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 1.84 by root, Thu Mar 27 06:37:35 2008 UTC vs.
Revision 1.128 by root, Fri Oct 7 05:18:48 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
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#define F_ALLOW_UNKNOWN 0x00002000UL 48#define F_ALLOW_UNKNOWN 0x00002000UL
37 49#define F_ALLOW_TAGS 0x00004000UL
38#define F_MAXDEPTH 0xf8000000UL
39#define S_MAXDEPTH 27
40#define F_MAXSIZE 0x01f00000UL
41#define S_MAXSIZE 20
42#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 50#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
43 51
44#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
45#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
46
47#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 52#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
48#define F_DEFAULT (9UL << S_MAXDEPTH)
49 53
50#define INIT_SIZE 32 // initial scalar size to be allocated 54#define INIT_SIZE 32 // initial scalar size to be allocated
51#define INDENT_STEP 3 // spaces per indentation level 55#define INDENT_STEP 3 // spaces per indentation level
52 56
53#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)
54 60
55#define SB do { 61#define SB do {
56#define SE } while (0) 62#define SE } while (0)
57 63
58#if __GNUC__ >= 3 64#if __GNUC__ >= 3
68 74
69#define IN_RANGE_INC(type,val,beg,end) \ 75#define IN_RANGE_INC(type,val,beg,end) \
70 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 76 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
71 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 77 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
72 78
79#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
80
73#ifdef USE_ITHREADS 81#ifdef USE_ITHREADS
74# define JSON_SLOW 1 82# define JSON_SLOW 1
75# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
84# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
76#else 85#else
77# define JSON_SLOW 0 86# define JSON_SLOW 0
78# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
79#endif 89#endif
80 90
81static HV *json_stash, *json_boolean_stash; // JSON::XS:: 91// the amount of HEs to allocate on the stack, when sorting keys
82static SV *json_true, *json_false; 92#define STACK_HES 64
93
94static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
95static SV *bool_true, *bool_false, *sv_json;
83 96
84enum { 97enum {
85 INCR_M_WS = 0, // initial whitespace skipping, must be 0 98 INCR_M_WS = 0, // initial whitespace skipping, must be 0
86 INCR_M_STR, // inside string 99 INCR_M_STR, // inside string
87 INCR_M_BS, // inside backslash 100 INCR_M_BS, // inside backslash
101 INCR_M_C0, // inside comment in initial whitespace sequence
102 INCR_M_C1, // inside comment in other places
88 INCR_M_JSON // outside anything, count nesting 103 INCR_M_JSON // outside anything, count nesting
89}; 104};
90 105
91#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 106#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
92 107
93typedef struct { 108typedef struct {
94 U32 flags; 109 U32 flags;
110 U32 max_depth;
111 STRLEN max_size;
112
95 SV *cb_object; 113 SV *cb_object;
96 HV *cb_sk_object; 114 HV *cb_sk_object;
97 115
98 // for the incremental parser 116 // for the incremental parser
99 SV *incr_text; // the source text so far 117 SV *incr_text; // the source text so far
100 STRLEN incr_pos; // the current offset into the text 118 STRLEN incr_pos; // the current offset into the text
101 int incr_nest; // {[]}-nesting level 119 int incr_nest; // {[]}-nesting level
102 int incr_mode; 120 unsigned char incr_mode;
103} JSON; 121} JSON;
122
123INLINE void
124json_init (JSON *json)
125{
126 Zero (json, 1, JSON);
127 json->max_depth = 512;
128}
104 129
105///////////////////////////////////////////////////////////////////////////// 130/////////////////////////////////////////////////////////////////////////////
106// utility functions 131// utility functions
107 132
108INLINE SV * 133INLINE SV *
118 143
119INLINE void 144INLINE void
120shrink (SV *sv) 145shrink (SV *sv)
121{ 146{
122 sv_utf8_downgrade (sv, 1); 147 sv_utf8_downgrade (sv, 1);
148
123 if (SvLEN (sv) > SvCUR (sv) + 1) 149 if (SvLEN (sv) > SvCUR (sv) + 1)
124 { 150 {
125#ifdef SvPV_shrink_to_cur 151#ifdef SvPV_shrink_to_cur
126 SvPV_shrink_to_cur (sv); 152 SvPV_shrink_to_cur (sv);
127#elif defined (SvPV_renew) 153#elif defined (SvPV_renew)
171 *s++ = 0x80 | ( ch & 0x3f); 197 *s++ = 0x80 | ( ch & 0x3f);
172 198
173 return s; 199 return s;
174} 200}
175 201
202// convert offset pointer to character index, sv must be string
203static STRLEN
204ptr_to_index (SV *sv, char *offset)
205{
206 return SvUTF8 (sv)
207 ? utf8_distance (offset, SvPVX (sv))
208 : offset - SvPVX (sv);
209}
210
211/////////////////////////////////////////////////////////////////////////////
212// fp hell
213
214// scan a group of digits, and a trailing exponent
215static void
216json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
217{
218 UV uaccum = 0;
219 int eaccum = 0;
220
221 // if we recurse too deep, skip all remaining digits
222 // to avoid a stack overflow attack
223 if (expect_false (--maxdepth <= 0))
224 while (((U8)*s - '0') < 10)
225 ++s;
226
227 for (;;)
228 {
229 U8 dig = (U8)*s - '0';
230
231 if (expect_false (dig >= 10))
232 {
233 if (dig == (U8)((U8)'.' - (U8)'0'))
234 {
235 ++s;
236 json_atof_scan1 (s, accum, expo, 1, maxdepth);
237 }
238 else if ((dig | ' ') == 'e' - '0')
239 {
240 int exp2 = 0;
241 int neg = 0;
242
243 ++s;
244
245 if (*s == '-')
246 {
247 ++s;
248 neg = 1;
249 }
250 else if (*s == '+')
251 ++s;
252
253 while ((dig = (U8)*s - '0') < 10)
254 exp2 = exp2 * 10 + *s++ - '0';
255
256 *expo += neg ? -exp2 : exp2;
257 }
258
259 break;
260 }
261
262 ++s;
263
264 uaccum = uaccum * 10 + dig;
265 ++eaccum;
266
267 // if we have too many digits, then recurse for more
268 // we actually do this for rather few digits
269 if (uaccum >= (UV_MAX - 9) / 10)
270 {
271 if (postdp) *expo -= eaccum;
272 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
273 if (postdp) *expo += eaccum;
274
275 break;
276 }
277 }
278
279 // this relies greatly on the quality of the pow ()
280 // implementation of the platform, but a good
281 // implementation is hard to beat.
282 // (IEEE 754 conformant ones are required to be exact)
283 if (postdp) *expo -= eaccum;
284 *accum += uaccum * Perl_pow (10., *expo);
285 *expo += eaccum;
286}
287
288static NV
289json_atof (const char *s)
290{
291 NV accum = 0.;
292 int expo = 0;
293 int neg = 0;
294
295 if (*s == '-')
296 {
297 ++s;
298 neg = 1;
299 }
300
301 // a recursion depth of ten gives us >>500 bits
302 json_atof_scan1 (s, &accum, &expo, 0, 10);
303
304 return neg ? -accum : accum;
305}
306
307// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
308static int
309ref_bool_type (SV *sv)
310{
311 svtype svt = SvTYPE (sv);
312
313 if (svt < SVt_PVAV)
314 {
315 STRLEN len = 0;
316 char *pv = svt ? SvPV (sv, len) : 0;
317
318 if (len == 1)
319 if (*pv == '1')
320 return 1;
321 else if (*pv == '0')
322 return 0;
323 }
324
325 return -1;
326}
327
328// returns whether scalar is not a reference in the sense of allow_nonref
329static int
330json_nonref (SV *scalar)
331{
332 if (!SvROK (scalar))
333 return 1;
334
335 scalar = SvRV (scalar);
336
337 if (SvTYPE (scalar) >= SVt_PVMG)
338 {
339 if (SvSTASH (scalar) == bool_stash)
340 return 1;
341
342 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
343 return 1;
344 }
345
346 return 0;
347}
348
176///////////////////////////////////////////////////////////////////////////// 349/////////////////////////////////////////////////////////////////////////////
177// encoder 350// encoder
178 351
179// structure used for encoding JSON 352// structure used for encoding JSON
180typedef struct 353typedef struct
182 char *cur; // SvPVX (sv) + current output position 355 char *cur; // SvPVX (sv) + current output position
183 char *end; // SvEND (sv) 356 char *end; // SvEND (sv)
184 SV *sv; // result scalar 357 SV *sv; // result scalar
185 JSON json; 358 JSON json;
186 U32 indent; // indentation level 359 U32 indent; // indentation level
187 U32 maxdepth; // max. indentation/recursion level
188 UV limit; // escape character values >= this value when encoding 360 UV limit; // escape character values >= this value when encoding
189} enc_t; 361} enc_t;
190 362
191INLINE void 363INLINE void
192need (enc_t *enc, STRLEN len) 364need (enc_t *enc, STRLEN len)
193{ 365{
194 if (expect_false (enc->cur + len >= enc->end)) 366 if (expect_false (enc->cur + len >= enc->end))
195 { 367 {
196 STRLEN cur = enc->cur - SvPVX (enc->sv); 368 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
197 SvGROW (enc->sv, cur + len + 1); 369 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
198 enc->cur = SvPVX (enc->sv) + cur; 370 enc->cur = SvPVX (enc->sv) + cur;
199 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 371 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
200 } 372 }
201} 373}
202 374
277 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 449 (int)((uch - 0x10000) % 0x400 + 0xDC00));
278 enc->cur += 12; 450 enc->cur += 12;
279 } 451 }
280 else 452 else
281 { 453 {
282 static char hexdigit [16] = "0123456789abcdef";
283 need (enc, len += 5); 454 need (enc, len += 5);
284 *enc->cur++ = '\\'; 455 *enc->cur++ = '\\';
285 *enc->cur++ = 'u'; 456 *enc->cur++ = 'u';
286 *enc->cur++ = hexdigit [ uch >> 12 ]; 457 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
287 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 458 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
288 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 459 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
289 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 460 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
290 } 461 }
291 462
292 str += clen; 463 str += clen;
293 } 464 }
294 else if (enc->json.flags & F_LATIN1) 465 else if (enc->json.flags & F_LATIN1)
365static void 536static void
366encode_av (enc_t *enc, AV *av) 537encode_av (enc_t *enc, AV *av)
367{ 538{
368 int i, len = av_len (av); 539 int i, len = av_len (av);
369 540
370 if (enc->indent >= enc->maxdepth) 541 if (enc->indent >= enc->json.max_depth)
371 croak ("data structure too deep (hit recursion limit)"); 542 croak (ERR_NESTING_EXCEEDED);
372 543
373 encode_ch (enc, '['); 544 encode_ch (enc, '[');
374 545
375 if (len >= 0) 546 if (len >= 0)
376 { 547 {
377 encode_nl (enc); ++enc->indent; 548 encode_nl (enc); ++enc->indent;
378 549
379 for (i = 0; i <= len; ++i) 550 for (i = 0; i <= len; ++i)
391 encode_comma (enc); 562 encode_comma (enc);
392 } 563 }
393 564
394 encode_nl (enc); --enc->indent; encode_indent (enc); 565 encode_nl (enc); --enc->indent; encode_indent (enc);
395 } 566 }
396 567
397 encode_ch (enc, ']'); 568 encode_ch (enc, ']');
398} 569}
399 570
400static void 571static void
401encode_hk (enc_t *enc, HE *he) 572encode_hk (enc_t *enc, HE *he)
405 if (HeKLEN (he) == HEf_SVKEY) 576 if (HeKLEN (he) == HEf_SVKEY)
406 { 577 {
407 SV *sv = HeSVKEY (he); 578 SV *sv = HeSVKEY (he);
408 STRLEN len; 579 STRLEN len;
409 char *str; 580 char *str;
410 581
411 SvGETMAGIC (sv); 582 SvGETMAGIC (sv);
412 str = SvPV (sv, len); 583 str = SvPV (sv, len);
413 584
414 encode_str (enc, str, len, SvUTF8 (sv)); 585 encode_str (enc, str, len, SvUTF8 (sv));
415 } 586 }
451static void 622static void
452encode_hv (enc_t *enc, HV *hv) 623encode_hv (enc_t *enc, HV *hv)
453{ 624{
454 HE *he; 625 HE *he;
455 626
456 if (enc->indent >= enc->maxdepth) 627 if (enc->indent >= enc->json.max_depth)
457 croak ("data structure too deep (hit recursion limit)"); 628 croak (ERR_NESTING_EXCEEDED);
458 629
459 encode_ch (enc, '{'); 630 encode_ch (enc, '{');
460 631
461 // for canonical output we have to sort by keys first 632 // for canonical output we have to sort by keys first
462 // actually, this is mostly due to the stupid so-called 633 // actually, this is mostly due to the stupid so-called
463 // security workaround added somewhere in 5.8.x. 634 // security workaround added somewhere in 5.8.x
464 // that randomises hash orderings 635 // that randomises hash orderings
465 if (enc->json.flags & F_CANONICAL) 636 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
466 { 637 {
467 int count = hv_iterinit (hv); 638 int count = hv_iterinit (hv);
468 639
469 if (SvMAGICAL (hv)) 640 if (SvMAGICAL (hv))
470 { 641 {
481 } 652 }
482 653
483 if (count) 654 if (count)
484 { 655 {
485 int i, fast = 1; 656 int i, fast = 1;
486#if defined(__BORLANDC__) || defined(_MSC_VER) 657 HE *hes_stack [STACK_HES];
487 HE **hes = _alloca (count * sizeof (HE)); 658 HE **hes = hes_stack;
488#else 659
489 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 660 // allocate larger arrays on the heap
490#endif 661 if (count > STACK_HES)
662 {
663 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
664 hes = (HE **)SvPVX (sv);
665 }
491 666
492 i = 0; 667 i = 0;
493 while ((he = hv_iternext (hv))) 668 while ((he = hv_iternext (hv)))
494 { 669 {
495 hes [i++] = he; 670 hes [i++] = he;
564// encode objects, arrays and special \0=false and \1=true values. 739// encode objects, arrays and special \0=false and \1=true values.
565static void 740static void
566encode_rv (enc_t *enc, SV *sv) 741encode_rv (enc_t *enc, SV *sv)
567{ 742{
568 svtype svt; 743 svtype svt;
744 GV *method;
569 745
570 SvGETMAGIC (sv); 746 SvGETMAGIC (sv);
571 svt = SvTYPE (sv); 747 svt = SvTYPE (sv);
572 748
573 if (expect_false (SvOBJECT (sv))) 749 if (expect_false (SvOBJECT (sv)))
574 { 750 {
575 HV *stash = !JSON_SLOW || json_boolean_stash 751 HV *stash = SvSTASH (sv);
576 ? json_boolean_stash
577 : gv_stashpv ("JSON::XS::Boolean", 1);
578 752
579 if (SvSTASH (sv) == stash) 753 if (stash == bool_stash)
580 { 754 {
581 if (SvIV (sv)) 755 if (SvIV (sv))
582 encode_str (enc, "true", 4, 0); 756 encode_str (enc, "true", 4, 0);
583 else 757 else
584 encode_str (enc, "false", 5, 0); 758 encode_str (enc, "false", 5, 0);
585 } 759 }
760 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
761 {
762 int count;
763 dSP;
764
765 ENTER; SAVETMPS;
766 SAVESTACK_POS ();
767 PUSHMARK (SP);
768 EXTEND (SP, 2);
769 // we re-bless the reference to get overload and other niceties right
770 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
771 PUSHs (sv_json);
772
773 PUTBACK;
774 count = call_sv ((SV *)GvCV (method), G_ARRAY);
775 SPAGAIN;
776
777 // catch this surprisingly common error
778 if (SvROK (TOPs) && SvRV (TOPs) == sv)
779 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
780
781 encode_ch (enc, '(');
782 encode_ch (enc, '"');
783 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
784 encode_ch (enc, '"');
785 encode_ch (enc, ')');
786 encode_ch (enc, '[');
787
788 while (count)
789 {
790 encode_sv (enc, SP[1 - count--]);
791
792 if (count)
793 encode_ch (enc, ',');
794 }
795
796 encode_ch (enc, ']');
797
798 FREETMPS; LEAVE;
799 }
800 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
801 {
802 dSP;
803
804 ENTER; SAVETMPS;
805 PUSHMARK (SP);
806 // we re-bless the reference to get overload and other niceties right
807 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
808
809 // calling with G_SCALAR ensures that we always get a 1 return value
810 PUTBACK;
811 call_sv ((SV *)GvCV (method), G_SCALAR);
812 SPAGAIN;
813
814 // catch this surprisingly common error
815 if (SvROK (TOPs) && SvRV (TOPs) == sv)
816 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
817
818 sv = POPs;
819 PUTBACK;
820
821 encode_sv (enc, sv);
822
823 FREETMPS; LEAVE;
824 }
825 else if (enc->json.flags & F_ALLOW_BLESSED)
826 encode_str (enc, "null", 4, 0);
586 else 827 else
587 {
588#if 0
589 if (0 && sv_derived_from (rv, "JSON::Literal"))
590 {
591 // not yet
592 }
593#endif
594 if (enc->json.flags & F_CONV_BLESSED)
595 {
596 // we re-bless the reference to get overload and other niceties right
597 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
598
599 if (to_json)
600 {
601 dSP;
602
603 ENTER; SAVETMPS; PUSHMARK (SP);
604 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
605
606 // calling with G_SCALAR ensures that we always get a 1 return value
607 PUTBACK;
608 call_sv ((SV *)GvCV (to_json), G_SCALAR);
609 SPAGAIN;
610
611 // catch this surprisingly common error
612 if (SvROK (TOPs) && SvRV (TOPs) == sv)
613 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
614
615 sv = POPs;
616 PUTBACK;
617
618 encode_sv (enc, sv);
619
620 FREETMPS; LEAVE;
621 }
622 else if (enc->json.flags & F_ALLOW_BLESSED)
623 encode_str (enc, "null", 4, 0);
624 else
625 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
626 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
627 }
628 else if (enc->json.flags & F_ALLOW_BLESSED)
629 encode_str (enc, "null", 4, 0);
630 else
631 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 828 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
632 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 829 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
633 }
634 } 830 }
635 else if (svt == SVt_PVHV) 831 else if (svt == SVt_PVHV)
636 encode_hv (enc, (HV *)sv); 832 encode_hv (enc, (HV *)sv);
637 else if (svt == SVt_PVAV) 833 else if (svt == SVt_PVAV)
638 encode_av (enc, (AV *)sv); 834 encode_av (enc, (AV *)sv);
639 else if (svt < SVt_PVAV) 835 else if (svt < SVt_PVAV)
640 { 836 {
641 STRLEN len = 0; 837 int bool_type = ref_bool_type (sv);
642 char *pv = svt ? SvPV (sv, len) : 0;
643 838
644 if (len == 1 && *pv == '1') 839 if (bool_type == 1)
645 encode_str (enc, "true", 4, 0); 840 encode_str (enc, "true", 4, 0);
646 else if (len == 1 && *pv == '0') 841 else if (bool_type == 0)
647 encode_str (enc, "false", 5, 0); 842 encode_str (enc, "false", 5, 0);
648 else if (enc->json.flags & F_ALLOW_UNKNOWN) 843 else if (enc->json.flags & F_ALLOW_UNKNOWN)
649 encode_str (enc, "null", 4, 0); 844 encode_str (enc, "null", 4, 0);
650 else 845 else
651 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 846 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
714 } 909 }
715 else 910 else
716 { 911 {
717 // large integer, use the (rather slow) snprintf way. 912 // large integer, use the (rather slow) snprintf way.
718 need (enc, IVUV_MAXCHARS); 913 need (enc, IVUV_MAXCHARS);
719 enc->cur += 914 enc->cur +=
720 SvIsUV(sv) 915 SvIsUV(sv)
721 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 916 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
722 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 917 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
723 } 918 }
724 } 919 }
725 else if (SvROK (sv)) 920 else if (SvROK (sv))
726 encode_rv (enc, SvRV (sv)); 921 encode_rv (enc, SvRV (sv));
727 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 922 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
728 encode_str (enc, "null", 4, 0); 923 encode_str (enc, "null", 4, 0);
729 else 924 else
730 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 925 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
731 SvPV_nolen (sv), SvFLAGS (sv)); 926 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
732} 927}
733 928
734static SV * 929static SV *
735encode_json (SV *scalar, JSON *json) 930encode_json (SV *scalar, JSON *json)
736{ 931{
737 enc_t enc; 932 enc_t enc;
738 933
739 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 934 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
740 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 935 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
741 936
742 enc.json = *json; 937 enc.json = *json;
743 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 938 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
744 enc.cur = SvPVX (enc.sv); 939 enc.cur = SvPVX (enc.sv);
745 enc.end = SvEND (enc.sv); 940 enc.end = SvEND (enc.sv);
746 enc.indent = 0; 941 enc.indent = 0;
747 enc.maxdepth = DEC_DEPTH (enc.json.flags);
748 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 942 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
749 : enc.json.flags & F_LATIN1 ? 0x000100UL 943 : enc.json.flags & F_LATIN1 ? 0x000100UL
750 : 0x110000UL; 944 : 0x110000UL;
751 945
752 SvPOK_only (enc.sv); 946 SvPOK_only (enc.sv);
753 encode_sv (&enc, scalar); 947 encode_sv (&enc, scalar);
948 encode_nl (&enc);
754 949
755 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 950 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
756 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 951 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
757 952
758 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 953 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
819 if (*dec->cur != ch) \ 1014 if (*dec->cur != ch) \
820 ERR (# ch " expected"); \ 1015 ERR (# ch " expected"); \
821 ++dec->cur; \ 1016 ++dec->cur; \
822 SE 1017 SE
823 1018
824#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1019#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
825#define DEC_DEC_DEPTH --dec->depth 1020#define DEC_DEC_DEPTH --dec->depth
826 1021
827static SV *decode_sv (dec_t *dec); 1022static SV *decode_sv (dec_t *dec);
828 1023
829static signed char decode_hexdigit[256]; 1024static signed char decode_hexdigit[256];
938 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1133 else if (expect_true (ch >= 0x20 && ch < 0x80))
939 *cur++ = ch; 1134 *cur++ = ch;
940 else if (ch >= 0x80) 1135 else if (ch >= 0x80)
941 { 1136 {
942 STRLEN clen; 1137 STRLEN clen;
943 UV uch;
944 1138
945 --dec_cur; 1139 --dec_cur;
946 1140
947 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1141 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
948 if (clen == (STRLEN)-1) 1142 if (clen == (STRLEN)-1)
949 ERR ("malformed UTF-8 character in JSON string"); 1143 ERR ("malformed UTF-8 character in JSON string");
950 1144
951 do 1145 do
952 *cur++ = *dec_cur++; 1146 *cur++ = *dec_cur++;
953 while (--clen); 1147 while (--clen);
954 1148
955 utf8 = 1; 1149 utf8 = 1;
956 } 1150 }
1151 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1152 *cur++ = ch;
957 else 1153 else
958 { 1154 {
959 --dec_cur; 1155 --dec_cur;
960 1156
961 if (!ch) 1157 if (!ch)
969 { 1165 {
970 STRLEN len = cur - buf; 1166 STRLEN len = cur - buf;
971 1167
972 if (sv) 1168 if (sv)
973 { 1169 {
974 SvGROW (sv, SvCUR (sv) + len + 1); 1170 STRLEN cur = SvCUR (sv);
1171
1172 if (SvLEN (sv) <= cur + len)
1173 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1174
975 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1175 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
976 SvCUR_set (sv, SvCUR (sv) + len); 1176 SvCUR_set (sv, SvCUR (sv) + len);
977 } 1177 }
978 else 1178 else
979 sv = newSVpvn (buf, len); 1179 sv = newSVpvn (buf, len);
1070 1270
1071 // special case the rather common 1..5-digit-int case 1271 // special case the rather common 1..5-digit-int case
1072 if (*start == '-') 1272 if (*start == '-')
1073 switch (len) 1273 switch (len)
1074 { 1274 {
1075 case 2: return newSViv (-( start [1] - '0' * 1)); 1275 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1076 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1276 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1077 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1277 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1078 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1278 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1079 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1279 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1080 } 1280 }
1081 else 1281 else
1082 switch (len) 1282 switch (len)
1083 { 1283 {
1084 case 1: return newSViv ( start [0] - '0' * 1); 1284 case 1: return newSViv ( start [0] - '0' * 1);
1085 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1285 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1086 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1286 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1087 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1287 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1088 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1288 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1089 } 1289 }
1090 1290
1091 { 1291 {
1092 UV uv; 1292 UV uv;
1093 int numtype = grok_number (start, len, &uv); 1293 int numtype = grok_number (start, len, &uv);
1102 } 1302 }
1103 1303
1104 len -= *start == '-' ? 1 : 0; 1304 len -= *start == '-' ? 1 : 0;
1105 1305
1106 // does not fit into IV or UV, try NV 1306 // does not fit into IV or UV, try NV
1107 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1307 if (len <= NV_DIG)
1108 #if defined (LDBL_DIG)
1109 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1110 #endif
1111 )
1112 // fits into NV without loss of precision 1308 // fits into NV without loss of precision
1113 return newSVnv (Atof (start)); 1309 return newSVnv (json_atof (start));
1114 1310
1115 // everything else fails, convert it to a string 1311 // everything else fails, convert it to a string
1116 return newSVpvn (start, dec->cur - start); 1312 return newSVpvn (start, dec->cur - start);
1117 } 1313 }
1118 1314
1119 // loss of precision here 1315 // loss of precision here
1120 return newSVnv (Atof (start)); 1316 return newSVnv (json_atof (start));
1121 1317
1122fail: 1318fail:
1123 return 0; 1319 return 0;
1124} 1320}
1125 1321
1149 if (*dec->cur == ']') 1345 if (*dec->cur == ']')
1150 { 1346 {
1151 ++dec->cur; 1347 ++dec->cur;
1152 break; 1348 break;
1153 } 1349 }
1154 1350
1155 if (*dec->cur != ',') 1351 if (*dec->cur != ',')
1156 ERR (", or ] expected while parsing array"); 1352 ERR (", or ] expected while parsing array");
1157 1353
1158 ++dec->cur; 1354 ++dec->cur;
1159 1355
1283 1479
1284 hv_iterinit (hv); 1480 hv_iterinit (hv);
1285 he = hv_iternext (hv); 1481 he = hv_iternext (hv);
1286 hv_iterinit (hv); 1482 hv_iterinit (hv);
1287 1483
1288 // the next line creates a mortal sv each time its called. 1484 // the next line creates a mortal sv each time it's called.
1289 // might want to optimise this for common cases. 1485 // might want to optimise this for common cases.
1290 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1486 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1291 1487
1292 if (cb) 1488 if (cb)
1293 { 1489 {
1294 dSP; 1490 dSP;
1295 int count; 1491 int count;
1296 1492
1297 ENTER; SAVETMPS; PUSHMARK (SP); 1493 ENTER; SAVETMPS;
1494 SAVESTACK_POS ();
1495 PUSHMARK (SP);
1298 XPUSHs (HeVAL (he)); 1496 XPUSHs (HeVAL (he));
1497 sv_2mortal (sv);
1299 1498
1300 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1499 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1301 1500
1302 if (count == 1) 1501 if (count == 1)
1303 { 1502 {
1304 sv = newSVsv (POPs); 1503 sv = newSVsv (POPs);
1305 FREETMPS; LEAVE; 1504 FREETMPS; LEAVE;
1306 return sv; 1505 return sv;
1307 } 1506 }
1308 1507
1508 SvREFCNT_inc (sv);
1309 FREETMPS; LEAVE; 1509 FREETMPS; LEAVE;
1310 } 1510 }
1311 } 1511 }
1312 1512
1313 if (dec->json.cb_object) 1513 if (dec->json.cb_object)
1314 { 1514 {
1315 dSP; 1515 dSP;
1316 int count; 1516 int count;
1317 1517
1318 ENTER; SAVETMPS; PUSHMARK (SP); 1518 ENTER; SAVETMPS;
1519 SAVESTACK_POS ();
1520 PUSHMARK (SP);
1319 XPUSHs (sv_2mortal (sv)); 1521 XPUSHs (sv_2mortal (sv));
1320 1522
1321 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1523 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1322 1524
1323 if (count == 1) 1525 if (count == 1)
1339 DEC_DEC_DEPTH; 1541 DEC_DEC_DEPTH;
1340 return 0; 1542 return 0;
1341} 1543}
1342 1544
1343static SV * 1545static SV *
1546decode_tag (dec_t *dec)
1547{
1548 SV *tag = 0;
1549 SV *val = 0;
1550
1551 if (!(dec->json.flags & F_ALLOW_TAGS))
1552 ERR ("malformed JSON string, neither array, object, number, string or atom");
1553
1554 ++dec->cur;
1555
1556 decode_ws (dec);
1557
1558 tag = decode_sv (dec);
1559 if (!tag)
1560 goto fail;
1561
1562 if (!SvPOK (tag))
1563 ERR ("malformed JSON string, (tag) must be a string");
1564
1565 decode_ws (dec);
1566
1567 if (*dec->cur != ')')
1568 ERR (") expected after tag");
1569
1570 ++dec->cur;
1571
1572 decode_ws (dec);
1573
1574 val = decode_sv (dec);
1575 if (!val)
1576 goto fail;
1577
1578 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1579 ERR ("malformed JSON string, tag value must be an array");
1580
1581 {
1582 AV *av = (AV *)SvRV (val);
1583 int i, len = av_len (av) + 1;
1584 HV *stash = gv_stashsv (tag, 0);
1585 SV *sv;
1586
1587 if (!stash)
1588 ERR ("cannot decode perl-object (package does not exist)");
1589
1590 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1591
1592 if (!method)
1593 ERR ("cannot decode perl-object (package does not have a THAW method)");
1594
1595 dSP;
1596
1597 ENTER; SAVETMPS;
1598 PUSHMARK (SP);
1599 EXTEND (SP, len + 2);
1600 // we re-bless the reference to get overload and other niceties right
1601 PUSHs (tag);
1602 PUSHs (sv_json);
1603
1604 for (i = 0; i < len; ++i)
1605 PUSHs (*av_fetch (av, i, 1));
1606
1607 PUTBACK;
1608 call_sv ((SV *)GvCV (method), G_SCALAR);
1609 SPAGAIN;
1610
1611 SvREFCNT_dec (tag);
1612 SvREFCNT_dec (val);
1613 sv = SvREFCNT_inc (POPs);
1614
1615 PUTBACK;
1616
1617 FREETMPS; LEAVE;
1618
1619 return sv;
1620 }
1621
1622fail:
1623 SvREFCNT_dec (tag);
1624 SvREFCNT_dec (val);
1625 return 0;
1626}
1627
1628static SV *
1344decode_sv (dec_t *dec) 1629decode_sv (dec_t *dec)
1345{ 1630{
1346 // the beauty of JSON: you need exactly one character lookahead 1631 // the beauty of JSON: you need exactly one character lookahead
1347 // to parse everything. 1632 // to parse everything.
1348 switch (*dec->cur) 1633 switch (*dec->cur)
1349 { 1634 {
1350 case '"': ++dec->cur; return decode_str (dec); 1635 case '"': ++dec->cur; return decode_str (dec);
1351 case '[': ++dec->cur; return decode_av (dec); 1636 case '[': ++dec->cur; return decode_av (dec);
1352 case '{': ++dec->cur; return decode_hv (dec); 1637 case '{': ++dec->cur; return decode_hv (dec);
1638 case '(': return decode_tag (dec);
1353 1639
1354 case '-': 1640 case '-':
1355 case '0': case '1': case '2': case '3': case '4': 1641 case '0': case '1': case '2': case '3': case '4':
1356 case '5': case '6': case '7': case '8': case '9': 1642 case '5': case '6': case '7': case '8': case '9':
1357 return decode_num (dec); 1643 return decode_num (dec);
1359 case 't': 1645 case 't':
1360 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1646 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1361 { 1647 {
1362 dec->cur += 4; 1648 dec->cur += 4;
1363#if JSON_SLOW 1649#if JSON_SLOW
1364 json_true = get_bool ("JSON::XS::true"); 1650 bool_true = get_bool ("Types::Serialiser::true");
1365#endif 1651#endif
1366 return newSVsv (json_true); 1652 return newSVsv (bool_true);
1367 } 1653 }
1368 else 1654 else
1369 ERR ("'true' expected"); 1655 ERR ("'true' expected");
1370 1656
1371 break; 1657 break;
1373 case 'f': 1659 case 'f':
1374 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1660 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1375 { 1661 {
1376 dec->cur += 5; 1662 dec->cur += 5;
1377#if JSON_SLOW 1663#if JSON_SLOW
1378 json_false = get_bool ("JSON::XS::false"); 1664 bool_false = get_bool ("Types::Serialiser::false");
1379#endif 1665#endif
1380 return newSVsv (json_false); 1666 return newSVsv (bool_false);
1381 } 1667 }
1382 else 1668 else
1383 ERR ("'false' expected"); 1669 ERR ("'false' expected");
1384 1670
1385 break; 1671 break;
1394 ERR ("'null' expected"); 1680 ERR ("'null' expected");
1395 1681
1396 break; 1682 break;
1397 1683
1398 default: 1684 default:
1399 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1685 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1400 break; 1686 break;
1401 } 1687 }
1402 1688
1403fail: 1689fail:
1404 return 0; 1690 return 0;
1405} 1691}
1406 1692
1407static SV * 1693static SV *
1408decode_json (SV *string, JSON *json, STRLEN *offset_return) 1694decode_json (SV *string, JSON *json, char **offset_return)
1409{ 1695{
1410 dec_t dec; 1696 dec_t dec;
1411 STRLEN offset;
1412 SV *sv; 1697 SV *sv;
1413 1698
1699 /* work around bugs in 5.10 where manipulating magic values
1700 * makes perl ignore the magic in subsequent accesses.
1701 * also make a copy of non-PV values, to get them into a clean
1702 * state (SvPV should do that, but it's buggy, see below).
1703 *
1704 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1705 * as reported by Reini Urban.
1706 */
1414 SvGETMAGIC (string); 1707 /*SvGETMAGIC (string);*/
1708 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1709 string = sv_2mortal (newSVsv (string));
1710
1415 SvUPGRADE (string, SVt_PV); 1711 SvUPGRADE (string, SVt_PV);
1416 1712
1417 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1713 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1714 * assertion with -DDEBUGGING, although SvCUR is documented to
1715 * return the xpv_cur field which certainly exists after upgrading.
1716 * according to nicholas clark, calling SvPOK fixes this.
1717 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1718 * and hope for the best.
1719 * Damnit, SvPV_nolen still trips over yet another assertion. This
1720 * assertion business is seriously broken, try yet another workaround
1721 * for the broken -DDEBUGGING.
1722 */
1723 {
1724#ifdef DEBUGGING
1725 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1726#else
1727 STRLEN offset = SvCUR (string);
1728#endif
1729
1730 if (offset > json->max_size && json->max_size)
1418 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1731 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1419 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1732 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1733 }
1420 1734
1421 if (json->flags & F_UTF8) 1735 if (DECODE_WANTS_OCTETS (json))
1422 sv_utf8_downgrade (string, 0); 1736 sv_utf8_downgrade (string, 0);
1423 else 1737 else
1424 sv_utf8_upgrade (string); 1738 sv_utf8_upgrade (string);
1425 1739
1426 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1740 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1427 1741
1428 dec.json = *json; 1742 dec.json = *json;
1429 dec.cur = SvPVX (string); 1743 dec.cur = SvPVX (string);
1430 dec.end = SvEND (string); 1744 dec.end = SvEND (string);
1431 dec.err = 0; 1745 dec.err = 0;
1432 dec.depth = 0; 1746 dec.depth = 0;
1433 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1434 1747
1435 if (dec.json.cb_object || dec.json.cb_sk_object) 1748 if (dec.json.cb_object || dec.json.cb_sk_object)
1436 dec.json.flags |= F_HOOK; 1749 dec.json.flags |= F_HOOK;
1437 1750
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1751 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439 1752
1440 decode_ws (&dec); 1753 decode_ws (&dec);
1441 sv = decode_sv (&dec); 1754 sv = decode_sv (&dec);
1755
1756 if (offset_return)
1757 *offset_return = dec.cur;
1442 1758
1443 if (!(offset_return || !sv)) 1759 if (!(offset_return || !sv))
1444 { 1760 {
1445 // check for trailing garbage 1761 // check for trailing garbage
1446 decode_ws (&dec); 1762 decode_ws (&dec);
1449 { 1765 {
1450 dec.err = "garbage after JSON object"; 1766 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv); 1767 SvREFCNT_dec (sv);
1452 sv = 0; 1768 sv = 0;
1453 } 1769 }
1454 }
1455
1456 if (offset_return || !sv)
1457 {
1458 offset = dec.json.flags & F_UTF8
1459 ? dec.cur - SvPVX (string)
1460 : utf8_distance (dec.cur, SvPVX (string));
1461
1462 if (offset_return)
1463 *offset_return = offset;
1464 } 1770 }
1465 1771
1466 if (!sv) 1772 if (!sv)
1467 { 1773 {
1468 SV *uni = sv_newmortal (); 1774 SV *uni = sv_newmortal ();
1474 SAVEVPTR (PL_curcop); 1780 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop; 1781 PL_curcop = &cop;
1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1782 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE; 1783 LEAVE;
1478 1784
1479 croak ("%s, at character offset %d [\"%s\"]", 1785 croak ("%s, at character offset %d (before \"%s\")",
1480 dec.err, 1786 dec.err,
1481 (int)offset, 1787 (int)ptr_to_index (string, dec.cur),
1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1788 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1483 } 1789 }
1484 1790
1485 sv = sv_2mortal (sv); 1791 sv = sv_2mortal (sv);
1486 1792
1487 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1793 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1488 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1794 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1489 1795
1490 return sv; 1796 return sv;
1491} 1797}
1492 1798
1495 1801
1496static void 1802static void
1497incr_parse (JSON *self) 1803incr_parse (JSON *self)
1498{ 1804{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1805 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1806
1807 // the state machine here is a bit convoluted and could be simplified a lot
1808 // but this would make it slower, so...
1500 1809
1501 for (;;) 1810 for (;;)
1502 { 1811 {
1503 //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 1812 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1504 switch (self->incr_mode) 1813 switch (self->incr_mode)
1505 { 1814 {
1506 // only used for intiial whitespace skipping 1815 // only used for initial whitespace skipping
1507 case INCR_M_WS: 1816 case INCR_M_WS:
1508 for (;;) 1817 for (;;)
1509 { 1818 {
1510 if (*p > 0x20) 1819 if (*p > 0x20)
1511 { 1820 {
1821 if (*p == '#')
1822 {
1823 self->incr_mode = INCR_M_C0;
1824 goto incr_m_c;
1825 }
1826 else
1827 {
1512 self->incr_mode = INCR_M_JSON; 1828 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json; 1829 goto incr_m_json;
1830 }
1514 } 1831 }
1515 else if (!*p) 1832 else if (!*p)
1516 goto interrupt; 1833 goto interrupt;
1517 1834
1518 ++p; 1835 ++p;
1524 goto interrupt; 1841 goto interrupt;
1525 1842
1526 ++p; 1843 ++p;
1527 self->incr_mode = INCR_M_STR; 1844 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str; 1845 goto incr_m_str;
1846
1847 // inside #-style comments
1848 case INCR_M_C0:
1849 case INCR_M_C1:
1850 incr_m_c:
1851 for (;;)
1852 {
1853 if (*p == '\n')
1854 {
1855 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1856 break;
1857 }
1858 else if (!*p)
1859 goto interrupt;
1860
1861 ++p;
1862 }
1863
1864 break;
1529 1865
1530 // inside a string 1866 // inside a string
1531 case INCR_M_STR: 1867 case INCR_M_STR:
1532 incr_m_str: 1868 incr_m_str:
1533 for (;;) 1869 for (;;)
1584 self->incr_mode = INCR_M_STR; 1920 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str; 1921 goto incr_m_str;
1586 1922
1587 case '[': 1923 case '[':
1588 case '{': 1924 case '{':
1925 case '(':
1589 ++self->incr_nest; 1926 if (++self->incr_nest > self->max_depth)
1927 croak (ERR_NESTING_EXCEEDED);
1590 break; 1928 break;
1591 1929
1592 case ']': 1930 case ']':
1593 case '}': 1931 case '}':
1594 if (!--self->incr_nest) 1932 if (--self->incr_nest <= 0)
1595 goto interrupt; 1933 goto interrupt;
1934 break;
1935
1936 case ')':
1937 --self->incr_nest;
1938 break;
1939
1940 case '#':
1941 self->incr_mode = INCR_M_C1;
1942 goto incr_m_c;
1596 } 1943 }
1597 } 1944 }
1598 } 1945 }
1599 1946
1600 modechange: 1947 modechange:
1601 ; 1948 ;
1602 } 1949 }
1603 1950
1604interrupt: 1951interrupt:
1605 self->incr_pos = p - SvPVX (self->incr_text); 1952 self->incr_pos = p - SvPVX (self->incr_text);
1953 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1606 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1954 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1607} 1955}
1608 1956
1609///////////////////////////////////////////////////////////////////////////// 1957/////////////////////////////////////////////////////////////////////////////
1610// XS interface functions 1958// XS interface functions
1620 i >= '0' && i <= '9' ? i - '0' 1968 i >= '0' && i <= '9' ? i - '0'
1621 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1969 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1622 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1970 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1623 : -1; 1971 : -1;
1624 1972
1625 json_stash = gv_stashpv ("JSON::XS" , 1); 1973 json_stash = gv_stashpv ("JSON::XS" , 1);
1626 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1974 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1627
1628 json_true = get_bool ("JSON::XS::true"); 1975 bool_true = get_bool ("Types::Serialiser::true");
1629 json_false = get_bool ("JSON::XS::false"); 1976 bool_false = get_bool ("Types::Serialiser::false");
1977
1978 sv_json = newSVpv ("JSON", 0);
1979 SvREADONLY_on (sv_json);
1980
1981 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1630} 1982}
1631 1983
1632PROTOTYPES: DISABLE 1984PROTOTYPES: DISABLE
1633 1985
1634void CLONE (...) 1986void CLONE (...)
1635 CODE: 1987 CODE:
1636 json_stash = 0; 1988 json_stash = 0;
1637 json_boolean_stash = 0; 1989 bool_stash = 0;
1638 1990
1639void new (char *klass) 1991void new (char *klass)
1640 PPCODE: 1992 PPCODE:
1641{ 1993{
1642 SV *pv = NEWSV (0, sizeof (JSON)); 1994 SV *pv = NEWSV (0, sizeof (JSON));
1643 SvPOK_only (pv); 1995 SvPOK_only (pv);
1644 Zero (SvPVX (pv), 1, JSON); 1996 json_init ((JSON *)SvPVX (pv));
1645 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1646 XPUSHs (sv_2mortal (sv_bless ( 1997 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv), 1998 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1999 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1649 ))); 2000 )));
1650} 2001}
1663 shrink = F_SHRINK 2014 shrink = F_SHRINK
1664 allow_blessed = F_ALLOW_BLESSED 2015 allow_blessed = F_ALLOW_BLESSED
1665 convert_blessed = F_CONV_BLESSED 2016 convert_blessed = F_CONV_BLESSED
1666 relaxed = F_RELAXED 2017 relaxed = F_RELAXED
1667 allow_unknown = F_ALLOW_UNKNOWN 2018 allow_unknown = F_ALLOW_UNKNOWN
2019 allow_tags = F_ALLOW_TAGS
1668 PPCODE: 2020 PPCODE:
1669{ 2021{
1670 if (enable) 2022 if (enable)
1671 self->flags |= ix; 2023 self->flags |= ix;
1672 else 2024 else
1688 get_shrink = F_SHRINK 2040 get_shrink = F_SHRINK
1689 get_allow_blessed = F_ALLOW_BLESSED 2041 get_allow_blessed = F_ALLOW_BLESSED
1690 get_convert_blessed = F_CONV_BLESSED 2042 get_convert_blessed = F_CONV_BLESSED
1691 get_relaxed = F_RELAXED 2043 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN 2044 get_allow_unknown = F_ALLOW_UNKNOWN
2045 get_allow_tags = F_ALLOW_TAGS
1693 PPCODE: 2046 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix)); 2047 XPUSHs (boolSV (self->flags & ix));
1695 2048
1696void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2049void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1697 PPCODE: 2050 PPCODE:
1698{ 2051 self->max_depth = max_depth;
1699 UV log2 = 0;
1700
1701 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1702
1703 while ((1UL << log2) < max_depth)
1704 ++log2;
1705
1706 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1707
1708 XPUSHs (ST (0)); 2052 XPUSHs (ST (0));
1709}
1710 2053
1711U32 get_max_depth (JSON *self) 2054U32 get_max_depth (JSON *self)
1712 CODE: 2055 CODE:
1713 RETVAL = DEC_DEPTH (self->flags); 2056 RETVAL = self->max_depth;
1714 OUTPUT: 2057 OUTPUT:
1715 RETVAL 2058 RETVAL
1716 2059
1717void max_size (JSON *self, UV max_size = 0) 2060void max_size (JSON *self, U32 max_size = 0)
1718 PPCODE: 2061 PPCODE:
1719{ 2062 self->max_size = max_size;
1720 UV log2 = 0;
1721
1722 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1723 if (max_size == 1) max_size = 2;
1724
1725 while ((1UL << log2) < max_size)
1726 ++log2;
1727
1728 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1729
1730 XPUSHs (ST (0)); 2063 XPUSHs (ST (0));
1731}
1732 2064
1733int get_max_size (JSON *self) 2065int get_max_size (JSON *self)
1734 CODE: 2066 CODE:
1735 RETVAL = DEC_SIZE (self->flags); 2067 RETVAL = self->max_size;
1736 OUTPUT: 2068 OUTPUT:
1737 RETVAL 2069 RETVAL
1738 2070
1739void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2071void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1740 PPCODE: 2072 PPCODE:
1746} 2078}
1747 2079
1748void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2080void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1749 PPCODE: 2081 PPCODE:
1750{ 2082{
1751 if (!self->cb_sk_object) 2083 if (!self->cb_sk_object)
1752 self->cb_sk_object = newHV (); 2084 self->cb_sk_object = newHV ();
1753 2085
1754 if (SvOK (cb)) 2086 if (SvOK (cb))
1755 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2087 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1756 else 2088 else
1767 XPUSHs (ST (0)); 2099 XPUSHs (ST (0));
1768} 2100}
1769 2101
1770void encode (JSON *self, SV *scalar) 2102void encode (JSON *self, SV *scalar)
1771 PPCODE: 2103 PPCODE:
1772 XPUSHs (encode_json (scalar, self)); 2104 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2105 XPUSHs (scalar);
1773 2106
1774void decode (JSON *self, SV *jsonstr) 2107void decode (JSON *self, SV *jsonstr)
1775 PPCODE: 2108 PPCODE:
1776 XPUSHs (decode_json (jsonstr, self, 0)); 2109 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2110 XPUSHs (jsonstr);
1777 2111
1778void decode_prefix (JSON *self, SV *jsonstr) 2112void decode_prefix (JSON *self, SV *jsonstr)
1779 PPCODE: 2113 PPCODE:
1780{ 2114{
1781 STRLEN offset; 2115 SV *sv;
2116 char *offset;
2117 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1782 EXTEND (SP, 2); 2118 EXTEND (SP, 2);
1783 PUSHs (decode_json (jsonstr, self, &offset)); 2119 PUSHs (sv);
1784 PUSHs (sv_2mortal (newSVuv (offset))); 2120 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1785} 2121}
1786 2122
1787void incr_parse (JSON *self, SV *jsonstr = 0) 2123void incr_parse (JSON *self, SV *jsonstr = 0)
1788 PPCODE: 2124 PPCODE:
1789{ 2125{
1790 if (!self->incr_text) 2126 if (!self->incr_text)
1791 self->incr_text = newSVpvn ("", 0); 2127 self->incr_text = newSVpvn ("", 0);
2128
2129 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2130 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2131 if (DECODE_WANTS_OCTETS (self))
2132 {
2133 if (self->incr_pos)
2134 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2135 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2136
2137 sv_utf8_downgrade (self->incr_text, 0);
2138 }
2139 else
2140 {
2141 sv_utf8_upgrade (self->incr_text);
2142
2143 if (self->incr_pos)
2144 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2145 - (U8 *)SvPVX (self->incr_text);
2146 }
1792 2147
1793 // append data, if any 2148 // append data, if any
1794 if (jsonstr) 2149 if (jsonstr)
1795 { 2150 {
2151 /* make sure both strings have same encoding */
1796 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2152 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2153 if (SvUTF8 (jsonstr))
2154 sv_utf8_downgrade (jsonstr, 0);
1797 { 2155 else
1798 /* utf-8-ness differs, need to upgrade */
1799 sv_utf8_upgrade (self->incr_text); 2156 sv_utf8_upgrade (jsonstr);
1800 2157
1801 if (self->incr_pos) 2158 /* and then just blindly append */
1802 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1803 - (U8 *)SvPVX (self->incr_text);
1804 }
1805
1806 { 2159 {
1807 STRLEN len; 2160 STRLEN len;
1808 const char *str = SvPV (jsonstr, len); 2161 const char *str = SvPV (jsonstr, len);
1809 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2162 STRLEN cur = SvCUR (self->incr_text);
2163
2164 if (SvLEN (self->incr_text) <= cur + len)
2165 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2166
1810 Move (str, SvEND (self->incr_text), len, char); 2167 Move (str, SvEND (self->incr_text), len, char);
1811 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2168 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1812 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2169 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1813 } 2170 }
1814 } 2171 }
1815 2172
1816 if (GIMME_V != G_VOID) 2173 if (GIMME_V != G_VOID)
1817 do 2174 do
1818 { 2175 {
2176 SV *sv;
1819 STRLEN offset; 2177 char *offset;
1820 2178
1821 if (!INCR_DONE (self)) 2179 if (!INCR_DONE (self))
1822 { 2180 {
1823 incr_parse (self); 2181 incr_parse (self);
2182
2183 if (self->incr_pos > self->max_size && self->max_size)
2184 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2185 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2186
1824 if (!INCR_DONE (self)) 2187 if (!INCR_DONE (self))
2188 {
2189 // as an optimisation, do not accumulate white space in the incr buffer
2190 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2191 {
2192 self->incr_pos = 0;
2193 SvCUR_set (self->incr_text, 0);
2194 }
2195
1825 break; 2196 break;
2197 }
1826 } 2198 }
1827 2199
1828 XPUSHs (decode_json (self->incr_text, self, &offset)); 2200 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2201 XPUSHs (sv);
1829 2202
1830 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1831 self->incr_pos -= offset; 2203 self->incr_pos -= offset - SvPVX (self->incr_text);
1832 self->incr_nest = 0; 2204 self->incr_nest = 0;
1833 self->incr_mode = 0; 2205 self->incr_mode = 0;
2206
2207 sv_chop (self->incr_text, offset);
1834 } 2208 }
1835 while (GIMME_V == G_ARRAY); 2209 while (GIMME_V == G_ARRAY);
1836} 2210}
1837 2211
1838SV *incr_text (JSON *self) 2212SV *incr_text (JSON *self)
1857 self->incr_nest = 0; 2231 self->incr_nest = 0;
1858 self->incr_mode = 0; 2232 self->incr_mode = 0;
1859 } 2233 }
1860} 2234}
1861 2235
2236void incr_reset (JSON *self)
2237 CODE:
2238{
2239 SvREFCNT_dec (self->incr_text);
2240 self->incr_text = 0;
2241 self->incr_pos = 0;
2242 self->incr_nest = 0;
2243 self->incr_mode = 0;
2244}
2245
1862void DESTROY (JSON *self) 2246void DESTROY (JSON *self)
1863 CODE: 2247 CODE:
1864 SvREFCNT_dec (self->cb_sk_object); 2248 SvREFCNT_dec (self->cb_sk_object);
1865 SvREFCNT_dec (self->cb_object); 2249 SvREFCNT_dec (self->cb_object);
1866 SvREFCNT_dec (self->incr_text); 2250 SvREFCNT_dec (self->incr_text);
1867 2251
1868PROTOTYPES: ENABLE 2252PROTOTYPES: ENABLE
1869 2253
1870void encode_json (SV *scalar) 2254void encode_json (SV *scalar)
1871 ALIAS:
1872 to_json_ = 0
1873 encode_json = F_UTF8
1874 PPCODE: 2255 PPCODE:
1875{ 2256{
1876 JSON json = { F_DEFAULT | ix }; 2257 JSON json;
1877 XPUSHs (encode_json (scalar, &json)); 2258 json_init (&json);
2259 json.flags |= F_UTF8;
2260 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2261 XPUSHs (scalar);
1878} 2262}
1879 2263
1880void decode_json (SV *jsonstr) 2264void decode_json (SV *jsonstr)
1881 ALIAS:
1882 from_json_ = 0
1883 decode_json = F_UTF8
1884 PPCODE: 2265 PPCODE:
1885{ 2266{
1886 JSON json = { F_DEFAULT | ix }; 2267 JSON json;
1887 XPUSHs (decode_json (jsonstr, &json, 0)); 2268 json_init (&json);
2269 json.flags |= F_UTF8;
2270 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2271 XPUSHs (jsonstr);
1888} 2272}
1889 2273
1890

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines