ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/JSON-XS/XS.xs
(Generate patch)

Comparing JSON-XS/XS.xs (file contents):
Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC vs.
Revision 1.123 by root, Sun Dec 8 20:44:23 2013 UTC

4 4
5#include <assert.h> 5#include <assert.h>
6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h>
9#include <float.h> 10#include <float.h>
10 11
11#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
12# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
13#endif 14#endif
14 15
15// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
16// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
17#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
18# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
19#endif 20#endif
21
22// 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
34#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20 35
21#define F_ASCII 0x00000001UL 36#define F_ASCII 0x00000001UL
22#define F_LATIN1 0x00000002UL 37#define F_LATIN1 0x00000002UL
23#define F_UTF8 0x00000004UL 38#define F_UTF8 0x00000004UL
24#define F_INDENT 0x00000008UL 39#define F_INDENT 0x00000008UL
28#define F_ALLOW_NONREF 0x00000100UL 43#define F_ALLOW_NONREF 0x00000100UL
29#define F_SHRINK 0x00000200UL 44#define F_SHRINK 0x00000200UL
30#define F_ALLOW_BLESSED 0x00000400UL 45#define F_ALLOW_BLESSED 0x00000400UL
31#define F_CONV_BLESSED 0x00000800UL 46#define F_CONV_BLESSED 0x00000800UL
32#define F_RELAXED 0x00001000UL 47#define F_RELAXED 0x00001000UL
33 48#define F_ALLOW_UNKNOWN 0x00002000UL
34#define F_MAXDEPTH 0xf8000000UL 49#define F_ALLOW_TAGS 0x00004000UL
35#define S_MAXDEPTH 27
36#define F_MAXSIZE 0x01f00000UL
37#define S_MAXSIZE 20
38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 50#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39 51
40#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 52#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
44#define F_DEFAULT (9UL << S_MAXDEPTH)
45 53
46#define INIT_SIZE 32 // initial scalar size to be allocated 54#define INIT_SIZE 32 // initial scalar size to be allocated
47#define INDENT_STEP 3 // spaces per indentation level 55#define INDENT_STEP 3 // spaces per indentation level
48 56
49#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)
50 60
51#define SB do { 61#define SB do {
52#define SE } while (0) 62#define SE } while (0)
53 63
54#if __GNUC__ >= 3 64#if __GNUC__ >= 3
64 74
65#define IN_RANGE_INC(type,val,beg,end) \ 75#define IN_RANGE_INC(type,val,beg,end) \
66 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 76 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
67 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 77 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
68 78
79#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
80
69#ifdef USE_ITHREADS 81#ifdef USE_ITHREADS
70# define JSON_SLOW 1 82# define JSON_SLOW 1
71# 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))
72#else 84#else
73# define JSON_SLOW 0 85# define JSON_SLOW 0
74# define JSON_STASH json_stash 86# define JSON_STASH json_stash
75#endif 87#endif
76 88
89// the amount of HEs to allocate on the stack, when sorting keys
90#define STACK_HES 64
91
77static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92static HV *json_stash, *types_boolean_stash; // JSON::XS::
78static SV *json_true, *json_false; 93static SV *types_true, *types_false, *sv_json;
94
95enum {
96 INCR_M_WS = 0, // initial whitespace skipping, must be 0
97 INCR_M_STR, // inside string
98 INCR_M_BS, // inside backslash
99 INCR_M_C0, // inside comment in initial whitespace sequence
100 INCR_M_C1, // inside comment in other places
101 INCR_M_JSON // outside anything, count nesting
102};
103
104#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
79 105
80typedef struct { 106typedef struct {
81 U32 flags; 107 U32 flags;
108 U32 max_depth;
109 STRLEN max_size;
110
82 SV *cb_object; 111 SV *cb_object;
83 HV *cb_sk_object; 112 HV *cb_sk_object;
113
114 // for the incremental parser
115 SV *incr_text; // the source text so far
116 STRLEN incr_pos; // the current offset into the text
117 int incr_nest; // {[]}-nesting level
118 unsigned char incr_mode;
84} JSON; 119} JSON;
120
121INLINE void
122json_init (JSON *json)
123{
124 Zero (json, 1, JSON);
125 json->max_depth = 512;
126}
85 127
86///////////////////////////////////////////////////////////////////////////// 128/////////////////////////////////////////////////////////////////////////////
87// utility functions 129// utility functions
88 130
131INLINE SV *
132get_bool (const char *name)
133{
134 SV *sv = get_sv (name, 1);
135
136 SvREADONLY_on (sv);
137 SvREADONLY_on (SvRV (sv));
138
139 return sv;
140}
141
89INLINE void 142INLINE void
90shrink (SV *sv) 143shrink (SV *sv)
91{ 144{
92 sv_utf8_downgrade (sv, 1); 145 sv_utf8_downgrade (sv, 1);
146
93 if (SvLEN (sv) > SvCUR (sv) + 1) 147 if (SvLEN (sv) > SvCUR (sv) + 1)
94 { 148 {
95#ifdef SvPV_shrink_to_cur 149#ifdef SvPV_shrink_to_cur
96 SvPV_shrink_to_cur (sv); 150 SvPV_shrink_to_cur (sv);
97#elif defined (SvPV_renew) 151#elif defined (SvPV_renew)
141 *s++ = 0x80 | ( ch & 0x3f); 195 *s++ = 0x80 | ( ch & 0x3f);
142 196
143 return s; 197 return s;
144} 198}
145 199
200// convert offset pointer to character index, sv must be string
201static STRLEN
202ptr_to_index (SV *sv, char *offset)
203{
204 return SvUTF8 (sv)
205 ? utf8_distance (offset, SvPVX (sv))
206 : offset - SvPVX (sv);
207}
208
209/////////////////////////////////////////////////////////////////////////////
210// fp hell
211
212// scan a group of digits, and a trailing exponent
213static void
214json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
215{
216 UV uaccum = 0;
217 int eaccum = 0;
218
219 // if we recurse too deep, skip all remaining digits
220 // to avoid a stack overflow attack
221 if (expect_false (--maxdepth <= 0))
222 while (((U8)*s - '0') < 10)
223 ++s;
224
225 for (;;)
226 {
227 U8 dig = (U8)*s - '0';
228
229 if (expect_false (dig >= 10))
230 {
231 if (dig == (U8)((U8)'.' - (U8)'0'))
232 {
233 ++s;
234 json_atof_scan1 (s, accum, expo, 1, maxdepth);
235 }
236 else if ((dig | ' ') == 'e' - '0')
237 {
238 int exp2 = 0;
239 int neg = 0;
240
241 ++s;
242
243 if (*s == '-')
244 {
245 ++s;
246 neg = 1;
247 }
248 else if (*s == '+')
249 ++s;
250
251 while ((dig = (U8)*s - '0') < 10)
252 exp2 = exp2 * 10 + *s++ - '0';
253
254 *expo += neg ? -exp2 : exp2;
255 }
256
257 break;
258 }
259
260 ++s;
261
262 uaccum = uaccum * 10 + dig;
263 ++eaccum;
264
265 // if we have too many digits, then recurse for more
266 // we actually do this for rather few digits
267 if (uaccum >= (UV_MAX - 9) / 10)
268 {
269 if (postdp) *expo -= eaccum;
270 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
271 if (postdp) *expo += eaccum;
272
273 break;
274 }
275 }
276
277 // this relies greatly on the quality of the pow ()
278 // implementation of the platform, but a good
279 // implementation is hard to beat.
280 // (IEEE 754 conformant ones are required to be exact)
281 if (postdp) *expo -= eaccum;
282 *accum += uaccum * Perl_pow (10., *expo);
283 *expo += eaccum;
284}
285
286static NV
287json_atof (const char *s)
288{
289 NV accum = 0.;
290 int expo = 0;
291 int neg = 0;
292
293 if (*s == '-')
294 {
295 ++s;
296 neg = 1;
297 }
298
299 // a recursion depth of ten gives us >>500 bits
300 json_atof_scan1 (s, &accum, &expo, 0, 10);
301
302 return neg ? -accum : accum;
303}
146///////////////////////////////////////////////////////////////////////////// 304/////////////////////////////////////////////////////////////////////////////
147// encoder 305// encoder
148 306
149// structure used for encoding JSON 307// structure used for encoding JSON
150typedef struct 308typedef struct
152 char *cur; // SvPVX (sv) + current output position 310 char *cur; // SvPVX (sv) + current output position
153 char *end; // SvEND (sv) 311 char *end; // SvEND (sv)
154 SV *sv; // result scalar 312 SV *sv; // result scalar
155 JSON json; 313 JSON json;
156 U32 indent; // indentation level 314 U32 indent; // indentation level
157 U32 maxdepth; // max. indentation/recursion level
158 UV limit; // escape character values >= this value when encoding 315 UV limit; // escape character values >= this value when encoding
159} enc_t; 316} enc_t;
160 317
161INLINE void 318INLINE void
162need (enc_t *enc, STRLEN len) 319need (enc_t *enc, STRLEN len)
163{ 320{
164 if (expect_false (enc->cur + len >= enc->end)) 321 if (expect_false (enc->cur + len >= enc->end))
165 { 322 {
166 STRLEN cur = enc->cur - SvPVX (enc->sv); 323 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
167 SvGROW (enc->sv, cur + len + 1); 324 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
168 enc->cur = SvPVX (enc->sv) + cur; 325 enc->cur = SvPVX (enc->sv) + cur;
169 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 326 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
170 } 327 }
171} 328}
172 329
247 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 404 (int)((uch - 0x10000) % 0x400 + 0xDC00));
248 enc->cur += 12; 405 enc->cur += 12;
249 } 406 }
250 else 407 else
251 { 408 {
252 static char hexdigit [16] = "0123456789abcdef";
253 need (enc, len += 5); 409 need (enc, len += 5);
254 *enc->cur++ = '\\'; 410 *enc->cur++ = '\\';
255 *enc->cur++ = 'u'; 411 *enc->cur++ = 'u';
256 *enc->cur++ = hexdigit [ uch >> 12 ]; 412 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
257 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 413 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
258 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 414 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
259 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 415 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
260 } 416 }
261 417
262 str += clen; 418 str += clen;
263 } 419 }
264 else if (enc->json.flags & F_LATIN1) 420 else if (enc->json.flags & F_LATIN1)
335static void 491static void
336encode_av (enc_t *enc, AV *av) 492encode_av (enc_t *enc, AV *av)
337{ 493{
338 int i, len = av_len (av); 494 int i, len = av_len (av);
339 495
340 if (enc->indent >= enc->maxdepth) 496 if (enc->indent >= enc->json.max_depth)
341 croak ("data structure too deep (hit recursion limit)"); 497 croak (ERR_NESTING_EXCEEDED);
342 498
343 encode_ch (enc, '['); 499 encode_ch (enc, '[');
344 500
345 if (len >= 0) 501 if (len >= 0)
346 { 502 {
347 encode_nl (enc); ++enc->indent; 503 encode_nl (enc); ++enc->indent;
348 504
349 for (i = 0; i <= len; ++i) 505 for (i = 0; i <= len; ++i)
361 encode_comma (enc); 517 encode_comma (enc);
362 } 518 }
363 519
364 encode_nl (enc); --enc->indent; encode_indent (enc); 520 encode_nl (enc); --enc->indent; encode_indent (enc);
365 } 521 }
366 522
367 encode_ch (enc, ']'); 523 encode_ch (enc, ']');
368} 524}
369 525
370static void 526static void
371encode_hk (enc_t *enc, HE *he) 527encode_hk (enc_t *enc, HE *he)
375 if (HeKLEN (he) == HEf_SVKEY) 531 if (HeKLEN (he) == HEf_SVKEY)
376 { 532 {
377 SV *sv = HeSVKEY (he); 533 SV *sv = HeSVKEY (he);
378 STRLEN len; 534 STRLEN len;
379 char *str; 535 char *str;
380 536
381 SvGETMAGIC (sv); 537 SvGETMAGIC (sv);
382 str = SvPV (sv, len); 538 str = SvPV (sv, len);
383 539
384 encode_str (enc, str, len, SvUTF8 (sv)); 540 encode_str (enc, str, len, SvUTF8 (sv));
385 } 541 }
420 576
421static void 577static void
422encode_hv (enc_t *enc, HV *hv) 578encode_hv (enc_t *enc, HV *hv)
423{ 579{
424 HE *he; 580 HE *he;
425 int count;
426 581
427 if (enc->indent >= enc->maxdepth) 582 if (enc->indent >= enc->json.max_depth)
428 croak ("data structure too deep (hit recursion limit)"); 583 croak (ERR_NESTING_EXCEEDED);
429 584
430 encode_ch (enc, '{'); 585 encode_ch (enc, '{');
431 586
432 // for canonical output we have to sort by keys first 587 // for canonical output we have to sort by keys first
433 // actually, this is mostly due to the stupid so-called 588 // actually, this is mostly due to the stupid so-called
434 // security workaround added somewhere in 5.8.x. 589 // security workaround added somewhere in 5.8.x
435 // that randomises hash orderings 590 // that randomises hash orderings
436 if (enc->json.flags & F_CANONICAL) 591 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
437 { 592 {
438 int count = hv_iterinit (hv); 593 int count = hv_iterinit (hv);
439 594
440 if (SvMAGICAL (hv)) 595 if (SvMAGICAL (hv))
441 { 596 {
452 } 607 }
453 608
454 if (count) 609 if (count)
455 { 610 {
456 int i, fast = 1; 611 int i, fast = 1;
457#if defined(__BORLANDC__) || defined(_MSC_VER) 612 HE *hes_stack [STACK_HES];
458 HE **hes = _alloca (count * sizeof (HE)); 613 HE **hes = hes_stack;
459#else 614
460 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 615 // allocate larger arrays on the heap
461#endif 616 if (count > STACK_HES)
617 {
618 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
619 hes = (HE **)SvPVX (sv);
620 }
462 621
463 i = 0; 622 i = 0;
464 while ((he = hv_iternext (hv))) 623 while ((he = hv_iternext (hv)))
465 { 624 {
466 hes [i++] = he; 625 hes [i++] = he;
535// encode objects, arrays and special \0=false and \1=true values. 694// encode objects, arrays and special \0=false and \1=true values.
536static void 695static void
537encode_rv (enc_t *enc, SV *sv) 696encode_rv (enc_t *enc, SV *sv)
538{ 697{
539 svtype svt; 698 svtype svt;
699 GV *method;
540 700
541 SvGETMAGIC (sv); 701 SvGETMAGIC (sv);
542 svt = SvTYPE (sv); 702 svt = SvTYPE (sv);
543 703
544 if (expect_false (SvOBJECT (sv))) 704 if (expect_false (SvOBJECT (sv)))
545 { 705 {
546 HV *stash = !JSON_SLOW || json_boolean_stash 706 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
547 ? json_boolean_stash 707 ? types_boolean_stash
548 : gv_stashpv ("JSON::XS::Boolean", 1); 708 : gv_stashpv ("Types::Serialiser::Boolean", 1);
709 HV *stash = SvSTASH (sv);
549 710
550 if (SvSTASH (sv) == stash) 711 if (stash == boolean_stash)
551 { 712 {
552 if (SvIV (sv)) 713 if (SvIV (sv))
553 encode_str (enc, "true", 4, 0); 714 encode_str (enc, "true", 4, 0);
554 else 715 else
555 encode_str (enc, "false", 5, 0); 716 encode_str (enc, "false", 5, 0);
556 } 717 }
718 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
719 {
720 int count;
721 dSP;
722
723 ENTER; SAVETMPS; PUSHMARK (SP);
724 EXTEND (SP, 2);
725 // we re-bless the reference to get overload and other niceties right
726 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
727 PUSHs (sv_json);
728
729 PUTBACK;
730 count = call_sv ((SV *)GvCV (method), G_ARRAY);
731 SPAGAIN;
732
733 // catch this surprisingly common error
734 if (SvROK (TOPs) && SvRV (TOPs) == sv)
735 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
736
737 encode_ch (enc, '(');
738 encode_ch (enc, '"');
739 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
740 encode_ch (enc, '"');
741 encode_ch (enc, ')');
742 encode_ch (enc, '[');
743
744 while (count)
745 {
746 encode_sv (enc, SP[1 - count--]);
747
748 if (count)
749 encode_ch (enc, ',');
750 }
751
752 encode_ch (enc, ']');
753
754 PUTBACK;
755
756 FREETMPS; LEAVE;
757 }
758 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
759 {
760 dSP;
761
762 ENTER; SAVETMPS; PUSHMARK (SP);
763 // we re-bless the reference to get overload and other niceties right
764 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
765
766 // calling with G_SCALAR ensures that we always get a 1 return value
767 PUTBACK;
768 call_sv ((SV *)GvCV (method), G_SCALAR);
769 SPAGAIN;
770
771 // catch this surprisingly common error
772 if (SvROK (TOPs) && SvRV (TOPs) == sv)
773 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
774
775 sv = POPs;
776 PUTBACK;
777
778 encode_sv (enc, sv);
779
780 FREETMPS; LEAVE;
781 }
782 else if (enc->json.flags & F_ALLOW_BLESSED)
783 encode_str (enc, "null", 4, 0);
557 else 784 else
558 {
559#if 0
560 if (0 && sv_derived_from (rv, "JSON::Literal"))
561 {
562 // not yet
563 }
564#endif
565 if (enc->json.flags & F_CONV_BLESSED)
566 {
567 // we re-bless the reference to get overload and other niceties right
568 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
569
570 if (to_json)
571 {
572 dSP;
573
574 ENTER; SAVETMPS; PUSHMARK (SP);
575 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
576
577 // calling with G_SCALAR ensures that we always get a 1 return value
578 PUTBACK;
579 call_sv ((SV *)GvCV (to_json), G_SCALAR);
580 SPAGAIN;
581
582 // catch this surprisingly common error
583 if (SvROK (TOPs) && SvRV (TOPs) == sv)
584 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
585
586 sv = POPs;
587 PUTBACK;
588
589 encode_sv (enc, sv);
590
591 FREETMPS; LEAVE;
592 }
593 else if (enc->json.flags & F_ALLOW_BLESSED)
594 encode_str (enc, "null", 4, 0);
595 else
596 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
597 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
598 }
599 else if (enc->json.flags & F_ALLOW_BLESSED)
600 encode_str (enc, "null", 4, 0);
601 else
602 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 785 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
603 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 786 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
604 }
605 } 787 }
606 else if (svt == SVt_PVHV) 788 else if (svt == SVt_PVHV)
607 encode_hv (enc, (HV *)sv); 789 encode_hv (enc, (HV *)sv);
608 else if (svt == SVt_PVAV) 790 else if (svt == SVt_PVAV)
609 encode_av (enc, (AV *)sv); 791 encode_av (enc, (AV *)sv);
614 796
615 if (len == 1 && *pv == '1') 797 if (len == 1 && *pv == '1')
616 encode_str (enc, "true", 4, 0); 798 encode_str (enc, "true", 4, 0);
617 else if (len == 1 && *pv == '0') 799 else if (len == 1 && *pv == '0')
618 encode_str (enc, "false", 5, 0); 800 encode_str (enc, "false", 5, 0);
801 else if (enc->json.flags & F_ALLOW_UNKNOWN)
802 encode_str (enc, "null", 4, 0);
619 else 803 else
620 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 804 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
621 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 805 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
622 } 806 }
807 else if (enc->json.flags & F_ALLOW_UNKNOWN)
808 encode_str (enc, "null", 4, 0);
623 else 809 else
624 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 810 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
625 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 811 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
626} 812}
627 813
648 else if (SvIOKp (sv)) 834 else if (SvIOKp (sv))
649 { 835 {
650 // we assume we can always read an IV as a UV and vice versa 836 // we assume we can always read an IV as a UV and vice versa
651 // we assume two's complement 837 // we assume two's complement
652 // we assume no aliasing issues in the union 838 // we assume no aliasing issues in the union
653 if (SvIsUV (sv) ? SvUVX (sv) > 59000 839 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
654 : SvIVX (sv) > 59000 || SvIVX (sv) < -59000) 840 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
655 {
656 // large integer, use the (rather slow) snprintf way.
657 need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8
658 enc->cur +=
659 SvIsUV(sv)
660 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
661 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
662 }
663 else
664 { 841 {
665 // optimise the "small number case" 842 // optimise the "small number case"
666 // code will likely be branchless and use only a single multiplication 843 // code will likely be branchless and use only a single multiplication
667 // works for numbers up to 59074 844 // works for numbers up to 59074
668 I32 i = SvIVX (sv); 845 I32 i = SvIVX (sv);
679 856
680 // now output digit by digit, each time masking out the integer part 857 // now output digit by digit, each time masking out the integer part
681 // and multiplying by 5 while moving the decimal point one to the right, 858 // and multiplying by 5 while moving the decimal point one to the right,
682 // resulting in a net multiplication by 10. 859 // resulting in a net multiplication by 10.
683 // we always write the digit to memory but conditionally increment 860 // we always write the digit to memory but conditionally increment
684 // the pointer, to ease the usage of conditional move instructions. 861 // the pointer, to enable the use of conditional move instructions.
685 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; 862 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
686 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; 863 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
687 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; 864 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
688 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; 865 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
689 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' 866 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
690 } 867 }
868 else
869 {
870 // large integer, use the (rather slow) snprintf way.
871 need (enc, IVUV_MAXCHARS);
872 enc->cur +=
873 SvIsUV(sv)
874 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
875 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
876 }
691 } 877 }
692 else if (SvROK (sv)) 878 else if (SvROK (sv))
693 encode_rv (enc, SvRV (sv)); 879 encode_rv (enc, SvRV (sv));
694 else if (!SvOK (sv)) 880 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
695 encode_str (enc, "null", 4, 0); 881 encode_str (enc, "null", 4, 0);
696 else 882 else
697 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 883 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
698 SvPV_nolen (sv), SvFLAGS (sv)); 884 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
699} 885}
700 886
701static SV * 887static SV *
702encode_json (SV *scalar, JSON *json) 888encode_json (SV *scalar, JSON *json)
703{ 889{
709 enc.json = *json; 895 enc.json = *json;
710 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 896 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
711 enc.cur = SvPVX (enc.sv); 897 enc.cur = SvPVX (enc.sv);
712 enc.end = SvEND (enc.sv); 898 enc.end = SvEND (enc.sv);
713 enc.indent = 0; 899 enc.indent = 0;
714 enc.maxdepth = DEC_DEPTH (enc.json.flags);
715 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 900 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
716 : enc.json.flags & F_LATIN1 ? 0x000100UL 901 : enc.json.flags & F_LATIN1 ? 0x000100UL
717 : 0x110000UL; 902 : 0x110000UL;
718 903
719 SvPOK_only (enc.sv); 904 SvPOK_only (enc.sv);
720 encode_sv (&enc, scalar); 905 encode_sv (&enc, scalar);
906 encode_nl (&enc);
721 907
722 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 908 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
723 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 909 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
724 910
725 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 911 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
786 if (*dec->cur != ch) \ 972 if (*dec->cur != ch) \
787 ERR (# ch " expected"); \ 973 ERR (# ch " expected"); \
788 ++dec->cur; \ 974 ++dec->cur; \
789 SE 975 SE
790 976
791#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 977#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
792#define DEC_DEC_DEPTH --dec->depth 978#define DEC_DEC_DEPTH --dec->depth
793 979
794static SV *decode_sv (dec_t *dec); 980static SV *decode_sv (dec_t *dec);
795 981
796static signed char decode_hexdigit[256]; 982static signed char decode_hexdigit[256];
905 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1091 else if (expect_true (ch >= 0x20 && ch < 0x80))
906 *cur++ = ch; 1092 *cur++ = ch;
907 else if (ch >= 0x80) 1093 else if (ch >= 0x80)
908 { 1094 {
909 STRLEN clen; 1095 STRLEN clen;
910 UV uch;
911 1096
912 --dec_cur; 1097 --dec_cur;
913 1098
914 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1099 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
915 if (clen == (STRLEN)-1) 1100 if (clen == (STRLEN)-1)
916 ERR ("malformed UTF-8 character in JSON string"); 1101 ERR ("malformed UTF-8 character in JSON string");
917 1102
918 do 1103 do
919 *cur++ = *dec_cur++; 1104 *cur++ = *dec_cur++;
936 { 1121 {
937 STRLEN len = cur - buf; 1122 STRLEN len = cur - buf;
938 1123
939 if (sv) 1124 if (sv)
940 { 1125 {
941 SvGROW (sv, SvCUR (sv) + len + 1); 1126 STRLEN cur = SvCUR (sv);
1127
1128 if (SvLEN (sv) <= cur + len)
1129 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1130
942 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1131 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
943 SvCUR_set (sv, SvCUR (sv) + len); 1132 SvCUR_set (sv, SvCUR (sv) + len);
944 } 1133 }
945 else 1134 else
946 sv = newSVpvn (buf, len); 1135 sv = newSVpvn (buf, len);
1037 1226
1038 // special case the rather common 1..5-digit-int case 1227 // special case the rather common 1..5-digit-int case
1039 if (*start == '-') 1228 if (*start == '-')
1040 switch (len) 1229 switch (len)
1041 { 1230 {
1042 case 2: return newSViv (-( start [1] - '0' * 1)); 1231 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1043 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1232 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1044 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1233 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1045 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1234 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1046 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1235 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1047 } 1236 }
1048 else 1237 else
1049 switch (len) 1238 switch (len)
1050 { 1239 {
1051 case 1: return newSViv ( start [0] - '0' * 1); 1240 case 1: return newSViv ( start [0] - '0' * 1);
1052 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1241 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1053 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1242 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1054 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1243 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1055 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1244 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1056 } 1245 }
1057 1246
1058 { 1247 {
1059 UV uv; 1248 UV uv;
1060 int numtype = grok_number (start, len, &uv); 1249 int numtype = grok_number (start, len, &uv);
1069 } 1258 }
1070 1259
1071 len -= *start == '-' ? 1 : 0; 1260 len -= *start == '-' ? 1 : 0;
1072 1261
1073 // does not fit into IV or UV, try NV 1262 // does not fit into IV or UV, try NV
1074 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1263 if (len <= NV_DIG)
1075 #if defined (LDBL_DIG)
1076 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1077 #endif
1078 )
1079 // fits into NV without loss of precision 1264 // fits into NV without loss of precision
1080 return newSVnv (Atof (start)); 1265 return newSVnv (json_atof (start));
1081 1266
1082 // everything else fails, convert it to a string 1267 // everything else fails, convert it to a string
1083 return newSVpvn (start, dec->cur - start); 1268 return newSVpvn (start, dec->cur - start);
1084 } 1269 }
1085 1270
1086 // loss of precision here 1271 // loss of precision here
1087 return newSVnv (Atof (start)); 1272 return newSVnv (json_atof (start));
1088 1273
1089fail: 1274fail:
1090 return 0; 1275 return 0;
1091} 1276}
1092 1277
1116 if (*dec->cur == ']') 1301 if (*dec->cur == ']')
1117 { 1302 {
1118 ++dec->cur; 1303 ++dec->cur;
1119 break; 1304 break;
1120 } 1305 }
1121 1306
1122 if (*dec->cur != ',') 1307 if (*dec->cur != ',')
1123 ERR (", or ] expected while parsing array"); 1308 ERR (", or ] expected while parsing array");
1124 1309
1125 ++dec->cur; 1310 ++dec->cur;
1126 1311
1168 char *p = dec->cur; 1353 char *p = dec->cur;
1169 char *e = p + 24; // only try up to 24 bytes 1354 char *e = p + 24; // only try up to 24 bytes
1170 1355
1171 for (;;) 1356 for (;;)
1172 { 1357 {
1173 // the >= 0x80 is true on most architectures 1358 // the >= 0x80 is false on most architectures
1174 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1359 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1175 { 1360 {
1176 // slow path, back up and use decode_str 1361 // slow path, back up and use decode_str
1177 SV *key = decode_str (dec); 1362 SV *key = decode_str (dec);
1178 if (!key) 1363 if (!key)
1261 dSP; 1446 dSP;
1262 int count; 1447 int count;
1263 1448
1264 ENTER; SAVETMPS; PUSHMARK (SP); 1449 ENTER; SAVETMPS; PUSHMARK (SP);
1265 XPUSHs (HeVAL (he)); 1450 XPUSHs (HeVAL (he));
1451 sv_2mortal (sv);
1266 1452
1267 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1453 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1268 1454
1269 if (count == 1) 1455 if (count == 1)
1270 { 1456 {
1271 sv = newSVsv (POPs); 1457 sv = newSVsv (POPs);
1272 FREETMPS; LEAVE; 1458 FREETMPS; LEAVE;
1273 return sv; 1459 return sv;
1274 } 1460 }
1275 1461
1462 SvREFCNT_inc (sv);
1276 FREETMPS; LEAVE; 1463 FREETMPS; LEAVE;
1277 } 1464 }
1278 } 1465 }
1279 1466
1280 if (dec->json.cb_object) 1467 if (dec->json.cb_object)
1306 DEC_DEC_DEPTH; 1493 DEC_DEC_DEPTH;
1307 return 0; 1494 return 0;
1308} 1495}
1309 1496
1310static SV * 1497static SV *
1498decode_tag (dec_t *dec)
1499{
1500 SV *tag = 0;
1501 SV *val = 0;
1502
1503 if (!(dec->json.flags & F_ALLOW_TAGS))
1504 ERR ("malformed JSON string, neither array, object, number, string or atom");
1505
1506 ++dec->cur;
1507
1508 decode_ws (dec);
1509
1510 tag = decode_sv (dec);
1511 if (!tag)
1512 goto fail;
1513
1514 if (!SvPOK (tag))
1515 ERR ("malformed JSON string, (tag) must be a string");
1516
1517 decode_ws (dec);
1518
1519 if (*dec->cur != ')')
1520 ERR (") expected after tag");
1521
1522 ++dec->cur;
1523
1524 decode_ws (dec);
1525
1526 val = decode_sv (dec);
1527 if (!val)
1528 goto fail;
1529
1530 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1531 ERR ("malformed JSON string, tag value must be an array");
1532
1533 {
1534 AV *av = (AV *)SvRV (val);
1535 int i, len = av_len (av) + 1;
1536 HV *stash = gv_stashsv (tag, 0);
1537 SV *sv;
1538
1539 if (!stash)
1540 ERR ("cannot decode perl-object (package does not exist)");
1541
1542 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1543
1544 if (!method)
1545 ERR ("cannot decode perl-object (package does not have a THAW method)");
1546
1547 dSP;
1548
1549 ENTER; SAVETMPS; PUSHMARK (SP);
1550 EXTEND (SP, len + 2);
1551 // we re-bless the reference to get overload and other niceties right
1552 PUSHs (tag);
1553 PUSHs (sv_json);
1554
1555 for (i = 0; i < len; ++i)
1556 PUSHs (*av_fetch (av, i, 1));
1557
1558 PUTBACK;
1559 call_sv ((SV *)GvCV (method), G_SCALAR);
1560 SPAGAIN;
1561
1562 SvREFCNT_dec (tag);
1563 SvREFCNT_dec (val);
1564 sv = SvREFCNT_inc (POPs);
1565
1566 PUTBACK;
1567
1568 FREETMPS; LEAVE;
1569
1570 return sv;
1571 }
1572
1573fail:
1574 SvREFCNT_dec (tag);
1575 SvREFCNT_dec (val);
1576 return 0;
1577}
1578
1579static SV *
1311decode_sv (dec_t *dec) 1580decode_sv (dec_t *dec)
1312{ 1581{
1313 // the beauty of JSON: you need exactly one character lookahead 1582 // the beauty of JSON: you need exactly one character lookahead
1314 // to parse everything. 1583 // to parse everything.
1315 switch (*dec->cur) 1584 switch (*dec->cur)
1316 { 1585 {
1317 case '"': ++dec->cur; return decode_str (dec); 1586 case '"': ++dec->cur; return decode_str (dec);
1318 case '[': ++dec->cur; return decode_av (dec); 1587 case '[': ++dec->cur; return decode_av (dec);
1319 case '{': ++dec->cur; return decode_hv (dec); 1588 case '{': ++dec->cur; return decode_hv (dec);
1589 case '(': return decode_tag (dec);
1320 1590
1321 case '-': 1591 case '-':
1322 case '0': case '1': case '2': case '3': case '4': 1592 case '0': case '1': case '2': case '3': case '4':
1323 case '5': case '6': case '7': case '8': case '9': 1593 case '5': case '6': case '7': case '8': case '9':
1324 return decode_num (dec); 1594 return decode_num (dec);
1326 case 't': 1596 case 't':
1327 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1597 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1328 { 1598 {
1329 dec->cur += 4; 1599 dec->cur += 4;
1330#if JSON_SLOW 1600#if JSON_SLOW
1331 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1601 types_true = get_bool ("Types::Serialiser::true");
1332#endif 1602#endif
1333 return SvREFCNT_inc (json_true); 1603 return newSVsv (types_true);
1334 } 1604 }
1335 else 1605 else
1336 ERR ("'true' expected"); 1606 ERR ("'true' expected");
1337 1607
1338 break; 1608 break;
1340 case 'f': 1610 case 'f':
1341 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1611 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1342 { 1612 {
1343 dec->cur += 5; 1613 dec->cur += 5;
1344#if JSON_SLOW 1614#if JSON_SLOW
1345 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1615 types_false = get_bool ("Types::Serialiser::false");
1346#endif 1616#endif
1347 return SvREFCNT_inc (json_false); 1617 return newSVsv (types_false);
1348 } 1618 }
1349 else 1619 else
1350 ERR ("'false' expected"); 1620 ERR ("'false' expected");
1351 1621
1352 break; 1622 break;
1361 ERR ("'null' expected"); 1631 ERR ("'null' expected");
1362 1632
1363 break; 1633 break;
1364 1634
1365 default: 1635 default:
1366 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1636 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1367 break; 1637 break;
1368 } 1638 }
1369 1639
1370fail: 1640fail:
1371 return 0; 1641 return 0;
1372} 1642}
1373 1643
1374static SV * 1644static SV *
1375decode_json (SV *string, JSON *json, UV *offset_return) 1645decode_json (SV *string, JSON *json, char **offset_return)
1376{ 1646{
1377 dec_t dec; 1647 dec_t dec;
1378 UV offset;
1379 SV *sv; 1648 SV *sv;
1380 1649
1650 /* work around bugs in 5.10 where manipulating magic values
1651 * makes perl ignore the magic in subsequent accesses.
1652 * also make a copy of non-PV values, to get them into a clean
1653 * state (SvPV should do that, but it's buggy, see below).
1654 */
1381 SvGETMAGIC (string); 1655 /*SvGETMAGIC (string);*/
1656 if (SvMAGICAL (string) || !SvPOK (string))
1657 string = sv_2mortal (newSVsv (string));
1658
1382 SvUPGRADE (string, SVt_PV); 1659 SvUPGRADE (string, SVt_PV);
1383 1660
1384 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1661 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1662 * assertion with -DDEBUGGING, although SvCUR is documented to
1663 * return the xpv_cur field which certainly exists after upgrading.
1664 * according to nicholas clark, calling SvPOK fixes this.
1665 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1666 * and hope for the best.
1667 * Damnit, SvPV_nolen still trips over yet another assertion. This
1668 * assertion business is seriously broken, try yet another workaround
1669 * for the broken -DDEBUGGING.
1670 */
1671 {
1672#ifdef DEBUGGING
1673 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1674#else
1675 STRLEN offset = SvCUR (string);
1676#endif
1677
1678 if (offset > json->max_size && json->max_size)
1385 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1679 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1386 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1680 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1681 }
1387 1682
1388 if (json->flags & F_UTF8) 1683 if (DECODE_WANTS_OCTETS (json))
1389 sv_utf8_downgrade (string, 0); 1684 sv_utf8_downgrade (string, 0);
1390 else 1685 else
1391 sv_utf8_upgrade (string); 1686 sv_utf8_upgrade (string);
1392 1687
1393 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1688 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1394 1689
1395 dec.json = *json; 1690 dec.json = *json;
1396 dec.cur = SvPVX (string); 1691 dec.cur = SvPVX (string);
1397 dec.end = SvEND (string); 1692 dec.end = SvEND (string);
1398 dec.err = 0; 1693 dec.err = 0;
1399 dec.depth = 0; 1694 dec.depth = 0;
1400 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1401 1695
1402 if (dec.json.cb_object || dec.json.cb_sk_object) 1696 if (dec.json.cb_object || dec.json.cb_sk_object)
1403 dec.json.flags |= F_HOOK; 1697 dec.json.flags |= F_HOOK;
1404 1698
1405 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1699 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1406 1700
1407 decode_ws (&dec); 1701 decode_ws (&dec);
1408 sv = decode_sv (&dec); 1702 sv = decode_sv (&dec);
1703
1704 if (offset_return)
1705 *offset_return = dec.cur;
1409 1706
1410 if (!(offset_return || !sv)) 1707 if (!(offset_return || !sv))
1411 { 1708 {
1412 // check for trailing garbage 1709 // check for trailing garbage
1413 decode_ws (&dec); 1710 decode_ws (&dec);
1416 { 1713 {
1417 dec.err = "garbage after JSON object"; 1714 dec.err = "garbage after JSON object";
1418 SvREFCNT_dec (sv); 1715 SvREFCNT_dec (sv);
1419 sv = 0; 1716 sv = 0;
1420 } 1717 }
1421 }
1422
1423 if (offset_return || !sv)
1424 {
1425 offset = dec.json.flags & F_UTF8
1426 ? dec.cur - SvPVX (string)
1427 : utf8_distance (dec.cur, SvPVX (string));
1428
1429 if (offset_return)
1430 *offset_return = offset;
1431 } 1718 }
1432 1719
1433 if (!sv) 1720 if (!sv)
1434 { 1721 {
1435 SV *uni = sv_newmortal (); 1722 SV *uni = sv_newmortal ();
1441 SAVEVPTR (PL_curcop); 1728 SAVEVPTR (PL_curcop);
1442 PL_curcop = &cop; 1729 PL_curcop = &cop;
1443 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1730 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1444 LEAVE; 1731 LEAVE;
1445 1732
1446 croak ("%s, at character offset %d [\"%s\"]", 1733 croak ("%s, at character offset %d (before \"%s\")",
1447 dec.err, 1734 dec.err,
1448 (int)offset, 1735 (int)ptr_to_index (string, dec.cur),
1449 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1736 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1450 } 1737 }
1451 1738
1452 sv = sv_2mortal (sv); 1739 sv = sv_2mortal (sv);
1453 1740
1454 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1741 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1455 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1742 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1456 1743
1457 return sv; 1744 return sv;
1745}
1746
1747/////////////////////////////////////////////////////////////////////////////
1748// incremental parser
1749
1750static void
1751incr_parse (JSON *self)
1752{
1753 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1754
1755 // the state machine here is a bit convoluted and could be simplified a lot
1756 // but this would make it slower, so...
1757
1758 for (;;)
1759 {
1760 //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
1761 switch (self->incr_mode)
1762 {
1763 // only used for initial whitespace skipping
1764 case INCR_M_WS:
1765 for (;;)
1766 {
1767 if (*p > 0x20)
1768 {
1769 if (*p == '#')
1770 {
1771 self->incr_mode = INCR_M_C0;
1772 goto incr_m_c;
1773 }
1774 else
1775 {
1776 self->incr_mode = INCR_M_JSON;
1777 goto incr_m_json;
1778 }
1779 }
1780 else if (!*p)
1781 goto interrupt;
1782
1783 ++p;
1784 }
1785
1786 // skip a single char inside a string (for \\-processing)
1787 case INCR_M_BS:
1788 if (!*p)
1789 goto interrupt;
1790
1791 ++p;
1792 self->incr_mode = INCR_M_STR;
1793 goto incr_m_str;
1794
1795 // inside #-style comments
1796 case INCR_M_C0:
1797 case INCR_M_C1:
1798 incr_m_c:
1799 for (;;)
1800 {
1801 if (*p == '\n')
1802 {
1803 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1804 break;
1805 }
1806 else if (!*p)
1807 goto interrupt;
1808
1809 ++p;
1810 }
1811
1812 break;
1813
1814 // inside a string
1815 case INCR_M_STR:
1816 incr_m_str:
1817 for (;;)
1818 {
1819 if (*p == '"')
1820 {
1821 ++p;
1822 self->incr_mode = INCR_M_JSON;
1823
1824 if (!self->incr_nest)
1825 goto interrupt;
1826
1827 goto incr_m_json;
1828 }
1829 else if (*p == '\\')
1830 {
1831 ++p; // "virtually" consumes character after \
1832
1833 if (!*p) // if at end of string we have to switch modes
1834 {
1835 self->incr_mode = INCR_M_BS;
1836 goto interrupt;
1837 }
1838 }
1839 else if (!*p)
1840 goto interrupt;
1841
1842 ++p;
1843 }
1844
1845 // after initial ws, outside string
1846 case INCR_M_JSON:
1847 incr_m_json:
1848 for (;;)
1849 {
1850 switch (*p++)
1851 {
1852 case 0:
1853 --p;
1854 goto interrupt;
1855
1856 case 0x09:
1857 case 0x0a:
1858 case 0x0d:
1859 case 0x20:
1860 if (!self->incr_nest)
1861 {
1862 --p; // do not eat the whitespace, let the next round do it
1863 goto interrupt;
1864 }
1865 break;
1866
1867 case '"':
1868 self->incr_mode = INCR_M_STR;
1869 goto incr_m_str;
1870
1871 case '[':
1872 case '{':
1873 case '(':
1874 if (++self->incr_nest > self->max_depth)
1875 croak (ERR_NESTING_EXCEEDED);
1876 break;
1877
1878 case ']':
1879 case '}':
1880 if (--self->incr_nest <= 0)
1881 goto interrupt;
1882 break;
1883
1884 case ')':
1885 --self->incr_nest;
1886 break;
1887
1888 case '#':
1889 self->incr_mode = INCR_M_C1;
1890 goto incr_m_c;
1891 }
1892 }
1893 }
1894
1895 modechange:
1896 ;
1897 }
1898
1899interrupt:
1900 self->incr_pos = p - SvPVX (self->incr_text);
1901 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1902 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1458} 1903}
1459 1904
1460///////////////////////////////////////////////////////////////////////////// 1905/////////////////////////////////////////////////////////////////////////////
1461// XS interface functions 1906// XS interface functions
1462 1907
1471 i >= '0' && i <= '9' ? i - '0' 1916 i >= '0' && i <= '9' ? i - '0'
1472 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1917 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1473 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1918 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1474 : -1; 1919 : -1;
1475 1920
1476 json_stash = gv_stashpv ("JSON::XS" , 1); 1921 json_stash = gv_stashpv ("JSON::XS" , 1);
1477 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1922 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1478 1923
1479 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1924 types_true = get_bool ("Types::Serialiser::true");
1480 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1925 types_false = get_bool ("Types::Serialiser::false");
1926
1927 sv_json = newSVpv ("JSON", 0);
1928 SvREADONLY_on (sv_json);
1929
1930 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1481} 1931}
1482 1932
1483PROTOTYPES: DISABLE 1933PROTOTYPES: DISABLE
1484 1934
1485void CLONE (...) 1935void CLONE (...)
1486 CODE: 1936 CODE:
1487 json_stash = 0; 1937 json_stash = 0;
1488 json_boolean_stash = 0; 1938 types_boolean_stash = 0;
1489 1939
1490void new (char *klass) 1940void new (char *klass)
1491 PPCODE: 1941 PPCODE:
1492{ 1942{
1493 SV *pv = NEWSV (0, sizeof (JSON)); 1943 SV *pv = NEWSV (0, sizeof (JSON));
1494 SvPOK_only (pv); 1944 SvPOK_only (pv);
1495 Zero (SvPVX (pv), 1, JSON); 1945 json_init ((JSON *)SvPVX (pv));
1496 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1497 XPUSHs (sv_2mortal (sv_bless ( 1946 XPUSHs (sv_2mortal (sv_bless (
1498 newRV_noinc (pv), 1947 newRV_noinc (pv),
1499 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1948 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1500 ))); 1949 )));
1501} 1950}
1513 allow_nonref = F_ALLOW_NONREF 1962 allow_nonref = F_ALLOW_NONREF
1514 shrink = F_SHRINK 1963 shrink = F_SHRINK
1515 allow_blessed = F_ALLOW_BLESSED 1964 allow_blessed = F_ALLOW_BLESSED
1516 convert_blessed = F_CONV_BLESSED 1965 convert_blessed = F_CONV_BLESSED
1517 relaxed = F_RELAXED 1966 relaxed = F_RELAXED
1967 allow_unknown = F_ALLOW_UNKNOWN
1968 allow_tags = F_ALLOW_TAGS
1518 PPCODE: 1969 PPCODE:
1519{ 1970{
1520 if (enable) 1971 if (enable)
1521 self->flags |= ix; 1972 self->flags |= ix;
1522 else 1973 else
1537 get_allow_nonref = F_ALLOW_NONREF 1988 get_allow_nonref = F_ALLOW_NONREF
1538 get_shrink = F_SHRINK 1989 get_shrink = F_SHRINK
1539 get_allow_blessed = F_ALLOW_BLESSED 1990 get_allow_blessed = F_ALLOW_BLESSED
1540 get_convert_blessed = F_CONV_BLESSED 1991 get_convert_blessed = F_CONV_BLESSED
1541 get_relaxed = F_RELAXED 1992 get_relaxed = F_RELAXED
1993 get_allow_unknown = F_ALLOW_UNKNOWN
1994 get_allow_tags = F_ALLOW_TAGS
1542 PPCODE: 1995 PPCODE:
1543 XPUSHs (boolSV (self->flags & ix)); 1996 XPUSHs (boolSV (self->flags & ix));
1544 1997
1545void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1998void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1546 PPCODE: 1999 PPCODE:
1547{ 2000 self->max_depth = max_depth;
1548 UV log2 = 0;
1549
1550 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1551
1552 while ((1UL << log2) < max_depth)
1553 ++log2;
1554
1555 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1556
1557 XPUSHs (ST (0)); 2001 XPUSHs (ST (0));
1558}
1559 2002
1560U32 get_max_depth (JSON *self) 2003U32 get_max_depth (JSON *self)
1561 CODE: 2004 CODE:
1562 RETVAL = DEC_DEPTH (self->flags); 2005 RETVAL = self->max_depth;
1563 OUTPUT: 2006 OUTPUT:
1564 RETVAL 2007 RETVAL
1565 2008
1566void max_size (JSON *self, UV max_size = 0) 2009void max_size (JSON *self, U32 max_size = 0)
1567 PPCODE: 2010 PPCODE:
1568{ 2011 self->max_size = max_size;
1569 UV log2 = 0;
1570
1571 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1572 if (max_size == 1) max_size = 2;
1573
1574 while ((1UL << log2) < max_size)
1575 ++log2;
1576
1577 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1578
1579 XPUSHs (ST (0)); 2012 XPUSHs (ST (0));
1580}
1581 2013
1582int get_max_size (JSON *self) 2014int get_max_size (JSON *self)
1583 CODE: 2015 CODE:
1584 RETVAL = DEC_SIZE (self->flags); 2016 RETVAL = self->max_size;
1585 OUTPUT: 2017 OUTPUT:
1586 RETVAL 2018 RETVAL
1587 2019
1588void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2020void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1589 PPCODE: 2021 PPCODE:
1595} 2027}
1596 2028
1597void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2029void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1598 PPCODE: 2030 PPCODE:
1599{ 2031{
1600 if (!self->cb_sk_object) 2032 if (!self->cb_sk_object)
1601 self->cb_sk_object = newHV (); 2033 self->cb_sk_object = newHV ();
1602 2034
1603 if (SvOK (cb)) 2035 if (SvOK (cb))
1604 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2036 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1605 else 2037 else
1616 XPUSHs (ST (0)); 2048 XPUSHs (ST (0));
1617} 2049}
1618 2050
1619void encode (JSON *self, SV *scalar) 2051void encode (JSON *self, SV *scalar)
1620 PPCODE: 2052 PPCODE:
1621 XPUSHs (encode_json (scalar, self)); 2053 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2054 XPUSHs (scalar);
1622 2055
1623void decode (JSON *self, SV *jsonstr) 2056void decode (JSON *self, SV *jsonstr)
1624 PPCODE: 2057 PPCODE:
1625 XPUSHs (decode_json (jsonstr, self, 0)); 2058 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2059 XPUSHs (jsonstr);
1626 2060
1627void decode_prefix (JSON *self, SV *jsonstr) 2061void decode_prefix (JSON *self, SV *jsonstr)
1628 PPCODE: 2062 PPCODE:
1629{ 2063{
1630 UV offset; 2064 SV *sv;
2065 char *offset;
2066 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1631 EXTEND (SP, 2); 2067 EXTEND (SP, 2);
1632 PUSHs (decode_json (jsonstr, self, &offset)); 2068 PUSHs (sv);
1633 PUSHs (sv_2mortal (newSVuv (offset))); 2069 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
2070}
2071
2072void incr_parse (JSON *self, SV *jsonstr = 0)
2073 PPCODE:
2074{
2075 if (!self->incr_text)
2076 self->incr_text = newSVpvn ("", 0);
2077
2078 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2079 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2080 if (DECODE_WANTS_OCTETS (self))
2081 {
2082 if (self->incr_pos)
2083 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2084 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2085
2086 sv_utf8_downgrade (self->incr_text, 0);
2087 }
2088 else
2089 {
2090 sv_utf8_upgrade (self->incr_text);
2091
2092 if (self->incr_pos)
2093 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2094 - (U8 *)SvPVX (self->incr_text);
2095 }
2096
2097 // append data, if any
2098 if (jsonstr)
2099 {
2100 /* make sure both strings have same encoding */
2101 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2102 if (SvUTF8 (jsonstr))
2103 sv_utf8_downgrade (jsonstr, 0);
2104 else
2105 sv_utf8_upgrade (jsonstr);
2106
2107 /* and then just blindly append */
2108 {
2109 STRLEN len;
2110 const char *str = SvPV (jsonstr, len);
2111 STRLEN cur = SvCUR (self->incr_text);
2112
2113 if (SvLEN (self->incr_text) <= cur + len)
2114 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2115
2116 Move (str, SvEND (self->incr_text), len, char);
2117 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2118 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2119 }
2120 }
2121
2122 if (GIMME_V != G_VOID)
2123 do
2124 {
2125 SV *sv;
2126 char *offset;
2127
2128 if (!INCR_DONE (self))
2129 {
2130 incr_parse (self);
2131
2132 if (self->incr_pos > self->max_size && self->max_size)
2133 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2134 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2135
2136 if (!INCR_DONE (self))
2137 {
2138 // as an optimisation, do not accumulate white space in the incr buffer
2139 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2140 {
2141 self->incr_pos = 0;
2142 SvCUR_set (self->incr_text, 0);
2143 }
2144
2145 break;
2146 }
2147 }
2148
2149 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2150 XPUSHs (sv);
2151
2152 self->incr_pos -= offset - SvPVX (self->incr_text);
2153 self->incr_nest = 0;
2154 self->incr_mode = 0;
2155
2156 sv_chop (self->incr_text, offset);
2157 }
2158 while (GIMME_V == G_ARRAY);
2159}
2160
2161SV *incr_text (JSON *self)
2162 ATTRS: lvalue
2163 CODE:
2164{
2165 if (self->incr_pos)
2166 croak ("incr_text can not be called when the incremental parser already started parsing");
2167
2168 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2169}
2170 OUTPUT:
2171 RETVAL
2172
2173void incr_skip (JSON *self)
2174 CODE:
2175{
2176 if (self->incr_pos)
2177 {
2178 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2179 self->incr_pos = 0;
2180 self->incr_nest = 0;
2181 self->incr_mode = 0;
2182 }
2183}
2184
2185void incr_reset (JSON *self)
2186 CODE:
2187{
2188 SvREFCNT_dec (self->incr_text);
2189 self->incr_text = 0;
2190 self->incr_pos = 0;
2191 self->incr_nest = 0;
2192 self->incr_mode = 0;
1634} 2193}
1635 2194
1636void DESTROY (JSON *self) 2195void DESTROY (JSON *self)
1637 CODE: 2196 CODE:
1638 SvREFCNT_dec (self->cb_sk_object); 2197 SvREFCNT_dec (self->cb_sk_object);
1639 SvREFCNT_dec (self->cb_object); 2198 SvREFCNT_dec (self->cb_object);
2199 SvREFCNT_dec (self->incr_text);
1640 2200
1641PROTOTYPES: ENABLE 2201PROTOTYPES: ENABLE
1642 2202
1643void encode_json (SV *scalar) 2203void encode_json (SV *scalar)
1644 PPCODE: 2204 PPCODE:
1645{ 2205{
1646 JSON json = { F_DEFAULT | F_UTF8 }; 2206 JSON json;
1647 XPUSHs (encode_json (scalar, &json)); 2207 json_init (&json);
2208 json.flags |= F_UTF8;
2209 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2210 XPUSHs (scalar);
1648} 2211}
1649 2212
1650void decode_json (SV *jsonstr) 2213void decode_json (SV *jsonstr)
1651 PPCODE: 2214 PPCODE:
1652{ 2215{
1653 JSON json = { F_DEFAULT | F_UTF8 }; 2216 JSON json;
1654 XPUSHs (decode_json (jsonstr, &json, 0)); 2217 json_init (&json);
2218 json.flags |= F_UTF8;
2219 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2220 XPUSHs (jsonstr);
1655} 2221}
1656 2222

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines