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 1.100 by root, Sat Sep 5 23:00:56 2009 UTC vs.
Revision 1.127 by root, Fri Feb 26 21:46:45 2016 UTC

17// guarantees, 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 <= 0 && (json)->incr_mode == INCR_M_JSON) 106#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
85 107
182ptr_to_index (SV *sv, char *offset) 204ptr_to_index (SV *sv, char *offset)
183{ 205{
184 return SvUTF8 (sv) 206 return SvUTF8 (sv)
185 ? utf8_distance (offset, SvPVX (sv)) 207 ? utf8_distance (offset, SvPVX (sv))
186 : 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
326 return -1;
327}
328
329// returns whether scalar is not a reference in the sense of allow_nonref
330static int
331json_nonref (SV *scalar)
332{
333 if (!SvROK (scalar))
334 return 1;
335
336 scalar = SvRV (scalar);
337
338 if (SvSTASH (scalar) == bool_stash)
339 return 1;
340
341 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
342 return 1;
343
344 return 0;
187} 345}
188 346
189///////////////////////////////////////////////////////////////////////////// 347/////////////////////////////////////////////////////////////////////////////
190// encoder 348// encoder
191 349
380 538
381 if (enc->indent >= enc->json.max_depth) 539 if (enc->indent >= enc->json.max_depth)
382 croak (ERR_NESTING_EXCEEDED); 540 croak (ERR_NESTING_EXCEEDED);
383 541
384 encode_ch (enc, '['); 542 encode_ch (enc, '[');
385 543
386 if (len >= 0) 544 if (len >= 0)
387 { 545 {
388 encode_nl (enc); ++enc->indent; 546 encode_nl (enc); ++enc->indent;
389 547
390 for (i = 0; i <= len; ++i) 548 for (i = 0; i <= len; ++i)
402 encode_comma (enc); 560 encode_comma (enc);
403 } 561 }
404 562
405 encode_nl (enc); --enc->indent; encode_indent (enc); 563 encode_nl (enc); --enc->indent; encode_indent (enc);
406 } 564 }
407 565
408 encode_ch (enc, ']'); 566 encode_ch (enc, ']');
409} 567}
410 568
411static void 569static void
412encode_hk (enc_t *enc, HE *he) 570encode_hk (enc_t *enc, HE *he)
416 if (HeKLEN (he) == HEf_SVKEY) 574 if (HeKLEN (he) == HEf_SVKEY)
417 { 575 {
418 SV *sv = HeSVKEY (he); 576 SV *sv = HeSVKEY (he);
419 STRLEN len; 577 STRLEN len;
420 char *str; 578 char *str;
421 579
422 SvGETMAGIC (sv); 580 SvGETMAGIC (sv);
423 str = SvPV (sv, len); 581 str = SvPV (sv, len);
424 582
425 encode_str (enc, str, len, SvUTF8 (sv)); 583 encode_str (enc, str, len, SvUTF8 (sv));
426 } 584 }
492 } 650 }
493 651
494 if (count) 652 if (count)
495 { 653 {
496 int i, fast = 1; 654 int i, fast = 1;
497#if defined(__BORLANDC__) || defined(_MSC_VER) 655 HE *hes_stack [STACK_HES];
498 HE **hes = _alloca (count * sizeof (HE)); 656 HE **hes = hes_stack;
499#else 657
500 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 658 // allocate larger arrays on the heap
501#endif 659 if (count > STACK_HES)
660 {
661 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
662 hes = (HE **)SvPVX (sv);
663 }
502 664
503 i = 0; 665 i = 0;
504 while ((he = hv_iternext (hv))) 666 while ((he = hv_iternext (hv)))
505 { 667 {
506 hes [i++] = he; 668 hes [i++] = he;
575// encode objects, arrays and special \0=false and \1=true values. 737// encode objects, arrays and special \0=false and \1=true values.
576static void 738static void
577encode_rv (enc_t *enc, SV *sv) 739encode_rv (enc_t *enc, SV *sv)
578{ 740{
579 svtype svt; 741 svtype svt;
742 GV *method;
580 743
581 SvGETMAGIC (sv); 744 SvGETMAGIC (sv);
582 svt = SvTYPE (sv); 745 svt = SvTYPE (sv);
583 746
584 if (expect_false (SvOBJECT (sv))) 747 if (expect_false (SvOBJECT (sv)))
585 { 748 {
586 HV *stash = !JSON_SLOW || json_boolean_stash 749 HV *stash = SvSTASH (sv);
587 ? json_boolean_stash
588 : gv_stashpv ("JSON::XS::Boolean", 1);
589 750
590 if (SvSTASH (sv) == stash) 751 if (stash == bool_stash)
591 { 752 {
592 if (SvIV (sv)) 753 if (SvIV (sv))
593 encode_str (enc, "true", 4, 0); 754 encode_str (enc, "true", 4, 0);
594 else 755 else
595 encode_str (enc, "false", 5, 0); 756 encode_str (enc, "false", 5, 0);
596 } 757 }
758 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
759 {
760 int count;
761 dSP;
762
763 ENTER; SAVETMPS;
764 SAVESTACK_POS ();
765 PUSHMARK (SP);
766 EXTEND (SP, 2);
767 // we re-bless the reference to get overload and other niceties right
768 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
769 PUSHs (sv_json);
770
771 PUTBACK;
772 count = call_sv ((SV *)GvCV (method), G_ARRAY);
773 SPAGAIN;
774
775 // catch this surprisingly common error
776 if (SvROK (TOPs) && SvRV (TOPs) == sv)
777 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
778
779 encode_ch (enc, '(');
780 encode_ch (enc, '"');
781 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
782 encode_ch (enc, '"');
783 encode_ch (enc, ')');
784 encode_ch (enc, '[');
785
786 while (count)
787 {
788 encode_sv (enc, SP[1 - count--]);
789
790 if (count)
791 encode_ch (enc, ',');
792 }
793
794 encode_ch (enc, ']');
795
796 FREETMPS; LEAVE;
797 }
798 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
799 {
800 dSP;
801
802 ENTER; SAVETMPS;
803 PUSHMARK (SP);
804 // we re-bless the reference to get overload and other niceties right
805 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
806
807 // calling with G_SCALAR ensures that we always get a 1 return value
808 PUTBACK;
809 call_sv ((SV *)GvCV (method), G_SCALAR);
810 SPAGAIN;
811
812 // catch this surprisingly common error
813 if (SvROK (TOPs) && SvRV (TOPs) == sv)
814 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
815
816 sv = POPs;
817 PUTBACK;
818
819 encode_sv (enc, sv);
820
821 FREETMPS; LEAVE;
822 }
823 else if (enc->json.flags & F_ALLOW_BLESSED)
824 encode_str (enc, "null", 4, 0);
597 else 825 else
598 {
599#if 0
600 if (0 && sv_derived_from (rv, "JSON::Literal"))
601 {
602 // not yet
603 }
604#endif
605 if (enc->json.flags & F_CONV_BLESSED)
606 {
607 // we re-bless the reference to get overload and other niceties right
608 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
609
610 if (to_json)
611 {
612 dSP;
613
614 ENTER; SAVETMPS; PUSHMARK (SP);
615 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
616
617 // calling with G_SCALAR ensures that we always get a 1 return value
618 PUTBACK;
619 call_sv ((SV *)GvCV (to_json), G_SCALAR);
620 SPAGAIN;
621
622 // catch this surprisingly common error
623 if (SvROK (TOPs) && SvRV (TOPs) == sv)
624 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
625
626 sv = POPs;
627 PUTBACK;
628
629 encode_sv (enc, sv);
630
631 FREETMPS; LEAVE;
632 }
633 else if (enc->json.flags & F_ALLOW_BLESSED)
634 encode_str (enc, "null", 4, 0);
635 else
636 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
637 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
638 }
639 else if (enc->json.flags & F_ALLOW_BLESSED)
640 encode_str (enc, "null", 4, 0);
641 else
642 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 826 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
643 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 827 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
644 }
645 } 828 }
646 else if (svt == SVt_PVHV) 829 else if (svt == SVt_PVHV)
647 encode_hv (enc, (HV *)sv); 830 encode_hv (enc, (HV *)sv);
648 else if (svt == SVt_PVAV) 831 else if (svt == SVt_PVAV)
649 encode_av (enc, (AV *)sv); 832 encode_av (enc, (AV *)sv);
650 else if (svt < SVt_PVAV) 833 else if (svt < SVt_PVAV)
651 { 834 {
652 STRLEN len = 0; 835 int bool_type = ref_bool_type (sv);
653 char *pv = svt ? SvPV (sv, len) : 0;
654 836
655 if (len == 1 && *pv == '1') 837 if (bool_type == 1)
656 encode_str (enc, "true", 4, 0); 838 encode_str (enc, "true", 4, 0);
657 else if (len == 1 && *pv == '0') 839 else if (bool_type == 0)
658 encode_str (enc, "false", 5, 0); 840 encode_str (enc, "false", 5, 0);
659 else if (enc->json.flags & F_ALLOW_UNKNOWN) 841 else if (enc->json.flags & F_ALLOW_UNKNOWN)
660 encode_str (enc, "null", 4, 0); 842 encode_str (enc, "null", 4, 0);
661 else 843 else
662 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 844 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
725 } 907 }
726 else 908 else
727 { 909 {
728 // large integer, use the (rather slow) snprintf way. 910 // large integer, use the (rather slow) snprintf way.
729 need (enc, IVUV_MAXCHARS); 911 need (enc, IVUV_MAXCHARS);
730 enc->cur += 912 enc->cur +=
731 SvIsUV(sv) 913 SvIsUV(sv)
732 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 914 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
733 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 915 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
734 } 916 }
735 } 917 }
736 else if (SvROK (sv)) 918 else if (SvROK (sv))
737 encode_rv (enc, SvRV (sv)); 919 encode_rv (enc, SvRV (sv));
738 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 920 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
739 encode_str (enc, "null", 4, 0); 921 encode_str (enc, "null", 4, 0);
740 else 922 else
741 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 923 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
742 SvPV_nolen (sv), SvFLAGS (sv)); 924 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
743} 925}
744 926
745static SV * 927static SV *
746encode_json (SV *scalar, JSON *json) 928encode_json (SV *scalar, JSON *json)
747{ 929{
748 enc_t enc; 930 enc_t enc;
749 931
750 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 932 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
751 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 933 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
752 934
753 enc.json = *json; 935 enc.json = *json;
754 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 936 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
755 enc.cur = SvPVX (enc.sv); 937 enc.cur = SvPVX (enc.sv);
949 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1131 else if (expect_true (ch >= 0x20 && ch < 0x80))
950 *cur++ = ch; 1132 *cur++ = ch;
951 else if (ch >= 0x80) 1133 else if (ch >= 0x80)
952 { 1134 {
953 STRLEN clen; 1135 STRLEN clen;
954 UV uch;
955 1136
956 --dec_cur; 1137 --dec_cur;
957 1138
958 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1139 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
959 if (clen == (STRLEN)-1) 1140 if (clen == (STRLEN)-1)
960 ERR ("malformed UTF-8 character in JSON string"); 1141 ERR ("malformed UTF-8 character in JSON string");
961 1142
962 do 1143 do
963 *cur++ = *dec_cur++; 1144 *cur++ = *dec_cur++;
964 while (--clen); 1145 while (--clen);
965 1146
966 utf8 = 1; 1147 utf8 = 1;
967 } 1148 }
1149 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1150 *cur++ = ch;
968 else 1151 else
969 { 1152 {
970 --dec_cur; 1153 --dec_cur;
971 1154
972 if (!ch) 1155 if (!ch)
1117 } 1300 }
1118 1301
1119 len -= *start == '-' ? 1 : 0; 1302 len -= *start == '-' ? 1 : 0;
1120 1303
1121 // does not fit into IV or UV, try NV 1304 // does not fit into IV or UV, try NV
1122 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1305 if (len <= NV_DIG)
1123 #if defined (LDBL_DIG)
1124 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1125 #endif
1126 )
1127 // fits into NV without loss of precision 1306 // fits into NV without loss of precision
1128 return newSVnv (Atof (start)); 1307 return newSVnv (json_atof (start));
1129 1308
1130 // everything else fails, convert it to a string 1309 // everything else fails, convert it to a string
1131 return newSVpvn (start, dec->cur - start); 1310 return newSVpvn (start, dec->cur - start);
1132 } 1311 }
1133 1312
1134 // loss of precision here 1313 // loss of precision here
1135 return newSVnv (Atof (start)); 1314 return newSVnv (json_atof (start));
1136 1315
1137fail: 1316fail:
1138 return 0; 1317 return 0;
1139} 1318}
1140 1319
1164 if (*dec->cur == ']') 1343 if (*dec->cur == ']')
1165 { 1344 {
1166 ++dec->cur; 1345 ++dec->cur;
1167 break; 1346 break;
1168 } 1347 }
1169 1348
1170 if (*dec->cur != ',') 1349 if (*dec->cur != ',')
1171 ERR (", or ] expected while parsing array"); 1350 ERR (", or ] expected while parsing array");
1172 1351
1173 ++dec->cur; 1352 ++dec->cur;
1174 1353
1298 1477
1299 hv_iterinit (hv); 1478 hv_iterinit (hv);
1300 he = hv_iternext (hv); 1479 he = hv_iternext (hv);
1301 hv_iterinit (hv); 1480 hv_iterinit (hv);
1302 1481
1303 // the next line creates a mortal sv each time its called. 1482 // the next line creates a mortal sv each time it's called.
1304 // might want to optimise this for common cases. 1483 // might want to optimise this for common cases.
1305 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1484 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1306 1485
1307 if (cb) 1486 if (cb)
1308 { 1487 {
1309 dSP; 1488 dSP;
1310 int count; 1489 int count;
1311 1490
1312 ENTER; SAVETMPS; PUSHMARK (SP); 1491 ENTER; SAVETMPS;
1492 SAVESTACK_POS ();
1493 PUSHMARK (SP);
1313 XPUSHs (HeVAL (he)); 1494 XPUSHs (HeVAL (he));
1495 sv_2mortal (sv);
1314 1496
1315 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1497 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1316 1498
1317 if (count == 1) 1499 if (count == 1)
1318 { 1500 {
1319 sv = newSVsv (POPs); 1501 sv = newSVsv (POPs);
1320 FREETMPS; LEAVE; 1502 FREETMPS; LEAVE;
1321 return sv; 1503 return sv;
1322 } 1504 }
1323 1505
1506 SvREFCNT_inc (sv);
1324 FREETMPS; LEAVE; 1507 FREETMPS; LEAVE;
1325 } 1508 }
1326 } 1509 }
1327 1510
1328 if (dec->json.cb_object) 1511 if (dec->json.cb_object)
1329 { 1512 {
1330 dSP; 1513 dSP;
1331 int count; 1514 int count;
1332 1515
1333 ENTER; SAVETMPS; PUSHMARK (SP); 1516 ENTER; SAVETMPS;
1517 SAVESTACK_POS ();
1518 PUSHMARK (SP);
1334 XPUSHs (sv_2mortal (sv)); 1519 XPUSHs (sv_2mortal (sv));
1335 1520
1336 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1521 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1337 1522
1338 if (count == 1) 1523 if (count == 1)
1354 DEC_DEC_DEPTH; 1539 DEC_DEC_DEPTH;
1355 return 0; 1540 return 0;
1356} 1541}
1357 1542
1358static SV * 1543static SV *
1544decode_tag (dec_t *dec)
1545{
1546 SV *tag = 0;
1547 SV *val = 0;
1548
1549 if (!(dec->json.flags & F_ALLOW_TAGS))
1550 ERR ("malformed JSON string, neither array, object, number, string or atom");
1551
1552 ++dec->cur;
1553
1554 decode_ws (dec);
1555
1556 tag = decode_sv (dec);
1557 if (!tag)
1558 goto fail;
1559
1560 if (!SvPOK (tag))
1561 ERR ("malformed JSON string, (tag) must be a string");
1562
1563 decode_ws (dec);
1564
1565 if (*dec->cur != ')')
1566 ERR (") expected after tag");
1567
1568 ++dec->cur;
1569
1570 decode_ws (dec);
1571
1572 val = decode_sv (dec);
1573 if (!val)
1574 goto fail;
1575
1576 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1577 ERR ("malformed JSON string, tag value must be an array");
1578
1579 {
1580 AV *av = (AV *)SvRV (val);
1581 int i, len = av_len (av) + 1;
1582 HV *stash = gv_stashsv (tag, 0);
1583 SV *sv;
1584
1585 if (!stash)
1586 ERR ("cannot decode perl-object (package does not exist)");
1587
1588 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1589
1590 if (!method)
1591 ERR ("cannot decode perl-object (package does not have a THAW method)");
1592
1593 dSP;
1594
1595 ENTER; SAVETMPS;
1596 PUSHMARK (SP);
1597 EXTEND (SP, len + 2);
1598 // we re-bless the reference to get overload and other niceties right
1599 PUSHs (tag);
1600 PUSHs (sv_json);
1601
1602 for (i = 0; i < len; ++i)
1603 PUSHs (*av_fetch (av, i, 1));
1604
1605 PUTBACK;
1606 call_sv ((SV *)GvCV (method), G_SCALAR);
1607 SPAGAIN;
1608
1609 SvREFCNT_dec (tag);
1610 SvREFCNT_dec (val);
1611 sv = SvREFCNT_inc (POPs);
1612
1613 PUTBACK;
1614
1615 FREETMPS; LEAVE;
1616
1617 return sv;
1618 }
1619
1620fail:
1621 SvREFCNT_dec (tag);
1622 SvREFCNT_dec (val);
1623 return 0;
1624}
1625
1626static SV *
1359decode_sv (dec_t *dec) 1627decode_sv (dec_t *dec)
1360{ 1628{
1361 // the beauty of JSON: you need exactly one character lookahead 1629 // the beauty of JSON: you need exactly one character lookahead
1362 // to parse everything. 1630 // to parse everything.
1363 switch (*dec->cur) 1631 switch (*dec->cur)
1364 { 1632 {
1365 case '"': ++dec->cur; return decode_str (dec); 1633 case '"': ++dec->cur; return decode_str (dec);
1366 case '[': ++dec->cur; return decode_av (dec); 1634 case '[': ++dec->cur; return decode_av (dec);
1367 case '{': ++dec->cur; return decode_hv (dec); 1635 case '{': ++dec->cur; return decode_hv (dec);
1636 case '(': return decode_tag (dec);
1368 1637
1369 case '-': 1638 case '-':
1370 case '0': case '1': case '2': case '3': case '4': 1639 case '0': case '1': case '2': case '3': case '4':
1371 case '5': case '6': case '7': case '8': case '9': 1640 case '5': case '6': case '7': case '8': case '9':
1372 return decode_num (dec); 1641 return decode_num (dec);
1374 case 't': 1643 case 't':
1375 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1644 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1376 { 1645 {
1377 dec->cur += 4; 1646 dec->cur += 4;
1378#if JSON_SLOW 1647#if JSON_SLOW
1379 json_true = get_bool ("JSON::XS::true"); 1648 bool_true = get_bool ("Types::Serialiser::true");
1380#endif 1649#endif
1381 return newSVsv (json_true); 1650 return newSVsv (bool_true);
1382 } 1651 }
1383 else 1652 else
1384 ERR ("'true' expected"); 1653 ERR ("'true' expected");
1385 1654
1386 break; 1655 break;
1388 case 'f': 1657 case 'f':
1389 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1658 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1390 { 1659 {
1391 dec->cur += 5; 1660 dec->cur += 5;
1392#if JSON_SLOW 1661#if JSON_SLOW
1393 json_false = get_bool ("JSON::XS::false"); 1662 bool_false = get_bool ("Types::Serialiser::false");
1394#endif 1663#endif
1395 return newSVsv (json_false); 1664 return newSVsv (bool_false);
1396 } 1665 }
1397 else 1666 else
1398 ERR ("'false' expected"); 1667 ERR ("'false' expected");
1399 1668
1400 break; 1669 break;
1409 ERR ("'null' expected"); 1678 ERR ("'null' expected");
1410 1679
1411 break; 1680 break;
1412 1681
1413 default: 1682 default:
1414 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1683 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1415 break; 1684 break;
1416 } 1685 }
1417 1686
1418fail: 1687fail:
1419 return 0; 1688 return 0;
1424{ 1693{
1425 dec_t dec; 1694 dec_t dec;
1426 SV *sv; 1695 SV *sv;
1427 1696
1428 /* work around bugs in 5.10 where manipulating magic values 1697 /* work around bugs in 5.10 where manipulating magic values
1429 * will perl ignore the magic in subsequent accesses 1698 * makes perl ignore the magic in subsequent accesses.
1699 * also make a copy of non-PV values, to get them into a clean
1700 * state (SvPV should do that, but it's buggy, see below).
1701 *
1702 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1703 * as reported by Reini Urban.
1430 */ 1704 */
1431 /*SvGETMAGIC (string);*/ 1705 /*SvGETMAGIC (string);*/
1432 if (SvMAGICAL (string)) 1706 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1433 string = sv_2mortal (newSVsv (string)); 1707 string = sv_2mortal (newSVsv (string));
1434 1708
1435 SvUPGRADE (string, SVt_PV); 1709 SvUPGRADE (string, SVt_PV);
1436 1710
1437 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1711 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1454 if (offset > json->max_size && json->max_size) 1728 if (offset > json->max_size && json->max_size)
1455 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1729 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1456 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1730 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1457 } 1731 }
1458 1732
1459 if (json->flags & F_UTF8) 1733 if (DECODE_WANTS_OCTETS (json))
1460 sv_utf8_downgrade (string, 0); 1734 sv_utf8_downgrade (string, 0);
1461 else 1735 else
1462 sv_utf8_upgrade (string); 1736 sv_utf8_upgrade (string);
1463 1737
1464 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1738 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1506 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1780 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1507 LEAVE; 1781 LEAVE;
1508 1782
1509 croak ("%s, at character offset %d (before \"%s\")", 1783 croak ("%s, at character offset %d (before \"%s\")",
1510 dec.err, 1784 dec.err,
1511 ptr_to_index (string, dec.cur), 1785 (int)ptr_to_index (string, dec.cur),
1512 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1786 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1513 } 1787 }
1514 1788
1515 sv = sv_2mortal (sv); 1789 sv = sv_2mortal (sv);
1516 1790
1517 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1791 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1518 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1792 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1519 1793
1520 return sv; 1794 return sv;
1521} 1795}
1522 1796
1525 1799
1526static void 1800static void
1527incr_parse (JSON *self) 1801incr_parse (JSON *self)
1528{ 1802{
1529 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1803 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1804
1805 // the state machine here is a bit convoluted and could be simplified a lot
1806 // but this would make it slower, so...
1530 1807
1531 for (;;) 1808 for (;;)
1532 { 1809 {
1533 //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 1810 //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
1534 switch (self->incr_mode) 1811 switch (self->incr_mode)
1535 { 1812 {
1536 // only used for intiial whitespace skipping 1813 // only used for initial whitespace skipping
1537 case INCR_M_WS: 1814 case INCR_M_WS:
1538 for (;;) 1815 for (;;)
1539 { 1816 {
1540 if (*p > 0x20) 1817 if (*p > 0x20)
1541 { 1818 {
1819 if (*p == '#')
1820 {
1821 self->incr_mode = INCR_M_C0;
1822 goto incr_m_c;
1823 }
1824 else
1825 {
1542 self->incr_mode = INCR_M_JSON; 1826 self->incr_mode = INCR_M_JSON;
1543 goto incr_m_json; 1827 goto incr_m_json;
1828 }
1544 } 1829 }
1545 else if (!*p) 1830 else if (!*p)
1546 goto interrupt; 1831 goto interrupt;
1547 1832
1548 ++p; 1833 ++p;
1554 goto interrupt; 1839 goto interrupt;
1555 1840
1556 ++p; 1841 ++p;
1557 self->incr_mode = INCR_M_STR; 1842 self->incr_mode = INCR_M_STR;
1558 goto incr_m_str; 1843 goto incr_m_str;
1844
1845 // inside #-style comments
1846 case INCR_M_C0:
1847 case INCR_M_C1:
1848 incr_m_c:
1849 for (;;)
1850 {
1851 if (*p == '\n')
1852 {
1853 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1854 break;
1855 }
1856 else if (!*p)
1857 goto interrupt;
1858
1859 ++p;
1860 }
1861
1862 break;
1559 1863
1560 // inside a string 1864 // inside a string
1561 case INCR_M_STR: 1865 case INCR_M_STR:
1562 incr_m_str: 1866 incr_m_str:
1563 for (;;) 1867 for (;;)
1614 self->incr_mode = INCR_M_STR; 1918 self->incr_mode = INCR_M_STR;
1615 goto incr_m_str; 1919 goto incr_m_str;
1616 1920
1617 case '[': 1921 case '[':
1618 case '{': 1922 case '{':
1923 case '(':
1619 if (++self->incr_nest > self->max_depth) 1924 if (++self->incr_nest > self->max_depth)
1620 croak (ERR_NESTING_EXCEEDED); 1925 croak (ERR_NESTING_EXCEEDED);
1621 break; 1926 break;
1622 1927
1623 case ']': 1928 case ']':
1624 case '}': 1929 case '}':
1625 if (--self->incr_nest <= 0) 1930 if (--self->incr_nest <= 0)
1626 goto interrupt; 1931 goto interrupt;
1932 break;
1933
1934 case ')':
1935 --self->incr_nest;
1936 break;
1937
1938 case '#':
1939 self->incr_mode = INCR_M_C1;
1940 goto incr_m_c;
1627 } 1941 }
1628 } 1942 }
1629 } 1943 }
1630 1944
1631 modechange: 1945 modechange:
1632 ; 1946 ;
1633 } 1947 }
1634 1948
1635interrupt: 1949interrupt:
1636 self->incr_pos = p - SvPVX (self->incr_text); 1950 self->incr_pos = p - SvPVX (self->incr_text);
1951 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1637 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1952 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1638} 1953}
1639 1954
1640///////////////////////////////////////////////////////////////////////////// 1955/////////////////////////////////////////////////////////////////////////////
1641// XS interface functions 1956// XS interface functions
1651 i >= '0' && i <= '9' ? i - '0' 1966 i >= '0' && i <= '9' ? i - '0'
1652 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1967 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1653 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1968 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1654 : -1; 1969 : -1;
1655 1970
1656 json_stash = gv_stashpv ("JSON::XS" , 1); 1971 json_stash = gv_stashpv ("JSON::XS" , 1);
1657 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1972 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1658
1659 json_true = get_bool ("JSON::XS::true"); 1973 bool_true = get_bool ("Types::Serialiser::true");
1660 json_false = get_bool ("JSON::XS::false"); 1974 bool_false = get_bool ("Types::Serialiser::false");
1975
1976 sv_json = newSVpv ("JSON", 0);
1977 SvREADONLY_on (sv_json);
1661 1978
1662 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ 1979 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1663} 1980}
1664 1981
1665PROTOTYPES: DISABLE 1982PROTOTYPES: DISABLE
1666 1983
1667void CLONE (...) 1984void CLONE (...)
1668 CODE: 1985 CODE:
1669 json_stash = 0; 1986 json_stash = 0;
1670 json_boolean_stash = 0; 1987 bool_stash = 0;
1671 1988
1672void new (char *klass) 1989void new (char *klass)
1673 PPCODE: 1990 PPCODE:
1674{ 1991{
1675 SV *pv = NEWSV (0, sizeof (JSON)); 1992 SV *pv = NEWSV (0, sizeof (JSON));
1676 SvPOK_only (pv); 1993 SvPOK_only (pv);
1677 json_init ((JSON *)SvPVX (pv)); 1994 json_init ((JSON *)SvPVX (pv));
1678 XPUSHs (sv_2mortal (sv_bless ( 1995 XPUSHs (sv_2mortal (sv_bless (
1679 newRV_noinc (pv), 1996 newRV_noinc (pv),
1680 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1997 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1695 shrink = F_SHRINK 2012 shrink = F_SHRINK
1696 allow_blessed = F_ALLOW_BLESSED 2013 allow_blessed = F_ALLOW_BLESSED
1697 convert_blessed = F_CONV_BLESSED 2014 convert_blessed = F_CONV_BLESSED
1698 relaxed = F_RELAXED 2015 relaxed = F_RELAXED
1699 allow_unknown = F_ALLOW_UNKNOWN 2016 allow_unknown = F_ALLOW_UNKNOWN
2017 allow_tags = F_ALLOW_TAGS
1700 PPCODE: 2018 PPCODE:
1701{ 2019{
1702 if (enable) 2020 if (enable)
1703 self->flags |= ix; 2021 self->flags |= ix;
1704 else 2022 else
1720 get_shrink = F_SHRINK 2038 get_shrink = F_SHRINK
1721 get_allow_blessed = F_ALLOW_BLESSED 2039 get_allow_blessed = F_ALLOW_BLESSED
1722 get_convert_blessed = F_CONV_BLESSED 2040 get_convert_blessed = F_CONV_BLESSED
1723 get_relaxed = F_RELAXED 2041 get_relaxed = F_RELAXED
1724 get_allow_unknown = F_ALLOW_UNKNOWN 2042 get_allow_unknown = F_ALLOW_UNKNOWN
2043 get_allow_tags = F_ALLOW_TAGS
1725 PPCODE: 2044 PPCODE:
1726 XPUSHs (boolSV (self->flags & ix)); 2045 XPUSHs (boolSV (self->flags & ix));
1727 2046
1728void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2047void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1729 PPCODE: 2048 PPCODE:
1757} 2076}
1758 2077
1759void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2078void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1760 PPCODE: 2079 PPCODE:
1761{ 2080{
1762 if (!self->cb_sk_object) 2081 if (!self->cb_sk_object)
1763 self->cb_sk_object = newHV (); 2082 self->cb_sk_object = newHV ();
1764 2083
1765 if (SvOK (cb)) 2084 if (SvOK (cb))
1766 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2085 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1767 else 2086 else
1778 XPUSHs (ST (0)); 2097 XPUSHs (ST (0));
1779} 2098}
1780 2099
1781void encode (JSON *self, SV *scalar) 2100void encode (JSON *self, SV *scalar)
1782 PPCODE: 2101 PPCODE:
1783 XPUSHs (encode_json (scalar, self)); 2102 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2103 XPUSHs (scalar);
1784 2104
1785void decode (JSON *self, SV *jsonstr) 2105void decode (JSON *self, SV *jsonstr)
1786 PPCODE: 2106 PPCODE:
1787 XPUSHs (decode_json (jsonstr, self, 0)); 2107 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2108 XPUSHs (jsonstr);
1788 2109
1789void decode_prefix (JSON *self, SV *jsonstr) 2110void decode_prefix (JSON *self, SV *jsonstr)
1790 PPCODE: 2111 PPCODE:
1791{ 2112{
2113 SV *sv;
1792 char *offset; 2114 char *offset;
2115 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1793 EXTEND (SP, 2); 2116 EXTEND (SP, 2);
1794 PUSHs (decode_json (jsonstr, self, &offset)); 2117 PUSHs (sv);
1795 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2118 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1796} 2119}
1797 2120
1798void incr_parse (JSON *self, SV *jsonstr = 0) 2121void incr_parse (JSON *self, SV *jsonstr = 0)
1799 PPCODE: 2122 PPCODE:
1800{ 2123{
1801 if (!self->incr_text) 2124 if (!self->incr_text)
1802 self->incr_text = newSVpvn ("", 0); 2125 self->incr_text = newSVpvn ("", 0);
2126
2127 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2128 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2129 if (DECODE_WANTS_OCTETS (self))
2130 {
2131 if (self->incr_pos)
2132 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2133 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2134
2135 sv_utf8_downgrade (self->incr_text, 0);
2136 }
2137 else
2138 {
2139 sv_utf8_upgrade (self->incr_text);
2140
2141 if (self->incr_pos)
2142 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2143 - (U8 *)SvPVX (self->incr_text);
2144 }
1803 2145
1804 // append data, if any 2146 // append data, if any
1805 if (jsonstr) 2147 if (jsonstr)
1806 { 2148 {
2149 /* make sure both strings have same encoding */
2150 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1807 if (SvUTF8 (jsonstr)) 2151 if (SvUTF8 (jsonstr))
2152 sv_utf8_downgrade (jsonstr, 0);
1808 { 2153 else
1809 if (!SvUTF8 (self->incr_text))
1810 {
1811 /* utf-8-ness differs, need to upgrade */
1812 sv_utf8_upgrade (self->incr_text);
1813
1814 if (self->incr_pos)
1815 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1816 - (U8 *)SvPVX (self->incr_text);
1817 }
1818 }
1819 else if (SvUTF8 (self->incr_text))
1820 sv_utf8_upgrade (jsonstr); 2154 sv_utf8_upgrade (jsonstr);
1821 2155
2156 /* and then just blindly append */
1822 { 2157 {
1823 STRLEN len; 2158 STRLEN len;
1824 const char *str = SvPV (jsonstr, len); 2159 const char *str = SvPV (jsonstr, len);
1825 STRLEN cur = SvCUR (self->incr_text); 2160 STRLEN cur = SvCUR (self->incr_text);
1826 2161
1834 } 2169 }
1835 2170
1836 if (GIMME_V != G_VOID) 2171 if (GIMME_V != G_VOID)
1837 do 2172 do
1838 { 2173 {
2174 SV *sv;
1839 char *offset; 2175 char *offset;
1840 2176
1841 if (!INCR_DONE (self)) 2177 if (!INCR_DONE (self))
1842 { 2178 {
1843 incr_parse (self); 2179 incr_parse (self);
1845 if (self->incr_pos > self->max_size && self->max_size) 2181 if (self->incr_pos > self->max_size && self->max_size)
1846 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2182 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1847 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2183 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1848 2184
1849 if (!INCR_DONE (self)) 2185 if (!INCR_DONE (self))
2186 {
2187 // as an optimisation, do not accumulate white space in the incr buffer
2188 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2189 {
2190 self->incr_pos = 0;
2191 SvCUR_set (self->incr_text, 0);
2192 }
2193
1850 break; 2194 break;
2195 }
1851 } 2196 }
1852 2197
1853 XPUSHs (decode_json (self->incr_text, self, &offset)); 2198 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2199 XPUSHs (sv);
1854 2200
1855 self->incr_pos -= offset - SvPVX (self->incr_text); 2201 self->incr_pos -= offset - SvPVX (self->incr_text);
1856 self->incr_nest = 0; 2202 self->incr_nest = 0;
1857 self->incr_mode = 0; 2203 self->incr_mode = 0;
1858 2204
1902 SvREFCNT_dec (self->incr_text); 2248 SvREFCNT_dec (self->incr_text);
1903 2249
1904PROTOTYPES: ENABLE 2250PROTOTYPES: ENABLE
1905 2251
1906void encode_json (SV *scalar) 2252void encode_json (SV *scalar)
1907 ALIAS:
1908 to_json_ = 0
1909 encode_json = F_UTF8
1910 PPCODE: 2253 PPCODE:
1911{ 2254{
1912 JSON json; 2255 JSON json;
1913 json_init (&json); 2256 json_init (&json);
1914 json.flags |= ix; 2257 json.flags |= F_UTF8;
1915 XPUSHs (encode_json (scalar, &json)); 2258 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2259 XPUSHs (scalar);
1916} 2260}
1917 2261
1918void decode_json (SV *jsonstr) 2262void decode_json (SV *jsonstr)
1919 ALIAS:
1920 from_json_ = 0
1921 decode_json = F_UTF8
1922 PPCODE: 2263 PPCODE:
1923{ 2264{
1924 JSON json; 2265 JSON json;
1925 json_init (&json); 2266 json_init (&json);
1926 json.flags |= ix; 2267 json.flags |= F_UTF8;
1927 XPUSHs (decode_json (jsonstr, &json, 0)); 2268 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2269 XPUSHs (jsonstr);
1928} 2270}
1929 2271
1930

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines