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 1.93 by root, Mon Sep 29 03:09:27 2008 UTC vs.
Revision 1.131 by root, Thu Aug 17 01:42:20 2017 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
130 SvPV_shrink_to_cur (sv); 153 SvPV_shrink_to_cur (sv);
131#elif defined (SvPV_renew) 154#elif defined (SvPV_renew)
132 SvPV_renew (sv, SvCUR (sv) + 1); 155 SvPV_renew (sv, SvCUR (sv) + 1);
133#endif 156#endif
134 } 157 }
158}
159
160/* adds two STRLENs together, slow, and with paranoia */
161STRLEN
162strlen_sum (STRLEN l1, STRLEN l2)
163{
164 size_t sum = l1 + l2;
165
166 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
167 croak ("JSON::XS: string size overflow");
168
169 return sum;
170}
171
172/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
173static char *
174json_sv_grow (SV *sv, size_t len1, size_t len2)
175{
176 len1 = strlen_sum (len1, len2);
177 len1 = strlen_sum (len1, len1 >> 1);
178
179 if (len1 > 4096 - 24)
180 len1 = (len1 | 4095) - 24;
181
182 return SvGROW (sv, len1);
135} 183}
136 184
137// decode an utf-8 character and return it, or (UV)-1 in 185// decode an utf-8 character and return it, or (UV)-1 in
138// case of an error. 186// case of an error.
139// we special-case "safe" characters from U+80 .. U+7FF, 187// we special-case "safe" characters from U+80 .. U+7FF,
175 *s++ = 0x80 | ( ch & 0x3f); 223 *s++ = 0x80 | ( ch & 0x3f);
176 224
177 return s; 225 return s;
178} 226}
179 227
228// convert offset pointer to character index, sv must be string
229static STRLEN
230ptr_to_index (SV *sv, char *offset)
231{
232 return SvUTF8 (sv)
233 ? utf8_distance (offset, SvPVX (sv))
234 : offset - SvPVX (sv);
235}
236
237/////////////////////////////////////////////////////////////////////////////
238// fp hell
239
240// scan a group of digits, and a trailing exponent
241static void
242json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
243{
244 UV uaccum = 0;
245 int eaccum = 0;
246
247 // if we recurse too deep, skip all remaining digits
248 // to avoid a stack overflow attack
249 if (expect_false (--maxdepth <= 0))
250 while (((U8)*s - '0') < 10)
251 ++s;
252
253 for (;;)
254 {
255 U8 dig = (U8)*s - '0';
256
257 if (expect_false (dig >= 10))
258 {
259 if (dig == (U8)((U8)'.' - (U8)'0'))
260 {
261 ++s;
262 json_atof_scan1 (s, accum, expo, 1, maxdepth);
263 }
264 else if ((dig | ' ') == 'e' - '0')
265 {
266 int exp2 = 0;
267 int neg = 0;
268
269 ++s;
270
271 if (*s == '-')
272 {
273 ++s;
274 neg = 1;
275 }
276 else if (*s == '+')
277 ++s;
278
279 while ((dig = (U8)*s - '0') < 10)
280 exp2 = exp2 * 10 + *s++ - '0';
281
282 *expo += neg ? -exp2 : exp2;
283 }
284
285 break;
286 }
287
288 ++s;
289
290 uaccum = uaccum * 10 + dig;
291 ++eaccum;
292
293 // if we have too many digits, then recurse for more
294 // we actually do this for rather few digits
295 if (uaccum >= (UV_MAX - 9) / 10)
296 {
297 if (postdp) *expo -= eaccum;
298 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
299 if (postdp) *expo += eaccum;
300
301 break;
302 }
303 }
304
305 // this relies greatly on the quality of the pow ()
306 // implementation of the platform, but a good
307 // implementation is hard to beat.
308 // (IEEE 754 conformant ones are required to be exact)
309 if (postdp) *expo -= eaccum;
310 *accum += uaccum * Perl_pow (10., *expo);
311 *expo += eaccum;
312}
313
314static NV
315json_atof (const char *s)
316{
317 NV accum = 0.;
318 int expo = 0;
319 int neg = 0;
320
321 if (*s == '-')
322 {
323 ++s;
324 neg = 1;
325 }
326
327 // a recursion depth of ten gives us >>500 bits
328 json_atof_scan1 (s, &accum, &expo, 0, 10);
329
330 return neg ? -accum : accum;
331}
332
333// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
334static int
335ref_bool_type (SV *sv)
336{
337 svtype svt = SvTYPE (sv);
338
339 if (svt < SVt_PVAV)
340 {
341 STRLEN len = 0;
342 char *pv = svt ? SvPV (sv, len) : 0;
343
344 if (len == 1)
345 if (*pv == '1')
346 return 1;
347 else if (*pv == '0')
348 return 0;
349 }
350
351 return -1;
352}
353
354// returns whether scalar is not a reference in the sense of allow_nonref
355static int
356json_nonref (SV *scalar)
357{
358 if (!SvROK (scalar))
359 return 1;
360
361 scalar = SvRV (scalar);
362
363 if (SvTYPE (scalar) >= SVt_PVMG)
364 {
365 if (SvSTASH (scalar) == bool_stash)
366 return 1;
367
368 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
369 return 1;
370 }
371
372 return 0;
373}
374
180///////////////////////////////////////////////////////////////////////////// 375/////////////////////////////////////////////////////////////////////////////
181// encoder 376// encoder
182 377
183// structure used for encoding JSON 378// structure used for encoding JSON
184typedef struct 379typedef struct
192} enc_t; 387} enc_t;
193 388
194INLINE void 389INLINE void
195need (enc_t *enc, STRLEN len) 390need (enc_t *enc, STRLEN len)
196{ 391{
197 if (expect_false (enc->cur + len >= enc->end)) 392 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
198 { 393 {
199 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 394 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
200 SvGROW (enc->sv, cur + len + 1); 395 char *buf = json_sv_grow (enc->sv, cur, len);
201 enc->cur = SvPVX (enc->sv) + cur; 396 enc->cur = buf + cur;
202 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 397 enc->end = buf + SvLEN (enc->sv) - 1;
203 } 398 }
204} 399}
205 400
206INLINE void 401INLINE void
207encode_ch (enc_t *enc, char ch) 402encode_ch (enc_t *enc, char ch)
223 418
224 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 419 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
225 { 420 {
226 if (expect_false (ch == '"')) // but with slow exceptions 421 if (expect_false (ch == '"')) // but with slow exceptions
227 { 422 {
228 need (enc, len += 1); 423 need (enc, len + 1);
229 *enc->cur++ = '\\'; 424 *enc->cur++ = '\\';
230 *enc->cur++ = '"'; 425 *enc->cur++ = '"';
231 } 426 }
232 else if (expect_false (ch == '\\')) 427 else if (expect_false (ch == '\\'))
233 { 428 {
234 need (enc, len += 1); 429 need (enc, len + 1);
235 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
236 *enc->cur++ = '\\'; 431 *enc->cur++ = '\\';
237 } 432 }
238 else 433 else
239 *enc->cur++ = ch; 434 *enc->cur++ = ch;
242 } 437 }
243 else 438 else
244 { 439 {
245 switch (ch) 440 switch (ch)
246 { 441 {
247 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 442 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
248 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 443 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
249 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 444 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
250 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 445 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
251 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 446 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
252 447
253 default: 448 default:
254 { 449 {
255 STRLEN clen; 450 STRLEN clen;
256 UV uch; 451 UV uch;
272 if (uch >= 0x10000UL) 467 if (uch >= 0x10000UL)
273 { 468 {
274 if (uch >= 0x110000UL) 469 if (uch >= 0x110000UL)
275 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 470 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
276 471
277 need (enc, len += 11); 472 need (enc, len + 11);
278 sprintf (enc->cur, "\\u%04x\\u%04x", 473 sprintf (enc->cur, "\\u%04x\\u%04x",
279 (int)((uch - 0x10000) / 0x400 + 0xD800), 474 (int)((uch - 0x10000) / 0x400 + 0xD800),
280 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 475 (int)((uch - 0x10000) % 0x400 + 0xDC00));
281 enc->cur += 12; 476 enc->cur += 12;
282 } 477 }
283 else 478 else
284 { 479 {
285 need (enc, len += 5); 480 need (enc, len + 5);
286 *enc->cur++ = '\\'; 481 *enc->cur++ = '\\';
287 *enc->cur++ = 'u'; 482 *enc->cur++ = 'u';
288 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 483 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
289 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 484 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
290 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 485 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
298 *enc->cur++ = uch; 493 *enc->cur++ = uch;
299 str += clen; 494 str += clen;
300 } 495 }
301 else if (is_utf8) 496 else if (is_utf8)
302 { 497 {
303 need (enc, len += clen); 498 need (enc, len + clen);
304 do 499 do
305 { 500 {
306 *enc->cur++ = *str++; 501 *enc->cur++ = *str++;
307 } 502 }
308 while (--clen); 503 while (--clen);
309 } 504 }
310 else 505 else
311 { 506 {
312 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 507 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
313 enc->cur = encode_utf8 (enc->cur, uch); 508 enc->cur = encode_utf8 (enc->cur, uch);
314 ++str; 509 ++str;
315 } 510 }
316 } 511 }
317 } 512 }
371 566
372 if (enc->indent >= enc->json.max_depth) 567 if (enc->indent >= enc->json.max_depth)
373 croak (ERR_NESTING_EXCEEDED); 568 croak (ERR_NESTING_EXCEEDED);
374 569
375 encode_ch (enc, '['); 570 encode_ch (enc, '[');
376 571
377 if (len >= 0) 572 if (len >= 0)
378 { 573 {
379 encode_nl (enc); ++enc->indent; 574 encode_nl (enc); ++enc->indent;
380 575
381 for (i = 0; i <= len; ++i) 576 for (i = 0; i <= len; ++i)
393 encode_comma (enc); 588 encode_comma (enc);
394 } 589 }
395 590
396 encode_nl (enc); --enc->indent; encode_indent (enc); 591 encode_nl (enc); --enc->indent; encode_indent (enc);
397 } 592 }
398 593
399 encode_ch (enc, ']'); 594 encode_ch (enc, ']');
400} 595}
401 596
402static void 597static void
403encode_hk (enc_t *enc, HE *he) 598encode_hk (enc_t *enc, HE *he)
407 if (HeKLEN (he) == HEf_SVKEY) 602 if (HeKLEN (he) == HEf_SVKEY)
408 { 603 {
409 SV *sv = HeSVKEY (he); 604 SV *sv = HeSVKEY (he);
410 STRLEN len; 605 STRLEN len;
411 char *str; 606 char *str;
412 607
413 SvGETMAGIC (sv); 608 SvGETMAGIC (sv);
414 str = SvPV (sv, len); 609 str = SvPV (sv, len);
415 610
416 encode_str (enc, str, len, SvUTF8 (sv)); 611 encode_str (enc, str, len, SvUTF8 (sv));
417 } 612 }
462 657
463 // for canonical output we have to sort by keys first 658 // for canonical output we have to sort by keys first
464 // actually, this is mostly due to the stupid so-called 659 // actually, this is mostly due to the stupid so-called
465 // security workaround added somewhere in 5.8.x 660 // security workaround added somewhere in 5.8.x
466 // that randomises hash orderings 661 // that randomises hash orderings
467 if (enc->json.flags & F_CANONICAL) 662 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
468 { 663 {
469 int count = hv_iterinit (hv); 664 int count = hv_iterinit (hv);
470 665
471 if (SvMAGICAL (hv)) 666 if (SvMAGICAL (hv))
472 { 667 {
483 } 678 }
484 679
485 if (count) 680 if (count)
486 { 681 {
487 int i, fast = 1; 682 int i, fast = 1;
488#if defined(__BORLANDC__) || defined(_MSC_VER) 683 HE *hes_stack [STACK_HES];
489 HE **hes = _alloca (count * sizeof (HE)); 684 HE **hes = hes_stack;
490#else 685
491 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 686 // allocate larger arrays on the heap
492#endif 687 if (count > STACK_HES)
688 {
689 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
690 hes = (HE **)SvPVX (sv);
691 }
493 692
494 i = 0; 693 i = 0;
495 while ((he = hv_iternext (hv))) 694 while ((he = hv_iternext (hv)))
496 { 695 {
497 hes [i++] = he; 696 hes [i++] = he;
566// encode objects, arrays and special \0=false and \1=true values. 765// encode objects, arrays and special \0=false and \1=true values.
567static void 766static void
568encode_rv (enc_t *enc, SV *sv) 767encode_rv (enc_t *enc, SV *sv)
569{ 768{
570 svtype svt; 769 svtype svt;
770 GV *method;
571 771
572 SvGETMAGIC (sv); 772 SvGETMAGIC (sv);
573 svt = SvTYPE (sv); 773 svt = SvTYPE (sv);
574 774
575 if (expect_false (SvOBJECT (sv))) 775 if (expect_false (SvOBJECT (sv)))
576 { 776 {
577 HV *stash = !JSON_SLOW || json_boolean_stash 777 HV *stash = SvSTASH (sv);
578 ? json_boolean_stash
579 : gv_stashpv ("JSON::XS::Boolean", 1);
580 778
581 if (SvSTASH (sv) == stash) 779 if (stash == bool_stash)
582 { 780 {
583 if (SvIV (sv)) 781 if (SvIV (sv))
584 encode_str (enc, "true", 4, 0); 782 encode_str (enc, "true", 4, 0);
585 else 783 else
586 encode_str (enc, "false", 5, 0); 784 encode_str (enc, "false", 5, 0);
587 } 785 }
786 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
787 {
788 int count;
789 dSP;
790
791 ENTER; SAVETMPS;
792 SAVESTACK_POS ();
793 PUSHMARK (SP);
794 EXTEND (SP, 2);
795 // we re-bless the reference to get overload and other niceties right
796 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
797 PUSHs (sv_json);
798
799 PUTBACK;
800 count = call_sv ((SV *)GvCV (method), G_ARRAY);
801 SPAGAIN;
802
803 // catch this surprisingly common error
804 if (SvROK (TOPs) && SvRV (TOPs) == sv)
805 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
806
807 encode_ch (enc, '(');
808 encode_ch (enc, '"');
809 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
810 encode_ch (enc, '"');
811 encode_ch (enc, ')');
812 encode_ch (enc, '[');
813
814 while (count)
815 {
816 encode_sv (enc, SP[1 - count--]);
817
818 if (count)
819 encode_ch (enc, ',');
820 }
821
822 encode_ch (enc, ']');
823
824 FREETMPS; LEAVE;
825 }
826 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
827 {
828 dSP;
829
830 ENTER; SAVETMPS;
831 PUSHMARK (SP);
832 // we re-bless the reference to get overload and other niceties right
833 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
834
835 // calling with G_SCALAR ensures that we always get a 1 return value
836 PUTBACK;
837 call_sv ((SV *)GvCV (method), G_SCALAR);
838 SPAGAIN;
839
840 // catch this surprisingly common error
841 if (SvROK (TOPs) && SvRV (TOPs) == sv)
842 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
843
844 sv = POPs;
845 PUTBACK;
846
847 encode_sv (enc, sv);
848
849 FREETMPS; LEAVE;
850 }
851 else if (enc->json.flags & F_ALLOW_BLESSED)
852 encode_str (enc, "null", 4, 0);
588 else 853 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", 854 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)))); 855 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
635 }
636 } 856 }
637 else if (svt == SVt_PVHV) 857 else if (svt == SVt_PVHV)
638 encode_hv (enc, (HV *)sv); 858 encode_hv (enc, (HV *)sv);
639 else if (svt == SVt_PVAV) 859 else if (svt == SVt_PVAV)
640 encode_av (enc, (AV *)sv); 860 encode_av (enc, (AV *)sv);
641 else if (svt < SVt_PVAV) 861 else if (svt < SVt_PVAV)
642 { 862 {
643 STRLEN len = 0; 863 int bool_type = ref_bool_type (sv);
644 char *pv = svt ? SvPV (sv, len) : 0;
645 864
646 if (len == 1 && *pv == '1') 865 if (bool_type == 1)
647 encode_str (enc, "true", 4, 0); 866 encode_str (enc, "true", 4, 0);
648 else if (len == 1 && *pv == '0') 867 else if (bool_type == 0)
649 encode_str (enc, "false", 5, 0); 868 encode_str (enc, "false", 5, 0);
650 else if (enc->json.flags & F_ALLOW_UNKNOWN) 869 else if (enc->json.flags & F_ALLOW_UNKNOWN)
651 encode_str (enc, "null", 4, 0); 870 encode_str (enc, "null", 4, 0);
652 else 871 else
653 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 872 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
716 } 935 }
717 else 936 else
718 { 937 {
719 // large integer, use the (rather slow) snprintf way. 938 // large integer, use the (rather slow) snprintf way.
720 need (enc, IVUV_MAXCHARS); 939 need (enc, IVUV_MAXCHARS);
721 enc->cur += 940 enc->cur +=
722 SvIsUV(sv) 941 SvIsUV(sv)
723 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 942 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
724 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 943 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
725 } 944 }
726 } 945 }
727 else if (SvROK (sv)) 946 else if (SvROK (sv))
728 encode_rv (enc, SvRV (sv)); 947 encode_rv (enc, SvRV (sv));
729 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 948 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
730 encode_str (enc, "null", 4, 0); 949 encode_str (enc, "null", 4, 0);
731 else 950 else
732 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 951 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
733 SvPV_nolen (sv), SvFLAGS (sv)); 952 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
734} 953}
735 954
736static SV * 955static SV *
737encode_json (SV *scalar, JSON *json) 956encode_json (SV *scalar, JSON *json)
738{ 957{
739 enc_t enc; 958 enc_t enc;
740 959
741 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 960 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)"); 961 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
743 962
744 enc.json = *json; 963 enc.json = *json;
745 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 964 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
746 enc.cur = SvPVX (enc.sv); 965 enc.cur = SvPVX (enc.sv);
750 : enc.json.flags & F_LATIN1 ? 0x000100UL 969 : enc.json.flags & F_LATIN1 ? 0x000100UL
751 : 0x110000UL; 970 : 0x110000UL;
752 971
753 SvPOK_only (enc.sv); 972 SvPOK_only (enc.sv);
754 encode_sv (&enc, scalar); 973 encode_sv (&enc, scalar);
974 encode_nl (&enc);
755 975
756 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 976 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
757 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 977 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
758 978
759 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 979 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
939 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1159 else if (expect_true (ch >= 0x20 && ch < 0x80))
940 *cur++ = ch; 1160 *cur++ = ch;
941 else if (ch >= 0x80) 1161 else if (ch >= 0x80)
942 { 1162 {
943 STRLEN clen; 1163 STRLEN clen;
944 UV uch;
945 1164
946 --dec_cur; 1165 --dec_cur;
947 1166
948 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1167 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
949 if (clen == (STRLEN)-1) 1168 if (clen == (STRLEN)-1)
950 ERR ("malformed UTF-8 character in JSON string"); 1169 ERR ("malformed UTF-8 character in JSON string");
951 1170
952 do 1171 do
953 *cur++ = *dec_cur++; 1172 *cur++ = *dec_cur++;
954 while (--clen); 1173 while (--clen);
955 1174
956 utf8 = 1; 1175 utf8 = 1;
957 } 1176 }
1177 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1178 *cur++ = ch;
958 else 1179 else
959 { 1180 {
960 --dec_cur; 1181 --dec_cur;
961 1182
962 if (!ch) 1183 if (!ch)
970 { 1191 {
971 STRLEN len = cur - buf; 1192 STRLEN len = cur - buf;
972 1193
973 if (sv) 1194 if (sv)
974 { 1195 {
975 SvGROW (sv, SvCUR (sv) + len + 1); 1196 STRLEN cur = SvCUR (sv);
1197
1198 if (SvLEN (sv) - cur <= len)
1199 json_sv_grow (sv, cur, len);
1200
976 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1201 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
977 SvCUR_set (sv, SvCUR (sv) + len); 1202 SvCUR_set (sv, SvCUR (sv) + len);
978 } 1203 }
979 else 1204 else
980 sv = newSVpvn (buf, len); 1205 sv = newSVpvn (buf, len);
1071 1296
1072 // special case the rather common 1..5-digit-int case 1297 // special case the rather common 1..5-digit-int case
1073 if (*start == '-') 1298 if (*start == '-')
1074 switch (len) 1299 switch (len)
1075 { 1300 {
1076 case 2: return newSViv (-( start [1] - '0' * 1)); 1301 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1077 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1302 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)); 1303 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)); 1304 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)); 1305 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1081 } 1306 }
1082 else 1307 else
1083 switch (len) 1308 switch (len)
1084 { 1309 {
1085 case 1: return newSViv ( start [0] - '0' * 1); 1310 case 1: return newSViv ( start [0] - '0' * 1);
1086 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1311 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); 1312 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); 1313 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); 1314 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1090 } 1315 }
1091 1316
1092 { 1317 {
1093 UV uv; 1318 UV uv;
1094 int numtype = grok_number (start, len, &uv); 1319 int numtype = grok_number (start, len, &uv);
1103 } 1328 }
1104 1329
1105 len -= *start == '-' ? 1 : 0; 1330 len -= *start == '-' ? 1 : 0;
1106 1331
1107 // does not fit into IV or UV, try NV 1332 // does not fit into IV or UV, try NV
1108 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1333 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 1334 // fits into NV without loss of precision
1114 return newSVnv (Atof (start)); 1335 return newSVnv (json_atof (start));
1115 1336
1116 // everything else fails, convert it to a string 1337 // everything else fails, convert it to a string
1117 return newSVpvn (start, dec->cur - start); 1338 return newSVpvn (start, dec->cur - start);
1118 } 1339 }
1119 1340
1120 // loss of precision here 1341 // loss of precision here
1121 return newSVnv (Atof (start)); 1342 return newSVnv (json_atof (start));
1122 1343
1123fail: 1344fail:
1124 return 0; 1345 return 0;
1125} 1346}
1126 1347
1150 if (*dec->cur == ']') 1371 if (*dec->cur == ']')
1151 { 1372 {
1152 ++dec->cur; 1373 ++dec->cur;
1153 break; 1374 break;
1154 } 1375 }
1155 1376
1156 if (*dec->cur != ',') 1377 if (*dec->cur != ',')
1157 ERR (", or ] expected while parsing array"); 1378 ERR (", or ] expected while parsing array");
1158 1379
1159 ++dec->cur; 1380 ++dec->cur;
1160 1381
1284 1505
1285 hv_iterinit (hv); 1506 hv_iterinit (hv);
1286 he = hv_iternext (hv); 1507 he = hv_iternext (hv);
1287 hv_iterinit (hv); 1508 hv_iterinit (hv);
1288 1509
1289 // the next line creates a mortal sv each time its called. 1510 // the next line creates a mortal sv each time it's called.
1290 // might want to optimise this for common cases. 1511 // might want to optimise this for common cases.
1291 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1512 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1292 1513
1293 if (cb) 1514 if (cb)
1294 { 1515 {
1295 dSP; 1516 dSP;
1296 int count; 1517 int count;
1297 1518
1298 ENTER; SAVETMPS; PUSHMARK (SP); 1519 ENTER; SAVETMPS;
1520 SAVESTACK_POS ();
1521 PUSHMARK (SP);
1299 XPUSHs (HeVAL (he)); 1522 XPUSHs (HeVAL (he));
1523 sv_2mortal (sv);
1300 1524
1301 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1525 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1302 1526
1303 if (count == 1) 1527 if (count == 1)
1304 { 1528 {
1305 sv = newSVsv (POPs); 1529 sv = newSVsv (POPs);
1306 FREETMPS; LEAVE; 1530 FREETMPS; LEAVE;
1307 return sv; 1531 return sv;
1308 } 1532 }
1309 1533
1534 SvREFCNT_inc (sv);
1310 FREETMPS; LEAVE; 1535 FREETMPS; LEAVE;
1311 } 1536 }
1312 } 1537 }
1313 1538
1314 if (dec->json.cb_object) 1539 if (dec->json.cb_object)
1315 { 1540 {
1316 dSP; 1541 dSP;
1317 int count; 1542 int count;
1318 1543
1319 ENTER; SAVETMPS; PUSHMARK (SP); 1544 ENTER; SAVETMPS;
1545 SAVESTACK_POS ();
1546 PUSHMARK (SP);
1320 XPUSHs (sv_2mortal (sv)); 1547 XPUSHs (sv_2mortal (sv));
1321 1548
1322 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1549 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1323 1550
1324 if (count == 1) 1551 if (count == 1)
1340 DEC_DEC_DEPTH; 1567 DEC_DEC_DEPTH;
1341 return 0; 1568 return 0;
1342} 1569}
1343 1570
1344static SV * 1571static SV *
1572decode_tag (dec_t *dec)
1573{
1574 SV *tag = 0;
1575 SV *val = 0;
1576
1577 if (!(dec->json.flags & F_ALLOW_TAGS))
1578 ERR ("malformed JSON string, neither array, object, number, string or atom");
1579
1580 ++dec->cur;
1581
1582 decode_ws (dec);
1583
1584 tag = decode_sv (dec);
1585 if (!tag)
1586 goto fail;
1587
1588 if (!SvPOK (tag))
1589 ERR ("malformed JSON string, (tag) must be a string");
1590
1591 decode_ws (dec);
1592
1593 if (*dec->cur != ')')
1594 ERR (") expected after tag");
1595
1596 ++dec->cur;
1597
1598 decode_ws (dec);
1599
1600 val = decode_sv (dec);
1601 if (!val)
1602 goto fail;
1603
1604 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1605 ERR ("malformed JSON string, tag value must be an array");
1606
1607 {
1608 AV *av = (AV *)SvRV (val);
1609 int i, len = av_len (av) + 1;
1610 HV *stash = gv_stashsv (tag, 0);
1611 SV *sv;
1612
1613 if (!stash)
1614 ERR ("cannot decode perl-object (package does not exist)");
1615
1616 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1617
1618 if (!method)
1619 ERR ("cannot decode perl-object (package does not have a THAW method)");
1620
1621 dSP;
1622
1623 ENTER; SAVETMPS;
1624 PUSHMARK (SP);
1625 EXTEND (SP, len + 2);
1626 // we re-bless the reference to get overload and other niceties right
1627 PUSHs (tag);
1628 PUSHs (sv_json);
1629
1630 for (i = 0; i < len; ++i)
1631 PUSHs (*av_fetch (av, i, 1));
1632
1633 PUTBACK;
1634 call_sv ((SV *)GvCV (method), G_SCALAR);
1635 SPAGAIN;
1636
1637 SvREFCNT_dec (tag);
1638 SvREFCNT_dec (val);
1639 sv = SvREFCNT_inc (POPs);
1640
1641 PUTBACK;
1642
1643 FREETMPS; LEAVE;
1644
1645 return sv;
1646 }
1647
1648fail:
1649 SvREFCNT_dec (tag);
1650 SvREFCNT_dec (val);
1651 return 0;
1652}
1653
1654static SV *
1345decode_sv (dec_t *dec) 1655decode_sv (dec_t *dec)
1346{ 1656{
1347 // the beauty of JSON: you need exactly one character lookahead 1657 // the beauty of JSON: you need exactly one character lookahead
1348 // to parse everything. 1658 // to parse everything.
1349 switch (*dec->cur) 1659 switch (*dec->cur)
1350 { 1660 {
1351 case '"': ++dec->cur; return decode_str (dec); 1661 case '"': ++dec->cur; return decode_str (dec);
1352 case '[': ++dec->cur; return decode_av (dec); 1662 case '[': ++dec->cur; return decode_av (dec);
1353 case '{': ++dec->cur; return decode_hv (dec); 1663 case '{': ++dec->cur; return decode_hv (dec);
1664 case '(': return decode_tag (dec);
1354 1665
1355 case '-': 1666 case '-':
1356 case '0': case '1': case '2': case '3': case '4': 1667 case '0': case '1': case '2': case '3': case '4':
1357 case '5': case '6': case '7': case '8': case '9': 1668 case '5': case '6': case '7': case '8': case '9':
1358 return decode_num (dec); 1669 return decode_num (dec);
1360 case 't': 1671 case 't':
1361 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1672 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1362 { 1673 {
1363 dec->cur += 4; 1674 dec->cur += 4;
1364#if JSON_SLOW 1675#if JSON_SLOW
1365 json_true = get_bool ("JSON::XS::true"); 1676 bool_true = get_bool ("Types::Serialiser::true");
1366#endif 1677#endif
1367 return newSVsv (json_true); 1678 return newSVsv (bool_true);
1368 } 1679 }
1369 else 1680 else
1370 ERR ("'true' expected"); 1681 ERR ("'true' expected");
1371 1682
1372 break; 1683 break;
1374 case 'f': 1685 case 'f':
1375 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1686 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1376 { 1687 {
1377 dec->cur += 5; 1688 dec->cur += 5;
1378#if JSON_SLOW 1689#if JSON_SLOW
1379 json_false = get_bool ("JSON::XS::false"); 1690 bool_false = get_bool ("Types::Serialiser::false");
1380#endif 1691#endif
1381 return newSVsv (json_false); 1692 return newSVsv (bool_false);
1382 } 1693 }
1383 else 1694 else
1384 ERR ("'false' expected"); 1695 ERR ("'false' expected");
1385 1696
1386 break; 1697 break;
1395 ERR ("'null' expected"); 1706 ERR ("'null' expected");
1396 1707
1397 break; 1708 break;
1398 1709
1399 default: 1710 default:
1400 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1711 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1401 break; 1712 break;
1402 } 1713 }
1403 1714
1404fail: 1715fail:
1405 return 0; 1716 return 0;
1407 1718
1408static SV * 1719static SV *
1409decode_json (SV *string, JSON *json, STRLEN *offset_return) 1720decode_json (SV *string, JSON *json, STRLEN *offset_return)
1410{ 1721{
1411 dec_t dec; 1722 dec_t dec;
1412 STRLEN offset;
1413 SV *sv; 1723 SV *sv;
1414 1724
1725 /* work around bugs in 5.10 where manipulating magic values
1726 * makes perl ignore the magic in subsequent accesses.
1727 * also make a copy of non-PV values, to get them into a clean
1728 * state (SvPV should do that, but it's buggy, see below).
1729 *
1730 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1731 * as reported by Reini Urban.
1732 */
1415 SvGETMAGIC (string); 1733 /*SvGETMAGIC (string);*/
1734 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1735 string = sv_2mortal (newSVsv (string));
1736
1416 SvUPGRADE (string, SVt_PV); 1737 SvUPGRADE (string, SVt_PV);
1417 1738
1418 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1739 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1419 * assertion with -DDEBUGGING, although SvCUR is documented to 1740 * assertion with -DDEBUGGING, although SvCUR is documented to
1420 * return the xpv_cur field which certainly exists after upgrading. 1741 * return the xpv_cur field which certainly exists after upgrading.
1423 * and hope for the best. 1744 * and hope for the best.
1424 * Damnit, SvPV_nolen still trips over yet another assertion. This 1745 * Damnit, SvPV_nolen still trips over yet another assertion. This
1425 * assertion business is seriously broken, try yet another workaround 1746 * assertion business is seriously broken, try yet another workaround
1426 * for the broken -DDEBUGGING. 1747 * for the broken -DDEBUGGING.
1427 */ 1748 */
1749 {
1428#ifdef DEBUGGING 1750#ifdef DEBUGGING
1429 offset = SvOK (string) ? sv_len (string) : 0; 1751 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1430#else 1752#else
1431 offset = SvCUR (string); 1753 STRLEN offset = SvCUR (string);
1432#endif 1754#endif
1433 1755
1434 if (offset > json->max_size && json->max_size) 1756 if (offset > json->max_size && json->max_size)
1435 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1757 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1436 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1758 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1759 }
1437 1760
1438 if (json->flags & F_UTF8) 1761 if (DECODE_WANTS_OCTETS (json))
1439 sv_utf8_downgrade (string, 0); 1762 sv_utf8_downgrade (string, 0);
1440 else 1763 else
1441 sv_utf8_upgrade (string); 1764 sv_utf8_upgrade (string);
1442 1765
1443 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1766 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1454 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1777 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1455 1778
1456 decode_ws (&dec); 1779 decode_ws (&dec);
1457 sv = decode_sv (&dec); 1780 sv = decode_sv (&dec);
1458 1781
1459 if (!(offset_return || !sv)) 1782 if (offset_return)
1783 *offset_return = dec.cur - SvPVX (string);
1784 else if (sv)
1460 { 1785 {
1461 // check for trailing garbage 1786 // check for trailing garbage
1462 decode_ws (&dec); 1787 decode_ws (&dec);
1463 1788
1464 if (*dec.cur) 1789 if (*dec.cur)
1465 { 1790 {
1466 dec.err = "garbage after JSON object"; 1791 dec.err = "garbage after JSON object";
1467 SvREFCNT_dec (sv); 1792 SvREFCNT_dec (sv);
1468 sv = 0; 1793 sv = 0;
1469 } 1794 }
1470 }
1471
1472 if (offset_return || !sv)
1473 {
1474 offset = dec.json.flags & F_UTF8
1475 ? dec.cur - SvPVX (string)
1476 : utf8_distance (dec.cur, SvPVX (string));
1477
1478 if (offset_return)
1479 *offset_return = offset;
1480 } 1795 }
1481 1796
1482 if (!sv) 1797 if (!sv)
1483 { 1798 {
1484 SV *uni = sv_newmortal (); 1799 SV *uni = sv_newmortal ();
1490 SAVEVPTR (PL_curcop); 1805 SAVEVPTR (PL_curcop);
1491 PL_curcop = &cop; 1806 PL_curcop = &cop;
1492 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1807 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1493 LEAVE; 1808 LEAVE;
1494 1809
1495 croak ("%s, at character offset %d [\"%s\"]", 1810 croak ("%s, at character offset %d (before \"%s\")",
1496 dec.err, 1811 dec.err,
1497 (int)offset, 1812 (int)ptr_to_index (string, dec.cur),
1498 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1813 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1499 } 1814 }
1500 1815
1501 sv = sv_2mortal (sv); 1816 sv = sv_2mortal (sv);
1502 1817
1503 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1818 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1504 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1819 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1505 1820
1506 return sv; 1821 return sv;
1507} 1822}
1508 1823
1511 1826
1512static void 1827static void
1513incr_parse (JSON *self) 1828incr_parse (JSON *self)
1514{ 1829{
1515 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1830 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1831
1832 // the state machine here is a bit convoluted and could be simplified a lot
1833 // but this would make it slower, so...
1516 1834
1517 for (;;) 1835 for (;;)
1518 { 1836 {
1519 //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 1837 //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
1520 switch (self->incr_mode) 1838 switch (self->incr_mode)
1521 { 1839 {
1522 // only used for intiial whitespace skipping 1840 // only used for initial whitespace skipping
1523 case INCR_M_WS: 1841 case INCR_M_WS:
1524 for (;;) 1842 for (;;)
1525 { 1843 {
1526 if (*p > 0x20) 1844 if (*p > 0x20)
1527 { 1845 {
1846 if (*p == '#')
1847 {
1848 self->incr_mode = INCR_M_C0;
1849 goto incr_m_c;
1850 }
1851 else
1852 {
1528 self->incr_mode = INCR_M_JSON; 1853 self->incr_mode = INCR_M_JSON;
1529 goto incr_m_json; 1854 goto incr_m_json;
1855 }
1530 } 1856 }
1531 else if (!*p) 1857 else if (!*p)
1532 goto interrupt; 1858 goto interrupt;
1533 1859
1534 ++p; 1860 ++p;
1540 goto interrupt; 1866 goto interrupt;
1541 1867
1542 ++p; 1868 ++p;
1543 self->incr_mode = INCR_M_STR; 1869 self->incr_mode = INCR_M_STR;
1544 goto incr_m_str; 1870 goto incr_m_str;
1871
1872 // inside #-style comments
1873 case INCR_M_C0:
1874 case INCR_M_C1:
1875 incr_m_c:
1876 for (;;)
1877 {
1878 if (*p == '\n')
1879 {
1880 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1881 break;
1882 }
1883 else if (!*p)
1884 goto interrupt;
1885
1886 ++p;
1887 }
1888
1889 break;
1545 1890
1546 // inside a string 1891 // inside a string
1547 case INCR_M_STR: 1892 case INCR_M_STR:
1548 incr_m_str: 1893 incr_m_str:
1549 for (;;) 1894 for (;;)
1600 self->incr_mode = INCR_M_STR; 1945 self->incr_mode = INCR_M_STR;
1601 goto incr_m_str; 1946 goto incr_m_str;
1602 1947
1603 case '[': 1948 case '[':
1604 case '{': 1949 case '{':
1950 case '(':
1605 if (++self->incr_nest > self->max_depth) 1951 if (++self->incr_nest > self->max_depth)
1606 croak (ERR_NESTING_EXCEEDED); 1952 croak (ERR_NESTING_EXCEEDED);
1607 break; 1953 break;
1608 1954
1609 case ']': 1955 case ']':
1610 case '}': 1956 case '}':
1611 if (--self->incr_nest <= 0) 1957 if (--self->incr_nest <= 0)
1612 goto interrupt; 1958 goto interrupt;
1959 break;
1960
1961 case ')':
1962 --self->incr_nest;
1963 break;
1964
1965 case '#':
1966 self->incr_mode = INCR_M_C1;
1967 goto incr_m_c;
1613 } 1968 }
1614 } 1969 }
1615 } 1970 }
1616 1971
1617 modechange: 1972 modechange:
1618 ; 1973 ;
1619 } 1974 }
1620 1975
1621interrupt: 1976interrupt:
1622 self->incr_pos = p - SvPVX (self->incr_text); 1977 self->incr_pos = p - SvPVX (self->incr_text);
1978 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1623 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1979 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1624} 1980}
1625 1981
1626///////////////////////////////////////////////////////////////////////////// 1982/////////////////////////////////////////////////////////////////////////////
1627// XS interface functions 1983// XS interface functions
1637 i >= '0' && i <= '9' ? i - '0' 1993 i >= '0' && i <= '9' ? i - '0'
1638 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1994 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1639 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1995 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1640 : -1; 1996 : -1;
1641 1997
1642 json_stash = gv_stashpv ("JSON::XS" , 1); 1998 json_stash = gv_stashpv ("JSON::XS" , 1);
1643 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1999 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1644
1645 json_true = get_bool ("JSON::XS::true"); 2000 bool_true = get_bool ("Types::Serialiser::true");
1646 json_false = get_bool ("JSON::XS::false"); 2001 bool_false = get_bool ("Types::Serialiser::false");
2002
2003 sv_json = newSVpv ("JSON", 0);
2004 SvREADONLY_on (sv_json);
2005
2006 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1647} 2007}
1648 2008
1649PROTOTYPES: DISABLE 2009PROTOTYPES: DISABLE
1650 2010
1651void CLONE (...) 2011void CLONE (...)
1652 CODE: 2012 CODE:
1653 json_stash = 0; 2013 json_stash = 0;
1654 json_boolean_stash = 0; 2014 bool_stash = 0;
1655 2015
1656void new (char *klass) 2016void new (char *klass)
1657 PPCODE: 2017 PPCODE:
1658{ 2018{
1659 SV *pv = NEWSV (0, sizeof (JSON)); 2019 SV *pv = NEWSV (0, sizeof (JSON));
1660 SvPOK_only (pv); 2020 SvPOK_only (pv);
1661 json_init ((JSON *)SvPVX (pv)); 2021 json_init ((JSON *)SvPVX (pv));
1662 XPUSHs (sv_2mortal (sv_bless ( 2022 XPUSHs (sv_2mortal (sv_bless (
1663 newRV_noinc (pv), 2023 newRV_noinc (pv),
1664 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2024 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1679 shrink = F_SHRINK 2039 shrink = F_SHRINK
1680 allow_blessed = F_ALLOW_BLESSED 2040 allow_blessed = F_ALLOW_BLESSED
1681 convert_blessed = F_CONV_BLESSED 2041 convert_blessed = F_CONV_BLESSED
1682 relaxed = F_RELAXED 2042 relaxed = F_RELAXED
1683 allow_unknown = F_ALLOW_UNKNOWN 2043 allow_unknown = F_ALLOW_UNKNOWN
2044 allow_tags = F_ALLOW_TAGS
1684 PPCODE: 2045 PPCODE:
1685{ 2046{
1686 if (enable) 2047 if (enable)
1687 self->flags |= ix; 2048 self->flags |= ix;
1688 else 2049 else
1704 get_shrink = F_SHRINK 2065 get_shrink = F_SHRINK
1705 get_allow_blessed = F_ALLOW_BLESSED 2066 get_allow_blessed = F_ALLOW_BLESSED
1706 get_convert_blessed = F_CONV_BLESSED 2067 get_convert_blessed = F_CONV_BLESSED
1707 get_relaxed = F_RELAXED 2068 get_relaxed = F_RELAXED
1708 get_allow_unknown = F_ALLOW_UNKNOWN 2069 get_allow_unknown = F_ALLOW_UNKNOWN
2070 get_allow_tags = F_ALLOW_TAGS
1709 PPCODE: 2071 PPCODE:
1710 XPUSHs (boolSV (self->flags & ix)); 2072 XPUSHs (boolSV (self->flags & ix));
1711 2073
1712void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2074void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1713 PPCODE: 2075 PPCODE:
1741} 2103}
1742 2104
1743void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2105void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1744 PPCODE: 2106 PPCODE:
1745{ 2107{
1746 if (!self->cb_sk_object) 2108 if (!self->cb_sk_object)
1747 self->cb_sk_object = newHV (); 2109 self->cb_sk_object = newHV ();
1748 2110
1749 if (SvOK (cb)) 2111 if (SvOK (cb))
1750 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2112 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1751 else 2113 else
1762 XPUSHs (ST (0)); 2124 XPUSHs (ST (0));
1763} 2125}
1764 2126
1765void encode (JSON *self, SV *scalar) 2127void encode (JSON *self, SV *scalar)
1766 PPCODE: 2128 PPCODE:
1767 XPUSHs (encode_json (scalar, self)); 2129 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2130 XPUSHs (scalar);
1768 2131
1769void decode (JSON *self, SV *jsonstr) 2132void decode (JSON *self, SV *jsonstr)
1770 PPCODE: 2133 PPCODE:
1771 XPUSHs (decode_json (jsonstr, self, 0)); 2134 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2135 XPUSHs (jsonstr);
1772 2136
1773void decode_prefix (JSON *self, SV *jsonstr) 2137void decode_prefix (JSON *self, SV *jsonstr)
1774 PPCODE: 2138 PPCODE:
1775{ 2139{
2140 SV *sv;
1776 STRLEN offset; 2141 STRLEN offset;
2142 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1777 EXTEND (SP, 2); 2143 EXTEND (SP, 2);
1778 PUSHs (decode_json (jsonstr, self, &offset)); 2144 PUSHs (sv);
1779 PUSHs (sv_2mortal (newSVuv (offset))); 2145 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1780} 2146}
1781 2147
1782void incr_parse (JSON *self, SV *jsonstr = 0) 2148void incr_parse (JSON *self, SV *jsonstr = 0)
1783 PPCODE: 2149 PPCODE:
1784{ 2150{
1785 if (!self->incr_text) 2151 if (!self->incr_text)
1786 self->incr_text = newSVpvn ("", 0); 2152 self->incr_text = newSVpvn ("", 0);
2153
2154 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2155 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2156 if (DECODE_WANTS_OCTETS (self))
2157 {
2158 if (self->incr_pos)
2159 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2160 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2161
2162 sv_utf8_downgrade (self->incr_text, 0);
2163 }
2164 else
2165 {
2166 sv_utf8_upgrade (self->incr_text);
2167
2168 if (self->incr_pos)
2169 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2170 - (U8 *)SvPVX (self->incr_text);
2171 }
1787 2172
1788 // append data, if any 2173 // append data, if any
1789 if (jsonstr) 2174 if (jsonstr)
1790 { 2175 {
2176 /* make sure both strings have same encoding */
1791 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2177 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2178 if (SvUTF8 (jsonstr))
2179 sv_utf8_downgrade (jsonstr, 0);
1792 { 2180 else
1793 /* utf-8-ness differs, need to upgrade */
1794 sv_utf8_upgrade (self->incr_text); 2181 sv_utf8_upgrade (jsonstr);
1795 2182
1796 if (self->incr_pos) 2183 /* and then just blindly append */
1797 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1798 - (U8 *)SvPVX (self->incr_text);
1799 }
1800
1801 { 2184 {
1802 STRLEN len; 2185 STRLEN len;
1803 const char *str = SvPV (jsonstr, len); 2186 const char *str = SvPV (jsonstr, len);
1804 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2187 STRLEN cur = SvCUR (self->incr_text);
2188
2189 if (SvLEN (self->incr_text) - cur <= len)
2190 json_sv_grow (self->incr_text, cur, len);
2191
1805 Move (str, SvEND (self->incr_text), len, char); 2192 Move (str, SvEND (self->incr_text), len, char);
1806 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2193 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1807 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2194 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1808 } 2195 }
1809 } 2196 }
1810 2197
1811 if (GIMME_V != G_VOID) 2198 if (GIMME_V != G_VOID)
1812 do 2199 do
1813 { 2200 {
2201 SV *sv;
1814 STRLEN offset; 2202 STRLEN offset;
1815 2203
1816 if (!INCR_DONE (self)) 2204 if (!INCR_DONE (self))
1817 { 2205 {
1818 incr_parse (self); 2206 incr_parse (self);
1820 if (self->incr_pos > self->max_size && self->max_size) 2208 if (self->incr_pos > self->max_size && self->max_size)
1821 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2209 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1822 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2210 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1823 2211
1824 if (!INCR_DONE (self)) 2212 if (!INCR_DONE (self))
2213 {
2214 // as an optimisation, do not accumulate white space in the incr buffer
2215 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2216 {
2217 self->incr_pos = 0;
2218 SvCUR_set (self->incr_text, 0);
2219 }
2220
1825 break; 2221 break;
2222 }
1826 } 2223 }
1827 2224
1828 XPUSHs (decode_json (self->incr_text, self, &offset)); 2225 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2226 XPUSHs (sv);
1829 2227
1830 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1831 self->incr_pos -= offset; 2228 self->incr_pos -= offset;
1832 self->incr_nest = 0; 2229 self->incr_nest = 0;
1833 self->incr_mode = 0; 2230 self->incr_mode = 0;
2231
2232 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1834 } 2233 }
1835 while (GIMME_V == G_ARRAY); 2234 while (GIMME_V == G_ARRAY);
1836} 2235}
1837 2236
1838SV *incr_text (JSON *self) 2237SV *incr_text (JSON *self)
1876 SvREFCNT_dec (self->incr_text); 2275 SvREFCNT_dec (self->incr_text);
1877 2276
1878PROTOTYPES: ENABLE 2277PROTOTYPES: ENABLE
1879 2278
1880void encode_json (SV *scalar) 2279void encode_json (SV *scalar)
1881 ALIAS:
1882 to_json_ = 0
1883 encode_json = F_UTF8
1884 PPCODE: 2280 PPCODE:
1885{ 2281{
1886 JSON json; 2282 JSON json;
1887 json_init (&json); 2283 json_init (&json);
1888 json.flags |= ix; 2284 json.flags |= F_UTF8;
1889 XPUSHs (encode_json (scalar, &json)); 2285 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2286 XPUSHs (scalar);
1890} 2287}
1891 2288
1892void decode_json (SV *jsonstr) 2289void decode_json (SV *jsonstr)
1893 ALIAS:
1894 from_json_ = 0
1895 decode_json = F_UTF8
1896 PPCODE: 2290 PPCODE:
1897{ 2291{
1898 JSON json; 2292 JSON json;
1899 json_init (&json); 2293 json_init (&json);
1900 json.flags |= ix; 2294 json.flags |= F_UTF8;
1901 XPUSHs (decode_json (jsonstr, &json, 0)); 2295 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2296 XPUSHs (jsonstr);
1902} 2297}
1903 2298
1904

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines