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 1.81 by root, Wed Mar 26 01:40:43 2008 UTC vs.
Revision 1.125 by root, Sun Feb 21 15:37:53 2016 UTC

12#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 14#endif
15 15
16// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
20#endif 20#endif
21 21
22// compatibility with perl <5.18
23#ifndef HvNAMELEN_get
24# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
25#endif
26#ifndef HvNAMELEN
27# define HvNAMELEN(hv) HvNAMELEN_get (hv)
28#endif
29#ifndef HvNAMEUTF8
30# define HvNAMEUTF8(hv) 0
31#endif
32
33// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 34#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 35
24#define F_ASCII 0x00000001UL 36#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 37#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 38#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 39#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 43#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL 44#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 45#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 46#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 47#define F_RELAXED 0x00001000UL
36 48#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL 49#define F_ALLOW_TAGS 0x00004000UL
38#define S_MAXDEPTH 27
39#define F_MAXSIZE 0x01f00000UL
40#define S_MAXSIZE 20
41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 50#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 51
43#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
44#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
45
46#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 52#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 53
49#define INIT_SIZE 32 // initial scalar size to be allocated 54#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 55#define INDENT_STEP 3 // spaces per indentation level
51 56
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 57#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
58
59#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
53 60
54#define SB do { 61#define SB do {
55#define SE } while (0) 62#define SE } while (0)
56 63
57#if __GNUC__ >= 3 64#if __GNUC__ >= 3
67 74
68#define IN_RANGE_INC(type,val,beg,end) \ 75#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 76 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 77 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 78
79#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
80
72#ifdef USE_ITHREADS 81#ifdef USE_ITHREADS
73# define JSON_SLOW 1 82# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
75#else 84#else
76# define JSON_SLOW 0 85# define JSON_SLOW 0
77# define JSON_STASH json_stash 86# define JSON_STASH json_stash
78#endif 87#endif
79 88
89// the amount of HEs to allocate on the stack, when sorting keys
90#define STACK_HES 64
91
80static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92static HV *json_stash, *types_boolean_stash; // JSON::XS::
81static SV *json_true, *json_false; 93static SV *types_true, *types_false, *sv_json;
82 94
83enum { 95enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0 96 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string 97 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash 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
87 INCR_M_JSON // outside anything, count nesting 101 INCR_M_JSON // outside anything, count nesting
88}; 102};
89 103
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 104#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
91 105
92typedef struct { 106typedef struct {
93 U32 flags; 107 U32 flags;
108 U32 max_depth;
109 STRLEN max_size;
110
94 SV *cb_object; 111 SV *cb_object;
95 HV *cb_sk_object; 112 HV *cb_sk_object;
96 113
97 // for the incremental parser 114 // for the incremental parser
98 SV *incr_text; // the source text so far 115 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 116 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 117 int incr_nest; // {[]}-nesting level
101 int incr_mode; 118 unsigned char incr_mode;
102} JSON; 119} JSON;
120
121INLINE void
122json_init (JSON *json)
123{
124 Zero (json, 1, JSON);
125 json->max_depth = 512;
126}
103 127
104///////////////////////////////////////////////////////////////////////////// 128/////////////////////////////////////////////////////////////////////////////
105// utility functions 129// utility functions
106 130
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
107INLINE void 142INLINE void
108shrink (SV *sv) 143shrink (SV *sv)
109{ 144{
110 sv_utf8_downgrade (sv, 1); 145 sv_utf8_downgrade (sv, 1);
146
111 if (SvLEN (sv) > SvCUR (sv) + 1) 147 if (SvLEN (sv) > SvCUR (sv) + 1)
112 { 148 {
113#ifdef SvPV_shrink_to_cur 149#ifdef SvPV_shrink_to_cur
114 SvPV_shrink_to_cur (sv); 150 SvPV_shrink_to_cur (sv);
115#elif defined (SvPV_renew) 151#elif defined (SvPV_renew)
159 *s++ = 0x80 | ( ch & 0x3f); 195 *s++ = 0x80 | ( ch & 0x3f);
160 196
161 return s; 197 return s;
162} 198}
163 199
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}
164///////////////////////////////////////////////////////////////////////////// 304/////////////////////////////////////////////////////////////////////////////
165// encoder 305// encoder
166 306
167// structure used for encoding JSON 307// structure used for encoding JSON
168typedef struct 308typedef struct
170 char *cur; // SvPVX (sv) + current output position 310 char *cur; // SvPVX (sv) + current output position
171 char *end; // SvEND (sv) 311 char *end; // SvEND (sv)
172 SV *sv; // result scalar 312 SV *sv; // result scalar
173 JSON json; 313 JSON json;
174 U32 indent; // indentation level 314 U32 indent; // indentation level
175 U32 maxdepth; // max. indentation/recursion level
176 UV limit; // escape character values >= this value when encoding 315 UV limit; // escape character values >= this value when encoding
177} enc_t; 316} enc_t;
178 317
179INLINE void 318INLINE void
180need (enc_t *enc, STRLEN len) 319need (enc_t *enc, STRLEN len)
181{ 320{
182 if (expect_false (enc->cur + len >= enc->end)) 321 if (expect_false (enc->cur + len >= enc->end))
183 { 322 {
184 STRLEN cur = enc->cur - SvPVX (enc->sv); 323 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
185 SvGROW (enc->sv, cur + len + 1); 324 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
186 enc->cur = SvPVX (enc->sv) + cur; 325 enc->cur = SvPVX (enc->sv) + cur;
187 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 326 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
188 } 327 }
189} 328}
190 329
265 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 404 (int)((uch - 0x10000) % 0x400 + 0xDC00));
266 enc->cur += 12; 405 enc->cur += 12;
267 } 406 }
268 else 407 else
269 { 408 {
270 static char hexdigit [16] = "0123456789abcdef";
271 need (enc, len += 5); 409 need (enc, len += 5);
272 *enc->cur++ = '\\'; 410 *enc->cur++ = '\\';
273 *enc->cur++ = 'u'; 411 *enc->cur++ = 'u';
274 *enc->cur++ = hexdigit [ uch >> 12 ]; 412 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
275 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 413 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
276 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 414 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
277 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 415 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
278 } 416 }
279 417
280 str += clen; 418 str += clen;
281 } 419 }
282 else if (enc->json.flags & F_LATIN1) 420 else if (enc->json.flags & F_LATIN1)
353static void 491static void
354encode_av (enc_t *enc, AV *av) 492encode_av (enc_t *enc, AV *av)
355{ 493{
356 int i, len = av_len (av); 494 int i, len = av_len (av);
357 495
358 if (enc->indent >= enc->maxdepth) 496 if (enc->indent >= enc->json.max_depth)
359 croak ("data structure too deep (hit recursion limit)"); 497 croak (ERR_NESTING_EXCEEDED);
360 498
361 encode_ch (enc, '['); 499 encode_ch (enc, '[');
362 500
363 if (len >= 0) 501 if (len >= 0)
364 { 502 {
365 encode_nl (enc); ++enc->indent; 503 encode_nl (enc); ++enc->indent;
366 504
367 for (i = 0; i <= len; ++i) 505 for (i = 0; i <= len; ++i)
379 encode_comma (enc); 517 encode_comma (enc);
380 } 518 }
381 519
382 encode_nl (enc); --enc->indent; encode_indent (enc); 520 encode_nl (enc); --enc->indent; encode_indent (enc);
383 } 521 }
384 522
385 encode_ch (enc, ']'); 523 encode_ch (enc, ']');
386} 524}
387 525
388static void 526static void
389encode_hk (enc_t *enc, HE *he) 527encode_hk (enc_t *enc, HE *he)
393 if (HeKLEN (he) == HEf_SVKEY) 531 if (HeKLEN (he) == HEf_SVKEY)
394 { 532 {
395 SV *sv = HeSVKEY (he); 533 SV *sv = HeSVKEY (he);
396 STRLEN len; 534 STRLEN len;
397 char *str; 535 char *str;
398 536
399 SvGETMAGIC (sv); 537 SvGETMAGIC (sv);
400 str = SvPV (sv, len); 538 str = SvPV (sv, len);
401 539
402 encode_str (enc, str, len, SvUTF8 (sv)); 540 encode_str (enc, str, len, SvUTF8 (sv));
403 } 541 }
439static void 577static void
440encode_hv (enc_t *enc, HV *hv) 578encode_hv (enc_t *enc, HV *hv)
441{ 579{
442 HE *he; 580 HE *he;
443 581
444 if (enc->indent >= enc->maxdepth) 582 if (enc->indent >= enc->json.max_depth)
445 croak ("data structure too deep (hit recursion limit)"); 583 croak (ERR_NESTING_EXCEEDED);
446 584
447 encode_ch (enc, '{'); 585 encode_ch (enc, '{');
448 586
449 // for canonical output we have to sort by keys first 587 // for canonical output we have to sort by keys first
450 // actually, this is mostly due to the stupid so-called 588 // actually, this is mostly due to the stupid so-called
451 // security workaround added somewhere in 5.8.x. 589 // security workaround added somewhere in 5.8.x
452 // that randomises hash orderings 590 // that randomises hash orderings
453 if (enc->json.flags & F_CANONICAL) 591 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
454 { 592 {
455 int count = hv_iterinit (hv); 593 int count = hv_iterinit (hv);
456 594
457 if (SvMAGICAL (hv)) 595 if (SvMAGICAL (hv))
458 { 596 {
469 } 607 }
470 608
471 if (count) 609 if (count)
472 { 610 {
473 int i, fast = 1; 611 int i, fast = 1;
474#if defined(__BORLANDC__) || defined(_MSC_VER) 612 HE *hes_stack [STACK_HES];
475 HE **hes = _alloca (count * sizeof (HE)); 613 HE **hes = hes_stack;
476#else 614
477 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 615 // allocate larger arrays on the heap
478#endif 616 if (count > STACK_HES)
617 {
618 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
619 hes = (HE **)SvPVX (sv);
620 }
479 621
480 i = 0; 622 i = 0;
481 while ((he = hv_iternext (hv))) 623 while ((he = hv_iternext (hv)))
482 { 624 {
483 hes [i++] = he; 625 hes [i++] = he;
552// encode objects, arrays and special \0=false and \1=true values. 694// encode objects, arrays and special \0=false and \1=true values.
553static void 695static void
554encode_rv (enc_t *enc, SV *sv) 696encode_rv (enc_t *enc, SV *sv)
555{ 697{
556 svtype svt; 698 svtype svt;
699 GV *method;
557 700
558 SvGETMAGIC (sv); 701 SvGETMAGIC (sv);
559 svt = SvTYPE (sv); 702 svt = SvTYPE (sv);
560 703
561 if (expect_false (SvOBJECT (sv))) 704 if (expect_false (SvOBJECT (sv)))
562 { 705 {
563 HV *stash = !JSON_SLOW || json_boolean_stash 706 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
564 ? json_boolean_stash 707 ? types_boolean_stash
565 : gv_stashpv ("JSON::XS::Boolean", 1); 708 : gv_stashpv ("Types::Serialiser::Boolean", 1);
709 HV *stash = SvSTASH (sv);
566 710
567 if (SvSTASH (sv) == stash) 711 if (stash == boolean_stash)
568 { 712 {
569 if (SvIV (sv)) 713 if (SvIV (sv))
570 encode_str (enc, "true", 4, 0); 714 encode_str (enc, "true", 4, 0);
571 else 715 else
572 encode_str (enc, "false", 5, 0); 716 encode_str (enc, "false", 5, 0);
573 } 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;
724 SAVESTACK_POS ();
725 PUSHMARK (SP);
726 EXTEND (SP, 2);
727 // we re-bless the reference to get overload and other niceties right
728 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
729 PUSHs (sv_json);
730
731 PUTBACK;
732 count = call_sv ((SV *)GvCV (method), G_ARRAY);
733 SPAGAIN;
734
735 // catch this surprisingly common error
736 if (SvROK (TOPs) && SvRV (TOPs) == sv)
737 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
738
739 encode_ch (enc, '(');
740 encode_ch (enc, '"');
741 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
742 encode_ch (enc, '"');
743 encode_ch (enc, ')');
744 encode_ch (enc, '[');
745
746 while (count)
747 {
748 encode_sv (enc, SP[1 - count--]);
749
750 if (count)
751 encode_ch (enc, ',');
752 }
753
754 encode_ch (enc, ']');
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;
763 PUSHMARK (SP);
764 // we re-bless the reference to get overload and other niceties right
765 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
766
767 // calling with G_SCALAR ensures that we always get a 1 return value
768 PUTBACK;
769 call_sv ((SV *)GvCV (method), G_SCALAR);
770 SPAGAIN;
771
772 // catch this surprisingly common error
773 if (SvROK (TOPs) && SvRV (TOPs) == sv)
774 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
775
776 sv = POPs;
777 PUTBACK;
778
779 encode_sv (enc, sv);
780
781 FREETMPS; LEAVE;
782 }
783 else if (enc->json.flags & F_ALLOW_BLESSED)
784 encode_str (enc, "null", 4, 0);
574 else 785 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", 786 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)))); 787 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
621 }
622 } 788 }
623 else if (svt == SVt_PVHV) 789 else if (svt == SVt_PVHV)
624 encode_hv (enc, (HV *)sv); 790 encode_hv (enc, (HV *)sv);
625 else if (svt == SVt_PVAV) 791 else if (svt == SVt_PVAV)
626 encode_av (enc, (AV *)sv); 792 encode_av (enc, (AV *)sv);
631 797
632 if (len == 1 && *pv == '1') 798 if (len == 1 && *pv == '1')
633 encode_str (enc, "true", 4, 0); 799 encode_str (enc, "true", 4, 0);
634 else if (len == 1 && *pv == '0') 800 else if (len == 1 && *pv == '0')
635 encode_str (enc, "false", 5, 0); 801 encode_str (enc, "false", 5, 0);
802 else if (enc->json.flags & F_ALLOW_UNKNOWN)
803 encode_str (enc, "null", 4, 0);
636 else 804 else
637 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 805 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
638 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 806 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
639 } 807 }
808 else if (enc->json.flags & F_ALLOW_UNKNOWN)
809 encode_str (enc, "null", 4, 0);
640 else 810 else
641 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 811 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
642 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 812 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
643} 813}
644 814
698 } 868 }
699 else 869 else
700 { 870 {
701 // large integer, use the (rather slow) snprintf way. 871 // large integer, use the (rather slow) snprintf way.
702 need (enc, IVUV_MAXCHARS); 872 need (enc, IVUV_MAXCHARS);
703 enc->cur += 873 enc->cur +=
704 SvIsUV(sv) 874 SvIsUV(sv)
705 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 875 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
706 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 876 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
707 } 877 }
708 } 878 }
709 else if (SvROK (sv)) 879 else if (SvROK (sv))
710 encode_rv (enc, SvRV (sv)); 880 encode_rv (enc, SvRV (sv));
711 else if (!SvOK (sv)) 881 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
712 encode_str (enc, "null", 4, 0); 882 encode_str (enc, "null", 4, 0);
713 else 883 else
714 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 884 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
715 SvPV_nolen (sv), SvFLAGS (sv)); 885 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
886}
887
888static int
889json_scalar (SV *scalar)
890{
891 return 0;//D
892 if (!SvROK (scalar))
893 return 1;
716} 894}
717 895
718static SV * 896static SV *
719encode_json (SV *scalar, JSON *json) 897encode_json (SV *scalar, JSON *json)
720{ 898{
721 enc_t enc; 899 enc_t enc;
722 900
723 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 901 if (!(json->flags & F_ALLOW_NONREF) && json_scalar (scalar))
724 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 902 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
725 903
726 enc.json = *json; 904 enc.json = *json;
727 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 905 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
728 enc.cur = SvPVX (enc.sv); 906 enc.cur = SvPVX (enc.sv);
729 enc.end = SvEND (enc.sv); 907 enc.end = SvEND (enc.sv);
730 enc.indent = 0; 908 enc.indent = 0;
731 enc.maxdepth = DEC_DEPTH (enc.json.flags);
732 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 909 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
733 : enc.json.flags & F_LATIN1 ? 0x000100UL 910 : enc.json.flags & F_LATIN1 ? 0x000100UL
734 : 0x110000UL; 911 : 0x110000UL;
735 912
736 SvPOK_only (enc.sv); 913 SvPOK_only (enc.sv);
737 encode_sv (&enc, scalar); 914 encode_sv (&enc, scalar);
915 encode_nl (&enc);
738 916
739 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 917 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
740 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 918 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
741 919
742 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 920 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
803 if (*dec->cur != ch) \ 981 if (*dec->cur != ch) \
804 ERR (# ch " expected"); \ 982 ERR (# ch " expected"); \
805 ++dec->cur; \ 983 ++dec->cur; \
806 SE 984 SE
807 985
808#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 986#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
809#define DEC_DEC_DEPTH --dec->depth 987#define DEC_DEC_DEPTH --dec->depth
810 988
811static SV *decode_sv (dec_t *dec); 989static SV *decode_sv (dec_t *dec);
812 990
813static signed char decode_hexdigit[256]; 991static signed char decode_hexdigit[256];
922 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1100 else if (expect_true (ch >= 0x20 && ch < 0x80))
923 *cur++ = ch; 1101 *cur++ = ch;
924 else if (ch >= 0x80) 1102 else if (ch >= 0x80)
925 { 1103 {
926 STRLEN clen; 1104 STRLEN clen;
927 UV uch;
928 1105
929 --dec_cur; 1106 --dec_cur;
930 1107
931 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1108 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
932 if (clen == (STRLEN)-1) 1109 if (clen == (STRLEN)-1)
933 ERR ("malformed UTF-8 character in JSON string"); 1110 ERR ("malformed UTF-8 character in JSON string");
934 1111
935 do 1112 do
936 *cur++ = *dec_cur++; 1113 *cur++ = *dec_cur++;
937 while (--clen); 1114 while (--clen);
938 1115
939 utf8 = 1; 1116 utf8 = 1;
940 } 1117 }
1118 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1119 *cur++ = ch;
941 else 1120 else
942 { 1121 {
943 --dec_cur; 1122 --dec_cur;
944 1123
945 if (!ch) 1124 if (!ch)
953 { 1132 {
954 STRLEN len = cur - buf; 1133 STRLEN len = cur - buf;
955 1134
956 if (sv) 1135 if (sv)
957 { 1136 {
958 SvGROW (sv, SvCUR (sv) + len + 1); 1137 STRLEN cur = SvCUR (sv);
1138
1139 if (SvLEN (sv) <= cur + len)
1140 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1141
959 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1142 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
960 SvCUR_set (sv, SvCUR (sv) + len); 1143 SvCUR_set (sv, SvCUR (sv) + len);
961 } 1144 }
962 else 1145 else
963 sv = newSVpvn (buf, len); 1146 sv = newSVpvn (buf, len);
1054 1237
1055 // special case the rather common 1..5-digit-int case 1238 // special case the rather common 1..5-digit-int case
1056 if (*start == '-') 1239 if (*start == '-')
1057 switch (len) 1240 switch (len)
1058 { 1241 {
1059 case 2: return newSViv (-( start [1] - '0' * 1)); 1242 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1060 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1243 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)); 1244 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)); 1245 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)); 1246 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1064 } 1247 }
1065 else 1248 else
1066 switch (len) 1249 switch (len)
1067 { 1250 {
1068 case 1: return newSViv ( start [0] - '0' * 1); 1251 case 1: return newSViv ( start [0] - '0' * 1);
1069 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1252 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); 1253 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); 1254 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); 1255 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1073 } 1256 }
1074 1257
1075 { 1258 {
1076 UV uv; 1259 UV uv;
1077 int numtype = grok_number (start, len, &uv); 1260 int numtype = grok_number (start, len, &uv);
1086 } 1269 }
1087 1270
1088 len -= *start == '-' ? 1 : 0; 1271 len -= *start == '-' ? 1 : 0;
1089 1272
1090 // does not fit into IV or UV, try NV 1273 // does not fit into IV or UV, try NV
1091 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1274 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 1275 // fits into NV without loss of precision
1097 return newSVnv (Atof (start)); 1276 return newSVnv (json_atof (start));
1098 1277
1099 // everything else fails, convert it to a string 1278 // everything else fails, convert it to a string
1100 return newSVpvn (start, dec->cur - start); 1279 return newSVpvn (start, dec->cur - start);
1101 } 1280 }
1102 1281
1103 // loss of precision here 1282 // loss of precision here
1104 return newSVnv (Atof (start)); 1283 return newSVnv (json_atof (start));
1105 1284
1106fail: 1285fail:
1107 return 0; 1286 return 0;
1108} 1287}
1109 1288
1133 if (*dec->cur == ']') 1312 if (*dec->cur == ']')
1134 { 1313 {
1135 ++dec->cur; 1314 ++dec->cur;
1136 break; 1315 break;
1137 } 1316 }
1138 1317
1139 if (*dec->cur != ',') 1318 if (*dec->cur != ',')
1140 ERR (", or ] expected while parsing array"); 1319 ERR (", or ] expected while parsing array");
1141 1320
1142 ++dec->cur; 1321 ++dec->cur;
1143 1322
1267 1446
1268 hv_iterinit (hv); 1447 hv_iterinit (hv);
1269 he = hv_iternext (hv); 1448 he = hv_iternext (hv);
1270 hv_iterinit (hv); 1449 hv_iterinit (hv);
1271 1450
1272 // the next line creates a mortal sv each time its called. 1451 // the next line creates a mortal sv each time it's called.
1273 // might want to optimise this for common cases. 1452 // might want to optimise this for common cases.
1274 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1453 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1275 1454
1276 if (cb) 1455 if (cb)
1277 { 1456 {
1278 dSP; 1457 dSP;
1279 int count; 1458 int count;
1280 1459
1281 ENTER; SAVETMPS; PUSHMARK (SP); 1460 ENTER; SAVETMPS;
1461 SAVESTACK_POS ();
1462 PUSHMARK (SP);
1282 XPUSHs (HeVAL (he)); 1463 XPUSHs (HeVAL (he));
1464 sv_2mortal (sv);
1283 1465
1284 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1466 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1285 1467
1286 if (count == 1) 1468 if (count == 1)
1287 { 1469 {
1288 sv = newSVsv (POPs); 1470 sv = newSVsv (POPs);
1289 FREETMPS; LEAVE; 1471 FREETMPS; LEAVE;
1290 return sv; 1472 return sv;
1291 } 1473 }
1292 1474
1475 SvREFCNT_inc (sv);
1293 FREETMPS; LEAVE; 1476 FREETMPS; LEAVE;
1294 } 1477 }
1295 } 1478 }
1296 1479
1297 if (dec->json.cb_object) 1480 if (dec->json.cb_object)
1298 { 1481 {
1299 dSP; 1482 dSP;
1300 int count; 1483 int count;
1301 1484
1302 ENTER; SAVETMPS; PUSHMARK (SP); 1485 ENTER; SAVETMPS;
1486 SAVESTACK_POS ();
1487 PUSHMARK (SP);
1303 XPUSHs (sv_2mortal (sv)); 1488 XPUSHs (sv_2mortal (sv));
1304 1489
1305 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1490 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1306 1491
1307 if (count == 1) 1492 if (count == 1)
1323 DEC_DEC_DEPTH; 1508 DEC_DEC_DEPTH;
1324 return 0; 1509 return 0;
1325} 1510}
1326 1511
1327static SV * 1512static SV *
1513decode_tag (dec_t *dec)
1514{
1515 SV *tag = 0;
1516 SV *val = 0;
1517
1518 if (!(dec->json.flags & F_ALLOW_TAGS))
1519 ERR ("malformed JSON string, neither array, object, number, string or atom");
1520
1521 ++dec->cur;
1522
1523 decode_ws (dec);
1524
1525 tag = decode_sv (dec);
1526 if (!tag)
1527 goto fail;
1528
1529 if (!SvPOK (tag))
1530 ERR ("malformed JSON string, (tag) must be a string");
1531
1532 decode_ws (dec);
1533
1534 if (*dec->cur != ')')
1535 ERR (") expected after tag");
1536
1537 ++dec->cur;
1538
1539 decode_ws (dec);
1540
1541 val = decode_sv (dec);
1542 if (!val)
1543 goto fail;
1544
1545 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1546 ERR ("malformed JSON string, tag value must be an array");
1547
1548 {
1549 AV *av = (AV *)SvRV (val);
1550 int i, len = av_len (av) + 1;
1551 HV *stash = gv_stashsv (tag, 0);
1552 SV *sv;
1553
1554 if (!stash)
1555 ERR ("cannot decode perl-object (package does not exist)");
1556
1557 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1558
1559 if (!method)
1560 ERR ("cannot decode perl-object (package does not have a THAW method)");
1561
1562 dSP;
1563
1564 ENTER; SAVETMPS;
1565 PUSHMARK (SP);
1566 EXTEND (SP, len + 2);
1567 // we re-bless the reference to get overload and other niceties right
1568 PUSHs (tag);
1569 PUSHs (sv_json);
1570
1571 for (i = 0; i < len; ++i)
1572 PUSHs (*av_fetch (av, i, 1));
1573
1574 PUTBACK;
1575 call_sv ((SV *)GvCV (method), G_SCALAR);
1576 SPAGAIN;
1577
1578 SvREFCNT_dec (tag);
1579 SvREFCNT_dec (val);
1580 sv = SvREFCNT_inc (POPs);
1581
1582 PUTBACK;
1583
1584 FREETMPS; LEAVE;
1585
1586 return sv;
1587 }
1588
1589fail:
1590 SvREFCNT_dec (tag);
1591 SvREFCNT_dec (val);
1592 return 0;
1593}
1594
1595static SV *
1328decode_sv (dec_t *dec) 1596decode_sv (dec_t *dec)
1329{ 1597{
1330 // the beauty of JSON: you need exactly one character lookahead 1598 // the beauty of JSON: you need exactly one character lookahead
1331 // to parse everything. 1599 // to parse everything.
1332 switch (*dec->cur) 1600 switch (*dec->cur)
1333 { 1601 {
1334 case '"': ++dec->cur; return decode_str (dec); 1602 case '"': ++dec->cur; return decode_str (dec);
1335 case '[': ++dec->cur; return decode_av (dec); 1603 case '[': ++dec->cur; return decode_av (dec);
1336 case '{': ++dec->cur; return decode_hv (dec); 1604 case '{': ++dec->cur; return decode_hv (dec);
1605 case '(': return decode_tag (dec);
1337 1606
1338 case '-': 1607 case '-':
1339 case '0': case '1': case '2': case '3': case '4': 1608 case '0': case '1': case '2': case '3': case '4':
1340 case '5': case '6': case '7': case '8': case '9': 1609 case '5': case '6': case '7': case '8': case '9':
1341 return decode_num (dec); 1610 return decode_num (dec);
1343 case 't': 1612 case 't':
1344 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1613 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1345 { 1614 {
1346 dec->cur += 4; 1615 dec->cur += 4;
1347#if JSON_SLOW 1616#if JSON_SLOW
1348 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1617 types_true = get_bool ("Types::Serialiser::true");
1349#endif 1618#endif
1350 return SvREFCNT_inc (json_true); 1619 return newSVsv (types_true);
1351 } 1620 }
1352 else 1621 else
1353 ERR ("'true' expected"); 1622 ERR ("'true' expected");
1354 1623
1355 break; 1624 break;
1357 case 'f': 1626 case 'f':
1358 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1627 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1359 { 1628 {
1360 dec->cur += 5; 1629 dec->cur += 5;
1361#if JSON_SLOW 1630#if JSON_SLOW
1362 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1631 types_false = get_bool ("Types::Serialiser::false");
1363#endif 1632#endif
1364 return SvREFCNT_inc (json_false); 1633 return newSVsv (types_false);
1365 } 1634 }
1366 else 1635 else
1367 ERR ("'false' expected"); 1636 ERR ("'false' expected");
1368 1637
1369 break; 1638 break;
1378 ERR ("'null' expected"); 1647 ERR ("'null' expected");
1379 1648
1380 break; 1649 break;
1381 1650
1382 default: 1651 default:
1383 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1652 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1384 break; 1653 break;
1385 } 1654 }
1386 1655
1387fail: 1656fail:
1388 return 0; 1657 return 0;
1389} 1658}
1390 1659
1391static SV * 1660static SV *
1392decode_json (SV *string, JSON *json, STRLEN *offset_return) 1661decode_json (SV *string, JSON *json, char **offset_return)
1393{ 1662{
1394 dec_t dec; 1663 dec_t dec;
1395 STRLEN offset;
1396 SV *sv; 1664 SV *sv;
1397 1665
1666 /* work around bugs in 5.10 where manipulating magic values
1667 * makes perl ignore the magic in subsequent accesses.
1668 * also make a copy of non-PV values, to get them into a clean
1669 * state (SvPV should do that, but it's buggy, see below).
1670 */
1398 SvGETMAGIC (string); 1671 /*SvGETMAGIC (string);*/
1672 if (SvMAGICAL (string) || !SvPOK (string))
1673 string = sv_2mortal (newSVsv (string));
1674
1399 SvUPGRADE (string, SVt_PV); 1675 SvUPGRADE (string, SVt_PV);
1400 1676
1401 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1677 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1678 * assertion with -DDEBUGGING, although SvCUR is documented to
1679 * return the xpv_cur field which certainly exists after upgrading.
1680 * according to nicholas clark, calling SvPOK fixes this.
1681 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1682 * and hope for the best.
1683 * Damnit, SvPV_nolen still trips over yet another assertion. This
1684 * assertion business is seriously broken, try yet another workaround
1685 * for the broken -DDEBUGGING.
1686 */
1687 {
1688#ifdef DEBUGGING
1689 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1690#else
1691 STRLEN offset = SvCUR (string);
1692#endif
1693
1694 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", 1695 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)); 1696 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1697 }
1404 1698
1405 if (json->flags & F_UTF8) 1699 if (DECODE_WANTS_OCTETS (json))
1406 sv_utf8_downgrade (string, 0); 1700 sv_utf8_downgrade (string, 0);
1407 else 1701 else
1408 sv_utf8_upgrade (string); 1702 sv_utf8_upgrade (string);
1409 1703
1410 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1704 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1411 1705
1412 dec.json = *json; 1706 dec.json = *json;
1413 dec.cur = SvPVX (string); 1707 dec.cur = SvPVX (string);
1414 dec.end = SvEND (string); 1708 dec.end = SvEND (string);
1415 dec.err = 0; 1709 dec.err = 0;
1416 dec.depth = 0; 1710 dec.depth = 0;
1417 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1418 1711
1419 if (dec.json.cb_object || dec.json.cb_sk_object) 1712 if (dec.json.cb_object || dec.json.cb_sk_object)
1420 dec.json.flags |= F_HOOK; 1713 dec.json.flags |= F_HOOK;
1421 1714
1422 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1715 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1423 1716
1424 decode_ws (&dec); 1717 decode_ws (&dec);
1425 sv = decode_sv (&dec); 1718 sv = decode_sv (&dec);
1719
1720 if (offset_return)
1721 *offset_return = dec.cur;
1426 1722
1427 if (!(offset_return || !sv)) 1723 if (!(offset_return || !sv))
1428 { 1724 {
1429 // check for trailing garbage 1725 // check for trailing garbage
1430 decode_ws (&dec); 1726 decode_ws (&dec);
1433 { 1729 {
1434 dec.err = "garbage after JSON object"; 1730 dec.err = "garbage after JSON object";
1435 SvREFCNT_dec (sv); 1731 SvREFCNT_dec (sv);
1436 sv = 0; 1732 sv = 0;
1437 } 1733 }
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 } 1734 }
1449 1735
1450 if (!sv) 1736 if (!sv)
1451 { 1737 {
1452 SV *uni = sv_newmortal (); 1738 SV *uni = sv_newmortal ();
1458 SAVEVPTR (PL_curcop); 1744 SAVEVPTR (PL_curcop);
1459 PL_curcop = &cop; 1745 PL_curcop = &cop;
1460 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1746 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1461 LEAVE; 1747 LEAVE;
1462 1748
1463 croak ("%s, at character offset %d [\"%s\"]", 1749 croak ("%s, at character offset %d (before \"%s\")",
1464 dec.err, 1750 dec.err,
1465 (int)offset, 1751 (int)ptr_to_index (string, dec.cur),
1466 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1752 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1467 } 1753 }
1468 1754
1469 sv = sv_2mortal (sv); 1755 sv = sv_2mortal (sv);
1470 1756
1479 1765
1480static void 1766static void
1481incr_parse (JSON *self) 1767incr_parse (JSON *self)
1482{ 1768{
1483 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1769 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1770
1771 // the state machine here is a bit convoluted and could be simplified a lot
1772 // but this would make it slower, so...
1484 1773
1485 for (;;) 1774 for (;;)
1486 { 1775 {
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 1776 //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) 1777 switch (self->incr_mode)
1489 { 1778 {
1490 // only used for intiial whitespace skipping 1779 // only used for initial whitespace skipping
1491 case INCR_M_WS: 1780 case INCR_M_WS:
1492 for (;;) 1781 for (;;)
1493 { 1782 {
1494 if (*p > 0x20) 1783 if (*p > 0x20)
1495 { 1784 {
1785 if (*p == '#')
1786 {
1787 self->incr_mode = INCR_M_C0;
1788 goto incr_m_c;
1789 }
1790 else
1791 {
1496 self->incr_mode = INCR_M_JSON; 1792 self->incr_mode = INCR_M_JSON;
1497 goto incr_m_json; 1793 goto incr_m_json;
1794 }
1498 } 1795 }
1499 else if (!*p) 1796 else if (!*p)
1500 goto interrupt; 1797 goto interrupt;
1501 1798
1502 ++p; 1799 ++p;
1508 goto interrupt; 1805 goto interrupt;
1509 1806
1510 ++p; 1807 ++p;
1511 self->incr_mode = INCR_M_STR; 1808 self->incr_mode = INCR_M_STR;
1512 goto incr_m_str; 1809 goto incr_m_str;
1810
1811 // inside #-style comments
1812 case INCR_M_C0:
1813 case INCR_M_C1:
1814 incr_m_c:
1815 for (;;)
1816 {
1817 if (*p == '\n')
1818 {
1819 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1820 break;
1821 }
1822 else if (!*p)
1823 goto interrupt;
1824
1825 ++p;
1826 }
1827
1828 break;
1513 1829
1514 // inside a string 1830 // inside a string
1515 case INCR_M_STR: 1831 case INCR_M_STR:
1516 incr_m_str: 1832 incr_m_str:
1517 for (;;) 1833 for (;;)
1568 self->incr_mode = INCR_M_STR; 1884 self->incr_mode = INCR_M_STR;
1569 goto incr_m_str; 1885 goto incr_m_str;
1570 1886
1571 case '[': 1887 case '[':
1572 case '{': 1888 case '{':
1889 case '(':
1573 ++self->incr_nest; 1890 if (++self->incr_nest > self->max_depth)
1891 croak (ERR_NESTING_EXCEEDED);
1574 break; 1892 break;
1575 1893
1576 case ']': 1894 case ']':
1577 case '}': 1895 case '}':
1578 if (!--self->incr_nest) 1896 if (--self->incr_nest <= 0)
1579 goto interrupt; 1897 goto interrupt;
1898 break;
1899
1900 case ')':
1901 --self->incr_nest;
1902 break;
1903
1904 case '#':
1905 self->incr_mode = INCR_M_C1;
1906 goto incr_m_c;
1580 } 1907 }
1581 } 1908 }
1582 } 1909 }
1583 1910
1584 modechange: 1911 modechange:
1585 ; 1912 ;
1586 } 1913 }
1587 1914
1588interrupt: 1915interrupt:
1589 self->incr_pos = p - SvPVX (self->incr_text); 1916 self->incr_pos = p - SvPVX (self->incr_text);
1917 //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 1918 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1591} 1919}
1592 1920
1593///////////////////////////////////////////////////////////////////////////// 1921/////////////////////////////////////////////////////////////////////////////
1594// XS interface functions 1922// XS interface functions
1604 i >= '0' && i <= '9' ? i - '0' 1932 i >= '0' && i <= '9' ? i - '0'
1605 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1933 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1606 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1934 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1607 : -1; 1935 : -1;
1608 1936
1609 json_stash = gv_stashpv ("JSON::XS" , 1); 1937 json_stash = gv_stashpv ("JSON::XS" , 1);
1610 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1938 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1611 1939
1612 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1940 types_true = get_bool ("Types::Serialiser::true");
1613 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1941 types_false = get_bool ("Types::Serialiser::false");
1942
1943 sv_json = newSVpv ("JSON", 0);
1944 SvREADONLY_on (sv_json);
1945
1946 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1614} 1947}
1615 1948
1616PROTOTYPES: DISABLE 1949PROTOTYPES: DISABLE
1617 1950
1618void CLONE (...) 1951void CLONE (...)
1619 CODE: 1952 CODE:
1620 json_stash = 0; 1953 json_stash = 0;
1621 json_boolean_stash = 0; 1954 types_boolean_stash = 0;
1622 1955
1623void new (char *klass) 1956void new (char *klass)
1624 PPCODE: 1957 PPCODE:
1625{ 1958{
1626 SV *pv = NEWSV (0, sizeof (JSON)); 1959 SV *pv = NEWSV (0, sizeof (JSON));
1627 SvPOK_only (pv); 1960 SvPOK_only (pv);
1628 Zero (SvPVX (pv), 1, JSON); 1961 json_init ((JSON *)SvPVX (pv));
1629 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1630 XPUSHs (sv_2mortal (sv_bless ( 1962 XPUSHs (sv_2mortal (sv_bless (
1631 newRV_noinc (pv), 1963 newRV_noinc (pv),
1632 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1964 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1633 ))); 1965 )));
1634} 1966}
1646 allow_nonref = F_ALLOW_NONREF 1978 allow_nonref = F_ALLOW_NONREF
1647 shrink = F_SHRINK 1979 shrink = F_SHRINK
1648 allow_blessed = F_ALLOW_BLESSED 1980 allow_blessed = F_ALLOW_BLESSED
1649 convert_blessed = F_CONV_BLESSED 1981 convert_blessed = F_CONV_BLESSED
1650 relaxed = F_RELAXED 1982 relaxed = F_RELAXED
1983 allow_unknown = F_ALLOW_UNKNOWN
1984 allow_tags = F_ALLOW_TAGS
1651 PPCODE: 1985 PPCODE:
1652{ 1986{
1653 if (enable) 1987 if (enable)
1654 self->flags |= ix; 1988 self->flags |= ix;
1655 else 1989 else
1670 get_allow_nonref = F_ALLOW_NONREF 2004 get_allow_nonref = F_ALLOW_NONREF
1671 get_shrink = F_SHRINK 2005 get_shrink = F_SHRINK
1672 get_allow_blessed = F_ALLOW_BLESSED 2006 get_allow_blessed = F_ALLOW_BLESSED
1673 get_convert_blessed = F_CONV_BLESSED 2007 get_convert_blessed = F_CONV_BLESSED
1674 get_relaxed = F_RELAXED 2008 get_relaxed = F_RELAXED
2009 get_allow_unknown = F_ALLOW_UNKNOWN
2010 get_allow_tags = F_ALLOW_TAGS
1675 PPCODE: 2011 PPCODE:
1676 XPUSHs (boolSV (self->flags & ix)); 2012 XPUSHs (boolSV (self->flags & ix));
1677 2013
1678void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2014void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1679 PPCODE: 2015 PPCODE:
1680{ 2016 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)); 2017 XPUSHs (ST (0));
1691}
1692 2018
1693U32 get_max_depth (JSON *self) 2019U32 get_max_depth (JSON *self)
1694 CODE: 2020 CODE:
1695 RETVAL = DEC_DEPTH (self->flags); 2021 RETVAL = self->max_depth;
1696 OUTPUT: 2022 OUTPUT:
1697 RETVAL 2023 RETVAL
1698 2024
1699void max_size (JSON *self, UV max_size = 0) 2025void max_size (JSON *self, U32 max_size = 0)
1700 PPCODE: 2026 PPCODE:
1701{ 2027 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)); 2028 XPUSHs (ST (0));
1713}
1714 2029
1715int get_max_size (JSON *self) 2030int get_max_size (JSON *self)
1716 CODE: 2031 CODE:
1717 RETVAL = DEC_SIZE (self->flags); 2032 RETVAL = self->max_size;
1718 OUTPUT: 2033 OUTPUT:
1719 RETVAL 2034 RETVAL
1720 2035
1721void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2036void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1722 PPCODE: 2037 PPCODE:
1728} 2043}
1729 2044
1730void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2045void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1731 PPCODE: 2046 PPCODE:
1732{ 2047{
1733 if (!self->cb_sk_object) 2048 if (!self->cb_sk_object)
1734 self->cb_sk_object = newHV (); 2049 self->cb_sk_object = newHV ();
1735 2050
1736 if (SvOK (cb)) 2051 if (SvOK (cb))
1737 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2052 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1738 else 2053 else
1749 XPUSHs (ST (0)); 2064 XPUSHs (ST (0));
1750} 2065}
1751 2066
1752void encode (JSON *self, SV *scalar) 2067void encode (JSON *self, SV *scalar)
1753 PPCODE: 2068 PPCODE:
1754 XPUSHs (encode_json (scalar, self)); 2069 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2070 XPUSHs (scalar);
1755 2071
1756void decode (JSON *self, SV *jsonstr) 2072void decode (JSON *self, SV *jsonstr)
1757 PPCODE: 2073 PPCODE:
1758 XPUSHs (decode_json (jsonstr, self, 0)); 2074 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2075 XPUSHs (jsonstr);
1759 2076
1760void decode_prefix (JSON *self, SV *jsonstr) 2077void decode_prefix (JSON *self, SV *jsonstr)
1761 PPCODE: 2078 PPCODE:
1762{ 2079{
1763 STRLEN offset; 2080 SV *sv;
2081 char *offset;
2082 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1764 EXTEND (SP, 2); 2083 EXTEND (SP, 2);
1765 PUSHs (decode_json (jsonstr, self, &offset)); 2084 PUSHs (sv);
1766 PUSHs (sv_2mortal (newSVuv (offset))); 2085 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1767} 2086}
1768 2087
1769void incr_parse (JSON *self, SV *jsonstr = 0) 2088void incr_parse (JSON *self, SV *jsonstr = 0)
1770 PPCODE: 2089 PPCODE:
1771{ 2090{
1772 if (!self->incr_text) 2091 if (!self->incr_text)
1773 self->incr_text = newSVpvn ("", 0); 2092 self->incr_text = newSVpvn ("", 0);
2093
2094 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2095 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2096 if (DECODE_WANTS_OCTETS (self))
2097 {
2098 if (self->incr_pos)
2099 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2100 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2101
2102 sv_utf8_downgrade (self->incr_text, 0);
2103 }
2104 else
2105 {
2106 sv_utf8_upgrade (self->incr_text);
2107
2108 if (self->incr_pos)
2109 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2110 - (U8 *)SvPVX (self->incr_text);
2111 }
1774 2112
1775 // append data, if any 2113 // append data, if any
1776 if (jsonstr) 2114 if (jsonstr)
1777 { 2115 {
2116 /* make sure both strings have same encoding */
1778 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2117 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2118 if (SvUTF8 (jsonstr))
2119 sv_utf8_downgrade (jsonstr, 0);
1779 { 2120 else
1780 /* utf-8-ness differs, need to upgrade */
1781 sv_utf8_upgrade (self->incr_text); 2121 sv_utf8_upgrade (jsonstr);
1782 2122
1783 if (self->incr_pos) 2123 /* 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 { 2124 {
1789 STRLEN len; 2125 STRLEN len;
1790 const char *str = SvPV (jsonstr, len); 2126 const char *str = SvPV (jsonstr, len);
1791 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2127 STRLEN cur = SvCUR (self->incr_text);
2128
2129 if (SvLEN (self->incr_text) <= cur + len)
2130 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2131
1792 Move (str, SvEND (self->incr_text), len, char); 2132 Move (str, SvEND (self->incr_text), len, char);
1793 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2133 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 2134 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1795 } 2135 }
1796 } 2136 }
1797 2137
1798 if (GIMME_V != G_VOID) 2138 if (GIMME_V != G_VOID)
1799 do 2139 do
1800 { 2140 {
2141 SV *sv;
1801 STRLEN offset; 2142 char *offset;
1802
1803 incr_parse (self);
1804 2143
1805 if (!INCR_DONE (self)) 2144 if (!INCR_DONE (self))
2145 {
2146 incr_parse (self);
2147
2148 if (self->incr_pos > self->max_size && self->max_size)
2149 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2150 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2151
2152 if (!INCR_DONE (self))
2153 {
2154 // as an optimisation, do not accumulate white space in the incr buffer
2155 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2156 {
2157 self->incr_pos = 0;
2158 SvCUR_set (self->incr_text, 0);
2159 }
2160
1806 break; 2161 break;
2162 }
2163 }
1807 2164
1808 XPUSHs (decode_json (self->incr_text, self, &offset)); 2165 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2166 XPUSHs (sv);
1809 2167
1810 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1811 self->incr_pos -= offset; 2168 self->incr_pos -= offset - SvPVX (self->incr_text);
1812 self->incr_nest = 0; 2169 self->incr_nest = 0;
1813 self->incr_mode = 0; 2170 self->incr_mode = 0;
2171
2172 sv_chop (self->incr_text, offset);
1814 } 2173 }
1815 while (GIMME_V == G_ARRAY); 2174 while (GIMME_V == G_ARRAY);
1816} 2175}
1817 2176
1818SV *incr_text (JSON *self) 2177SV *incr_text (JSON *self)
1819 ATTRS: lvalue 2178 ATTRS: lvalue
1820 CODE: 2179 CODE:
1821{ 2180{
1822 if (self->incr_pos) 2181 if (self->incr_pos)
1823 croak ("incr_text can only be called after a successful incr_parse call in scalar context"); 2182 croak ("incr_text can not be called when the incremental parser already started parsing");
1824 2183
1825 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; 2184 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1826} 2185}
1827 OUTPUT: 2186 OUTPUT:
1828 RETVAL 2187 RETVAL
1837 self->incr_nest = 0; 2196 self->incr_nest = 0;
1838 self->incr_mode = 0; 2197 self->incr_mode = 0;
1839 } 2198 }
1840} 2199}
1841 2200
2201void incr_reset (JSON *self)
2202 CODE:
2203{
2204 SvREFCNT_dec (self->incr_text);
2205 self->incr_text = 0;
2206 self->incr_pos = 0;
2207 self->incr_nest = 0;
2208 self->incr_mode = 0;
2209}
2210
1842void DESTROY (JSON *self) 2211void DESTROY (JSON *self)
1843 CODE: 2212 CODE:
1844 SvREFCNT_dec (self->cb_sk_object); 2213 SvREFCNT_dec (self->cb_sk_object);
1845 SvREFCNT_dec (self->cb_object); 2214 SvREFCNT_dec (self->cb_object);
1846 SvREFCNT_dec (self->incr_text); 2215 SvREFCNT_dec (self->incr_text);
1847 2216
1848PROTOTYPES: ENABLE 2217PROTOTYPES: ENABLE
1849 2218
1850void encode_json (SV *scalar) 2219void encode_json (SV *scalar)
1851 ALIAS:
1852 to_json_ = 0
1853 encode_json = F_UTF8
1854 PPCODE: 2220 PPCODE:
1855{ 2221{
1856 JSON json = { F_DEFAULT | ix }; 2222 JSON json;
1857 XPUSHs (encode_json (scalar, &json)); 2223 json_init (&json);
2224 json.flags |= F_UTF8;
2225 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2226 XPUSHs (scalar);
1858} 2227}
1859 2228
1860void decode_json (SV *jsonstr) 2229void decode_json (SV *jsonstr)
1861 ALIAS:
1862 from_json_ = 0
1863 decode_json = F_UTF8
1864 PPCODE: 2230 PPCODE:
1865{ 2231{
1866 JSON json = { F_DEFAULT | ix }; 2232 JSON json;
1867 XPUSHs (decode_json (jsonstr, &json, 0)); 2233 json_init (&json);
2234 json.flags |= F_UTF8;
2235 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2236 XPUSHs (jsonstr);
1868} 2237}
1869 2238
1870

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines