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 1.130 by root, Sat Nov 26 06:09:29 2016 UTC

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 <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 15#endif
15 16
16// some old perls do not have this, try to make it work, no 17// 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. 18// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
20#endif 21#endif
21 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 36
24#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 44#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
36 49#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL 50#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 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 52
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 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 54
49#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
51 57
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
59
60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
53 61
54#define SB do { 62#define SB do {
55#define SE } while (0) 63#define SE } while (0)
56 64
57#if __GNUC__ >= 3 65#if __GNUC__ >= 3
67 75
68#define IN_RANGE_INC(type,val,beg,end) \ 76#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 77 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 79
80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
81
72#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
73# define JSON_SLOW 1 83# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 84# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
85# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
75#else 86#else
76# define JSON_SLOW 0 87# define JSON_SLOW 0
77# define JSON_STASH json_stash 88# define JSON_STASH json_stash
89# define BOOL_STASH bool_stash
78#endif 90#endif
79 91
80static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
81static SV *json_true, *json_false; 93#define STACK_HES 64
94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_true, *bool_false, *sv_json;
97
98enum {
99 INCR_M_WS = 0, // initial whitespace skipping, must be 0
100 INCR_M_STR, // inside string
101 INCR_M_BS, // inside backslash
102 INCR_M_C0, // inside comment in initial whitespace sequence
103 INCR_M_C1, // inside comment in other places
104 INCR_M_JSON // outside anything, count nesting
105};
106
107#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
82 108
83typedef struct { 109typedef struct {
84 U32 flags; 110 U32 flags;
111 U32 max_depth;
112 STRLEN max_size;
113
85 SV *cb_object; 114 SV *cb_object;
86 HV *cb_sk_object; 115 HV *cb_sk_object;
116
117 // for the incremental parser
118 SV *incr_text; // the source text so far
119 STRLEN incr_pos; // the current offset into the text
120 int incr_nest; // {[]}-nesting level
121 unsigned char incr_mode;
87} JSON; 122} JSON;
123
124INLINE void
125json_init (JSON *json)
126{
127 Zero (json, 1, JSON);
128 json->max_depth = 512;
129}
88 130
89///////////////////////////////////////////////////////////////////////////// 131/////////////////////////////////////////////////////////////////////////////
90// utility functions 132// utility functions
91 133
134INLINE SV *
135get_bool (const char *name)
136{
137 SV *sv = get_sv (name, 1);
138
139 SvREADONLY_on (sv);
140 SvREADONLY_on (SvRV (sv));
141
142 return sv;
143}
144
92INLINE void 145INLINE void
93shrink (SV *sv) 146shrink (SV *sv)
94{ 147{
95 sv_utf8_downgrade (sv, 1); 148 sv_utf8_downgrade (sv, 1);
149
96 if (SvLEN (sv) > SvCUR (sv) + 1) 150 if (SvLEN (sv) > SvCUR (sv) + 1)
97 { 151 {
98#ifdef SvPV_shrink_to_cur 152#ifdef SvPV_shrink_to_cur
99 SvPV_shrink_to_cur (sv); 153 SvPV_shrink_to_cur (sv);
100#elif defined (SvPV_renew) 154#elif defined (SvPV_renew)
144 *s++ = 0x80 | ( ch & 0x3f); 198 *s++ = 0x80 | ( ch & 0x3f);
145 199
146 return s; 200 return s;
147} 201}
148 202
203// convert offset pointer to character index, sv must be string
204static STRLEN
205ptr_to_index (SV *sv, char *offset)
206{
207 return SvUTF8 (sv)
208 ? utf8_distance (offset, SvPVX (sv))
209 : offset - SvPVX (sv);
210}
211
212/////////////////////////////////////////////////////////////////////////////
213// fp hell
214
215// scan a group of digits, and a trailing exponent
216static void
217json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
218{
219 UV uaccum = 0;
220 int eaccum = 0;
221
222 // if we recurse too deep, skip all remaining digits
223 // to avoid a stack overflow attack
224 if (expect_false (--maxdepth <= 0))
225 while (((U8)*s - '0') < 10)
226 ++s;
227
228 for (;;)
229 {
230 U8 dig = (U8)*s - '0';
231
232 if (expect_false (dig >= 10))
233 {
234 if (dig == (U8)((U8)'.' - (U8)'0'))
235 {
236 ++s;
237 json_atof_scan1 (s, accum, expo, 1, maxdepth);
238 }
239 else if ((dig | ' ') == 'e' - '0')
240 {
241 int exp2 = 0;
242 int neg = 0;
243
244 ++s;
245
246 if (*s == '-')
247 {
248 ++s;
249 neg = 1;
250 }
251 else if (*s == '+')
252 ++s;
253
254 while ((dig = (U8)*s - '0') < 10)
255 exp2 = exp2 * 10 + *s++ - '0';
256
257 *expo += neg ? -exp2 : exp2;
258 }
259
260 break;
261 }
262
263 ++s;
264
265 uaccum = uaccum * 10 + dig;
266 ++eaccum;
267
268 // if we have too many digits, then recurse for more
269 // we actually do this for rather few digits
270 if (uaccum >= (UV_MAX - 9) / 10)
271 {
272 if (postdp) *expo -= eaccum;
273 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
274 if (postdp) *expo += eaccum;
275
276 break;
277 }
278 }
279
280 // this relies greatly on the quality of the pow ()
281 // implementation of the platform, but a good
282 // implementation is hard to beat.
283 // (IEEE 754 conformant ones are required to be exact)
284 if (postdp) *expo -= eaccum;
285 *accum += uaccum * Perl_pow (10., *expo);
286 *expo += eaccum;
287}
288
289static NV
290json_atof (const char *s)
291{
292 NV accum = 0.;
293 int expo = 0;
294 int neg = 0;
295
296 if (*s == '-')
297 {
298 ++s;
299 neg = 1;
300 }
301
302 // a recursion depth of ten gives us >>500 bits
303 json_atof_scan1 (s, &accum, &expo, 0, 10);
304
305 return neg ? -accum : accum;
306}
307
308// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
309static int
310ref_bool_type (SV *sv)
311{
312 svtype svt = SvTYPE (sv);
313
314 if (svt < SVt_PVAV)
315 {
316 STRLEN len = 0;
317 char *pv = svt ? SvPV (sv, len) : 0;
318
319 if (len == 1)
320 if (*pv == '1')
321 return 1;
322 else if (*pv == '0')
323 return 0;
324 }
325
326 return -1;
327}
328
329// returns whether scalar is not a reference in the sense of allow_nonref
330static int
331json_nonref (SV *scalar)
332{
333 if (!SvROK (scalar))
334 return 1;
335
336 scalar = SvRV (scalar);
337
338 if (SvTYPE (scalar) >= SVt_PVMG)
339 {
340 if (SvSTASH (scalar) == bool_stash)
341 return 1;
342
343 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
344 return 1;
345 }
346
347 return 0;
348}
349
149///////////////////////////////////////////////////////////////////////////// 350/////////////////////////////////////////////////////////////////////////////
150// encoder 351// encoder
151 352
152// structure used for encoding JSON 353// structure used for encoding JSON
153typedef struct 354typedef struct
155 char *cur; // SvPVX (sv) + current output position 356 char *cur; // SvPVX (sv) + current output position
156 char *end; // SvEND (sv) 357 char *end; // SvEND (sv)
157 SV *sv; // result scalar 358 SV *sv; // result scalar
158 JSON json; 359 JSON json;
159 U32 indent; // indentation level 360 U32 indent; // indentation level
160 U32 maxdepth; // max. indentation/recursion level
161 UV limit; // escape character values >= this value when encoding 361 UV limit; // escape character values >= this value when encoding
162} enc_t; 362} enc_t;
163 363
164INLINE void 364INLINE void
165need (enc_t *enc, STRLEN len) 365need (enc_t *enc, STRLEN len)
166{ 366{
167 if (expect_false (enc->cur + len >= enc->end)) 367 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
168 { 368 {
169 STRLEN cur = enc->cur - SvPVX (enc->sv); 369 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
170 SvGROW (enc->sv, cur + len + 1); 370 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
171 enc->cur = SvPVX (enc->sv) + cur; 371 enc->cur = SvPVX (enc->sv) + cur;
172 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 372 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
173 } 373 }
174} 374}
175 375
250 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 450 (int)((uch - 0x10000) % 0x400 + 0xDC00));
251 enc->cur += 12; 451 enc->cur += 12;
252 } 452 }
253 else 453 else
254 { 454 {
255 static char hexdigit [16] = "0123456789abcdef";
256 need (enc, len += 5); 455 need (enc, len += 5);
257 *enc->cur++ = '\\'; 456 *enc->cur++ = '\\';
258 *enc->cur++ = 'u'; 457 *enc->cur++ = 'u';
259 *enc->cur++ = hexdigit [ uch >> 12 ]; 458 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
260 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 459 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
261 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 460 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
262 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 461 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
263 } 462 }
264 463
265 str += clen; 464 str += clen;
266 } 465 }
267 else if (enc->json.flags & F_LATIN1) 466 else if (enc->json.flags & F_LATIN1)
338static void 537static void
339encode_av (enc_t *enc, AV *av) 538encode_av (enc_t *enc, AV *av)
340{ 539{
341 int i, len = av_len (av); 540 int i, len = av_len (av);
342 541
343 if (enc->indent >= enc->maxdepth) 542 if (enc->indent >= enc->json.max_depth)
344 croak ("data structure too deep (hit recursion limit)"); 543 croak (ERR_NESTING_EXCEEDED);
345 544
346 encode_ch (enc, '['); 545 encode_ch (enc, '[');
347 546
348 if (len >= 0) 547 if (len >= 0)
349 { 548 {
350 encode_nl (enc); ++enc->indent; 549 encode_nl (enc); ++enc->indent;
351 550
352 for (i = 0; i <= len; ++i) 551 for (i = 0; i <= len; ++i)
364 encode_comma (enc); 563 encode_comma (enc);
365 } 564 }
366 565
367 encode_nl (enc); --enc->indent; encode_indent (enc); 566 encode_nl (enc); --enc->indent; encode_indent (enc);
368 } 567 }
369 568
370 encode_ch (enc, ']'); 569 encode_ch (enc, ']');
371} 570}
372 571
373static void 572static void
374encode_hk (enc_t *enc, HE *he) 573encode_hk (enc_t *enc, HE *he)
378 if (HeKLEN (he) == HEf_SVKEY) 577 if (HeKLEN (he) == HEf_SVKEY)
379 { 578 {
380 SV *sv = HeSVKEY (he); 579 SV *sv = HeSVKEY (he);
381 STRLEN len; 580 STRLEN len;
382 char *str; 581 char *str;
383 582
384 SvGETMAGIC (sv); 583 SvGETMAGIC (sv);
385 str = SvPV (sv, len); 584 str = SvPV (sv, len);
386 585
387 encode_str (enc, str, len, SvUTF8 (sv)); 586 encode_str (enc, str, len, SvUTF8 (sv));
388 } 587 }
423 622
424static void 623static void
425encode_hv (enc_t *enc, HV *hv) 624encode_hv (enc_t *enc, HV *hv)
426{ 625{
427 HE *he; 626 HE *he;
428 int count;
429 627
430 if (enc->indent >= enc->maxdepth) 628 if (enc->indent >= enc->json.max_depth)
431 croak ("data structure too deep (hit recursion limit)"); 629 croak (ERR_NESTING_EXCEEDED);
432 630
433 encode_ch (enc, '{'); 631 encode_ch (enc, '{');
434 632
435 // for canonical output we have to sort by keys first 633 // for canonical output we have to sort by keys first
436 // actually, this is mostly due to the stupid so-called 634 // actually, this is mostly due to the stupid so-called
437 // security workaround added somewhere in 5.8.x. 635 // security workaround added somewhere in 5.8.x
438 // that randomises hash orderings 636 // that randomises hash orderings
439 if (enc->json.flags & F_CANONICAL) 637 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
440 { 638 {
441 int count = hv_iterinit (hv); 639 int count = hv_iterinit (hv);
442 640
443 if (SvMAGICAL (hv)) 641 if (SvMAGICAL (hv))
444 { 642 {
455 } 653 }
456 654
457 if (count) 655 if (count)
458 { 656 {
459 int i, fast = 1; 657 int i, fast = 1;
460#if defined(__BORLANDC__) || defined(_MSC_VER) 658 HE *hes_stack [STACK_HES];
461 HE **hes = _alloca (count * sizeof (HE)); 659 HE **hes = hes_stack;
462#else 660
463 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 661 // allocate larger arrays on the heap
464#endif 662 if (count > STACK_HES)
663 {
664 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
665 hes = (HE **)SvPVX (sv);
666 }
465 667
466 i = 0; 668 i = 0;
467 while ((he = hv_iternext (hv))) 669 while ((he = hv_iternext (hv)))
468 { 670 {
469 hes [i++] = he; 671 hes [i++] = he;
538// encode objects, arrays and special \0=false and \1=true values. 740// encode objects, arrays and special \0=false and \1=true values.
539static void 741static void
540encode_rv (enc_t *enc, SV *sv) 742encode_rv (enc_t *enc, SV *sv)
541{ 743{
542 svtype svt; 744 svtype svt;
745 GV *method;
543 746
544 SvGETMAGIC (sv); 747 SvGETMAGIC (sv);
545 svt = SvTYPE (sv); 748 svt = SvTYPE (sv);
546 749
547 if (expect_false (SvOBJECT (sv))) 750 if (expect_false (SvOBJECT (sv)))
548 { 751 {
549 HV *stash = !JSON_SLOW || json_boolean_stash 752 HV *stash = SvSTASH (sv);
550 ? json_boolean_stash
551 : gv_stashpv ("JSON::XS::Boolean", 1);
552 753
553 if (SvSTASH (sv) == stash) 754 if (stash == bool_stash)
554 { 755 {
555 if (SvIV (sv)) 756 if (SvIV (sv))
556 encode_str (enc, "true", 4, 0); 757 encode_str (enc, "true", 4, 0);
557 else 758 else
558 encode_str (enc, "false", 5, 0); 759 encode_str (enc, "false", 5, 0);
559 } 760 }
761 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
762 {
763 int count;
764 dSP;
765
766 ENTER; SAVETMPS;
767 SAVESTACK_POS ();
768 PUSHMARK (SP);
769 EXTEND (SP, 2);
770 // we re-bless the reference to get overload and other niceties right
771 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
772 PUSHs (sv_json);
773
774 PUTBACK;
775 count = call_sv ((SV *)GvCV (method), G_ARRAY);
776 SPAGAIN;
777
778 // catch this surprisingly common error
779 if (SvROK (TOPs) && SvRV (TOPs) == sv)
780 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
781
782 encode_ch (enc, '(');
783 encode_ch (enc, '"');
784 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
785 encode_ch (enc, '"');
786 encode_ch (enc, ')');
787 encode_ch (enc, '[');
788
789 while (count)
790 {
791 encode_sv (enc, SP[1 - count--]);
792
793 if (count)
794 encode_ch (enc, ',');
795 }
796
797 encode_ch (enc, ']');
798
799 FREETMPS; LEAVE;
800 }
801 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
802 {
803 dSP;
804
805 ENTER; SAVETMPS;
806 PUSHMARK (SP);
807 // we re-bless the reference to get overload and other niceties right
808 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
809
810 // calling with G_SCALAR ensures that we always get a 1 return value
811 PUTBACK;
812 call_sv ((SV *)GvCV (method), G_SCALAR);
813 SPAGAIN;
814
815 // catch this surprisingly common error
816 if (SvROK (TOPs) && SvRV (TOPs) == sv)
817 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
818
819 sv = POPs;
820 PUTBACK;
821
822 encode_sv (enc, sv);
823
824 FREETMPS; LEAVE;
825 }
826 else if (enc->json.flags & F_ALLOW_BLESSED)
827 encode_str (enc, "null", 4, 0);
560 else 828 else
561 {
562#if 0
563 if (0 && sv_derived_from (rv, "JSON::Literal"))
564 {
565 // not yet
566 }
567#endif
568 if (enc->json.flags & F_CONV_BLESSED)
569 {
570 // 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)));
579
580 // calling with G_SCALAR ensures that we always get a 1 return value
581 PUTBACK;
582 call_sv ((SV *)GvCV (to_json), G_SCALAR);
583 SPAGAIN;
584
585 // catch this surprisingly common error
586 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)));
588
589 sv = POPs;
590 PUTBACK;
591
592 encode_sv (enc, sv);
593
594 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 }
602 else if (enc->json.flags & F_ALLOW_BLESSED)
603 encode_str (enc, "null", 4, 0);
604 else
605 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 829 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
606 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 830 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
607 }
608 } 831 }
609 else if (svt == SVt_PVHV) 832 else if (svt == SVt_PVHV)
610 encode_hv (enc, (HV *)sv); 833 encode_hv (enc, (HV *)sv);
611 else if (svt == SVt_PVAV) 834 else if (svt == SVt_PVAV)
612 encode_av (enc, (AV *)sv); 835 encode_av (enc, (AV *)sv);
613 else if (svt < SVt_PVAV) 836 else if (svt < SVt_PVAV)
614 { 837 {
615 STRLEN len = 0; 838 int bool_type = ref_bool_type (sv);
616 char *pv = svt ? SvPV (sv, len) : 0;
617 839
618 if (len == 1 && *pv == '1') 840 if (bool_type == 1)
619 encode_str (enc, "true", 4, 0); 841 encode_str (enc, "true", 4, 0);
620 else if (len == 1 && *pv == '0') 842 else if (bool_type == 0)
621 encode_str (enc, "false", 5, 0); 843 encode_str (enc, "false", 5, 0);
844 else if (enc->json.flags & F_ALLOW_UNKNOWN)
845 encode_str (enc, "null", 4, 0);
622 else 846 else
623 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 847 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
624 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 848 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
625 } 849 }
850 else if (enc->json.flags & F_ALLOW_UNKNOWN)
851 encode_str (enc, "null", 4, 0);
626 else 852 else
627 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 853 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
628 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 854 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
629} 855}
630 856
684 } 910 }
685 else 911 else
686 { 912 {
687 // large integer, use the (rather slow) snprintf way. 913 // large integer, use the (rather slow) snprintf way.
688 need (enc, IVUV_MAXCHARS); 914 need (enc, IVUV_MAXCHARS);
689 enc->cur += 915 enc->cur +=
690 SvIsUV(sv) 916 SvIsUV(sv)
691 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 917 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
692 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 918 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
693 } 919 }
694 } 920 }
695 else if (SvROK (sv)) 921 else if (SvROK (sv))
696 encode_rv (enc, SvRV (sv)); 922 encode_rv (enc, SvRV (sv));
697 else if (!SvOK (sv)) 923 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
698 encode_str (enc, "null", 4, 0); 924 encode_str (enc, "null", 4, 0);
699 else 925 else
700 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 926 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
701 SvPV_nolen (sv), SvFLAGS (sv)); 927 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
702} 928}
703 929
704static SV * 930static SV *
705encode_json (SV *scalar, JSON *json) 931encode_json (SV *scalar, JSON *json)
706{ 932{
707 enc_t enc; 933 enc_t enc;
708 934
709 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 935 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
710 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 936 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
711 937
712 enc.json = *json; 938 enc.json = *json;
713 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 939 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
714 enc.cur = SvPVX (enc.sv); 940 enc.cur = SvPVX (enc.sv);
715 enc.end = SvEND (enc.sv); 941 enc.end = SvEND (enc.sv);
716 enc.indent = 0; 942 enc.indent = 0;
717 enc.maxdepth = DEC_DEPTH (enc.json.flags);
718 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 943 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
719 : enc.json.flags & F_LATIN1 ? 0x000100UL 944 : enc.json.flags & F_LATIN1 ? 0x000100UL
720 : 0x110000UL; 945 : 0x110000UL;
721 946
722 SvPOK_only (enc.sv); 947 SvPOK_only (enc.sv);
723 encode_sv (&enc, scalar); 948 encode_sv (&enc, scalar);
949 encode_nl (&enc);
724 950
725 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 951 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
726 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 952 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
727 953
728 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 954 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
789 if (*dec->cur != ch) \ 1015 if (*dec->cur != ch) \
790 ERR (# ch " expected"); \ 1016 ERR (# ch " expected"); \
791 ++dec->cur; \ 1017 ++dec->cur; \
792 SE 1018 SE
793 1019
794#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1020#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
795#define DEC_DEC_DEPTH --dec->depth 1021#define DEC_DEC_DEPTH --dec->depth
796 1022
797static SV *decode_sv (dec_t *dec); 1023static SV *decode_sv (dec_t *dec);
798 1024
799static signed char decode_hexdigit[256]; 1025static signed char decode_hexdigit[256];
908 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1134 else if (expect_true (ch >= 0x20 && ch < 0x80))
909 *cur++ = ch; 1135 *cur++ = ch;
910 else if (ch >= 0x80) 1136 else if (ch >= 0x80)
911 { 1137 {
912 STRLEN clen; 1138 STRLEN clen;
913 UV uch;
914 1139
915 --dec_cur; 1140 --dec_cur;
916 1141
917 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1142 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
918 if (clen == (STRLEN)-1) 1143 if (clen == (STRLEN)-1)
919 ERR ("malformed UTF-8 character in JSON string"); 1144 ERR ("malformed UTF-8 character in JSON string");
920 1145
921 do 1146 do
922 *cur++ = *dec_cur++; 1147 *cur++ = *dec_cur++;
923 while (--clen); 1148 while (--clen);
924 1149
925 utf8 = 1; 1150 utf8 = 1;
926 } 1151 }
1152 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1153 *cur++ = ch;
927 else 1154 else
928 { 1155 {
929 --dec_cur; 1156 --dec_cur;
930 1157
931 if (!ch) 1158 if (!ch)
939 { 1166 {
940 STRLEN len = cur - buf; 1167 STRLEN len = cur - buf;
941 1168
942 if (sv) 1169 if (sv)
943 { 1170 {
944 SvGROW (sv, SvCUR (sv) + len + 1); 1171 STRLEN cur = SvCUR (sv);
1172
1173 if (SvLEN (sv) <= cur + len)
1174 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1175
945 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1176 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
946 SvCUR_set (sv, SvCUR (sv) + len); 1177 SvCUR_set (sv, SvCUR (sv) + len);
947 } 1178 }
948 else 1179 else
949 sv = newSVpvn (buf, len); 1180 sv = newSVpvn (buf, len);
1040 1271
1041 // special case the rather common 1..5-digit-int case 1272 // special case the rather common 1..5-digit-int case
1042 if (*start == '-') 1273 if (*start == '-')
1043 switch (len) 1274 switch (len)
1044 { 1275 {
1045 case 2: return newSViv (-( start [1] - '0' * 1)); 1276 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1046 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1277 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)); 1278 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)); 1279 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)); 1280 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1050 } 1281 }
1051 else 1282 else
1052 switch (len) 1283 switch (len)
1053 { 1284 {
1054 case 1: return newSViv ( start [0] - '0' * 1); 1285 case 1: return newSViv ( start [0] - '0' * 1);
1055 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1286 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); 1287 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); 1288 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); 1289 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1059 } 1290 }
1060 1291
1061 { 1292 {
1062 UV uv; 1293 UV uv;
1063 int numtype = grok_number (start, len, &uv); 1294 int numtype = grok_number (start, len, &uv);
1072 } 1303 }
1073 1304
1074 len -= *start == '-' ? 1 : 0; 1305 len -= *start == '-' ? 1 : 0;
1075 1306
1076 // does not fit into IV or UV, try NV 1307 // does not fit into IV or UV, try NV
1077 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1308 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 1309 // fits into NV without loss of precision
1083 return newSVnv (Atof (start)); 1310 return newSVnv (json_atof (start));
1084 1311
1085 // everything else fails, convert it to a string 1312 // everything else fails, convert it to a string
1086 return newSVpvn (start, dec->cur - start); 1313 return newSVpvn (start, dec->cur - start);
1087 } 1314 }
1088 1315
1089 // loss of precision here 1316 // loss of precision here
1090 return newSVnv (Atof (start)); 1317 return newSVnv (json_atof (start));
1091 1318
1092fail: 1319fail:
1093 return 0; 1320 return 0;
1094} 1321}
1095 1322
1119 if (*dec->cur == ']') 1346 if (*dec->cur == ']')
1120 { 1347 {
1121 ++dec->cur; 1348 ++dec->cur;
1122 break; 1349 break;
1123 } 1350 }
1124 1351
1125 if (*dec->cur != ',') 1352 if (*dec->cur != ',')
1126 ERR (", or ] expected while parsing array"); 1353 ERR (", or ] expected while parsing array");
1127 1354
1128 ++dec->cur; 1355 ++dec->cur;
1129 1356
1171 char *p = dec->cur; 1398 char *p = dec->cur;
1172 char *e = p + 24; // only try up to 24 bytes 1399 char *e = p + 24; // only try up to 24 bytes
1173 1400
1174 for (;;) 1401 for (;;)
1175 { 1402 {
1176 // the >= 0x80 is true on most architectures 1403 // the >= 0x80 is false on most architectures
1177 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1404 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1178 { 1405 {
1179 // slow path, back up and use decode_str 1406 // slow path, back up and use decode_str
1180 SV *key = decode_str (dec); 1407 SV *key = decode_str (dec);
1181 if (!key) 1408 if (!key)
1253 1480
1254 hv_iterinit (hv); 1481 hv_iterinit (hv);
1255 he = hv_iternext (hv); 1482 he = hv_iternext (hv);
1256 hv_iterinit (hv); 1483 hv_iterinit (hv);
1257 1484
1258 // the next line creates a mortal sv each time its called. 1485 // the next line creates a mortal sv each time it's called.
1259 // might want to optimise this for common cases. 1486 // might want to optimise this for common cases.
1260 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1487 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1261 1488
1262 if (cb) 1489 if (cb)
1263 { 1490 {
1264 dSP; 1491 dSP;
1265 int count; 1492 int count;
1266 1493
1267 ENTER; SAVETMPS; PUSHMARK (SP); 1494 ENTER; SAVETMPS;
1495 SAVESTACK_POS ();
1496 PUSHMARK (SP);
1268 XPUSHs (HeVAL (he)); 1497 XPUSHs (HeVAL (he));
1498 sv_2mortal (sv);
1269 1499
1270 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1500 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1271 1501
1272 if (count == 1) 1502 if (count == 1)
1273 { 1503 {
1274 sv = newSVsv (POPs); 1504 sv = newSVsv (POPs);
1275 FREETMPS; LEAVE; 1505 FREETMPS; LEAVE;
1276 return sv; 1506 return sv;
1277 } 1507 }
1278 1508
1509 SvREFCNT_inc (sv);
1279 FREETMPS; LEAVE; 1510 FREETMPS; LEAVE;
1280 } 1511 }
1281 } 1512 }
1282 1513
1283 if (dec->json.cb_object) 1514 if (dec->json.cb_object)
1284 { 1515 {
1285 dSP; 1516 dSP;
1286 int count; 1517 int count;
1287 1518
1288 ENTER; SAVETMPS; PUSHMARK (SP); 1519 ENTER; SAVETMPS;
1520 SAVESTACK_POS ();
1521 PUSHMARK (SP);
1289 XPUSHs (sv_2mortal (sv)); 1522 XPUSHs (sv_2mortal (sv));
1290 1523
1291 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1524 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1292 1525
1293 if (count == 1) 1526 if (count == 1)
1309 DEC_DEC_DEPTH; 1542 DEC_DEC_DEPTH;
1310 return 0; 1543 return 0;
1311} 1544}
1312 1545
1313static SV * 1546static SV *
1547decode_tag (dec_t *dec)
1548{
1549 SV *tag = 0;
1550 SV *val = 0;
1551
1552 if (!(dec->json.flags & F_ALLOW_TAGS))
1553 ERR ("malformed JSON string, neither array, object, number, string or atom");
1554
1555 ++dec->cur;
1556
1557 decode_ws (dec);
1558
1559 tag = decode_sv (dec);
1560 if (!tag)
1561 goto fail;
1562
1563 if (!SvPOK (tag))
1564 ERR ("malformed JSON string, (tag) must be a string");
1565
1566 decode_ws (dec);
1567
1568 if (*dec->cur != ')')
1569 ERR (") expected after tag");
1570
1571 ++dec->cur;
1572
1573 decode_ws (dec);
1574
1575 val = decode_sv (dec);
1576 if (!val)
1577 goto fail;
1578
1579 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1580 ERR ("malformed JSON string, tag value must be an array");
1581
1582 {
1583 AV *av = (AV *)SvRV (val);
1584 int i, len = av_len (av) + 1;
1585 HV *stash = gv_stashsv (tag, 0);
1586 SV *sv;
1587
1588 if (!stash)
1589 ERR ("cannot decode perl-object (package does not exist)");
1590
1591 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1592
1593 if (!method)
1594 ERR ("cannot decode perl-object (package does not have a THAW method)");
1595
1596 dSP;
1597
1598 ENTER; SAVETMPS;
1599 PUSHMARK (SP);
1600 EXTEND (SP, len + 2);
1601 // we re-bless the reference to get overload and other niceties right
1602 PUSHs (tag);
1603 PUSHs (sv_json);
1604
1605 for (i = 0; i < len; ++i)
1606 PUSHs (*av_fetch (av, i, 1));
1607
1608 PUTBACK;
1609 call_sv ((SV *)GvCV (method), G_SCALAR);
1610 SPAGAIN;
1611
1612 SvREFCNT_dec (tag);
1613 SvREFCNT_dec (val);
1614 sv = SvREFCNT_inc (POPs);
1615
1616 PUTBACK;
1617
1618 FREETMPS; LEAVE;
1619
1620 return sv;
1621 }
1622
1623fail:
1624 SvREFCNT_dec (tag);
1625 SvREFCNT_dec (val);
1626 return 0;
1627}
1628
1629static SV *
1314decode_sv (dec_t *dec) 1630decode_sv (dec_t *dec)
1315{ 1631{
1316 // the beauty of JSON: you need exactly one character lookahead 1632 // the beauty of JSON: you need exactly one character lookahead
1317 // to parse everything. 1633 // to parse everything.
1318 switch (*dec->cur) 1634 switch (*dec->cur)
1319 { 1635 {
1320 case '"': ++dec->cur; return decode_str (dec); 1636 case '"': ++dec->cur; return decode_str (dec);
1321 case '[': ++dec->cur; return decode_av (dec); 1637 case '[': ++dec->cur; return decode_av (dec);
1322 case '{': ++dec->cur; return decode_hv (dec); 1638 case '{': ++dec->cur; return decode_hv (dec);
1639 case '(': return decode_tag (dec);
1323 1640
1324 case '-': 1641 case '-':
1325 case '0': case '1': case '2': case '3': case '4': 1642 case '0': case '1': case '2': case '3': case '4':
1326 case '5': case '6': case '7': case '8': case '9': 1643 case '5': case '6': case '7': case '8': case '9':
1327 return decode_num (dec); 1644 return decode_num (dec);
1329 case 't': 1646 case 't':
1330 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1647 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1331 { 1648 {
1332 dec->cur += 4; 1649 dec->cur += 4;
1333#if JSON_SLOW 1650#if JSON_SLOW
1334 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1651 bool_true = get_bool ("Types::Serialiser::true");
1335#endif 1652#endif
1336 return SvREFCNT_inc (json_true); 1653 return newSVsv (bool_true);
1337 } 1654 }
1338 else 1655 else
1339 ERR ("'true' expected"); 1656 ERR ("'true' expected");
1340 1657
1341 break; 1658 break;
1343 case 'f': 1660 case 'f':
1344 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1661 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1345 { 1662 {
1346 dec->cur += 5; 1663 dec->cur += 5;
1347#if JSON_SLOW 1664#if JSON_SLOW
1348 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1665 bool_false = get_bool ("Types::Serialiser::false");
1349#endif 1666#endif
1350 return SvREFCNT_inc (json_false); 1667 return newSVsv (bool_false);
1351 } 1668 }
1352 else 1669 else
1353 ERR ("'false' expected"); 1670 ERR ("'false' expected");
1354 1671
1355 break; 1672 break;
1364 ERR ("'null' expected"); 1681 ERR ("'null' expected");
1365 1682
1366 break; 1683 break;
1367 1684
1368 default: 1685 default:
1369 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1686 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1370 break; 1687 break;
1371 } 1688 }
1372 1689
1373fail: 1690fail:
1374 return 0; 1691 return 0;
1375} 1692}
1376 1693
1377static SV * 1694static SV *
1378decode_json (SV *string, JSON *json, UV *offset_return) 1695decode_json (SV *string, JSON *json, STRLEN *offset_return)
1379{ 1696{
1380 dec_t dec; 1697 dec_t dec;
1381 UV offset;
1382 SV *sv; 1698 SV *sv;
1383 1699
1700 /* work around bugs in 5.10 where manipulating magic values
1701 * makes perl ignore the magic in subsequent accesses.
1702 * also make a copy of non-PV values, to get them into a clean
1703 * state (SvPV should do that, but it's buggy, see below).
1704 *
1705 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1706 * as reported by Reini Urban.
1707 */
1384 SvGETMAGIC (string); 1708 /*SvGETMAGIC (string);*/
1709 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1710 string = sv_2mortal (newSVsv (string));
1711
1385 SvUPGRADE (string, SVt_PV); 1712 SvUPGRADE (string, SVt_PV);
1386 1713
1387 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1714 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1715 * assertion with -DDEBUGGING, although SvCUR is documented to
1716 * return the xpv_cur field which certainly exists after upgrading.
1717 * according to nicholas clark, calling SvPOK fixes this.
1718 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1719 * and hope for the best.
1720 * Damnit, SvPV_nolen still trips over yet another assertion. This
1721 * assertion business is seriously broken, try yet another workaround
1722 * for the broken -DDEBUGGING.
1723 */
1724 {
1725#ifdef DEBUGGING
1726 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1727#else
1728 STRLEN offset = SvCUR (string);
1729#endif
1730
1731 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", 1732 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)); 1733 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1734 }
1390 1735
1391 if (json->flags & F_UTF8) 1736 if (DECODE_WANTS_OCTETS (json))
1392 sv_utf8_downgrade (string, 0); 1737 sv_utf8_downgrade (string, 0);
1393 else 1738 else
1394 sv_utf8_upgrade (string); 1739 sv_utf8_upgrade (string);
1395 1740
1396 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1741 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1397 1742
1398 dec.json = *json; 1743 dec.json = *json;
1399 dec.cur = SvPVX (string); 1744 dec.cur = SvPVX (string);
1400 dec.end = SvEND (string); 1745 dec.end = SvEND (string);
1401 dec.err = 0; 1746 dec.err = 0;
1402 dec.depth = 0; 1747 dec.depth = 0;
1403 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1404 1748
1405 if (dec.json.cb_object || dec.json.cb_sk_object) 1749 if (dec.json.cb_object || dec.json.cb_sk_object)
1406 dec.json.flags |= F_HOOK; 1750 dec.json.flags |= F_HOOK;
1407 1751
1408 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1752 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1409 1753
1410 decode_ws (&dec); 1754 decode_ws (&dec);
1411 sv = decode_sv (&dec); 1755 sv = decode_sv (&dec);
1412 1756
1413 if (!(offset_return || !sv)) 1757 if (offset_return)
1758 *offset_return = dec.cur - SvPVX (string);
1759 else if (sv)
1414 { 1760 {
1415 // check for trailing garbage 1761 // check for trailing garbage
1416 decode_ws (&dec); 1762 decode_ws (&dec);
1417 1763
1418 if (*dec.cur) 1764 if (*dec.cur)
1419 { 1765 {
1420 dec.err = "garbage after JSON object"; 1766 dec.err = "garbage after JSON object";
1421 SvREFCNT_dec (sv); 1767 SvREFCNT_dec (sv);
1422 sv = 0; 1768 sv = 0;
1423 } 1769 }
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 } 1770 }
1435 1771
1436 if (!sv) 1772 if (!sv)
1437 { 1773 {
1438 SV *uni = sv_newmortal (); 1774 SV *uni = sv_newmortal ();
1444 SAVEVPTR (PL_curcop); 1780 SAVEVPTR (PL_curcop);
1445 PL_curcop = &cop; 1781 PL_curcop = &cop;
1446 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1782 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1447 LEAVE; 1783 LEAVE;
1448 1784
1449 croak ("%s, at character offset %d [\"%s\"]", 1785 croak ("%s, at character offset %d (before \"%s\")",
1450 dec.err, 1786 dec.err,
1451 (int)offset, 1787 (int)ptr_to_index (string, dec.cur),
1452 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1788 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1453 } 1789 }
1454 1790
1455 sv = sv_2mortal (sv); 1791 sv = sv_2mortal (sv);
1456 1792
1457 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1793 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (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)"); 1794 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1459 1795
1460 return sv; 1796 return sv;
1797}
1798
1799/////////////////////////////////////////////////////////////////////////////
1800// incremental parser
1801
1802static void
1803incr_parse (JSON *self)
1804{
1805 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1806
1807 // the state machine here is a bit convoluted and could be simplified a lot
1808 // but this would make it slower, so...
1809
1810 for (;;)
1811 {
1812 //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
1813 switch (self->incr_mode)
1814 {
1815 // only used for initial whitespace skipping
1816 case INCR_M_WS:
1817 for (;;)
1818 {
1819 if (*p > 0x20)
1820 {
1821 if (*p == '#')
1822 {
1823 self->incr_mode = INCR_M_C0;
1824 goto incr_m_c;
1825 }
1826 else
1827 {
1828 self->incr_mode = INCR_M_JSON;
1829 goto incr_m_json;
1830 }
1831 }
1832 else if (!*p)
1833 goto interrupt;
1834
1835 ++p;
1836 }
1837
1838 // skip a single char inside a string (for \\-processing)
1839 case INCR_M_BS:
1840 if (!*p)
1841 goto interrupt;
1842
1843 ++p;
1844 self->incr_mode = INCR_M_STR;
1845 goto incr_m_str;
1846
1847 // inside #-style comments
1848 case INCR_M_C0:
1849 case INCR_M_C1:
1850 incr_m_c:
1851 for (;;)
1852 {
1853 if (*p == '\n')
1854 {
1855 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1856 break;
1857 }
1858 else if (!*p)
1859 goto interrupt;
1860
1861 ++p;
1862 }
1863
1864 break;
1865
1866 // inside a string
1867 case INCR_M_STR:
1868 incr_m_str:
1869 for (;;)
1870 {
1871 if (*p == '"')
1872 {
1873 ++p;
1874 self->incr_mode = INCR_M_JSON;
1875
1876 if (!self->incr_nest)
1877 goto interrupt;
1878
1879 goto incr_m_json;
1880 }
1881 else if (*p == '\\')
1882 {
1883 ++p; // "virtually" consumes character after \
1884
1885 if (!*p) // if at end of string we have to switch modes
1886 {
1887 self->incr_mode = INCR_M_BS;
1888 goto interrupt;
1889 }
1890 }
1891 else if (!*p)
1892 goto interrupt;
1893
1894 ++p;
1895 }
1896
1897 // after initial ws, outside string
1898 case INCR_M_JSON:
1899 incr_m_json:
1900 for (;;)
1901 {
1902 switch (*p++)
1903 {
1904 case 0:
1905 --p;
1906 goto interrupt;
1907
1908 case 0x09:
1909 case 0x0a:
1910 case 0x0d:
1911 case 0x20:
1912 if (!self->incr_nest)
1913 {
1914 --p; // do not eat the whitespace, let the next round do it
1915 goto interrupt;
1916 }
1917 break;
1918
1919 case '"':
1920 self->incr_mode = INCR_M_STR;
1921 goto incr_m_str;
1922
1923 case '[':
1924 case '{':
1925 case '(':
1926 if (++self->incr_nest > self->max_depth)
1927 croak (ERR_NESTING_EXCEEDED);
1928 break;
1929
1930 case ']':
1931 case '}':
1932 if (--self->incr_nest <= 0)
1933 goto interrupt;
1934 break;
1935
1936 case ')':
1937 --self->incr_nest;
1938 break;
1939
1940 case '#':
1941 self->incr_mode = INCR_M_C1;
1942 goto incr_m_c;
1943 }
1944 }
1945 }
1946
1947 modechange:
1948 ;
1949 }
1950
1951interrupt:
1952 self->incr_pos = p - SvPVX (self->incr_text);
1953 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1954 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1461} 1955}
1462 1956
1463///////////////////////////////////////////////////////////////////////////// 1957/////////////////////////////////////////////////////////////////////////////
1464// XS interface functions 1958// XS interface functions
1465 1959
1474 i >= '0' && i <= '9' ? i - '0' 1968 i >= '0' && i <= '9' ? i - '0'
1475 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1969 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1476 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1970 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1477 : -1; 1971 : -1;
1478 1972
1479 json_stash = gv_stashpv ("JSON::XS" , 1); 1973 json_stash = gv_stashpv ("JSON::XS" , 1);
1480 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1974 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1975 bool_true = get_bool ("Types::Serialiser::true");
1976 bool_false = get_bool ("Types::Serialiser::false");
1481 1977
1482 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1978 sv_json = newSVpv ("JSON", 0);
1483 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1979 SvREADONLY_on (sv_json);
1980
1981 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1484} 1982}
1485 1983
1486PROTOTYPES: DISABLE 1984PROTOTYPES: DISABLE
1487 1985
1488void CLONE (...) 1986void CLONE (...)
1489 CODE: 1987 CODE:
1490 json_stash = 0; 1988 json_stash = 0;
1491 json_boolean_stash = 0; 1989 bool_stash = 0;
1492 1990
1493void new (char *klass) 1991void new (char *klass)
1494 PPCODE: 1992 PPCODE:
1495{ 1993{
1496 SV *pv = NEWSV (0, sizeof (JSON)); 1994 SV *pv = NEWSV (0, sizeof (JSON));
1497 SvPOK_only (pv); 1995 SvPOK_only (pv);
1498 Zero (SvPVX (pv), 1, JSON); 1996 json_init ((JSON *)SvPVX (pv));
1499 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1500 XPUSHs (sv_2mortal (sv_bless ( 1997 XPUSHs (sv_2mortal (sv_bless (
1501 newRV_noinc (pv), 1998 newRV_noinc (pv),
1502 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1999 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1503 ))); 2000 )));
1504} 2001}
1516 allow_nonref = F_ALLOW_NONREF 2013 allow_nonref = F_ALLOW_NONREF
1517 shrink = F_SHRINK 2014 shrink = F_SHRINK
1518 allow_blessed = F_ALLOW_BLESSED 2015 allow_blessed = F_ALLOW_BLESSED
1519 convert_blessed = F_CONV_BLESSED 2016 convert_blessed = F_CONV_BLESSED
1520 relaxed = F_RELAXED 2017 relaxed = F_RELAXED
2018 allow_unknown = F_ALLOW_UNKNOWN
2019 allow_tags = F_ALLOW_TAGS
1521 PPCODE: 2020 PPCODE:
1522{ 2021{
1523 if (enable) 2022 if (enable)
1524 self->flags |= ix; 2023 self->flags |= ix;
1525 else 2024 else
1540 get_allow_nonref = F_ALLOW_NONREF 2039 get_allow_nonref = F_ALLOW_NONREF
1541 get_shrink = F_SHRINK 2040 get_shrink = F_SHRINK
1542 get_allow_blessed = F_ALLOW_BLESSED 2041 get_allow_blessed = F_ALLOW_BLESSED
1543 get_convert_blessed = F_CONV_BLESSED 2042 get_convert_blessed = F_CONV_BLESSED
1544 get_relaxed = F_RELAXED 2043 get_relaxed = F_RELAXED
2044 get_allow_unknown = F_ALLOW_UNKNOWN
2045 get_allow_tags = F_ALLOW_TAGS
1545 PPCODE: 2046 PPCODE:
1546 XPUSHs (boolSV (self->flags & ix)); 2047 XPUSHs (boolSV (self->flags & ix));
1547 2048
1548void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2049void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1549 PPCODE: 2050 PPCODE:
1550{ 2051 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)); 2052 XPUSHs (ST (0));
1561}
1562 2053
1563U32 get_max_depth (JSON *self) 2054U32 get_max_depth (JSON *self)
1564 CODE: 2055 CODE:
1565 RETVAL = DEC_DEPTH (self->flags); 2056 RETVAL = self->max_depth;
1566 OUTPUT: 2057 OUTPUT:
1567 RETVAL 2058 RETVAL
1568 2059
1569void max_size (JSON *self, UV max_size = 0) 2060void max_size (JSON *self, U32 max_size = 0)
1570 PPCODE: 2061 PPCODE:
1571{ 2062 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)); 2063 XPUSHs (ST (0));
1583}
1584 2064
1585int get_max_size (JSON *self) 2065int get_max_size (JSON *self)
1586 CODE: 2066 CODE:
1587 RETVAL = DEC_SIZE (self->flags); 2067 RETVAL = self->max_size;
1588 OUTPUT: 2068 OUTPUT:
1589 RETVAL 2069 RETVAL
1590 2070
1591void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2071void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1592 PPCODE: 2072 PPCODE:
1598} 2078}
1599 2079
1600void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2080void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1601 PPCODE: 2081 PPCODE:
1602{ 2082{
1603 if (!self->cb_sk_object) 2083 if (!self->cb_sk_object)
1604 self->cb_sk_object = newHV (); 2084 self->cb_sk_object = newHV ();
1605 2085
1606 if (SvOK (cb)) 2086 if (SvOK (cb))
1607 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2087 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1608 else 2088 else
1619 XPUSHs (ST (0)); 2099 XPUSHs (ST (0));
1620} 2100}
1621 2101
1622void encode (JSON *self, SV *scalar) 2102void encode (JSON *self, SV *scalar)
1623 PPCODE: 2103 PPCODE:
1624 XPUSHs (encode_json (scalar, self)); 2104 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2105 XPUSHs (scalar);
1625 2106
1626void decode (JSON *self, SV *jsonstr) 2107void decode (JSON *self, SV *jsonstr)
1627 PPCODE: 2108 PPCODE:
1628 XPUSHs (decode_json (jsonstr, self, 0)); 2109 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2110 XPUSHs (jsonstr);
1629 2111
1630void decode_prefix (JSON *self, SV *jsonstr) 2112void decode_prefix (JSON *self, SV *jsonstr)
1631 PPCODE: 2113 PPCODE:
1632{ 2114{
1633 UV offset; 2115 SV *sv;
2116 STRLEN offset;
2117 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1634 EXTEND (SP, 2); 2118 EXTEND (SP, 2);
1635 PUSHs (decode_json (jsonstr, self, &offset)); 2119 PUSHs (sv);
1636 PUSHs (sv_2mortal (newSVuv (offset))); 2120 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
2121}
2122
2123void incr_parse (JSON *self, SV *jsonstr = 0)
2124 PPCODE:
2125{
2126 if (!self->incr_text)
2127 self->incr_text = newSVpvn ("", 0);
2128
2129 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2130 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2131 if (DECODE_WANTS_OCTETS (self))
2132 {
2133 if (self->incr_pos)
2134 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2135 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2136
2137 sv_utf8_downgrade (self->incr_text, 0);
2138 }
2139 else
2140 {
2141 sv_utf8_upgrade (self->incr_text);
2142
2143 if (self->incr_pos)
2144 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2145 - (U8 *)SvPVX (self->incr_text);
2146 }
2147
2148 // append data, if any
2149 if (jsonstr)
2150 {
2151 /* make sure both strings have same encoding */
2152 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2153 if (SvUTF8 (jsonstr))
2154 sv_utf8_downgrade (jsonstr, 0);
2155 else
2156 sv_utf8_upgrade (jsonstr);
2157
2158 /* and then just blindly append */
2159 {
2160 STRLEN len;
2161 const char *str = SvPV (jsonstr, len);
2162 STRLEN cur = SvCUR (self->incr_text);
2163
2164 if (SvLEN (self->incr_text) <= cur + len)
2165 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2166
2167 Move (str, SvEND (self->incr_text), len, char);
2168 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2169 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2170 }
2171 }
2172
2173 if (GIMME_V != G_VOID)
2174 do
2175 {
2176 SV *sv;
2177 STRLEN offset;
2178
2179 if (!INCR_DONE (self))
2180 {
2181 incr_parse (self);
2182
2183 if (self->incr_pos > self->max_size && self->max_size)
2184 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2185 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2186
2187 if (!INCR_DONE (self))
2188 {
2189 // as an optimisation, do not accumulate white space in the incr buffer
2190 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2191 {
2192 self->incr_pos = 0;
2193 SvCUR_set (self->incr_text, 0);
2194 }
2195
2196 break;
2197 }
2198 }
2199
2200 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2201 XPUSHs (sv);
2202
2203 self->incr_pos -= offset;
2204 self->incr_nest = 0;
2205 self->incr_mode = 0;
2206
2207 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
2208 }
2209 while (GIMME_V == G_ARRAY);
2210}
2211
2212SV *incr_text (JSON *self)
2213 ATTRS: lvalue
2214 CODE:
2215{
2216 if (self->incr_pos)
2217 croak ("incr_text can not be called when the incremental parser already started parsing");
2218
2219 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2220}
2221 OUTPUT:
2222 RETVAL
2223
2224void incr_skip (JSON *self)
2225 CODE:
2226{
2227 if (self->incr_pos)
2228 {
2229 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2230 self->incr_pos = 0;
2231 self->incr_nest = 0;
2232 self->incr_mode = 0;
2233 }
2234}
2235
2236void incr_reset (JSON *self)
2237 CODE:
2238{
2239 SvREFCNT_dec (self->incr_text);
2240 self->incr_text = 0;
2241 self->incr_pos = 0;
2242 self->incr_nest = 0;
2243 self->incr_mode = 0;
1637} 2244}
1638 2245
1639void DESTROY (JSON *self) 2246void DESTROY (JSON *self)
1640 CODE: 2247 CODE:
1641 SvREFCNT_dec (self->cb_sk_object); 2248 SvREFCNT_dec (self->cb_sk_object);
1642 SvREFCNT_dec (self->cb_object); 2249 SvREFCNT_dec (self->cb_object);
2250 SvREFCNT_dec (self->incr_text);
1643 2251
1644PROTOTYPES: ENABLE 2252PROTOTYPES: ENABLE
1645 2253
1646void encode_json (SV *scalar) 2254void encode_json (SV *scalar)
1647 PPCODE: 2255 PPCODE:
1648{ 2256{
1649 JSON json = { F_DEFAULT | F_UTF8 }; 2257 JSON json;
1650 XPUSHs (encode_json (scalar, &json)); 2258 json_init (&json);
2259 json.flags |= F_UTF8;
2260 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2261 XPUSHs (scalar);
1651} 2262}
1652 2263
1653void decode_json (SV *jsonstr) 2264void decode_json (SV *jsonstr)
1654 PPCODE: 2265 PPCODE:
1655{ 2266{
1656 JSON json = { F_DEFAULT | F_UTF8 }; 2267 JSON json;
1657 XPUSHs (decode_json (jsonstr, &json, 0)); 2268 json_init (&json);
2269 json.flags |= F_UTF8;
2270 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2271 XPUSHs (jsonstr);
1658} 2272}
1659 2273

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines