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 1.127 by root, Fri Feb 26 21:46:45 2016 UTC

12#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 14#endif
15 15
16// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
20#endif 20#endif
21 21
22// compatibility with perl <5.18
23#ifndef HvNAMELEN_get
24# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
25#endif
26#ifndef HvNAMELEN
27# define HvNAMELEN(hv) HvNAMELEN_get (hv)
28#endif
29#ifndef HvNAMEUTF8
30# define HvNAMEUTF8(hv) 0
31#endif
32
33// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 34#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 35
24#define F_ASCII 0x00000001UL 36#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 37#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 38#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 39#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 43#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL 44#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 45#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 46#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 47#define F_RELAXED 0x00001000UL
36 48#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL 49#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 50#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 51
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 52#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 53
49#define INIT_SIZE 32 // initial scalar size to be allocated 54#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 55#define INDENT_STEP 3 // spaces per indentation level
51 56
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 57#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
58
59#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
53 60
54#define SB do { 61#define SB do {
55#define SE } while (0) 62#define SE } while (0)
56 63
57#if __GNUC__ >= 3 64#if __GNUC__ >= 3
67 74
68#define IN_RANGE_INC(type,val,beg,end) \ 75#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 76 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 77 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 78
79#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
80
72#ifdef USE_ITHREADS 81#ifdef USE_ITHREADS
73# define JSON_SLOW 1 82# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
84# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
75#else 85#else
76# define JSON_SLOW 0 86# define JSON_SLOW 0
77# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
78#endif 89#endif
79 90
80static HV *json_stash, *json_boolean_stash; // JSON::XS:: 91// the amount of HEs to allocate on the stack, when sorting keys
81static SV *json_true, *json_false; 92#define STACK_HES 64
93
94static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
95static SV *bool_true, *bool_false, *sv_json;
96
97enum {
98 INCR_M_WS = 0, // initial whitespace skipping, must be 0
99 INCR_M_STR, // inside string
100 INCR_M_BS, // inside backslash
101 INCR_M_C0, // inside comment in initial whitespace sequence
102 INCR_M_C1, // inside comment in other places
103 INCR_M_JSON // outside anything, count nesting
104};
105
106#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
82 107
83typedef struct { 108typedef struct {
84 U32 flags; 109 U32 flags;
110 U32 max_depth;
111 STRLEN max_size;
112
85 SV *cb_object; 113 SV *cb_object;
86 HV *cb_sk_object; 114 HV *cb_sk_object;
115
116 // for the incremental parser
117 SV *incr_text; // the source text so far
118 STRLEN incr_pos; // the current offset into the text
119 int incr_nest; // {[]}-nesting level
120 unsigned char incr_mode;
87} JSON; 121} JSON;
122
123INLINE void
124json_init (JSON *json)
125{
126 Zero (json, 1, JSON);
127 json->max_depth = 512;
128}
88 129
89///////////////////////////////////////////////////////////////////////////// 130/////////////////////////////////////////////////////////////////////////////
90// utility functions 131// utility functions
91 132
133INLINE SV *
134get_bool (const char *name)
135{
136 SV *sv = get_sv (name, 1);
137
138 SvREADONLY_on (sv);
139 SvREADONLY_on (SvRV (sv));
140
141 return sv;
142}
143
92INLINE void 144INLINE void
93shrink (SV *sv) 145shrink (SV *sv)
94{ 146{
95 sv_utf8_downgrade (sv, 1); 147 sv_utf8_downgrade (sv, 1);
148
96 if (SvLEN (sv) > SvCUR (sv) + 1) 149 if (SvLEN (sv) > SvCUR (sv) + 1)
97 { 150 {
98#ifdef SvPV_shrink_to_cur 151#ifdef SvPV_shrink_to_cur
99 SvPV_shrink_to_cur (sv); 152 SvPV_shrink_to_cur (sv);
100#elif defined (SvPV_renew) 153#elif defined (SvPV_renew)
144 *s++ = 0x80 | ( ch & 0x3f); 197 *s++ = 0x80 | ( ch & 0x3f);
145 198
146 return s; 199 return s;
147} 200}
148 201
202// convert offset pointer to character index, sv must be string
203static STRLEN
204ptr_to_index (SV *sv, char *offset)
205{
206 return SvUTF8 (sv)
207 ? utf8_distance (offset, SvPVX (sv))
208 : offset - SvPVX (sv);
209}
210
211/////////////////////////////////////////////////////////////////////////////
212// fp hell
213
214// scan a group of digits, and a trailing exponent
215static void
216json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
217{
218 UV uaccum = 0;
219 int eaccum = 0;
220
221 // if we recurse too deep, skip all remaining digits
222 // to avoid a stack overflow attack
223 if (expect_false (--maxdepth <= 0))
224 while (((U8)*s - '0') < 10)
225 ++s;
226
227 for (;;)
228 {
229 U8 dig = (U8)*s - '0';
230
231 if (expect_false (dig >= 10))
232 {
233 if (dig == (U8)((U8)'.' - (U8)'0'))
234 {
235 ++s;
236 json_atof_scan1 (s, accum, expo, 1, maxdepth);
237 }
238 else if ((dig | ' ') == 'e' - '0')
239 {
240 int exp2 = 0;
241 int neg = 0;
242
243 ++s;
244
245 if (*s == '-')
246 {
247 ++s;
248 neg = 1;
249 }
250 else if (*s == '+')
251 ++s;
252
253 while ((dig = (U8)*s - '0') < 10)
254 exp2 = exp2 * 10 + *s++ - '0';
255
256 *expo += neg ? -exp2 : exp2;
257 }
258
259 break;
260 }
261
262 ++s;
263
264 uaccum = uaccum * 10 + dig;
265 ++eaccum;
266
267 // if we have too many digits, then recurse for more
268 // we actually do this for rather few digits
269 if (uaccum >= (UV_MAX - 9) / 10)
270 {
271 if (postdp) *expo -= eaccum;
272 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
273 if (postdp) *expo += eaccum;
274
275 break;
276 }
277 }
278
279 // this relies greatly on the quality of the pow ()
280 // implementation of the platform, but a good
281 // implementation is hard to beat.
282 // (IEEE 754 conformant ones are required to be exact)
283 if (postdp) *expo -= eaccum;
284 *accum += uaccum * Perl_pow (10., *expo);
285 *expo += eaccum;
286}
287
288static NV
289json_atof (const char *s)
290{
291 NV accum = 0.;
292 int expo = 0;
293 int neg = 0;
294
295 if (*s == '-')
296 {
297 ++s;
298 neg = 1;
299 }
300
301 // a recursion depth of ten gives us >>500 bits
302 json_atof_scan1 (s, &accum, &expo, 0, 10);
303
304 return neg ? -accum : accum;
305}
306
307// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
308static int
309ref_bool_type (SV *sv)
310{
311 svtype svt = SvTYPE (sv);
312
313 if (svt < SVt_PVAV)
314 {
315 STRLEN len = 0;
316 char *pv = svt ? SvPV (sv, len) : 0;
317
318 if (len == 1)
319 if (*pv == '1')
320 return 1;
321 else if (*pv == '0')
322 return 0;
323
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 (SvSTASH (scalar) == bool_stash)
339 return 1;
340
341 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
342 return 1;
343
344 return 0;
345}
346
149///////////////////////////////////////////////////////////////////////////// 347/////////////////////////////////////////////////////////////////////////////
150// encoder 348// encoder
151 349
152// structure used for encoding JSON 350// structure used for encoding JSON
153typedef struct 351typedef struct
155 char *cur; // SvPVX (sv) + current output position 353 char *cur; // SvPVX (sv) + current output position
156 char *end; // SvEND (sv) 354 char *end; // SvEND (sv)
157 SV *sv; // result scalar 355 SV *sv; // result scalar
158 JSON json; 356 JSON json;
159 U32 indent; // indentation level 357 U32 indent; // indentation level
160 U32 maxdepth; // max. indentation/recursion level
161 UV limit; // escape character values >= this value when encoding 358 UV limit; // escape character values >= this value when encoding
162} enc_t; 359} enc_t;
163 360
164INLINE void 361INLINE void
165need (enc_t *enc, STRLEN len) 362need (enc_t *enc, STRLEN len)
166{ 363{
167 if (expect_false (enc->cur + len >= enc->end)) 364 if (expect_false (enc->cur + len >= enc->end))
168 { 365 {
169 STRLEN cur = enc->cur - SvPVX (enc->sv); 366 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
170 SvGROW (enc->sv, cur + len + 1); 367 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
171 enc->cur = SvPVX (enc->sv) + cur; 368 enc->cur = SvPVX (enc->sv) + cur;
172 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 369 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
173 } 370 }
174} 371}
175 372
250 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 447 (int)((uch - 0x10000) % 0x400 + 0xDC00));
251 enc->cur += 12; 448 enc->cur += 12;
252 } 449 }
253 else 450 else
254 { 451 {
255 static char hexdigit [16] = "0123456789abcdef";
256 need (enc, len += 5); 452 need (enc, len += 5);
257 *enc->cur++ = '\\'; 453 *enc->cur++ = '\\';
258 *enc->cur++ = 'u'; 454 *enc->cur++ = 'u';
259 *enc->cur++ = hexdigit [ uch >> 12 ]; 455 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
260 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 456 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
261 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 457 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
262 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 458 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
263 } 459 }
264 460
265 str += clen; 461 str += clen;
266 } 462 }
267 else if (enc->json.flags & F_LATIN1) 463 else if (enc->json.flags & F_LATIN1)
338static void 534static void
339encode_av (enc_t *enc, AV *av) 535encode_av (enc_t *enc, AV *av)
340{ 536{
341 int i, len = av_len (av); 537 int i, len = av_len (av);
342 538
343 if (enc->indent >= enc->maxdepth) 539 if (enc->indent >= enc->json.max_depth)
344 croak ("data structure too deep (hit recursion limit)"); 540 croak (ERR_NESTING_EXCEEDED);
345 541
346 encode_ch (enc, '['); 542 encode_ch (enc, '[');
347 543
348 if (len >= 0) 544 if (len >= 0)
349 { 545 {
350 encode_nl (enc); ++enc->indent; 546 encode_nl (enc); ++enc->indent;
351 547
352 for (i = 0; i <= len; ++i) 548 for (i = 0; i <= len; ++i)
364 encode_comma (enc); 560 encode_comma (enc);
365 } 561 }
366 562
367 encode_nl (enc); --enc->indent; encode_indent (enc); 563 encode_nl (enc); --enc->indent; encode_indent (enc);
368 } 564 }
369 565
370 encode_ch (enc, ']'); 566 encode_ch (enc, ']');
371} 567}
372 568
373static void 569static void
374encode_hk (enc_t *enc, HE *he) 570encode_hk (enc_t *enc, HE *he)
378 if (HeKLEN (he) == HEf_SVKEY) 574 if (HeKLEN (he) == HEf_SVKEY)
379 { 575 {
380 SV *sv = HeSVKEY (he); 576 SV *sv = HeSVKEY (he);
381 STRLEN len; 577 STRLEN len;
382 char *str; 578 char *str;
383 579
384 SvGETMAGIC (sv); 580 SvGETMAGIC (sv);
385 str = SvPV (sv, len); 581 str = SvPV (sv, len);
386 582
387 encode_str (enc, str, len, SvUTF8 (sv)); 583 encode_str (enc, str, len, SvUTF8 (sv));
388 } 584 }
423 619
424static void 620static void
425encode_hv (enc_t *enc, HV *hv) 621encode_hv (enc_t *enc, HV *hv)
426{ 622{
427 HE *he; 623 HE *he;
428 int count;
429 624
430 if (enc->indent >= enc->maxdepth) 625 if (enc->indent >= enc->json.max_depth)
431 croak ("data structure too deep (hit recursion limit)"); 626 croak (ERR_NESTING_EXCEEDED);
432 627
433 encode_ch (enc, '{'); 628 encode_ch (enc, '{');
434 629
435 // for canonical output we have to sort by keys first 630 // for canonical output we have to sort by keys first
436 // actually, this is mostly due to the stupid so-called 631 // actually, this is mostly due to the stupid so-called
437 // security workaround added somewhere in 5.8.x. 632 // security workaround added somewhere in 5.8.x
438 // that randomises hash orderings 633 // that randomises hash orderings
439 if (enc->json.flags & F_CANONICAL) 634 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
440 { 635 {
441 int count = hv_iterinit (hv); 636 int count = hv_iterinit (hv);
442 637
443 if (SvMAGICAL (hv)) 638 if (SvMAGICAL (hv))
444 { 639 {
455 } 650 }
456 651
457 if (count) 652 if (count)
458 { 653 {
459 int i, fast = 1; 654 int i, fast = 1;
460#if defined(__BORLANDC__) || defined(_MSC_VER) 655 HE *hes_stack [STACK_HES];
461 HE **hes = _alloca (count * sizeof (HE)); 656 HE **hes = hes_stack;
462#else 657
463 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 658 // allocate larger arrays on the heap
464#endif 659 if (count > STACK_HES)
660 {
661 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
662 hes = (HE **)SvPVX (sv);
663 }
465 664
466 i = 0; 665 i = 0;
467 while ((he = hv_iternext (hv))) 666 while ((he = hv_iternext (hv)))
468 { 667 {
469 hes [i++] = he; 668 hes [i++] = he;
538// encode objects, arrays and special \0=false and \1=true values. 737// encode objects, arrays and special \0=false and \1=true values.
539static void 738static void
540encode_rv (enc_t *enc, SV *sv) 739encode_rv (enc_t *enc, SV *sv)
541{ 740{
542 svtype svt; 741 svtype svt;
742 GV *method;
543 743
544 SvGETMAGIC (sv); 744 SvGETMAGIC (sv);
545 svt = SvTYPE (sv); 745 svt = SvTYPE (sv);
546 746
547 if (expect_false (SvOBJECT (sv))) 747 if (expect_false (SvOBJECT (sv)))
548 { 748 {
549 HV *stash = !JSON_SLOW || json_boolean_stash 749 HV *stash = SvSTASH (sv);
550 ? json_boolean_stash
551 : gv_stashpv ("JSON::XS::Boolean", 1);
552 750
553 if (SvSTASH (sv) == stash) 751 if (stash == bool_stash)
554 { 752 {
555 if (SvIV (sv)) 753 if (SvIV (sv))
556 encode_str (enc, "true", 4, 0); 754 encode_str (enc, "true", 4, 0);
557 else 755 else
558 encode_str (enc, "false", 5, 0); 756 encode_str (enc, "false", 5, 0);
559 } 757 }
758 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
759 {
760 int count;
761 dSP;
762
763 ENTER; SAVETMPS;
764 SAVESTACK_POS ();
765 PUSHMARK (SP);
766 EXTEND (SP, 2);
767 // we re-bless the reference to get overload and other niceties right
768 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
769 PUSHs (sv_json);
770
771 PUTBACK;
772 count = call_sv ((SV *)GvCV (method), G_ARRAY);
773 SPAGAIN;
774
775 // catch this surprisingly common error
776 if (SvROK (TOPs) && SvRV (TOPs) == sv)
777 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
778
779 encode_ch (enc, '(');
780 encode_ch (enc, '"');
781 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
782 encode_ch (enc, '"');
783 encode_ch (enc, ')');
784 encode_ch (enc, '[');
785
786 while (count)
787 {
788 encode_sv (enc, SP[1 - count--]);
789
790 if (count)
791 encode_ch (enc, ',');
792 }
793
794 encode_ch (enc, ']');
795
796 FREETMPS; LEAVE;
797 }
798 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
799 {
800 dSP;
801
802 ENTER; SAVETMPS;
803 PUSHMARK (SP);
804 // we re-bless the reference to get overload and other niceties right
805 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
806
807 // calling with G_SCALAR ensures that we always get a 1 return value
808 PUTBACK;
809 call_sv ((SV *)GvCV (method), G_SCALAR);
810 SPAGAIN;
811
812 // catch this surprisingly common error
813 if (SvROK (TOPs) && SvRV (TOPs) == sv)
814 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
815
816 sv = POPs;
817 PUTBACK;
818
819 encode_sv (enc, sv);
820
821 FREETMPS; LEAVE;
822 }
823 else if (enc->json.flags & F_ALLOW_BLESSED)
824 encode_str (enc, "null", 4, 0);
560 else 825 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", 826 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)))); 827 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
607 }
608 } 828 }
609 else if (svt == SVt_PVHV) 829 else if (svt == SVt_PVHV)
610 encode_hv (enc, (HV *)sv); 830 encode_hv (enc, (HV *)sv);
611 else if (svt == SVt_PVAV) 831 else if (svt == SVt_PVAV)
612 encode_av (enc, (AV *)sv); 832 encode_av (enc, (AV *)sv);
613 else if (svt < SVt_PVAV) 833 else if (svt < SVt_PVAV)
614 { 834 {
615 STRLEN len = 0; 835 int bool_type = ref_bool_type (sv);
616 char *pv = svt ? SvPV (sv, len) : 0;
617 836
618 if (len == 1 && *pv == '1') 837 if (bool_type == 1)
619 encode_str (enc, "true", 4, 0); 838 encode_str (enc, "true", 4, 0);
620 else if (len == 1 && *pv == '0') 839 else if (bool_type == 0)
621 encode_str (enc, "false", 5, 0); 840 encode_str (enc, "false", 5, 0);
841 else if (enc->json.flags & F_ALLOW_UNKNOWN)
842 encode_str (enc, "null", 4, 0);
622 else 843 else
623 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 844 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
624 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 845 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
625 } 846 }
847 else if (enc->json.flags & F_ALLOW_UNKNOWN)
848 encode_str (enc, "null", 4, 0);
626 else 849 else
627 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 850 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
628 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 851 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
629} 852}
630 853
684 } 907 }
685 else 908 else
686 { 909 {
687 // large integer, use the (rather slow) snprintf way. 910 // large integer, use the (rather slow) snprintf way.
688 need (enc, IVUV_MAXCHARS); 911 need (enc, IVUV_MAXCHARS);
689 enc->cur += 912 enc->cur +=
690 SvIsUV(sv) 913 SvIsUV(sv)
691 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 914 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
692 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 915 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
693 } 916 }
694 } 917 }
695 else if (SvROK (sv)) 918 else if (SvROK (sv))
696 encode_rv (enc, SvRV (sv)); 919 encode_rv (enc, SvRV (sv));
697 else if (!SvOK (sv)) 920 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
698 encode_str (enc, "null", 4, 0); 921 encode_str (enc, "null", 4, 0);
699 else 922 else
700 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 923 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
701 SvPV_nolen (sv), SvFLAGS (sv)); 924 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
702} 925}
703 926
704static SV * 927static SV *
705encode_json (SV *scalar, JSON *json) 928encode_json (SV *scalar, JSON *json)
706{ 929{
707 enc_t enc; 930 enc_t enc;
708 931
709 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 932 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)"); 933 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
711 934
712 enc.json = *json; 935 enc.json = *json;
713 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 936 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
714 enc.cur = SvPVX (enc.sv); 937 enc.cur = SvPVX (enc.sv);
715 enc.end = SvEND (enc.sv); 938 enc.end = SvEND (enc.sv);
716 enc.indent = 0; 939 enc.indent = 0;
717 enc.maxdepth = DEC_DEPTH (enc.json.flags);
718 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 940 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
719 : enc.json.flags & F_LATIN1 ? 0x000100UL 941 : enc.json.flags & F_LATIN1 ? 0x000100UL
720 : 0x110000UL; 942 : 0x110000UL;
721 943
722 SvPOK_only (enc.sv); 944 SvPOK_only (enc.sv);
723 encode_sv (&enc, scalar); 945 encode_sv (&enc, scalar);
946 encode_nl (&enc);
724 947
725 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 948 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
726 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 949 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
727 950
728 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 951 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
789 if (*dec->cur != ch) \ 1012 if (*dec->cur != ch) \
790 ERR (# ch " expected"); \ 1013 ERR (# ch " expected"); \
791 ++dec->cur; \ 1014 ++dec->cur; \
792 SE 1015 SE
793 1016
794#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1017#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
795#define DEC_DEC_DEPTH --dec->depth 1018#define DEC_DEC_DEPTH --dec->depth
796 1019
797static SV *decode_sv (dec_t *dec); 1020static SV *decode_sv (dec_t *dec);
798 1021
799static signed char decode_hexdigit[256]; 1022static signed char decode_hexdigit[256];
908 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1131 else if (expect_true (ch >= 0x20 && ch < 0x80))
909 *cur++ = ch; 1132 *cur++ = ch;
910 else if (ch >= 0x80) 1133 else if (ch >= 0x80)
911 { 1134 {
912 STRLEN clen; 1135 STRLEN clen;
913 UV uch;
914 1136
915 --dec_cur; 1137 --dec_cur;
916 1138
917 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1139 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
918 if (clen == (STRLEN)-1) 1140 if (clen == (STRLEN)-1)
919 ERR ("malformed UTF-8 character in JSON string"); 1141 ERR ("malformed UTF-8 character in JSON string");
920 1142
921 do 1143 do
922 *cur++ = *dec_cur++; 1144 *cur++ = *dec_cur++;
923 while (--clen); 1145 while (--clen);
924 1146
925 utf8 = 1; 1147 utf8 = 1;
926 } 1148 }
1149 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1150 *cur++ = ch;
927 else 1151 else
928 { 1152 {
929 --dec_cur; 1153 --dec_cur;
930 1154
931 if (!ch) 1155 if (!ch)
939 { 1163 {
940 STRLEN len = cur - buf; 1164 STRLEN len = cur - buf;
941 1165
942 if (sv) 1166 if (sv)
943 { 1167 {
944 SvGROW (sv, SvCUR (sv) + len + 1); 1168 STRLEN cur = SvCUR (sv);
1169
1170 if (SvLEN (sv) <= cur + len)
1171 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1172
945 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1173 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
946 SvCUR_set (sv, SvCUR (sv) + len); 1174 SvCUR_set (sv, SvCUR (sv) + len);
947 } 1175 }
948 else 1176 else
949 sv = newSVpvn (buf, len); 1177 sv = newSVpvn (buf, len);
1040 1268
1041 // special case the rather common 1..5-digit-int case 1269 // special case the rather common 1..5-digit-int case
1042 if (*start == '-') 1270 if (*start == '-')
1043 switch (len) 1271 switch (len)
1044 { 1272 {
1045 case 2: return newSViv (-( start [1] - '0' * 1)); 1273 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1046 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1274 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)); 1275 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)); 1276 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)); 1277 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1050 } 1278 }
1051 else 1279 else
1052 switch (len) 1280 switch (len)
1053 { 1281 {
1054 case 1: return newSViv ( start [0] - '0' * 1); 1282 case 1: return newSViv ( start [0] - '0' * 1);
1055 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1283 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); 1284 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); 1285 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); 1286 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1059 } 1287 }
1060 1288
1061 { 1289 {
1062 UV uv; 1290 UV uv;
1063 int numtype = grok_number (start, len, &uv); 1291 int numtype = grok_number (start, len, &uv);
1072 } 1300 }
1073 1301
1074 len -= *start == '-' ? 1 : 0; 1302 len -= *start == '-' ? 1 : 0;
1075 1303
1076 // does not fit into IV or UV, try NV 1304 // does not fit into IV or UV, try NV
1077 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1305 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 1306 // fits into NV without loss of precision
1083 return newSVnv (Atof (start)); 1307 return newSVnv (json_atof (start));
1084 1308
1085 // everything else fails, convert it to a string 1309 // everything else fails, convert it to a string
1086 return newSVpvn (start, dec->cur - start); 1310 return newSVpvn (start, dec->cur - start);
1087 } 1311 }
1088 1312
1089 // loss of precision here 1313 // loss of precision here
1090 return newSVnv (Atof (start)); 1314 return newSVnv (json_atof (start));
1091 1315
1092fail: 1316fail:
1093 return 0; 1317 return 0;
1094} 1318}
1095 1319
1119 if (*dec->cur == ']') 1343 if (*dec->cur == ']')
1120 { 1344 {
1121 ++dec->cur; 1345 ++dec->cur;
1122 break; 1346 break;
1123 } 1347 }
1124 1348
1125 if (*dec->cur != ',') 1349 if (*dec->cur != ',')
1126 ERR (", or ] expected while parsing array"); 1350 ERR (", or ] expected while parsing array");
1127 1351
1128 ++dec->cur; 1352 ++dec->cur;
1129 1353
1171 char *p = dec->cur; 1395 char *p = dec->cur;
1172 char *e = p + 24; // only try up to 24 bytes 1396 char *e = p + 24; // only try up to 24 bytes
1173 1397
1174 for (;;) 1398 for (;;)
1175 { 1399 {
1176 // the >= 0x80 is true on most architectures 1400 // the >= 0x80 is false on most architectures
1177 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1401 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1178 { 1402 {
1179 // slow path, back up and use decode_str 1403 // slow path, back up and use decode_str
1180 SV *key = decode_str (dec); 1404 SV *key = decode_str (dec);
1181 if (!key) 1405 if (!key)
1253 1477
1254 hv_iterinit (hv); 1478 hv_iterinit (hv);
1255 he = hv_iternext (hv); 1479 he = hv_iternext (hv);
1256 hv_iterinit (hv); 1480 hv_iterinit (hv);
1257 1481
1258 // the next line creates a mortal sv each time its called. 1482 // the next line creates a mortal sv each time it's called.
1259 // might want to optimise this for common cases. 1483 // might want to optimise this for common cases.
1260 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1484 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1261 1485
1262 if (cb) 1486 if (cb)
1263 { 1487 {
1264 dSP; 1488 dSP;
1265 int count; 1489 int count;
1266 1490
1267 ENTER; SAVETMPS; PUSHMARK (SP); 1491 ENTER; SAVETMPS;
1492 SAVESTACK_POS ();
1493 PUSHMARK (SP);
1268 XPUSHs (HeVAL (he)); 1494 XPUSHs (HeVAL (he));
1495 sv_2mortal (sv);
1269 1496
1270 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1497 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1271 1498
1272 if (count == 1) 1499 if (count == 1)
1273 { 1500 {
1274 sv = newSVsv (POPs); 1501 sv = newSVsv (POPs);
1275 FREETMPS; LEAVE; 1502 FREETMPS; LEAVE;
1276 return sv; 1503 return sv;
1277 } 1504 }
1278 1505
1506 SvREFCNT_inc (sv);
1279 FREETMPS; LEAVE; 1507 FREETMPS; LEAVE;
1280 } 1508 }
1281 } 1509 }
1282 1510
1283 if (dec->json.cb_object) 1511 if (dec->json.cb_object)
1284 { 1512 {
1285 dSP; 1513 dSP;
1286 int count; 1514 int count;
1287 1515
1288 ENTER; SAVETMPS; PUSHMARK (SP); 1516 ENTER; SAVETMPS;
1517 SAVESTACK_POS ();
1518 PUSHMARK (SP);
1289 XPUSHs (sv_2mortal (sv)); 1519 XPUSHs (sv_2mortal (sv));
1290 1520
1291 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1521 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1292 1522
1293 if (count == 1) 1523 if (count == 1)
1309 DEC_DEC_DEPTH; 1539 DEC_DEC_DEPTH;
1310 return 0; 1540 return 0;
1311} 1541}
1312 1542
1313static SV * 1543static SV *
1544decode_tag (dec_t *dec)
1545{
1546 SV *tag = 0;
1547 SV *val = 0;
1548
1549 if (!(dec->json.flags & F_ALLOW_TAGS))
1550 ERR ("malformed JSON string, neither array, object, number, string or atom");
1551
1552 ++dec->cur;
1553
1554 decode_ws (dec);
1555
1556 tag = decode_sv (dec);
1557 if (!tag)
1558 goto fail;
1559
1560 if (!SvPOK (tag))
1561 ERR ("malformed JSON string, (tag) must be a string");
1562
1563 decode_ws (dec);
1564
1565 if (*dec->cur != ')')
1566 ERR (") expected after tag");
1567
1568 ++dec->cur;
1569
1570 decode_ws (dec);
1571
1572 val = decode_sv (dec);
1573 if (!val)
1574 goto fail;
1575
1576 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1577 ERR ("malformed JSON string, tag value must be an array");
1578
1579 {
1580 AV *av = (AV *)SvRV (val);
1581 int i, len = av_len (av) + 1;
1582 HV *stash = gv_stashsv (tag, 0);
1583 SV *sv;
1584
1585 if (!stash)
1586 ERR ("cannot decode perl-object (package does not exist)");
1587
1588 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1589
1590 if (!method)
1591 ERR ("cannot decode perl-object (package does not have a THAW method)");
1592
1593 dSP;
1594
1595 ENTER; SAVETMPS;
1596 PUSHMARK (SP);
1597 EXTEND (SP, len + 2);
1598 // we re-bless the reference to get overload and other niceties right
1599 PUSHs (tag);
1600 PUSHs (sv_json);
1601
1602 for (i = 0; i < len; ++i)
1603 PUSHs (*av_fetch (av, i, 1));
1604
1605 PUTBACK;
1606 call_sv ((SV *)GvCV (method), G_SCALAR);
1607 SPAGAIN;
1608
1609 SvREFCNT_dec (tag);
1610 SvREFCNT_dec (val);
1611 sv = SvREFCNT_inc (POPs);
1612
1613 PUTBACK;
1614
1615 FREETMPS; LEAVE;
1616
1617 return sv;
1618 }
1619
1620fail:
1621 SvREFCNT_dec (tag);
1622 SvREFCNT_dec (val);
1623 return 0;
1624}
1625
1626static SV *
1314decode_sv (dec_t *dec) 1627decode_sv (dec_t *dec)
1315{ 1628{
1316 // the beauty of JSON: you need exactly one character lookahead 1629 // the beauty of JSON: you need exactly one character lookahead
1317 // to parse everything. 1630 // to parse everything.
1318 switch (*dec->cur) 1631 switch (*dec->cur)
1319 { 1632 {
1320 case '"': ++dec->cur; return decode_str (dec); 1633 case '"': ++dec->cur; return decode_str (dec);
1321 case '[': ++dec->cur; return decode_av (dec); 1634 case '[': ++dec->cur; return decode_av (dec);
1322 case '{': ++dec->cur; return decode_hv (dec); 1635 case '{': ++dec->cur; return decode_hv (dec);
1636 case '(': return decode_tag (dec);
1323 1637
1324 case '-': 1638 case '-':
1325 case '0': case '1': case '2': case '3': case '4': 1639 case '0': case '1': case '2': case '3': case '4':
1326 case '5': case '6': case '7': case '8': case '9': 1640 case '5': case '6': case '7': case '8': case '9':
1327 return decode_num (dec); 1641 return decode_num (dec);
1329 case 't': 1643 case 't':
1330 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1644 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1331 { 1645 {
1332 dec->cur += 4; 1646 dec->cur += 4;
1333#if JSON_SLOW 1647#if JSON_SLOW
1334 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1648 bool_true = get_bool ("Types::Serialiser::true");
1335#endif 1649#endif
1336 return SvREFCNT_inc (json_true); 1650 return newSVsv (bool_true);
1337 } 1651 }
1338 else 1652 else
1339 ERR ("'true' expected"); 1653 ERR ("'true' expected");
1340 1654
1341 break; 1655 break;
1343 case 'f': 1657 case 'f':
1344 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1658 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1345 { 1659 {
1346 dec->cur += 5; 1660 dec->cur += 5;
1347#if JSON_SLOW 1661#if JSON_SLOW
1348 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1662 bool_false = get_bool ("Types::Serialiser::false");
1349#endif 1663#endif
1350 return SvREFCNT_inc (json_false); 1664 return newSVsv (bool_false);
1351 } 1665 }
1352 else 1666 else
1353 ERR ("'false' expected"); 1667 ERR ("'false' expected");
1354 1668
1355 break; 1669 break;
1364 ERR ("'null' expected"); 1678 ERR ("'null' expected");
1365 1679
1366 break; 1680 break;
1367 1681
1368 default: 1682 default:
1369 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1683 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1370 break; 1684 break;
1371 } 1685 }
1372 1686
1373fail: 1687fail:
1374 return 0; 1688 return 0;
1375} 1689}
1376 1690
1377static SV * 1691static SV *
1378decode_json (SV *string, JSON *json, UV *offset_return) 1692decode_json (SV *string, JSON *json, char **offset_return)
1379{ 1693{
1380 dec_t dec; 1694 dec_t dec;
1381 UV offset;
1382 SV *sv; 1695 SV *sv;
1383 1696
1697 /* work around bugs in 5.10 where manipulating magic values
1698 * makes perl ignore the magic in subsequent accesses.
1699 * also make a copy of non-PV values, to get them into a clean
1700 * state (SvPV should do that, but it's buggy, see below).
1701 *
1702 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1703 * as reported by Reini Urban.
1704 */
1384 SvGETMAGIC (string); 1705 /*SvGETMAGIC (string);*/
1706 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1707 string = sv_2mortal (newSVsv (string));
1708
1385 SvUPGRADE (string, SVt_PV); 1709 SvUPGRADE (string, SVt_PV);
1386 1710
1387 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1711 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1712 * assertion with -DDEBUGGING, although SvCUR is documented to
1713 * return the xpv_cur field which certainly exists after upgrading.
1714 * according to nicholas clark, calling SvPOK fixes this.
1715 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1716 * and hope for the best.
1717 * Damnit, SvPV_nolen still trips over yet another assertion. This
1718 * assertion business is seriously broken, try yet another workaround
1719 * for the broken -DDEBUGGING.
1720 */
1721 {
1722#ifdef DEBUGGING
1723 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1724#else
1725 STRLEN offset = SvCUR (string);
1726#endif
1727
1728 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", 1729 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)); 1730 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1731 }
1390 1732
1391 if (json->flags & F_UTF8) 1733 if (DECODE_WANTS_OCTETS (json))
1392 sv_utf8_downgrade (string, 0); 1734 sv_utf8_downgrade (string, 0);
1393 else 1735 else
1394 sv_utf8_upgrade (string); 1736 sv_utf8_upgrade (string);
1395 1737
1396 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1738 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1397 1739
1398 dec.json = *json; 1740 dec.json = *json;
1399 dec.cur = SvPVX (string); 1741 dec.cur = SvPVX (string);
1400 dec.end = SvEND (string); 1742 dec.end = SvEND (string);
1401 dec.err = 0; 1743 dec.err = 0;
1402 dec.depth = 0; 1744 dec.depth = 0;
1403 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1404 1745
1405 if (dec.json.cb_object || dec.json.cb_sk_object) 1746 if (dec.json.cb_object || dec.json.cb_sk_object)
1406 dec.json.flags |= F_HOOK; 1747 dec.json.flags |= F_HOOK;
1407 1748
1408 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1749 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1409 1750
1410 decode_ws (&dec); 1751 decode_ws (&dec);
1411 sv = decode_sv (&dec); 1752 sv = decode_sv (&dec);
1753
1754 if (offset_return)
1755 *offset_return = dec.cur;
1412 1756
1413 if (!(offset_return || !sv)) 1757 if (!(offset_return || !sv))
1414 { 1758 {
1415 // check for trailing garbage 1759 // check for trailing garbage
1416 decode_ws (&dec); 1760 decode_ws (&dec);
1419 { 1763 {
1420 dec.err = "garbage after JSON object"; 1764 dec.err = "garbage after JSON object";
1421 SvREFCNT_dec (sv); 1765 SvREFCNT_dec (sv);
1422 sv = 0; 1766 sv = 0;
1423 } 1767 }
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 } 1768 }
1435 1769
1436 if (!sv) 1770 if (!sv)
1437 { 1771 {
1438 SV *uni = sv_newmortal (); 1772 SV *uni = sv_newmortal ();
1444 SAVEVPTR (PL_curcop); 1778 SAVEVPTR (PL_curcop);
1445 PL_curcop = &cop; 1779 PL_curcop = &cop;
1446 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1780 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1447 LEAVE; 1781 LEAVE;
1448 1782
1449 croak ("%s, at character offset %d [\"%s\"]", 1783 croak ("%s, at character offset %d (before \"%s\")",
1450 dec.err, 1784 dec.err,
1451 (int)offset, 1785 (int)ptr_to_index (string, dec.cur),
1452 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1786 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1453 } 1787 }
1454 1788
1455 sv = sv_2mortal (sv); 1789 sv = sv_2mortal (sv);
1456 1790
1457 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1791 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)"); 1792 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1459 1793
1460 return sv; 1794 return sv;
1795}
1796
1797/////////////////////////////////////////////////////////////////////////////
1798// incremental parser
1799
1800static void
1801incr_parse (JSON *self)
1802{
1803 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1804
1805 // the state machine here is a bit convoluted and could be simplified a lot
1806 // but this would make it slower, so...
1807
1808 for (;;)
1809 {
1810 //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
1811 switch (self->incr_mode)
1812 {
1813 // only used for initial whitespace skipping
1814 case INCR_M_WS:
1815 for (;;)
1816 {
1817 if (*p > 0x20)
1818 {
1819 if (*p == '#')
1820 {
1821 self->incr_mode = INCR_M_C0;
1822 goto incr_m_c;
1823 }
1824 else
1825 {
1826 self->incr_mode = INCR_M_JSON;
1827 goto incr_m_json;
1828 }
1829 }
1830 else if (!*p)
1831 goto interrupt;
1832
1833 ++p;
1834 }
1835
1836 // skip a single char inside a string (for \\-processing)
1837 case INCR_M_BS:
1838 if (!*p)
1839 goto interrupt;
1840
1841 ++p;
1842 self->incr_mode = INCR_M_STR;
1843 goto incr_m_str;
1844
1845 // inside #-style comments
1846 case INCR_M_C0:
1847 case INCR_M_C1:
1848 incr_m_c:
1849 for (;;)
1850 {
1851 if (*p == '\n')
1852 {
1853 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1854 break;
1855 }
1856 else if (!*p)
1857 goto interrupt;
1858
1859 ++p;
1860 }
1861
1862 break;
1863
1864 // inside a string
1865 case INCR_M_STR:
1866 incr_m_str:
1867 for (;;)
1868 {
1869 if (*p == '"')
1870 {
1871 ++p;
1872 self->incr_mode = INCR_M_JSON;
1873
1874 if (!self->incr_nest)
1875 goto interrupt;
1876
1877 goto incr_m_json;
1878 }
1879 else if (*p == '\\')
1880 {
1881 ++p; // "virtually" consumes character after \
1882
1883 if (!*p) // if at end of string we have to switch modes
1884 {
1885 self->incr_mode = INCR_M_BS;
1886 goto interrupt;
1887 }
1888 }
1889 else if (!*p)
1890 goto interrupt;
1891
1892 ++p;
1893 }
1894
1895 // after initial ws, outside string
1896 case INCR_M_JSON:
1897 incr_m_json:
1898 for (;;)
1899 {
1900 switch (*p++)
1901 {
1902 case 0:
1903 --p;
1904 goto interrupt;
1905
1906 case 0x09:
1907 case 0x0a:
1908 case 0x0d:
1909 case 0x20:
1910 if (!self->incr_nest)
1911 {
1912 --p; // do not eat the whitespace, let the next round do it
1913 goto interrupt;
1914 }
1915 break;
1916
1917 case '"':
1918 self->incr_mode = INCR_M_STR;
1919 goto incr_m_str;
1920
1921 case '[':
1922 case '{':
1923 case '(':
1924 if (++self->incr_nest > self->max_depth)
1925 croak (ERR_NESTING_EXCEEDED);
1926 break;
1927
1928 case ']':
1929 case '}':
1930 if (--self->incr_nest <= 0)
1931 goto interrupt;
1932 break;
1933
1934 case ')':
1935 --self->incr_nest;
1936 break;
1937
1938 case '#':
1939 self->incr_mode = INCR_M_C1;
1940 goto incr_m_c;
1941 }
1942 }
1943 }
1944
1945 modechange:
1946 ;
1947 }
1948
1949interrupt:
1950 self->incr_pos = p - SvPVX (self->incr_text);
1951 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1952 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1461} 1953}
1462 1954
1463///////////////////////////////////////////////////////////////////////////// 1955/////////////////////////////////////////////////////////////////////////////
1464// XS interface functions 1956// XS interface functions
1465 1957
1474 i >= '0' && i <= '9' ? i - '0' 1966 i >= '0' && i <= '9' ? i - '0'
1475 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1967 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1476 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1968 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1477 : -1; 1969 : -1;
1478 1970
1479 json_stash = gv_stashpv ("JSON::XS" , 1); 1971 json_stash = gv_stashpv ("JSON::XS" , 1);
1480 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1972 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1973 bool_true = get_bool ("Types::Serialiser::true");
1974 bool_false = get_bool ("Types::Serialiser::false");
1481 1975
1482 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1976 sv_json = newSVpv ("JSON", 0);
1483 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1977 SvREADONLY_on (sv_json);
1978
1979 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1484} 1980}
1485 1981
1486PROTOTYPES: DISABLE 1982PROTOTYPES: DISABLE
1487 1983
1488void CLONE (...) 1984void CLONE (...)
1489 CODE: 1985 CODE:
1490 json_stash = 0; 1986 json_stash = 0;
1491 json_boolean_stash = 0; 1987 bool_stash = 0;
1492 1988
1493void new (char *klass) 1989void new (char *klass)
1494 PPCODE: 1990 PPCODE:
1495{ 1991{
1496 SV *pv = NEWSV (0, sizeof (JSON)); 1992 SV *pv = NEWSV (0, sizeof (JSON));
1497 SvPOK_only (pv); 1993 SvPOK_only (pv);
1498 Zero (SvPVX (pv), 1, JSON); 1994 json_init ((JSON *)SvPVX (pv));
1499 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1500 XPUSHs (sv_2mortal (sv_bless ( 1995 XPUSHs (sv_2mortal (sv_bless (
1501 newRV_noinc (pv), 1996 newRV_noinc (pv),
1502 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1997 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1503 ))); 1998 )));
1504} 1999}
1516 allow_nonref = F_ALLOW_NONREF 2011 allow_nonref = F_ALLOW_NONREF
1517 shrink = F_SHRINK 2012 shrink = F_SHRINK
1518 allow_blessed = F_ALLOW_BLESSED 2013 allow_blessed = F_ALLOW_BLESSED
1519 convert_blessed = F_CONV_BLESSED 2014 convert_blessed = F_CONV_BLESSED
1520 relaxed = F_RELAXED 2015 relaxed = F_RELAXED
2016 allow_unknown = F_ALLOW_UNKNOWN
2017 allow_tags = F_ALLOW_TAGS
1521 PPCODE: 2018 PPCODE:
1522{ 2019{
1523 if (enable) 2020 if (enable)
1524 self->flags |= ix; 2021 self->flags |= ix;
1525 else 2022 else
1540 get_allow_nonref = F_ALLOW_NONREF 2037 get_allow_nonref = F_ALLOW_NONREF
1541 get_shrink = F_SHRINK 2038 get_shrink = F_SHRINK
1542 get_allow_blessed = F_ALLOW_BLESSED 2039 get_allow_blessed = F_ALLOW_BLESSED
1543 get_convert_blessed = F_CONV_BLESSED 2040 get_convert_blessed = F_CONV_BLESSED
1544 get_relaxed = F_RELAXED 2041 get_relaxed = F_RELAXED
2042 get_allow_unknown = F_ALLOW_UNKNOWN
2043 get_allow_tags = F_ALLOW_TAGS
1545 PPCODE: 2044 PPCODE:
1546 XPUSHs (boolSV (self->flags & ix)); 2045 XPUSHs (boolSV (self->flags & ix));
1547 2046
1548void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2047void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1549 PPCODE: 2048 PPCODE:
1550{ 2049 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)); 2050 XPUSHs (ST (0));
1561}
1562 2051
1563U32 get_max_depth (JSON *self) 2052U32 get_max_depth (JSON *self)
1564 CODE: 2053 CODE:
1565 RETVAL = DEC_DEPTH (self->flags); 2054 RETVAL = self->max_depth;
1566 OUTPUT: 2055 OUTPUT:
1567 RETVAL 2056 RETVAL
1568 2057
1569void max_size (JSON *self, UV max_size = 0) 2058void max_size (JSON *self, U32 max_size = 0)
1570 PPCODE: 2059 PPCODE:
1571{ 2060 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)); 2061 XPUSHs (ST (0));
1583}
1584 2062
1585int get_max_size (JSON *self) 2063int get_max_size (JSON *self)
1586 CODE: 2064 CODE:
1587 RETVAL = DEC_SIZE (self->flags); 2065 RETVAL = self->max_size;
1588 OUTPUT: 2066 OUTPUT:
1589 RETVAL 2067 RETVAL
1590 2068
1591void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2069void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1592 PPCODE: 2070 PPCODE:
1598} 2076}
1599 2077
1600void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2078void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1601 PPCODE: 2079 PPCODE:
1602{ 2080{
1603 if (!self->cb_sk_object) 2081 if (!self->cb_sk_object)
1604 self->cb_sk_object = newHV (); 2082 self->cb_sk_object = newHV ();
1605 2083
1606 if (SvOK (cb)) 2084 if (SvOK (cb))
1607 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2085 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1608 else 2086 else
1619 XPUSHs (ST (0)); 2097 XPUSHs (ST (0));
1620} 2098}
1621 2099
1622void encode (JSON *self, SV *scalar) 2100void encode (JSON *self, SV *scalar)
1623 PPCODE: 2101 PPCODE:
1624 XPUSHs (encode_json (scalar, self)); 2102 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2103 XPUSHs (scalar);
1625 2104
1626void decode (JSON *self, SV *jsonstr) 2105void decode (JSON *self, SV *jsonstr)
1627 PPCODE: 2106 PPCODE:
1628 XPUSHs (decode_json (jsonstr, self, 0)); 2107 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2108 XPUSHs (jsonstr);
1629 2109
1630void decode_prefix (JSON *self, SV *jsonstr) 2110void decode_prefix (JSON *self, SV *jsonstr)
1631 PPCODE: 2111 PPCODE:
1632{ 2112{
1633 UV offset; 2113 SV *sv;
2114 char *offset;
2115 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1634 EXTEND (SP, 2); 2116 EXTEND (SP, 2);
1635 PUSHs (decode_json (jsonstr, self, &offset)); 2117 PUSHs (sv);
1636 PUSHs (sv_2mortal (newSVuv (offset))); 2118 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
2119}
2120
2121void incr_parse (JSON *self, SV *jsonstr = 0)
2122 PPCODE:
2123{
2124 if (!self->incr_text)
2125 self->incr_text = newSVpvn ("", 0);
2126
2127 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2128 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2129 if (DECODE_WANTS_OCTETS (self))
2130 {
2131 if (self->incr_pos)
2132 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2133 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2134
2135 sv_utf8_downgrade (self->incr_text, 0);
2136 }
2137 else
2138 {
2139 sv_utf8_upgrade (self->incr_text);
2140
2141 if (self->incr_pos)
2142 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2143 - (U8 *)SvPVX (self->incr_text);
2144 }
2145
2146 // append data, if any
2147 if (jsonstr)
2148 {
2149 /* make sure both strings have same encoding */
2150 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2151 if (SvUTF8 (jsonstr))
2152 sv_utf8_downgrade (jsonstr, 0);
2153 else
2154 sv_utf8_upgrade (jsonstr);
2155
2156 /* and then just blindly append */
2157 {
2158 STRLEN len;
2159 const char *str = SvPV (jsonstr, len);
2160 STRLEN cur = SvCUR (self->incr_text);
2161
2162 if (SvLEN (self->incr_text) <= cur + len)
2163 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2164
2165 Move (str, SvEND (self->incr_text), len, char);
2166 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2167 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2168 }
2169 }
2170
2171 if (GIMME_V != G_VOID)
2172 do
2173 {
2174 SV *sv;
2175 char *offset;
2176
2177 if (!INCR_DONE (self))
2178 {
2179 incr_parse (self);
2180
2181 if (self->incr_pos > self->max_size && self->max_size)
2182 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2183 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2184
2185 if (!INCR_DONE (self))
2186 {
2187 // as an optimisation, do not accumulate white space in the incr buffer
2188 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2189 {
2190 self->incr_pos = 0;
2191 SvCUR_set (self->incr_text, 0);
2192 }
2193
2194 break;
2195 }
2196 }
2197
2198 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2199 XPUSHs (sv);
2200
2201 self->incr_pos -= offset - SvPVX (self->incr_text);
2202 self->incr_nest = 0;
2203 self->incr_mode = 0;
2204
2205 sv_chop (self->incr_text, offset);
2206 }
2207 while (GIMME_V == G_ARRAY);
2208}
2209
2210SV *incr_text (JSON *self)
2211 ATTRS: lvalue
2212 CODE:
2213{
2214 if (self->incr_pos)
2215 croak ("incr_text can not be called when the incremental parser already started parsing");
2216
2217 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2218}
2219 OUTPUT:
2220 RETVAL
2221
2222void incr_skip (JSON *self)
2223 CODE:
2224{
2225 if (self->incr_pos)
2226 {
2227 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2228 self->incr_pos = 0;
2229 self->incr_nest = 0;
2230 self->incr_mode = 0;
2231 }
2232}
2233
2234void incr_reset (JSON *self)
2235 CODE:
2236{
2237 SvREFCNT_dec (self->incr_text);
2238 self->incr_text = 0;
2239 self->incr_pos = 0;
2240 self->incr_nest = 0;
2241 self->incr_mode = 0;
1637} 2242}
1638 2243
1639void DESTROY (JSON *self) 2244void DESTROY (JSON *self)
1640 CODE: 2245 CODE:
1641 SvREFCNT_dec (self->cb_sk_object); 2246 SvREFCNT_dec (self->cb_sk_object);
1642 SvREFCNT_dec (self->cb_object); 2247 SvREFCNT_dec (self->cb_object);
2248 SvREFCNT_dec (self->incr_text);
1643 2249
1644PROTOTYPES: ENABLE 2250PROTOTYPES: ENABLE
1645 2251
1646void encode_json (SV *scalar) 2252void encode_json (SV *scalar)
1647 PPCODE: 2253 PPCODE:
1648{ 2254{
1649 JSON json = { F_DEFAULT | F_UTF8 }; 2255 JSON json;
1650 XPUSHs (encode_json (scalar, &json)); 2256 json_init (&json);
2257 json.flags |= F_UTF8;
2258 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2259 XPUSHs (scalar);
1651} 2260}
1652 2261
1653void decode_json (SV *jsonstr) 2262void decode_json (SV *jsonstr)
1654 PPCODE: 2263 PPCODE:
1655{ 2264{
1656 JSON json = { F_DEFAULT | F_UTF8 }; 2265 JSON json;
1657 XPUSHs (decode_json (jsonstr, &json, 0)); 2266 json_init (&json);
2267 json.flags |= F_UTF8;
2268 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2269 XPUSHs (jsonstr);
1658} 2270}
1659 2271

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines