCODE: 1999 PPCODE:
1568{ 2000 self->max_size = max_size;
1569 UV log2 = 0;
1570
1571 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1572 if (max_size == 1) max_size = 2;
1573
1574 while ((1UL << log2) < max_size)
1575 ++log2;
1576
1577 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1578
1579 XPUSHs (ST (0)); 2001 XPUSHs (ST (0));
1580}
1581 2002
1582int get_max_size (JSON *self) 2003int get_max_size (JSON *self)
1583 CODE: 2004 CODE:
1584 RETVAL = DEC_SIZE (self->flags); 2005 RETVAL = self->max_size;
1585 OUTPUT: 2006 OUTPUT:
1586 RETVAL 2007 RETVAL
1587 2008
1588void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2009void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1589 PPCODE: 2010 PPCODE:
1595} 2016}
1596 2017
1597void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2018void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1598 PPCODE: 2019 PPCODE:
1599{ 2020{
1600 if (!self->cb_sk_object) 2021 if (!self->cb_sk_object)
1601 self->cb_sk_object = newHV (); 2022 self->cb_sk_object = newHV ();
1602 2023
1603 if (SvOK (cb)) 2024 if (SvOK (cb))
1604 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2025 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1605 else 2026 else
1616 XPUSHs (ST (0)); 2037 XPUSHs (ST (0));
1617} 2038}
1618 2039
1619void encode (JSON *self, SV *scalar) 2040void encode (JSON *self, SV *scalar)
1620 PPCODE: 2041 PPCODE:
1621 XPUSHs (encode_json (scalar, self)); 2042 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2043 XPUSHs (scalar);
1622 2044
1623void decode (JSON *self, SV *jsonstr) 2045void decode (JSON *self, SV *jsonstr)
1624 PPCODE: 2046 PPCODE:
1625 XPUSHs (decode_json (jsonstr, self, 0)); 2047 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2048 XPUSHs (jsonstr);
1626 2049
1627void decode_prefix (JSON *self, SV *jsonstr) 2050void decode_prefix (JSON *self, SV *jsonstr)
1628 PPCODE: 2051 PPCODE:
1629{ 2052{
1630 UV offset; 2053 SV *sv;
2054 char *offset;
2055 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1631 EXTEND (SP, 2); 2056 EXTEND (SP, 2);
1632 PUSHs (decode_json (jsonstr, self, &offset)); 2057 PUSHs (sv);
1633 PUSHs (sv_2mortal (newSVuv (offset))); 2058 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
2059}
2060
2061void incr_parse (JSON *self, SV *jsonstr = 0)
2062 PPCODE:
2063{
2064 if (!self->incr_text)
2065 self->incr_text = newSVpvn ("", 0);
2066
2067 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2068 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2069 if (DECODE_WANTS_OCTETS (self))
2070 {
2071 if (self->incr_pos)
2072 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2073 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2074
2075 sv_utf8_downgrade (self->incr_text, 0);
2076 }
2077 else
2078 {
2079 sv_utf8_upgrade (self->incr_text);
2080
2081 if (self->incr_pos)
2082 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2083 - (U8 *)SvPVX (self->incr_text);
2084 }
2085
2086 // append data, if any
2087 if (jsonstr)
2088 {
2089 /* make sure both strings have same encoding */
2090 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2091 if (SvUTF8 (jsonstr))
2092 sv_utf8_downgrade (jsonstr, 0);
2093 else
2094 sv_utf8_upgrade (jsonstr);
2095
2096 /* and then just blindly append */
2097 {
2098 STRLEN len;
2099 const char *str = SvPV (jsonstr, len);
2100 STRLEN cur = SvCUR (self->incr_text);
2101
2102 if (SvLEN (self->incr_text) <= cur + len)
2103 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2104
2105 Move (str, SvEND (self->incr_text), len, char);
2106 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2107 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2108 }
2109 }
2110
2111 if (GIMME_V != G_VOID)
2112 do
2113 {
2114 SV *sv;
2115 char *offset;
2116
2117 if (!INCR_DONE (self))
2118 {
2119 incr_parse (self);
2120
2121 if (self->incr_pos > self->max_size && self->max_size)
2122 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2123 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2124
2125 if (!INCR_DONE (self))
2126 {
2127 // as an optimisation, do not accumulate white space in the incr buffer
2128 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2129 {
2130 self->incr_pos = 0;
2131 SvCUR_set (self->incr_text, 0);
2132 }
2133
2134 break;
2135 }
2136 }
2137
2138 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2139 XPUSHs (sv);
2140
2141 self->incr_pos -= offset - SvPVX (self->incr_text);
2142 self->incr_nest = 0;
2143 self->incr_mode = 0;
2144
2145 sv_chop (self->incr_text, offset);
2146 }
2147 while (GIMME_V == G_ARRAY);
2148}
2149
2150SV *incr_text (JSON *self)
2151 ATTRS: lvalue
2152 CODE:
2153{
2154 if (self->incr_pos)
2155 croak ("incr_text can not be called when the incremental parser already started parsing");
2156
2157 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2158}
2159 OUTPUT:
2160 RETVAL
2161
2162void incr_skip (JSON *self)
2163 CODE:
2164{
2165 if (self->incr_pos)
2166 {
2167 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2168 self->incr_pos = 0;
2169 self->incr_nest = 0;
2170 self->incr_mode = 0;
2171 }
2172}
2173
2174void incr_reset (JSON *self)
2175 CODE:
2176{
2177 SvREFCNT_dec (self->incr_text);
2178 self->incr_text = 0;
2179 self->incr_pos = 0;
2180 self->incr_nest = 0;
2181 self->incr_mode = 0;
1634} 2182}
1635 2183
1636void DESTROY (JSON *self) 2184void DESTROY (JSON *self)
1637 CODE: 2185 CODE:
1638 SvREFCNT_dec (self->cb_sk_object); 2186 SvREFCNT_dec (self->cb_sk_object);
1639 SvREFCNT_dec (self->cb_object); 2187 SvREFCNT_dec (self->cb_object);
2188 SvREFCNT_dec (self->incr_text);
1640 2189
1641PROTOTYPES: ENABLE 2190PROTOTYPES: ENABLE
1642 2191
1643void encode_json (SV *scalar) 2192void encode_json (SV *scalar)
1644 PPCODE: 2193 PPCODE:
1645{ 2194{
1646 JSON json = { F_DEFAULT | F_UTF8 }; 2195 JSON json;
1647 XPUSHs (encode_json (scalar, &json)); 2196 json_init (&json);
2197 json.flags |= F_UTF8;
2198 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2199 XPUSHs (scalar);
1648} 2200}
1649 2201
1650void decode_json (SV *jsonstr) 2202void decode_json (SV *jsonstr)
1651 PPCODE: 2203 PPCODE:
1652{ 2204{
1653 JSON json = { F_DEFAULT | F_UTF8 }; 2205 JSON json;
1654 XPUSHs (decode_json (jsonstr, &json, 0)); 2206 json_init (&json);
2207 json.flags |= F_UTF8;
2208 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2209 XPUSHs (jsonstr);
1655} 2210}
1656 2211

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines