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 1.85 by root, Sat Apr 5 18:15:46 2008 UTC vs.
Revision 1.119 by root, Mon Oct 28 23:19:54 2013 UTC

12#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 14#endif
15 15
16// some old perls do not have this, try to make it work, no 16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 17// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
20#endif 20#endif
21 21
22// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 24
24#define F_ASCII 0x00000001UL 25#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 26#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 27#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 28#define F_INDENT 0x00000008UL
32#define F_SHRINK 0x00000200UL 33#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 34#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 35#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 36#define F_RELAXED 0x00001000UL
36#define F_ALLOW_UNKNOWN 0x00002000UL 37#define F_ALLOW_UNKNOWN 0x00002000UL
38#define F_ALLOW_TAGS 0x00004000UL
37#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 39#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
38 40
39#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 41#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
40 42
41#define INIT_SIZE 32 // initial scalar size to be allocated 43#define INIT_SIZE 32 // initial scalar size to be allocated
42#define INDENT_STEP 3 // spaces per indentation level 44#define INDENT_STEP 3 // spaces per indentation level
43 45
44#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 46#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
47
48#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
45 49
46#define SB do { 50#define SB do {
47#define SE } while (0) 51#define SE } while (0)
48 52
49#if __GNUC__ >= 3 53#if __GNUC__ >= 3
69#else 73#else
70# define JSON_SLOW 0 74# define JSON_SLOW 0
71# define JSON_STASH json_stash 75# define JSON_STASH json_stash
72#endif 76#endif
73 77
78// the amount of HEs to allocate on the stack, when sorting keys
79#define STACK_HES 64
80
74static HV *json_stash, *json_boolean_stash; // JSON::XS:: 81static HV *json_stash, *types_boolean_stash; // JSON::XS::
75static SV *json_true, *json_false; 82static SV *types_true, *types_false, *sv_json;
76 83
77enum { 84enum {
78 INCR_M_WS = 0, // initial whitespace skipping, must be 0 85 INCR_M_WS = 0, // initial whitespace skipping, must be 0
79 INCR_M_STR, // inside string 86 INCR_M_STR, // inside string
80 INCR_M_BS, // inside backslash 87 INCR_M_BS, // inside backslash
88 INCR_M_C0, // inside comment in initial whitespace sequence
89 INCR_M_C1, // inside comment in other places
81 INCR_M_JSON // outside anything, count nesting 90 INCR_M_JSON // outside anything, count nesting
82}; 91};
83 92
84#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 93#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
85 94
86typedef struct { 95typedef struct {
87 U32 flags; 96 U32 flags;
88 U32 max_depth; 97 U32 max_depth;
89 STRLEN max_size; 98 STRLEN max_size;
92 HV *cb_sk_object; 101 HV *cb_sk_object;
93 102
94 // for the incremental parser 103 // for the incremental parser
95 SV *incr_text; // the source text so far 104 SV *incr_text; // the source text so far
96 STRLEN incr_pos; // the current offset into the text 105 STRLEN incr_pos; // the current offset into the text
97 unsigned char incr_nest; // {[]}-nesting level 106 int incr_nest; // {[]}-nesting level
98 unsigned char incr_mode; 107 unsigned char incr_mode;
99} JSON; 108} JSON;
100 109
101INLINE void 110INLINE void
102json_init (JSON *json) 111json_init (JSON *json)
121 130
122INLINE void 131INLINE void
123shrink (SV *sv) 132shrink (SV *sv)
124{ 133{
125 sv_utf8_downgrade (sv, 1); 134 sv_utf8_downgrade (sv, 1);
135
126 if (SvLEN (sv) > SvCUR (sv) + 1) 136 if (SvLEN (sv) > SvCUR (sv) + 1)
127 { 137 {
128#ifdef SvPV_shrink_to_cur 138#ifdef SvPV_shrink_to_cur
129 SvPV_shrink_to_cur (sv); 139 SvPV_shrink_to_cur (sv);
130#elif defined (SvPV_renew) 140#elif defined (SvPV_renew)
174 *s++ = 0x80 | ( ch & 0x3f); 184 *s++ = 0x80 | ( ch & 0x3f);
175 185
176 return s; 186 return s;
177} 187}
178 188
189// convert offset pointer to character index, sv must be string
190static STRLEN
191ptr_to_index (SV *sv, char *offset)
192{
193 return SvUTF8 (sv)
194 ? utf8_distance (offset, SvPVX (sv))
195 : offset - SvPVX (sv);
196}
197
198/////////////////////////////////////////////////////////////////////////////
199// fp hell
200
201// scan a group of digits, and a trailing exponent
202static void
203json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
204{
205 UV uaccum = 0;
206 int eaccum = 0;
207
208 // if we recurse too deep, skip all remaining digits
209 // to avoid a stack overflow attack
210 if (expect_false (--maxdepth <= 0))
211 while (((U8)*s - '0') < 10)
212 ++s;
213
214 for (;;)
215 {
216 U8 dig = (U8)*s - '0';
217
218 if (expect_false (dig >= 10))
219 {
220 if (dig == (U8)((U8)'.' - (U8)'0'))
221 {
222 ++s;
223 json_atof_scan1 (s, accum, expo, 1, maxdepth);
224 }
225 else if ((dig | ' ') == 'e' - '0')
226 {
227 int exp2 = 0;
228 int neg = 0;
229
230 ++s;
231
232 if (*s == '-')
233 {
234 ++s;
235 neg = 1;
236 }
237 else if (*s == '+')
238 ++s;
239
240 while ((dig = (U8)*s - '0') < 10)
241 exp2 = exp2 * 10 + *s++ - '0';
242
243 *expo += neg ? -exp2 : exp2;
244 }
245
246 break;
247 }
248
249 ++s;
250
251 uaccum = uaccum * 10 + dig;
252 ++eaccum;
253
254 // if we have too many digits, then recurse for more
255 // we actually do this for rather few digits
256 if (uaccum >= (UV_MAX - 9) / 10)
257 {
258 if (postdp) *expo -= eaccum;
259 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
260 if (postdp) *expo += eaccum;
261
262 break;
263 }
264 }
265
266 // this relies greatly on the quality of the pow ()
267 // implementation of the platform, but a good
268 // implementation is hard to beat.
269 // (IEEE 754 conformant ones are required to be exact)
270 if (postdp) *expo -= eaccum;
271 *accum += uaccum * Perl_pow (10., *expo);
272 *expo += eaccum;
273}
274
275static NV
276json_atof (const char *s)
277{
278 NV accum = 0.;
279 int expo = 0;
280 int neg = 0;
281
282 if (*s == '-')
283 {
284 ++s;
285 neg = 1;
286 }
287
288 // a recursion depth of ten gives us >>500 bits
289 json_atof_scan1 (s, &accum, &expo, 0, 10);
290
291 return neg ? -accum : accum;
292}
179///////////////////////////////////////////////////////////////////////////// 293/////////////////////////////////////////////////////////////////////////////
180// encoder 294// encoder
181 295
182// structure used for encoding JSON 296// structure used for encoding JSON
183typedef struct 297typedef struct
193INLINE void 307INLINE void
194need (enc_t *enc, STRLEN len) 308need (enc_t *enc, STRLEN len)
195{ 309{
196 if (expect_false (enc->cur + len >= enc->end)) 310 if (expect_false (enc->cur + len >= enc->end))
197 { 311 {
198 STRLEN cur = enc->cur - SvPVX (enc->sv); 312 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
199 SvGROW (enc->sv, cur + len + 1); 313 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
200 enc->cur = SvPVX (enc->sv) + cur; 314 enc->cur = SvPVX (enc->sv) + cur;
201 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 315 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
202 } 316 }
203} 317}
204 318
279 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 393 (int)((uch - 0x10000) % 0x400 + 0xDC00));
280 enc->cur += 12; 394 enc->cur += 12;
281 } 395 }
282 else 396 else
283 { 397 {
284 static char hexdigit [16] = "0123456789abcdef";
285 need (enc, len += 5); 398 need (enc, len += 5);
286 *enc->cur++ = '\\'; 399 *enc->cur++ = '\\';
287 *enc->cur++ = 'u'; 400 *enc->cur++ = 'u';
288 *enc->cur++ = hexdigit [ uch >> 12 ]; 401 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
289 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 402 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
290 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 403 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
291 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 404 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
292 } 405 }
293 406
294 str += clen; 407 str += clen;
295 } 408 }
296 else if (enc->json.flags & F_LATIN1) 409 else if (enc->json.flags & F_LATIN1)
371 484
372 if (enc->indent >= enc->json.max_depth) 485 if (enc->indent >= enc->json.max_depth)
373 croak (ERR_NESTING_EXCEEDED); 486 croak (ERR_NESTING_EXCEEDED);
374 487
375 encode_ch (enc, '['); 488 encode_ch (enc, '[');
376 489
377 if (len >= 0) 490 if (len >= 0)
378 { 491 {
379 encode_nl (enc); ++enc->indent; 492 encode_nl (enc); ++enc->indent;
380 493
381 for (i = 0; i <= len; ++i) 494 for (i = 0; i <= len; ++i)
393 encode_comma (enc); 506 encode_comma (enc);
394 } 507 }
395 508
396 encode_nl (enc); --enc->indent; encode_indent (enc); 509 encode_nl (enc); --enc->indent; encode_indent (enc);
397 } 510 }
398 511
399 encode_ch (enc, ']'); 512 encode_ch (enc, ']');
400} 513}
401 514
402static void 515static void
403encode_hk (enc_t *enc, HE *he) 516encode_hk (enc_t *enc, HE *he)
407 if (HeKLEN (he) == HEf_SVKEY) 520 if (HeKLEN (he) == HEf_SVKEY)
408 { 521 {
409 SV *sv = HeSVKEY (he); 522 SV *sv = HeSVKEY (he);
410 STRLEN len; 523 STRLEN len;
411 char *str; 524 char *str;
412 525
413 SvGETMAGIC (sv); 526 SvGETMAGIC (sv);
414 str = SvPV (sv, len); 527 str = SvPV (sv, len);
415 528
416 encode_str (enc, str, len, SvUTF8 (sv)); 529 encode_str (enc, str, len, SvUTF8 (sv));
417 } 530 }
460 573
461 encode_ch (enc, '{'); 574 encode_ch (enc, '{');
462 575
463 // for canonical output we have to sort by keys first 576 // for canonical output we have to sort by keys first
464 // actually, this is mostly due to the stupid so-called 577 // actually, this is mostly due to the stupid so-called
465 // security workaround added somewhere in 5.8.x. 578 // security workaround added somewhere in 5.8.x
466 // that randomises hash orderings 579 // that randomises hash orderings
467 if (enc->json.flags & F_CANONICAL) 580 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
468 { 581 {
469 int count = hv_iterinit (hv); 582 int count = hv_iterinit (hv);
470 583
471 if (SvMAGICAL (hv)) 584 if (SvMAGICAL (hv))
472 { 585 {
483 } 596 }
484 597
485 if (count) 598 if (count)
486 { 599 {
487 int i, fast = 1; 600 int i, fast = 1;
488#if defined(__BORLANDC__) || defined(_MSC_VER) 601 HE *hes_stack [STACK_HES];
489 HE **hes = _alloca (count * sizeof (HE)); 602 HE **hes = hes_stack;
490#else 603
491 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 604 // allocate larger arrays on the heap
492#endif 605 if (count > STACK_HES)
606 {
607 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
608 hes = (HE **)SvPVX (sv);
609 }
493 610
494 i = 0; 611 i = 0;
495 while ((he = hv_iternext (hv))) 612 while ((he = hv_iternext (hv)))
496 { 613 {
497 hes [i++] = he; 614 hes [i++] = he;
572 SvGETMAGIC (sv); 689 SvGETMAGIC (sv);
573 svt = SvTYPE (sv); 690 svt = SvTYPE (sv);
574 691
575 if (expect_false (SvOBJECT (sv))) 692 if (expect_false (SvOBJECT (sv)))
576 { 693 {
577 HV *stash = !JSON_SLOW || json_boolean_stash 694 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
578 ? json_boolean_stash 695 ? types_boolean_stash
579 : gv_stashpv ("JSON::XS::Boolean", 1); 696 : gv_stashpv ("Types::Serialiser::Boolean", 1);
697 HV *stash = SvSTASH (sv);
580 698
581 if (SvSTASH (sv) == stash) 699 if (stash == boolean_stash)
582 { 700 {
583 if (SvIV (sv)) 701 if (SvIV (sv))
584 encode_str (enc, "true", 4, 0); 702 encode_str (enc, "true", 4, 0);
585 else 703 else
586 encode_str (enc, "false", 5, 0); 704 encode_str (enc, "false", 5, 0);
587 } 705 }
588 else 706 else if (enc->json.flags & F_CONV_BLESSED)
589 { 707 {
590#if 0 708 GV *to_json = gv_fetchmethod_autoload (stash, "TO_JSON", 0);
591 if (0 && sv_derived_from (rv, "JSON::Literal")) 709
592 { 710 if (to_json)
593 // not yet
594 } 711 {
595#endif
596 if (enc->json.flags & F_CONV_BLESSED)
597 { 712 dSP;
713
714 ENTER; SAVETMPS; PUSHMARK (SP);
598 // we re-bless the reference to get overload and other niceties right 715 // we re-bless the reference to get overload and other niceties right
599 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
600
601 if (to_json)
602 {
603 dSP;
604
605 ENTER; SAVETMPS; PUSHMARK (SP);
606 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); 716 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
607 717
608 // calling with G_SCALAR ensures that we always get a 1 return value 718 // calling with G_SCALAR ensures that we always get a 1 return value
609 PUTBACK; 719 PUTBACK;
610 call_sv ((SV *)GvCV (to_json), G_SCALAR); 720 call_sv ((SV *)GvCV (to_json), G_SCALAR);
611 SPAGAIN; 721 SPAGAIN;
612 722
613 // catch this surprisingly common error 723 // catch this surprisingly common error
614 if (SvROK (TOPs) && SvRV (TOPs) == sv) 724 if (SvROK (TOPs) && SvRV (TOPs) == sv)
615 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); 725 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
616 726
617 sv = POPs; 727 sv = POPs;
618 PUTBACK; 728 PUTBACK;
619 729
620 encode_sv (enc, sv); 730 encode_sv (enc, sv);
621 731
622 FREETMPS; LEAVE; 732 FREETMPS; LEAVE;
623 }
624 else if (enc->json.flags & F_ALLOW_BLESSED)
625 encode_str (enc, "null", 4, 0);
626 else
627 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
628 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
629 } 733 }
630 else if (enc->json.flags & F_ALLOW_BLESSED) 734 else if (enc->json.flags & F_ALLOW_BLESSED)
631 encode_str (enc, "null", 4, 0); 735 encode_str (enc, "null", 4, 0);
632 else 736 else
633 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 737 croak ("encountered object '%s' when conv_object is enabled, but TO_JSON is missing and allow_blessed disabled",
634 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 738 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
635 } 739 }
740 else if (enc->json.flags & F_ALLOW_TAGS)
741 {
742 GV *method = gv_fetchmethod_autoload (stash, "FREEZE", 0);
743
744 if (method)
745 {
746 int count;
747 dSP;
748
749 ENTER; SAVETMPS; PUSHMARK (SP);
750 EXTEND (SP, 2);
751 // we re-bless the reference to get overload and other niceties right
752 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
753 PUSHs (sv_json);
754
755 PUTBACK;
756 count = call_sv ((SV *)GvCV (method), G_ARRAY);
757 SPAGAIN;
758
759 // catch this surprisingly common error
760 if (SvROK (TOPs) && SvRV (TOPs) == sv)
761 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
762
763 encode_ch (enc, '(');
764 encode_ch (enc, '"');
765 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
766 encode_ch (enc, '"');
767 encode_ch (enc, ')');
768 encode_ch (enc, '[');
769
770 while (count)
771 {
772 encode_sv (enc, SP[1 - count--]);
773
774 if (count)
775 encode_ch (enc, ',');
776 }
777
778 encode_ch (enc, ']');
779
780 PUTBACK;
781
782 FREETMPS; LEAVE;
783 }
784 else if (enc->json.flags & F_ALLOW_BLESSED)
785 encode_str (enc, "null", 4, 0);
786 else
787 croak ("encountered object '%s' when allow_tags is enabled, but FREEZE is missing and allow_blessed disabled",
788 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
789 }
790 else if (enc->json.flags & F_ALLOW_BLESSED)
791 encode_str (enc, "null", 4, 0);
792 else
793 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled",
794 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
636 } 795 }
637 else if (svt == SVt_PVHV) 796 else if (svt == SVt_PVHV)
638 encode_hv (enc, (HV *)sv); 797 encode_hv (enc, (HV *)sv);
639 else if (svt == SVt_PVAV) 798 else if (svt == SVt_PVAV)
640 encode_av (enc, (AV *)sv); 799 encode_av (enc, (AV *)sv);
716 } 875 }
717 else 876 else
718 { 877 {
719 // large integer, use the (rather slow) snprintf way. 878 // large integer, use the (rather slow) snprintf way.
720 need (enc, IVUV_MAXCHARS); 879 need (enc, IVUV_MAXCHARS);
721 enc->cur += 880 enc->cur +=
722 SvIsUV(sv) 881 SvIsUV(sv)
723 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 882 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
724 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 883 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
725 } 884 }
726 } 885 }
727 else if (SvROK (sv)) 886 else if (SvROK (sv))
728 encode_rv (enc, SvRV (sv)); 887 encode_rv (enc, SvRV (sv));
729 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 888 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
730 encode_str (enc, "null", 4, 0); 889 encode_str (enc, "null", 4, 0);
731 else 890 else
732 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 891 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
733 SvPV_nolen (sv), SvFLAGS (sv)); 892 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
734} 893}
735 894
736static SV * 895static SV *
737encode_json (SV *scalar, JSON *json) 896encode_json (SV *scalar, JSON *json)
738{ 897{
750 : enc.json.flags & F_LATIN1 ? 0x000100UL 909 : enc.json.flags & F_LATIN1 ? 0x000100UL
751 : 0x110000UL; 910 : 0x110000UL;
752 911
753 SvPOK_only (enc.sv); 912 SvPOK_only (enc.sv);
754 encode_sv (&enc, scalar); 913 encode_sv (&enc, scalar);
914 encode_nl (&enc);
755 915
756 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 916 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
757 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 917 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
758 918
759 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 919 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
939 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1099 else if (expect_true (ch >= 0x20 && ch < 0x80))
940 *cur++ = ch; 1100 *cur++ = ch;
941 else if (ch >= 0x80) 1101 else if (ch >= 0x80)
942 { 1102 {
943 STRLEN clen; 1103 STRLEN clen;
944 UV uch;
945 1104
946 --dec_cur; 1105 --dec_cur;
947 1106
948 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1107 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
949 if (clen == (STRLEN)-1) 1108 if (clen == (STRLEN)-1)
950 ERR ("malformed UTF-8 character in JSON string"); 1109 ERR ("malformed UTF-8 character in JSON string");
951 1110
952 do 1111 do
953 *cur++ = *dec_cur++; 1112 *cur++ = *dec_cur++;
970 { 1129 {
971 STRLEN len = cur - buf; 1130 STRLEN len = cur - buf;
972 1131
973 if (sv) 1132 if (sv)
974 { 1133 {
975 SvGROW (sv, SvCUR (sv) + len + 1); 1134 STRLEN cur = SvCUR (sv);
1135
1136 if (SvLEN (sv) <= cur + len)
1137 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1138
976 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1139 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
977 SvCUR_set (sv, SvCUR (sv) + len); 1140 SvCUR_set (sv, SvCUR (sv) + len);
978 } 1141 }
979 else 1142 else
980 sv = newSVpvn (buf, len); 1143 sv = newSVpvn (buf, len);
1071 1234
1072 // special case the rather common 1..5-digit-int case 1235 // special case the rather common 1..5-digit-int case
1073 if (*start == '-') 1236 if (*start == '-')
1074 switch (len) 1237 switch (len)
1075 { 1238 {
1076 case 2: return newSViv (-( start [1] - '0' * 1)); 1239 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1077 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1240 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1078 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1241 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1079 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1242 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1080 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1243 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1081 } 1244 }
1082 else 1245 else
1083 switch (len) 1246 switch (len)
1084 { 1247 {
1085 case 1: return newSViv ( start [0] - '0' * 1); 1248 case 1: return newSViv ( start [0] - '0' * 1);
1086 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1249 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1087 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1250 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1088 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1251 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1089 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1252 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1090 } 1253 }
1091 1254
1092 { 1255 {
1093 UV uv; 1256 UV uv;
1094 int numtype = grok_number (start, len, &uv); 1257 int numtype = grok_number (start, len, &uv);
1103 } 1266 }
1104 1267
1105 len -= *start == '-' ? 1 : 0; 1268 len -= *start == '-' ? 1 : 0;
1106 1269
1107 // does not fit into IV or UV, try NV 1270 // does not fit into IV or UV, try NV
1108 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1271 if (len <= NV_DIG)
1109 #if defined (LDBL_DIG)
1110 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1111 #endif
1112 )
1113 // fits into NV without loss of precision 1272 // fits into NV without loss of precision
1114 return newSVnv (Atof (start)); 1273 return newSVnv (json_atof (start));
1115 1274
1116 // everything else fails, convert it to a string 1275 // everything else fails, convert it to a string
1117 return newSVpvn (start, dec->cur - start); 1276 return newSVpvn (start, dec->cur - start);
1118 } 1277 }
1119 1278
1120 // loss of precision here 1279 // loss of precision here
1121 return newSVnv (Atof (start)); 1280 return newSVnv (json_atof (start));
1122 1281
1123fail: 1282fail:
1124 return 0; 1283 return 0;
1125} 1284}
1126 1285
1150 if (*dec->cur == ']') 1309 if (*dec->cur == ']')
1151 { 1310 {
1152 ++dec->cur; 1311 ++dec->cur;
1153 break; 1312 break;
1154 } 1313 }
1155 1314
1156 if (*dec->cur != ',') 1315 if (*dec->cur != ',')
1157 ERR (", or ] expected while parsing array"); 1316 ERR (", or ] expected while parsing array");
1158 1317
1159 ++dec->cur; 1318 ++dec->cur;
1160 1319
1295 dSP; 1454 dSP;
1296 int count; 1455 int count;
1297 1456
1298 ENTER; SAVETMPS; PUSHMARK (SP); 1457 ENTER; SAVETMPS; PUSHMARK (SP);
1299 XPUSHs (HeVAL (he)); 1458 XPUSHs (HeVAL (he));
1459 sv_2mortal (sv);
1300 1460
1301 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1461 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1302 1462
1303 if (count == 1) 1463 if (count == 1)
1304 { 1464 {
1305 sv = newSVsv (POPs); 1465 sv = newSVsv (POPs);
1306 FREETMPS; LEAVE; 1466 FREETMPS; LEAVE;
1307 return sv; 1467 return sv;
1308 } 1468 }
1309 1469
1470 SvREFCNT_inc (sv);
1310 FREETMPS; LEAVE; 1471 FREETMPS; LEAVE;
1311 } 1472 }
1312 } 1473 }
1313 1474
1314 if (dec->json.cb_object) 1475 if (dec->json.cb_object)
1340 DEC_DEC_DEPTH; 1501 DEC_DEC_DEPTH;
1341 return 0; 1502 return 0;
1342} 1503}
1343 1504
1344static SV * 1505static SV *
1506decode_tag (dec_t *dec)
1507{
1508 SV *tag = 0;
1509 SV *val = 0;
1510
1511 if (!(dec->json.flags & F_ALLOW_TAGS))
1512 ERR ("malformed JSON string, neither array, object, number, string or atom");
1513
1514 ++dec->cur;
1515
1516 tag = decode_sv (dec);
1517 if (!tag)
1518 goto fail;
1519
1520 if (!SvPOK (tag))
1521 ERR ("malformed JSON string, (tag) must be a string");
1522
1523 if (*dec->cur != ')')
1524 ERR (") expected after tag");
1525
1526 ++dec->cur;
1527
1528 val = decode_sv (dec);
1529 if (!val)
1530 goto fail;
1531
1532 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1533 ERR ("malformed JSON string, tag value must be an array");
1534
1535 {
1536 AV *av = (AV *)SvRV (val);
1537 int i, len = av_len (av) + 1;
1538 HV *stash = gv_stashsv (tag, 0);
1539 SV *sv;
1540
1541 if (!stash)
1542 ERR ("cannot decode perl-object (package does not exist)");
1543
1544 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1545
1546 if (!method)
1547 ERR ("cannot decode perl-object (package does not have a THAW method)");
1548
1549 dSP;
1550
1551 ENTER; SAVETMPS; PUSHMARK (SP);
1552 EXTEND (SP, len + 2);
1553 // we re-bless the reference to get overload and other niceties right
1554 PUSHs (tag);
1555 PUSHs (sv_json);
1556
1557 for (i = 0; i < len; ++i)
1558 PUSHs (*av_fetch (av, i, 1));
1559
1560 PUTBACK;
1561 call_sv ((SV *)GvCV (method), G_SCALAR);
1562 SPAGAIN;
1563
1564 SvREFCNT_dec (tag);
1565 SvREFCNT_dec (val);
1566 sv = SvREFCNT_inc (POPs);
1567
1568 PUTBACK;
1569
1570 FREETMPS; LEAVE;
1571
1572 return sv;
1573 }
1574
1575fail:
1576 SvREFCNT_dec (tag);
1577 SvREFCNT_dec (val);
1578 return 0;
1579}
1580
1581static SV *
1345decode_sv (dec_t *dec) 1582decode_sv (dec_t *dec)
1346{ 1583{
1347 // the beauty of JSON: you need exactly one character lookahead 1584 // the beauty of JSON: you need exactly one character lookahead
1348 // to parse everything. 1585 // to parse everything.
1349 switch (*dec->cur) 1586 switch (*dec->cur)
1350 { 1587 {
1351 case '"': ++dec->cur; return decode_str (dec); 1588 case '"': ++dec->cur; return decode_str (dec);
1352 case '[': ++dec->cur; return decode_av (dec); 1589 case '[': ++dec->cur; return decode_av (dec);
1353 case '{': ++dec->cur; return decode_hv (dec); 1590 case '{': ++dec->cur; return decode_hv (dec);
1591 case '(': return decode_tag (dec);
1354 1592
1355 case '-': 1593 case '-':
1356 case '0': case '1': case '2': case '3': case '4': 1594 case '0': case '1': case '2': case '3': case '4':
1357 case '5': case '6': case '7': case '8': case '9': 1595 case '5': case '6': case '7': case '8': case '9':
1358 return decode_num (dec); 1596 return decode_num (dec);
1360 case 't': 1598 case 't':
1361 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1599 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1362 { 1600 {
1363 dec->cur += 4; 1601 dec->cur += 4;
1364#if JSON_SLOW 1602#if JSON_SLOW
1365 json_true = get_bool ("JSON::XS::true"); 1603 types_true = get_bool ("Types::Serialiser::true");
1366#endif 1604#endif
1367 return newSVsv (json_true); 1605 return newSVsv (types_true);
1368 } 1606 }
1369 else 1607 else
1370 ERR ("'true' expected"); 1608 ERR ("'true' expected");
1371 1609
1372 break; 1610 break;
1374 case 'f': 1612 case 'f':
1375 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1613 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1376 { 1614 {
1377 dec->cur += 5; 1615 dec->cur += 5;
1378#if JSON_SLOW 1616#if JSON_SLOW
1379 json_false = get_bool ("JSON::XS::false"); 1617 types_false = get_bool ("Types::Serialiser::false");
1380#endif 1618#endif
1381 return newSVsv (json_false); 1619 return newSVsv (types_false);
1382 } 1620 }
1383 else 1621 else
1384 ERR ("'false' expected"); 1622 ERR ("'false' expected");
1385 1623
1386 break; 1624 break;
1395 ERR ("'null' expected"); 1633 ERR ("'null' expected");
1396 1634
1397 break; 1635 break;
1398 1636
1399 default: 1637 default:
1400 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1638 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1401 break; 1639 break;
1402 } 1640 }
1403 1641
1404fail: 1642fail:
1405 return 0; 1643 return 0;
1406} 1644}
1407 1645
1408static SV * 1646static SV *
1409decode_json (SV *string, JSON *json, STRLEN *offset_return) 1647decode_json (SV *string, JSON *json, char **offset_return)
1410{ 1648{
1411 dec_t dec; 1649 dec_t dec;
1412 STRLEN offset;
1413 SV *sv; 1650 SV *sv;
1414 1651
1652 /* work around bugs in 5.10 where manipulating magic values
1653 * makes perl ignore the magic in subsequent accesses.
1654 * also make a copy of non-PV values, to get them into a clean
1655 * state (SvPV should do that, but it's buggy, see below).
1656 */
1415 SvGETMAGIC (string); 1657 /*SvGETMAGIC (string);*/
1658 if (SvMAGICAL (string) || !SvPOK (string))
1659 string = sv_2mortal (newSVsv (string));
1660
1416 SvUPGRADE (string, SVt_PV); 1661 SvUPGRADE (string, SVt_PV);
1417 1662
1663 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1664 * assertion with -DDEBUGGING, although SvCUR is documented to
1665 * return the xpv_cur field which certainly exists after upgrading.
1666 * according to nicholas clark, calling SvPOK fixes this.
1667 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1668 * and hope for the best.
1669 * Damnit, SvPV_nolen still trips over yet another assertion. This
1670 * assertion business is seriously broken, try yet another workaround
1671 * for the broken -DDEBUGGING.
1672 */
1673 {
1674#ifdef DEBUGGING
1675 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1676#else
1677 STRLEN offset = SvCUR (string);
1678#endif
1679
1418 if (SvCUR (string) > json->max_size && json->max_size) 1680 if (offset > json->max_size && json->max_size)
1419 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1681 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1420 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1682 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1683 }
1421 1684
1422 if (json->flags & F_UTF8) 1685 if (DECODE_WANTS_OCTETS (json))
1423 sv_utf8_downgrade (string, 0); 1686 sv_utf8_downgrade (string, 0);
1424 else 1687 else
1425 sv_utf8_upgrade (string); 1688 sv_utf8_upgrade (string);
1426 1689
1427 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1690 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1437 1700
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1701 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439 1702
1440 decode_ws (&dec); 1703 decode_ws (&dec);
1441 sv = decode_sv (&dec); 1704 sv = decode_sv (&dec);
1705
1706 if (offset_return)
1707 *offset_return = dec.cur;
1442 1708
1443 if (!(offset_return || !sv)) 1709 if (!(offset_return || !sv))
1444 { 1710 {
1445 // check for trailing garbage 1711 // check for trailing garbage
1446 decode_ws (&dec); 1712 decode_ws (&dec);
1449 { 1715 {
1450 dec.err = "garbage after JSON object"; 1716 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv); 1717 SvREFCNT_dec (sv);
1452 sv = 0; 1718 sv = 0;
1453 } 1719 }
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 } 1720 }
1465 1721
1466 if (!sv) 1722 if (!sv)
1467 { 1723 {
1468 SV *uni = sv_newmortal (); 1724 SV *uni = sv_newmortal ();
1474 SAVEVPTR (PL_curcop); 1730 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop; 1731 PL_curcop = &cop;
1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1732 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE; 1733 LEAVE;
1478 1734
1479 croak ("%s, at character offset %d [\"%s\"]", 1735 croak ("%s, at character offset %d (before \"%s\")",
1480 dec.err, 1736 dec.err,
1481 (int)offset, 1737 (int)ptr_to_index (string, dec.cur),
1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1738 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1483 } 1739 }
1484 1740
1485 sv = sv_2mortal (sv); 1741 sv = sv_2mortal (sv);
1486 1742
1495 1751
1496static void 1752static void
1497incr_parse (JSON *self) 1753incr_parse (JSON *self)
1498{ 1754{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1755 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1756
1757 // the state machine here is a bit convoluted and could be simplified a lot
1758 // but this would make it slower, so...
1500 1759
1501 for (;;) 1760 for (;;)
1502 { 1761 {
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 1762 //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) 1763 switch (self->incr_mode)
1505 { 1764 {
1506 // only used for intiial whitespace skipping 1765 // only used for initial whitespace skipping
1507 case INCR_M_WS: 1766 case INCR_M_WS:
1508 for (;;) 1767 for (;;)
1509 { 1768 {
1510 if (*p > 0x20) 1769 if (*p > 0x20)
1511 { 1770 {
1771 if (*p == '#')
1772 {
1773 self->incr_mode = INCR_M_C0;
1774 goto incr_m_c;
1775 }
1776 else
1777 {
1512 self->incr_mode = INCR_M_JSON; 1778 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json; 1779 goto incr_m_json;
1780 }
1514 } 1781 }
1515 else if (!*p) 1782 else if (!*p)
1516 goto interrupt; 1783 goto interrupt;
1517 1784
1518 ++p; 1785 ++p;
1524 goto interrupt; 1791 goto interrupt;
1525 1792
1526 ++p; 1793 ++p;
1527 self->incr_mode = INCR_M_STR; 1794 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str; 1795 goto incr_m_str;
1796
1797 // inside #-style comments
1798 case INCR_M_C0:
1799 case INCR_M_C1:
1800 incr_m_c:
1801 for (;;)
1802 {
1803 if (*p == '\n')
1804 {
1805 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1806 break;
1807 }
1808 else if (!*p)
1809 goto interrupt;
1810
1811 ++p;
1812 }
1813
1814 break;
1529 1815
1530 // inside a string 1816 // inside a string
1531 case INCR_M_STR: 1817 case INCR_M_STR:
1532 incr_m_str: 1818 incr_m_str:
1533 for (;;) 1819 for (;;)
1584 self->incr_mode = INCR_M_STR; 1870 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str; 1871 goto incr_m_str;
1586 1872
1587 case '[': 1873 case '[':
1588 case '{': 1874 case '{':
1875 case '(':
1589 if (++self->incr_nest > self->max_depth) 1876 if (++self->incr_nest > self->max_depth)
1590 croak (ERR_NESTING_EXCEEDED); 1877 croak (ERR_NESTING_EXCEEDED);
1591 break; 1878 break;
1592 1879
1593 case ']': 1880 case ']':
1594 case '}': 1881 case '}':
1595 if (!--self->incr_nest) 1882 if (--self->incr_nest <= 0)
1596 goto interrupt; 1883 goto interrupt;
1884 break;
1885
1886 case ')':
1887 --self->incr_nest;
1888 break;
1889
1890 case '#':
1891 self->incr_mode = INCR_M_C1;
1892 goto incr_m_c;
1597 } 1893 }
1598 } 1894 }
1599 } 1895 }
1600 1896
1601 modechange: 1897 modechange:
1602 ; 1898 ;
1603 } 1899 }
1604 1900
1605interrupt: 1901interrupt:
1606 self->incr_pos = p - SvPVX (self->incr_text); 1902 self->incr_pos = p - SvPVX (self->incr_text);
1903 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1607 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1904 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1608} 1905}
1609 1906
1610///////////////////////////////////////////////////////////////////////////// 1907/////////////////////////////////////////////////////////////////////////////
1611// XS interface functions 1908// XS interface functions
1621 i >= '0' && i <= '9' ? i - '0' 1918 i >= '0' && i <= '9' ? i - '0'
1622 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1919 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1623 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1920 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1624 : -1; 1921 : -1;
1625 1922
1626 json_stash = gv_stashpv ("JSON::XS" , 1); 1923 json_stash = gv_stashpv ("JSON::XS" , 1);
1627 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1924 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1628 1925
1629 json_true = get_bool ("JSON::XS::true"); 1926 types_true = get_bool ("Types::Serialiser::true");
1630 json_false = get_bool ("JSON::XS::false"); 1927 types_false = get_bool ("Types::Serialiser::false");
1928
1929 sv_json = newSVpv ("JSON", 0);
1930 SvREADONLY_on (sv_json);
1931
1932 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1631} 1933}
1632 1934
1633PROTOTYPES: DISABLE 1935PROTOTYPES: DISABLE
1634 1936
1635void CLONE (...) 1937void CLONE (...)
1636 CODE: 1938 CODE:
1637 json_stash = 0; 1939 json_stash = 0;
1638 json_boolean_stash = 0; 1940 types_boolean_stash = 0;
1639 1941
1640void new (char *klass) 1942void new (char *klass)
1641 PPCODE: 1943 PPCODE:
1642{ 1944{
1643 SV *pv = NEWSV (0, sizeof (JSON)); 1945 SV *pv = NEWSV (0, sizeof (JSON));
1644 SvPOK_only (pv); 1946 SvPOK_only (pv);
1645 json_init ((JSON *)SvPVX (pv)); 1947 json_init ((JSON *)SvPVX (pv));
1646 XPUSHs (sv_2mortal (sv_bless ( 1948 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv), 1949 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1950 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1663 shrink = F_SHRINK 1965 shrink = F_SHRINK
1664 allow_blessed = F_ALLOW_BLESSED 1966 allow_blessed = F_ALLOW_BLESSED
1665 convert_blessed = F_CONV_BLESSED 1967 convert_blessed = F_CONV_BLESSED
1666 relaxed = F_RELAXED 1968 relaxed = F_RELAXED
1667 allow_unknown = F_ALLOW_UNKNOWN 1969 allow_unknown = F_ALLOW_UNKNOWN
1970 allow_tags = F_ALLOW_TAGS
1668 PPCODE: 1971 PPCODE:
1669{ 1972{
1670 if (enable) 1973 if (enable)
1671 self->flags |= ix; 1974 self->flags |= ix;
1672 else 1975 else
1688 get_shrink = F_SHRINK 1991 get_shrink = F_SHRINK
1689 get_allow_blessed = F_ALLOW_BLESSED 1992 get_allow_blessed = F_ALLOW_BLESSED
1690 get_convert_blessed = F_CONV_BLESSED 1993 get_convert_blessed = F_CONV_BLESSED
1691 get_relaxed = F_RELAXED 1994 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN 1995 get_allow_unknown = F_ALLOW_UNKNOWN
1996 get_allow_tags = F_ALLOW_TAGS
1693 PPCODE: 1997 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix)); 1998 XPUSHs (boolSV (self->flags & ix));
1695 1999
1696void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2000void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1697 PPCODE: 2001 PPCODE:
1725} 2029}
1726 2030
1727void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2031void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1728 PPCODE: 2032 PPCODE:
1729{ 2033{
1730 if (!self->cb_sk_object) 2034 if (!self->cb_sk_object)
1731 self->cb_sk_object = newHV (); 2035 self->cb_sk_object = newHV ();
1732 2036
1733 if (SvOK (cb)) 2037 if (SvOK (cb))
1734 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2038 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1735 else 2039 else
1746 XPUSHs (ST (0)); 2050 XPUSHs (ST (0));
1747} 2051}
1748 2052
1749void encode (JSON *self, SV *scalar) 2053void encode (JSON *self, SV *scalar)
1750 PPCODE: 2054 PPCODE:
1751 XPUSHs (encode_json (scalar, self)); 2055 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2056 XPUSHs (scalar);
1752 2057
1753void decode (JSON *self, SV *jsonstr) 2058void decode (JSON *self, SV *jsonstr)
1754 PPCODE: 2059 PPCODE:
1755 XPUSHs (decode_json (jsonstr, self, 0)); 2060 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2061 XPUSHs (jsonstr);
1756 2062
1757void decode_prefix (JSON *self, SV *jsonstr) 2063void decode_prefix (JSON *self, SV *jsonstr)
1758 PPCODE: 2064 PPCODE:
1759{ 2065{
1760 STRLEN offset; 2066 SV *sv;
2067 char *offset;
2068 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1761 EXTEND (SP, 2); 2069 EXTEND (SP, 2);
1762 PUSHs (decode_json (jsonstr, self, &offset)); 2070 PUSHs (sv);
1763 PUSHs (sv_2mortal (newSVuv (offset))); 2071 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1764} 2072}
1765 2073
1766void incr_parse (JSON *self, SV *jsonstr = 0) 2074void incr_parse (JSON *self, SV *jsonstr = 0)
1767 PPCODE: 2075 PPCODE:
1768{ 2076{
1769 if (!self->incr_text) 2077 if (!self->incr_text)
1770 self->incr_text = newSVpvn ("", 0); 2078 self->incr_text = newSVpvn ("", 0);
2079
2080 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2081 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2082 if (DECODE_WANTS_OCTETS (self))
2083 {
2084 if (self->incr_pos)
2085 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2086 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2087
2088 sv_utf8_downgrade (self->incr_text, 0);
2089 }
2090 else
2091 {
2092 sv_utf8_upgrade (self->incr_text);
2093
2094 if (self->incr_pos)
2095 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2096 - (U8 *)SvPVX (self->incr_text);
2097 }
1771 2098
1772 // append data, if any 2099 // append data, if any
1773 if (jsonstr) 2100 if (jsonstr)
1774 { 2101 {
2102 /* make sure both strings have same encoding */
1775 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2103 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2104 if (SvUTF8 (jsonstr))
2105 sv_utf8_downgrade (jsonstr, 0);
1776 { 2106 else
1777 /* utf-8-ness differs, need to upgrade */
1778 sv_utf8_upgrade (self->incr_text); 2107 sv_utf8_upgrade (jsonstr);
1779 2108
1780 if (self->incr_pos) 2109 /* and then just blindly append */
1781 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1782 - (U8 *)SvPVX (self->incr_text);
1783 }
1784
1785 { 2110 {
1786 STRLEN len; 2111 STRLEN len;
1787 const char *str = SvPV (jsonstr, len); 2112 const char *str = SvPV (jsonstr, len);
1788 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2113 STRLEN cur = SvCUR (self->incr_text);
2114
2115 if (SvLEN (self->incr_text) <= cur + len)
2116 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2117
1789 Move (str, SvEND (self->incr_text), len, char); 2118 Move (str, SvEND (self->incr_text), len, char);
1790 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2119 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1791 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2120 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1792 } 2121 }
1793 } 2122 }
1794 2123
1795 if (GIMME_V != G_VOID) 2124 if (GIMME_V != G_VOID)
1796 do 2125 do
1797 { 2126 {
2127 SV *sv;
1798 STRLEN offset; 2128 char *offset;
1799 2129
1800 if (!INCR_DONE (self)) 2130 if (!INCR_DONE (self))
1801 { 2131 {
1802 incr_parse (self); 2132 incr_parse (self);
1803 2133
1804 if (self->incr_pos > self->max_size && self->max_size) 2134 if (self->incr_pos > self->max_size && self->max_size)
1805 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2135 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1806 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2136 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1807 2137
1808 if (!INCR_DONE (self)) 2138 if (!INCR_DONE (self))
2139 {
2140 // as an optimisation, do not accumulate white space in the incr buffer
2141 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2142 {
2143 self->incr_pos = 0;
2144 SvCUR_set (self->incr_text, 0);
2145 }
2146
1809 break; 2147 break;
2148 }
1810 } 2149 }
1811 2150
1812 XPUSHs (decode_json (self->incr_text, self, &offset)); 2151 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2152 XPUSHs (sv);
1813 2153
1814 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1815 self->incr_pos -= offset; 2154 self->incr_pos -= offset - SvPVX (self->incr_text);
1816 self->incr_nest = 0; 2155 self->incr_nest = 0;
1817 self->incr_mode = 0; 2156 self->incr_mode = 0;
2157
2158 sv_chop (self->incr_text, offset);
1818 } 2159 }
1819 while (GIMME_V == G_ARRAY); 2160 while (GIMME_V == G_ARRAY);
1820} 2161}
1821 2162
1822SV *incr_text (JSON *self) 2163SV *incr_text (JSON *self)
1841 self->incr_nest = 0; 2182 self->incr_nest = 0;
1842 self->incr_mode = 0; 2183 self->incr_mode = 0;
1843 } 2184 }
1844} 2185}
1845 2186
2187void incr_reset (JSON *self)
2188 CODE:
2189{
2190 SvREFCNT_dec (self->incr_text);
2191 self->incr_text = 0;
2192 self->incr_pos = 0;
2193 self->incr_nest = 0;
2194 self->incr_mode = 0;
2195}
2196
1846void DESTROY (JSON *self) 2197void DESTROY (JSON *self)
1847 CODE: 2198 CODE:
1848 SvREFCNT_dec (self->cb_sk_object); 2199 SvREFCNT_dec (self->cb_sk_object);
1849 SvREFCNT_dec (self->cb_object); 2200 SvREFCNT_dec (self->cb_object);
1850 SvREFCNT_dec (self->incr_text); 2201 SvREFCNT_dec (self->incr_text);
1851 2202
1852PROTOTYPES: ENABLE 2203PROTOTYPES: ENABLE
1853 2204
1854void encode_json (SV *scalar) 2205void encode_json (SV *scalar)
1855 ALIAS:
1856 to_json_ = 0
1857 encode_json = F_UTF8
1858 PPCODE: 2206 PPCODE:
1859{ 2207{
1860 JSON json; 2208 JSON json;
1861 json_init (&json); 2209 json_init (&json);
1862 json.flags |= ix; 2210 json.flags |= F_UTF8;
1863 XPUSHs (encode_json (scalar, &json)); 2211 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2212 XPUSHs (scalar);
1864} 2213}
1865 2214
1866void decode_json (SV *jsonstr) 2215void decode_json (SV *jsonstr)
1867 ALIAS:
1868 from_json_ = 0
1869 decode_json = F_UTF8
1870 PPCODE: 2216 PPCODE:
1871{ 2217{
1872 JSON json; 2218 JSON json;
1873 json_init (&json); 2219 json_init (&json);
1874 json.flags |= ix; 2220 json.flags |= F_UTF8;
1875 XPUSHs (decode_json (jsonstr, &json, 0)); 2221 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2222 XPUSHs (jsonstr);
1876} 2223}
1877 2224
1878

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines