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 1.102 by root, Sat Oct 10 01:48:50 2009 UTC vs.
Revision 1.134 by root, Thu Nov 15 20:13:03 2018 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// guarantees, 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
21#endif
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
20#endif 32#endif
21 33
22// three extra for rounding, sign, and end of string 34// three extra for rounding, sign, and end of string
23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
24 36
33#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
34#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
35#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
36#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
37#define F_ALLOW_UNKNOWN 0x00002000UL 49#define F_ALLOW_UNKNOWN 0x00002000UL
50#define F_ALLOW_TAGS 0x00004000UL
38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39 52
40#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
41 54
42#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
43#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
44 57
45#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)
46 61
47#define SB do { 62#define SB do {
48#define SE } while (0) 63#define SE } while (0)
49 64
50#if __GNUC__ >= 3 65#if __GNUC__ >= 3
63 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
64 79
65#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?)"
66 81
67#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
68# define JSON_SLOW 1
69# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
84# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
85# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
70#else 86#else
71# define JSON_SLOW 0
72# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
73#endif 90#endif
74 91
75static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
76static 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_false, *bool_true;
97static SV *sv_json;
77 98
78enum { 99enum {
79 INCR_M_WS = 0, // initial whitespace skipping, must be 0 100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
80 INCR_M_STR, // inside string 101 INCR_M_STR, // inside string
81 INCR_M_BS, // inside backslash 102 INCR_M_BS, // inside backslash
103 INCR_M_C0, // inside comment in initial whitespace sequence
104 INCR_M_C1, // inside comment in other places
82 INCR_M_JSON // outside anything, count nesting 105 INCR_M_JSON // outside anything, count nesting
83}; 106};
84 107
85#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) 108#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
86 109
95 // for the incremental parser 118 // for the incremental parser
96 SV *incr_text; // the source text so far 119 SV *incr_text; // the source text so far
97 STRLEN incr_pos; // the current offset into the text 120 STRLEN incr_pos; // the current offset into the text
98 int incr_nest; // {[]}-nesting level 121 int incr_nest; // {[]}-nesting level
99 unsigned char incr_mode; 122 unsigned char incr_mode;
123
124 SV *v_false, *v_true;
100} JSON; 125} JSON;
101 126
102INLINE void 127INLINE void
103json_init (JSON *json) 128json_init (JSON *json)
104{ 129{
133 SvPV_renew (sv, SvCUR (sv) + 1); 158 SvPV_renew (sv, SvCUR (sv) + 1);
134#endif 159#endif
135 } 160 }
136} 161}
137 162
163/* adds two STRLENs together, slow, and with paranoia */
164STRLEN
165strlen_sum (STRLEN l1, STRLEN l2)
166{
167 size_t sum = l1 + l2;
168
169 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
170 croak ("JSON::XS: string size overflow");
171
172 return sum;
173}
174
175/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
176static char *
177json_sv_grow (SV *sv, size_t len1, size_t len2)
178{
179 len1 = strlen_sum (len1, len2);
180 len1 = strlen_sum (len1, len1 >> 1);
181
182 if (len1 > 4096 - 24)
183 len1 = (len1 | 4095) - 24;
184
185 return SvGROW (sv, len1);
186}
187
138// decode an utf-8 character and return it, or (UV)-1 in 188// decode a utf-8 character and return it, or (UV)-1 in
139// case of an error. 189// case of an error.
140// we special-case "safe" characters from U+80 .. U+7FF, 190// we special-case "safe" characters from U+80 .. U+7FF,
141// but use the very good perl function to parse anything else. 191// but use the very good perl function to parse anything else.
142// note that we never call this function for a ascii codepoints 192// note that we never call this function for a ascii codepoints
143INLINE UV 193INLINE UV
186 ? utf8_distance (offset, SvPVX (sv)) 236 ? utf8_distance (offset, SvPVX (sv))
187 : offset - SvPVX (sv); 237 : offset - SvPVX (sv);
188} 238}
189 239
190///////////////////////////////////////////////////////////////////////////// 240/////////////////////////////////////////////////////////////////////////////
241// fp hell
242
243// scan a group of digits, and a trailing exponent
244static void
245json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
246{
247 UV uaccum = 0;
248 int eaccum = 0;
249
250 // if we recurse too deep, skip all remaining digits
251 // to avoid a stack overflow attack
252 if (expect_false (--maxdepth <= 0))
253 while (((U8)*s - '0') < 10)
254 ++s;
255
256 for (;;)
257 {
258 U8 dig = (U8)*s - '0';
259
260 if (expect_false (dig >= 10))
261 {
262 if (dig == (U8)((U8)'.' - (U8)'0'))
263 {
264 ++s;
265 json_atof_scan1 (s, accum, expo, 1, maxdepth);
266 }
267 else if ((dig | ' ') == 'e' - '0')
268 {
269 int exp2 = 0;
270 int neg = 0;
271
272 ++s;
273
274 if (*s == '-')
275 {
276 ++s;
277 neg = 1;
278 }
279 else if (*s == '+')
280 ++s;
281
282 while ((dig = (U8)*s - '0') < 10)
283 exp2 = exp2 * 10 + *s++ - '0';
284
285 *expo += neg ? -exp2 : exp2;
286 }
287
288 break;
289 }
290
291 ++s;
292
293 uaccum = uaccum * 10 + dig;
294 ++eaccum;
295
296 // if we have too many digits, then recurse for more
297 // we actually do this for rather few digits
298 if (uaccum >= (UV_MAX - 9) / 10)
299 {
300 if (postdp) *expo -= eaccum;
301 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
302 if (postdp) *expo += eaccum;
303
304 break;
305 }
306 }
307
308 // this relies greatly on the quality of the pow ()
309 // implementation of the platform, but a good
310 // implementation is hard to beat.
311 // (IEEE 754 conformant ones are required to be exact)
312 if (postdp) *expo -= eaccum;
313 *accum += uaccum * Perl_pow (10., *expo);
314 *expo += eaccum;
315}
316
317static NV
318json_atof (const char *s)
319{
320 NV accum = 0.;
321 int expo = 0;
322 int neg = 0;
323
324 if (*s == '-')
325 {
326 ++s;
327 neg = 1;
328 }
329
330 // a recursion depth of ten gives us >>500 bits
331 json_atof_scan1 (s, &accum, &expo, 0, 10);
332
333 return neg ? -accum : accum;
334}
335
336// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
337static int
338ref_bool_type (SV *sv)
339{
340 svtype svt = SvTYPE (sv);
341
342 if (svt < SVt_PVAV)
343 {
344 STRLEN len = 0;
345 char *pv = svt ? SvPV (sv, len) : 0;
346
347 if (len == 1)
348 if (*pv == '1')
349 return 1;
350 else if (*pv == '0')
351 return 0;
352 }
353
354 return -1;
355}
356
357// returns whether scalar is not a reference in the sense of allow_nonref
358static int
359json_nonref (SV *scalar)
360{
361 if (!SvROK (scalar))
362 return 1;
363
364 scalar = SvRV (scalar);
365
366 if (SvTYPE (scalar) >= SVt_PVMG)
367 {
368 if (SvSTASH (scalar) == bool_stash)
369 return 1;
370
371 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
372 return 1;
373 }
374
375 return 0;
376}
377
378/////////////////////////////////////////////////////////////////////////////
191// encoder 379// encoder
192 380
193// structure used for encoding JSON 381// structure used for encoding JSON
194typedef struct 382typedef struct
195{ 383{
202} enc_t; 390} enc_t;
203 391
204INLINE void 392INLINE void
205need (enc_t *enc, STRLEN len) 393need (enc_t *enc, STRLEN len)
206{ 394{
207 if (expect_false (enc->cur + len >= enc->end)) 395 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
208 { 396 {
209 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 397 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
210 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 398 char *buf = json_sv_grow (enc->sv, cur, len);
211 enc->cur = SvPVX (enc->sv) + cur; 399 enc->cur = buf + cur;
212 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 400 enc->end = buf + SvLEN (enc->sv) - 1;
213 } 401 }
214} 402}
215 403
216INLINE void 404INLINE void
217encode_ch (enc_t *enc, char ch) 405encode_ch (enc_t *enc, char ch)
233 421
234 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 422 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
235 { 423 {
236 if (expect_false (ch == '"')) // but with slow exceptions 424 if (expect_false (ch == '"')) // but with slow exceptions
237 { 425 {
238 need (enc, len += 1); 426 need (enc, len + 1);
239 *enc->cur++ = '\\'; 427 *enc->cur++ = '\\';
240 *enc->cur++ = '"'; 428 *enc->cur++ = '"';
241 } 429 }
242 else if (expect_false (ch == '\\')) 430 else if (expect_false (ch == '\\'))
243 { 431 {
244 need (enc, len += 1); 432 need (enc, len + 1);
245 *enc->cur++ = '\\'; 433 *enc->cur++ = '\\';
246 *enc->cur++ = '\\'; 434 *enc->cur++ = '\\';
247 } 435 }
248 else 436 else
249 *enc->cur++ = ch; 437 *enc->cur++ = ch;
252 } 440 }
253 else 441 else
254 { 442 {
255 switch (ch) 443 switch (ch)
256 { 444 {
257 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 445 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
258 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 446 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
259 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 447 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
260 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 448 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
261 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 449 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
262 450
263 default: 451 default:
264 { 452 {
265 STRLEN clen; 453 STRLEN clen;
266 UV uch; 454 UV uch;
282 if (uch >= 0x10000UL) 470 if (uch >= 0x10000UL)
283 { 471 {
284 if (uch >= 0x110000UL) 472 if (uch >= 0x110000UL)
285 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 473 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
286 474
287 need (enc, len += 11); 475 need (enc, len + 11);
288 sprintf (enc->cur, "\\u%04x\\u%04x", 476 sprintf (enc->cur, "\\u%04x\\u%04x",
289 (int)((uch - 0x10000) / 0x400 + 0xD800), 477 (int)((uch - 0x10000) / 0x400 + 0xD800),
290 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 478 (int)((uch - 0x10000) % 0x400 + 0xDC00));
291 enc->cur += 12; 479 enc->cur += 12;
292 } 480 }
293 else 481 else
294 { 482 {
295 need (enc, len += 5); 483 need (enc, len + 5);
296 *enc->cur++ = '\\'; 484 *enc->cur++ = '\\';
297 *enc->cur++ = 'u'; 485 *enc->cur++ = 'u';
298 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 486 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
299 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 487 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
300 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 488 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
308 *enc->cur++ = uch; 496 *enc->cur++ = uch;
309 str += clen; 497 str += clen;
310 } 498 }
311 else if (is_utf8) 499 else if (is_utf8)
312 { 500 {
313 need (enc, len += clen); 501 need (enc, len + clen);
314 do 502 do
315 { 503 {
316 *enc->cur++ = *str++; 504 *enc->cur++ = *str++;
317 } 505 }
318 while (--clen); 506 while (--clen);
319 } 507 }
320 else 508 else
321 { 509 {
322 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 510 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
323 enc->cur = encode_utf8 (enc->cur, uch); 511 enc->cur = encode_utf8 (enc->cur, uch);
324 ++str; 512 ++str;
325 } 513 }
326 } 514 }
327 } 515 }
381 569
382 if (enc->indent >= enc->json.max_depth) 570 if (enc->indent >= enc->json.max_depth)
383 croak (ERR_NESTING_EXCEEDED); 571 croak (ERR_NESTING_EXCEEDED);
384 572
385 encode_ch (enc, '['); 573 encode_ch (enc, '[');
386 574
387 if (len >= 0) 575 if (len >= 0)
388 { 576 {
389 encode_nl (enc); ++enc->indent; 577 encode_nl (enc); ++enc->indent;
390 578
391 for (i = 0; i <= len; ++i) 579 for (i = 0; i <= len; ++i)
403 encode_comma (enc); 591 encode_comma (enc);
404 } 592 }
405 593
406 encode_nl (enc); --enc->indent; encode_indent (enc); 594 encode_nl (enc); --enc->indent; encode_indent (enc);
407 } 595 }
408 596
409 encode_ch (enc, ']'); 597 encode_ch (enc, ']');
410} 598}
411 599
412static void 600static void
413encode_hk (enc_t *enc, HE *he) 601encode_hk (enc_t *enc, HE *he)
417 if (HeKLEN (he) == HEf_SVKEY) 605 if (HeKLEN (he) == HEf_SVKEY)
418 { 606 {
419 SV *sv = HeSVKEY (he); 607 SV *sv = HeSVKEY (he);
420 STRLEN len; 608 STRLEN len;
421 char *str; 609 char *str;
422 610
423 SvGETMAGIC (sv); 611 SvGETMAGIC (sv);
424 str = SvPV (sv, len); 612 str = SvPV (sv, len);
425 613
426 encode_str (enc, str, len, SvUTF8 (sv)); 614 encode_str (enc, str, len, SvUTF8 (sv));
427 } 615 }
493 } 681 }
494 682
495 if (count) 683 if (count)
496 { 684 {
497 int i, fast = 1; 685 int i, fast = 1;
498#if defined(__BORLANDC__) || defined(_MSC_VER) 686 HE *hes_stack [STACK_HES];
499 HE **hes = _alloca (count * sizeof (HE)); 687 HE **hes = hes_stack;
500#else 688
501 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 689 // allocate larger arrays on the heap
502#endif 690 if (count > STACK_HES)
691 {
692 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
693 hes = (HE **)SvPVX (sv);
694 }
503 695
504 i = 0; 696 i = 0;
505 while ((he = hv_iternext (hv))) 697 while ((he = hv_iternext (hv)))
506 { 698 {
507 hes [i++] = he; 699 hes [i++] = he;
576// encode objects, arrays and special \0=false and \1=true values. 768// encode objects, arrays and special \0=false and \1=true values.
577static void 769static void
578encode_rv (enc_t *enc, SV *sv) 770encode_rv (enc_t *enc, SV *sv)
579{ 771{
580 svtype svt; 772 svtype svt;
773 GV *method;
581 774
582 SvGETMAGIC (sv); 775 SvGETMAGIC (sv);
583 svt = SvTYPE (sv); 776 svt = SvTYPE (sv);
584 777
585 if (expect_false (SvOBJECT (sv))) 778 if (expect_false (SvOBJECT (sv)))
586 { 779 {
587 HV *stash = !JSON_SLOW || json_boolean_stash 780 HV *stash = SvSTASH (sv);
588 ? json_boolean_stash
589 : gv_stashpv ("JSON::XS::Boolean", 1);
590 781
591 if (SvSTASH (sv) == stash) 782 if (stash == bool_stash)
592 { 783 {
593 if (SvIV (sv))
594 encode_str (enc, "true", 4, 0); 784 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
595 else
596 encode_str (enc, "false", 5, 0); 785 else encode_str (enc, "false", 5, 0);
597 } 786 }
787 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
788 {
789 int count;
790 dSP;
791
792 ENTER; SAVETMPS;
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 if (count)
815 {
816 int i;
817
818 for (i = 0; i < count - 1; ++i)
819 {
820 encode_sv (enc, SP[i + 1 - count]);
821 encode_ch (enc, ',');
822 }
823
824 encode_sv (enc, TOPs);
825 SP -= count;
826 }
827
828 encode_ch (enc, ']');
829
830 FREETMPS; LEAVE;
831 }
832 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
833 {
834 dSP;
835
836 ENTER; SAVETMPS;
837 PUSHMARK (SP);
838 // we re-bless the reference to get overload and other niceties right
839 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
840
841 // calling with G_SCALAR ensures that we always get a 1 return value
842 PUTBACK;
843 call_sv ((SV *)GvCV (method), G_SCALAR);
844 SPAGAIN;
845
846 // catch this surprisingly common error
847 if (SvROK (TOPs) && SvRV (TOPs) == sv)
848 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
849
850 sv = POPs;
851 PUTBACK;
852
853 encode_sv (enc, sv);
854
855 FREETMPS; LEAVE;
856 }
857 else if (enc->json.flags & F_ALLOW_BLESSED)
858 encode_str (enc, "null", 4, 0);
598 else 859 else
599 {
600#if 0
601 if (0 && sv_derived_from (rv, "JSON::Literal"))
602 {
603 // not yet
604 }
605#endif
606 if (enc->json.flags & F_CONV_BLESSED)
607 {
608 // we re-bless the reference to get overload and other niceties right
609 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
610
611 if (to_json)
612 {
613 dSP;
614
615 ENTER; SAVETMPS; PUSHMARK (SP);
616 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
617
618 // calling with G_SCALAR ensures that we always get a 1 return value
619 PUTBACK;
620 call_sv ((SV *)GvCV (to_json), G_SCALAR);
621 SPAGAIN;
622
623 // catch this surprisingly common error
624 if (SvROK (TOPs) && SvRV (TOPs) == sv)
625 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
626
627 sv = POPs;
628 PUTBACK;
629
630 encode_sv (enc, sv);
631
632 FREETMPS; LEAVE;
633 }
634 else if (enc->json.flags & F_ALLOW_BLESSED)
635 encode_str (enc, "null", 4, 0);
636 else
637 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
638 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
639 }
640 else if (enc->json.flags & F_ALLOW_BLESSED)
641 encode_str (enc, "null", 4, 0);
642 else
643 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 860 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
644 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 861 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
645 }
646 } 862 }
647 else if (svt == SVt_PVHV) 863 else if (svt == SVt_PVHV)
648 encode_hv (enc, (HV *)sv); 864 encode_hv (enc, (HV *)sv);
649 else if (svt == SVt_PVAV) 865 else if (svt == SVt_PVAV)
650 encode_av (enc, (AV *)sv); 866 encode_av (enc, (AV *)sv);
651 else if (svt < SVt_PVAV) 867 else if (svt < SVt_PVAV)
652 { 868 {
653 STRLEN len = 0; 869 int bool_type = ref_bool_type (sv);
654 char *pv = svt ? SvPV (sv, len) : 0;
655 870
656 if (len == 1 && *pv == '1') 871 if (bool_type == 1)
657 encode_str (enc, "true", 4, 0); 872 encode_str (enc, "true", 4, 0);
658 else if (len == 1 && *pv == '0') 873 else if (bool_type == 0)
659 encode_str (enc, "false", 5, 0); 874 encode_str (enc, "false", 5, 0);
660 else if (enc->json.flags & F_ALLOW_UNKNOWN) 875 else if (enc->json.flags & F_ALLOW_UNKNOWN)
661 encode_str (enc, "null", 4, 0); 876 encode_str (enc, "null", 4, 0);
662 else 877 else
663 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 878 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
726 } 941 }
727 else 942 else
728 { 943 {
729 // large integer, use the (rather slow) snprintf way. 944 // large integer, use the (rather slow) snprintf way.
730 need (enc, IVUV_MAXCHARS); 945 need (enc, IVUV_MAXCHARS);
731 enc->cur += 946 enc->cur +=
732 SvIsUV(sv) 947 SvIsUV(sv)
733 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 948 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
734 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 949 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
735 } 950 }
736 } 951 }
737 else if (SvROK (sv)) 952 else if (SvROK (sv))
738 encode_rv (enc, SvRV (sv)); 953 encode_rv (enc, SvRV (sv));
739 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 954 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
740 encode_str (enc, "null", 4, 0); 955 encode_str (enc, "null", 4, 0);
741 else 956 else
742 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 957 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
743 SvPV_nolen (sv), SvFLAGS (sv)); 958 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
744} 959}
745 960
746static SV * 961static SV *
747encode_json (SV *scalar, JSON *json) 962encode_json (SV *scalar, JSON *json)
748{ 963{
749 enc_t enc; 964 enc_t enc;
750 965
751 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 966 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
752 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 967 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
753 968
754 enc.json = *json; 969 enc.json = *json;
755 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 970 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
756 enc.cur = SvPVX (enc.sv); 971 enc.cur = SvPVX (enc.sv);
963 *cur++ = *dec_cur++; 1178 *cur++ = *dec_cur++;
964 while (--clen); 1179 while (--clen);
965 1180
966 utf8 = 1; 1181 utf8 = 1;
967 } 1182 }
1183 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1184 *cur++ = ch;
968 else 1185 else
969 { 1186 {
970 --dec_cur; 1187 --dec_cur;
971 1188
972 if (!ch) 1189 if (!ch)
982 1199
983 if (sv) 1200 if (sv)
984 { 1201 {
985 STRLEN cur = SvCUR (sv); 1202 STRLEN cur = SvCUR (sv);
986 1203
987 if (SvLEN (sv) <= cur + len) 1204 if (SvLEN (sv) - cur <= len)
988 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1205 json_sv_grow (sv, cur, len);
989 1206
990 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1207 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
991 SvCUR_set (sv, SvCUR (sv) + len); 1208 SvCUR_set (sv, SvCUR (sv) + len);
992 } 1209 }
993 else 1210 else
1117 } 1334 }
1118 1335
1119 len -= *start == '-' ? 1 : 0; 1336 len -= *start == '-' ? 1 : 0;
1120 1337
1121 // does not fit into IV or UV, try NV 1338 // does not fit into IV or UV, try NV
1122 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1339 if (len <= NV_DIG)
1123 #if defined (LDBL_DIG)
1124 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1125 #endif
1126 )
1127 // fits into NV without loss of precision 1340 // fits into NV without loss of precision
1128 return newSVnv (Atof (start)); 1341 return newSVnv (json_atof (start));
1129 1342
1130 // everything else fails, convert it to a string 1343 // everything else fails, convert it to a string
1131 return newSVpvn (start, dec->cur - start); 1344 return newSVpvn (start, dec->cur - start);
1132 } 1345 }
1133 1346
1134 // loss of precision here 1347 // loss of precision here
1135 return newSVnv (Atof (start)); 1348 return newSVnv (json_atof (start));
1136 1349
1137fail: 1350fail:
1138 return 0; 1351 return 0;
1139} 1352}
1140 1353
1164 if (*dec->cur == ']') 1377 if (*dec->cur == ']')
1165 { 1378 {
1166 ++dec->cur; 1379 ++dec->cur;
1167 break; 1380 break;
1168 } 1381 }
1169 1382
1170 if (*dec->cur != ',') 1383 if (*dec->cur != ',')
1171 ERR (", or ] expected while parsing array"); 1384 ERR (", or ] expected while parsing array");
1172 1385
1173 ++dec->cur; 1386 ++dec->cur;
1174 1387
1298 1511
1299 hv_iterinit (hv); 1512 hv_iterinit (hv);
1300 he = hv_iternext (hv); 1513 he = hv_iternext (hv);
1301 hv_iterinit (hv); 1514 hv_iterinit (hv);
1302 1515
1303 // the next line creates a mortal sv each time its called. 1516 // the next line creates a mortal sv each time it's called.
1304 // might want to optimise this for common cases. 1517 // might want to optimise this for common cases.
1305 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1518 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1306 1519
1307 if (cb) 1520 if (cb)
1308 { 1521 {
1309 dSP; 1522 dSP;
1310 int count; 1523 int count;
1311 1524
1312 ENTER; SAVETMPS; PUSHMARK (SP); 1525 ENTER; SAVETMPS;
1526 PUSHMARK (SP);
1313 XPUSHs (HeVAL (he)); 1527 XPUSHs (HeVAL (he));
1528 sv_2mortal (sv);
1314 1529
1315 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1530 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1316 1531
1317 if (count == 1) 1532 if (count == 1)
1318 { 1533 {
1319 sv = newSVsv (POPs); 1534 sv = newSVsv (POPs);
1320 FREETMPS; LEAVE; 1535 FREETMPS; LEAVE;
1321 return sv; 1536 return sv;
1322 } 1537 }
1538 else if (count)
1539 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1323 1540
1541 SvREFCNT_inc (sv);
1324 FREETMPS; LEAVE; 1542 FREETMPS; LEAVE;
1325 } 1543 }
1326 } 1544 }
1327 1545
1328 if (dec->json.cb_object) 1546 if (dec->json.cb_object)
1329 { 1547 {
1330 dSP; 1548 dSP;
1331 int count; 1549 int count;
1332 1550
1333 ENTER; SAVETMPS; PUSHMARK (SP); 1551 ENTER; SAVETMPS;
1552 PUSHMARK (SP);
1334 XPUSHs (sv_2mortal (sv)); 1553 XPUSHs (sv_2mortal (sv));
1335 1554
1336 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1555 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1337 1556
1338 if (count == 1) 1557 if (count == 1)
1339 { 1558 {
1340 sv = newSVsv (POPs); 1559 sv = newSVsv (POPs);
1341 FREETMPS; LEAVE; 1560 FREETMPS; LEAVE;
1342 return sv; 1561 return sv;
1343 } 1562 }
1563 else if (count)
1564 croak ("filter_json_object callbacks must not return more than one scalar");
1344 1565
1345 SvREFCNT_inc (sv); 1566 SvREFCNT_inc (sv);
1346 FREETMPS; LEAVE; 1567 FREETMPS; LEAVE;
1347 } 1568 }
1348 } 1569 }
1354 DEC_DEC_DEPTH; 1575 DEC_DEC_DEPTH;
1355 return 0; 1576 return 0;
1356} 1577}
1357 1578
1358static SV * 1579static SV *
1580decode_tag (dec_t *dec)
1581{
1582 SV *tag = 0;
1583 SV *val = 0;
1584
1585 if (!(dec->json.flags & F_ALLOW_TAGS))
1586 ERR ("malformed JSON string, neither array, object, number, string or atom");
1587
1588 ++dec->cur;
1589
1590 decode_ws (dec);
1591
1592 tag = decode_sv (dec);
1593 if (!tag)
1594 goto fail;
1595
1596 if (!SvPOK (tag))
1597 ERR ("malformed JSON string, (tag) must be a string");
1598
1599 decode_ws (dec);
1600
1601 if (*dec->cur != ')')
1602 ERR (") expected after tag");
1603
1604 ++dec->cur;
1605
1606 decode_ws (dec);
1607
1608 val = decode_sv (dec);
1609 if (!val)
1610 goto fail;
1611
1612 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1613 ERR ("malformed JSON string, tag value must be an array");
1614
1615 {
1616 AV *av = (AV *)SvRV (val);
1617 int i, len = av_len (av) + 1;
1618 HV *stash = gv_stashsv (tag, 0);
1619 SV *sv;
1620
1621 if (!stash)
1622 ERR ("cannot decode perl-object (package does not exist)");
1623
1624 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1625
1626 if (!method)
1627 ERR ("cannot decode perl-object (package does not have a THAW method)");
1628
1629 dSP;
1630
1631 ENTER; SAVETMPS;
1632 PUSHMARK (SP);
1633 EXTEND (SP, len + 2);
1634 // we re-bless the reference to get overload and other niceties right
1635 PUSHs (tag);
1636 PUSHs (sv_json);
1637
1638 for (i = 0; i < len; ++i)
1639 PUSHs (*av_fetch (av, i, 1));
1640
1641 PUTBACK;
1642 call_sv ((SV *)GvCV (method), G_SCALAR);
1643 SPAGAIN;
1644
1645 SvREFCNT_dec (tag);
1646 SvREFCNT_dec (val);
1647 sv = SvREFCNT_inc (POPs);
1648
1649 PUTBACK;
1650
1651 FREETMPS; LEAVE;
1652
1653 return sv;
1654 }
1655
1656fail:
1657 SvREFCNT_dec (tag);
1658 SvREFCNT_dec (val);
1659 return 0;
1660}
1661
1662static SV *
1359decode_sv (dec_t *dec) 1663decode_sv (dec_t *dec)
1360{ 1664{
1361 // the beauty of JSON: you need exactly one character lookahead 1665 // the beauty of JSON: you need exactly one character lookahead
1362 // to parse everything. 1666 // to parse everything.
1363 switch (*dec->cur) 1667 switch (*dec->cur)
1364 { 1668 {
1365 case '"': ++dec->cur; return decode_str (dec); 1669 case '"': ++dec->cur; return decode_str (dec);
1366 case '[': ++dec->cur; return decode_av (dec); 1670 case '[': ++dec->cur; return decode_av (dec);
1367 case '{': ++dec->cur; return decode_hv (dec); 1671 case '{': ++dec->cur; return decode_hv (dec);
1672 case '(': return decode_tag (dec);
1368 1673
1369 case '-': 1674 case '-':
1370 case '0': case '1': case '2': case '3': case '4': 1675 case '0': case '1': case '2': case '3': case '4':
1371 case '5': case '6': case '7': case '8': case '9': 1676 case '5': case '6': case '7': case '8': case '9':
1372 return decode_num (dec); 1677 return decode_num (dec);
1373 1678
1679 case 'f':
1680 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1681 {
1682 dec->cur += 5;
1683
1684 if (expect_false (!dec->json.v_false))
1685 dec->json.v_false = GET_BOOL (false);
1686
1687 return newSVsv (dec->json.v_false);
1688 }
1689 else
1690 ERR ("'false' expected");
1691
1692 break;
1693
1374 case 't': 1694 case 't':
1375 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1695 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1376 { 1696 {
1377 dec->cur += 4; 1697 dec->cur += 4;
1378#if JSON_SLOW 1698
1379 json_true = get_bool ("JSON::XS::true"); 1699 if (expect_false (!dec->json.v_true))
1380#endif 1700 dec->json.v_true = GET_BOOL (true);
1701
1381 return newSVsv (json_true); 1702 return newSVsv (dec->json.v_true);
1382 } 1703 }
1383 else 1704 else
1384 ERR ("'true' expected"); 1705 ERR ("'true' expected");
1385
1386 break;
1387
1388 case 'f':
1389 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1390 {
1391 dec->cur += 5;
1392#if JSON_SLOW
1393 json_false = get_bool ("JSON::XS::false");
1394#endif
1395 return newSVsv (json_false);
1396 }
1397 else
1398 ERR ("'false' expected");
1399 1706
1400 break; 1707 break;
1401 1708
1402 case 'n': 1709 case 'n':
1403 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1710 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1409 ERR ("'null' expected"); 1716 ERR ("'null' expected");
1410 1717
1411 break; 1718 break;
1412 1719
1413 default: 1720 default:
1414 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1721 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1415 break; 1722 break;
1416 } 1723 }
1417 1724
1418fail: 1725fail:
1419 return 0; 1726 return 0;
1420} 1727}
1421 1728
1422static SV * 1729static SV *
1423decode_json (SV *string, JSON *json, char **offset_return) 1730decode_json (SV *string, JSON *json, STRLEN *offset_return)
1424{ 1731{
1425 dec_t dec; 1732 dec_t dec;
1426 SV *sv; 1733 SV *sv;
1427 1734
1428 /* work around bugs in 5.10 where manipulating magic values 1735 /* work around bugs in 5.10 where manipulating magic values
1429 * will perl ignore the magic in subsequent accesses 1736 * makes perl ignore the magic in subsequent accesses.
1737 * also make a copy of non-PV values, to get them into a clean
1738 * state (SvPV should do that, but it's buggy, see below).
1739 *
1740 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1741 * as reported by Reini Urban.
1430 */ 1742 */
1431 /*SvGETMAGIC (string);*/ 1743 /*SvGETMAGIC (string);*/
1432 if (SvMAGICAL (string)) 1744 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1433 string = sv_2mortal (newSVsv (string)); 1745 string = sv_2mortal (newSVsv (string));
1434 1746
1435 SvUPGRADE (string, SVt_PV); 1747 SvUPGRADE (string, SVt_PV);
1436 1748
1437 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1749 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1454 if (offset > json->max_size && json->max_size) 1766 if (offset > json->max_size && json->max_size)
1455 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1767 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1456 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1768 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1457 } 1769 }
1458 1770
1459 if (json->flags & F_UTF8) 1771 if (DECODE_WANTS_OCTETS (json))
1460 sv_utf8_downgrade (string, 0); 1772 sv_utf8_downgrade (string, 0);
1461 else 1773 else
1462 sv_utf8_upgrade (string); 1774 sv_utf8_upgrade (string);
1463 1775
1464 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1776 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1476 1788
1477 decode_ws (&dec); 1789 decode_ws (&dec);
1478 sv = decode_sv (&dec); 1790 sv = decode_sv (&dec);
1479 1791
1480 if (offset_return) 1792 if (offset_return)
1481 *offset_return = dec.cur; 1793 *offset_return = dec.cur - SvPVX (string);
1482 1794 else if (sv)
1483 if (!(offset_return || !sv))
1484 { 1795 {
1485 // check for trailing garbage 1796 // check for trailing garbage
1486 decode_ws (&dec); 1797 decode_ws (&dec);
1487 1798
1488 if (*dec.cur) 1799 if (*dec.cur)
1506 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1817 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1507 LEAVE; 1818 LEAVE;
1508 1819
1509 croak ("%s, at character offset %d (before \"%s\")", 1820 croak ("%s, at character offset %d (before \"%s\")",
1510 dec.err, 1821 dec.err,
1511 ptr_to_index (string, dec.cur), 1822 (int)ptr_to_index (string, dec.cur),
1512 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1823 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1513 } 1824 }
1514 1825
1515 sv = sv_2mortal (sv); 1826 sv = sv_2mortal (sv);
1516 1827
1517 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1828 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1518 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1829 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1519 1830
1520 return sv; 1831 return sv;
1521} 1832}
1522 1833
1525 1836
1526static void 1837static void
1527incr_parse (JSON *self) 1838incr_parse (JSON *self)
1528{ 1839{
1529 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1840 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1841
1842 // the state machine here is a bit convoluted and could be simplified a lot
1843 // but this would make it slower, so...
1530 1844
1531 for (;;) 1845 for (;;)
1532 { 1846 {
1533 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D 1847 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1534 switch (self->incr_mode) 1848 switch (self->incr_mode)
1535 { 1849 {
1536 // only used for intiial whitespace skipping 1850 // only used for initial whitespace skipping
1537 case INCR_M_WS: 1851 case INCR_M_WS:
1538 for (;;) 1852 for (;;)
1539 { 1853 {
1540 if (*p > 0x20) 1854 if (*p > 0x20)
1541 { 1855 {
1856 if (*p == '#')
1857 {
1858 self->incr_mode = INCR_M_C0;
1859 goto incr_m_c;
1860 }
1861 else
1862 {
1542 self->incr_mode = INCR_M_JSON; 1863 self->incr_mode = INCR_M_JSON;
1543 goto incr_m_json; 1864 goto incr_m_json;
1865 }
1544 } 1866 }
1545 else if (!*p) 1867 else if (!*p)
1546 goto interrupt; 1868 goto interrupt;
1547 1869
1548 ++p; 1870 ++p;
1554 goto interrupt; 1876 goto interrupt;
1555 1877
1556 ++p; 1878 ++p;
1557 self->incr_mode = INCR_M_STR; 1879 self->incr_mode = INCR_M_STR;
1558 goto incr_m_str; 1880 goto incr_m_str;
1881
1882 // inside #-style comments
1883 case INCR_M_C0:
1884 case INCR_M_C1:
1885 incr_m_c:
1886 for (;;)
1887 {
1888 if (*p == '\n')
1889 {
1890 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1891 break;
1892 }
1893 else if (!*p)
1894 goto interrupt;
1895
1896 ++p;
1897 }
1898
1899 break;
1559 1900
1560 // inside a string 1901 // inside a string
1561 case INCR_M_STR: 1902 case INCR_M_STR:
1562 incr_m_str: 1903 incr_m_str:
1563 for (;;) 1904 for (;;)
1614 self->incr_mode = INCR_M_STR; 1955 self->incr_mode = INCR_M_STR;
1615 goto incr_m_str; 1956 goto incr_m_str;
1616 1957
1617 case '[': 1958 case '[':
1618 case '{': 1959 case '{':
1960 case '(':
1619 if (++self->incr_nest > self->max_depth) 1961 if (++self->incr_nest > self->max_depth)
1620 croak (ERR_NESTING_EXCEEDED); 1962 croak (ERR_NESTING_EXCEEDED);
1621 break; 1963 break;
1622 1964
1623 case ']': 1965 case ']':
1624 case '}': 1966 case '}':
1625 if (--self->incr_nest <= 0) 1967 if (--self->incr_nest <= 0)
1626 goto interrupt; 1968 goto interrupt;
1969 break;
1970
1971 case ')':
1972 --self->incr_nest;
1973 break;
1974
1975 case '#':
1976 self->incr_mode = INCR_M_C1;
1977 goto incr_m_c;
1627 } 1978 }
1628 } 1979 }
1629 } 1980 }
1630 1981
1631 modechange: 1982 modechange:
1632 ; 1983 ;
1633 } 1984 }
1634 1985
1635interrupt: 1986interrupt:
1636 self->incr_pos = p - SvPVX (self->incr_text); 1987 self->incr_pos = p - SvPVX (self->incr_text);
1988 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1637 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1989 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1638} 1990}
1639 1991
1640///////////////////////////////////////////////////////////////////////////// 1992/////////////////////////////////////////////////////////////////////////////
1641// XS interface functions 1993// XS interface functions
1651 i >= '0' && i <= '9' ? i - '0' 2003 i >= '0' && i <= '9' ? i - '0'
1652 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2004 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1653 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2005 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1654 : -1; 2006 : -1;
1655 2007
1656 json_stash = gv_stashpv ("JSON::XS" , 1); 2008 json_stash = gv_stashpv ("JSON::XS" , 1);
1657 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2009 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1658
1659 json_true = get_bool ("JSON::XS::true");
1660 json_false = get_bool ("JSON::XS::false"); 2010 bool_false = get_bool ("Types::Serialiser::false");
2011 bool_true = get_bool ("Types::Serialiser::true");
2012
2013 sv_json = newSVpv ("JSON", 0);
2014 SvREADONLY_on (sv_json);
1661 2015
1662 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ 2016 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1663} 2017}
1664 2018
1665PROTOTYPES: DISABLE 2019PROTOTYPES: DISABLE
1666 2020
1667void CLONE (...) 2021void CLONE (...)
1668 CODE: 2022 CODE:
2023 // as long as these writes are atomic, the race should not matter
2024 // as existing threads either already use 0, or use the old value,
2025 // which is sitll correct for the initial thread.
1669 json_stash = 0; 2026 json_stash = 0;
1670 json_boolean_stash = 0; 2027 bool_stash = 0;
2028 bool_false = 0;
2029 bool_true = 0;
1671 2030
1672void new (char *klass) 2031void new (char *klass)
1673 PPCODE: 2032 PPCODE:
1674{ 2033{
1675 SV *pv = NEWSV (0, sizeof (JSON)); 2034 SV *pv = NEWSV (0, sizeof (JSON));
1676 SvPOK_only (pv); 2035 SvPOK_only (pv);
1677 json_init ((JSON *)SvPVX (pv)); 2036 json_init ((JSON *)SvPVX (pv));
1678 XPUSHs (sv_2mortal (sv_bless ( 2037 XPUSHs (sv_2mortal (sv_bless (
1679 newRV_noinc (pv), 2038 newRV_noinc (pv),
1680 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2039 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1681 ))); 2040 )));
1682} 2041}
2042
2043void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2044 PPCODE:
2045 self->v_false = newSVsv (v_false);
2046 self->v_true = newSVsv (v_true);
2047 XPUSHs (ST (0));
2048
2049void get_boolean_values (JSON *self)
2050 PPCODE:
2051 if (self->v_false && self->v_true)
2052 {
2053 EXTEND (SP, 2);
2054 PUSHs (self->v_false);
2055 PUSHs (self->v_true);
2056 }
1683 2057
1684void ascii (JSON *self, int enable = 1) 2058void ascii (JSON *self, int enable = 1)
1685 ALIAS: 2059 ALIAS:
1686 ascii = F_ASCII 2060 ascii = F_ASCII
1687 latin1 = F_LATIN1 2061 latin1 = F_LATIN1
1695 shrink = F_SHRINK 2069 shrink = F_SHRINK
1696 allow_blessed = F_ALLOW_BLESSED 2070 allow_blessed = F_ALLOW_BLESSED
1697 convert_blessed = F_CONV_BLESSED 2071 convert_blessed = F_CONV_BLESSED
1698 relaxed = F_RELAXED 2072 relaxed = F_RELAXED
1699 allow_unknown = F_ALLOW_UNKNOWN 2073 allow_unknown = F_ALLOW_UNKNOWN
2074 allow_tags = F_ALLOW_TAGS
1700 PPCODE: 2075 PPCODE:
1701{ 2076{
1702 if (enable) 2077 if (enable)
1703 self->flags |= ix; 2078 self->flags |= ix;
1704 else 2079 else
1720 get_shrink = F_SHRINK 2095 get_shrink = F_SHRINK
1721 get_allow_blessed = F_ALLOW_BLESSED 2096 get_allow_blessed = F_ALLOW_BLESSED
1722 get_convert_blessed = F_CONV_BLESSED 2097 get_convert_blessed = F_CONV_BLESSED
1723 get_relaxed = F_RELAXED 2098 get_relaxed = F_RELAXED
1724 get_allow_unknown = F_ALLOW_UNKNOWN 2099 get_allow_unknown = F_ALLOW_UNKNOWN
2100 get_allow_tags = F_ALLOW_TAGS
1725 PPCODE: 2101 PPCODE:
1726 XPUSHs (boolSV (self->flags & ix)); 2102 XPUSHs (boolSV (self->flags & ix));
1727 2103
1728void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2104void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1729 PPCODE: 2105 PPCODE:
1757} 2133}
1758 2134
1759void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2135void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1760 PPCODE: 2136 PPCODE:
1761{ 2137{
1762 if (!self->cb_sk_object) 2138 if (!self->cb_sk_object)
1763 self->cb_sk_object = newHV (); 2139 self->cb_sk_object = newHV ();
1764 2140
1765 if (SvOK (cb)) 2141 if (SvOK (cb))
1766 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2142 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1767 else 2143 else
1778 XPUSHs (ST (0)); 2154 XPUSHs (ST (0));
1779} 2155}
1780 2156
1781void encode (JSON *self, SV *scalar) 2157void encode (JSON *self, SV *scalar)
1782 PPCODE: 2158 PPCODE:
1783 XPUSHs (encode_json (scalar, self)); 2159 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2160 XPUSHs (scalar);
1784 2161
1785void decode (JSON *self, SV *jsonstr) 2162void decode (JSON *self, SV *jsonstr)
1786 PPCODE: 2163 PPCODE:
1787 XPUSHs (decode_json (jsonstr, self, 0)); 2164 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2165 XPUSHs (jsonstr);
1788 2166
1789void decode_prefix (JSON *self, SV *jsonstr) 2167void decode_prefix (JSON *self, SV *jsonstr)
1790 PPCODE: 2168 PPCODE:
1791{ 2169{
1792 char *offset; 2170 SV *sv;
2171 STRLEN offset;
2172 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1793 EXTEND (SP, 2); 2173 EXTEND (SP, 2);
1794 PUSHs (decode_json (jsonstr, self, &offset)); 2174 PUSHs (sv);
1795 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2175 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1796} 2176}
1797 2177
1798void incr_parse (JSON *self, SV *jsonstr = 0) 2178void incr_parse (JSON *self, SV *jsonstr = 0)
1799 PPCODE: 2179 PPCODE:
1800{ 2180{
1801 if (!self->incr_text) 2181 if (!self->incr_text)
1802 self->incr_text = newSVpvn ("", 0); 2182 self->incr_text = newSVpvn ("", 0);
2183
2184 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2185 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2186 if (DECODE_WANTS_OCTETS (self))
2187 {
2188 if (self->incr_pos)
2189 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2190 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2191
2192 sv_utf8_downgrade (self->incr_text, 0);
2193 }
2194 else
2195 {
2196 sv_utf8_upgrade (self->incr_text);
2197
2198 if (self->incr_pos)
2199 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2200 - (U8 *)SvPVX (self->incr_text);
2201 }
1803 2202
1804 // append data, if any 2203 // append data, if any
1805 if (jsonstr) 2204 if (jsonstr)
1806 { 2205 {
2206 /* make sure both strings have same encoding */
2207 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1807 if (SvUTF8 (jsonstr)) 2208 if (SvUTF8 (jsonstr))
2209 sv_utf8_downgrade (jsonstr, 0);
1808 { 2210 else
1809 if (!SvUTF8 (self->incr_text))
1810 {
1811 /* utf-8-ness differs, need to upgrade */
1812 sv_utf8_upgrade (self->incr_text);
1813
1814 if (self->incr_pos)
1815 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1816 - (U8 *)SvPVX (self->incr_text);
1817 }
1818 }
1819 else if (SvUTF8 (self->incr_text))
1820 sv_utf8_upgrade (jsonstr); 2211 sv_utf8_upgrade (jsonstr);
1821 2212
2213 /* and then just blindly append */
1822 { 2214 {
1823 STRLEN len; 2215 STRLEN len;
1824 const char *str = SvPV (jsonstr, len); 2216 const char *str = SvPV (jsonstr, len);
1825 STRLEN cur = SvCUR (self->incr_text); 2217 STRLEN cur = SvCUR (self->incr_text);
1826 2218
1827 if (SvLEN (self->incr_text) <= cur + len) 2219 if (SvLEN (self->incr_text) - cur <= len)
1828 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2220 json_sv_grow (self->incr_text, cur, len);
1829 2221
1830 Move (str, SvEND (self->incr_text), len, char); 2222 Move (str, SvEND (self->incr_text), len, char);
1831 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2223 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1832 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2224 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1833 } 2225 }
1834 } 2226 }
1835 2227
1836 if (GIMME_V != G_VOID) 2228 if (GIMME_V != G_VOID)
1837 do 2229 do
1838 { 2230 {
2231 SV *sv;
1839 char *offset; 2232 STRLEN offset;
1840 2233
1841 if (!INCR_DONE (self)) 2234 if (!INCR_DONE (self))
1842 { 2235 {
1843 incr_parse (self); 2236 incr_parse (self);
1844 2237
1845 if (self->incr_pos > self->max_size && self->max_size) 2238 if (self->incr_pos > self->max_size && self->max_size)
1846 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2239 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1847 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2240 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1848 2241
1849 if (!INCR_DONE (self)) 2242 if (!INCR_DONE (self))
2243 {
2244 // as an optimisation, do not accumulate white space in the incr buffer
2245 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2246 {
2247 self->incr_pos = 0;
2248 SvCUR_set (self->incr_text, 0);
2249 }
2250
1850 break; 2251 break;
2252 }
1851 } 2253 }
1852 2254
1853 XPUSHs (decode_json (self->incr_text, self, &offset)); 2255 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2256 XPUSHs (sv);
1854 2257
1855 self->incr_pos -= offset - SvPVX (self->incr_text); 2258 self->incr_pos -= offset;
1856 self->incr_nest = 0; 2259 self->incr_nest = 0;
1857 self->incr_mode = 0; 2260 self->incr_mode = 0;
1858 2261
1859 sv_chop (self->incr_text, offset); 2262 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1860 } 2263 }
1861 while (GIMME_V == G_ARRAY); 2264 while (GIMME_V == G_ARRAY);
1862} 2265}
1863 2266
1864SV *incr_text (JSON *self) 2267SV *incr_text (JSON *self)
1895 self->incr_mode = 0; 2298 self->incr_mode = 0;
1896} 2299}
1897 2300
1898void DESTROY (JSON *self) 2301void DESTROY (JSON *self)
1899 CODE: 2302 CODE:
2303 SvREFCNT_dec (self->v_false);
2304 SvREFCNT_dec (self->v_true);
1900 SvREFCNT_dec (self->cb_sk_object); 2305 SvREFCNT_dec (self->cb_sk_object);
1901 SvREFCNT_dec (self->cb_object); 2306 SvREFCNT_dec (self->cb_object);
1902 SvREFCNT_dec (self->incr_text); 2307 SvREFCNT_dec (self->incr_text);
1903 2308
1904PROTOTYPES: ENABLE 2309PROTOTYPES: ENABLE
1905 2310
1906void encode_json (SV *scalar) 2311void encode_json (SV *scalar)
1907 ALIAS:
1908 to_json_ = 0
1909 encode_json = F_UTF8
1910 PPCODE: 2312 PPCODE:
1911{ 2313{
1912 JSON json; 2314 JSON json;
1913 json_init (&json); 2315 json_init (&json);
1914 json.flags |= ix; 2316 json.flags |= F_UTF8;
1915 XPUSHs (encode_json (scalar, &json)); 2317 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2318 XPUSHs (scalar);
1916} 2319}
1917 2320
1918void decode_json (SV *jsonstr) 2321void decode_json (SV *jsonstr)
1919 ALIAS:
1920 from_json_ = 0
1921 decode_json = F_UTF8
1922 PPCODE: 2322 PPCODE:
1923{ 2323{
1924 JSON json; 2324 JSON json;
1925 json_init (&json); 2325 json_init (&json);
1926 json.flags |= ix; 2326 json.flags |= F_UTF8;
1927 XPUSHs (decode_json (jsonstr, &json, 0)); 2327 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2328 XPUSHs (jsonstr);
1928} 2329}
1929 2330
1930

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines