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 1.89 by root, Sat Jul 19 04:21:32 2008 UTC vs.
Revision 1.130 by root, Sat Nov 26 06:09:29 2016 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 15#endif
15 16
16// some old perls do not have this, try to make it work, no 17// 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. 18// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
20#endif 21#endif
21 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 36
24#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
32#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
36#define F_ALLOW_UNKNOWN 0x00002000UL 49#define F_ALLOW_UNKNOWN 0x00002000UL
50#define F_ALLOW_TAGS 0x00004000UL
37#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
38 52
39#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
40 54
41#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 32 // initial scalar size to be allocated
42#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
43 57
44#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
59
60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
45 61
46#define SB do { 62#define SB do {
47#define SE } while (0) 63#define SE } while (0)
48 64
49#if __GNUC__ >= 3 65#if __GNUC__ >= 3
64#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" 80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
65 81
66#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
67# define JSON_SLOW 1 83# define JSON_SLOW 1
68# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 84# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
85# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
69#else 86#else
70# define JSON_SLOW 0 87# define JSON_SLOW 0
71# define JSON_STASH json_stash 88# define JSON_STASH json_stash
89# define BOOL_STASH bool_stash
72#endif 90#endif
73 91
74static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
75static SV *json_true, *json_false; 93#define STACK_HES 64
94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_true, *bool_false, *sv_json;
76 97
77enum { 98enum {
78 INCR_M_WS = 0, // initial whitespace skipping, must be 0 99 INCR_M_WS = 0, // initial whitespace skipping, must be 0
79 INCR_M_STR, // inside string 100 INCR_M_STR, // inside string
80 INCR_M_BS, // inside backslash 101 INCR_M_BS, // inside backslash
102 INCR_M_C0, // inside comment in initial whitespace sequence
103 INCR_M_C1, // inside comment in other places
81 INCR_M_JSON // outside anything, count nesting 104 INCR_M_JSON // outside anything, count nesting
82}; 105};
83 106
84#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) 107#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
85 108
175 *s++ = 0x80 | ( ch & 0x3f); 198 *s++ = 0x80 | ( ch & 0x3f);
176 199
177 return s; 200 return s;
178} 201}
179 202
203// convert offset pointer to character index, sv must be string
204static STRLEN
205ptr_to_index (SV *sv, char *offset)
206{
207 return SvUTF8 (sv)
208 ? utf8_distance (offset, SvPVX (sv))
209 : offset - SvPVX (sv);
210}
211
212/////////////////////////////////////////////////////////////////////////////
213// fp hell
214
215// scan a group of digits, and a trailing exponent
216static void
217json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
218{
219 UV uaccum = 0;
220 int eaccum = 0;
221
222 // if we recurse too deep, skip all remaining digits
223 // to avoid a stack overflow attack
224 if (expect_false (--maxdepth <= 0))
225 while (((U8)*s - '0') < 10)
226 ++s;
227
228 for (;;)
229 {
230 U8 dig = (U8)*s - '0';
231
232 if (expect_false (dig >= 10))
233 {
234 if (dig == (U8)((U8)'.' - (U8)'0'))
235 {
236 ++s;
237 json_atof_scan1 (s, accum, expo, 1, maxdepth);
238 }
239 else if ((dig | ' ') == 'e' - '0')
240 {
241 int exp2 = 0;
242 int neg = 0;
243
244 ++s;
245
246 if (*s == '-')
247 {
248 ++s;
249 neg = 1;
250 }
251 else if (*s == '+')
252 ++s;
253
254 while ((dig = (U8)*s - '0') < 10)
255 exp2 = exp2 * 10 + *s++ - '0';
256
257 *expo += neg ? -exp2 : exp2;
258 }
259
260 break;
261 }
262
263 ++s;
264
265 uaccum = uaccum * 10 + dig;
266 ++eaccum;
267
268 // if we have too many digits, then recurse for more
269 // we actually do this for rather few digits
270 if (uaccum >= (UV_MAX - 9) / 10)
271 {
272 if (postdp) *expo -= eaccum;
273 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
274 if (postdp) *expo += eaccum;
275
276 break;
277 }
278 }
279
280 // this relies greatly on the quality of the pow ()
281 // implementation of the platform, but a good
282 // implementation is hard to beat.
283 // (IEEE 754 conformant ones are required to be exact)
284 if (postdp) *expo -= eaccum;
285 *accum += uaccum * Perl_pow (10., *expo);
286 *expo += eaccum;
287}
288
289static NV
290json_atof (const char *s)
291{
292 NV accum = 0.;
293 int expo = 0;
294 int neg = 0;
295
296 if (*s == '-')
297 {
298 ++s;
299 neg = 1;
300 }
301
302 // a recursion depth of ten gives us >>500 bits
303 json_atof_scan1 (s, &accum, &expo, 0, 10);
304
305 return neg ? -accum : accum;
306}
307
308// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
309static int
310ref_bool_type (SV *sv)
311{
312 svtype svt = SvTYPE (sv);
313
314 if (svt < SVt_PVAV)
315 {
316 STRLEN len = 0;
317 char *pv = svt ? SvPV (sv, len) : 0;
318
319 if (len == 1)
320 if (*pv == '1')
321 return 1;
322 else if (*pv == '0')
323 return 0;
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 (SvTYPE (scalar) >= SVt_PVMG)
339 {
340 if (SvSTASH (scalar) == bool_stash)
341 return 1;
342
343 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
344 return 1;
345 }
346
347 return 0;
348}
349
180///////////////////////////////////////////////////////////////////////////// 350/////////////////////////////////////////////////////////////////////////////
181// encoder 351// encoder
182 352
183// structure used for encoding JSON 353// structure used for encoding JSON
184typedef struct 354typedef struct
192} enc_t; 362} enc_t;
193 363
194INLINE void 364INLINE void
195need (enc_t *enc, STRLEN len) 365need (enc_t *enc, STRLEN len)
196{ 366{
197 if (expect_false (enc->cur + len >= enc->end)) 367 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
198 { 368 {
199 STRLEN cur = enc->cur - SvPVX (enc->sv); 369 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
200 SvGROW (enc->sv, cur + len + 1); 370 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
201 enc->cur = SvPVX (enc->sv) + cur; 371 enc->cur = SvPVX (enc->sv) + cur;
202 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 372 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
203 } 373 }
204} 374}
205 375
280 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 450 (int)((uch - 0x10000) % 0x400 + 0xDC00));
281 enc->cur += 12; 451 enc->cur += 12;
282 } 452 }
283 else 453 else
284 { 454 {
285 static char hexdigit [16] = "0123456789abcdef";
286 need (enc, len += 5); 455 need (enc, len += 5);
287 *enc->cur++ = '\\'; 456 *enc->cur++ = '\\';
288 *enc->cur++ = 'u'; 457 *enc->cur++ = 'u';
289 *enc->cur++ = hexdigit [ uch >> 12 ]; 458 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
290 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 459 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
291 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 460 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
292 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 461 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
293 } 462 }
294 463
295 str += clen; 464 str += clen;
296 } 465 }
297 else if (enc->json.flags & F_LATIN1) 466 else if (enc->json.flags & F_LATIN1)
372 541
373 if (enc->indent >= enc->json.max_depth) 542 if (enc->indent >= enc->json.max_depth)
374 croak (ERR_NESTING_EXCEEDED); 543 croak (ERR_NESTING_EXCEEDED);
375 544
376 encode_ch (enc, '['); 545 encode_ch (enc, '[');
377 546
378 if (len >= 0) 547 if (len >= 0)
379 { 548 {
380 encode_nl (enc); ++enc->indent; 549 encode_nl (enc); ++enc->indent;
381 550
382 for (i = 0; i <= len; ++i) 551 for (i = 0; i <= len; ++i)
394 encode_comma (enc); 563 encode_comma (enc);
395 } 564 }
396 565
397 encode_nl (enc); --enc->indent; encode_indent (enc); 566 encode_nl (enc); --enc->indent; encode_indent (enc);
398 } 567 }
399 568
400 encode_ch (enc, ']'); 569 encode_ch (enc, ']');
401} 570}
402 571
403static void 572static void
404encode_hk (enc_t *enc, HE *he) 573encode_hk (enc_t *enc, HE *he)
408 if (HeKLEN (he) == HEf_SVKEY) 577 if (HeKLEN (he) == HEf_SVKEY)
409 { 578 {
410 SV *sv = HeSVKEY (he); 579 SV *sv = HeSVKEY (he);
411 STRLEN len; 580 STRLEN len;
412 char *str; 581 char *str;
413 582
414 SvGETMAGIC (sv); 583 SvGETMAGIC (sv);
415 str = SvPV (sv, len); 584 str = SvPV (sv, len);
416 585
417 encode_str (enc, str, len, SvUTF8 (sv)); 586 encode_str (enc, str, len, SvUTF8 (sv));
418 } 587 }
461 630
462 encode_ch (enc, '{'); 631 encode_ch (enc, '{');
463 632
464 // for canonical output we have to sort by keys first 633 // for canonical output we have to sort by keys first
465 // actually, this is mostly due to the stupid so-called 634 // actually, this is mostly due to the stupid so-called
466 // security workaround added somewhere in 5.8.x. 635 // security workaround added somewhere in 5.8.x
467 // that randomises hash orderings 636 // that randomises hash orderings
468 if (enc->json.flags & F_CANONICAL) 637 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
469 { 638 {
470 int count = hv_iterinit (hv); 639 int count = hv_iterinit (hv);
471 640
472 if (SvMAGICAL (hv)) 641 if (SvMAGICAL (hv))
473 { 642 {
484 } 653 }
485 654
486 if (count) 655 if (count)
487 { 656 {
488 int i, fast = 1; 657 int i, fast = 1;
489#if defined(__BORLANDC__) || defined(_MSC_VER) 658 HE *hes_stack [STACK_HES];
490 HE **hes = _alloca (count * sizeof (HE)); 659 HE **hes = hes_stack;
491#else 660
492 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 661 // allocate larger arrays on the heap
493#endif 662 if (count > STACK_HES)
663 {
664 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
665 hes = (HE **)SvPVX (sv);
666 }
494 667
495 i = 0; 668 i = 0;
496 while ((he = hv_iternext (hv))) 669 while ((he = hv_iternext (hv)))
497 { 670 {
498 hes [i++] = he; 671 hes [i++] = he;
567// encode objects, arrays and special \0=false and \1=true values. 740// encode objects, arrays and special \0=false and \1=true values.
568static void 741static void
569encode_rv (enc_t *enc, SV *sv) 742encode_rv (enc_t *enc, SV *sv)
570{ 743{
571 svtype svt; 744 svtype svt;
745 GV *method;
572 746
573 SvGETMAGIC (sv); 747 SvGETMAGIC (sv);
574 svt = SvTYPE (sv); 748 svt = SvTYPE (sv);
575 749
576 if (expect_false (SvOBJECT (sv))) 750 if (expect_false (SvOBJECT (sv)))
577 { 751 {
578 HV *stash = !JSON_SLOW || json_boolean_stash 752 HV *stash = SvSTASH (sv);
579 ? json_boolean_stash
580 : gv_stashpv ("JSON::XS::Boolean", 1);
581 753
582 if (SvSTASH (sv) == stash) 754 if (stash == bool_stash)
583 { 755 {
584 if (SvIV (sv)) 756 if (SvIV (sv))
585 encode_str (enc, "true", 4, 0); 757 encode_str (enc, "true", 4, 0);
586 else 758 else
587 encode_str (enc, "false", 5, 0); 759 encode_str (enc, "false", 5, 0);
588 } 760 }
761 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
762 {
763 int count;
764 dSP;
765
766 ENTER; SAVETMPS;
767 SAVESTACK_POS ();
768 PUSHMARK (SP);
769 EXTEND (SP, 2);
770 // we re-bless the reference to get overload and other niceties right
771 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
772 PUSHs (sv_json);
773
774 PUTBACK;
775 count = call_sv ((SV *)GvCV (method), G_ARRAY);
776 SPAGAIN;
777
778 // catch this surprisingly common error
779 if (SvROK (TOPs) && SvRV (TOPs) == sv)
780 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
781
782 encode_ch (enc, '(');
783 encode_ch (enc, '"');
784 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
785 encode_ch (enc, '"');
786 encode_ch (enc, ')');
787 encode_ch (enc, '[');
788
789 while (count)
790 {
791 encode_sv (enc, SP[1 - count--]);
792
793 if (count)
794 encode_ch (enc, ',');
795 }
796
797 encode_ch (enc, ']');
798
799 FREETMPS; LEAVE;
800 }
801 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
802 {
803 dSP;
804
805 ENTER; SAVETMPS;
806 PUSHMARK (SP);
807 // we re-bless the reference to get overload and other niceties right
808 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
809
810 // calling with G_SCALAR ensures that we always get a 1 return value
811 PUTBACK;
812 call_sv ((SV *)GvCV (method), G_SCALAR);
813 SPAGAIN;
814
815 // catch this surprisingly common error
816 if (SvROK (TOPs) && SvRV (TOPs) == sv)
817 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
818
819 sv = POPs;
820 PUTBACK;
821
822 encode_sv (enc, sv);
823
824 FREETMPS; LEAVE;
825 }
826 else if (enc->json.flags & F_ALLOW_BLESSED)
827 encode_str (enc, "null", 4, 0);
589 else 828 else
590 {
591#if 0
592 if (0 && sv_derived_from (rv, "JSON::Literal"))
593 {
594 // not yet
595 }
596#endif
597 if (enc->json.flags & F_CONV_BLESSED)
598 {
599 // we re-bless the reference to get overload and other niceties right
600 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
601
602 if (to_json)
603 {
604 dSP;
605
606 ENTER; SAVETMPS; PUSHMARK (SP);
607 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
608
609 // calling with G_SCALAR ensures that we always get a 1 return value
610 PUTBACK;
611 call_sv ((SV *)GvCV (to_json), G_SCALAR);
612 SPAGAIN;
613
614 // catch this surprisingly common error
615 if (SvROK (TOPs) && SvRV (TOPs) == sv)
616 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
617
618 sv = POPs;
619 PUTBACK;
620
621 encode_sv (enc, sv);
622
623 FREETMPS; LEAVE;
624 }
625 else if (enc->json.flags & F_ALLOW_BLESSED)
626 encode_str (enc, "null", 4, 0);
627 else
628 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
629 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
630 }
631 else if (enc->json.flags & F_ALLOW_BLESSED)
632 encode_str (enc, "null", 4, 0);
633 else
634 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 829 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
635 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 830 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
636 }
637 } 831 }
638 else if (svt == SVt_PVHV) 832 else if (svt == SVt_PVHV)
639 encode_hv (enc, (HV *)sv); 833 encode_hv (enc, (HV *)sv);
640 else if (svt == SVt_PVAV) 834 else if (svt == SVt_PVAV)
641 encode_av (enc, (AV *)sv); 835 encode_av (enc, (AV *)sv);
642 else if (svt < SVt_PVAV) 836 else if (svt < SVt_PVAV)
643 { 837 {
644 STRLEN len = 0; 838 int bool_type = ref_bool_type (sv);
645 char *pv = svt ? SvPV (sv, len) : 0;
646 839
647 if (len == 1 && *pv == '1') 840 if (bool_type == 1)
648 encode_str (enc, "true", 4, 0); 841 encode_str (enc, "true", 4, 0);
649 else if (len == 1 && *pv == '0') 842 else if (bool_type == 0)
650 encode_str (enc, "false", 5, 0); 843 encode_str (enc, "false", 5, 0);
651 else if (enc->json.flags & F_ALLOW_UNKNOWN) 844 else if (enc->json.flags & F_ALLOW_UNKNOWN)
652 encode_str (enc, "null", 4, 0); 845 encode_str (enc, "null", 4, 0);
653 else 846 else
654 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 847 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
717 } 910 }
718 else 911 else
719 { 912 {
720 // large integer, use the (rather slow) snprintf way. 913 // large integer, use the (rather slow) snprintf way.
721 need (enc, IVUV_MAXCHARS); 914 need (enc, IVUV_MAXCHARS);
722 enc->cur += 915 enc->cur +=
723 SvIsUV(sv) 916 SvIsUV(sv)
724 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 917 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
725 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 918 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
726 } 919 }
727 } 920 }
728 else if (SvROK (sv)) 921 else if (SvROK (sv))
729 encode_rv (enc, SvRV (sv)); 922 encode_rv (enc, SvRV (sv));
730 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 923 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
731 encode_str (enc, "null", 4, 0); 924 encode_str (enc, "null", 4, 0);
732 else 925 else
733 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 926 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
734 SvPV_nolen (sv), SvFLAGS (sv)); 927 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
735} 928}
736 929
737static SV * 930static SV *
738encode_json (SV *scalar, JSON *json) 931encode_json (SV *scalar, JSON *json)
739{ 932{
740 enc_t enc; 933 enc_t enc;
741 934
742 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 935 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
743 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 936 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
744 937
745 enc.json = *json; 938 enc.json = *json;
746 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 939 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
747 enc.cur = SvPVX (enc.sv); 940 enc.cur = SvPVX (enc.sv);
751 : enc.json.flags & F_LATIN1 ? 0x000100UL 944 : enc.json.flags & F_LATIN1 ? 0x000100UL
752 : 0x110000UL; 945 : 0x110000UL;
753 946
754 SvPOK_only (enc.sv); 947 SvPOK_only (enc.sv);
755 encode_sv (&enc, scalar); 948 encode_sv (&enc, scalar);
949 encode_nl (&enc);
756 950
757 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 951 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
758 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 952 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
759 953
760 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 954 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
940 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1134 else if (expect_true (ch >= 0x20 && ch < 0x80))
941 *cur++ = ch; 1135 *cur++ = ch;
942 else if (ch >= 0x80) 1136 else if (ch >= 0x80)
943 { 1137 {
944 STRLEN clen; 1138 STRLEN clen;
945 UV uch;
946 1139
947 --dec_cur; 1140 --dec_cur;
948 1141
949 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1142 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
950 if (clen == (STRLEN)-1) 1143 if (clen == (STRLEN)-1)
951 ERR ("malformed UTF-8 character in JSON string"); 1144 ERR ("malformed UTF-8 character in JSON string");
952 1145
953 do 1146 do
954 *cur++ = *dec_cur++; 1147 *cur++ = *dec_cur++;
955 while (--clen); 1148 while (--clen);
956 1149
957 utf8 = 1; 1150 utf8 = 1;
958 } 1151 }
1152 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1153 *cur++ = ch;
959 else 1154 else
960 { 1155 {
961 --dec_cur; 1156 --dec_cur;
962 1157
963 if (!ch) 1158 if (!ch)
971 { 1166 {
972 STRLEN len = cur - buf; 1167 STRLEN len = cur - buf;
973 1168
974 if (sv) 1169 if (sv)
975 { 1170 {
976 SvGROW (sv, SvCUR (sv) + len + 1); 1171 STRLEN cur = SvCUR (sv);
1172
1173 if (SvLEN (sv) <= cur + len)
1174 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1175
977 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1176 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
978 SvCUR_set (sv, SvCUR (sv) + len); 1177 SvCUR_set (sv, SvCUR (sv) + len);
979 } 1178 }
980 else 1179 else
981 sv = newSVpvn (buf, len); 1180 sv = newSVpvn (buf, len);
1072 1271
1073 // special case the rather common 1..5-digit-int case 1272 // special case the rather common 1..5-digit-int case
1074 if (*start == '-') 1273 if (*start == '-')
1075 switch (len) 1274 switch (len)
1076 { 1275 {
1077 case 2: return newSViv (-( start [1] - '0' * 1)); 1276 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1078 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1277 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1079 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1278 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1080 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1279 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1081 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1280 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1082 } 1281 }
1083 else 1282 else
1084 switch (len) 1283 switch (len)
1085 { 1284 {
1086 case 1: return newSViv ( start [0] - '0' * 1); 1285 case 1: return newSViv ( start [0] - '0' * 1);
1087 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1286 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1088 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1287 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1089 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1288 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1090 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1289 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1091 } 1290 }
1092 1291
1093 { 1292 {
1094 UV uv; 1293 UV uv;
1095 int numtype = grok_number (start, len, &uv); 1294 int numtype = grok_number (start, len, &uv);
1104 } 1303 }
1105 1304
1106 len -= *start == '-' ? 1 : 0; 1305 len -= *start == '-' ? 1 : 0;
1107 1306
1108 // does not fit into IV or UV, try NV 1307 // does not fit into IV or UV, try NV
1109 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1308 if (len <= NV_DIG)
1110 #if defined (LDBL_DIG)
1111 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1112 #endif
1113 )
1114 // fits into NV without loss of precision 1309 // fits into NV without loss of precision
1115 return newSVnv (Atof (start)); 1310 return newSVnv (json_atof (start));
1116 1311
1117 // everything else fails, convert it to a string 1312 // everything else fails, convert it to a string
1118 return newSVpvn (start, dec->cur - start); 1313 return newSVpvn (start, dec->cur - start);
1119 } 1314 }
1120 1315
1121 // loss of precision here 1316 // loss of precision here
1122 return newSVnv (Atof (start)); 1317 return newSVnv (json_atof (start));
1123 1318
1124fail: 1319fail:
1125 return 0; 1320 return 0;
1126} 1321}
1127 1322
1151 if (*dec->cur == ']') 1346 if (*dec->cur == ']')
1152 { 1347 {
1153 ++dec->cur; 1348 ++dec->cur;
1154 break; 1349 break;
1155 } 1350 }
1156 1351
1157 if (*dec->cur != ',') 1352 if (*dec->cur != ',')
1158 ERR (", or ] expected while parsing array"); 1353 ERR (", or ] expected while parsing array");
1159 1354
1160 ++dec->cur; 1355 ++dec->cur;
1161 1356
1285 1480
1286 hv_iterinit (hv); 1481 hv_iterinit (hv);
1287 he = hv_iternext (hv); 1482 he = hv_iternext (hv);
1288 hv_iterinit (hv); 1483 hv_iterinit (hv);
1289 1484
1290 // the next line creates a mortal sv each time its called. 1485 // the next line creates a mortal sv each time it's called.
1291 // might want to optimise this for common cases. 1486 // might want to optimise this for common cases.
1292 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1487 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1293 1488
1294 if (cb) 1489 if (cb)
1295 { 1490 {
1296 dSP; 1491 dSP;
1297 int count; 1492 int count;
1298 1493
1299 ENTER; SAVETMPS; PUSHMARK (SP); 1494 ENTER; SAVETMPS;
1495 SAVESTACK_POS ();
1496 PUSHMARK (SP);
1300 XPUSHs (HeVAL (he)); 1497 XPUSHs (HeVAL (he));
1498 sv_2mortal (sv);
1301 1499
1302 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1500 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1303 1501
1304 if (count == 1) 1502 if (count == 1)
1305 { 1503 {
1306 sv = newSVsv (POPs); 1504 sv = newSVsv (POPs);
1307 FREETMPS; LEAVE; 1505 FREETMPS; LEAVE;
1308 return sv; 1506 return sv;
1309 } 1507 }
1310 1508
1509 SvREFCNT_inc (sv);
1311 FREETMPS; LEAVE; 1510 FREETMPS; LEAVE;
1312 } 1511 }
1313 } 1512 }
1314 1513
1315 if (dec->json.cb_object) 1514 if (dec->json.cb_object)
1316 { 1515 {
1317 dSP; 1516 dSP;
1318 int count; 1517 int count;
1319 1518
1320 ENTER; SAVETMPS; PUSHMARK (SP); 1519 ENTER; SAVETMPS;
1520 SAVESTACK_POS ();
1521 PUSHMARK (SP);
1321 XPUSHs (sv_2mortal (sv)); 1522 XPUSHs (sv_2mortal (sv));
1322 1523
1323 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1524 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1324 1525
1325 if (count == 1) 1526 if (count == 1)
1341 DEC_DEC_DEPTH; 1542 DEC_DEC_DEPTH;
1342 return 0; 1543 return 0;
1343} 1544}
1344 1545
1345static SV * 1546static SV *
1547decode_tag (dec_t *dec)
1548{
1549 SV *tag = 0;
1550 SV *val = 0;
1551
1552 if (!(dec->json.flags & F_ALLOW_TAGS))
1553 ERR ("malformed JSON string, neither array, object, number, string or atom");
1554
1555 ++dec->cur;
1556
1557 decode_ws (dec);
1558
1559 tag = decode_sv (dec);
1560 if (!tag)
1561 goto fail;
1562
1563 if (!SvPOK (tag))
1564 ERR ("malformed JSON string, (tag) must be a string");
1565
1566 decode_ws (dec);
1567
1568 if (*dec->cur != ')')
1569 ERR (") expected after tag");
1570
1571 ++dec->cur;
1572
1573 decode_ws (dec);
1574
1575 val = decode_sv (dec);
1576 if (!val)
1577 goto fail;
1578
1579 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1580 ERR ("malformed JSON string, tag value must be an array");
1581
1582 {
1583 AV *av = (AV *)SvRV (val);
1584 int i, len = av_len (av) + 1;
1585 HV *stash = gv_stashsv (tag, 0);
1586 SV *sv;
1587
1588 if (!stash)
1589 ERR ("cannot decode perl-object (package does not exist)");
1590
1591 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1592
1593 if (!method)
1594 ERR ("cannot decode perl-object (package does not have a THAW method)");
1595
1596 dSP;
1597
1598 ENTER; SAVETMPS;
1599 PUSHMARK (SP);
1600 EXTEND (SP, len + 2);
1601 // we re-bless the reference to get overload and other niceties right
1602 PUSHs (tag);
1603 PUSHs (sv_json);
1604
1605 for (i = 0; i < len; ++i)
1606 PUSHs (*av_fetch (av, i, 1));
1607
1608 PUTBACK;
1609 call_sv ((SV *)GvCV (method), G_SCALAR);
1610 SPAGAIN;
1611
1612 SvREFCNT_dec (tag);
1613 SvREFCNT_dec (val);
1614 sv = SvREFCNT_inc (POPs);
1615
1616 PUTBACK;
1617
1618 FREETMPS; LEAVE;
1619
1620 return sv;
1621 }
1622
1623fail:
1624 SvREFCNT_dec (tag);
1625 SvREFCNT_dec (val);
1626 return 0;
1627}
1628
1629static SV *
1346decode_sv (dec_t *dec) 1630decode_sv (dec_t *dec)
1347{ 1631{
1348 // the beauty of JSON: you need exactly one character lookahead 1632 // the beauty of JSON: you need exactly one character lookahead
1349 // to parse everything. 1633 // to parse everything.
1350 switch (*dec->cur) 1634 switch (*dec->cur)
1351 { 1635 {
1352 case '"': ++dec->cur; return decode_str (dec); 1636 case '"': ++dec->cur; return decode_str (dec);
1353 case '[': ++dec->cur; return decode_av (dec); 1637 case '[': ++dec->cur; return decode_av (dec);
1354 case '{': ++dec->cur; return decode_hv (dec); 1638 case '{': ++dec->cur; return decode_hv (dec);
1639 case '(': return decode_tag (dec);
1355 1640
1356 case '-': 1641 case '-':
1357 case '0': case '1': case '2': case '3': case '4': 1642 case '0': case '1': case '2': case '3': case '4':
1358 case '5': case '6': case '7': case '8': case '9': 1643 case '5': case '6': case '7': case '8': case '9':
1359 return decode_num (dec); 1644 return decode_num (dec);
1361 case 't': 1646 case 't':
1362 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1647 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1363 { 1648 {
1364 dec->cur += 4; 1649 dec->cur += 4;
1365#if JSON_SLOW 1650#if JSON_SLOW
1366 json_true = get_bool ("JSON::XS::true"); 1651 bool_true = get_bool ("Types::Serialiser::true");
1367#endif 1652#endif
1368 return newSVsv (json_true); 1653 return newSVsv (bool_true);
1369 } 1654 }
1370 else 1655 else
1371 ERR ("'true' expected"); 1656 ERR ("'true' expected");
1372 1657
1373 break; 1658 break;
1375 case 'f': 1660 case 'f':
1376 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1661 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1377 { 1662 {
1378 dec->cur += 5; 1663 dec->cur += 5;
1379#if JSON_SLOW 1664#if JSON_SLOW
1380 json_false = get_bool ("JSON::XS::false"); 1665 bool_false = get_bool ("Types::Serialiser::false");
1381#endif 1666#endif
1382 return newSVsv (json_false); 1667 return newSVsv (bool_false);
1383 } 1668 }
1384 else 1669 else
1385 ERR ("'false' expected"); 1670 ERR ("'false' expected");
1386 1671
1387 break; 1672 break;
1396 ERR ("'null' expected"); 1681 ERR ("'null' expected");
1397 1682
1398 break; 1683 break;
1399 1684
1400 default: 1685 default:
1401 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1686 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1402 break; 1687 break;
1403 } 1688 }
1404 1689
1405fail: 1690fail:
1406 return 0; 1691 return 0;
1408 1693
1409static SV * 1694static SV *
1410decode_json (SV *string, JSON *json, STRLEN *offset_return) 1695decode_json (SV *string, JSON *json, STRLEN *offset_return)
1411{ 1696{
1412 dec_t dec; 1697 dec_t dec;
1413 STRLEN offset;
1414 SV *sv; 1698 SV *sv;
1415 1699
1700 /* work around bugs in 5.10 where manipulating magic values
1701 * makes perl ignore the magic in subsequent accesses.
1702 * also make a copy of non-PV values, to get them into a clean
1703 * state (SvPV should do that, but it's buggy, see below).
1704 *
1705 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1706 * as reported by Reini Urban.
1707 */
1416 SvGETMAGIC (string); 1708 /*SvGETMAGIC (string);*/
1709 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1710 string = sv_2mortal (newSVsv (string));
1711
1417 SvUPGRADE (string, SVt_PV); 1712 SvUPGRADE (string, SVt_PV);
1418 1713
1419 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1714 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1420 * assertion with -DDEBUGGING, although SvCUR is documented to 1715 * assertion with -DDEBUGGING, although SvCUR is documented to
1421 * return the xpv_cur field which certainly exists after upgrading. 1716 * return the xpv_cur field which certainly exists after upgrading.
1422 * according to nicholas clark, calling SvPOK fixes this. 1717 * according to nicholas clark, calling SvPOK fixes this.
1423 * But it doesn't fix it, so try another workaround, call SvPV_nolen 1718 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1424 * and hope for the best. 1719 * and hope for the best.
1720 * Damnit, SvPV_nolen still trips over yet another assertion. This
1721 * assertion business is seriously broken, try yet another workaround
1722 * for the broken -DDEBUGGING.
1425 */ 1723 */
1724 {
1426#ifdef DEBUGGING 1725#ifdef DEBUGGING
1427 SvPV_nolen (string); 1726 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1727#else
1728 STRLEN offset = SvCUR (string);
1428#endif 1729#endif
1429 1730
1430 if (SvCUR (string) > json->max_size && json->max_size) 1731 if (offset > json->max_size && json->max_size)
1431 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1732 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1432 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1733 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1734 }
1433 1735
1434 if (json->flags & F_UTF8) 1736 if (DECODE_WANTS_OCTETS (json))
1435 sv_utf8_downgrade (string, 0); 1737 sv_utf8_downgrade (string, 0);
1436 else 1738 else
1437 sv_utf8_upgrade (string); 1739 sv_utf8_upgrade (string);
1438 1740
1439 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1741 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1450 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1752 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1451 1753
1452 decode_ws (&dec); 1754 decode_ws (&dec);
1453 sv = decode_sv (&dec); 1755 sv = decode_sv (&dec);
1454 1756
1455 if (!(offset_return || !sv)) 1757 if (offset_return)
1758 *offset_return = dec.cur - SvPVX (string);
1759 else if (sv)
1456 { 1760 {
1457 // check for trailing garbage 1761 // check for trailing garbage
1458 decode_ws (&dec); 1762 decode_ws (&dec);
1459 1763
1460 if (*dec.cur) 1764 if (*dec.cur)
1461 { 1765 {
1462 dec.err = "garbage after JSON object"; 1766 dec.err = "garbage after JSON object";
1463 SvREFCNT_dec (sv); 1767 SvREFCNT_dec (sv);
1464 sv = 0; 1768 sv = 0;
1465 } 1769 }
1466 }
1467
1468 if (offset_return || !sv)
1469 {
1470 offset = dec.json.flags & F_UTF8
1471 ? dec.cur - SvPVX (string)
1472 : utf8_distance (dec.cur, SvPVX (string));
1473
1474 if (offset_return)
1475 *offset_return = offset;
1476 } 1770 }
1477 1771
1478 if (!sv) 1772 if (!sv)
1479 { 1773 {
1480 SV *uni = sv_newmortal (); 1774 SV *uni = sv_newmortal ();
1486 SAVEVPTR (PL_curcop); 1780 SAVEVPTR (PL_curcop);
1487 PL_curcop = &cop; 1781 PL_curcop = &cop;
1488 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1782 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1489 LEAVE; 1783 LEAVE;
1490 1784
1491 croak ("%s, at character offset %d [\"%s\"]", 1785 croak ("%s, at character offset %d (before \"%s\")",
1492 dec.err, 1786 dec.err,
1493 (int)offset, 1787 (int)ptr_to_index (string, dec.cur),
1494 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1788 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1495 } 1789 }
1496 1790
1497 sv = sv_2mortal (sv); 1791 sv = sv_2mortal (sv);
1498 1792
1499 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1793 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1500 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1794 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1501 1795
1502 return sv; 1796 return sv;
1503} 1797}
1504 1798
1507 1801
1508static void 1802static void
1509incr_parse (JSON *self) 1803incr_parse (JSON *self)
1510{ 1804{
1511 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1805 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1806
1807 // the state machine here is a bit convoluted and could be simplified a lot
1808 // but this would make it slower, so...
1512 1809
1513 for (;;) 1810 for (;;)
1514 { 1811 {
1515 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D 1812 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1516 switch (self->incr_mode) 1813 switch (self->incr_mode)
1517 { 1814 {
1518 // only used for intiial whitespace skipping 1815 // only used for initial whitespace skipping
1519 case INCR_M_WS: 1816 case INCR_M_WS:
1520 for (;;) 1817 for (;;)
1521 { 1818 {
1522 if (*p > 0x20) 1819 if (*p > 0x20)
1523 { 1820 {
1821 if (*p == '#')
1822 {
1823 self->incr_mode = INCR_M_C0;
1824 goto incr_m_c;
1825 }
1826 else
1827 {
1524 self->incr_mode = INCR_M_JSON; 1828 self->incr_mode = INCR_M_JSON;
1525 goto incr_m_json; 1829 goto incr_m_json;
1830 }
1526 } 1831 }
1527 else if (!*p) 1832 else if (!*p)
1528 goto interrupt; 1833 goto interrupt;
1529 1834
1530 ++p; 1835 ++p;
1536 goto interrupt; 1841 goto interrupt;
1537 1842
1538 ++p; 1843 ++p;
1539 self->incr_mode = INCR_M_STR; 1844 self->incr_mode = INCR_M_STR;
1540 goto incr_m_str; 1845 goto incr_m_str;
1846
1847 // inside #-style comments
1848 case INCR_M_C0:
1849 case INCR_M_C1:
1850 incr_m_c:
1851 for (;;)
1852 {
1853 if (*p == '\n')
1854 {
1855 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1856 break;
1857 }
1858 else if (!*p)
1859 goto interrupt;
1860
1861 ++p;
1862 }
1863
1864 break;
1541 1865
1542 // inside a string 1866 // inside a string
1543 case INCR_M_STR: 1867 case INCR_M_STR:
1544 incr_m_str: 1868 incr_m_str:
1545 for (;;) 1869 for (;;)
1596 self->incr_mode = INCR_M_STR; 1920 self->incr_mode = INCR_M_STR;
1597 goto incr_m_str; 1921 goto incr_m_str;
1598 1922
1599 case '[': 1923 case '[':
1600 case '{': 1924 case '{':
1925 case '(':
1601 if (++self->incr_nest > self->max_depth) 1926 if (++self->incr_nest > self->max_depth)
1602 croak (ERR_NESTING_EXCEEDED); 1927 croak (ERR_NESTING_EXCEEDED);
1603 break; 1928 break;
1604 1929
1605 case ']': 1930 case ']':
1606 case '}': 1931 case '}':
1607 if (--self->incr_nest <= 0) 1932 if (--self->incr_nest <= 0)
1608 goto interrupt; 1933 goto interrupt;
1934 break;
1935
1936 case ')':
1937 --self->incr_nest;
1938 break;
1939
1940 case '#':
1941 self->incr_mode = INCR_M_C1;
1942 goto incr_m_c;
1609 } 1943 }
1610 } 1944 }
1611 } 1945 }
1612 1946
1613 modechange: 1947 modechange:
1614 ; 1948 ;
1615 } 1949 }
1616 1950
1617interrupt: 1951interrupt:
1618 self->incr_pos = p - SvPVX (self->incr_text); 1952 self->incr_pos = p - SvPVX (self->incr_text);
1953 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1619 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1954 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1620} 1955}
1621 1956
1622///////////////////////////////////////////////////////////////////////////// 1957/////////////////////////////////////////////////////////////////////////////
1623// XS interface functions 1958// XS interface functions
1633 i >= '0' && i <= '9' ? i - '0' 1968 i >= '0' && i <= '9' ? i - '0'
1634 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1969 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1635 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1970 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1636 : -1; 1971 : -1;
1637 1972
1638 json_stash = gv_stashpv ("JSON::XS" , 1); 1973 json_stash = gv_stashpv ("JSON::XS" , 1);
1639 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1974 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1640
1641 json_true = get_bool ("JSON::XS::true"); 1975 bool_true = get_bool ("Types::Serialiser::true");
1642 json_false = get_bool ("JSON::XS::false"); 1976 bool_false = get_bool ("Types::Serialiser::false");
1977
1978 sv_json = newSVpv ("JSON", 0);
1979 SvREADONLY_on (sv_json);
1980
1981 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1643} 1982}
1644 1983
1645PROTOTYPES: DISABLE 1984PROTOTYPES: DISABLE
1646 1985
1647void CLONE (...) 1986void CLONE (...)
1648 CODE: 1987 CODE:
1649 json_stash = 0; 1988 json_stash = 0;
1650 json_boolean_stash = 0; 1989 bool_stash = 0;
1651 1990
1652void new (char *klass) 1991void new (char *klass)
1653 PPCODE: 1992 PPCODE:
1654{ 1993{
1655 SV *pv = NEWSV (0, sizeof (JSON)); 1994 SV *pv = NEWSV (0, sizeof (JSON));
1656 SvPOK_only (pv); 1995 SvPOK_only (pv);
1657 json_init ((JSON *)SvPVX (pv)); 1996 json_init ((JSON *)SvPVX (pv));
1658 XPUSHs (sv_2mortal (sv_bless ( 1997 XPUSHs (sv_2mortal (sv_bless (
1659 newRV_noinc (pv), 1998 newRV_noinc (pv),
1660 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 1999 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1675 shrink = F_SHRINK 2014 shrink = F_SHRINK
1676 allow_blessed = F_ALLOW_BLESSED 2015 allow_blessed = F_ALLOW_BLESSED
1677 convert_blessed = F_CONV_BLESSED 2016 convert_blessed = F_CONV_BLESSED
1678 relaxed = F_RELAXED 2017 relaxed = F_RELAXED
1679 allow_unknown = F_ALLOW_UNKNOWN 2018 allow_unknown = F_ALLOW_UNKNOWN
2019 allow_tags = F_ALLOW_TAGS
1680 PPCODE: 2020 PPCODE:
1681{ 2021{
1682 if (enable) 2022 if (enable)
1683 self->flags |= ix; 2023 self->flags |= ix;
1684 else 2024 else
1700 get_shrink = F_SHRINK 2040 get_shrink = F_SHRINK
1701 get_allow_blessed = F_ALLOW_BLESSED 2041 get_allow_blessed = F_ALLOW_BLESSED
1702 get_convert_blessed = F_CONV_BLESSED 2042 get_convert_blessed = F_CONV_BLESSED
1703 get_relaxed = F_RELAXED 2043 get_relaxed = F_RELAXED
1704 get_allow_unknown = F_ALLOW_UNKNOWN 2044 get_allow_unknown = F_ALLOW_UNKNOWN
2045 get_allow_tags = F_ALLOW_TAGS
1705 PPCODE: 2046 PPCODE:
1706 XPUSHs (boolSV (self->flags & ix)); 2047 XPUSHs (boolSV (self->flags & ix));
1707 2048
1708void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2049void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1709 PPCODE: 2050 PPCODE:
1737} 2078}
1738 2079
1739void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2080void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1740 PPCODE: 2081 PPCODE:
1741{ 2082{
1742 if (!self->cb_sk_object) 2083 if (!self->cb_sk_object)
1743 self->cb_sk_object = newHV (); 2084 self->cb_sk_object = newHV ();
1744 2085
1745 if (SvOK (cb)) 2086 if (SvOK (cb))
1746 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2087 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1747 else 2088 else
1758 XPUSHs (ST (0)); 2099 XPUSHs (ST (0));
1759} 2100}
1760 2101
1761void encode (JSON *self, SV *scalar) 2102void encode (JSON *self, SV *scalar)
1762 PPCODE: 2103 PPCODE:
1763 XPUSHs (encode_json (scalar, self)); 2104 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2105 XPUSHs (scalar);
1764 2106
1765void decode (JSON *self, SV *jsonstr) 2107void decode (JSON *self, SV *jsonstr)
1766 PPCODE: 2108 PPCODE:
1767 XPUSHs (decode_json (jsonstr, self, 0)); 2109 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2110 XPUSHs (jsonstr);
1768 2111
1769void decode_prefix (JSON *self, SV *jsonstr) 2112void decode_prefix (JSON *self, SV *jsonstr)
1770 PPCODE: 2113 PPCODE:
1771{ 2114{
2115 SV *sv;
1772 STRLEN offset; 2116 STRLEN offset;
2117 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1773 EXTEND (SP, 2); 2118 EXTEND (SP, 2);
1774 PUSHs (decode_json (jsonstr, self, &offset)); 2119 PUSHs (sv);
1775 PUSHs (sv_2mortal (newSVuv (offset))); 2120 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1776} 2121}
1777 2122
1778void incr_parse (JSON *self, SV *jsonstr = 0) 2123void incr_parse (JSON *self, SV *jsonstr = 0)
1779 PPCODE: 2124 PPCODE:
1780{ 2125{
1781 if (!self->incr_text) 2126 if (!self->incr_text)
1782 self->incr_text = newSVpvn ("", 0); 2127 self->incr_text = newSVpvn ("", 0);
2128
2129 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2130 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2131 if (DECODE_WANTS_OCTETS (self))
2132 {
2133 if (self->incr_pos)
2134 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2135 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2136
2137 sv_utf8_downgrade (self->incr_text, 0);
2138 }
2139 else
2140 {
2141 sv_utf8_upgrade (self->incr_text);
2142
2143 if (self->incr_pos)
2144 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2145 - (U8 *)SvPVX (self->incr_text);
2146 }
1783 2147
1784 // append data, if any 2148 // append data, if any
1785 if (jsonstr) 2149 if (jsonstr)
1786 { 2150 {
2151 /* make sure both strings have same encoding */
1787 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2152 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2153 if (SvUTF8 (jsonstr))
2154 sv_utf8_downgrade (jsonstr, 0);
1788 { 2155 else
1789 /* utf-8-ness differs, need to upgrade */
1790 sv_utf8_upgrade (self->incr_text); 2156 sv_utf8_upgrade (jsonstr);
1791 2157
1792 if (self->incr_pos) 2158 /* and then just blindly append */
1793 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1794 - (U8 *)SvPVX (self->incr_text);
1795 }
1796
1797 { 2159 {
1798 STRLEN len; 2160 STRLEN len;
1799 const char *str = SvPV (jsonstr, len); 2161 const char *str = SvPV (jsonstr, len);
1800 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2162 STRLEN cur = SvCUR (self->incr_text);
2163
2164 if (SvLEN (self->incr_text) <= cur + len)
2165 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2166
1801 Move (str, SvEND (self->incr_text), len, char); 2167 Move (str, SvEND (self->incr_text), len, char);
1802 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2168 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1803 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2169 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1804 } 2170 }
1805 } 2171 }
1806 2172
1807 if (GIMME_V != G_VOID) 2173 if (GIMME_V != G_VOID)
1808 do 2174 do
1809 { 2175 {
2176 SV *sv;
1810 STRLEN offset; 2177 STRLEN offset;
1811 2178
1812 if (!INCR_DONE (self)) 2179 if (!INCR_DONE (self))
1813 { 2180 {
1814 incr_parse (self); 2181 incr_parse (self);
1816 if (self->incr_pos > self->max_size && self->max_size) 2183 if (self->incr_pos > self->max_size && self->max_size)
1817 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2184 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1818 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2185 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1819 2186
1820 if (!INCR_DONE (self)) 2187 if (!INCR_DONE (self))
2188 {
2189 // as an optimisation, do not accumulate white space in the incr buffer
2190 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2191 {
2192 self->incr_pos = 0;
2193 SvCUR_set (self->incr_text, 0);
2194 }
2195
1821 break; 2196 break;
2197 }
1822 } 2198 }
1823 2199
1824 XPUSHs (decode_json (self->incr_text, self, &offset)); 2200 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2201 XPUSHs (sv);
1825 2202
1826 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1827 self->incr_pos -= offset; 2203 self->incr_pos -= offset;
1828 self->incr_nest = 0; 2204 self->incr_nest = 0;
1829 self->incr_mode = 0; 2205 self->incr_mode = 0;
2206
2207 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1830 } 2208 }
1831 while (GIMME_V == G_ARRAY); 2209 while (GIMME_V == G_ARRAY);
1832} 2210}
1833 2211
1834SV *incr_text (JSON *self) 2212SV *incr_text (JSON *self)
1872 SvREFCNT_dec (self->incr_text); 2250 SvREFCNT_dec (self->incr_text);
1873 2251
1874PROTOTYPES: ENABLE 2252PROTOTYPES: ENABLE
1875 2253
1876void encode_json (SV *scalar) 2254void encode_json (SV *scalar)
1877 ALIAS:
1878 to_json_ = 0
1879 encode_json = F_UTF8
1880 PPCODE: 2255 PPCODE:
1881{ 2256{
1882 JSON json; 2257 JSON json;
1883 json_init (&json); 2258 json_init (&json);
1884 json.flags |= ix; 2259 json.flags |= F_UTF8;
1885 XPUSHs (encode_json (scalar, &json)); 2260 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2261 XPUSHs (scalar);
1886} 2262}
1887 2263
1888void decode_json (SV *jsonstr) 2264void decode_json (SV *jsonstr)
1889 ALIAS:
1890 from_json_ = 0
1891 decode_json = F_UTF8
1892 PPCODE: 2265 PPCODE:
1893{ 2266{
1894 JSON json; 2267 JSON json;
1895 json_init (&json); 2268 json_init (&json);
1896 json.flags |= ix; 2269 json.flags |= F_UTF8;
1897 XPUSHs (decode_json (jsonstr, &json, 0)); 2270 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2271 XPUSHs (jsonstr);
1898} 2272}
1899 2273
1900

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines