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 1.78 by root, Tue Mar 25 19:31:02 2008 UTC vs.
Revision 1.120 by root, Mon Oct 28 23:23:10 2013 UTC

12#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 14#endif
15 15
16// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
20#endif 20#endif
21 21
22// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 24
24#define F_ASCII 0x00000001UL 25#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 26#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 27#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 28#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 32#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL 33#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 34#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 35#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 36#define F_RELAXED 0x00001000UL
36 37#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL 38#define F_ALLOW_TAGS 0x00004000UL
38#define S_MAXDEPTH 27
39#define F_MAXSIZE 0x01f00000UL
40#define S_MAXSIZE 20
41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 39#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 40
43#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
44#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
45
46#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 41#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 42
49#define INIT_SIZE 32 // initial scalar size to be allocated 43#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 44#define INDENT_STEP 3 // spaces per indentation level
51 45
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 46#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
47
48#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
53 49
54#define SB do { 50#define SB do {
55#define SE } while (0) 51#define SE } while (0)
56 52
57#if __GNUC__ >= 3 53#if __GNUC__ >= 3
67 63
68#define IN_RANGE_INC(type,val,beg,end) \ 64#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 65 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 66 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 67
68#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
69
72#ifdef USE_ITHREADS 70#ifdef USE_ITHREADS
73# define JSON_SLOW 1 71# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 72# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
75#else 73#else
76# define JSON_SLOW 0 74# define JSON_SLOW 0
77# define JSON_STASH json_stash 75# define JSON_STASH json_stash
78#endif 76#endif
79 77
78// the amount of HEs to allocate on the stack, when sorting keys
79#define STACK_HES 64
80
80static HV *json_stash, *json_boolean_stash; // JSON::XS:: 81static HV *json_stash, *types_boolean_stash; // JSON::XS::
81static SV *json_true, *json_false; 82static SV *types_true, *types_false, *sv_json;
82 83
83enum { 84enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0 85 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string 86 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash 87 INCR_M_BS, // inside backslash
88 INCR_M_C0, // inside comment in initial whitespace sequence
89 INCR_M_C1, // inside comment in other places
87 INCR_M_JSON // outside anything, count nesting 90 INCR_M_JSON // outside anything, count nesting
88}; 91};
89 92
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 93#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
91 94
92typedef struct { 95typedef struct {
93 U32 flags; 96 U32 flags;
97 U32 max_depth;
98 STRLEN max_size;
99
94 SV *cb_object; 100 SV *cb_object;
95 HV *cb_sk_object; 101 HV *cb_sk_object;
96 102
97 // for the incremental parser 103 // for the incremental parser
98 SV *incr_text; // the source text so far 104 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 105 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 106 int incr_nest; // {[]}-nesting level
101 int incr_mode; 107 unsigned char incr_mode;
102} JSON; 108} JSON;
109
110INLINE void
111json_init (JSON *json)
112{
113 Zero (json, 1, JSON);
114 json->max_depth = 512;
115}
103 116
104///////////////////////////////////////////////////////////////////////////// 117/////////////////////////////////////////////////////////////////////////////
105// utility functions 118// utility functions
106 119
120INLINE SV *
121get_bool (const char *name)
122{
123 SV *sv = get_sv (name, 1);
124
125 SvREADONLY_on (sv);
126 SvREADONLY_on (SvRV (sv));
127
128 return sv;
129}
130
107INLINE void 131INLINE void
108shrink (SV *sv) 132shrink (SV *sv)
109{ 133{
110 sv_utf8_downgrade (sv, 1); 134 sv_utf8_downgrade (sv, 1);
135
111 if (SvLEN (sv) > SvCUR (sv) + 1) 136 if (SvLEN (sv) > SvCUR (sv) + 1)
112 { 137 {
113#ifdef SvPV_shrink_to_cur 138#ifdef SvPV_shrink_to_cur
114 SvPV_shrink_to_cur (sv); 139 SvPV_shrink_to_cur (sv);
115#elif defined (SvPV_renew) 140#elif defined (SvPV_renew)
159 *s++ = 0x80 | ( ch & 0x3f); 184 *s++ = 0x80 | ( ch & 0x3f);
160 185
161 return s; 186 return s;
162} 187}
163 188
189// convert offset pointer to character index, sv must be string
190static STRLEN
191ptr_to_index (SV *sv, char *offset)
192{
193 return SvUTF8 (sv)
194 ? utf8_distance (offset, SvPVX (sv))
195 : offset - SvPVX (sv);
196}
197
198/////////////////////////////////////////////////////////////////////////////
199// fp hell
200
201// scan a group of digits, and a trailing exponent
202static void
203json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
204{
205 UV uaccum = 0;
206 int eaccum = 0;
207
208 // if we recurse too deep, skip all remaining digits
209 // to avoid a stack overflow attack
210 if (expect_false (--maxdepth <= 0))
211 while (((U8)*s - '0') < 10)
212 ++s;
213
214 for (;;)
215 {
216 U8 dig = (U8)*s - '0';
217
218 if (expect_false (dig >= 10))
219 {
220 if (dig == (U8)((U8)'.' - (U8)'0'))
221 {
222 ++s;
223 json_atof_scan1 (s, accum, expo, 1, maxdepth);
224 }
225 else if ((dig | ' ') == 'e' - '0')
226 {
227 int exp2 = 0;
228 int neg = 0;
229
230 ++s;
231
232 if (*s == '-')
233 {
234 ++s;
235 neg = 1;
236 }
237 else if (*s == '+')
238 ++s;
239
240 while ((dig = (U8)*s - '0') < 10)
241 exp2 = exp2 * 10 + *s++ - '0';
242
243 *expo += neg ? -exp2 : exp2;
244 }
245
246 break;
247 }
248
249 ++s;
250
251 uaccum = uaccum * 10 + dig;
252 ++eaccum;
253
254 // if we have too many digits, then recurse for more
255 // we actually do this for rather few digits
256 if (uaccum >= (UV_MAX - 9) / 10)
257 {
258 if (postdp) *expo -= eaccum;
259 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
260 if (postdp) *expo += eaccum;
261
262 break;
263 }
264 }
265
266 // this relies greatly on the quality of the pow ()
267 // implementation of the platform, but a good
268 // implementation is hard to beat.
269 // (IEEE 754 conformant ones are required to be exact)
270 if (postdp) *expo -= eaccum;
271 *accum += uaccum * Perl_pow (10., *expo);
272 *expo += eaccum;
273}
274
275static NV
276json_atof (const char *s)
277{
278 NV accum = 0.;
279 int expo = 0;
280 int neg = 0;
281
282 if (*s == '-')
283 {
284 ++s;
285 neg = 1;
286 }
287
288 // a recursion depth of ten gives us >>500 bits
289 json_atof_scan1 (s, &accum, &expo, 0, 10);
290
291 return neg ? -accum : accum;
292}
164///////////////////////////////////////////////////////////////////////////// 293/////////////////////////////////////////////////////////////////////////////
165// encoder 294// encoder
166 295
167// structure used for encoding JSON 296// structure used for encoding JSON
168typedef struct 297typedef struct
170 char *cur; // SvPVX (sv) + current output position 299 char *cur; // SvPVX (sv) + current output position
171 char *end; // SvEND (sv) 300 char *end; // SvEND (sv)
172 SV *sv; // result scalar 301 SV *sv; // result scalar
173 JSON json; 302 JSON json;
174 U32 indent; // indentation level 303 U32 indent; // indentation level
175 U32 maxdepth; // max. indentation/recursion level
176 UV limit; // escape character values >= this value when encoding 304 UV limit; // escape character values >= this value when encoding
177} enc_t; 305} enc_t;
178 306
179INLINE void 307INLINE void
180need (enc_t *enc, STRLEN len) 308need (enc_t *enc, STRLEN len)
181{ 309{
182 if (expect_false (enc->cur + len >= enc->end)) 310 if (expect_false (enc->cur + len >= enc->end))
183 { 311 {
184 STRLEN cur = enc->cur - SvPVX (enc->sv); 312 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
185 SvGROW (enc->sv, cur + len + 1); 313 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
186 enc->cur = SvPVX (enc->sv) + cur; 314 enc->cur = SvPVX (enc->sv) + cur;
187 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 315 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
188 } 316 }
189} 317}
190 318
265 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 393 (int)((uch - 0x10000) % 0x400 + 0xDC00));
266 enc->cur += 12; 394 enc->cur += 12;
267 } 395 }
268 else 396 else
269 { 397 {
270 static char hexdigit [16] = "0123456789abcdef";
271 need (enc, len += 5); 398 need (enc, len += 5);
272 *enc->cur++ = '\\'; 399 *enc->cur++ = '\\';
273 *enc->cur++ = 'u'; 400 *enc->cur++ = 'u';
274 *enc->cur++ = hexdigit [ uch >> 12 ]; 401 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
275 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 402 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
276 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 403 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
277 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 404 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
278 } 405 }
279 406
280 str += clen; 407 str += clen;
281 } 408 }
282 else if (enc->json.flags & F_LATIN1) 409 else if (enc->json.flags & F_LATIN1)
353static void 480static void
354encode_av (enc_t *enc, AV *av) 481encode_av (enc_t *enc, AV *av)
355{ 482{
356 int i, len = av_len (av); 483 int i, len = av_len (av);
357 484
358 if (enc->indent >= enc->maxdepth) 485 if (enc->indent >= enc->json.max_depth)
359 croak ("data structure too deep (hit recursion limit)"); 486 croak (ERR_NESTING_EXCEEDED);
360 487
361 encode_ch (enc, '['); 488 encode_ch (enc, '[');
362 489
363 if (len >= 0) 490 if (len >= 0)
364 { 491 {
365 encode_nl (enc); ++enc->indent; 492 encode_nl (enc); ++enc->indent;
366 493
367 for (i = 0; i <= len; ++i) 494 for (i = 0; i <= len; ++i)
379 encode_comma (enc); 506 encode_comma (enc);
380 } 507 }
381 508
382 encode_nl (enc); --enc->indent; encode_indent (enc); 509 encode_nl (enc); --enc->indent; encode_indent (enc);
383 } 510 }
384 511
385 encode_ch (enc, ']'); 512 encode_ch (enc, ']');
386} 513}
387 514
388static void 515static void
389encode_hk (enc_t *enc, HE *he) 516encode_hk (enc_t *enc, HE *he)
393 if (HeKLEN (he) == HEf_SVKEY) 520 if (HeKLEN (he) == HEf_SVKEY)
394 { 521 {
395 SV *sv = HeSVKEY (he); 522 SV *sv = HeSVKEY (he);
396 STRLEN len; 523 STRLEN len;
397 char *str; 524 char *str;
398 525
399 SvGETMAGIC (sv); 526 SvGETMAGIC (sv);
400 str = SvPV (sv, len); 527 str = SvPV (sv, len);
401 528
402 encode_str (enc, str, len, SvUTF8 (sv)); 529 encode_str (enc, str, len, SvUTF8 (sv));
403 } 530 }
439static void 566static void
440encode_hv (enc_t *enc, HV *hv) 567encode_hv (enc_t *enc, HV *hv)
441{ 568{
442 HE *he; 569 HE *he;
443 570
444 if (enc->indent >= enc->maxdepth) 571 if (enc->indent >= enc->json.max_depth)
445 croak ("data structure too deep (hit recursion limit)"); 572 croak (ERR_NESTING_EXCEEDED);
446 573
447 encode_ch (enc, '{'); 574 encode_ch (enc, '{');
448 575
449 // for canonical output we have to sort by keys first 576 // for canonical output we have to sort by keys first
450 // actually, this is mostly due to the stupid so-called 577 // actually, this is mostly due to the stupid so-called
451 // security workaround added somewhere in 5.8.x. 578 // security workaround added somewhere in 5.8.x
452 // that randomises hash orderings 579 // that randomises hash orderings
453 if (enc->json.flags & F_CANONICAL) 580 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
454 { 581 {
455 int count = hv_iterinit (hv); 582 int count = hv_iterinit (hv);
456 583
457 if (SvMAGICAL (hv)) 584 if (SvMAGICAL (hv))
458 { 585 {
469 } 596 }
470 597
471 if (count) 598 if (count)
472 { 599 {
473 int i, fast = 1; 600 int i, fast = 1;
474#if defined(__BORLANDC__) || defined(_MSC_VER) 601 HE *hes_stack [STACK_HES];
475 HE **hes = _alloca (count * sizeof (HE)); 602 HE **hes = hes_stack;
476#else 603
477 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 604 // allocate larger arrays on the heap
478#endif 605 if (count > STACK_HES)
606 {
607 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
608 hes = (HE **)SvPVX (sv);
609 }
479 610
480 i = 0; 611 i = 0;
481 while ((he = hv_iternext (hv))) 612 while ((he = hv_iternext (hv)))
482 { 613 {
483 hes [i++] = he; 614 hes [i++] = he;
552// encode objects, arrays and special \0=false and \1=true values. 683// encode objects, arrays and special \0=false and \1=true values.
553static void 684static void
554encode_rv (enc_t *enc, SV *sv) 685encode_rv (enc_t *enc, SV *sv)
555{ 686{
556 svtype svt; 687 svtype svt;
688 GV *method;
557 689
558 SvGETMAGIC (sv); 690 SvGETMAGIC (sv);
559 svt = SvTYPE (sv); 691 svt = SvTYPE (sv);
560 692
561 if (expect_false (SvOBJECT (sv))) 693 if (expect_false (SvOBJECT (sv)))
562 { 694 {
563 HV *stash = !JSON_SLOW || json_boolean_stash 695 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
564 ? json_boolean_stash 696 ? types_boolean_stash
565 : gv_stashpv ("JSON::XS::Boolean", 1); 697 : gv_stashpv ("Types::Serialiser::Boolean", 1);
698 HV *stash = SvSTASH (sv);
566 699
567 if (SvSTASH (sv) == stash) 700 if (stash == boolean_stash)
568 { 701 {
569 if (SvIV (sv)) 702 if (SvIV (sv))
570 encode_str (enc, "true", 4, 0); 703 encode_str (enc, "true", 4, 0);
571 else 704 else
572 encode_str (enc, "false", 5, 0); 705 encode_str (enc, "false", 5, 0);
573 } 706 }
707 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
708 {
709 int count;
710 dSP;
711
712 ENTER; SAVETMPS; PUSHMARK (SP);
713 EXTEND (SP, 2);
714 // we re-bless the reference to get overload and other niceties right
715 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
716 PUSHs (sv_json);
717
718 PUTBACK;
719 count = call_sv ((SV *)GvCV (method), G_ARRAY);
720 SPAGAIN;
721
722 // catch this surprisingly common error
723 if (SvROK (TOPs) && SvRV (TOPs) == sv)
724 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
725
726 encode_ch (enc, '(');
727 encode_ch (enc, '"');
728 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
729 encode_ch (enc, '"');
730 encode_ch (enc, ')');
731 encode_ch (enc, '[');
732
733 while (count)
734 {
735 encode_sv (enc, SP[1 - count--]);
736
737 if (count)
738 encode_ch (enc, ',');
739 }
740
741 encode_ch (enc, ']');
742
743 PUTBACK;
744
745 FREETMPS; LEAVE;
746 }
747 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
748 {
749 dSP;
750
751 ENTER; SAVETMPS; PUSHMARK (SP);
752 // we re-bless the reference to get overload and other niceties right
753 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
754
755 // calling with G_SCALAR ensures that we always get a 1 return value
756 PUTBACK;
757 call_sv ((SV *)GvCV (method), G_SCALAR);
758 SPAGAIN;
759
760 // catch this surprisingly common error
761 if (SvROK (TOPs) && SvRV (TOPs) == sv)
762 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
763
764 sv = POPs;
765 PUTBACK;
766
767 encode_sv (enc, sv);
768
769 FREETMPS; LEAVE;
770 }
771 else if (enc->json.flags & F_ALLOW_BLESSED)
772 encode_str (enc, "null", 4, 0);
574 else 773 else
575 {
576#if 0
577 if (0 && sv_derived_from (rv, "JSON::Literal"))
578 {
579 // not yet
580 }
581#endif
582 if (enc->json.flags & F_CONV_BLESSED)
583 {
584 // we re-bless the reference to get overload and other niceties right
585 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
586
587 if (to_json)
588 {
589 dSP;
590
591 ENTER; SAVETMPS; PUSHMARK (SP);
592 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
593
594 // calling with G_SCALAR ensures that we always get a 1 return value
595 PUTBACK;
596 call_sv ((SV *)GvCV (to_json), G_SCALAR);
597 SPAGAIN;
598
599 // catch this surprisingly common error
600 if (SvROK (TOPs) && SvRV (TOPs) == sv)
601 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
602
603 sv = POPs;
604 PUTBACK;
605
606 encode_sv (enc, sv);
607
608 FREETMPS; LEAVE;
609 }
610 else if (enc->json.flags & F_ALLOW_BLESSED)
611 encode_str (enc, "null", 4, 0);
612 else
613 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
614 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
615 }
616 else if (enc->json.flags & F_ALLOW_BLESSED)
617 encode_str (enc, "null", 4, 0);
618 else
619 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 774 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
620 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 775 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
621 }
622 } 776 }
623 else if (svt == SVt_PVHV) 777 else if (svt == SVt_PVHV)
624 encode_hv (enc, (HV *)sv); 778 encode_hv (enc, (HV *)sv);
625 else if (svt == SVt_PVAV) 779 else if (svt == SVt_PVAV)
626 encode_av (enc, (AV *)sv); 780 encode_av (enc, (AV *)sv);
631 785
632 if (len == 1 && *pv == '1') 786 if (len == 1 && *pv == '1')
633 encode_str (enc, "true", 4, 0); 787 encode_str (enc, "true", 4, 0);
634 else if (len == 1 && *pv == '0') 788 else if (len == 1 && *pv == '0')
635 encode_str (enc, "false", 5, 0); 789 encode_str (enc, "false", 5, 0);
790 else if (enc->json.flags & F_ALLOW_UNKNOWN)
791 encode_str (enc, "null", 4, 0);
636 else 792 else
637 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 793 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
638 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 794 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
639 } 795 }
796 else if (enc->json.flags & F_ALLOW_UNKNOWN)
797 encode_str (enc, "null", 4, 0);
640 else 798 else
641 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 799 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
642 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 800 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
643} 801}
644 802
698 } 856 }
699 else 857 else
700 { 858 {
701 // large integer, use the (rather slow) snprintf way. 859 // large integer, use the (rather slow) snprintf way.
702 need (enc, IVUV_MAXCHARS); 860 need (enc, IVUV_MAXCHARS);
703 enc->cur += 861 enc->cur +=
704 SvIsUV(sv) 862 SvIsUV(sv)
705 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 863 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
706 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 864 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
707 } 865 }
708 } 866 }
709 else if (SvROK (sv)) 867 else if (SvROK (sv))
710 encode_rv (enc, SvRV (sv)); 868 encode_rv (enc, SvRV (sv));
711 else if (!SvOK (sv)) 869 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
712 encode_str (enc, "null", 4, 0); 870 encode_str (enc, "null", 4, 0);
713 else 871 else
714 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 872 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
715 SvPV_nolen (sv), SvFLAGS (sv)); 873 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
716} 874}
717 875
718static SV * 876static SV *
719encode_json (SV *scalar, JSON *json) 877encode_json (SV *scalar, JSON *json)
720{ 878{
726 enc.json = *json; 884 enc.json = *json;
727 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 885 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
728 enc.cur = SvPVX (enc.sv); 886 enc.cur = SvPVX (enc.sv);
729 enc.end = SvEND (enc.sv); 887 enc.end = SvEND (enc.sv);
730 enc.indent = 0; 888 enc.indent = 0;
731 enc.maxdepth = DEC_DEPTH (enc.json.flags);
732 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 889 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
733 : enc.json.flags & F_LATIN1 ? 0x000100UL 890 : enc.json.flags & F_LATIN1 ? 0x000100UL
734 : 0x110000UL; 891 : 0x110000UL;
735 892
736 SvPOK_only (enc.sv); 893 SvPOK_only (enc.sv);
737 encode_sv (&enc, scalar); 894 encode_sv (&enc, scalar);
895 encode_nl (&enc);
738 896
739 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 897 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
740 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 898 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
741 899
742 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 900 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
803 if (*dec->cur != ch) \ 961 if (*dec->cur != ch) \
804 ERR (# ch " expected"); \ 962 ERR (# ch " expected"); \
805 ++dec->cur; \ 963 ++dec->cur; \
806 SE 964 SE
807 965
808#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 966#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
809#define DEC_DEC_DEPTH --dec->depth 967#define DEC_DEC_DEPTH --dec->depth
810 968
811static SV *decode_sv (dec_t *dec); 969static SV *decode_sv (dec_t *dec);
812 970
813static signed char decode_hexdigit[256]; 971static signed char decode_hexdigit[256];
922 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1080 else if (expect_true (ch >= 0x20 && ch < 0x80))
923 *cur++ = ch; 1081 *cur++ = ch;
924 else if (ch >= 0x80) 1082 else if (ch >= 0x80)
925 { 1083 {
926 STRLEN clen; 1084 STRLEN clen;
927 UV uch;
928 1085
929 --dec_cur; 1086 --dec_cur;
930 1087
931 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1088 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
932 if (clen == (STRLEN)-1) 1089 if (clen == (STRLEN)-1)
933 ERR ("malformed UTF-8 character in JSON string"); 1090 ERR ("malformed UTF-8 character in JSON string");
934 1091
935 do 1092 do
936 *cur++ = *dec_cur++; 1093 *cur++ = *dec_cur++;
953 { 1110 {
954 STRLEN len = cur - buf; 1111 STRLEN len = cur - buf;
955 1112
956 if (sv) 1113 if (sv)
957 { 1114 {
958 SvGROW (sv, SvCUR (sv) + len + 1); 1115 STRLEN cur = SvCUR (sv);
1116
1117 if (SvLEN (sv) <= cur + len)
1118 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1119
959 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1120 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
960 SvCUR_set (sv, SvCUR (sv) + len); 1121 SvCUR_set (sv, SvCUR (sv) + len);
961 } 1122 }
962 else 1123 else
963 sv = newSVpvn (buf, len); 1124 sv = newSVpvn (buf, len);
1054 1215
1055 // special case the rather common 1..5-digit-int case 1216 // special case the rather common 1..5-digit-int case
1056 if (*start == '-') 1217 if (*start == '-')
1057 switch (len) 1218 switch (len)
1058 { 1219 {
1059 case 2: return newSViv (-( start [1] - '0' * 1)); 1220 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1060 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1221 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1061 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1222 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1062 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1223 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1063 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1224 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1064 } 1225 }
1065 else 1226 else
1066 switch (len) 1227 switch (len)
1067 { 1228 {
1068 case 1: return newSViv ( start [0] - '0' * 1); 1229 case 1: return newSViv ( start [0] - '0' * 1);
1069 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1230 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1070 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1231 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1071 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1232 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1072 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1233 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1073 } 1234 }
1074 1235
1075 { 1236 {
1076 UV uv; 1237 UV uv;
1077 int numtype = grok_number (start, len, &uv); 1238 int numtype = grok_number (start, len, &uv);
1086 } 1247 }
1087 1248
1088 len -= *start == '-' ? 1 : 0; 1249 len -= *start == '-' ? 1 : 0;
1089 1250
1090 // does not fit into IV or UV, try NV 1251 // does not fit into IV or UV, try NV
1091 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1252 if (len <= NV_DIG)
1092 #if defined (LDBL_DIG)
1093 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1094 #endif
1095 )
1096 // fits into NV without loss of precision 1253 // fits into NV without loss of precision
1097 return newSVnv (Atof (start)); 1254 return newSVnv (json_atof (start));
1098 1255
1099 // everything else fails, convert it to a string 1256 // everything else fails, convert it to a string
1100 return newSVpvn (start, dec->cur - start); 1257 return newSVpvn (start, dec->cur - start);
1101 } 1258 }
1102 1259
1103 // loss of precision here 1260 // loss of precision here
1104 return newSVnv (Atof (start)); 1261 return newSVnv (json_atof (start));
1105 1262
1106fail: 1263fail:
1107 return 0; 1264 return 0;
1108} 1265}
1109 1266
1133 if (*dec->cur == ']') 1290 if (*dec->cur == ']')
1134 { 1291 {
1135 ++dec->cur; 1292 ++dec->cur;
1136 break; 1293 break;
1137 } 1294 }
1138 1295
1139 if (*dec->cur != ',') 1296 if (*dec->cur != ',')
1140 ERR (", or ] expected while parsing array"); 1297 ERR (", or ] expected while parsing array");
1141 1298
1142 ++dec->cur; 1299 ++dec->cur;
1143 1300
1278 dSP; 1435 dSP;
1279 int count; 1436 int count;
1280 1437
1281 ENTER; SAVETMPS; PUSHMARK (SP); 1438 ENTER; SAVETMPS; PUSHMARK (SP);
1282 XPUSHs (HeVAL (he)); 1439 XPUSHs (HeVAL (he));
1440 sv_2mortal (sv);
1283 1441
1284 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1442 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1285 1443
1286 if (count == 1) 1444 if (count == 1)
1287 { 1445 {
1288 sv = newSVsv (POPs); 1446 sv = newSVsv (POPs);
1289 FREETMPS; LEAVE; 1447 FREETMPS; LEAVE;
1290 return sv; 1448 return sv;
1291 } 1449 }
1292 1450
1451 SvREFCNT_inc (sv);
1293 FREETMPS; LEAVE; 1452 FREETMPS; LEAVE;
1294 } 1453 }
1295 } 1454 }
1296 1455
1297 if (dec->json.cb_object) 1456 if (dec->json.cb_object)
1323 DEC_DEC_DEPTH; 1482 DEC_DEC_DEPTH;
1324 return 0; 1483 return 0;
1325} 1484}
1326 1485
1327static SV * 1486static SV *
1487decode_tag (dec_t *dec)
1488{
1489 SV *tag = 0;
1490 SV *val = 0;
1491
1492 if (!(dec->json.flags & F_ALLOW_TAGS))
1493 ERR ("malformed JSON string, neither array, object, number, string or atom");
1494
1495 ++dec->cur;
1496
1497 tag = decode_sv (dec);
1498 if (!tag)
1499 goto fail;
1500
1501 if (!SvPOK (tag))
1502 ERR ("malformed JSON string, (tag) must be a string");
1503
1504 if (*dec->cur != ')')
1505 ERR (") expected after tag");
1506
1507 ++dec->cur;
1508
1509 val = decode_sv (dec);
1510 if (!val)
1511 goto fail;
1512
1513 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1514 ERR ("malformed JSON string, tag value must be an array");
1515
1516 {
1517 AV *av = (AV *)SvRV (val);
1518 int i, len = av_len (av) + 1;
1519 HV *stash = gv_stashsv (tag, 0);
1520 SV *sv;
1521
1522 if (!stash)
1523 ERR ("cannot decode perl-object (package does not exist)");
1524
1525 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1526
1527 if (!method)
1528 ERR ("cannot decode perl-object (package does not have a THAW method)");
1529
1530 dSP;
1531
1532 ENTER; SAVETMPS; PUSHMARK (SP);
1533 EXTEND (SP, len + 2);
1534 // we re-bless the reference to get overload and other niceties right
1535 PUSHs (tag);
1536 PUSHs (sv_json);
1537
1538 for (i = 0; i < len; ++i)
1539 PUSHs (*av_fetch (av, i, 1));
1540
1541 PUTBACK;
1542 call_sv ((SV *)GvCV (method), G_SCALAR);
1543 SPAGAIN;
1544
1545 SvREFCNT_dec (tag);
1546 SvREFCNT_dec (val);
1547 sv = SvREFCNT_inc (POPs);
1548
1549 PUTBACK;
1550
1551 FREETMPS; LEAVE;
1552
1553 return sv;
1554 }
1555
1556fail:
1557 SvREFCNT_dec (tag);
1558 SvREFCNT_dec (val);
1559 return 0;
1560}
1561
1562static SV *
1328decode_sv (dec_t *dec) 1563decode_sv (dec_t *dec)
1329{ 1564{
1330 // the beauty of JSON: you need exactly one character lookahead 1565 // the beauty of JSON: you need exactly one character lookahead
1331 // to parse everything. 1566 // to parse everything.
1332 switch (*dec->cur) 1567 switch (*dec->cur)
1333 { 1568 {
1334 case '"': ++dec->cur; return decode_str (dec); 1569 case '"': ++dec->cur; return decode_str (dec);
1335 case '[': ++dec->cur; return decode_av (dec); 1570 case '[': ++dec->cur; return decode_av (dec);
1336 case '{': ++dec->cur; return decode_hv (dec); 1571 case '{': ++dec->cur; return decode_hv (dec);
1572 case '(': return decode_tag (dec);
1337 1573
1338 case '-': 1574 case '-':
1339 case '0': case '1': case '2': case '3': case '4': 1575 case '0': case '1': case '2': case '3': case '4':
1340 case '5': case '6': case '7': case '8': case '9': 1576 case '5': case '6': case '7': case '8': case '9':
1341 return decode_num (dec); 1577 return decode_num (dec);
1343 case 't': 1579 case 't':
1344 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1580 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1345 { 1581 {
1346 dec->cur += 4; 1582 dec->cur += 4;
1347#if JSON_SLOW 1583#if JSON_SLOW
1348 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1584 types_true = get_bool ("Types::Serialiser::true");
1349#endif 1585#endif
1350 return SvREFCNT_inc (json_true); 1586 return newSVsv (types_true);
1351 } 1587 }
1352 else 1588 else
1353 ERR ("'true' expected"); 1589 ERR ("'true' expected");
1354 1590
1355 break; 1591 break;
1357 case 'f': 1593 case 'f':
1358 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1594 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1359 { 1595 {
1360 dec->cur += 5; 1596 dec->cur += 5;
1361#if JSON_SLOW 1597#if JSON_SLOW
1362 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1598 types_false = get_bool ("Types::Serialiser::false");
1363#endif 1599#endif
1364 return SvREFCNT_inc (json_false); 1600 return newSVsv (types_false);
1365 } 1601 }
1366 else 1602 else
1367 ERR ("'false' expected"); 1603 ERR ("'false' expected");
1368 1604
1369 break; 1605 break;
1378 ERR ("'null' expected"); 1614 ERR ("'null' expected");
1379 1615
1380 break; 1616 break;
1381 1617
1382 default: 1618 default:
1383 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1619 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1384 break; 1620 break;
1385 } 1621 }
1386 1622
1387fail: 1623fail:
1388 return 0; 1624 return 0;
1389} 1625}
1390 1626
1391static SV * 1627static SV *
1392decode_json (SV *string, JSON *json, STRLEN *offset_return) 1628decode_json (SV *string, JSON *json, char **offset_return)
1393{ 1629{
1394 dec_t dec; 1630 dec_t dec;
1395 STRLEN offset;
1396 SV *sv; 1631 SV *sv;
1397 1632
1633 /* work around bugs in 5.10 where manipulating magic values
1634 * makes perl ignore the magic in subsequent accesses.
1635 * also make a copy of non-PV values, to get them into a clean
1636 * state (SvPV should do that, but it's buggy, see below).
1637 */
1398 SvGETMAGIC (string); 1638 /*SvGETMAGIC (string);*/
1639 if (SvMAGICAL (string) || !SvPOK (string))
1640 string = sv_2mortal (newSVsv (string));
1641
1399 SvUPGRADE (string, SVt_PV); 1642 SvUPGRADE (string, SVt_PV);
1400 1643
1401 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1644 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1645 * assertion with -DDEBUGGING, although SvCUR is documented to
1646 * return the xpv_cur field which certainly exists after upgrading.
1647 * according to nicholas clark, calling SvPOK fixes this.
1648 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1649 * and hope for the best.
1650 * Damnit, SvPV_nolen still trips over yet another assertion. This
1651 * assertion business is seriously broken, try yet another workaround
1652 * for the broken -DDEBUGGING.
1653 */
1654 {
1655#ifdef DEBUGGING
1656 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1657#else
1658 STRLEN offset = SvCUR (string);
1659#endif
1660
1661 if (offset > json->max_size && json->max_size)
1402 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1662 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1403 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1663 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1664 }
1404 1665
1405 if (json->flags & F_UTF8) 1666 if (DECODE_WANTS_OCTETS (json))
1406 sv_utf8_downgrade (string, 0); 1667 sv_utf8_downgrade (string, 0);
1407 else 1668 else
1408 sv_utf8_upgrade (string); 1669 sv_utf8_upgrade (string);
1409 1670
1410 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1671 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1411 1672
1412 dec.json = *json; 1673 dec.json = *json;
1413 dec.cur = SvPVX (string); 1674 dec.cur = SvPVX (string);
1414 dec.end = SvEND (string); 1675 dec.end = SvEND (string);
1415 dec.err = 0; 1676 dec.err = 0;
1416 dec.depth = 0; 1677 dec.depth = 0;
1417 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1418 1678
1419 if (dec.json.cb_object || dec.json.cb_sk_object) 1679 if (dec.json.cb_object || dec.json.cb_sk_object)
1420 dec.json.flags |= F_HOOK; 1680 dec.json.flags |= F_HOOK;
1421 1681
1422 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1682 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1423 1683
1424 decode_ws (&dec); 1684 decode_ws (&dec);
1425 sv = decode_sv (&dec); 1685 sv = decode_sv (&dec);
1686
1687 if (offset_return)
1688 *offset_return = dec.cur;
1426 1689
1427 if (!(offset_return || !sv)) 1690 if (!(offset_return || !sv))
1428 { 1691 {
1429 // check for trailing garbage 1692 // check for trailing garbage
1430 decode_ws (&dec); 1693 decode_ws (&dec);
1433 { 1696 {
1434 dec.err = "garbage after JSON object"; 1697 dec.err = "garbage after JSON object";
1435 SvREFCNT_dec (sv); 1698 SvREFCNT_dec (sv);
1436 sv = 0; 1699 sv = 0;
1437 } 1700 }
1438 }
1439
1440 if (offset_return || !sv)
1441 {
1442 offset = dec.json.flags & F_UTF8
1443 ? dec.cur - SvPVX (string)
1444 : utf8_distance (dec.cur, SvPVX (string));
1445
1446 if (offset_return)
1447 *offset_return = offset;
1448 } 1701 }
1449 1702
1450 if (!sv) 1703 if (!sv)
1451 { 1704 {
1452 SV *uni = sv_newmortal (); 1705 SV *uni = sv_newmortal ();
1458 SAVEVPTR (PL_curcop); 1711 SAVEVPTR (PL_curcop);
1459 PL_curcop = &cop; 1712 PL_curcop = &cop;
1460 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1713 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1461 LEAVE; 1714 LEAVE;
1462 1715
1463 croak ("%s, at character offset %d [\"%s\"]", 1716 croak ("%s, at character offset %d (before \"%s\")",
1464 dec.err, 1717 dec.err,
1465 (int)offset, 1718 (int)ptr_to_index (string, dec.cur),
1466 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1719 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1467 } 1720 }
1468 1721
1469 sv = sv_2mortal (sv); 1722 sv = sv_2mortal (sv);
1470 1723
1479 1732
1480static void 1733static void
1481incr_parse (JSON *self) 1734incr_parse (JSON *self)
1482{ 1735{
1483 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1736 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1737
1738 // the state machine here is a bit convoluted and could be simplified a lot
1739 // but this would make it slower, so...
1484 1740
1485 for (;;) 1741 for (;;)
1486 { 1742 {
1487 //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 1743 //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
1488 switch (self->incr_mode) 1744 switch (self->incr_mode)
1489 { 1745 {
1490 // only used for intiial whitespace skipping 1746 // only used for initial whitespace skipping
1491 case INCR_M_WS: 1747 case INCR_M_WS:
1492 for (;;) 1748 for (;;)
1493 { 1749 {
1494 if (*p > 0x20) 1750 if (*p > 0x20)
1495 { 1751 {
1752 if (*p == '#')
1753 {
1754 self->incr_mode = INCR_M_C0;
1755 goto incr_m_c;
1756 }
1757 else
1758 {
1496 self->incr_mode = INCR_M_JSON; 1759 self->incr_mode = INCR_M_JSON;
1497 goto incr_m_json; 1760 goto incr_m_json;
1761 }
1498 } 1762 }
1499 else if (!*p) 1763 else if (!*p)
1500 goto interrupt; 1764 goto interrupt;
1501 1765
1502 ++p; 1766 ++p;
1508 goto interrupt; 1772 goto interrupt;
1509 1773
1510 ++p; 1774 ++p;
1511 self->incr_mode = INCR_M_STR; 1775 self->incr_mode = INCR_M_STR;
1512 goto incr_m_str; 1776 goto incr_m_str;
1777
1778 // inside #-style comments
1779 case INCR_M_C0:
1780 case INCR_M_C1:
1781 incr_m_c:
1782 for (;;)
1783 {
1784 if (*p == '\n')
1785 {
1786 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1787 break;
1788 }
1789 else if (!*p)
1790 goto interrupt;
1791
1792 ++p;
1793 }
1794
1795 break;
1513 1796
1514 // inside a string 1797 // inside a string
1515 case INCR_M_STR: 1798 case INCR_M_STR:
1516 incr_m_str: 1799 incr_m_str:
1517 for (;;) 1800 for (;;)
1568 self->incr_mode = INCR_M_STR; 1851 self->incr_mode = INCR_M_STR;
1569 goto incr_m_str; 1852 goto incr_m_str;
1570 1853
1571 case '[': 1854 case '[':
1572 case '{': 1855 case '{':
1856 case '(':
1573 ++self->incr_nest; 1857 if (++self->incr_nest > self->max_depth)
1858 croak (ERR_NESTING_EXCEEDED);
1574 break; 1859 break;
1575 1860
1576 case ']': 1861 case ']':
1577 case '}': 1862 case '}':
1578 if (!--self->incr_nest) 1863 if (--self->incr_nest <= 0)
1579 goto interrupt; 1864 goto interrupt;
1865 break;
1866
1867 case ')':
1868 --self->incr_nest;
1869 break;
1870
1871 case '#':
1872 self->incr_mode = INCR_M_C1;
1873 goto incr_m_c;
1580 } 1874 }
1581 } 1875 }
1582 } 1876 }
1583 1877
1584 modechange: 1878 modechange:
1585 ; 1879 ;
1586 } 1880 }
1587 1881
1588interrupt: 1882interrupt:
1589 self->incr_pos = p - SvPVX (self->incr_text); 1883 self->incr_pos = p - SvPVX (self->incr_text);
1884 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1590 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1885 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1591} 1886}
1592 1887
1593///////////////////////////////////////////////////////////////////////////// 1888/////////////////////////////////////////////////////////////////////////////
1594// XS interface functions 1889// XS interface functions
1604 i >= '0' && i <= '9' ? i - '0' 1899 i >= '0' && i <= '9' ? i - '0'
1605 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1900 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1606 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1901 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1607 : -1; 1902 : -1;
1608 1903
1609 json_stash = gv_stashpv ("JSON::XS" , 1); 1904 json_stash = gv_stashpv ("JSON::XS" , 1);
1610 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1905 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1611 1906
1612 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1907 types_true = get_bool ("Types::Serialiser::true");
1613 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1908 types_false = get_bool ("Types::Serialiser::false");
1909
1910 sv_json = newSVpv ("JSON", 0);
1911 SvREADONLY_on (sv_json);
1912
1913 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1614} 1914}
1615 1915
1616PROTOTYPES: DISABLE 1916PROTOTYPES: DISABLE
1617 1917
1618void CLONE (...) 1918void CLONE (...)
1619 CODE: 1919 CODE:
1620 json_stash = 0; 1920 json_stash = 0;
1621 json_boolean_stash = 0; 1921 types_boolean_stash = 0;
1622 1922
1623void new (char *klass) 1923void new (char *klass)
1624 PPCODE: 1924 PPCODE:
1625{ 1925{
1626 SV *pv = NEWSV (0, sizeof (JSON)); 1926 SV *pv = NEWSV (0, sizeof (JSON));
1627 SvPOK_only (pv); 1927 SvPOK_only (pv);
1628 Zero (SvPVX (pv), 1, JSON); 1928 json_init ((JSON *)SvPVX (pv));
1629 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1630 XPUSHs (sv_2mortal (sv_bless ( 1929 XPUSHs (sv_2mortal (sv_bless (
1631 newRV_noinc (pv), 1930 newRV_noinc (pv),
1632 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1931 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1633 ))); 1932 )));
1634} 1933}
1646 allow_nonref = F_ALLOW_NONREF 1945 allow_nonref = F_ALLOW_NONREF
1647 shrink = F_SHRINK 1946 shrink = F_SHRINK
1648 allow_blessed = F_ALLOW_BLESSED 1947 allow_blessed = F_ALLOW_BLESSED
1649 convert_blessed = F_CONV_BLESSED 1948 convert_blessed = F_CONV_BLESSED
1650 relaxed = F_RELAXED 1949 relaxed = F_RELAXED
1950 allow_unknown = F_ALLOW_UNKNOWN
1951 allow_tags = F_ALLOW_TAGS
1651 PPCODE: 1952 PPCODE:
1652{ 1953{
1653 if (enable) 1954 if (enable)
1654 self->flags |= ix; 1955 self->flags |= ix;
1655 else 1956 else
1670 get_allow_nonref = F_ALLOW_NONREF 1971 get_allow_nonref = F_ALLOW_NONREF
1671 get_shrink = F_SHRINK 1972 get_shrink = F_SHRINK
1672 get_allow_blessed = F_ALLOW_BLESSED 1973 get_allow_blessed = F_ALLOW_BLESSED
1673 get_convert_blessed = F_CONV_BLESSED 1974 get_convert_blessed = F_CONV_BLESSED
1674 get_relaxed = F_RELAXED 1975 get_relaxed = F_RELAXED
1976 get_allow_unknown = F_ALLOW_UNKNOWN
1977 get_allow_tags = F_ALLOW_TAGS
1675 PPCODE: 1978 PPCODE:
1676 XPUSHs (boolSV (self->flags & ix)); 1979 XPUSHs (boolSV (self->flags & ix));
1677 1980
1678void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1981void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1679 PPCODE: 1982 PPCODE:
1680{ 1983 self->max_depth = max_depth;
1681 UV log2 = 0;
1682
1683 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1684
1685 while ((1UL << log2) < max_depth)
1686 ++log2;
1687
1688 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1689
1690 XPUSHs (ST (0)); 1984 XPUSHs (ST (0));
1691}
1692 1985
1693U32 get_max_depth (JSON *self) 1986U32 get_max_depth (JSON *self)
1694 CODE: 1987 CODE:
1695 RETVAL = DEC_DEPTH (self->flags); 1988 RETVAL = self->max_depth;
1696 OUTPUT: 1989 OUTPUT:
1697 RETVAL 1990 RETVAL
1698 1991
1699void max_size (JSON *self, UV max_size = 0) 1992void max_size (JSON *self, U32 max_size = 0)
1700 PPCODE: 1993 PPCODE:
1701{ 1994 self->max_size = max_size;
1702 UV log2 = 0;
1703
1704 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1705 if (max_size == 1) max_size = 2;
1706
1707 while ((1UL << log2) < max_size)
1708 ++log2;
1709
1710 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1711
1712 XPUSHs (ST (0)); 1995 XPUSHs (ST (0));
1713}
1714 1996
1715int get_max_size (JSON *self) 1997int get_max_size (JSON *self)
1716 CODE: 1998 CODE:
1717 RETVAL = DEC_SIZE (self->flags); 1999 RETVAL = self->max_size;
1718 OUTPUT: 2000 OUTPUT:
1719 RETVAL 2001 RETVAL
1720 2002
1721void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2003void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1722 PPCODE: 2004 PPCODE:
1728} 2010}
1729 2011
1730void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2012void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1731 PPCODE: 2013 PPCODE:
1732{ 2014{
1733 if (!self->cb_sk_object) 2015 if (!self->cb_sk_object)
1734 self->cb_sk_object = newHV (); 2016 self->cb_sk_object = newHV ();
1735 2017
1736 if (SvOK (cb)) 2018 if (SvOK (cb))
1737 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2019 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1738 else 2020 else
1749 XPUSHs (ST (0)); 2031 XPUSHs (ST (0));
1750} 2032}
1751 2033
1752void encode (JSON *self, SV *scalar) 2034void encode (JSON *self, SV *scalar)
1753 PPCODE: 2035 PPCODE:
1754 XPUSHs (encode_json (scalar, self)); 2036 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2037 XPUSHs (scalar);
1755 2038
1756void decode (JSON *self, SV *jsonstr) 2039void decode (JSON *self, SV *jsonstr)
1757 PPCODE: 2040 PPCODE:
1758 XPUSHs (decode_json (jsonstr, self, 0)); 2041 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2042 XPUSHs (jsonstr);
1759 2043
1760void decode_prefix (JSON *self, SV *jsonstr) 2044void decode_prefix (JSON *self, SV *jsonstr)
1761 PPCODE: 2045 PPCODE:
1762{ 2046{
1763 STRLEN offset; 2047 SV *sv;
2048 char *offset;
2049 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1764 EXTEND (SP, 2); 2050 EXTEND (SP, 2);
1765 PUSHs (decode_json (jsonstr, self, &offset)); 2051 PUSHs (sv);
1766 PUSHs (sv_2mortal (newSVuv (offset))); 2052 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1767} 2053}
1768 2054
1769void incr_parse (JSON *self, SV *jsonstr = 0) 2055void incr_parse (JSON *self, SV *jsonstr = 0)
1770 PPCODE: 2056 PPCODE:
1771{ 2057{
1772 if (!self->incr_text) 2058 if (!self->incr_text)
1773 self->incr_text = newSVpvn ("", 0); 2059 self->incr_text = newSVpvn ("", 0);
2060
2061 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2062 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2063 if (DECODE_WANTS_OCTETS (self))
2064 {
2065 if (self->incr_pos)
2066 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2067 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2068
2069 sv_utf8_downgrade (self->incr_text, 0);
2070 }
2071 else
2072 {
2073 sv_utf8_upgrade (self->incr_text);
2074
2075 if (self->incr_pos)
2076 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2077 - (U8 *)SvPVX (self->incr_text);
2078 }
1774 2079
1775 // append data, if any 2080 // append data, if any
1776 if (jsonstr) 2081 if (jsonstr)
1777 { 2082 {
2083 /* make sure both strings have same encoding */
1778 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2084 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2085 if (SvUTF8 (jsonstr))
2086 sv_utf8_downgrade (jsonstr, 0);
1779 { 2087 else
1780 /* utf-8-ness differs, need to upgrade */
1781 sv_utf8_upgrade (self->incr_text); 2088 sv_utf8_upgrade (jsonstr);
1782 2089
1783 if (self->incr_pos) 2090 /* and then just blindly append */
1784 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1785 - (U8 *)SvPVX (self->incr_text);
1786 }
1787
1788 { 2091 {
1789 STRLEN len; 2092 STRLEN len;
1790 const char *str = SvPV (jsonstr, len); 2093 const char *str = SvPV (jsonstr, len);
1791 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2094 STRLEN cur = SvCUR (self->incr_text);
2095
2096 if (SvLEN (self->incr_text) <= cur + len)
2097 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2098
1792 Move (str, SvEND (self->incr_text), len, char); 2099 Move (str, SvEND (self->incr_text), len, char);
1793 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2100 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1794 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2101 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1795 } 2102 }
1796 } 2103 }
1797 2104
1798 if (GIMME_V != G_VOID) 2105 if (GIMME_V != G_VOID)
1799 do 2106 do
1800 { 2107 {
2108 SV *sv;
1801 STRLEN offset; 2109 char *offset;
1802
1803 incr_parse (self);
1804 2110
1805 if (!INCR_DONE (self)) 2111 if (!INCR_DONE (self))
2112 {
2113 incr_parse (self);
2114
2115 if (self->incr_pos > self->max_size && self->max_size)
2116 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2117 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2118
2119 if (!INCR_DONE (self))
2120 {
2121 // as an optimisation, do not accumulate white space in the incr buffer
2122 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2123 {
2124 self->incr_pos = 0;
2125 SvCUR_set (self->incr_text, 0);
2126 }
2127
1806 break; 2128 break;
2129 }
2130 }
1807 2131
1808 XPUSHs (decode_json (self->incr_text, self, &offset)); 2132 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2133 XPUSHs (sv);
1809 2134
1810 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1811 self->incr_pos -= offset; 2135 self->incr_pos -= offset - SvPVX (self->incr_text);
1812 self->incr_nest = 0; 2136 self->incr_nest = 0;
1813 self->incr_mode = 0; 2137 self->incr_mode = 0;
2138
2139 sv_chop (self->incr_text, offset);
1814 } 2140 }
1815 while (GIMME_V == G_ARRAY); 2141 while (GIMME_V == G_ARRAY);
1816} 2142}
1817 2143
1818SV *incr_text (JSON *self) 2144SV *incr_text (JSON *self)
1819 ATTRS: lvalue 2145 ATTRS: lvalue
1820 CODE: 2146 CODE:
1821{ 2147{
1822 if (self->incr_pos) 2148 if (self->incr_pos)
1823 croak ("incr_text can only be called after a successful incr_parse call in scalar context %d", self->incr_pos);//D 2149 croak ("incr_text can not be called when the incremental parser already started parsing");
1824 2150
1825 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; 2151 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1826} 2152}
1827 OUTPUT: 2153 OUTPUT:
1828 RETVAL 2154 RETVAL
2155
2156void incr_skip (JSON *self)
2157 CODE:
2158{
2159 if (self->incr_pos)
2160 {
2161 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2162 self->incr_pos = 0;
2163 self->incr_nest = 0;
2164 self->incr_mode = 0;
2165 }
2166}
2167
2168void incr_reset (JSON *self)
2169 CODE:
2170{
2171 SvREFCNT_dec (self->incr_text);
2172 self->incr_text = 0;
2173 self->incr_pos = 0;
2174 self->incr_nest = 0;
2175 self->incr_mode = 0;
2176}
1829 2177
1830void DESTROY (JSON *self) 2178void DESTROY (JSON *self)
1831 CODE: 2179 CODE:
1832 SvREFCNT_dec (self->cb_sk_object); 2180 SvREFCNT_dec (self->cb_sk_object);
1833 SvREFCNT_dec (self->cb_object); 2181 SvREFCNT_dec (self->cb_object);
1834 SvREFCNT_dec (self->incr_text); 2182 SvREFCNT_dec (self->incr_text);
1835 2183
1836PROTOTYPES: ENABLE 2184PROTOTYPES: ENABLE
1837 2185
1838void encode_json (SV *scalar) 2186void encode_json (SV *scalar)
1839 ALIAS:
1840 to_json_ = 0
1841 encode_json = F_UTF8
1842 PPCODE: 2187 PPCODE:
1843{ 2188{
1844 JSON json = { F_DEFAULT | ix }; 2189 JSON json;
1845 XPUSHs (encode_json (scalar, &json)); 2190 json_init (&json);
2191 json.flags |= F_UTF8;
2192 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2193 XPUSHs (scalar);
1846} 2194}
1847 2195
1848void decode_json (SV *jsonstr) 2196void decode_json (SV *jsonstr)
1849 ALIAS:
1850 from_json_ = 0
1851 decode_json = F_UTF8
1852 PPCODE: 2197 PPCODE:
1853{ 2198{
1854 JSON json = { F_DEFAULT | ix }; 2199 JSON json;
1855 XPUSHs (decode_json (jsonstr, &json, 0)); 2200 json_init (&json);
2201 json.flags |= F_UTF8;
2202 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2203 XPUSHs (jsonstr);
1856} 2204}
1857 2205
1858

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines