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 1.86 by root, Tue May 27 05:31:39 2008 UTC vs.
Revision 1.126 by root, Sun Feb 21 16:18:19 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)
121 143
122INLINE void 144INLINE void
123shrink (SV *sv) 145shrink (SV *sv)
124{ 146{
125 sv_utf8_downgrade (sv, 1); 147 sv_utf8_downgrade (sv, 1);
148
126 if (SvLEN (sv) > SvCUR (sv) + 1) 149 if (SvLEN (sv) > SvCUR (sv) + 1)
127 { 150 {
128#ifdef SvPV_shrink_to_cur 151#ifdef SvPV_shrink_to_cur
129 SvPV_shrink_to_cur (sv); 152 SvPV_shrink_to_cur (sv);
130#elif defined (SvPV_renew) 153#elif defined (SvPV_renew)
174 *s++ = 0x80 | ( ch & 0x3f); 197 *s++ = 0x80 | ( ch & 0x3f);
175 198
176 return s; 199 return s;
177} 200}
178 201
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
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;
345}
346
179///////////////////////////////////////////////////////////////////////////// 347/////////////////////////////////////////////////////////////////////////////
180// encoder 348// encoder
181 349
182// structure used for encoding JSON 350// structure used for encoding JSON
183typedef struct 351typedef struct
193INLINE void 361INLINE void
194need (enc_t *enc, STRLEN len) 362need (enc_t *enc, STRLEN len)
195{ 363{
196 if (expect_false (enc->cur + len >= enc->end)) 364 if (expect_false (enc->cur + len >= enc->end))
197 { 365 {
198 STRLEN cur = enc->cur - SvPVX (enc->sv); 366 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
199 SvGROW (enc->sv, cur + len + 1); 367 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
200 enc->cur = SvPVX (enc->sv) + cur; 368 enc->cur = SvPVX (enc->sv) + cur;
201 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 369 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
202 } 370 }
203} 371}
204 372
279 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 447 (int)((uch - 0x10000) % 0x400 + 0xDC00));
280 enc->cur += 12; 448 enc->cur += 12;
281 } 449 }
282 else 450 else
283 { 451 {
284 static char hexdigit [16] = "0123456789abcdef";
285 need (enc, len += 5); 452 need (enc, len += 5);
286 *enc->cur++ = '\\'; 453 *enc->cur++ = '\\';
287 *enc->cur++ = 'u'; 454 *enc->cur++ = 'u';
288 *enc->cur++ = hexdigit [ uch >> 12 ]; 455 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
289 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 456 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
290 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 457 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
291 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 458 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
292 } 459 }
293 460
294 str += clen; 461 str += clen;
295 } 462 }
296 else if (enc->json.flags & F_LATIN1) 463 else if (enc->json.flags & F_LATIN1)
371 538
372 if (enc->indent >= enc->json.max_depth) 539 if (enc->indent >= enc->json.max_depth)
373 croak (ERR_NESTING_EXCEEDED); 540 croak (ERR_NESTING_EXCEEDED);
374 541
375 encode_ch (enc, '['); 542 encode_ch (enc, '[');
376 543
377 if (len >= 0) 544 if (len >= 0)
378 { 545 {
379 encode_nl (enc); ++enc->indent; 546 encode_nl (enc); ++enc->indent;
380 547
381 for (i = 0; i <= len; ++i) 548 for (i = 0; i <= len; ++i)
393 encode_comma (enc); 560 encode_comma (enc);
394 } 561 }
395 562
396 encode_nl (enc); --enc->indent; encode_indent (enc); 563 encode_nl (enc); --enc->indent; encode_indent (enc);
397 } 564 }
398 565
399 encode_ch (enc, ']'); 566 encode_ch (enc, ']');
400} 567}
401 568
402static void 569static void
403encode_hk (enc_t *enc, HE *he) 570encode_hk (enc_t *enc, HE *he)
407 if (HeKLEN (he) == HEf_SVKEY) 574 if (HeKLEN (he) == HEf_SVKEY)
408 { 575 {
409 SV *sv = HeSVKEY (he); 576 SV *sv = HeSVKEY (he);
410 STRLEN len; 577 STRLEN len;
411 char *str; 578 char *str;
412 579
413 SvGETMAGIC (sv); 580 SvGETMAGIC (sv);
414 str = SvPV (sv, len); 581 str = SvPV (sv, len);
415 582
416 encode_str (enc, str, len, SvUTF8 (sv)); 583 encode_str (enc, str, len, SvUTF8 (sv));
417 } 584 }
460 627
461 encode_ch (enc, '{'); 628 encode_ch (enc, '{');
462 629
463 // for canonical output we have to sort by keys first 630 // for canonical output we have to sort by keys first
464 // actually, this is mostly due to the stupid so-called 631 // actually, this is mostly due to the stupid so-called
465 // security workaround added somewhere in 5.8.x. 632 // security workaround added somewhere in 5.8.x
466 // that randomises hash orderings 633 // that randomises hash orderings
467 if (enc->json.flags & F_CANONICAL) 634 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
468 { 635 {
469 int count = hv_iterinit (hv); 636 int count = hv_iterinit (hv);
470 637
471 if (SvMAGICAL (hv)) 638 if (SvMAGICAL (hv))
472 { 639 {
483 } 650 }
484 651
485 if (count) 652 if (count)
486 { 653 {
487 int i, fast = 1; 654 int i, fast = 1;
488#if defined(__BORLANDC__) || defined(_MSC_VER) 655 HE *hes_stack [STACK_HES];
489 HE **hes = _alloca (count * sizeof (HE)); 656 HE **hes = hes_stack;
490#else 657
491 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 658 // allocate larger arrays on the heap
492#endif 659 if (count > STACK_HES)
660 {
661 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
662 hes = (HE **)SvPVX (sv);
663 }
493 664
494 i = 0; 665 i = 0;
495 while ((he = hv_iternext (hv))) 666 while ((he = hv_iternext (hv)))
496 { 667 {
497 hes [i++] = he; 668 hes [i++] = he;
566// encode objects, arrays and special \0=false and \1=true values. 737// encode objects, arrays and special \0=false and \1=true values.
567static void 738static void
568encode_rv (enc_t *enc, SV *sv) 739encode_rv (enc_t *enc, SV *sv)
569{ 740{
570 svtype svt; 741 svtype svt;
742 GV *method;
571 743
572 SvGETMAGIC (sv); 744 SvGETMAGIC (sv);
573 svt = SvTYPE (sv); 745 svt = SvTYPE (sv);
574 746
575 if (expect_false (SvOBJECT (sv))) 747 if (expect_false (SvOBJECT (sv)))
576 { 748 {
577 HV *stash = !JSON_SLOW || json_boolean_stash 749 HV *stash = SvSTASH (sv);
578 ? json_boolean_stash
579 : gv_stashpv ("JSON::XS::Boolean", 1);
580 750
581 if (SvSTASH (sv) == stash) 751 if (stash == bool_stash)
582 { 752 {
583 if (SvIV (sv)) 753 if (SvIV (sv))
584 encode_str (enc, "true", 4, 0); 754 encode_str (enc, "true", 4, 0);
585 else 755 else
586 encode_str (enc, "false", 5, 0); 756 encode_str (enc, "false", 5, 0);
587 } 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);
588 else 825 else
589 {
590#if 0
591 if (0 && sv_derived_from (rv, "JSON::Literal"))
592 {
593 // not yet
594 }
595#endif
596 if (enc->json.flags & F_CONV_BLESSED)
597 {
598 // 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)));
607
608 // calling with G_SCALAR ensures that we always get a 1 return value
609 PUTBACK;
610 call_sv ((SV *)GvCV (to_json), G_SCALAR);
611 SPAGAIN;
612
613 // catch this surprisingly common error
614 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)));
616
617 sv = POPs;
618 PUTBACK;
619
620 encode_sv (enc, sv);
621
622 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 }
630 else if (enc->json.flags & F_ALLOW_BLESSED)
631 encode_str (enc, "null", 4, 0);
632 else
633 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)",
634 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 827 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
635 }
636 } 828 }
637 else if (svt == SVt_PVHV) 829 else if (svt == SVt_PVHV)
638 encode_hv (enc, (HV *)sv); 830 encode_hv (enc, (HV *)sv);
639 else if (svt == SVt_PVAV) 831 else if (svt == SVt_PVAV)
640 encode_av (enc, (AV *)sv); 832 encode_av (enc, (AV *)sv);
641 else if (svt < SVt_PVAV) 833 else if (svt < SVt_PVAV)
642 { 834 {
643 STRLEN len = 0; 835 int bool_type = ref_bool_type (sv);
644 char *pv = svt ? SvPV (sv, len) : 0;
645 836
646 if (len == 1 && *pv == '1') 837 if (bool_type == 1)
647 encode_str (enc, "true", 4, 0); 838 encode_str (enc, "true", 4, 0);
648 else if (len == 1 && *pv == '0') 839 else if (bool_type == 0)
649 encode_str (enc, "false", 5, 0); 840 encode_str (enc, "false", 5, 0);
650 else if (enc->json.flags & F_ALLOW_UNKNOWN) 841 else if (enc->json.flags & F_ALLOW_UNKNOWN)
651 encode_str (enc, "null", 4, 0); 842 encode_str (enc, "null", 4, 0);
652 else 843 else
653 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",
716 } 907 }
717 else 908 else
718 { 909 {
719 // large integer, use the (rather slow) snprintf way. 910 // large integer, use the (rather slow) snprintf way.
720 need (enc, IVUV_MAXCHARS); 911 need (enc, IVUV_MAXCHARS);
721 enc->cur += 912 enc->cur +=
722 SvIsUV(sv) 913 SvIsUV(sv)
723 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 914 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
724 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 915 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
725 } 916 }
726 } 917 }
727 else if (SvROK (sv)) 918 else if (SvROK (sv))
728 encode_rv (enc, SvRV (sv)); 919 encode_rv (enc, SvRV (sv));
729 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 920 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
730 encode_str (enc, "null", 4, 0); 921 encode_str (enc, "null", 4, 0);
731 else 922 else
732 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",
733 SvPV_nolen (sv), SvFLAGS (sv)); 924 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
734} 925}
735 926
736static SV * 927static SV *
737encode_json (SV *scalar, JSON *json) 928encode_json (SV *scalar, JSON *json)
738{ 929{
739 enc_t enc; 930 enc_t enc;
740 931
741 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 932 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
742 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)");
743 934
744 enc.json = *json; 935 enc.json = *json;
745 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 936 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
746 enc.cur = SvPVX (enc.sv); 937 enc.cur = SvPVX (enc.sv);
750 : enc.json.flags & F_LATIN1 ? 0x000100UL 941 : enc.json.flags & F_LATIN1 ? 0x000100UL
751 : 0x110000UL; 942 : 0x110000UL;
752 943
753 SvPOK_only (enc.sv); 944 SvPOK_only (enc.sv);
754 encode_sv (&enc, scalar); 945 encode_sv (&enc, scalar);
946 encode_nl (&enc);
755 947
756 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 948 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
757 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 949 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
758 950
759 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 951 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
939 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1131 else if (expect_true (ch >= 0x20 && ch < 0x80))
940 *cur++ = ch; 1132 *cur++ = ch;
941 else if (ch >= 0x80) 1133 else if (ch >= 0x80)
942 { 1134 {
943 STRLEN clen; 1135 STRLEN clen;
944 UV uch;
945 1136
946 --dec_cur; 1137 --dec_cur;
947 1138
948 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1139 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
949 if (clen == (STRLEN)-1) 1140 if (clen == (STRLEN)-1)
950 ERR ("malformed UTF-8 character in JSON string"); 1141 ERR ("malformed UTF-8 character in JSON string");
951 1142
952 do 1143 do
953 *cur++ = *dec_cur++; 1144 *cur++ = *dec_cur++;
954 while (--clen); 1145 while (--clen);
955 1146
956 utf8 = 1; 1147 utf8 = 1;
957 } 1148 }
1149 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1150 *cur++ = ch;
958 else 1151 else
959 { 1152 {
960 --dec_cur; 1153 --dec_cur;
961 1154
962 if (!ch) 1155 if (!ch)
970 { 1163 {
971 STRLEN len = cur - buf; 1164 STRLEN len = cur - buf;
972 1165
973 if (sv) 1166 if (sv)
974 { 1167 {
975 SvGROW (sv, SvCUR (sv) + len + 1); 1168 STRLEN cur = SvCUR (sv);
1169
1170 if (SvLEN (sv) <= cur + len)
1171 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1172
976 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1173 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
977 SvCUR_set (sv, SvCUR (sv) + len); 1174 SvCUR_set (sv, SvCUR (sv) + len);
978 } 1175 }
979 else 1176 else
980 sv = newSVpvn (buf, len); 1177 sv = newSVpvn (buf, len);
1071 1268
1072 // special case the rather common 1..5-digit-int case 1269 // special case the rather common 1..5-digit-int case
1073 if (*start == '-') 1270 if (*start == '-')
1074 switch (len) 1271 switch (len)
1075 { 1272 {
1076 case 2: return newSViv (-( start [1] - '0' * 1)); 1273 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1077 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1274 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)); 1275 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)); 1276 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)); 1277 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1081 } 1278 }
1082 else 1279 else
1083 switch (len) 1280 switch (len)
1084 { 1281 {
1085 case 1: return newSViv ( start [0] - '0' * 1); 1282 case 1: return newSViv ( start [0] - '0' * 1);
1086 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1283 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); 1284 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); 1285 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); 1286 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1090 } 1287 }
1091 1288
1092 { 1289 {
1093 UV uv; 1290 UV uv;
1094 int numtype = grok_number (start, len, &uv); 1291 int numtype = grok_number (start, len, &uv);
1103 } 1300 }
1104 1301
1105 len -= *start == '-' ? 1 : 0; 1302 len -= *start == '-' ? 1 : 0;
1106 1303
1107 // does not fit into IV or UV, try NV 1304 // does not fit into IV or UV, try NV
1108 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1305 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 1306 // fits into NV without loss of precision
1114 return newSVnv (Atof (start)); 1307 return newSVnv (json_atof (start));
1115 1308
1116 // everything else fails, convert it to a string 1309 // everything else fails, convert it to a string
1117 return newSVpvn (start, dec->cur - start); 1310 return newSVpvn (start, dec->cur - start);
1118 } 1311 }
1119 1312
1120 // loss of precision here 1313 // loss of precision here
1121 return newSVnv (Atof (start)); 1314 return newSVnv (json_atof (start));
1122 1315
1123fail: 1316fail:
1124 return 0; 1317 return 0;
1125} 1318}
1126 1319
1150 if (*dec->cur == ']') 1343 if (*dec->cur == ']')
1151 { 1344 {
1152 ++dec->cur; 1345 ++dec->cur;
1153 break; 1346 break;
1154 } 1347 }
1155 1348
1156 if (*dec->cur != ',') 1349 if (*dec->cur != ',')
1157 ERR (", or ] expected while parsing array"); 1350 ERR (", or ] expected while parsing array");
1158 1351
1159 ++dec->cur; 1352 ++dec->cur;
1160 1353
1284 1477
1285 hv_iterinit (hv); 1478 hv_iterinit (hv);
1286 he = hv_iternext (hv); 1479 he = hv_iternext (hv);
1287 hv_iterinit (hv); 1480 hv_iterinit (hv);
1288 1481
1289 // the next line creates a mortal sv each time its called. 1482 // the next line creates a mortal sv each time it's called.
1290 // might want to optimise this for common cases. 1483 // might want to optimise this for common cases.
1291 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);
1292 1485
1293 if (cb) 1486 if (cb)
1294 { 1487 {
1295 dSP; 1488 dSP;
1296 int count; 1489 int count;
1297 1490
1298 ENTER; SAVETMPS; PUSHMARK (SP); 1491 ENTER; SAVETMPS;
1492 SAVESTACK_POS ();
1493 PUSHMARK (SP);
1299 XPUSHs (HeVAL (he)); 1494 XPUSHs (HeVAL (he));
1495 sv_2mortal (sv);
1300 1496
1301 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1497 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1302 1498
1303 if (count == 1) 1499 if (count == 1)
1304 { 1500 {
1305 sv = newSVsv (POPs); 1501 sv = newSVsv (POPs);
1306 FREETMPS; LEAVE; 1502 FREETMPS; LEAVE;
1307 return sv; 1503 return sv;
1308 } 1504 }
1309 1505
1506 SvREFCNT_inc (sv);
1310 FREETMPS; LEAVE; 1507 FREETMPS; LEAVE;
1311 } 1508 }
1312 } 1509 }
1313 1510
1314 if (dec->json.cb_object) 1511 if (dec->json.cb_object)
1315 { 1512 {
1316 dSP; 1513 dSP;
1317 int count; 1514 int count;
1318 1515
1319 ENTER; SAVETMPS; PUSHMARK (SP); 1516 ENTER; SAVETMPS;
1517 SAVESTACK_POS ();
1518 PUSHMARK (SP);
1320 XPUSHs (sv_2mortal (sv)); 1519 XPUSHs (sv_2mortal (sv));
1321 1520
1322 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1521 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1323 1522
1324 if (count == 1) 1523 if (count == 1)
1340 DEC_DEC_DEPTH; 1539 DEC_DEC_DEPTH;
1341 return 0; 1540 return 0;
1342} 1541}
1343 1542
1344static 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 *
1345decode_sv (dec_t *dec) 1627decode_sv (dec_t *dec)
1346{ 1628{
1347 // the beauty of JSON: you need exactly one character lookahead 1629 // the beauty of JSON: you need exactly one character lookahead
1348 // to parse everything. 1630 // to parse everything.
1349 switch (*dec->cur) 1631 switch (*dec->cur)
1350 { 1632 {
1351 case '"': ++dec->cur; return decode_str (dec); 1633 case '"': ++dec->cur; return decode_str (dec);
1352 case '[': ++dec->cur; return decode_av (dec); 1634 case '[': ++dec->cur; return decode_av (dec);
1353 case '{': ++dec->cur; return decode_hv (dec); 1635 case '{': ++dec->cur; return decode_hv (dec);
1636 case '(': return decode_tag (dec);
1354 1637
1355 case '-': 1638 case '-':
1356 case '0': case '1': case '2': case '3': case '4': 1639 case '0': case '1': case '2': case '3': case '4':
1357 case '5': case '6': case '7': case '8': case '9': 1640 case '5': case '6': case '7': case '8': case '9':
1358 return decode_num (dec); 1641 return decode_num (dec);
1360 case 't': 1643 case 't':
1361 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1644 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1362 { 1645 {
1363 dec->cur += 4; 1646 dec->cur += 4;
1364#if JSON_SLOW 1647#if JSON_SLOW
1365 json_true = get_bool ("JSON::XS::true"); 1648 bool_true = get_bool ("Types::Serialiser::true");
1366#endif 1649#endif
1367 return newSVsv (json_true); 1650 return newSVsv (bool_true);
1368 } 1651 }
1369 else 1652 else
1370 ERR ("'true' expected"); 1653 ERR ("'true' expected");
1371 1654
1372 break; 1655 break;
1374 case 'f': 1657 case 'f':
1375 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1658 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1376 { 1659 {
1377 dec->cur += 5; 1660 dec->cur += 5;
1378#if JSON_SLOW 1661#if JSON_SLOW
1379 json_false = get_bool ("JSON::XS::false"); 1662 bool_false = get_bool ("Types::Serialiser::false");
1380#endif 1663#endif
1381 return newSVsv (json_false); 1664 return newSVsv (bool_false);
1382 } 1665 }
1383 else 1666 else
1384 ERR ("'false' expected"); 1667 ERR ("'false' expected");
1385 1668
1386 break; 1669 break;
1395 ERR ("'null' expected"); 1678 ERR ("'null' expected");
1396 1679
1397 break; 1680 break;
1398 1681
1399 default: 1682 default:
1400 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1683 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1401 break; 1684 break;
1402 } 1685 }
1403 1686
1404fail: 1687fail:
1405 return 0; 1688 return 0;
1406} 1689}
1407 1690
1408static SV * 1691static SV *
1409decode_json (SV *string, JSON *json, STRLEN *offset_return) 1692decode_json (SV *string, JSON *json, char **offset_return)
1410{ 1693{
1411 dec_t dec; 1694 dec_t dec;
1412 STRLEN offset;
1413 SV *sv; 1695 SV *sv;
1414 1696
1697 /* work around bugs in 5.10 where manipulating magic values
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 */
1415 SvGETMAGIC (string); 1702 /*SvGETMAGIC (string);*/
1703 if (SvMAGICAL (string) || !SvPOK (string))
1704 string = sv_2mortal (newSVsv (string));
1705
1416 SvUPGRADE (string, SVt_PV); 1706 SvUPGRADE (string, SVt_PV);
1417 1707
1708 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1709 * assertion with -DDEBUGGING, although SvCUR is documented to
1710 * return the xpv_cur field which certainly exists after upgrading.
1711 * according to nicholas clark, calling SvPOK fixes this.
1712 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1713 * and hope for the best.
1714 * Damnit, SvPV_nolen still trips over yet another assertion. This
1715 * assertion business is seriously broken, try yet another workaround
1716 * for the broken -DDEBUGGING.
1717 */
1718 {
1719#ifdef DEBUGGING
1720 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1721#else
1722 STRLEN offset = SvCUR (string);
1723#endif
1724
1418 if (SvCUR (string) > json->max_size && json->max_size) 1725 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", 1726 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); 1727 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1728 }
1421 1729
1422 if (json->flags & F_UTF8) 1730 if (DECODE_WANTS_OCTETS (json))
1423 sv_utf8_downgrade (string, 0); 1731 sv_utf8_downgrade (string, 0);
1424 else 1732 else
1425 sv_utf8_upgrade (string); 1733 sv_utf8_upgrade (string);
1426 1734
1427 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1735 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1437 1745
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1746 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439 1747
1440 decode_ws (&dec); 1748 decode_ws (&dec);
1441 sv = decode_sv (&dec); 1749 sv = decode_sv (&dec);
1750
1751 if (offset_return)
1752 *offset_return = dec.cur;
1442 1753
1443 if (!(offset_return || !sv)) 1754 if (!(offset_return || !sv))
1444 { 1755 {
1445 // check for trailing garbage 1756 // check for trailing garbage
1446 decode_ws (&dec); 1757 decode_ws (&dec);
1449 { 1760 {
1450 dec.err = "garbage after JSON object"; 1761 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv); 1762 SvREFCNT_dec (sv);
1452 sv = 0; 1763 sv = 0;
1453 } 1764 }
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 } 1765 }
1465 1766
1466 if (!sv) 1767 if (!sv)
1467 { 1768 {
1468 SV *uni = sv_newmortal (); 1769 SV *uni = sv_newmortal ();
1474 SAVEVPTR (PL_curcop); 1775 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop; 1776 PL_curcop = &cop;
1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1777 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE; 1778 LEAVE;
1478 1779
1479 croak ("%s, at character offset %d [\"%s\"]", 1780 croak ("%s, at character offset %d (before \"%s\")",
1480 dec.err, 1781 dec.err,
1481 (int)offset, 1782 (int)ptr_to_index (string, dec.cur),
1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1783 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1483 } 1784 }
1484 1785
1485 sv = sv_2mortal (sv); 1786 sv = sv_2mortal (sv);
1486 1787
1487 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1788 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1488 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1789 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1489 1790
1490 return sv; 1791 return sv;
1491} 1792}
1492 1793
1495 1796
1496static void 1797static void
1497incr_parse (JSON *self) 1798incr_parse (JSON *self)
1498{ 1799{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1800 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1801
1802 // the state machine here is a bit convoluted and could be simplified a lot
1803 // but this would make it slower, so...
1500 1804
1501 for (;;) 1805 for (;;)
1502 { 1806 {
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 1807 //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) 1808 switch (self->incr_mode)
1505 { 1809 {
1506 // only used for intiial whitespace skipping 1810 // only used for initial whitespace skipping
1507 case INCR_M_WS: 1811 case INCR_M_WS:
1508 for (;;) 1812 for (;;)
1509 { 1813 {
1510 if (*p > 0x20) 1814 if (*p > 0x20)
1511 { 1815 {
1816 if (*p == '#')
1817 {
1818 self->incr_mode = INCR_M_C0;
1819 goto incr_m_c;
1820 }
1821 else
1822 {
1512 self->incr_mode = INCR_M_JSON; 1823 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json; 1824 goto incr_m_json;
1825 }
1514 } 1826 }
1515 else if (!*p) 1827 else if (!*p)
1516 goto interrupt; 1828 goto interrupt;
1517 1829
1518 ++p; 1830 ++p;
1524 goto interrupt; 1836 goto interrupt;
1525 1837
1526 ++p; 1838 ++p;
1527 self->incr_mode = INCR_M_STR; 1839 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str; 1840 goto incr_m_str;
1841
1842 // inside #-style comments
1843 case INCR_M_C0:
1844 case INCR_M_C1:
1845 incr_m_c:
1846 for (;;)
1847 {
1848 if (*p == '\n')
1849 {
1850 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1851 break;
1852 }
1853 else if (!*p)
1854 goto interrupt;
1855
1856 ++p;
1857 }
1858
1859 break;
1529 1860
1530 // inside a string 1861 // inside a string
1531 case INCR_M_STR: 1862 case INCR_M_STR:
1532 incr_m_str: 1863 incr_m_str:
1533 for (;;) 1864 for (;;)
1584 self->incr_mode = INCR_M_STR; 1915 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str; 1916 goto incr_m_str;
1586 1917
1587 case '[': 1918 case '[':
1588 case '{': 1919 case '{':
1920 case '(':
1589 if (++self->incr_nest > self->max_depth) 1921 if (++self->incr_nest > self->max_depth)
1590 croak (ERR_NESTING_EXCEEDED); 1922 croak (ERR_NESTING_EXCEEDED);
1591 break; 1923 break;
1592 1924
1593 case ']': 1925 case ']':
1594 case '}': 1926 case '}':
1595 if (!--self->incr_nest) 1927 if (--self->incr_nest <= 0)
1596 goto interrupt; 1928 goto interrupt;
1929 break;
1930
1931 case ')':
1932 --self->incr_nest;
1933 break;
1934
1935 case '#':
1936 self->incr_mode = INCR_M_C1;
1937 goto incr_m_c;
1597 } 1938 }
1598 } 1939 }
1599 } 1940 }
1600 1941
1601 modechange: 1942 modechange:
1602 ; 1943 ;
1603 } 1944 }
1604 1945
1605interrupt: 1946interrupt:
1606 self->incr_pos = p - SvPVX (self->incr_text); 1947 self->incr_pos = p - SvPVX (self->incr_text);
1948 //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 1949 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1608} 1950}
1609 1951
1610///////////////////////////////////////////////////////////////////////////// 1952/////////////////////////////////////////////////////////////////////////////
1611// XS interface functions 1953// XS interface functions
1621 i >= '0' && i <= '9' ? i - '0' 1963 i >= '0' && i <= '9' ? i - '0'
1622 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1964 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1623 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1965 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1624 : -1; 1966 : -1;
1625 1967
1626 json_stash = gv_stashpv ("JSON::XS" , 1); 1968 json_stash = gv_stashpv ("JSON::XS" , 1);
1627 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1969 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1628
1629 json_true = get_bool ("JSON::XS::true"); 1970 bool_true = get_bool ("Types::Serialiser::true");
1630 json_false = get_bool ("JSON::XS::false"); 1971 bool_false = get_bool ("Types::Serialiser::false");
1972
1973 sv_json = newSVpv ("JSON", 0);
1974 SvREADONLY_on (sv_json);
1975
1976 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1631} 1977}
1632 1978
1633PROTOTYPES: DISABLE 1979PROTOTYPES: DISABLE
1634 1980
1635void CLONE (...) 1981void CLONE (...)
1636 CODE: 1982 CODE:
1637 json_stash = 0; 1983 json_stash = 0;
1638 json_boolean_stash = 0; 1984 bool_stash = 0;
1639 1985
1640void new (char *klass) 1986void new (char *klass)
1641 PPCODE: 1987 PPCODE:
1642{ 1988{
1643 SV *pv = NEWSV (0, sizeof (JSON)); 1989 SV *pv = NEWSV (0, sizeof (JSON));
1644 SvPOK_only (pv); 1990 SvPOK_only (pv);
1645 json_init ((JSON *)SvPVX (pv)); 1991 json_init ((JSON *)SvPVX (pv));
1646 XPUSHs (sv_2mortal (sv_bless ( 1992 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv), 1993 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1994 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1663 shrink = F_SHRINK 2009 shrink = F_SHRINK
1664 allow_blessed = F_ALLOW_BLESSED 2010 allow_blessed = F_ALLOW_BLESSED
1665 convert_blessed = F_CONV_BLESSED 2011 convert_blessed = F_CONV_BLESSED
1666 relaxed = F_RELAXED 2012 relaxed = F_RELAXED
1667 allow_unknown = F_ALLOW_UNKNOWN 2013 allow_unknown = F_ALLOW_UNKNOWN
2014 allow_tags = F_ALLOW_TAGS
1668 PPCODE: 2015 PPCODE:
1669{ 2016{
1670 if (enable) 2017 if (enable)
1671 self->flags |= ix; 2018 self->flags |= ix;
1672 else 2019 else
1688 get_shrink = F_SHRINK 2035 get_shrink = F_SHRINK
1689 get_allow_blessed = F_ALLOW_BLESSED 2036 get_allow_blessed = F_ALLOW_BLESSED
1690 get_convert_blessed = F_CONV_BLESSED 2037 get_convert_blessed = F_CONV_BLESSED
1691 get_relaxed = F_RELAXED 2038 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN 2039 get_allow_unknown = F_ALLOW_UNKNOWN
2040 get_allow_tags = F_ALLOW_TAGS
1693 PPCODE: 2041 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix)); 2042 XPUSHs (boolSV (self->flags & ix));
1695 2043
1696void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2044void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1697 PPCODE: 2045 PPCODE:
1725} 2073}
1726 2074
1727void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2075void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1728 PPCODE: 2076 PPCODE:
1729{ 2077{
1730 if (!self->cb_sk_object) 2078 if (!self->cb_sk_object)
1731 self->cb_sk_object = newHV (); 2079 self->cb_sk_object = newHV ();
1732 2080
1733 if (SvOK (cb)) 2081 if (SvOK (cb))
1734 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2082 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1735 else 2083 else
1746 XPUSHs (ST (0)); 2094 XPUSHs (ST (0));
1747} 2095}
1748 2096
1749void encode (JSON *self, SV *scalar) 2097void encode (JSON *self, SV *scalar)
1750 PPCODE: 2098 PPCODE:
1751 XPUSHs (encode_json (scalar, self)); 2099 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2100 XPUSHs (scalar);
1752 2101
1753void decode (JSON *self, SV *jsonstr) 2102void decode (JSON *self, SV *jsonstr)
1754 PPCODE: 2103 PPCODE:
1755 XPUSHs (decode_json (jsonstr, self, 0)); 2104 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2105 XPUSHs (jsonstr);
1756 2106
1757void decode_prefix (JSON *self, SV *jsonstr) 2107void decode_prefix (JSON *self, SV *jsonstr)
1758 PPCODE: 2108 PPCODE:
1759{ 2109{
1760 STRLEN offset; 2110 SV *sv;
2111 char *offset;
2112 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1761 EXTEND (SP, 2); 2113 EXTEND (SP, 2);
1762 PUSHs (decode_json (jsonstr, self, &offset)); 2114 PUSHs (sv);
1763 PUSHs (sv_2mortal (newSVuv (offset))); 2115 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1764} 2116}
1765 2117
1766void incr_parse (JSON *self, SV *jsonstr = 0) 2118void incr_parse (JSON *self, SV *jsonstr = 0)
1767 PPCODE: 2119 PPCODE:
1768{ 2120{
1769 if (!self->incr_text) 2121 if (!self->incr_text)
1770 self->incr_text = newSVpvn ("", 0); 2122 self->incr_text = newSVpvn ("", 0);
2123
2124 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2125 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2126 if (DECODE_WANTS_OCTETS (self))
2127 {
2128 if (self->incr_pos)
2129 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2130 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2131
2132 sv_utf8_downgrade (self->incr_text, 0);
2133 }
2134 else
2135 {
2136 sv_utf8_upgrade (self->incr_text);
2137
2138 if (self->incr_pos)
2139 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2140 - (U8 *)SvPVX (self->incr_text);
2141 }
1771 2142
1772 // append data, if any 2143 // append data, if any
1773 if (jsonstr) 2144 if (jsonstr)
1774 { 2145 {
2146 /* make sure both strings have same encoding */
1775 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2147 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2148 if (SvUTF8 (jsonstr))
2149 sv_utf8_downgrade (jsonstr, 0);
1776 { 2150 else
1777 /* utf-8-ness differs, need to upgrade */
1778 sv_utf8_upgrade (self->incr_text); 2151 sv_utf8_upgrade (jsonstr);
1779 2152
1780 if (self->incr_pos) 2153 /* 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 { 2154 {
1786 STRLEN len; 2155 STRLEN len;
1787 const char *str = SvPV (jsonstr, len); 2156 const char *str = SvPV (jsonstr, len);
1788 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2157 STRLEN cur = SvCUR (self->incr_text);
2158
2159 if (SvLEN (self->incr_text) <= cur + len)
2160 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2161
1789 Move (str, SvEND (self->incr_text), len, char); 2162 Move (str, SvEND (self->incr_text), len, char);
1790 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2163 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 2164 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1792 } 2165 }
1793 } 2166 }
1794 2167
1795 if (GIMME_V != G_VOID) 2168 if (GIMME_V != G_VOID)
1796 do 2169 do
1797 { 2170 {
2171 SV *sv;
1798 STRLEN offset; 2172 char *offset;
1799 2173
1800 if (!INCR_DONE (self)) 2174 if (!INCR_DONE (self))
1801 { 2175 {
1802 incr_parse (self); 2176 incr_parse (self);
1803 2177
1804 if (self->incr_pos > self->max_size && self->max_size) 2178 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", 2179 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); 2180 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1807 2181
1808 if (!INCR_DONE (self)) 2182 if (!INCR_DONE (self))
2183 {
2184 // as an optimisation, do not accumulate white space in the incr buffer
2185 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2186 {
2187 self->incr_pos = 0;
2188 SvCUR_set (self->incr_text, 0);
2189 }
2190
1809 break; 2191 break;
2192 }
1810 } 2193 }
1811 2194
1812 XPUSHs (decode_json (self->incr_text, self, &offset)); 2195 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2196 XPUSHs (sv);
1813 2197
1814 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1815 self->incr_pos -= offset; 2198 self->incr_pos -= offset - SvPVX (self->incr_text);
1816 self->incr_nest = 0; 2199 self->incr_nest = 0;
1817 self->incr_mode = 0; 2200 self->incr_mode = 0;
2201
2202 sv_chop (self->incr_text, offset);
1818 } 2203 }
1819 while (GIMME_V == G_ARRAY); 2204 while (GIMME_V == G_ARRAY);
1820} 2205}
1821 2206
1822SV *incr_text (JSON *self) 2207SV *incr_text (JSON *self)
1860 SvREFCNT_dec (self->incr_text); 2245 SvREFCNT_dec (self->incr_text);
1861 2246
1862PROTOTYPES: ENABLE 2247PROTOTYPES: ENABLE
1863 2248
1864void encode_json (SV *scalar) 2249void encode_json (SV *scalar)
1865 ALIAS:
1866 to_json_ = 0
1867 encode_json = F_UTF8
1868 PPCODE: 2250 PPCODE:
1869{ 2251{
1870 JSON json; 2252 JSON json;
1871 json_init (&json); 2253 json_init (&json);
1872 json.flags |= ix; 2254 json.flags |= F_UTF8;
1873 XPUSHs (encode_json (scalar, &json)); 2255 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2256 XPUSHs (scalar);
1874} 2257}
1875 2258
1876void decode_json (SV *jsonstr) 2259void decode_json (SV *jsonstr)
1877 ALIAS:
1878 from_json_ = 0
1879 decode_json = F_UTF8
1880 PPCODE: 2260 PPCODE:
1881{ 2261{
1882 JSON json; 2262 JSON json;
1883 json_init (&json); 2263 json_init (&json);
1884 json.flags |= ix; 2264 json.flags |= F_UTF8;
1885 XPUSHs (decode_json (jsonstr, &json, 0)); 2265 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2266 XPUSHs (jsonstr);
1886} 2267}
1887 2268
1888

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines