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 1.98 by root, Fri Jul 17 14:33:45 2009 UTC vs.
Revision 1.133 by root, Tue Aug 28 16:16:17 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// 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
62 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
63 79
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
68# 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))
69#else 86#else
70# define JSON_SLOW 0
71# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
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_false, *bool_true;
97static SV *sv_json;
76 98
77enum { 99enum {
78 INCR_M_WS = 0, // initial whitespace skipping, must be 0 100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
79 INCR_M_STR, // inside string 101 INCR_M_STR, // inside string
80 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
81 INCR_M_JSON // outside anything, count nesting 105 INCR_M_JSON // outside anything, count nesting
82}; 106};
83 107
84#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)
85 109
94 // for the incremental parser 118 // for the incremental parser
95 SV *incr_text; // the source text so far 119 SV *incr_text; // the source text so far
96 STRLEN incr_pos; // the current offset into the text 120 STRLEN incr_pos; // the current offset into the text
97 int incr_nest; // {[]}-nesting level 121 int incr_nest; // {[]}-nesting level
98 unsigned char incr_mode; 122 unsigned char incr_mode;
123
124 SV *v_false, *v_true;
99} JSON; 125} JSON;
100 126
101INLINE void 127INLINE void
102json_init (JSON *json) 128json_init (JSON *json)
103{ 129{
132 SvPV_renew (sv, SvCUR (sv) + 1); 158 SvPV_renew (sv, SvCUR (sv) + 1);
133#endif 159#endif
134 } 160 }
135} 161}
136 162
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
137// 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
138// case of an error. 189// case of an error.
139// we special-case "safe" characters from U+80 .. U+7FF, 190// we special-case "safe" characters from U+80 .. U+7FF,
140// but use the very good perl function to parse anything else. 191// but use the very good perl function to parse anything else.
141// note that we never call this function for a ascii codepoints 192// note that we never call this function for a ascii codepoints
142INLINE UV 193INLINE UV
185 ? utf8_distance (offset, SvPVX (sv)) 236 ? utf8_distance (offset, SvPVX (sv))
186 : offset - SvPVX (sv); 237 : offset - SvPVX (sv);
187} 238}
188 239
189///////////////////////////////////////////////////////////////////////////// 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/////////////////////////////////////////////////////////////////////////////
190// encoder 379// encoder
191 380
192// structure used for encoding JSON 381// structure used for encoding JSON
193typedef struct 382typedef struct
194{ 383{
201} enc_t; 390} enc_t;
202 391
203INLINE void 392INLINE void
204need (enc_t *enc, STRLEN len) 393need (enc_t *enc, STRLEN len)
205{ 394{
206 if (expect_false (enc->cur + len >= enc->end)) 395 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
207 { 396 {
208 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 397 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
209 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 398 char *buf = json_sv_grow (enc->sv, cur, len);
210 enc->cur = SvPVX (enc->sv) + cur; 399 enc->cur = buf + cur;
211 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 400 enc->end = buf + SvLEN (enc->sv) - 1;
212 } 401 }
213} 402}
214 403
215INLINE void 404INLINE void
216encode_ch (enc_t *enc, char ch) 405encode_ch (enc_t *enc, char ch)
232 421
233 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 422 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
234 { 423 {
235 if (expect_false (ch == '"')) // but with slow exceptions 424 if (expect_false (ch == '"')) // but with slow exceptions
236 { 425 {
237 need (enc, len += 1); 426 need (enc, len + 1);
238 *enc->cur++ = '\\'; 427 *enc->cur++ = '\\';
239 *enc->cur++ = '"'; 428 *enc->cur++ = '"';
240 } 429 }
241 else if (expect_false (ch == '\\')) 430 else if (expect_false (ch == '\\'))
242 { 431 {
243 need (enc, len += 1); 432 need (enc, len + 1);
244 *enc->cur++ = '\\'; 433 *enc->cur++ = '\\';
245 *enc->cur++ = '\\'; 434 *enc->cur++ = '\\';
246 } 435 }
247 else 436 else
248 *enc->cur++ = ch; 437 *enc->cur++ = ch;
251 } 440 }
252 else 441 else
253 { 442 {
254 switch (ch) 443 switch (ch)
255 { 444 {
256 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;
257 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;
258 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;
259 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;
260 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;
261 450
262 default: 451 default:
263 { 452 {
264 STRLEN clen; 453 STRLEN clen;
265 UV uch; 454 UV uch;
281 if (uch >= 0x10000UL) 470 if (uch >= 0x10000UL)
282 { 471 {
283 if (uch >= 0x110000UL) 472 if (uch >= 0x110000UL)
284 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);
285 474
286 need (enc, len += 11); 475 need (enc, len + 11);
287 sprintf (enc->cur, "\\u%04x\\u%04x", 476 sprintf (enc->cur, "\\u%04x\\u%04x",
288 (int)((uch - 0x10000) / 0x400 + 0xD800), 477 (int)((uch - 0x10000) / 0x400 + 0xD800),
289 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 478 (int)((uch - 0x10000) % 0x400 + 0xDC00));
290 enc->cur += 12; 479 enc->cur += 12;
291 } 480 }
292 else 481 else
293 { 482 {
294 need (enc, len += 5); 483 need (enc, len + 5);
295 *enc->cur++ = '\\'; 484 *enc->cur++ = '\\';
296 *enc->cur++ = 'u'; 485 *enc->cur++ = 'u';
297 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 486 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
298 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 487 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
299 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 488 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
307 *enc->cur++ = uch; 496 *enc->cur++ = uch;
308 str += clen; 497 str += clen;
309 } 498 }
310 else if (is_utf8) 499 else if (is_utf8)
311 { 500 {
312 need (enc, len += clen); 501 need (enc, len + clen);
313 do 502 do
314 { 503 {
315 *enc->cur++ = *str++; 504 *enc->cur++ = *str++;
316 } 505 }
317 while (--clen); 506 while (--clen);
318 } 507 }
319 else 508 else
320 { 509 {
321 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
322 enc->cur = encode_utf8 (enc->cur, uch); 511 enc->cur = encode_utf8 (enc->cur, uch);
323 ++str; 512 ++str;
324 } 513 }
325 } 514 }
326 } 515 }
380 569
381 if (enc->indent >= enc->json.max_depth) 570 if (enc->indent >= enc->json.max_depth)
382 croak (ERR_NESTING_EXCEEDED); 571 croak (ERR_NESTING_EXCEEDED);
383 572
384 encode_ch (enc, '['); 573 encode_ch (enc, '[');
385 574
386 if (len >= 0) 575 if (len >= 0)
387 { 576 {
388 encode_nl (enc); ++enc->indent; 577 encode_nl (enc); ++enc->indent;
389 578
390 for (i = 0; i <= len; ++i) 579 for (i = 0; i <= len; ++i)
402 encode_comma (enc); 591 encode_comma (enc);
403 } 592 }
404 593
405 encode_nl (enc); --enc->indent; encode_indent (enc); 594 encode_nl (enc); --enc->indent; encode_indent (enc);
406 } 595 }
407 596
408 encode_ch (enc, ']'); 597 encode_ch (enc, ']');
409} 598}
410 599
411static void 600static void
412encode_hk (enc_t *enc, HE *he) 601encode_hk (enc_t *enc, HE *he)
416 if (HeKLEN (he) == HEf_SVKEY) 605 if (HeKLEN (he) == HEf_SVKEY)
417 { 606 {
418 SV *sv = HeSVKEY (he); 607 SV *sv = HeSVKEY (he);
419 STRLEN len; 608 STRLEN len;
420 char *str; 609 char *str;
421 610
422 SvGETMAGIC (sv); 611 SvGETMAGIC (sv);
423 str = SvPV (sv, len); 612 str = SvPV (sv, len);
424 613
425 encode_str (enc, str, len, SvUTF8 (sv)); 614 encode_str (enc, str, len, SvUTF8 (sv));
426 } 615 }
492 } 681 }
493 682
494 if (count) 683 if (count)
495 { 684 {
496 int i, fast = 1; 685 int i, fast = 1;
497#if defined(__BORLANDC__) || defined(_MSC_VER) 686 HE *hes_stack [STACK_HES];
498 HE **hes = _alloca (count * sizeof (HE)); 687 HE **hes = hes_stack;
499#else 688
500 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 689 // allocate larger arrays on the heap
501#endif 690 if (count > STACK_HES)
691 {
692 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
693 hes = (HE **)SvPVX (sv);
694 }
502 695
503 i = 0; 696 i = 0;
504 while ((he = hv_iternext (hv))) 697 while ((he = hv_iternext (hv)))
505 { 698 {
506 hes [i++] = he; 699 hes [i++] = he;
575// encode objects, arrays and special \0=false and \1=true values. 768// encode objects, arrays and special \0=false and \1=true values.
576static void 769static void
577encode_rv (enc_t *enc, SV *sv) 770encode_rv (enc_t *enc, SV *sv)
578{ 771{
579 svtype svt; 772 svtype svt;
773 GV *method;
580 774
581 SvGETMAGIC (sv); 775 SvGETMAGIC (sv);
582 svt = SvTYPE (sv); 776 svt = SvTYPE (sv);
583 777
584 if (expect_false (SvOBJECT (sv))) 778 if (expect_false (SvOBJECT (sv)))
585 { 779 {
586 HV *stash = !JSON_SLOW || json_boolean_stash 780 HV *stash = SvSTASH (sv);
587 ? json_boolean_stash
588 : gv_stashpv ("JSON::XS::Boolean", 1);
589 781
590 if (SvSTASH (sv) == stash) 782 if (stash == bool_stash)
591 { 783 {
592 if (SvIV (sv))
593 encode_str (enc, "true", 4, 0); 784 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
594 else
595 encode_str (enc, "false", 5, 0); 785 else encode_str (enc, "false", 5, 0);
596 } 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 SAVESTACK_POS ();
794 PUSHMARK (SP);
795 EXTEND (SP, 2);
796 // we re-bless the reference to get overload and other niceties right
797 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
798 PUSHs (sv_json);
799
800 PUTBACK;
801 count = call_sv ((SV *)GvCV (method), G_ARRAY);
802 SPAGAIN;
803
804 // catch this surprisingly common error
805 if (SvROK (TOPs) && SvRV (TOPs) == sv)
806 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
807
808 encode_ch (enc, '(');
809 encode_ch (enc, '"');
810 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
811 encode_ch (enc, '"');
812 encode_ch (enc, ')');
813 encode_ch (enc, '[');
814
815 while (count)
816 {
817 encode_sv (enc, SP[1 - count--]);
818
819 if (count)
820 encode_ch (enc, ',');
821 }
822
823 encode_ch (enc, ']');
824
825 FREETMPS; LEAVE;
826 }
827 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
828 {
829 dSP;
830
831 ENTER; SAVETMPS;
832 PUSHMARK (SP);
833 // we re-bless the reference to get overload and other niceties right
834 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
835
836 // calling with G_SCALAR ensures that we always get a 1 return value
837 PUTBACK;
838 call_sv ((SV *)GvCV (method), G_SCALAR);
839 SPAGAIN;
840
841 // catch this surprisingly common error
842 if (SvROK (TOPs) && SvRV (TOPs) == sv)
843 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
844
845 sv = POPs;
846 PUTBACK;
847
848 encode_sv (enc, sv);
849
850 FREETMPS; LEAVE;
851 }
852 else if (enc->json.flags & F_ALLOW_BLESSED)
853 encode_str (enc, "null", 4, 0);
597 else 854 else
598 {
599#if 0
600 if (0 && sv_derived_from (rv, "JSON::Literal"))
601 {
602 // not yet
603 }
604#endif
605 if (enc->json.flags & F_CONV_BLESSED)
606 {
607 // we re-bless the reference to get overload and other niceties right
608 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
609
610 if (to_json)
611 {
612 dSP;
613
614 ENTER; SAVETMPS; PUSHMARK (SP);
615 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
616
617 // calling with G_SCALAR ensures that we always get a 1 return value
618 PUTBACK;
619 call_sv ((SV *)GvCV (to_json), G_SCALAR);
620 SPAGAIN;
621
622 // catch this surprisingly common error
623 if (SvROK (TOPs) && SvRV (TOPs) == sv)
624 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
625
626 sv = POPs;
627 PUTBACK;
628
629 encode_sv (enc, sv);
630
631 FREETMPS; LEAVE;
632 }
633 else if (enc->json.flags & F_ALLOW_BLESSED)
634 encode_str (enc, "null", 4, 0);
635 else
636 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
637 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
638 }
639 else if (enc->json.flags & F_ALLOW_BLESSED)
640 encode_str (enc, "null", 4, 0);
641 else
642 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 855 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
643 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 856 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
644 }
645 } 857 }
646 else if (svt == SVt_PVHV) 858 else if (svt == SVt_PVHV)
647 encode_hv (enc, (HV *)sv); 859 encode_hv (enc, (HV *)sv);
648 else if (svt == SVt_PVAV) 860 else if (svt == SVt_PVAV)
649 encode_av (enc, (AV *)sv); 861 encode_av (enc, (AV *)sv);
650 else if (svt < SVt_PVAV) 862 else if (svt < SVt_PVAV)
651 { 863 {
652 STRLEN len = 0; 864 int bool_type = ref_bool_type (sv);
653 char *pv = svt ? SvPV (sv, len) : 0;
654 865
655 if (len == 1 && *pv == '1') 866 if (bool_type == 1)
656 encode_str (enc, "true", 4, 0); 867 encode_str (enc, "true", 4, 0);
657 else if (len == 1 && *pv == '0') 868 else if (bool_type == 0)
658 encode_str (enc, "false", 5, 0); 869 encode_str (enc, "false", 5, 0);
659 else if (enc->json.flags & F_ALLOW_UNKNOWN) 870 else if (enc->json.flags & F_ALLOW_UNKNOWN)
660 encode_str (enc, "null", 4, 0); 871 encode_str (enc, "null", 4, 0);
661 else 872 else
662 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 873 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
725 } 936 }
726 else 937 else
727 { 938 {
728 // large integer, use the (rather slow) snprintf way. 939 // large integer, use the (rather slow) snprintf way.
729 need (enc, IVUV_MAXCHARS); 940 need (enc, IVUV_MAXCHARS);
730 enc->cur += 941 enc->cur +=
731 SvIsUV(sv) 942 SvIsUV(sv)
732 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 943 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
733 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 944 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
734 } 945 }
735 } 946 }
736 else if (SvROK (sv)) 947 else if (SvROK (sv))
737 encode_rv (enc, SvRV (sv)); 948 encode_rv (enc, SvRV (sv));
738 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 949 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
739 encode_str (enc, "null", 4, 0); 950 encode_str (enc, "null", 4, 0);
740 else 951 else
741 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 952 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
742 SvPV_nolen (sv), SvFLAGS (sv)); 953 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
743} 954}
744 955
745static SV * 956static SV *
746encode_json (SV *scalar, JSON *json) 957encode_json (SV *scalar, JSON *json)
747{ 958{
748 enc_t enc; 959 enc_t enc;
749 960
750 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 961 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
751 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 962 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
752 963
753 enc.json = *json; 964 enc.json = *json;
754 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 965 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
755 enc.cur = SvPVX (enc.sv); 966 enc.cur = SvPVX (enc.sv);
759 : enc.json.flags & F_LATIN1 ? 0x000100UL 970 : enc.json.flags & F_LATIN1 ? 0x000100UL
760 : 0x110000UL; 971 : 0x110000UL;
761 972
762 SvPOK_only (enc.sv); 973 SvPOK_only (enc.sv);
763 encode_sv (&enc, scalar); 974 encode_sv (&enc, scalar);
975 encode_nl (&enc);
764 976
765 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 977 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
766 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 978 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
767 979
768 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 980 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
948 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1160 else if (expect_true (ch >= 0x20 && ch < 0x80))
949 *cur++ = ch; 1161 *cur++ = ch;
950 else if (ch >= 0x80) 1162 else if (ch >= 0x80)
951 { 1163 {
952 STRLEN clen; 1164 STRLEN clen;
953 UV uch;
954 1165
955 --dec_cur; 1166 --dec_cur;
956 1167
957 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1168 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
958 if (clen == (STRLEN)-1) 1169 if (clen == (STRLEN)-1)
959 ERR ("malformed UTF-8 character in JSON string"); 1170 ERR ("malformed UTF-8 character in JSON string");
960 1171
961 do 1172 do
962 *cur++ = *dec_cur++; 1173 *cur++ = *dec_cur++;
963 while (--clen); 1174 while (--clen);
964 1175
965 utf8 = 1; 1176 utf8 = 1;
966 } 1177 }
1178 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1179 *cur++ = ch;
967 else 1180 else
968 { 1181 {
969 --dec_cur; 1182 --dec_cur;
970 1183
971 if (!ch) 1184 if (!ch)
981 1194
982 if (sv) 1195 if (sv)
983 { 1196 {
984 STRLEN cur = SvCUR (sv); 1197 STRLEN cur = SvCUR (sv);
985 1198
986 if (SvLEN (sv) <= cur + len) 1199 if (SvLEN (sv) - cur <= len)
987 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1200 json_sv_grow (sv, cur, len);
988 1201
989 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1202 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
990 SvCUR_set (sv, SvCUR (sv) + len); 1203 SvCUR_set (sv, SvCUR (sv) + len);
991 } 1204 }
992 else 1205 else
1116 } 1329 }
1117 1330
1118 len -= *start == '-' ? 1 : 0; 1331 len -= *start == '-' ? 1 : 0;
1119 1332
1120 // does not fit into IV or UV, try NV 1333 // does not fit into IV or UV, try NV
1121 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1334 if (len <= NV_DIG)
1122 #if defined (LDBL_DIG)
1123 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1124 #endif
1125 )
1126 // fits into NV without loss of precision 1335 // fits into NV without loss of precision
1127 return newSVnv (Atof (start)); 1336 return newSVnv (json_atof (start));
1128 1337
1129 // everything else fails, convert it to a string 1338 // everything else fails, convert it to a string
1130 return newSVpvn (start, dec->cur - start); 1339 return newSVpvn (start, dec->cur - start);
1131 } 1340 }
1132 1341
1133 // loss of precision here 1342 // loss of precision here
1134 return newSVnv (Atof (start)); 1343 return newSVnv (json_atof (start));
1135 1344
1136fail: 1345fail:
1137 return 0; 1346 return 0;
1138} 1347}
1139 1348
1163 if (*dec->cur == ']') 1372 if (*dec->cur == ']')
1164 { 1373 {
1165 ++dec->cur; 1374 ++dec->cur;
1166 break; 1375 break;
1167 } 1376 }
1168 1377
1169 if (*dec->cur != ',') 1378 if (*dec->cur != ',')
1170 ERR (", or ] expected while parsing array"); 1379 ERR (", or ] expected while parsing array");
1171 1380
1172 ++dec->cur; 1381 ++dec->cur;
1173 1382
1297 1506
1298 hv_iterinit (hv); 1507 hv_iterinit (hv);
1299 he = hv_iternext (hv); 1508 he = hv_iternext (hv);
1300 hv_iterinit (hv); 1509 hv_iterinit (hv);
1301 1510
1302 // the next line creates a mortal sv each time its called. 1511 // the next line creates a mortal sv each time it's called.
1303 // might want to optimise this for common cases. 1512 // might want to optimise this for common cases.
1304 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1513 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1305 1514
1306 if (cb) 1515 if (cb)
1307 { 1516 {
1308 dSP; 1517 dSP;
1309 int count; 1518 int count;
1310 1519
1311 ENTER; SAVETMPS; PUSHMARK (SP); 1520 ENTER; SAVETMPS;
1521 SAVESTACK_POS ();
1522 PUSHMARK (SP);
1312 XPUSHs (HeVAL (he)); 1523 XPUSHs (HeVAL (he));
1524 sv_2mortal (sv);
1313 1525
1314 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1526 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1315 1527
1316 if (count == 1) 1528 if (count == 1)
1317 { 1529 {
1318 sv = newSVsv (POPs); 1530 sv = newSVsv (POPs);
1319 FREETMPS; LEAVE; 1531 FREETMPS; LEAVE;
1320 return sv; 1532 return sv;
1321 } 1533 }
1322 1534
1535 SvREFCNT_inc (sv);
1323 FREETMPS; LEAVE; 1536 FREETMPS; LEAVE;
1324 } 1537 }
1325 } 1538 }
1326 1539
1327 if (dec->json.cb_object) 1540 if (dec->json.cb_object)
1328 { 1541 {
1329 dSP; 1542 dSP;
1330 int count; 1543 int count;
1331 1544
1332 ENTER; SAVETMPS; PUSHMARK (SP); 1545 ENTER; SAVETMPS;
1546 SAVESTACK_POS ();
1547 PUSHMARK (SP);
1333 XPUSHs (sv_2mortal (sv)); 1548 XPUSHs (sv_2mortal (sv));
1334 1549
1335 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1550 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1336 1551
1337 if (count == 1) 1552 if (count == 1)
1353 DEC_DEC_DEPTH; 1568 DEC_DEC_DEPTH;
1354 return 0; 1569 return 0;
1355} 1570}
1356 1571
1357static SV * 1572static SV *
1573decode_tag (dec_t *dec)
1574{
1575 SV *tag = 0;
1576 SV *val = 0;
1577
1578 if (!(dec->json.flags & F_ALLOW_TAGS))
1579 ERR ("malformed JSON string, neither array, object, number, string or atom");
1580
1581 ++dec->cur;
1582
1583 decode_ws (dec);
1584
1585 tag = decode_sv (dec);
1586 if (!tag)
1587 goto fail;
1588
1589 if (!SvPOK (tag))
1590 ERR ("malformed JSON string, (tag) must be a string");
1591
1592 decode_ws (dec);
1593
1594 if (*dec->cur != ')')
1595 ERR (") expected after tag");
1596
1597 ++dec->cur;
1598
1599 decode_ws (dec);
1600
1601 val = decode_sv (dec);
1602 if (!val)
1603 goto fail;
1604
1605 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1606 ERR ("malformed JSON string, tag value must be an array");
1607
1608 {
1609 AV *av = (AV *)SvRV (val);
1610 int i, len = av_len (av) + 1;
1611 HV *stash = gv_stashsv (tag, 0);
1612 SV *sv;
1613
1614 if (!stash)
1615 ERR ("cannot decode perl-object (package does not exist)");
1616
1617 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1618
1619 if (!method)
1620 ERR ("cannot decode perl-object (package does not have a THAW method)");
1621
1622 dSP;
1623
1624 ENTER; SAVETMPS;
1625 PUSHMARK (SP);
1626 EXTEND (SP, len + 2);
1627 // we re-bless the reference to get overload and other niceties right
1628 PUSHs (tag);
1629 PUSHs (sv_json);
1630
1631 for (i = 0; i < len; ++i)
1632 PUSHs (*av_fetch (av, i, 1));
1633
1634 PUTBACK;
1635 call_sv ((SV *)GvCV (method), G_SCALAR);
1636 SPAGAIN;
1637
1638 SvREFCNT_dec (tag);
1639 SvREFCNT_dec (val);
1640 sv = SvREFCNT_inc (POPs);
1641
1642 PUTBACK;
1643
1644 FREETMPS; LEAVE;
1645
1646 return sv;
1647 }
1648
1649fail:
1650 SvREFCNT_dec (tag);
1651 SvREFCNT_dec (val);
1652 return 0;
1653}
1654
1655static SV *
1358decode_sv (dec_t *dec) 1656decode_sv (dec_t *dec)
1359{ 1657{
1360 // the beauty of JSON: you need exactly one character lookahead 1658 // the beauty of JSON: you need exactly one character lookahead
1361 // to parse everything. 1659 // to parse everything.
1362 switch (*dec->cur) 1660 switch (*dec->cur)
1363 { 1661 {
1364 case '"': ++dec->cur; return decode_str (dec); 1662 case '"': ++dec->cur; return decode_str (dec);
1365 case '[': ++dec->cur; return decode_av (dec); 1663 case '[': ++dec->cur; return decode_av (dec);
1366 case '{': ++dec->cur; return decode_hv (dec); 1664 case '{': ++dec->cur; return decode_hv (dec);
1665 case '(': return decode_tag (dec);
1367 1666
1368 case '-': 1667 case '-':
1369 case '0': case '1': case '2': case '3': case '4': 1668 case '0': case '1': case '2': case '3': case '4':
1370 case '5': case '6': case '7': case '8': case '9': 1669 case '5': case '6': case '7': case '8': case '9':
1371 return decode_num (dec); 1670 return decode_num (dec);
1372 1671
1672 case 'f':
1673 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1674 {
1675 dec->cur += 5;
1676
1677 if (expect_false (!dec->json.v_false))
1678 dec->json.v_false = GET_BOOL (false);
1679
1680 return newSVsv (dec->json.v_false);
1681 }
1682 else
1683 ERR ("'false' expected");
1684
1685 break;
1686
1373 case 't': 1687 case 't':
1374 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1688 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1375 { 1689 {
1376 dec->cur += 4; 1690 dec->cur += 4;
1377#if JSON_SLOW 1691
1378 json_true = get_bool ("JSON::XS::true"); 1692 if (expect_false (!dec->json.v_true))
1379#endif 1693 dec->json.v_true = GET_BOOL (true);
1694
1380 return newSVsv (json_true); 1695 return newSVsv (dec->json.v_true);
1381 } 1696 }
1382 else 1697 else
1383 ERR ("'true' expected"); 1698 ERR ("'true' expected");
1384
1385 break;
1386
1387 case 'f':
1388 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1389 {
1390 dec->cur += 5;
1391#if JSON_SLOW
1392 json_false = get_bool ("JSON::XS::false");
1393#endif
1394 return newSVsv (json_false);
1395 }
1396 else
1397 ERR ("'false' expected");
1398 1699
1399 break; 1700 break;
1400 1701
1401 case 'n': 1702 case 'n':
1402 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1703 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1408 ERR ("'null' expected"); 1709 ERR ("'null' expected");
1409 1710
1410 break; 1711 break;
1411 1712
1412 default: 1713 default:
1413 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1714 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1414 break; 1715 break;
1415 } 1716 }
1416 1717
1417fail: 1718fail:
1418 return 0; 1719 return 0;
1419} 1720}
1420 1721
1421static SV * 1722static SV *
1422decode_json (SV *string, JSON *json, char **offset_return) 1723decode_json (SV *string, JSON *json, STRLEN *offset_return)
1423{ 1724{
1424 dec_t dec; 1725 dec_t dec;
1425 SV *sv; 1726 SV *sv;
1426 1727
1427 /* work around bugs in 5.10 where manipulating magic values 1728 /* work around bugs in 5.10 where manipulating magic values
1428 * will perl ignore the magic in subsequent accesses 1729 * makes perl ignore the magic in subsequent accesses.
1730 * also make a copy of non-PV values, to get them into a clean
1731 * state (SvPV should do that, but it's buggy, see below).
1732 *
1733 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1734 * as reported by Reini Urban.
1429 */ 1735 */
1430 /*SvGETMAGIC (string);*/ 1736 /*SvGETMAGIC (string);*/
1431 if (SvMAGICAL (string)) 1737 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1432 string = sv_2mortal (newSVsv (string)); 1738 string = sv_2mortal (newSVsv (string));
1433 1739
1434 SvUPGRADE (string, SVt_PV); 1740 SvUPGRADE (string, SVt_PV);
1435 1741
1436 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1742 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1453 if (offset > json->max_size && json->max_size) 1759 if (offset > json->max_size && json->max_size)
1454 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1760 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1455 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1761 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1456 } 1762 }
1457 1763
1458 if (json->flags & F_UTF8) 1764 if (DECODE_WANTS_OCTETS (json))
1459 sv_utf8_downgrade (string, 0); 1765 sv_utf8_downgrade (string, 0);
1460 else 1766 else
1461 sv_utf8_upgrade (string); 1767 sv_utf8_upgrade (string);
1462 1768
1463 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1769 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1475 1781
1476 decode_ws (&dec); 1782 decode_ws (&dec);
1477 sv = decode_sv (&dec); 1783 sv = decode_sv (&dec);
1478 1784
1479 if (offset_return) 1785 if (offset_return)
1480 *offset_return = dec.cur; 1786 *offset_return = dec.cur - SvPVX (string);
1481 1787 else if (sv)
1482 if (!(offset_return || !sv))
1483 { 1788 {
1484 // check for trailing garbage 1789 // check for trailing garbage
1485 decode_ws (&dec); 1790 decode_ws (&dec);
1486 1791
1487 if (*dec.cur) 1792 if (*dec.cur)
1505 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1810 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1506 LEAVE; 1811 LEAVE;
1507 1812
1508 croak ("%s, at character offset %d (before \"%s\")", 1813 croak ("%s, at character offset %d (before \"%s\")",
1509 dec.err, 1814 dec.err,
1510 ptr_to_index (string, dec.cur), 1815 (int)ptr_to_index (string, dec.cur),
1511 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1816 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1512 } 1817 }
1513 1818
1514 sv = sv_2mortal (sv); 1819 sv = sv_2mortal (sv);
1515 1820
1516 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1821 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1517 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1822 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1518 1823
1519 return sv; 1824 return sv;
1520} 1825}
1521 1826
1524 1829
1525static void 1830static void
1526incr_parse (JSON *self) 1831incr_parse (JSON *self)
1527{ 1832{
1528 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1833 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1834
1835 // the state machine here is a bit convoluted and could be simplified a lot
1836 // but this would make it slower, so...
1529 1837
1530 for (;;) 1838 for (;;)
1531 { 1839 {
1532 //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 1840 //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
1533 switch (self->incr_mode) 1841 switch (self->incr_mode)
1534 { 1842 {
1535 // only used for intiial whitespace skipping 1843 // only used for initial whitespace skipping
1536 case INCR_M_WS: 1844 case INCR_M_WS:
1537 for (;;) 1845 for (;;)
1538 { 1846 {
1539 if (*p > 0x20) 1847 if (*p > 0x20)
1540 { 1848 {
1849 if (*p == '#')
1850 {
1851 self->incr_mode = INCR_M_C0;
1852 goto incr_m_c;
1853 }
1854 else
1855 {
1541 self->incr_mode = INCR_M_JSON; 1856 self->incr_mode = INCR_M_JSON;
1542 goto incr_m_json; 1857 goto incr_m_json;
1858 }
1543 } 1859 }
1544 else if (!*p) 1860 else if (!*p)
1545 goto interrupt; 1861 goto interrupt;
1546 1862
1547 ++p; 1863 ++p;
1553 goto interrupt; 1869 goto interrupt;
1554 1870
1555 ++p; 1871 ++p;
1556 self->incr_mode = INCR_M_STR; 1872 self->incr_mode = INCR_M_STR;
1557 goto incr_m_str; 1873 goto incr_m_str;
1874
1875 // inside #-style comments
1876 case INCR_M_C0:
1877 case INCR_M_C1:
1878 incr_m_c:
1879 for (;;)
1880 {
1881 if (*p == '\n')
1882 {
1883 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1884 break;
1885 }
1886 else if (!*p)
1887 goto interrupt;
1888
1889 ++p;
1890 }
1891
1892 break;
1558 1893
1559 // inside a string 1894 // inside a string
1560 case INCR_M_STR: 1895 case INCR_M_STR:
1561 incr_m_str: 1896 incr_m_str:
1562 for (;;) 1897 for (;;)
1613 self->incr_mode = INCR_M_STR; 1948 self->incr_mode = INCR_M_STR;
1614 goto incr_m_str; 1949 goto incr_m_str;
1615 1950
1616 case '[': 1951 case '[':
1617 case '{': 1952 case '{':
1953 case '(':
1618 if (++self->incr_nest > self->max_depth) 1954 if (++self->incr_nest > self->max_depth)
1619 croak (ERR_NESTING_EXCEEDED); 1955 croak (ERR_NESTING_EXCEEDED);
1620 break; 1956 break;
1621 1957
1622 case ']': 1958 case ']':
1623 case '}': 1959 case '}':
1624 if (--self->incr_nest <= 0) 1960 if (--self->incr_nest <= 0)
1625 goto interrupt; 1961 goto interrupt;
1962 break;
1963
1964 case ')':
1965 --self->incr_nest;
1966 break;
1967
1968 case '#':
1969 self->incr_mode = INCR_M_C1;
1970 goto incr_m_c;
1626 } 1971 }
1627 } 1972 }
1628 } 1973 }
1629 1974
1630 modechange: 1975 modechange:
1631 ; 1976 ;
1632 } 1977 }
1633 1978
1634interrupt: 1979interrupt:
1635 self->incr_pos = p - SvPVX (self->incr_text); 1980 self->incr_pos = p - SvPVX (self->incr_text);
1981 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1636 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1982 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1637} 1983}
1638 1984
1639///////////////////////////////////////////////////////////////////////////// 1985/////////////////////////////////////////////////////////////////////////////
1640// XS interface functions 1986// XS interface functions
1650 i >= '0' && i <= '9' ? i - '0' 1996 i >= '0' && i <= '9' ? i - '0'
1651 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1997 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1652 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1998 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1653 : -1; 1999 : -1;
1654 2000
1655 json_stash = gv_stashpv ("JSON::XS" , 1); 2001 json_stash = gv_stashpv ("JSON::XS" , 1);
1656 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2002 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1657
1658 json_true = get_bool ("JSON::XS::true");
1659 json_false = get_bool ("JSON::XS::false"); 2003 bool_false = get_bool ("Types::Serialiser::false");
2004 bool_true = get_bool ("Types::Serialiser::true");
2005
2006 sv_json = newSVpv ("JSON", 0);
2007 SvREADONLY_on (sv_json);
2008
2009 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1660} 2010}
1661 2011
1662PROTOTYPES: DISABLE 2012PROTOTYPES: DISABLE
1663 2013
1664void CLONE (...) 2014void CLONE (...)
1665 CODE: 2015 CODE:
2016 // as long as these writes are atomic, the race should not matter
2017 // as existing threads either already use 0, or use the old value,
2018 // which is sitll correct for the initial thread.
1666 json_stash = 0; 2019 json_stash = 0;
1667 json_boolean_stash = 0; 2020 bool_stash = 0;
2021 bool_false = 0;
2022 bool_true = 0;
1668 2023
1669void new (char *klass) 2024void new (char *klass)
1670 PPCODE: 2025 PPCODE:
1671{ 2026{
1672 SV *pv = NEWSV (0, sizeof (JSON)); 2027 SV *pv = NEWSV (0, sizeof (JSON));
1673 SvPOK_only (pv); 2028 SvPOK_only (pv);
1674 json_init ((JSON *)SvPVX (pv)); 2029 json_init ((JSON *)SvPVX (pv));
1675 XPUSHs (sv_2mortal (sv_bless ( 2030 XPUSHs (sv_2mortal (sv_bless (
1676 newRV_noinc (pv), 2031 newRV_noinc (pv),
1677 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2032 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1678 ))); 2033 )));
1679} 2034}
2035
2036void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2037 PPCODE:
2038 self->v_false = newSVsv (v_false);
2039 self->v_true = newSVsv (v_true);
2040 XPUSHs (ST (0));
2041
2042void get_boolean_values (JSON *self)
2043 PPCODE:
2044 if (self->v_false && self->v_true)
2045 {
2046 EXTEND (SP, 2);
2047 PUSHs (self->v_false);
2048 PUSHs (self->v_true);
2049 }
1680 2050
1681void ascii (JSON *self, int enable = 1) 2051void ascii (JSON *self, int enable = 1)
1682 ALIAS: 2052 ALIAS:
1683 ascii = F_ASCII 2053 ascii = F_ASCII
1684 latin1 = F_LATIN1 2054 latin1 = F_LATIN1
1692 shrink = F_SHRINK 2062 shrink = F_SHRINK
1693 allow_blessed = F_ALLOW_BLESSED 2063 allow_blessed = F_ALLOW_BLESSED
1694 convert_blessed = F_CONV_BLESSED 2064 convert_blessed = F_CONV_BLESSED
1695 relaxed = F_RELAXED 2065 relaxed = F_RELAXED
1696 allow_unknown = F_ALLOW_UNKNOWN 2066 allow_unknown = F_ALLOW_UNKNOWN
2067 allow_tags = F_ALLOW_TAGS
1697 PPCODE: 2068 PPCODE:
1698{ 2069{
1699 if (enable) 2070 if (enable)
1700 self->flags |= ix; 2071 self->flags |= ix;
1701 else 2072 else
1717 get_shrink = F_SHRINK 2088 get_shrink = F_SHRINK
1718 get_allow_blessed = F_ALLOW_BLESSED 2089 get_allow_blessed = F_ALLOW_BLESSED
1719 get_convert_blessed = F_CONV_BLESSED 2090 get_convert_blessed = F_CONV_BLESSED
1720 get_relaxed = F_RELAXED 2091 get_relaxed = F_RELAXED
1721 get_allow_unknown = F_ALLOW_UNKNOWN 2092 get_allow_unknown = F_ALLOW_UNKNOWN
2093 get_allow_tags = F_ALLOW_TAGS
1722 PPCODE: 2094 PPCODE:
1723 XPUSHs (boolSV (self->flags & ix)); 2095 XPUSHs (boolSV (self->flags & ix));
1724 2096
1725void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2097void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1726 PPCODE: 2098 PPCODE:
1754} 2126}
1755 2127
1756void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2128void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1757 PPCODE: 2129 PPCODE:
1758{ 2130{
1759 if (!self->cb_sk_object) 2131 if (!self->cb_sk_object)
1760 self->cb_sk_object = newHV (); 2132 self->cb_sk_object = newHV ();
1761 2133
1762 if (SvOK (cb)) 2134 if (SvOK (cb))
1763 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2135 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1764 else 2136 else
1775 XPUSHs (ST (0)); 2147 XPUSHs (ST (0));
1776} 2148}
1777 2149
1778void encode (JSON *self, SV *scalar) 2150void encode (JSON *self, SV *scalar)
1779 PPCODE: 2151 PPCODE:
1780 XPUSHs (encode_json (scalar, self)); 2152 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2153 XPUSHs (scalar);
1781 2154
1782void decode (JSON *self, SV *jsonstr) 2155void decode (JSON *self, SV *jsonstr)
1783 PPCODE: 2156 PPCODE:
1784 XPUSHs (decode_json (jsonstr, self, 0)); 2157 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2158 XPUSHs (jsonstr);
1785 2159
1786void decode_prefix (JSON *self, SV *jsonstr) 2160void decode_prefix (JSON *self, SV *jsonstr)
1787 PPCODE: 2161 PPCODE:
1788{ 2162{
1789 char *offset; 2163 SV *sv;
2164 STRLEN offset;
2165 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1790 EXTEND (SP, 2); 2166 EXTEND (SP, 2);
1791 PUSHs (decode_json (jsonstr, self, &offset)); 2167 PUSHs (sv);
1792 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2168 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1793} 2169}
1794 2170
1795void incr_parse (JSON *self, SV *jsonstr = 0) 2171void incr_parse (JSON *self, SV *jsonstr = 0)
1796 PPCODE: 2172 PPCODE:
1797{ 2173{
1798 if (!self->incr_text) 2174 if (!self->incr_text)
1799 self->incr_text = newSVpvn ("", 0); 2175 self->incr_text = newSVpvn ("", 0);
2176
2177 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2178 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2179 if (DECODE_WANTS_OCTETS (self))
2180 {
2181 if (self->incr_pos)
2182 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2183 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2184
2185 sv_utf8_downgrade (self->incr_text, 0);
2186 }
2187 else
2188 {
2189 sv_utf8_upgrade (self->incr_text);
2190
2191 if (self->incr_pos)
2192 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2193 - (U8 *)SvPVX (self->incr_text);
2194 }
1800 2195
1801 // append data, if any 2196 // append data, if any
1802 if (jsonstr) 2197 if (jsonstr)
1803 { 2198 {
2199 /* make sure both strings have same encoding */
2200 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1804 if (SvUTF8 (jsonstr)) 2201 if (SvUTF8 (jsonstr))
2202 sv_utf8_downgrade (jsonstr, 0);
1805 { 2203 else
1806 if (!SvUTF8 (self->incr_text))
1807 {
1808 /* utf-8-ness differs, need to upgrade */
1809 sv_utf8_upgrade (self->incr_text);
1810
1811 if (self->incr_pos)
1812 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1813 - (U8 *)SvPVX (self->incr_text);
1814 }
1815 }
1816 else if (SvUTF8 (self->incr_text))
1817 sv_utf8_upgrade (jsonstr); 2204 sv_utf8_upgrade (jsonstr);
1818 2205
2206 /* and then just blindly append */
1819 { 2207 {
1820 STRLEN len; 2208 STRLEN len;
1821 const char *str = SvPV (jsonstr, len); 2209 const char *str = SvPV (jsonstr, len);
1822 STRLEN cur = SvCUR (self->incr_text); 2210 STRLEN cur = SvCUR (self->incr_text);
1823 2211
1824 if (SvLEN (self->incr_text) <= cur + len) 2212 if (SvLEN (self->incr_text) - cur <= len)
1825 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2213 json_sv_grow (self->incr_text, cur, len);
1826 2214
1827 Move (str, SvEND (self->incr_text), len, char); 2215 Move (str, SvEND (self->incr_text), len, char);
1828 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2216 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1829 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2217 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1830 } 2218 }
1831 } 2219 }
1832 2220
1833 if (GIMME_V != G_VOID) 2221 if (GIMME_V != G_VOID)
1834 do 2222 do
1835 { 2223 {
2224 SV *sv;
1836 char *offset; 2225 STRLEN offset;
1837 2226
1838 if (!INCR_DONE (self)) 2227 if (!INCR_DONE (self))
1839 { 2228 {
1840 incr_parse (self); 2229 incr_parse (self);
1841 2230
1842 if (self->incr_pos > self->max_size && self->max_size) 2231 if (self->incr_pos > self->max_size && self->max_size)
1843 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2232 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1844 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2233 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1845 2234
1846 if (!INCR_DONE (self)) 2235 if (!INCR_DONE (self))
2236 {
2237 // as an optimisation, do not accumulate white space in the incr buffer
2238 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2239 {
2240 self->incr_pos = 0;
2241 SvCUR_set (self->incr_text, 0);
2242 }
2243
1847 break; 2244 break;
2245 }
1848 } 2246 }
1849 2247
1850 XPUSHs (decode_json (self->incr_text, self, &offset)); 2248 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2249 XPUSHs (sv);
1851 2250
1852 self->incr_pos -= offset - SvPVX (self->incr_text); 2251 self->incr_pos -= offset;
1853 self->incr_nest = 0; 2252 self->incr_nest = 0;
1854 self->incr_mode = 0; 2253 self->incr_mode = 0;
1855 2254
1856 sv_chop (self->incr_text, offset); 2255 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1857 } 2256 }
1858 while (GIMME_V == G_ARRAY); 2257 while (GIMME_V == G_ARRAY);
1859} 2258}
1860 2259
1861SV *incr_text (JSON *self) 2260SV *incr_text (JSON *self)
1892 self->incr_mode = 0; 2291 self->incr_mode = 0;
1893} 2292}
1894 2293
1895void DESTROY (JSON *self) 2294void DESTROY (JSON *self)
1896 CODE: 2295 CODE:
2296 SvREFCNT_dec (self->v_false);
2297 SvREFCNT_dec (self->v_true);
1897 SvREFCNT_dec (self->cb_sk_object); 2298 SvREFCNT_dec (self->cb_sk_object);
1898 SvREFCNT_dec (self->cb_object); 2299 SvREFCNT_dec (self->cb_object);
1899 SvREFCNT_dec (self->incr_text); 2300 SvREFCNT_dec (self->incr_text);
1900 2301
1901PROTOTYPES: ENABLE 2302PROTOTYPES: ENABLE
1902 2303
1903void encode_json (SV *scalar) 2304void encode_json (SV *scalar)
1904 ALIAS:
1905 to_json_ = 0
1906 encode_json = F_UTF8
1907 PPCODE: 2305 PPCODE:
1908{ 2306{
1909 JSON json; 2307 JSON json;
1910 json_init (&json); 2308 json_init (&json);
1911 json.flags |= ix; 2309 json.flags |= F_UTF8;
1912 XPUSHs (encode_json (scalar, &json)); 2310 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2311 XPUSHs (scalar);
1913} 2312}
1914 2313
1915void decode_json (SV *jsonstr) 2314void decode_json (SV *jsonstr)
1916 ALIAS:
1917 from_json_ = 0
1918 decode_json = F_UTF8
1919 PPCODE: 2315 PPCODE:
1920{ 2316{
1921 JSON json; 2317 JSON json;
1922 json_init (&json); 2318 json_init (&json);
1923 json.flags |= ix; 2319 json.flags |= F_UTF8;
1924 XPUSHs (decode_json (jsonstr, &json, 0)); 2320 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2321 XPUSHs (jsonstr);
1925} 2322}
1926 2323
1927

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines