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 1.87 by root, Tue Jun 3 06:43:45 2008 UTC vs.
Revision 1.129 by root, Wed Nov 16 18:06:34 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
49#define F_ALLOW_TAGS 0x00004000UL
37#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 50#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
38 51
39#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 52#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
40 53
41#define INIT_SIZE 32 // initial scalar size to be allocated 54#define INIT_SIZE 32 // initial scalar size to be allocated
42#define INDENT_STEP 3 // spaces per indentation level 55#define INDENT_STEP 3 // spaces per indentation level
43 56
44#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)
45 60
46#define SB do { 61#define SB do {
47#define SE } while (0) 62#define SE } while (0)
48 63
49#if __GNUC__ >= 3 64#if __GNUC__ >= 3
64#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" 79#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
65 80
66#ifdef USE_ITHREADS 81#ifdef USE_ITHREADS
67# define JSON_SLOW 1 82# define JSON_SLOW 1
68# 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))
69#else 85#else
70# define JSON_SLOW 0 86# define JSON_SLOW 0
71# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
72#endif 89#endif
73 90
74static HV *json_stash, *json_boolean_stash; // JSON::XS:: 91// the amount of HEs to allocate on the stack, when sorting keys
75static 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;
76 96
77enum { 97enum {
78 INCR_M_WS = 0, // initial whitespace skipping, must be 0 98 INCR_M_WS = 0, // initial whitespace skipping, must be 0
79 INCR_M_STR, // inside string 99 INCR_M_STR, // inside string
80 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
81 INCR_M_JSON // outside anything, count nesting 103 INCR_M_JSON // outside anything, count nesting
82}; 104};
83 105
84#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)
85 107
86typedef struct { 108typedef struct {
87 U32 flags; 109 U32 flags;
88 U32 max_depth; 110 U32 max_depth;
89 STRLEN max_size; 111 STRLEN max_size;
92 HV *cb_sk_object; 114 HV *cb_sk_object;
93 115
94 // for the incremental parser 116 // for the incremental parser
95 SV *incr_text; // the source text so far 117 SV *incr_text; // the source text so far
96 STRLEN incr_pos; // the current offset into the text 118 STRLEN incr_pos; // the current offset into the text
97 unsigned char incr_nest; // {[]}-nesting level 119 int incr_nest; // {[]}-nesting level
98 unsigned char incr_mode; 120 unsigned char incr_mode;
99} JSON; 121} JSON;
100 122
101INLINE void 123INLINE void
102json_init (JSON *json) 124json_init (JSON *json)
175 *s++ = 0x80 | ( ch & 0x3f); 197 *s++ = 0x80 | ( ch & 0x3f);
176 198
177 return s; 199 return s;
178} 200}
179 201
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
180///////////////////////////////////////////////////////////////////////////// 349/////////////////////////////////////////////////////////////////////////////
181// encoder 350// encoder
182 351
183// structure used for encoding JSON 352// structure used for encoding JSON
184typedef struct 353typedef struct
194INLINE void 363INLINE void
195need (enc_t *enc, STRLEN len) 364need (enc_t *enc, STRLEN len)
196{ 365{
197 if (expect_false (enc->cur + len >= enc->end)) 366 if (expect_false (enc->cur + len >= enc->end))
198 { 367 {
199 STRLEN cur = enc->cur - SvPVX (enc->sv); 368 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
200 SvGROW (enc->sv, cur + len + 1); 369 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
201 enc->cur = SvPVX (enc->sv) + cur; 370 enc->cur = SvPVX (enc->sv) + cur;
202 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 371 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
203 } 372 }
204} 373}
205 374
280 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 449 (int)((uch - 0x10000) % 0x400 + 0xDC00));
281 enc->cur += 12; 450 enc->cur += 12;
282 } 451 }
283 else 452 else
284 { 453 {
285 static char hexdigit [16] = "0123456789abcdef";
286 need (enc, len += 5); 454 need (enc, len += 5);
287 *enc->cur++ = '\\'; 455 *enc->cur++ = '\\';
288 *enc->cur++ = 'u'; 456 *enc->cur++ = 'u';
289 *enc->cur++ = hexdigit [ uch >> 12 ]; 457 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
290 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 458 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
291 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 459 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
292 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 460 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
293 } 461 }
294 462
295 str += clen; 463 str += clen;
296 } 464 }
297 else if (enc->json.flags & F_LATIN1) 465 else if (enc->json.flags & F_LATIN1)
372 540
373 if (enc->indent >= enc->json.max_depth) 541 if (enc->indent >= enc->json.max_depth)
374 croak (ERR_NESTING_EXCEEDED); 542 croak (ERR_NESTING_EXCEEDED);
375 543
376 encode_ch (enc, '['); 544 encode_ch (enc, '[');
377 545
378 if (len >= 0) 546 if (len >= 0)
379 { 547 {
380 encode_nl (enc); ++enc->indent; 548 encode_nl (enc); ++enc->indent;
381 549
382 for (i = 0; i <= len; ++i) 550 for (i = 0; i <= len; ++i)
394 encode_comma (enc); 562 encode_comma (enc);
395 } 563 }
396 564
397 encode_nl (enc); --enc->indent; encode_indent (enc); 565 encode_nl (enc); --enc->indent; encode_indent (enc);
398 } 566 }
399 567
400 encode_ch (enc, ']'); 568 encode_ch (enc, ']');
401} 569}
402 570
403static void 571static void
404encode_hk (enc_t *enc, HE *he) 572encode_hk (enc_t *enc, HE *he)
408 if (HeKLEN (he) == HEf_SVKEY) 576 if (HeKLEN (he) == HEf_SVKEY)
409 { 577 {
410 SV *sv = HeSVKEY (he); 578 SV *sv = HeSVKEY (he);
411 STRLEN len; 579 STRLEN len;
412 char *str; 580 char *str;
413 581
414 SvGETMAGIC (sv); 582 SvGETMAGIC (sv);
415 str = SvPV (sv, len); 583 str = SvPV (sv, len);
416 584
417 encode_str (enc, str, len, SvUTF8 (sv)); 585 encode_str (enc, str, len, SvUTF8 (sv));
418 } 586 }
461 629
462 encode_ch (enc, '{'); 630 encode_ch (enc, '{');
463 631
464 // for canonical output we have to sort by keys first 632 // for canonical output we have to sort by keys first
465 // actually, this is mostly due to the stupid so-called 633 // actually, this is mostly due to the stupid so-called
466 // security workaround added somewhere in 5.8.x. 634 // security workaround added somewhere in 5.8.x
467 // that randomises hash orderings 635 // that randomises hash orderings
468 if (enc->json.flags & F_CANONICAL) 636 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
469 { 637 {
470 int count = hv_iterinit (hv); 638 int count = hv_iterinit (hv);
471 639
472 if (SvMAGICAL (hv)) 640 if (SvMAGICAL (hv))
473 { 641 {
484 } 652 }
485 653
486 if (count) 654 if (count)
487 { 655 {
488 int i, fast = 1; 656 int i, fast = 1;
489#if defined(__BORLANDC__) || defined(_MSC_VER) 657 HE *hes_stack [STACK_HES];
490 HE **hes = _alloca (count * sizeof (HE)); 658 HE **hes = hes_stack;
491#else 659
492 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 660 // allocate larger arrays on the heap
493#endif 661 if (count > STACK_HES)
662 {
663 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
664 hes = (HE **)SvPVX (sv);
665 }
494 666
495 i = 0; 667 i = 0;
496 while ((he = hv_iternext (hv))) 668 while ((he = hv_iternext (hv)))
497 { 669 {
498 hes [i++] = he; 670 hes [i++] = he;
567// encode objects, arrays and special \0=false and \1=true values. 739// encode objects, arrays and special \0=false and \1=true values.
568static void 740static void
569encode_rv (enc_t *enc, SV *sv) 741encode_rv (enc_t *enc, SV *sv)
570{ 742{
571 svtype svt; 743 svtype svt;
744 GV *method;
572 745
573 SvGETMAGIC (sv); 746 SvGETMAGIC (sv);
574 svt = SvTYPE (sv); 747 svt = SvTYPE (sv);
575 748
576 if (expect_false (SvOBJECT (sv))) 749 if (expect_false (SvOBJECT (sv)))
577 { 750 {
578 HV *stash = !JSON_SLOW || json_boolean_stash 751 HV *stash = SvSTASH (sv);
579 ? json_boolean_stash
580 : gv_stashpv ("JSON::XS::Boolean", 1);
581 752
582 if (SvSTASH (sv) == stash) 753 if (stash == bool_stash)
583 { 754 {
584 if (SvIV (sv)) 755 if (SvIV (sv))
585 encode_str (enc, "true", 4, 0); 756 encode_str (enc, "true", 4, 0);
586 else 757 else
587 encode_str (enc, "false", 5, 0); 758 encode_str (enc, "false", 5, 0);
588 } 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);
589 else 827 else
590 {
591#if 0
592 if (0 && sv_derived_from (rv, "JSON::Literal"))
593 {
594 // not yet
595 }
596#endif
597 if (enc->json.flags & F_CONV_BLESSED)
598 {
599 // we re-bless the reference to get overload and other niceties right
600 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
601
602 if (to_json)
603 {
604 dSP;
605
606 ENTER; SAVETMPS; PUSHMARK (SP);
607 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
608
609 // calling with G_SCALAR ensures that we always get a 1 return value
610 PUTBACK;
611 call_sv ((SV *)GvCV (to_json), G_SCALAR);
612 SPAGAIN;
613
614 // catch this surprisingly common error
615 if (SvROK (TOPs) && SvRV (TOPs) == sv)
616 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
617
618 sv = POPs;
619 PUTBACK;
620
621 encode_sv (enc, sv);
622
623 FREETMPS; LEAVE;
624 }
625 else if (enc->json.flags & F_ALLOW_BLESSED)
626 encode_str (enc, "null", 4, 0);
627 else
628 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
629 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
630 }
631 else if (enc->json.flags & F_ALLOW_BLESSED)
632 encode_str (enc, "null", 4, 0);
633 else
634 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)",
635 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 829 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
636 }
637 } 830 }
638 else if (svt == SVt_PVHV) 831 else if (svt == SVt_PVHV)
639 encode_hv (enc, (HV *)sv); 832 encode_hv (enc, (HV *)sv);
640 else if (svt == SVt_PVAV) 833 else if (svt == SVt_PVAV)
641 encode_av (enc, (AV *)sv); 834 encode_av (enc, (AV *)sv);
642 else if (svt < SVt_PVAV) 835 else if (svt < SVt_PVAV)
643 { 836 {
644 STRLEN len = 0; 837 int bool_type = ref_bool_type (sv);
645 char *pv = svt ? SvPV (sv, len) : 0;
646 838
647 if (len == 1 && *pv == '1') 839 if (bool_type == 1)
648 encode_str (enc, "true", 4, 0); 840 encode_str (enc, "true", 4, 0);
649 else if (len == 1 && *pv == '0') 841 else if (bool_type == 0)
650 encode_str (enc, "false", 5, 0); 842 encode_str (enc, "false", 5, 0);
651 else if (enc->json.flags & F_ALLOW_UNKNOWN) 843 else if (enc->json.flags & F_ALLOW_UNKNOWN)
652 encode_str (enc, "null", 4, 0); 844 encode_str (enc, "null", 4, 0);
653 else 845 else
654 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",
717 } 909 }
718 else 910 else
719 { 911 {
720 // large integer, use the (rather slow) snprintf way. 912 // large integer, use the (rather slow) snprintf way.
721 need (enc, IVUV_MAXCHARS); 913 need (enc, IVUV_MAXCHARS);
722 enc->cur += 914 enc->cur +=
723 SvIsUV(sv) 915 SvIsUV(sv)
724 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 916 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
725 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 917 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
726 } 918 }
727 } 919 }
728 else if (SvROK (sv)) 920 else if (SvROK (sv))
729 encode_rv (enc, SvRV (sv)); 921 encode_rv (enc, SvRV (sv));
730 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 922 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
731 encode_str (enc, "null", 4, 0); 923 encode_str (enc, "null", 4, 0);
732 else 924 else
733 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",
734 SvPV_nolen (sv), SvFLAGS (sv)); 926 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
735} 927}
736 928
737static SV * 929static SV *
738encode_json (SV *scalar, JSON *json) 930encode_json (SV *scalar, JSON *json)
739{ 931{
740 enc_t enc; 932 enc_t enc;
741 933
742 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 934 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
743 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)");
744 936
745 enc.json = *json; 937 enc.json = *json;
746 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 938 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
747 enc.cur = SvPVX (enc.sv); 939 enc.cur = SvPVX (enc.sv);
751 : enc.json.flags & F_LATIN1 ? 0x000100UL 943 : enc.json.flags & F_LATIN1 ? 0x000100UL
752 : 0x110000UL; 944 : 0x110000UL;
753 945
754 SvPOK_only (enc.sv); 946 SvPOK_only (enc.sv);
755 encode_sv (&enc, scalar); 947 encode_sv (&enc, scalar);
948 encode_nl (&enc);
756 949
757 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 950 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
758 *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
759 952
760 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 953 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
940 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1133 else if (expect_true (ch >= 0x20 && ch < 0x80))
941 *cur++ = ch; 1134 *cur++ = ch;
942 else if (ch >= 0x80) 1135 else if (ch >= 0x80)
943 { 1136 {
944 STRLEN clen; 1137 STRLEN clen;
945 UV uch;
946 1138
947 --dec_cur; 1139 --dec_cur;
948 1140
949 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1141 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
950 if (clen == (STRLEN)-1) 1142 if (clen == (STRLEN)-1)
951 ERR ("malformed UTF-8 character in JSON string"); 1143 ERR ("malformed UTF-8 character in JSON string");
952 1144
953 do 1145 do
954 *cur++ = *dec_cur++; 1146 *cur++ = *dec_cur++;
955 while (--clen); 1147 while (--clen);
956 1148
957 utf8 = 1; 1149 utf8 = 1;
958 } 1150 }
1151 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1152 *cur++ = ch;
959 else 1153 else
960 { 1154 {
961 --dec_cur; 1155 --dec_cur;
962 1156
963 if (!ch) 1157 if (!ch)
971 { 1165 {
972 STRLEN len = cur - buf; 1166 STRLEN len = cur - buf;
973 1167
974 if (sv) 1168 if (sv)
975 { 1169 {
976 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
977 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1175 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
978 SvCUR_set (sv, SvCUR (sv) + len); 1176 SvCUR_set (sv, SvCUR (sv) + len);
979 } 1177 }
980 else 1178 else
981 sv = newSVpvn (buf, len); 1179 sv = newSVpvn (buf, len);
1072 1270
1073 // special case the rather common 1..5-digit-int case 1271 // special case the rather common 1..5-digit-int case
1074 if (*start == '-') 1272 if (*start == '-')
1075 switch (len) 1273 switch (len)
1076 { 1274 {
1077 case 2: return newSViv (-( start [1] - '0' * 1)); 1275 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1078 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1276 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1079 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));
1080 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));
1081 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));
1082 } 1280 }
1083 else 1281 else
1084 switch (len) 1282 switch (len)
1085 { 1283 {
1086 case 1: return newSViv ( start [0] - '0' * 1); 1284 case 1: return newSViv ( start [0] - '0' * 1);
1087 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1285 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1088 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);
1089 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);
1090 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);
1091 } 1289 }
1092 1290
1093 { 1291 {
1094 UV uv; 1292 UV uv;
1095 int numtype = grok_number (start, len, &uv); 1293 int numtype = grok_number (start, len, &uv);
1104 } 1302 }
1105 1303
1106 len -= *start == '-' ? 1 : 0; 1304 len -= *start == '-' ? 1 : 0;
1107 1305
1108 // does not fit into IV or UV, try NV 1306 // does not fit into IV or UV, try NV
1109 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1307 if (len <= NV_DIG)
1110 #if defined (LDBL_DIG)
1111 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1112 #endif
1113 )
1114 // fits into NV without loss of precision 1308 // fits into NV without loss of precision
1115 return newSVnv (Atof (start)); 1309 return newSVnv (json_atof (start));
1116 1310
1117 // everything else fails, convert it to a string 1311 // everything else fails, convert it to a string
1118 return newSVpvn (start, dec->cur - start); 1312 return newSVpvn (start, dec->cur - start);
1119 } 1313 }
1120 1314
1121 // loss of precision here 1315 // loss of precision here
1122 return newSVnv (Atof (start)); 1316 return newSVnv (json_atof (start));
1123 1317
1124fail: 1318fail:
1125 return 0; 1319 return 0;
1126} 1320}
1127 1321
1151 if (*dec->cur == ']') 1345 if (*dec->cur == ']')
1152 { 1346 {
1153 ++dec->cur; 1347 ++dec->cur;
1154 break; 1348 break;
1155 } 1349 }
1156 1350
1157 if (*dec->cur != ',') 1351 if (*dec->cur != ',')
1158 ERR (", or ] expected while parsing array"); 1352 ERR (", or ] expected while parsing array");
1159 1353
1160 ++dec->cur; 1354 ++dec->cur;
1161 1355
1285 1479
1286 hv_iterinit (hv); 1480 hv_iterinit (hv);
1287 he = hv_iternext (hv); 1481 he = hv_iternext (hv);
1288 hv_iterinit (hv); 1482 hv_iterinit (hv);
1289 1483
1290 // the next line creates a mortal sv each time its called. 1484 // the next line creates a mortal sv each time it's called.
1291 // might want to optimise this for common cases. 1485 // might want to optimise this for common cases.
1292 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);
1293 1487
1294 if (cb) 1488 if (cb)
1295 { 1489 {
1296 dSP; 1490 dSP;
1297 int count; 1491 int count;
1298 1492
1299 ENTER; SAVETMPS; PUSHMARK (SP); 1493 ENTER; SAVETMPS;
1494 SAVESTACK_POS ();
1495 PUSHMARK (SP);
1300 XPUSHs (HeVAL (he)); 1496 XPUSHs (HeVAL (he));
1497 sv_2mortal (sv);
1301 1498
1302 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1499 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1303 1500
1304 if (count == 1) 1501 if (count == 1)
1305 { 1502 {
1306 sv = newSVsv (POPs); 1503 sv = newSVsv (POPs);
1307 FREETMPS; LEAVE; 1504 FREETMPS; LEAVE;
1308 return sv; 1505 return sv;
1309 } 1506 }
1310 1507
1508 SvREFCNT_inc (sv);
1311 FREETMPS; LEAVE; 1509 FREETMPS; LEAVE;
1312 } 1510 }
1313 } 1511 }
1314 1512
1315 if (dec->json.cb_object) 1513 if (dec->json.cb_object)
1316 { 1514 {
1317 dSP; 1515 dSP;
1318 int count; 1516 int count;
1319 1517
1320 ENTER; SAVETMPS; PUSHMARK (SP); 1518 ENTER; SAVETMPS;
1519 SAVESTACK_POS ();
1520 PUSHMARK (SP);
1321 XPUSHs (sv_2mortal (sv)); 1521 XPUSHs (sv_2mortal (sv));
1322 1522
1323 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1523 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1324 1524
1325 if (count == 1) 1525 if (count == 1)
1341 DEC_DEC_DEPTH; 1541 DEC_DEC_DEPTH;
1342 return 0; 1542 return 0;
1343} 1543}
1344 1544
1345static 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 *
1346decode_sv (dec_t *dec) 1629decode_sv (dec_t *dec)
1347{ 1630{
1348 // the beauty of JSON: you need exactly one character lookahead 1631 // the beauty of JSON: you need exactly one character lookahead
1349 // to parse everything. 1632 // to parse everything.
1350 switch (*dec->cur) 1633 switch (*dec->cur)
1351 { 1634 {
1352 case '"': ++dec->cur; return decode_str (dec); 1635 case '"': ++dec->cur; return decode_str (dec);
1353 case '[': ++dec->cur; return decode_av (dec); 1636 case '[': ++dec->cur; return decode_av (dec);
1354 case '{': ++dec->cur; return decode_hv (dec); 1637 case '{': ++dec->cur; return decode_hv (dec);
1638 case '(': return decode_tag (dec);
1355 1639
1356 case '-': 1640 case '-':
1357 case '0': case '1': case '2': case '3': case '4': 1641 case '0': case '1': case '2': case '3': case '4':
1358 case '5': case '6': case '7': case '8': case '9': 1642 case '5': case '6': case '7': case '8': case '9':
1359 return decode_num (dec); 1643 return decode_num (dec);
1361 case 't': 1645 case 't':
1362 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1646 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1363 { 1647 {
1364 dec->cur += 4; 1648 dec->cur += 4;
1365#if JSON_SLOW 1649#if JSON_SLOW
1366 json_true = get_bool ("JSON::XS::true"); 1650 bool_true = get_bool ("Types::Serialiser::true");
1367#endif 1651#endif
1368 return newSVsv (json_true); 1652 return newSVsv (bool_true);
1369 } 1653 }
1370 else 1654 else
1371 ERR ("'true' expected"); 1655 ERR ("'true' expected");
1372 1656
1373 break; 1657 break;
1375 case 'f': 1659 case 'f':
1376 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1660 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1377 { 1661 {
1378 dec->cur += 5; 1662 dec->cur += 5;
1379#if JSON_SLOW 1663#if JSON_SLOW
1380 json_false = get_bool ("JSON::XS::false"); 1664 bool_false = get_bool ("Types::Serialiser::false");
1381#endif 1665#endif
1382 return newSVsv (json_false); 1666 return newSVsv (bool_false);
1383 } 1667 }
1384 else 1668 else
1385 ERR ("'false' expected"); 1669 ERR ("'false' expected");
1386 1670
1387 break; 1671 break;
1396 ERR ("'null' expected"); 1680 ERR ("'null' expected");
1397 1681
1398 break; 1682 break;
1399 1683
1400 default: 1684 default:
1401 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1685 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1402 break; 1686 break;
1403 } 1687 }
1404 1688
1405fail: 1689fail:
1406 return 0; 1690 return 0;
1408 1692
1409static SV * 1693static SV *
1410decode_json (SV *string, JSON *json, STRLEN *offset_return) 1694decode_json (SV *string, JSON *json, STRLEN *offset_return)
1411{ 1695{
1412 dec_t dec; 1696 dec_t dec;
1413 STRLEN offset;
1414 SV *sv; 1697 SV *sv;
1415 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 */
1416 SvGETMAGIC (string); 1707 /*SvGETMAGIC (string);*/
1708 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1709 string = sv_2mortal (newSVsv (string));
1710
1417 SvUPGRADE (string, SVt_PV); 1711 SvUPGRADE (string, SVt_PV);
1418 1712
1419 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1713 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1420 * assertion with -DDEBUGGING, although SvCUR is documented to 1714 * assertion with -DDEBUGGING, although SvCUR is documented to
1421 * return the xpv_cur field which certainly exists after upgrading. 1715 * return the xpv_cur field which certainly exists after upgrading.
1422 * according to nicholas clark, calling SvPOK fixes this. 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.
1423 */ 1722 */
1424 SvPOK (string); 1723 {
1724#ifdef DEBUGGING
1725 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1726#else
1727 STRLEN offset = SvCUR (string);
1728#endif
1425 1729
1426 if (SvCUR (string) > json->max_size && json->max_size) 1730 if (offset > json->max_size && json->max_size)
1427 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",
1428 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1732 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1733 }
1429 1734
1430 if (json->flags & F_UTF8) 1735 if (DECODE_WANTS_OCTETS (json))
1431 sv_utf8_downgrade (string, 0); 1736 sv_utf8_downgrade (string, 0);
1432 else 1737 else
1433 sv_utf8_upgrade (string); 1738 sv_utf8_upgrade (string);
1434 1739
1435 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1740 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1446 *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
1447 1752
1448 decode_ws (&dec); 1753 decode_ws (&dec);
1449 sv = decode_sv (&dec); 1754 sv = decode_sv (&dec);
1450 1755
1451 if (!(offset_return || !sv)) 1756 if (offset_return)
1757 *offset_return = dec.cur - SvPVX (string);
1758 else if (sv)
1452 { 1759 {
1453 // check for trailing garbage 1760 // check for trailing garbage
1454 decode_ws (&dec); 1761 decode_ws (&dec);
1455 1762
1456 if (*dec.cur) 1763 if (*dec.cur)
1457 { 1764 {
1458 dec.err = "garbage after JSON object"; 1765 dec.err = "garbage after JSON object";
1459 SvREFCNT_dec (sv); 1766 SvREFCNT_dec (sv);
1460 sv = 0; 1767 sv = 0;
1461 } 1768 }
1462 }
1463
1464 if (offset_return || !sv)
1465 {
1466 offset = dec.json.flags & F_UTF8
1467 ? dec.cur - SvPVX (string)
1468 : utf8_distance (dec.cur, SvPVX (string));
1469
1470 if (offset_return)
1471 *offset_return = offset;
1472 } 1769 }
1473 1770
1474 if (!sv) 1771 if (!sv)
1475 { 1772 {
1476 SV *uni = sv_newmortal (); 1773 SV *uni = sv_newmortal ();
1482 SAVEVPTR (PL_curcop); 1779 SAVEVPTR (PL_curcop);
1483 PL_curcop = &cop; 1780 PL_curcop = &cop;
1484 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1781 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1485 LEAVE; 1782 LEAVE;
1486 1783
1487 croak ("%s, at character offset %d [\"%s\"]", 1784 croak ("%s, at character offset %d (before \"%s\")",
1488 dec.err, 1785 dec.err,
1489 (int)offset, 1786 (int)ptr_to_index (string, dec.cur),
1490 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1787 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1491 } 1788 }
1492 1789
1493 sv = sv_2mortal (sv); 1790 sv = sv_2mortal (sv);
1494 1791
1495 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1792 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1496 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1793 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1497 1794
1498 return sv; 1795 return sv;
1499} 1796}
1500 1797
1503 1800
1504static void 1801static void
1505incr_parse (JSON *self) 1802incr_parse (JSON *self)
1506{ 1803{
1507 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1804 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1805
1806 // the state machine here is a bit convoluted and could be simplified a lot
1807 // but this would make it slower, so...
1508 1808
1509 for (;;) 1809 for (;;)
1510 { 1810 {
1511 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D 1811 //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
1512 switch (self->incr_mode) 1812 switch (self->incr_mode)
1513 { 1813 {
1514 // only used for intiial whitespace skipping 1814 // only used for initial whitespace skipping
1515 case INCR_M_WS: 1815 case INCR_M_WS:
1516 for (;;) 1816 for (;;)
1517 { 1817 {
1518 if (*p > 0x20) 1818 if (*p > 0x20)
1519 { 1819 {
1820 if (*p == '#')
1821 {
1822 self->incr_mode = INCR_M_C0;
1823 goto incr_m_c;
1824 }
1825 else
1826 {
1520 self->incr_mode = INCR_M_JSON; 1827 self->incr_mode = INCR_M_JSON;
1521 goto incr_m_json; 1828 goto incr_m_json;
1829 }
1522 } 1830 }
1523 else if (!*p) 1831 else if (!*p)
1524 goto interrupt; 1832 goto interrupt;
1525 1833
1526 ++p; 1834 ++p;
1532 goto interrupt; 1840 goto interrupt;
1533 1841
1534 ++p; 1842 ++p;
1535 self->incr_mode = INCR_M_STR; 1843 self->incr_mode = INCR_M_STR;
1536 goto incr_m_str; 1844 goto incr_m_str;
1845
1846 // inside #-style comments
1847 case INCR_M_C0:
1848 case INCR_M_C1:
1849 incr_m_c:
1850 for (;;)
1851 {
1852 if (*p == '\n')
1853 {
1854 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1855 break;
1856 }
1857 else if (!*p)
1858 goto interrupt;
1859
1860 ++p;
1861 }
1862
1863 break;
1537 1864
1538 // inside a string 1865 // inside a string
1539 case INCR_M_STR: 1866 case INCR_M_STR:
1540 incr_m_str: 1867 incr_m_str:
1541 for (;;) 1868 for (;;)
1592 self->incr_mode = INCR_M_STR; 1919 self->incr_mode = INCR_M_STR;
1593 goto incr_m_str; 1920 goto incr_m_str;
1594 1921
1595 case '[': 1922 case '[':
1596 case '{': 1923 case '{':
1924 case '(':
1597 if (++self->incr_nest > self->max_depth) 1925 if (++self->incr_nest > self->max_depth)
1598 croak (ERR_NESTING_EXCEEDED); 1926 croak (ERR_NESTING_EXCEEDED);
1599 break; 1927 break;
1600 1928
1601 case ']': 1929 case ']':
1602 case '}': 1930 case '}':
1603 if (!--self->incr_nest) 1931 if (--self->incr_nest <= 0)
1604 goto interrupt; 1932 goto interrupt;
1933 break;
1934
1935 case ')':
1936 --self->incr_nest;
1937 break;
1938
1939 case '#':
1940 self->incr_mode = INCR_M_C1;
1941 goto incr_m_c;
1605 } 1942 }
1606 } 1943 }
1607 } 1944 }
1608 1945
1609 modechange: 1946 modechange:
1610 ; 1947 ;
1611 } 1948 }
1612 1949
1613interrupt: 1950interrupt:
1614 self->incr_pos = p - SvPVX (self->incr_text); 1951 self->incr_pos = p - SvPVX (self->incr_text);
1952 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1615 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1953 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1616} 1954}
1617 1955
1618///////////////////////////////////////////////////////////////////////////// 1956/////////////////////////////////////////////////////////////////////////////
1619// XS interface functions 1957// XS interface functions
1629 i >= '0' && i <= '9' ? i - '0' 1967 i >= '0' && i <= '9' ? i - '0'
1630 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1968 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1631 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1969 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1632 : -1; 1970 : -1;
1633 1971
1634 json_stash = gv_stashpv ("JSON::XS" , 1); 1972 json_stash = gv_stashpv ("JSON::XS" , 1);
1635 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1973 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1636
1637 json_true = get_bool ("JSON::XS::true"); 1974 bool_true = get_bool ("Types::Serialiser::true");
1638 json_false = get_bool ("JSON::XS::false"); 1975 bool_false = get_bool ("Types::Serialiser::false");
1976
1977 sv_json = newSVpv ("JSON", 0);
1978 SvREADONLY_on (sv_json);
1979
1980 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1639} 1981}
1640 1982
1641PROTOTYPES: DISABLE 1983PROTOTYPES: DISABLE
1642 1984
1643void CLONE (...) 1985void CLONE (...)
1644 CODE: 1986 CODE:
1645 json_stash = 0; 1987 json_stash = 0;
1646 json_boolean_stash = 0; 1988 bool_stash = 0;
1647 1989
1648void new (char *klass) 1990void new (char *klass)
1649 PPCODE: 1991 PPCODE:
1650{ 1992{
1651 SV *pv = NEWSV (0, sizeof (JSON)); 1993 SV *pv = NEWSV (0, sizeof (JSON));
1652 SvPOK_only (pv); 1994 SvPOK_only (pv);
1653 json_init ((JSON *)SvPVX (pv)); 1995 json_init ((JSON *)SvPVX (pv));
1654 XPUSHs (sv_2mortal (sv_bless ( 1996 XPUSHs (sv_2mortal (sv_bless (
1655 newRV_noinc (pv), 1997 newRV_noinc (pv),
1656 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1998 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1671 shrink = F_SHRINK 2013 shrink = F_SHRINK
1672 allow_blessed = F_ALLOW_BLESSED 2014 allow_blessed = F_ALLOW_BLESSED
1673 convert_blessed = F_CONV_BLESSED 2015 convert_blessed = F_CONV_BLESSED
1674 relaxed = F_RELAXED 2016 relaxed = F_RELAXED
1675 allow_unknown = F_ALLOW_UNKNOWN 2017 allow_unknown = F_ALLOW_UNKNOWN
2018 allow_tags = F_ALLOW_TAGS
1676 PPCODE: 2019 PPCODE:
1677{ 2020{
1678 if (enable) 2021 if (enable)
1679 self->flags |= ix; 2022 self->flags |= ix;
1680 else 2023 else
1696 get_shrink = F_SHRINK 2039 get_shrink = F_SHRINK
1697 get_allow_blessed = F_ALLOW_BLESSED 2040 get_allow_blessed = F_ALLOW_BLESSED
1698 get_convert_blessed = F_CONV_BLESSED 2041 get_convert_blessed = F_CONV_BLESSED
1699 get_relaxed = F_RELAXED 2042 get_relaxed = F_RELAXED
1700 get_allow_unknown = F_ALLOW_UNKNOWN 2043 get_allow_unknown = F_ALLOW_UNKNOWN
2044 get_allow_tags = F_ALLOW_TAGS
1701 PPCODE: 2045 PPCODE:
1702 XPUSHs (boolSV (self->flags & ix)); 2046 XPUSHs (boolSV (self->flags & ix));
1703 2047
1704void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2048void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1705 PPCODE: 2049 PPCODE:
1733} 2077}
1734 2078
1735void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2079void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1736 PPCODE: 2080 PPCODE:
1737{ 2081{
1738 if (!self->cb_sk_object) 2082 if (!self->cb_sk_object)
1739 self->cb_sk_object = newHV (); 2083 self->cb_sk_object = newHV ();
1740 2084
1741 if (SvOK (cb)) 2085 if (SvOK (cb))
1742 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2086 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1743 else 2087 else
1754 XPUSHs (ST (0)); 2098 XPUSHs (ST (0));
1755} 2099}
1756 2100
1757void encode (JSON *self, SV *scalar) 2101void encode (JSON *self, SV *scalar)
1758 PPCODE: 2102 PPCODE:
1759 XPUSHs (encode_json (scalar, self)); 2103 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2104 XPUSHs (scalar);
1760 2105
1761void decode (JSON *self, SV *jsonstr) 2106void decode (JSON *self, SV *jsonstr)
1762 PPCODE: 2107 PPCODE:
1763 XPUSHs (decode_json (jsonstr, self, 0)); 2108 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2109 XPUSHs (jsonstr);
1764 2110
1765void decode_prefix (JSON *self, SV *jsonstr) 2111void decode_prefix (JSON *self, SV *jsonstr)
1766 PPCODE: 2112 PPCODE:
1767{ 2113{
2114 SV *sv;
1768 STRLEN offset; 2115 STRLEN offset;
2116 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1769 EXTEND (SP, 2); 2117 EXTEND (SP, 2);
1770 PUSHs (decode_json (jsonstr, self, &offset)); 2118 PUSHs (sv);
1771 PUSHs (sv_2mortal (newSVuv (offset))); 2119 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1772} 2120}
1773 2121
1774void incr_parse (JSON *self, SV *jsonstr = 0) 2122void incr_parse (JSON *self, SV *jsonstr = 0)
1775 PPCODE: 2123 PPCODE:
1776{ 2124{
1777 if (!self->incr_text) 2125 if (!self->incr_text)
1778 self->incr_text = newSVpvn ("", 0); 2126 self->incr_text = newSVpvn ("", 0);
2127
2128 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2129 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2130 if (DECODE_WANTS_OCTETS (self))
2131 {
2132 if (self->incr_pos)
2133 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2134 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2135
2136 sv_utf8_downgrade (self->incr_text, 0);
2137 }
2138 else
2139 {
2140 sv_utf8_upgrade (self->incr_text);
2141
2142 if (self->incr_pos)
2143 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2144 - (U8 *)SvPVX (self->incr_text);
2145 }
1779 2146
1780 // append data, if any 2147 // append data, if any
1781 if (jsonstr) 2148 if (jsonstr)
1782 { 2149 {
2150 /* make sure both strings have same encoding */
1783 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2151 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2152 if (SvUTF8 (jsonstr))
2153 sv_utf8_downgrade (jsonstr, 0);
1784 { 2154 else
1785 /* utf-8-ness differs, need to upgrade */
1786 sv_utf8_upgrade (self->incr_text); 2155 sv_utf8_upgrade (jsonstr);
1787 2156
1788 if (self->incr_pos) 2157 /* and then just blindly append */
1789 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1790 - (U8 *)SvPVX (self->incr_text);
1791 }
1792
1793 { 2158 {
1794 STRLEN len; 2159 STRLEN len;
1795 const char *str = SvPV (jsonstr, len); 2160 const char *str = SvPV (jsonstr, len);
1796 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2161 STRLEN cur = SvCUR (self->incr_text);
2162
2163 if (SvLEN (self->incr_text) <= cur + len)
2164 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2165
1797 Move (str, SvEND (self->incr_text), len, char); 2166 Move (str, SvEND (self->incr_text), len, char);
1798 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2167 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1799 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2168 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1800 } 2169 }
1801 } 2170 }
1802 2171
1803 if (GIMME_V != G_VOID) 2172 if (GIMME_V != G_VOID)
1804 do 2173 do
1805 { 2174 {
2175 SV *sv;
1806 STRLEN offset; 2176 STRLEN offset;
1807 2177
1808 if (!INCR_DONE (self)) 2178 if (!INCR_DONE (self))
1809 { 2179 {
1810 incr_parse (self); 2180 incr_parse (self);
1812 if (self->incr_pos > self->max_size && self->max_size) 2182 if (self->incr_pos > self->max_size && self->max_size)
1813 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2183 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1814 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2184 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1815 2185
1816 if (!INCR_DONE (self)) 2186 if (!INCR_DONE (self))
2187 {
2188 // as an optimisation, do not accumulate white space in the incr buffer
2189 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2190 {
2191 self->incr_pos = 0;
2192 SvCUR_set (self->incr_text, 0);
2193 }
2194
1817 break; 2195 break;
2196 }
1818 } 2197 }
1819 2198
1820 XPUSHs (decode_json (self->incr_text, self, &offset)); 2199 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2200 XPUSHs (sv);
1821 2201
1822 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1823 self->incr_pos -= offset; 2202 self->incr_pos -= offset;
1824 self->incr_nest = 0; 2203 self->incr_nest = 0;
1825 self->incr_mode = 0; 2204 self->incr_mode = 0;
2205
2206 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1826 } 2207 }
1827 while (GIMME_V == G_ARRAY); 2208 while (GIMME_V == G_ARRAY);
1828} 2209}
1829 2210
1830SV *incr_text (JSON *self) 2211SV *incr_text (JSON *self)
1868 SvREFCNT_dec (self->incr_text); 2249 SvREFCNT_dec (self->incr_text);
1869 2250
1870PROTOTYPES: ENABLE 2251PROTOTYPES: ENABLE
1871 2252
1872void encode_json (SV *scalar) 2253void encode_json (SV *scalar)
1873 ALIAS:
1874 to_json_ = 0
1875 encode_json = F_UTF8
1876 PPCODE: 2254 PPCODE:
1877{ 2255{
1878 JSON json; 2256 JSON json;
1879 json_init (&json); 2257 json_init (&json);
1880 json.flags |= ix; 2258 json.flags |= F_UTF8;
1881 XPUSHs (encode_json (scalar, &json)); 2259 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2260 XPUSHs (scalar);
1882} 2261}
1883 2262
1884void decode_json (SV *jsonstr) 2263void decode_json (SV *jsonstr)
1885 ALIAS:
1886 from_json_ = 0
1887 decode_json = F_UTF8
1888 PPCODE: 2264 PPCODE:
1889{ 2265{
1890 JSON json; 2266 JSON json;
1891 json_init (&json); 2267 json_init (&json);
1892 json.flags |= ix; 2268 json.flags |= F_UTF8;
1893 XPUSHs (decode_json (jsonstr, &json, 0)); 2269 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2270 XPUSHs (jsonstr);
1894} 2271}
1895 2272
1896

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines