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 1.83 by root, Wed Mar 26 22:54:38 2008 UTC vs.
Revision 1.131 by root, Thu Aug 17 01:42:20 2017 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 15#endif
15 16
16// some old perls do not have this, try to make it work, no 17// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 18// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
20#endif 21#endif
21 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 36
24#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
31#define F_ALLOW_NONREF 0x00000100UL 44#define F_ALLOW_NONREF 0x00000100UL
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 49#define F_ALLOW_UNKNOWN 0x00002000UL
37#define F_MAXDEPTH 0xf8000000UL 50#define F_ALLOW_TAGS 0x00004000UL
38#define S_MAXDEPTH 27
39#define F_MAXSIZE 0x01f00000UL
40#define S_MAXSIZE 20
41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42 52
43#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
44#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
45
46#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
47#define F_DEFAULT (9UL << S_MAXDEPTH)
48 54
49#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
51 57
52#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)
53 61
54#define SB do { 62#define SB do {
55#define SE } while (0) 63#define SE } while (0)
56 64
57#if __GNUC__ >= 3 65#if __GNUC__ >= 3
67 75
68#define IN_RANGE_INC(type,val,beg,end) \ 76#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 77 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71 79
80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
81
72#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
73# define JSON_SLOW 1 83# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 84# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
85# define BOOL_STASH (bool_stash ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
75#else 86#else
76# define JSON_SLOW 0 87# define JSON_SLOW 0
77# define JSON_STASH json_stash 88# define JSON_STASH json_stash
89# define BOOL_STASH bool_stash
78#endif 90#endif
79 91
80static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
81static SV *json_true, *json_false; 93#define STACK_HES 64
94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_true, *bool_false, *sv_json;
82 97
83enum { 98enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0 99 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string 100 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash 101 INCR_M_BS, // inside backslash
102 INCR_M_C0, // inside comment in initial whitespace sequence
103 INCR_M_C1, // inside comment in other places
87 INCR_M_JSON // outside anything, count nesting 104 INCR_M_JSON // outside anything, count nesting
88}; 105};
89 106
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 107#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
91 108
92typedef struct { 109typedef struct {
93 U32 flags; 110 U32 flags;
111 U32 max_depth;
112 STRLEN max_size;
113
94 SV *cb_object; 114 SV *cb_object;
95 HV *cb_sk_object; 115 HV *cb_sk_object;
96 116
97 // for the incremental parser 117 // for the incremental parser
98 SV *incr_text; // the source text so far 118 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 119 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 120 int incr_nest; // {[]}-nesting level
101 int incr_mode; 121 unsigned char incr_mode;
102} JSON; 122} JSON;
123
124INLINE void
125json_init (JSON *json)
126{
127 Zero (json, 1, JSON);
128 json->max_depth = 512;
129}
103 130
104///////////////////////////////////////////////////////////////////////////// 131/////////////////////////////////////////////////////////////////////////////
105// utility functions 132// utility functions
106 133
107INLINE SV * 134INLINE SV *
117 144
118INLINE void 145INLINE void
119shrink (SV *sv) 146shrink (SV *sv)
120{ 147{
121 sv_utf8_downgrade (sv, 1); 148 sv_utf8_downgrade (sv, 1);
149
122 if (SvLEN (sv) > SvCUR (sv) + 1) 150 if (SvLEN (sv) > SvCUR (sv) + 1)
123 { 151 {
124#ifdef SvPV_shrink_to_cur 152#ifdef SvPV_shrink_to_cur
125 SvPV_shrink_to_cur (sv); 153 SvPV_shrink_to_cur (sv);
126#elif defined (SvPV_renew) 154#elif defined (SvPV_renew)
127 SvPV_renew (sv, SvCUR (sv) + 1); 155 SvPV_renew (sv, SvCUR (sv) + 1);
128#endif 156#endif
129 } 157 }
158}
159
160/* adds two STRLENs together, slow, and with paranoia */
161STRLEN
162strlen_sum (STRLEN l1, STRLEN l2)
163{
164 size_t sum = l1 + l2;
165
166 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
167 croak ("JSON::XS: string size overflow");
168
169 return sum;
170}
171
172/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
173static char *
174json_sv_grow (SV *sv, size_t len1, size_t len2)
175{
176 len1 = strlen_sum (len1, len2);
177 len1 = strlen_sum (len1, len1 >> 1);
178
179 if (len1 > 4096 - 24)
180 len1 = (len1 | 4095) - 24;
181
182 return SvGROW (sv, len1);
130} 183}
131 184
132// decode an utf-8 character and return it, or (UV)-1 in 185// decode an utf-8 character and return it, or (UV)-1 in
133// case of an error. 186// case of an error.
134// we special-case "safe" characters from U+80 .. U+7FF, 187// we special-case "safe" characters from U+80 .. U+7FF,
170 *s++ = 0x80 | ( ch & 0x3f); 223 *s++ = 0x80 | ( ch & 0x3f);
171 224
172 return s; 225 return s;
173} 226}
174 227
228// convert offset pointer to character index, sv must be string
229static STRLEN
230ptr_to_index (SV *sv, char *offset)
231{
232 return SvUTF8 (sv)
233 ? utf8_distance (offset, SvPVX (sv))
234 : offset - SvPVX (sv);
235}
236
237/////////////////////////////////////////////////////////////////////////////
238// fp hell
239
240// scan a group of digits, and a trailing exponent
241static void
242json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
243{
244 UV uaccum = 0;
245 int eaccum = 0;
246
247 // if we recurse too deep, skip all remaining digits
248 // to avoid a stack overflow attack
249 if (expect_false (--maxdepth <= 0))
250 while (((U8)*s - '0') < 10)
251 ++s;
252
253 for (;;)
254 {
255 U8 dig = (U8)*s - '0';
256
257 if (expect_false (dig >= 10))
258 {
259 if (dig == (U8)((U8)'.' - (U8)'0'))
260 {
261 ++s;
262 json_atof_scan1 (s, accum, expo, 1, maxdepth);
263 }
264 else if ((dig | ' ') == 'e' - '0')
265 {
266 int exp2 = 0;
267 int neg = 0;
268
269 ++s;
270
271 if (*s == '-')
272 {
273 ++s;
274 neg = 1;
275 }
276 else if (*s == '+')
277 ++s;
278
279 while ((dig = (U8)*s - '0') < 10)
280 exp2 = exp2 * 10 + *s++ - '0';
281
282 *expo += neg ? -exp2 : exp2;
283 }
284
285 break;
286 }
287
288 ++s;
289
290 uaccum = uaccum * 10 + dig;
291 ++eaccum;
292
293 // if we have too many digits, then recurse for more
294 // we actually do this for rather few digits
295 if (uaccum >= (UV_MAX - 9) / 10)
296 {
297 if (postdp) *expo -= eaccum;
298 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
299 if (postdp) *expo += eaccum;
300
301 break;
302 }
303 }
304
305 // this relies greatly on the quality of the pow ()
306 // implementation of the platform, but a good
307 // implementation is hard to beat.
308 // (IEEE 754 conformant ones are required to be exact)
309 if (postdp) *expo -= eaccum;
310 *accum += uaccum * Perl_pow (10., *expo);
311 *expo += eaccum;
312}
313
314static NV
315json_atof (const char *s)
316{
317 NV accum = 0.;
318 int expo = 0;
319 int neg = 0;
320
321 if (*s == '-')
322 {
323 ++s;
324 neg = 1;
325 }
326
327 // a recursion depth of ten gives us >>500 bits
328 json_atof_scan1 (s, &accum, &expo, 0, 10);
329
330 return neg ? -accum : accum;
331}
332
333// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
334static int
335ref_bool_type (SV *sv)
336{
337 svtype svt = SvTYPE (sv);
338
339 if (svt < SVt_PVAV)
340 {
341 STRLEN len = 0;
342 char *pv = svt ? SvPV (sv, len) : 0;
343
344 if (len == 1)
345 if (*pv == '1')
346 return 1;
347 else if (*pv == '0')
348 return 0;
349 }
350
351 return -1;
352}
353
354// returns whether scalar is not a reference in the sense of allow_nonref
355static int
356json_nonref (SV *scalar)
357{
358 if (!SvROK (scalar))
359 return 1;
360
361 scalar = SvRV (scalar);
362
363 if (SvTYPE (scalar) >= SVt_PVMG)
364 {
365 if (SvSTASH (scalar) == bool_stash)
366 return 1;
367
368 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
369 return 1;
370 }
371
372 return 0;
373}
374
175///////////////////////////////////////////////////////////////////////////// 375/////////////////////////////////////////////////////////////////////////////
176// encoder 376// encoder
177 377
178// structure used for encoding JSON 378// structure used for encoding JSON
179typedef struct 379typedef struct
181 char *cur; // SvPVX (sv) + current output position 381 char *cur; // SvPVX (sv) + current output position
182 char *end; // SvEND (sv) 382 char *end; // SvEND (sv)
183 SV *sv; // result scalar 383 SV *sv; // result scalar
184 JSON json; 384 JSON json;
185 U32 indent; // indentation level 385 U32 indent; // indentation level
186 U32 maxdepth; // max. indentation/recursion level
187 UV limit; // escape character values >= this value when encoding 386 UV limit; // escape character values >= this value when encoding
188} enc_t; 387} enc_t;
189 388
190INLINE void 389INLINE void
191need (enc_t *enc, STRLEN len) 390need (enc_t *enc, STRLEN len)
192{ 391{
193 if (expect_false (enc->cur + len >= enc->end)) 392 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
194 { 393 {
195 STRLEN cur = enc->cur - SvPVX (enc->sv); 394 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
196 SvGROW (enc->sv, cur + len + 1); 395 char *buf = json_sv_grow (enc->sv, cur, len);
197 enc->cur = SvPVX (enc->sv) + cur; 396 enc->cur = buf + cur;
198 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 397 enc->end = buf + SvLEN (enc->sv) - 1;
199 } 398 }
200} 399}
201 400
202INLINE void 401INLINE void
203encode_ch (enc_t *enc, char ch) 402encode_ch (enc_t *enc, char ch)
219 418
220 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 419 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
221 { 420 {
222 if (expect_false (ch == '"')) // but with slow exceptions 421 if (expect_false (ch == '"')) // but with slow exceptions
223 { 422 {
224 need (enc, len += 1); 423 need (enc, len + 1);
225 *enc->cur++ = '\\'; 424 *enc->cur++ = '\\';
226 *enc->cur++ = '"'; 425 *enc->cur++ = '"';
227 } 426 }
228 else if (expect_false (ch == '\\')) 427 else if (expect_false (ch == '\\'))
229 { 428 {
230 need (enc, len += 1); 429 need (enc, len + 1);
231 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
232 *enc->cur++ = '\\'; 431 *enc->cur++ = '\\';
233 } 432 }
234 else 433 else
235 *enc->cur++ = ch; 434 *enc->cur++ = ch;
238 } 437 }
239 else 438 else
240 { 439 {
241 switch (ch) 440 switch (ch)
242 { 441 {
243 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 442 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
244 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 443 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
245 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 444 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
246 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 445 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
247 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 446 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
248 447
249 default: 448 default:
250 { 449 {
251 STRLEN clen; 450 STRLEN clen;
252 UV uch; 451 UV uch;
268 if (uch >= 0x10000UL) 467 if (uch >= 0x10000UL)
269 { 468 {
270 if (uch >= 0x110000UL) 469 if (uch >= 0x110000UL)
271 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 470 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
272 471
273 need (enc, len += 11); 472 need (enc, len + 11);
274 sprintf (enc->cur, "\\u%04x\\u%04x", 473 sprintf (enc->cur, "\\u%04x\\u%04x",
275 (int)((uch - 0x10000) / 0x400 + 0xD800), 474 (int)((uch - 0x10000) / 0x400 + 0xD800),
276 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 475 (int)((uch - 0x10000) % 0x400 + 0xDC00));
277 enc->cur += 12; 476 enc->cur += 12;
278 } 477 }
279 else 478 else
280 { 479 {
281 static char hexdigit [16] = "0123456789abcdef";
282 need (enc, len += 5); 480 need (enc, len + 5);
283 *enc->cur++ = '\\'; 481 *enc->cur++ = '\\';
284 *enc->cur++ = 'u'; 482 *enc->cur++ = 'u';
285 *enc->cur++ = hexdigit [ uch >> 12 ]; 483 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
286 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 484 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
287 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 485 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
288 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 486 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
289 } 487 }
290 488
291 str += clen; 489 str += clen;
292 } 490 }
293 else if (enc->json.flags & F_LATIN1) 491 else if (enc->json.flags & F_LATIN1)
295 *enc->cur++ = uch; 493 *enc->cur++ = uch;
296 str += clen; 494 str += clen;
297 } 495 }
298 else if (is_utf8) 496 else if (is_utf8)
299 { 497 {
300 need (enc, len += clen); 498 need (enc, len + clen);
301 do 499 do
302 { 500 {
303 *enc->cur++ = *str++; 501 *enc->cur++ = *str++;
304 } 502 }
305 while (--clen); 503 while (--clen);
306 } 504 }
307 else 505 else
308 { 506 {
309 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 507 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
310 enc->cur = encode_utf8 (enc->cur, uch); 508 enc->cur = encode_utf8 (enc->cur, uch);
311 ++str; 509 ++str;
312 } 510 }
313 } 511 }
314 } 512 }
364static void 562static void
365encode_av (enc_t *enc, AV *av) 563encode_av (enc_t *enc, AV *av)
366{ 564{
367 int i, len = av_len (av); 565 int i, len = av_len (av);
368 566
369 if (enc->indent >= enc->maxdepth) 567 if (enc->indent >= enc->json.max_depth)
370 croak ("data structure too deep (hit recursion limit)"); 568 croak (ERR_NESTING_EXCEEDED);
371 569
372 encode_ch (enc, '['); 570 encode_ch (enc, '[');
373 571
374 if (len >= 0) 572 if (len >= 0)
375 { 573 {
376 encode_nl (enc); ++enc->indent; 574 encode_nl (enc); ++enc->indent;
377 575
378 for (i = 0; i <= len; ++i) 576 for (i = 0; i <= len; ++i)
390 encode_comma (enc); 588 encode_comma (enc);
391 } 589 }
392 590
393 encode_nl (enc); --enc->indent; encode_indent (enc); 591 encode_nl (enc); --enc->indent; encode_indent (enc);
394 } 592 }
395 593
396 encode_ch (enc, ']'); 594 encode_ch (enc, ']');
397} 595}
398 596
399static void 597static void
400encode_hk (enc_t *enc, HE *he) 598encode_hk (enc_t *enc, HE *he)
404 if (HeKLEN (he) == HEf_SVKEY) 602 if (HeKLEN (he) == HEf_SVKEY)
405 { 603 {
406 SV *sv = HeSVKEY (he); 604 SV *sv = HeSVKEY (he);
407 STRLEN len; 605 STRLEN len;
408 char *str; 606 char *str;
409 607
410 SvGETMAGIC (sv); 608 SvGETMAGIC (sv);
411 str = SvPV (sv, len); 609 str = SvPV (sv, len);
412 610
413 encode_str (enc, str, len, SvUTF8 (sv)); 611 encode_str (enc, str, len, SvUTF8 (sv));
414 } 612 }
450static void 648static void
451encode_hv (enc_t *enc, HV *hv) 649encode_hv (enc_t *enc, HV *hv)
452{ 650{
453 HE *he; 651 HE *he;
454 652
455 if (enc->indent >= enc->maxdepth) 653 if (enc->indent >= enc->json.max_depth)
456 croak ("data structure too deep (hit recursion limit)"); 654 croak (ERR_NESTING_EXCEEDED);
457 655
458 encode_ch (enc, '{'); 656 encode_ch (enc, '{');
459 657
460 // for canonical output we have to sort by keys first 658 // for canonical output we have to sort by keys first
461 // actually, this is mostly due to the stupid so-called 659 // actually, this is mostly due to the stupid so-called
462 // security workaround added somewhere in 5.8.x. 660 // security workaround added somewhere in 5.8.x
463 // that randomises hash orderings 661 // that randomises hash orderings
464 if (enc->json.flags & F_CANONICAL) 662 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
465 { 663 {
466 int count = hv_iterinit (hv); 664 int count = hv_iterinit (hv);
467 665
468 if (SvMAGICAL (hv)) 666 if (SvMAGICAL (hv))
469 { 667 {
480 } 678 }
481 679
482 if (count) 680 if (count)
483 { 681 {
484 int i, fast = 1; 682 int i, fast = 1;
485#if defined(__BORLANDC__) || defined(_MSC_VER) 683 HE *hes_stack [STACK_HES];
486 HE **hes = _alloca (count * sizeof (HE)); 684 HE **hes = hes_stack;
487#else 685
488 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 686 // allocate larger arrays on the heap
489#endif 687 if (count > STACK_HES)
688 {
689 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
690 hes = (HE **)SvPVX (sv);
691 }
490 692
491 i = 0; 693 i = 0;
492 while ((he = hv_iternext (hv))) 694 while ((he = hv_iternext (hv)))
493 { 695 {
494 hes [i++] = he; 696 hes [i++] = he;
563// encode objects, arrays and special \0=false and \1=true values. 765// encode objects, arrays and special \0=false and \1=true values.
564static void 766static void
565encode_rv (enc_t *enc, SV *sv) 767encode_rv (enc_t *enc, SV *sv)
566{ 768{
567 svtype svt; 769 svtype svt;
770 GV *method;
568 771
569 SvGETMAGIC (sv); 772 SvGETMAGIC (sv);
570 svt = SvTYPE (sv); 773 svt = SvTYPE (sv);
571 774
572 if (expect_false (SvOBJECT (sv))) 775 if (expect_false (SvOBJECT (sv)))
573 { 776 {
574 HV *stash = !JSON_SLOW || json_boolean_stash 777 HV *stash = SvSTASH (sv);
575 ? json_boolean_stash
576 : gv_stashpv ("JSON::XS::Boolean", 1);
577 778
578 if (SvSTASH (sv) == stash) 779 if (stash == bool_stash)
579 { 780 {
580 if (SvIV (sv)) 781 if (SvIV (sv))
581 encode_str (enc, "true", 4, 0); 782 encode_str (enc, "true", 4, 0);
582 else 783 else
583 encode_str (enc, "false", 5, 0); 784 encode_str (enc, "false", 5, 0);
584 } 785 }
786 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
787 {
788 int count;
789 dSP;
790
791 ENTER; SAVETMPS;
792 SAVESTACK_POS ();
793 PUSHMARK (SP);
794 EXTEND (SP, 2);
795 // we re-bless the reference to get overload and other niceties right
796 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
797 PUSHs (sv_json);
798
799 PUTBACK;
800 count = call_sv ((SV *)GvCV (method), G_ARRAY);
801 SPAGAIN;
802
803 // catch this surprisingly common error
804 if (SvROK (TOPs) && SvRV (TOPs) == sv)
805 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
806
807 encode_ch (enc, '(');
808 encode_ch (enc, '"');
809 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
810 encode_ch (enc, '"');
811 encode_ch (enc, ')');
812 encode_ch (enc, '[');
813
814 while (count)
815 {
816 encode_sv (enc, SP[1 - count--]);
817
818 if (count)
819 encode_ch (enc, ',');
820 }
821
822 encode_ch (enc, ']');
823
824 FREETMPS; LEAVE;
825 }
826 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
827 {
828 dSP;
829
830 ENTER; SAVETMPS;
831 PUSHMARK (SP);
832 // we re-bless the reference to get overload and other niceties right
833 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
834
835 // calling with G_SCALAR ensures that we always get a 1 return value
836 PUTBACK;
837 call_sv ((SV *)GvCV (method), G_SCALAR);
838 SPAGAIN;
839
840 // catch this surprisingly common error
841 if (SvROK (TOPs) && SvRV (TOPs) == sv)
842 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
843
844 sv = POPs;
845 PUTBACK;
846
847 encode_sv (enc, sv);
848
849 FREETMPS; LEAVE;
850 }
851 else if (enc->json.flags & F_ALLOW_BLESSED)
852 encode_str (enc, "null", 4, 0);
585 else 853 else
586 {
587#if 0
588 if (0 && sv_derived_from (rv, "JSON::Literal"))
589 {
590 // not yet
591 }
592#endif
593 if (enc->json.flags & F_CONV_BLESSED)
594 {
595 // we re-bless the reference to get overload and other niceties right
596 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
597
598 if (to_json)
599 {
600 dSP;
601
602 ENTER; SAVETMPS; PUSHMARK (SP);
603 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
604
605 // calling with G_SCALAR ensures that we always get a 1 return value
606 PUTBACK;
607 call_sv ((SV *)GvCV (to_json), G_SCALAR);
608 SPAGAIN;
609
610 // catch this surprisingly common error
611 if (SvROK (TOPs) && SvRV (TOPs) == sv)
612 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
613
614 sv = POPs;
615 PUTBACK;
616
617 encode_sv (enc, sv);
618
619 FREETMPS; LEAVE;
620 }
621 else if (enc->json.flags & F_ALLOW_BLESSED)
622 encode_str (enc, "null", 4, 0);
623 else
624 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
625 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
626 }
627 else if (enc->json.flags & F_ALLOW_BLESSED)
628 encode_str (enc, "null", 4, 0);
629 else
630 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 854 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
631 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 855 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
632 }
633 } 856 }
634 else if (svt == SVt_PVHV) 857 else if (svt == SVt_PVHV)
635 encode_hv (enc, (HV *)sv); 858 encode_hv (enc, (HV *)sv);
636 else if (svt == SVt_PVAV) 859 else if (svt == SVt_PVAV)
637 encode_av (enc, (AV *)sv); 860 encode_av (enc, (AV *)sv);
638 else if (svt < SVt_PVAV) 861 else if (svt < SVt_PVAV)
639 { 862 {
640 STRLEN len = 0; 863 int bool_type = ref_bool_type (sv);
641 char *pv = svt ? SvPV (sv, len) : 0;
642 864
643 if (len == 1 && *pv == '1') 865 if (bool_type == 1)
644 encode_str (enc, "true", 4, 0); 866 encode_str (enc, "true", 4, 0);
645 else if (len == 1 && *pv == '0') 867 else if (bool_type == 0)
646 encode_str (enc, "false", 5, 0); 868 encode_str (enc, "false", 5, 0);
869 else if (enc->json.flags & F_ALLOW_UNKNOWN)
870 encode_str (enc, "null", 4, 0);
647 else 871 else
648 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 872 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
649 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 873 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
650 } 874 }
875 else if (enc->json.flags & F_ALLOW_UNKNOWN)
876 encode_str (enc, "null", 4, 0);
651 else 877 else
652 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 878 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
653 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 879 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
654} 880}
655 881
709 } 935 }
710 else 936 else
711 { 937 {
712 // large integer, use the (rather slow) snprintf way. 938 // large integer, use the (rather slow) snprintf way.
713 need (enc, IVUV_MAXCHARS); 939 need (enc, IVUV_MAXCHARS);
714 enc->cur += 940 enc->cur +=
715 SvIsUV(sv) 941 SvIsUV(sv)
716 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 942 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
717 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 943 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
718 } 944 }
719 } 945 }
720 else if (SvROK (sv)) 946 else if (SvROK (sv))
721 encode_rv (enc, SvRV (sv)); 947 encode_rv (enc, SvRV (sv));
722 else if (!SvOK (sv)) 948 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
723 encode_str (enc, "null", 4, 0); 949 encode_str (enc, "null", 4, 0);
724 else 950 else
725 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 951 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
726 SvPV_nolen (sv), SvFLAGS (sv)); 952 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
727} 953}
728 954
729static SV * 955static SV *
730encode_json (SV *scalar, JSON *json) 956encode_json (SV *scalar, JSON *json)
731{ 957{
732 enc_t enc; 958 enc_t enc;
733 959
734 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 960 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
735 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 961 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
736 962
737 enc.json = *json; 963 enc.json = *json;
738 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 964 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
739 enc.cur = SvPVX (enc.sv); 965 enc.cur = SvPVX (enc.sv);
740 enc.end = SvEND (enc.sv); 966 enc.end = SvEND (enc.sv);
741 enc.indent = 0; 967 enc.indent = 0;
742 enc.maxdepth = DEC_DEPTH (enc.json.flags);
743 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 968 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
744 : enc.json.flags & F_LATIN1 ? 0x000100UL 969 : enc.json.flags & F_LATIN1 ? 0x000100UL
745 : 0x110000UL; 970 : 0x110000UL;
746 971
747 SvPOK_only (enc.sv); 972 SvPOK_only (enc.sv);
748 encode_sv (&enc, scalar); 973 encode_sv (&enc, scalar);
974 encode_nl (&enc);
749 975
750 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 976 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
751 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 977 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
752 978
753 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 979 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
814 if (*dec->cur != ch) \ 1040 if (*dec->cur != ch) \
815 ERR (# ch " expected"); \ 1041 ERR (# ch " expected"); \
816 ++dec->cur; \ 1042 ++dec->cur; \
817 SE 1043 SE
818 1044
819#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1045#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
820#define DEC_DEC_DEPTH --dec->depth 1046#define DEC_DEC_DEPTH --dec->depth
821 1047
822static SV *decode_sv (dec_t *dec); 1048static SV *decode_sv (dec_t *dec);
823 1049
824static signed char decode_hexdigit[256]; 1050static signed char decode_hexdigit[256];
933 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1159 else if (expect_true (ch >= 0x20 && ch < 0x80))
934 *cur++ = ch; 1160 *cur++ = ch;
935 else if (ch >= 0x80) 1161 else if (ch >= 0x80)
936 { 1162 {
937 STRLEN clen; 1163 STRLEN clen;
938 UV uch;
939 1164
940 --dec_cur; 1165 --dec_cur;
941 1166
942 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1167 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
943 if (clen == (STRLEN)-1) 1168 if (clen == (STRLEN)-1)
944 ERR ("malformed UTF-8 character in JSON string"); 1169 ERR ("malformed UTF-8 character in JSON string");
945 1170
946 do 1171 do
947 *cur++ = *dec_cur++; 1172 *cur++ = *dec_cur++;
948 while (--clen); 1173 while (--clen);
949 1174
950 utf8 = 1; 1175 utf8 = 1;
951 } 1176 }
1177 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1178 *cur++ = ch;
952 else 1179 else
953 { 1180 {
954 --dec_cur; 1181 --dec_cur;
955 1182
956 if (!ch) 1183 if (!ch)
964 { 1191 {
965 STRLEN len = cur - buf; 1192 STRLEN len = cur - buf;
966 1193
967 if (sv) 1194 if (sv)
968 { 1195 {
969 SvGROW (sv, SvCUR (sv) + len + 1); 1196 STRLEN cur = SvCUR (sv);
1197
1198 if (SvLEN (sv) - cur <= len)
1199 json_sv_grow (sv, cur, len);
1200
970 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1201 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
971 SvCUR_set (sv, SvCUR (sv) + len); 1202 SvCUR_set (sv, SvCUR (sv) + len);
972 } 1203 }
973 else 1204 else
974 sv = newSVpvn (buf, len); 1205 sv = newSVpvn (buf, len);
1065 1296
1066 // special case the rather common 1..5-digit-int case 1297 // special case the rather common 1..5-digit-int case
1067 if (*start == '-') 1298 if (*start == '-')
1068 switch (len) 1299 switch (len)
1069 { 1300 {
1070 case 2: return newSViv (-( start [1] - '0' * 1)); 1301 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1071 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1302 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1072 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1303 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1073 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1304 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1074 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1305 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1075 } 1306 }
1076 else 1307 else
1077 switch (len) 1308 switch (len)
1078 { 1309 {
1079 case 1: return newSViv ( start [0] - '0' * 1); 1310 case 1: return newSViv ( start [0] - '0' * 1);
1080 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1311 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1081 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1312 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1082 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1313 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1083 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1314 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1084 } 1315 }
1085 1316
1086 { 1317 {
1087 UV uv; 1318 UV uv;
1088 int numtype = grok_number (start, len, &uv); 1319 int numtype = grok_number (start, len, &uv);
1097 } 1328 }
1098 1329
1099 len -= *start == '-' ? 1 : 0; 1330 len -= *start == '-' ? 1 : 0;
1100 1331
1101 // does not fit into IV or UV, try NV 1332 // does not fit into IV or UV, try NV
1102 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1333 if (len <= NV_DIG)
1103 #if defined (LDBL_DIG)
1104 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1105 #endif
1106 )
1107 // fits into NV without loss of precision 1334 // fits into NV without loss of precision
1108 return newSVnv (Atof (start)); 1335 return newSVnv (json_atof (start));
1109 1336
1110 // everything else fails, convert it to a string 1337 // everything else fails, convert it to a string
1111 return newSVpvn (start, dec->cur - start); 1338 return newSVpvn (start, dec->cur - start);
1112 } 1339 }
1113 1340
1114 // loss of precision here 1341 // loss of precision here
1115 return newSVnv (Atof (start)); 1342 return newSVnv (json_atof (start));
1116 1343
1117fail: 1344fail:
1118 return 0; 1345 return 0;
1119} 1346}
1120 1347
1144 if (*dec->cur == ']') 1371 if (*dec->cur == ']')
1145 { 1372 {
1146 ++dec->cur; 1373 ++dec->cur;
1147 break; 1374 break;
1148 } 1375 }
1149 1376
1150 if (*dec->cur != ',') 1377 if (*dec->cur != ',')
1151 ERR (", or ] expected while parsing array"); 1378 ERR (", or ] expected while parsing array");
1152 1379
1153 ++dec->cur; 1380 ++dec->cur;
1154 1381
1278 1505
1279 hv_iterinit (hv); 1506 hv_iterinit (hv);
1280 he = hv_iternext (hv); 1507 he = hv_iternext (hv);
1281 hv_iterinit (hv); 1508 hv_iterinit (hv);
1282 1509
1283 // the next line creates a mortal sv each time its called. 1510 // the next line creates a mortal sv each time it's called.
1284 // might want to optimise this for common cases. 1511 // might want to optimise this for common cases.
1285 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1512 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1286 1513
1287 if (cb) 1514 if (cb)
1288 { 1515 {
1289 dSP; 1516 dSP;
1290 int count; 1517 int count;
1291 1518
1292 ENTER; SAVETMPS; PUSHMARK (SP); 1519 ENTER; SAVETMPS;
1520 SAVESTACK_POS ();
1521 PUSHMARK (SP);
1293 XPUSHs (HeVAL (he)); 1522 XPUSHs (HeVAL (he));
1523 sv_2mortal (sv);
1294 1524
1295 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1525 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1296 1526
1297 if (count == 1) 1527 if (count == 1)
1298 { 1528 {
1299 sv = newSVsv (POPs); 1529 sv = newSVsv (POPs);
1300 FREETMPS; LEAVE; 1530 FREETMPS; LEAVE;
1301 return sv; 1531 return sv;
1302 } 1532 }
1303 1533
1534 SvREFCNT_inc (sv);
1304 FREETMPS; LEAVE; 1535 FREETMPS; LEAVE;
1305 } 1536 }
1306 } 1537 }
1307 1538
1308 if (dec->json.cb_object) 1539 if (dec->json.cb_object)
1309 { 1540 {
1310 dSP; 1541 dSP;
1311 int count; 1542 int count;
1312 1543
1313 ENTER; SAVETMPS; PUSHMARK (SP); 1544 ENTER; SAVETMPS;
1545 SAVESTACK_POS ();
1546 PUSHMARK (SP);
1314 XPUSHs (sv_2mortal (sv)); 1547 XPUSHs (sv_2mortal (sv));
1315 1548
1316 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1549 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1317 1550
1318 if (count == 1) 1551 if (count == 1)
1334 DEC_DEC_DEPTH; 1567 DEC_DEC_DEPTH;
1335 return 0; 1568 return 0;
1336} 1569}
1337 1570
1338static SV * 1571static SV *
1572decode_tag (dec_t *dec)
1573{
1574 SV *tag = 0;
1575 SV *val = 0;
1576
1577 if (!(dec->json.flags & F_ALLOW_TAGS))
1578 ERR ("malformed JSON string, neither array, object, number, string or atom");
1579
1580 ++dec->cur;
1581
1582 decode_ws (dec);
1583
1584 tag = decode_sv (dec);
1585 if (!tag)
1586 goto fail;
1587
1588 if (!SvPOK (tag))
1589 ERR ("malformed JSON string, (tag) must be a string");
1590
1591 decode_ws (dec);
1592
1593 if (*dec->cur != ')')
1594 ERR (") expected after tag");
1595
1596 ++dec->cur;
1597
1598 decode_ws (dec);
1599
1600 val = decode_sv (dec);
1601 if (!val)
1602 goto fail;
1603
1604 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1605 ERR ("malformed JSON string, tag value must be an array");
1606
1607 {
1608 AV *av = (AV *)SvRV (val);
1609 int i, len = av_len (av) + 1;
1610 HV *stash = gv_stashsv (tag, 0);
1611 SV *sv;
1612
1613 if (!stash)
1614 ERR ("cannot decode perl-object (package does not exist)");
1615
1616 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1617
1618 if (!method)
1619 ERR ("cannot decode perl-object (package does not have a THAW method)");
1620
1621 dSP;
1622
1623 ENTER; SAVETMPS;
1624 PUSHMARK (SP);
1625 EXTEND (SP, len + 2);
1626 // we re-bless the reference to get overload and other niceties right
1627 PUSHs (tag);
1628 PUSHs (sv_json);
1629
1630 for (i = 0; i < len; ++i)
1631 PUSHs (*av_fetch (av, i, 1));
1632
1633 PUTBACK;
1634 call_sv ((SV *)GvCV (method), G_SCALAR);
1635 SPAGAIN;
1636
1637 SvREFCNT_dec (tag);
1638 SvREFCNT_dec (val);
1639 sv = SvREFCNT_inc (POPs);
1640
1641 PUTBACK;
1642
1643 FREETMPS; LEAVE;
1644
1645 return sv;
1646 }
1647
1648fail:
1649 SvREFCNT_dec (tag);
1650 SvREFCNT_dec (val);
1651 return 0;
1652}
1653
1654static SV *
1339decode_sv (dec_t *dec) 1655decode_sv (dec_t *dec)
1340{ 1656{
1341 // the beauty of JSON: you need exactly one character lookahead 1657 // the beauty of JSON: you need exactly one character lookahead
1342 // to parse everything. 1658 // to parse everything.
1343 switch (*dec->cur) 1659 switch (*dec->cur)
1344 { 1660 {
1345 case '"': ++dec->cur; return decode_str (dec); 1661 case '"': ++dec->cur; return decode_str (dec);
1346 case '[': ++dec->cur; return decode_av (dec); 1662 case '[': ++dec->cur; return decode_av (dec);
1347 case '{': ++dec->cur; return decode_hv (dec); 1663 case '{': ++dec->cur; return decode_hv (dec);
1664 case '(': return decode_tag (dec);
1348 1665
1349 case '-': 1666 case '-':
1350 case '0': case '1': case '2': case '3': case '4': 1667 case '0': case '1': case '2': case '3': case '4':
1351 case '5': case '6': case '7': case '8': case '9': 1668 case '5': case '6': case '7': case '8': case '9':
1352 return decode_num (dec); 1669 return decode_num (dec);
1354 case 't': 1671 case 't':
1355 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1672 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1356 { 1673 {
1357 dec->cur += 4; 1674 dec->cur += 4;
1358#if JSON_SLOW 1675#if JSON_SLOW
1359 json_true = get_bool ("JSON::XS::true"); 1676 bool_true = get_bool ("Types::Serialiser::true");
1360#endif 1677#endif
1361 return newSVsv (json_true); 1678 return newSVsv (bool_true);
1362 } 1679 }
1363 else 1680 else
1364 ERR ("'true' expected"); 1681 ERR ("'true' expected");
1365 1682
1366 break; 1683 break;
1368 case 'f': 1685 case 'f':
1369 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1686 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1370 { 1687 {
1371 dec->cur += 5; 1688 dec->cur += 5;
1372#if JSON_SLOW 1689#if JSON_SLOW
1373 json_false = get_bool ("JSON::XS::false"); 1690 bool_false = get_bool ("Types::Serialiser::false");
1374#endif 1691#endif
1375 return newSVsv (json_false); 1692 return newSVsv (bool_false);
1376 } 1693 }
1377 else 1694 else
1378 ERR ("'false' expected"); 1695 ERR ("'false' expected");
1379 1696
1380 break; 1697 break;
1389 ERR ("'null' expected"); 1706 ERR ("'null' expected");
1390 1707
1391 break; 1708 break;
1392 1709
1393 default: 1710 default:
1394 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1711 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1395 break; 1712 break;
1396 } 1713 }
1397 1714
1398fail: 1715fail:
1399 return 0; 1716 return 0;
1401 1718
1402static SV * 1719static SV *
1403decode_json (SV *string, JSON *json, STRLEN *offset_return) 1720decode_json (SV *string, JSON *json, STRLEN *offset_return)
1404{ 1721{
1405 dec_t dec; 1722 dec_t dec;
1406 STRLEN offset;
1407 SV *sv; 1723 SV *sv;
1408 1724
1725 /* work around bugs in 5.10 where manipulating magic values
1726 * makes perl ignore the magic in subsequent accesses.
1727 * also make a copy of non-PV values, to get them into a clean
1728 * state (SvPV should do that, but it's buggy, see below).
1729 *
1730 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1731 * as reported by Reini Urban.
1732 */
1409 SvGETMAGIC (string); 1733 /*SvGETMAGIC (string);*/
1734 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1735 string = sv_2mortal (newSVsv (string));
1736
1410 SvUPGRADE (string, SVt_PV); 1737 SvUPGRADE (string, SVt_PV);
1411 1738
1412 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1739 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1740 * assertion with -DDEBUGGING, although SvCUR is documented to
1741 * return the xpv_cur field which certainly exists after upgrading.
1742 * according to nicholas clark, calling SvPOK fixes this.
1743 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1744 * and hope for the best.
1745 * Damnit, SvPV_nolen still trips over yet another assertion. This
1746 * assertion business is seriously broken, try yet another workaround
1747 * for the broken -DDEBUGGING.
1748 */
1749 {
1750#ifdef DEBUGGING
1751 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1752#else
1753 STRLEN offset = SvCUR (string);
1754#endif
1755
1756 if (offset > json->max_size && json->max_size)
1413 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1757 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1414 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1758 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1759 }
1415 1760
1416 if (json->flags & F_UTF8) 1761 if (DECODE_WANTS_OCTETS (json))
1417 sv_utf8_downgrade (string, 0); 1762 sv_utf8_downgrade (string, 0);
1418 else 1763 else
1419 sv_utf8_upgrade (string); 1764 sv_utf8_upgrade (string);
1420 1765
1421 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1766 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1422 1767
1423 dec.json = *json; 1768 dec.json = *json;
1424 dec.cur = SvPVX (string); 1769 dec.cur = SvPVX (string);
1425 dec.end = SvEND (string); 1770 dec.end = SvEND (string);
1426 dec.err = 0; 1771 dec.err = 0;
1427 dec.depth = 0; 1772 dec.depth = 0;
1428 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1429 1773
1430 if (dec.json.cb_object || dec.json.cb_sk_object) 1774 if (dec.json.cb_object || dec.json.cb_sk_object)
1431 dec.json.flags |= F_HOOK; 1775 dec.json.flags |= F_HOOK;
1432 1776
1433 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1777 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1434 1778
1435 decode_ws (&dec); 1779 decode_ws (&dec);
1436 sv = decode_sv (&dec); 1780 sv = decode_sv (&dec);
1437 1781
1438 if (!(offset_return || !sv)) 1782 if (offset_return)
1783 *offset_return = dec.cur - SvPVX (string);
1784 else if (sv)
1439 { 1785 {
1440 // check for trailing garbage 1786 // check for trailing garbage
1441 decode_ws (&dec); 1787 decode_ws (&dec);
1442 1788
1443 if (*dec.cur) 1789 if (*dec.cur)
1444 { 1790 {
1445 dec.err = "garbage after JSON object"; 1791 dec.err = "garbage after JSON object";
1446 SvREFCNT_dec (sv); 1792 SvREFCNT_dec (sv);
1447 sv = 0; 1793 sv = 0;
1448 } 1794 }
1449 }
1450
1451 if (offset_return || !sv)
1452 {
1453 offset = dec.json.flags & F_UTF8
1454 ? dec.cur - SvPVX (string)
1455 : utf8_distance (dec.cur, SvPVX (string));
1456
1457 if (offset_return)
1458 *offset_return = offset;
1459 } 1795 }
1460 1796
1461 if (!sv) 1797 if (!sv)
1462 { 1798 {
1463 SV *uni = sv_newmortal (); 1799 SV *uni = sv_newmortal ();
1469 SAVEVPTR (PL_curcop); 1805 SAVEVPTR (PL_curcop);
1470 PL_curcop = &cop; 1806 PL_curcop = &cop;
1471 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1807 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1472 LEAVE; 1808 LEAVE;
1473 1809
1474 croak ("%s, at character offset %d [\"%s\"]", 1810 croak ("%s, at character offset %d (before \"%s\")",
1475 dec.err, 1811 dec.err,
1476 (int)offset, 1812 (int)ptr_to_index (string, dec.cur),
1477 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1813 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1478 } 1814 }
1479 1815
1480 sv = sv_2mortal (sv); 1816 sv = sv_2mortal (sv);
1481 1817
1482 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1818 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1483 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1819 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1484 1820
1485 return sv; 1821 return sv;
1486} 1822}
1487 1823
1490 1826
1491static void 1827static void
1492incr_parse (JSON *self) 1828incr_parse (JSON *self)
1493{ 1829{
1494 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1830 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1831
1832 // the state machine here is a bit convoluted and could be simplified a lot
1833 // but this would make it slower, so...
1495 1834
1496 for (;;) 1835 for (;;)
1497 { 1836 {
1498 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D 1837 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1499 switch (self->incr_mode) 1838 switch (self->incr_mode)
1500 { 1839 {
1501 // only used for intiial whitespace skipping 1840 // only used for initial whitespace skipping
1502 case INCR_M_WS: 1841 case INCR_M_WS:
1503 for (;;) 1842 for (;;)
1504 { 1843 {
1505 if (*p > 0x20) 1844 if (*p > 0x20)
1506 { 1845 {
1846 if (*p == '#')
1847 {
1848 self->incr_mode = INCR_M_C0;
1849 goto incr_m_c;
1850 }
1851 else
1852 {
1507 self->incr_mode = INCR_M_JSON; 1853 self->incr_mode = INCR_M_JSON;
1508 goto incr_m_json; 1854 goto incr_m_json;
1855 }
1509 } 1856 }
1510 else if (!*p) 1857 else if (!*p)
1511 goto interrupt; 1858 goto interrupt;
1512 1859
1513 ++p; 1860 ++p;
1519 goto interrupt; 1866 goto interrupt;
1520 1867
1521 ++p; 1868 ++p;
1522 self->incr_mode = INCR_M_STR; 1869 self->incr_mode = INCR_M_STR;
1523 goto incr_m_str; 1870 goto incr_m_str;
1871
1872 // inside #-style comments
1873 case INCR_M_C0:
1874 case INCR_M_C1:
1875 incr_m_c:
1876 for (;;)
1877 {
1878 if (*p == '\n')
1879 {
1880 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1881 break;
1882 }
1883 else if (!*p)
1884 goto interrupt;
1885
1886 ++p;
1887 }
1888
1889 break;
1524 1890
1525 // inside a string 1891 // inside a string
1526 case INCR_M_STR: 1892 case INCR_M_STR:
1527 incr_m_str: 1893 incr_m_str:
1528 for (;;) 1894 for (;;)
1579 self->incr_mode = INCR_M_STR; 1945 self->incr_mode = INCR_M_STR;
1580 goto incr_m_str; 1946 goto incr_m_str;
1581 1947
1582 case '[': 1948 case '[':
1583 case '{': 1949 case '{':
1950 case '(':
1584 ++self->incr_nest; 1951 if (++self->incr_nest > self->max_depth)
1952 croak (ERR_NESTING_EXCEEDED);
1585 break; 1953 break;
1586 1954
1587 case ']': 1955 case ']':
1588 case '}': 1956 case '}':
1589 if (!--self->incr_nest) 1957 if (--self->incr_nest <= 0)
1590 goto interrupt; 1958 goto interrupt;
1959 break;
1960
1961 case ')':
1962 --self->incr_nest;
1963 break;
1964
1965 case '#':
1966 self->incr_mode = INCR_M_C1;
1967 goto incr_m_c;
1591 } 1968 }
1592 } 1969 }
1593 } 1970 }
1594 1971
1595 modechange: 1972 modechange:
1596 ; 1973 ;
1597 } 1974 }
1598 1975
1599interrupt: 1976interrupt:
1600 self->incr_pos = p - SvPVX (self->incr_text); 1977 self->incr_pos = p - SvPVX (self->incr_text);
1978 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1601 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1979 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1602} 1980}
1603 1981
1604///////////////////////////////////////////////////////////////////////////// 1982/////////////////////////////////////////////////////////////////////////////
1605// XS interface functions 1983// XS interface functions
1615 i >= '0' && i <= '9' ? i - '0' 1993 i >= '0' && i <= '9' ? i - '0'
1616 : i >= 'a' && i <= 'f' ? i - 'a' + 10 1994 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1617 : i >= 'A' && i <= 'F' ? i - 'A' + 10 1995 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1618 : -1; 1996 : -1;
1619 1997
1620 json_stash = gv_stashpv ("JSON::XS" , 1); 1998 json_stash = gv_stashpv ("JSON::XS" , 1);
1621 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1999 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1622
1623 json_true = get_bool ("JSON::XS::true"); 2000 bool_true = get_bool ("Types::Serialiser::true");
1624 json_false = get_bool ("JSON::XS::false"); 2001 bool_false = get_bool ("Types::Serialiser::false");
2002
2003 sv_json = newSVpv ("JSON", 0);
2004 SvREADONLY_on (sv_json);
2005
2006 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1625} 2007}
1626 2008
1627PROTOTYPES: DISABLE 2009PROTOTYPES: DISABLE
1628 2010
1629void CLONE (...) 2011void CLONE (...)
1630 CODE: 2012 CODE:
1631 json_stash = 0; 2013 json_stash = 0;
1632 json_boolean_stash = 0; 2014 bool_stash = 0;
1633 2015
1634void new (char *klass) 2016void new (char *klass)
1635 PPCODE: 2017 PPCODE:
1636{ 2018{
1637 SV *pv = NEWSV (0, sizeof (JSON)); 2019 SV *pv = NEWSV (0, sizeof (JSON));
1638 SvPOK_only (pv); 2020 SvPOK_only (pv);
1639 Zero (SvPVX (pv), 1, JSON); 2021 json_init ((JSON *)SvPVX (pv));
1640 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1641 XPUSHs (sv_2mortal (sv_bless ( 2022 XPUSHs (sv_2mortal (sv_bless (
1642 newRV_noinc (pv), 2023 newRV_noinc (pv),
1643 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2024 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1644 ))); 2025 )));
1645} 2026}
1657 allow_nonref = F_ALLOW_NONREF 2038 allow_nonref = F_ALLOW_NONREF
1658 shrink = F_SHRINK 2039 shrink = F_SHRINK
1659 allow_blessed = F_ALLOW_BLESSED 2040 allow_blessed = F_ALLOW_BLESSED
1660 convert_blessed = F_CONV_BLESSED 2041 convert_blessed = F_CONV_BLESSED
1661 relaxed = F_RELAXED 2042 relaxed = F_RELAXED
2043 allow_unknown = F_ALLOW_UNKNOWN
2044 allow_tags = F_ALLOW_TAGS
1662 PPCODE: 2045 PPCODE:
1663{ 2046{
1664 if (enable) 2047 if (enable)
1665 self->flags |= ix; 2048 self->flags |= ix;
1666 else 2049 else
1681 get_allow_nonref = F_ALLOW_NONREF 2064 get_allow_nonref = F_ALLOW_NONREF
1682 get_shrink = F_SHRINK 2065 get_shrink = F_SHRINK
1683 get_allow_blessed = F_ALLOW_BLESSED 2066 get_allow_blessed = F_ALLOW_BLESSED
1684 get_convert_blessed = F_CONV_BLESSED 2067 get_convert_blessed = F_CONV_BLESSED
1685 get_relaxed = F_RELAXED 2068 get_relaxed = F_RELAXED
2069 get_allow_unknown = F_ALLOW_UNKNOWN
2070 get_allow_tags = F_ALLOW_TAGS
1686 PPCODE: 2071 PPCODE:
1687 XPUSHs (boolSV (self->flags & ix)); 2072 XPUSHs (boolSV (self->flags & ix));
1688 2073
1689void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2074void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1690 PPCODE: 2075 PPCODE:
1691{ 2076 self->max_depth = max_depth;
1692 UV log2 = 0;
1693
1694 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1695
1696 while ((1UL << log2) < max_depth)
1697 ++log2;
1698
1699 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1700
1701 XPUSHs (ST (0)); 2077 XPUSHs (ST (0));
1702}
1703 2078
1704U32 get_max_depth (JSON *self) 2079U32 get_max_depth (JSON *self)
1705 CODE: 2080 CODE:
1706 RETVAL = DEC_DEPTH (self->flags); 2081 RETVAL = self->max_depth;
1707 OUTPUT: 2082 OUTPUT:
1708 RETVAL 2083 RETVAL
1709 2084
1710void max_size (JSON *self, UV max_size = 0) 2085void max_size (JSON *self, U32 max_size = 0)
1711 PPCODE: 2086 PPCODE:
1712{ 2087 self->max_size = max_size;
1713 UV log2 = 0;
1714
1715 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1716 if (max_size == 1) max_size = 2;
1717
1718 while ((1UL << log2) < max_size)
1719 ++log2;
1720
1721 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1722
1723 XPUSHs (ST (0)); 2088 XPUSHs (ST (0));
1724}
1725 2089
1726int get_max_size (JSON *self) 2090int get_max_size (JSON *self)
1727 CODE: 2091 CODE:
1728 RETVAL = DEC_SIZE (self->flags); 2092 RETVAL = self->max_size;
1729 OUTPUT: 2093 OUTPUT:
1730 RETVAL 2094 RETVAL
1731 2095
1732void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2096void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1733 PPCODE: 2097 PPCODE:
1739} 2103}
1740 2104
1741void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2105void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1742 PPCODE: 2106 PPCODE:
1743{ 2107{
1744 if (!self->cb_sk_object) 2108 if (!self->cb_sk_object)
1745 self->cb_sk_object = newHV (); 2109 self->cb_sk_object = newHV ();
1746 2110
1747 if (SvOK (cb)) 2111 if (SvOK (cb))
1748 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2112 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1749 else 2113 else
1760 XPUSHs (ST (0)); 2124 XPUSHs (ST (0));
1761} 2125}
1762 2126
1763void encode (JSON *self, SV *scalar) 2127void encode (JSON *self, SV *scalar)
1764 PPCODE: 2128 PPCODE:
1765 XPUSHs (encode_json (scalar, self)); 2129 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2130 XPUSHs (scalar);
1766 2131
1767void decode (JSON *self, SV *jsonstr) 2132void decode (JSON *self, SV *jsonstr)
1768 PPCODE: 2133 PPCODE:
1769 XPUSHs (decode_json (jsonstr, self, 0)); 2134 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2135 XPUSHs (jsonstr);
1770 2136
1771void decode_prefix (JSON *self, SV *jsonstr) 2137void decode_prefix (JSON *self, SV *jsonstr)
1772 PPCODE: 2138 PPCODE:
1773{ 2139{
2140 SV *sv;
1774 STRLEN offset; 2141 STRLEN offset;
2142 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1775 EXTEND (SP, 2); 2143 EXTEND (SP, 2);
1776 PUSHs (decode_json (jsonstr, self, &offset)); 2144 PUSHs (sv);
1777 PUSHs (sv_2mortal (newSVuv (offset))); 2145 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1778} 2146}
1779 2147
1780void incr_parse (JSON *self, SV *jsonstr = 0) 2148void incr_parse (JSON *self, SV *jsonstr = 0)
1781 PPCODE: 2149 PPCODE:
1782{ 2150{
1783 if (!self->incr_text) 2151 if (!self->incr_text)
1784 self->incr_text = newSVpvn ("", 0); 2152 self->incr_text = newSVpvn ("", 0);
2153
2154 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2155 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2156 if (DECODE_WANTS_OCTETS (self))
2157 {
2158 if (self->incr_pos)
2159 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2160 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2161
2162 sv_utf8_downgrade (self->incr_text, 0);
2163 }
2164 else
2165 {
2166 sv_utf8_upgrade (self->incr_text);
2167
2168 if (self->incr_pos)
2169 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2170 - (U8 *)SvPVX (self->incr_text);
2171 }
1785 2172
1786 // append data, if any 2173 // append data, if any
1787 if (jsonstr) 2174 if (jsonstr)
1788 { 2175 {
2176 /* make sure both strings have same encoding */
1789 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2177 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2178 if (SvUTF8 (jsonstr))
2179 sv_utf8_downgrade (jsonstr, 0);
1790 { 2180 else
1791 /* utf-8-ness differs, need to upgrade */
1792 sv_utf8_upgrade (self->incr_text); 2181 sv_utf8_upgrade (jsonstr);
1793 2182
1794 if (self->incr_pos) 2183 /* and then just blindly append */
1795 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1796 - (U8 *)SvPVX (self->incr_text);
1797 }
1798
1799 { 2184 {
1800 STRLEN len; 2185 STRLEN len;
1801 const char *str = SvPV (jsonstr, len); 2186 const char *str = SvPV (jsonstr, len);
1802 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2187 STRLEN cur = SvCUR (self->incr_text);
2188
2189 if (SvLEN (self->incr_text) - cur <= len)
2190 json_sv_grow (self->incr_text, cur, len);
2191
1803 Move (str, SvEND (self->incr_text), len, char); 2192 Move (str, SvEND (self->incr_text), len, char);
1804 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2193 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1805 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2194 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1806 } 2195 }
1807 } 2196 }
1808 2197
1809 if (GIMME_V != G_VOID) 2198 if (GIMME_V != G_VOID)
1810 do 2199 do
1811 { 2200 {
2201 SV *sv;
1812 STRLEN offset; 2202 STRLEN offset;
1813 2203
1814 if (!INCR_DONE (self)) 2204 if (!INCR_DONE (self))
1815 { 2205 {
1816 incr_parse (self); 2206 incr_parse (self);
2207
2208 if (self->incr_pos > self->max_size && self->max_size)
2209 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2210 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2211
1817 if (!INCR_DONE (self)) 2212 if (!INCR_DONE (self))
2213 {
2214 // as an optimisation, do not accumulate white space in the incr buffer
2215 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2216 {
2217 self->incr_pos = 0;
2218 SvCUR_set (self->incr_text, 0);
2219 }
2220
1818 break; 2221 break;
2222 }
1819 } 2223 }
1820 2224
1821 XPUSHs (decode_json (self->incr_text, self, &offset)); 2225 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2226 XPUSHs (sv);
1822 2227
1823 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1824 self->incr_pos -= offset; 2228 self->incr_pos -= offset;
1825 self->incr_nest = 0; 2229 self->incr_nest = 0;
1826 self->incr_mode = 0; 2230 self->incr_mode = 0;
2231
2232 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1827 } 2233 }
1828 while (GIMME_V == G_ARRAY); 2234 while (GIMME_V == G_ARRAY);
1829} 2235}
1830 2236
1831SV *incr_text (JSON *self) 2237SV *incr_text (JSON *self)
1850 self->incr_nest = 0; 2256 self->incr_nest = 0;
1851 self->incr_mode = 0; 2257 self->incr_mode = 0;
1852 } 2258 }
1853} 2259}
1854 2260
2261void incr_reset (JSON *self)
2262 CODE:
2263{
2264 SvREFCNT_dec (self->incr_text);
2265 self->incr_text = 0;
2266 self->incr_pos = 0;
2267 self->incr_nest = 0;
2268 self->incr_mode = 0;
2269}
2270
1855void DESTROY (JSON *self) 2271void DESTROY (JSON *self)
1856 CODE: 2272 CODE:
1857 SvREFCNT_dec (self->cb_sk_object); 2273 SvREFCNT_dec (self->cb_sk_object);
1858 SvREFCNT_dec (self->cb_object); 2274 SvREFCNT_dec (self->cb_object);
1859 SvREFCNT_dec (self->incr_text); 2275 SvREFCNT_dec (self->incr_text);
1860 2276
1861PROTOTYPES: ENABLE 2277PROTOTYPES: ENABLE
1862 2278
1863void encode_json (SV *scalar) 2279void encode_json (SV *scalar)
1864 ALIAS:
1865 to_json_ = 0
1866 encode_json = F_UTF8
1867 PPCODE: 2280 PPCODE:
1868{ 2281{
1869 JSON json = { F_DEFAULT | ix }; 2282 JSON json;
1870 XPUSHs (encode_json (scalar, &json)); 2283 json_init (&json);
2284 json.flags |= F_UTF8;
2285 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2286 XPUSHs (scalar);
1871} 2287}
1872 2288
1873void decode_json (SV *jsonstr) 2289void decode_json (SV *jsonstr)
1874 ALIAS:
1875 from_json_ = 0
1876 decode_json = F_UTF8
1877 PPCODE: 2290 PPCODE:
1878{ 2291{
1879 JSON json = { F_DEFAULT | ix }; 2292 JSON json;
1880 XPUSHs (decode_json (jsonstr, &json, 0)); 2293 json_init (&json);
2294 json.flags |= F_UTF8;
2295 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2296 XPUSHs (jsonstr);
1881} 2297}
1882 2298
1883

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines