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 1.135 by root, Thu Nov 15 20:49:12 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
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 64 // 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
74# 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))
75#else 86#else
76# define JSON_SLOW 0
77# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
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_false, *bool_true;
97static SV *sv_json;
82 98
83enum { 99enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0 100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string 101 INCR_M_STR, // inside string
86 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
87 INCR_M_JSON // outside anything, count nesting 105 INCR_M_JSON // outside anything, count nesting
88}; 106};
89 107
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 108#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
91 109
92typedef struct { 110typedef struct {
93 U32 flags; 111 U32 flags;
112 U32 max_depth;
113 STRLEN max_size;
114
94 SV *cb_object; 115 SV *cb_object;
95 HV *cb_sk_object; 116 HV *cb_sk_object;
96 117
97 // for the incremental parser 118 // for the incremental parser
98 SV *incr_text; // the source text so far 119 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 120 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 121 int incr_nest; // {[]}-nesting level
101 int incr_mode; 122 unsigned char incr_mode;
123
124 SV *v_false, *v_true;
102} JSON; 125} JSON;
126
127INLINE void
128json_init (JSON *json)
129{
130 static const JSON init = { F_ALLOW_NONREF, 512 };
131
132 *json = init;
133}
103 134
104///////////////////////////////////////////////////////////////////////////// 135/////////////////////////////////////////////////////////////////////////////
105// utility functions 136// utility functions
106 137
107INLINE SV * 138INLINE SV *
117 148
118INLINE void 149INLINE void
119shrink (SV *sv) 150shrink (SV *sv)
120{ 151{
121 sv_utf8_downgrade (sv, 1); 152 sv_utf8_downgrade (sv, 1);
153
122 if (SvLEN (sv) > SvCUR (sv) + 1) 154 if (SvLEN (sv) > SvCUR (sv) + 1)
123 { 155 {
124#ifdef SvPV_shrink_to_cur 156#ifdef SvPV_shrink_to_cur
125 SvPV_shrink_to_cur (sv); 157 SvPV_shrink_to_cur (sv);
126#elif defined (SvPV_renew) 158#elif defined (SvPV_renew)
127 SvPV_renew (sv, SvCUR (sv) + 1); 159 SvPV_renew (sv, SvCUR (sv) + 1);
128#endif 160#endif
129 } 161 }
130} 162}
131 163
164/* adds two STRLENs together, slow, and with paranoia */
165STRLEN
166strlen_sum (STRLEN l1, STRLEN l2)
167{
168 size_t sum = l1 + l2;
169
170 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
171 croak ("JSON::XS: string size overflow");
172
173 return sum;
174}
175
176/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
177static char *
178json_sv_grow (SV *sv, size_t len1, size_t len2)
179{
180 len1 = strlen_sum (len1, len2);
181 len1 = strlen_sum (len1, len1 >> 1);
182
183 if (len1 > 4096 - 24)
184 len1 = (len1 | 4095) - 24;
185
186 return SvGROW (sv, len1);
187}
188
132// decode an utf-8 character and return it, or (UV)-1 in 189// decode a utf-8 character and return it, or (UV)-1 in
133// case of an error. 190// case of an error.
134// we special-case "safe" characters from U+80 .. U+7FF, 191// we special-case "safe" characters from U+80 .. U+7FF,
135// but use the very good perl function to parse anything else. 192// but use the very good perl function to parse anything else.
136// note that we never call this function for a ascii codepoints 193// note that we never call this function for a ascii codepoints
137INLINE UV 194INLINE UV
170 *s++ = 0x80 | ( ch & 0x3f); 227 *s++ = 0x80 | ( ch & 0x3f);
171 228
172 return s; 229 return s;
173} 230}
174 231
232// convert offset pointer to character index, sv must be string
233static STRLEN
234ptr_to_index (SV *sv, char *offset)
235{
236 return SvUTF8 (sv)
237 ? utf8_distance (offset, SvPVX (sv))
238 : offset - SvPVX (sv);
239}
240
241/////////////////////////////////////////////////////////////////////////////
242// fp hell
243
244// scan a group of digits, and a trailing exponent
245static void
246json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
247{
248 UV uaccum = 0;
249 int eaccum = 0;
250
251 // if we recurse too deep, skip all remaining digits
252 // to avoid a stack overflow attack
253 if (expect_false (--maxdepth <= 0))
254 while (((U8)*s - '0') < 10)
255 ++s;
256
257 for (;;)
258 {
259 U8 dig = (U8)*s - '0';
260
261 if (expect_false (dig >= 10))
262 {
263 if (dig == (U8)((U8)'.' - (U8)'0'))
264 {
265 ++s;
266 json_atof_scan1 (s, accum, expo, 1, maxdepth);
267 }
268 else if ((dig | ' ') == 'e' - '0')
269 {
270 int exp2 = 0;
271 int neg = 0;
272
273 ++s;
274
275 if (*s == '-')
276 {
277 ++s;
278 neg = 1;
279 }
280 else if (*s == '+')
281 ++s;
282
283 while ((dig = (U8)*s - '0') < 10)
284 exp2 = exp2 * 10 + *s++ - '0';
285
286 *expo += neg ? -exp2 : exp2;
287 }
288
289 break;
290 }
291
292 ++s;
293
294 uaccum = uaccum * 10 + dig;
295 ++eaccum;
296
297 // if we have too many digits, then recurse for more
298 // we actually do this for rather few digits
299 if (uaccum >= (UV_MAX - 9) / 10)
300 {
301 if (postdp) *expo -= eaccum;
302 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
303 if (postdp) *expo += eaccum;
304
305 break;
306 }
307 }
308
309 // this relies greatly on the quality of the pow ()
310 // implementation of the platform, but a good
311 // implementation is hard to beat.
312 // (IEEE 754 conformant ones are required to be exact)
313 if (postdp) *expo -= eaccum;
314 *accum += uaccum * Perl_pow (10., *expo);
315 *expo += eaccum;
316}
317
318static NV
319json_atof (const char *s)
320{
321 NV accum = 0.;
322 int expo = 0;
323 int neg = 0;
324
325 if (*s == '-')
326 {
327 ++s;
328 neg = 1;
329 }
330
331 // a recursion depth of ten gives us >>500 bits
332 json_atof_scan1 (s, &accum, &expo, 0, 10);
333
334 return neg ? -accum : accum;
335}
336
337// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
338static int
339ref_bool_type (SV *sv)
340{
341 svtype svt = SvTYPE (sv);
342
343 if (svt < SVt_PVAV)
344 {
345 STRLEN len = 0;
346 char *pv = svt ? SvPV (sv, len) : 0;
347
348 if (len == 1)
349 if (*pv == '1')
350 return 1;
351 else if (*pv == '0')
352 return 0;
353 }
354
355 return -1;
356}
357
358// returns whether scalar is not a reference in the sense of allow_nonref
359static int
360json_nonref (SV *scalar)
361{
362 if (!SvROK (scalar))
363 return 1;
364
365 scalar = SvRV (scalar);
366
367 if (SvTYPE (scalar) >= SVt_PVMG)
368 {
369 if (SvSTASH (scalar) == bool_stash)
370 return 1;
371
372 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
373 return 1;
374 }
375
376 return 0;
377}
378
175///////////////////////////////////////////////////////////////////////////// 379/////////////////////////////////////////////////////////////////////////////
176// encoder 380// encoder
177 381
178// structure used for encoding JSON 382// structure used for encoding JSON
179typedef struct 383typedef struct
181 char *cur; // SvPVX (sv) + current output position 385 char *cur; // SvPVX (sv) + current output position
182 char *end; // SvEND (sv) 386 char *end; // SvEND (sv)
183 SV *sv; // result scalar 387 SV *sv; // result scalar
184 JSON json; 388 JSON json;
185 U32 indent; // indentation level 389 U32 indent; // indentation level
186 U32 maxdepth; // max. indentation/recursion level
187 UV limit; // escape character values >= this value when encoding 390 UV limit; // escape character values >= this value when encoding
188} enc_t; 391} enc_t;
189 392
190INLINE void 393INLINE void
191need (enc_t *enc, STRLEN len) 394need (enc_t *enc, STRLEN len)
192{ 395{
193 if (expect_false (enc->cur + len >= enc->end)) 396 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
194 { 397 {
195 STRLEN cur = enc->cur - SvPVX (enc->sv); 398 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
196 SvGROW (enc->sv, cur + len + 1); 399 char *buf = json_sv_grow (enc->sv, cur, len);
197 enc->cur = SvPVX (enc->sv) + cur; 400 enc->cur = buf + cur;
198 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 401 enc->end = buf + SvLEN (enc->sv) - 1;
199 } 402 }
200} 403}
201 404
202INLINE void 405INLINE void
203encode_ch (enc_t *enc, char ch) 406encode_ch (enc_t *enc, char ch)
219 422
220 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 423 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
221 { 424 {
222 if (expect_false (ch == '"')) // but with slow exceptions 425 if (expect_false (ch == '"')) // but with slow exceptions
223 { 426 {
224 need (enc, len += 1); 427 need (enc, len + 1);
225 *enc->cur++ = '\\'; 428 *enc->cur++ = '\\';
226 *enc->cur++ = '"'; 429 *enc->cur++ = '"';
227 } 430 }
228 else if (expect_false (ch == '\\')) 431 else if (expect_false (ch == '\\'))
229 { 432 {
230 need (enc, len += 1); 433 need (enc, len + 1);
231 *enc->cur++ = '\\'; 434 *enc->cur++ = '\\';
232 *enc->cur++ = '\\'; 435 *enc->cur++ = '\\';
233 } 436 }
234 else 437 else
235 *enc->cur++ = ch; 438 *enc->cur++ = ch;
238 } 441 }
239 else 442 else
240 { 443 {
241 switch (ch) 444 switch (ch)
242 { 445 {
243 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 446 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; 447 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; 448 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; 449 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; 450 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
248 451
249 default: 452 default:
250 { 453 {
251 STRLEN clen; 454 STRLEN clen;
252 UV uch; 455 UV uch;
268 if (uch >= 0x10000UL) 471 if (uch >= 0x10000UL)
269 { 472 {
270 if (uch >= 0x110000UL) 473 if (uch >= 0x110000UL)
271 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 474 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
272 475
273 need (enc, len += 11); 476 need (enc, len + 11);
274 sprintf (enc->cur, "\\u%04x\\u%04x", 477 sprintf (enc->cur, "\\u%04x\\u%04x",
275 (int)((uch - 0x10000) / 0x400 + 0xD800), 478 (int)((uch - 0x10000) / 0x400 + 0xD800),
276 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 479 (int)((uch - 0x10000) % 0x400 + 0xDC00));
277 enc->cur += 12; 480 enc->cur += 12;
278 } 481 }
279 else 482 else
280 { 483 {
281 static char hexdigit [16] = "0123456789abcdef";
282 need (enc, len += 5); 484 need (enc, len + 5);
283 *enc->cur++ = '\\'; 485 *enc->cur++ = '\\';
284 *enc->cur++ = 'u'; 486 *enc->cur++ = 'u';
285 *enc->cur++ = hexdigit [ uch >> 12 ]; 487 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
286 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 488 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
287 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 489 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
288 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
289 } 491 }
290 492
291 str += clen; 493 str += clen;
292 } 494 }
293 else if (enc->json.flags & F_LATIN1) 495 else if (enc->json.flags & F_LATIN1)
295 *enc->cur++ = uch; 497 *enc->cur++ = uch;
296 str += clen; 498 str += clen;
297 } 499 }
298 else if (is_utf8) 500 else if (is_utf8)
299 { 501 {
300 need (enc, len += clen); 502 need (enc, len + clen);
301 do 503 do
302 { 504 {
303 *enc->cur++ = *str++; 505 *enc->cur++ = *str++;
304 } 506 }
305 while (--clen); 507 while (--clen);
306 } 508 }
307 else 509 else
308 { 510 {
309 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 511 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
310 enc->cur = encode_utf8 (enc->cur, uch); 512 enc->cur = encode_utf8 (enc->cur, uch);
311 ++str; 513 ++str;
312 } 514 }
313 } 515 }
314 } 516 }
364static void 566static void
365encode_av (enc_t *enc, AV *av) 567encode_av (enc_t *enc, AV *av)
366{ 568{
367 int i, len = av_len (av); 569 int i, len = av_len (av);
368 570
369 if (enc->indent >= enc->maxdepth) 571 if (enc->indent >= enc->json.max_depth)
370 croak ("data structure too deep (hit recursion limit)"); 572 croak (ERR_NESTING_EXCEEDED);
371 573
372 encode_ch (enc, '['); 574 encode_ch (enc, '[');
373 575
374 if (len >= 0) 576 if (len >= 0)
375 { 577 {
376 encode_nl (enc); ++enc->indent; 578 encode_nl (enc); ++enc->indent;
377 579
378 for (i = 0; i <= len; ++i) 580 for (i = 0; i <= len; ++i)
390 encode_comma (enc); 592 encode_comma (enc);
391 } 593 }
392 594
393 encode_nl (enc); --enc->indent; encode_indent (enc); 595 encode_nl (enc); --enc->indent; encode_indent (enc);
394 } 596 }
395 597
396 encode_ch (enc, ']'); 598 encode_ch (enc, ']');
397} 599}
398 600
399static void 601static void
400encode_hk (enc_t *enc, HE *he) 602encode_hk (enc_t *enc, HE *he)
404 if (HeKLEN (he) == HEf_SVKEY) 606 if (HeKLEN (he) == HEf_SVKEY)
405 { 607 {
406 SV *sv = HeSVKEY (he); 608 SV *sv = HeSVKEY (he);
407 STRLEN len; 609 STRLEN len;
408 char *str; 610 char *str;
409 611
410 SvGETMAGIC (sv); 612 SvGETMAGIC (sv);
411 str = SvPV (sv, len); 613 str = SvPV (sv, len);
412 614
413 encode_str (enc, str, len, SvUTF8 (sv)); 615 encode_str (enc, str, len, SvUTF8 (sv));
414 } 616 }
450static void 652static void
451encode_hv (enc_t *enc, HV *hv) 653encode_hv (enc_t *enc, HV *hv)
452{ 654{
453 HE *he; 655 HE *he;
454 656
455 if (enc->indent >= enc->maxdepth) 657 if (enc->indent >= enc->json.max_depth)
456 croak ("data structure too deep (hit recursion limit)"); 658 croak (ERR_NESTING_EXCEEDED);
457 659
458 encode_ch (enc, '{'); 660 encode_ch (enc, '{');
459 661
460 // for canonical output we have to sort by keys first 662 // for canonical output we have to sort by keys first
461 // actually, this is mostly due to the stupid so-called 663 // actually, this is mostly due to the stupid so-called
462 // security workaround added somewhere in 5.8.x. 664 // security workaround added somewhere in 5.8.x
463 // that randomises hash orderings 665 // that randomises hash orderings
464 if (enc->json.flags & F_CANONICAL) 666 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
465 { 667 {
466 int count = hv_iterinit (hv); 668 int count = hv_iterinit (hv);
467 669
468 if (SvMAGICAL (hv)) 670 if (SvMAGICAL (hv))
469 { 671 {
480 } 682 }
481 683
482 if (count) 684 if (count)
483 { 685 {
484 int i, fast = 1; 686 int i, fast = 1;
485#if defined(__BORLANDC__) || defined(_MSC_VER) 687 HE *hes_stack [STACK_HES];
486 HE **hes = _alloca (count * sizeof (HE)); 688 HE **hes = hes_stack;
487#else 689
488 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 690 // allocate larger arrays on the heap
489#endif 691 if (count > STACK_HES)
692 {
693 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
694 hes = (HE **)SvPVX (sv);
695 }
490 696
491 i = 0; 697 i = 0;
492 while ((he = hv_iternext (hv))) 698 while ((he = hv_iternext (hv)))
493 { 699 {
494 hes [i++] = he; 700 hes [i++] = he;
563// encode objects, arrays and special \0=false and \1=true values. 769// encode objects, arrays and special \0=false and \1=true values.
564static void 770static void
565encode_rv (enc_t *enc, SV *sv) 771encode_rv (enc_t *enc, SV *sv)
566{ 772{
567 svtype svt; 773 svtype svt;
774 GV *method;
568 775
569 SvGETMAGIC (sv); 776 SvGETMAGIC (sv);
570 svt = SvTYPE (sv); 777 svt = SvTYPE (sv);
571 778
572 if (expect_false (SvOBJECT (sv))) 779 if (expect_false (SvOBJECT (sv)))
573 { 780 {
574 HV *stash = !JSON_SLOW || json_boolean_stash 781 HV *stash = SvSTASH (sv);
575 ? json_boolean_stash
576 : gv_stashpv ("JSON::XS::Boolean", 1);
577 782
578 if (SvSTASH (sv) == stash) 783 if (stash == bool_stash)
579 { 784 {
580 if (SvIV (sv))
581 encode_str (enc, "true", 4, 0); 785 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
582 else
583 encode_str (enc, "false", 5, 0); 786 else encode_str (enc, "false", 5, 0);
584 } 787 }
788 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
789 {
790 int count;
791 dSP;
792
793 ENTER; SAVETMPS;
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 if (count)
816 {
817 int i;
818
819 for (i = 0; i < count - 1; ++i)
820 {
821 encode_sv (enc, SP[i + 1 - count]);
822 encode_ch (enc, ',');
823 }
824
825 encode_sv (enc, TOPs);
826 SP -= count;
827 }
828
829 encode_ch (enc, ']');
830
831 FREETMPS; LEAVE;
832 }
833 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
834 {
835 dSP;
836
837 ENTER; SAVETMPS;
838 PUSHMARK (SP);
839 // we re-bless the reference to get overload and other niceties right
840 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
841
842 // calling with G_SCALAR ensures that we always get a 1 return value
843 PUTBACK;
844 call_sv ((SV *)GvCV (method), G_SCALAR);
845 SPAGAIN;
846
847 // catch this surprisingly common error
848 if (SvROK (TOPs) && SvRV (TOPs) == sv)
849 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
850
851 sv = POPs;
852 PUTBACK;
853
854 encode_sv (enc, sv);
855
856 FREETMPS; LEAVE;
857 }
858 else if (enc->json.flags & F_ALLOW_BLESSED)
859 encode_str (enc, "null", 4, 0);
585 else 860 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", 861 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)))); 862 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
632 }
633 } 863 }
634 else if (svt == SVt_PVHV) 864 else if (svt == SVt_PVHV)
635 encode_hv (enc, (HV *)sv); 865 encode_hv (enc, (HV *)sv);
636 else if (svt == SVt_PVAV) 866 else if (svt == SVt_PVAV)
637 encode_av (enc, (AV *)sv); 867 encode_av (enc, (AV *)sv);
638 else if (svt < SVt_PVAV) 868 else if (svt < SVt_PVAV)
639 { 869 {
640 STRLEN len = 0; 870 int bool_type = ref_bool_type (sv);
641 char *pv = svt ? SvPV (sv, len) : 0;
642 871
643 if (len == 1 && *pv == '1') 872 if (bool_type == 1)
644 encode_str (enc, "true", 4, 0); 873 encode_str (enc, "true", 4, 0);
645 else if (len == 1 && *pv == '0') 874 else if (bool_type == 0)
646 encode_str (enc, "false", 5, 0); 875 encode_str (enc, "false", 5, 0);
876 else if (enc->json.flags & F_ALLOW_UNKNOWN)
877 encode_str (enc, "null", 4, 0);
647 else 878 else
648 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 879 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
649 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 880 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
650 } 881 }
882 else if (enc->json.flags & F_ALLOW_UNKNOWN)
883 encode_str (enc, "null", 4, 0);
651 else 884 else
652 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 885 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
653 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 886 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
654} 887}
655 888
709 } 942 }
710 else 943 else
711 { 944 {
712 // large integer, use the (rather slow) snprintf way. 945 // large integer, use the (rather slow) snprintf way.
713 need (enc, IVUV_MAXCHARS); 946 need (enc, IVUV_MAXCHARS);
714 enc->cur += 947 enc->cur +=
715 SvIsUV(sv) 948 SvIsUV(sv)
716 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 949 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
717 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 950 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
718 } 951 }
719 } 952 }
720 else if (SvROK (sv)) 953 else if (SvROK (sv))
721 encode_rv (enc, SvRV (sv)); 954 encode_rv (enc, SvRV (sv));
722 else if (!SvOK (sv)) 955 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
723 encode_str (enc, "null", 4, 0); 956 encode_str (enc, "null", 4, 0);
724 else 957 else
725 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 958 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
726 SvPV_nolen (sv), SvFLAGS (sv)); 959 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
727} 960}
728 961
729static SV * 962static SV *
730encode_json (SV *scalar, JSON *json) 963encode_json (SV *scalar, JSON *json)
731{ 964{
732 enc_t enc; 965 enc_t enc;
733 966
734 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 967 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)"); 968 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
736 969
737 enc.json = *json; 970 enc.json = *json;
738 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 971 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
739 enc.cur = SvPVX (enc.sv); 972 enc.cur = SvPVX (enc.sv);
740 enc.end = SvEND (enc.sv); 973 enc.end = SvEND (enc.sv);
741 enc.indent = 0; 974 enc.indent = 0;
742 enc.maxdepth = DEC_DEPTH (enc.json.flags);
743 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 975 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
744 : enc.json.flags & F_LATIN1 ? 0x000100UL 976 : enc.json.flags & F_LATIN1 ? 0x000100UL
745 : 0x110000UL; 977 : 0x110000UL;
746 978
747 SvPOK_only (enc.sv); 979 SvPOK_only (enc.sv);
748 encode_sv (&enc, scalar); 980 encode_sv (&enc, scalar);
981 encode_nl (&enc);
749 982
750 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 983 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
751 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 984 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
752 985
753 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 986 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
814 if (*dec->cur != ch) \ 1047 if (*dec->cur != ch) \
815 ERR (# ch " expected"); \ 1048 ERR (# ch " expected"); \
816 ++dec->cur; \ 1049 ++dec->cur; \
817 SE 1050 SE
818 1051
819#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1052#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
820#define DEC_DEC_DEPTH --dec->depth 1053#define DEC_DEC_DEPTH --dec->depth
821 1054
822static SV *decode_sv (dec_t *dec); 1055static SV *decode_sv (dec_t *dec);
823 1056
824static signed char decode_hexdigit[256]; 1057static signed char decode_hexdigit[256];
933 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1166 else if (expect_true (ch >= 0x20 && ch < 0x80))
934 *cur++ = ch; 1167 *cur++ = ch;
935 else if (ch >= 0x80) 1168 else if (ch >= 0x80)
936 { 1169 {
937 STRLEN clen; 1170 STRLEN clen;
938 UV uch;
939 1171
940 --dec_cur; 1172 --dec_cur;
941 1173
942 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1174 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
943 if (clen == (STRLEN)-1) 1175 if (clen == (STRLEN)-1)
944 ERR ("malformed UTF-8 character in JSON string"); 1176 ERR ("malformed UTF-8 character in JSON string");
945 1177
946 do 1178 do
947 *cur++ = *dec_cur++; 1179 *cur++ = *dec_cur++;
948 while (--clen); 1180 while (--clen);
949 1181
950 utf8 = 1; 1182 utf8 = 1;
951 } 1183 }
1184 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1185 *cur++ = ch;
952 else 1186 else
953 { 1187 {
954 --dec_cur; 1188 --dec_cur;
955 1189
956 if (!ch) 1190 if (!ch)
964 { 1198 {
965 STRLEN len = cur - buf; 1199 STRLEN len = cur - buf;
966 1200
967 if (sv) 1201 if (sv)
968 { 1202 {
969 SvGROW (sv, SvCUR (sv) + len + 1); 1203 STRLEN cur = SvCUR (sv);
1204
1205 if (SvLEN (sv) - cur <= len)
1206 json_sv_grow (sv, cur, len);
1207
970 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1208 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
971 SvCUR_set (sv, SvCUR (sv) + len); 1209 SvCUR_set (sv, SvCUR (sv) + len);
972 } 1210 }
973 else 1211 else
974 sv = newSVpvn (buf, len); 1212 sv = newSVpvn (buf, len);
1065 1303
1066 // special case the rather common 1..5-digit-int case 1304 // special case the rather common 1..5-digit-int case
1067 if (*start == '-') 1305 if (*start == '-')
1068 switch (len) 1306 switch (len)
1069 { 1307 {
1070 case 2: return newSViv (-( start [1] - '0' * 1)); 1308 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1071 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1309 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)); 1310 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)); 1311 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)); 1312 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1075 } 1313 }
1076 else 1314 else
1077 switch (len) 1315 switch (len)
1078 { 1316 {
1079 case 1: return newSViv ( start [0] - '0' * 1); 1317 case 1: return newSViv ( start [0] - '0' * 1);
1080 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1318 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); 1319 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); 1320 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); 1321 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1084 } 1322 }
1085 1323
1086 { 1324 {
1087 UV uv; 1325 UV uv;
1088 int numtype = grok_number (start, len, &uv); 1326 int numtype = grok_number (start, len, &uv);
1097 } 1335 }
1098 1336
1099 len -= *start == '-' ? 1 : 0; 1337 len -= *start == '-' ? 1 : 0;
1100 1338
1101 // does not fit into IV or UV, try NV 1339 // does not fit into IV or UV, try NV
1102 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1340 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 1341 // fits into NV without loss of precision
1108 return newSVnv (Atof (start)); 1342 return newSVnv (json_atof (start));
1109 1343
1110 // everything else fails, convert it to a string 1344 // everything else fails, convert it to a string
1111 return newSVpvn (start, dec->cur - start); 1345 return newSVpvn (start, dec->cur - start);
1112 } 1346 }
1113 1347
1114 // loss of precision here 1348 // loss of precision here
1115 return newSVnv (Atof (start)); 1349 return newSVnv (json_atof (start));
1116 1350
1117fail: 1351fail:
1118 return 0; 1352 return 0;
1119} 1353}
1120 1354
1144 if (*dec->cur == ']') 1378 if (*dec->cur == ']')
1145 { 1379 {
1146 ++dec->cur; 1380 ++dec->cur;
1147 break; 1381 break;
1148 } 1382 }
1149 1383
1150 if (*dec->cur != ',') 1384 if (*dec->cur != ',')
1151 ERR (", or ] expected while parsing array"); 1385 ERR (", or ] expected while parsing array");
1152 1386
1153 ++dec->cur; 1387 ++dec->cur;
1154 1388
1278 1512
1279 hv_iterinit (hv); 1513 hv_iterinit (hv);
1280 he = hv_iternext (hv); 1514 he = hv_iternext (hv);
1281 hv_iterinit (hv); 1515 hv_iterinit (hv);
1282 1516
1283 // the next line creates a mortal sv each time its called. 1517 // the next line creates a mortal sv each time it's called.
1284 // might want to optimise this for common cases. 1518 // might want to optimise this for common cases.
1285 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1519 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1286 1520
1287 if (cb) 1521 if (cb)
1288 { 1522 {
1289 dSP; 1523 dSP;
1290 int count; 1524 int count;
1291 1525
1292 ENTER; SAVETMPS; PUSHMARK (SP); 1526 ENTER; SAVETMPS;
1527 PUSHMARK (SP);
1293 XPUSHs (HeVAL (he)); 1528 XPUSHs (HeVAL (he));
1529 sv_2mortal (sv);
1294 1530
1295 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1531 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1296 1532
1297 if (count == 1) 1533 if (count == 1)
1298 { 1534 {
1299 sv = newSVsv (POPs); 1535 sv = newSVsv (POPs);
1300 FREETMPS; LEAVE; 1536 FREETMPS; LEAVE;
1301 return sv; 1537 return sv;
1302 } 1538 }
1539 else if (count)
1540 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1303 1541
1542 SvREFCNT_inc (sv);
1304 FREETMPS; LEAVE; 1543 FREETMPS; LEAVE;
1305 } 1544 }
1306 } 1545 }
1307 1546
1308 if (dec->json.cb_object) 1547 if (dec->json.cb_object)
1309 { 1548 {
1310 dSP; 1549 dSP;
1311 int count; 1550 int count;
1312 1551
1313 ENTER; SAVETMPS; PUSHMARK (SP); 1552 ENTER; SAVETMPS;
1553 PUSHMARK (SP);
1314 XPUSHs (sv_2mortal (sv)); 1554 XPUSHs (sv_2mortal (sv));
1315 1555
1316 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1556 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1317 1557
1318 if (count == 1) 1558 if (count == 1)
1319 { 1559 {
1320 sv = newSVsv (POPs); 1560 sv = newSVsv (POPs);
1321 FREETMPS; LEAVE; 1561 FREETMPS; LEAVE;
1322 return sv; 1562 return sv;
1323 } 1563 }
1564 else if (count)
1565 croak ("filter_json_object callbacks must not return more than one scalar");
1324 1566
1325 SvREFCNT_inc (sv); 1567 SvREFCNT_inc (sv);
1326 FREETMPS; LEAVE; 1568 FREETMPS; LEAVE;
1327 } 1569 }
1328 } 1570 }
1334 DEC_DEC_DEPTH; 1576 DEC_DEC_DEPTH;
1335 return 0; 1577 return 0;
1336} 1578}
1337 1579
1338static SV * 1580static SV *
1581decode_tag (dec_t *dec)
1582{
1583 SV *tag = 0;
1584 SV *val = 0;
1585
1586 if (!(dec->json.flags & F_ALLOW_TAGS))
1587 ERR ("malformed JSON string, neither array, object, number, string or atom");
1588
1589 ++dec->cur;
1590
1591 decode_ws (dec);
1592
1593 tag = decode_sv (dec);
1594 if (!tag)
1595 goto fail;
1596
1597 if (!SvPOK (tag))
1598 ERR ("malformed JSON string, (tag) must be a string");
1599
1600 decode_ws (dec);
1601
1602 if (*dec->cur != ')')
1603 ERR (") expected after tag");
1604
1605 ++dec->cur;
1606
1607 decode_ws (dec);
1608
1609 val = decode_sv (dec);
1610 if (!val)
1611 goto fail;
1612
1613 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1614 ERR ("malformed JSON string, tag value must be an array");
1615
1616 {
1617 AV *av = (AV *)SvRV (val);
1618 int i, len = av_len (av) + 1;
1619 HV *stash = gv_stashsv (tag, 0);
1620 SV *sv;
1621
1622 if (!stash)
1623 ERR ("cannot decode perl-object (package does not exist)");
1624
1625 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1626
1627 if (!method)
1628 ERR ("cannot decode perl-object (package does not have a THAW method)");
1629
1630 dSP;
1631
1632 ENTER; SAVETMPS;
1633 PUSHMARK (SP);
1634 EXTEND (SP, len + 2);
1635 // we re-bless the reference to get overload and other niceties right
1636 PUSHs (tag);
1637 PUSHs (sv_json);
1638
1639 for (i = 0; i < len; ++i)
1640 PUSHs (*av_fetch (av, i, 1));
1641
1642 PUTBACK;
1643 call_sv ((SV *)GvCV (method), G_SCALAR);
1644 SPAGAIN;
1645
1646 SvREFCNT_dec (tag);
1647 SvREFCNT_dec (val);
1648 sv = SvREFCNT_inc (POPs);
1649
1650 PUTBACK;
1651
1652 FREETMPS; LEAVE;
1653
1654 return sv;
1655 }
1656
1657fail:
1658 SvREFCNT_dec (tag);
1659 SvREFCNT_dec (val);
1660 return 0;
1661}
1662
1663static SV *
1339decode_sv (dec_t *dec) 1664decode_sv (dec_t *dec)
1340{ 1665{
1341 // the beauty of JSON: you need exactly one character lookahead 1666 // the beauty of JSON: you need exactly one character lookahead
1342 // to parse everything. 1667 // to parse everything.
1343 switch (*dec->cur) 1668 switch (*dec->cur)
1344 { 1669 {
1345 case '"': ++dec->cur; return decode_str (dec); 1670 case '"': ++dec->cur; return decode_str (dec);
1346 case '[': ++dec->cur; return decode_av (dec); 1671 case '[': ++dec->cur; return decode_av (dec);
1347 case '{': ++dec->cur; return decode_hv (dec); 1672 case '{': ++dec->cur; return decode_hv (dec);
1673 case '(': return decode_tag (dec);
1348 1674
1349 case '-': 1675 case '-':
1350 case '0': case '1': case '2': case '3': case '4': 1676 case '0': case '1': case '2': case '3': case '4':
1351 case '5': case '6': case '7': case '8': case '9': 1677 case '5': case '6': case '7': case '8': case '9':
1352 return decode_num (dec); 1678 return decode_num (dec);
1353 1679
1680 case 'f':
1681 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1682 {
1683 dec->cur += 5;
1684
1685 if (expect_false (!dec->json.v_false))
1686 dec->json.v_false = GET_BOOL (false);
1687
1688 return newSVsv (dec->json.v_false);
1689 }
1690 else
1691 ERR ("'false' expected");
1692
1693 break;
1694
1354 case 't': 1695 case 't':
1355 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1696 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1356 { 1697 {
1357 dec->cur += 4; 1698 dec->cur += 4;
1358#if JSON_SLOW 1699
1359 json_true = get_bool ("JSON::XS::true"); 1700 if (expect_false (!dec->json.v_true))
1360#endif 1701 dec->json.v_true = GET_BOOL (true);
1702
1361 return newSVsv (json_true); 1703 return newSVsv (dec->json.v_true);
1362 } 1704 }
1363 else 1705 else
1364 ERR ("'true' expected"); 1706 ERR ("'true' expected");
1365
1366 break;
1367
1368 case 'f':
1369 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1370 {
1371 dec->cur += 5;
1372#if JSON_SLOW
1373 json_false = get_bool ("JSON::XS::false");
1374#endif
1375 return newSVsv (json_false);
1376 }
1377 else
1378 ERR ("'false' expected");
1379 1707
1380 break; 1708 break;
1381 1709
1382 case 'n': 1710 case 'n':
1383 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1711 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1389 ERR ("'null' expected"); 1717 ERR ("'null' expected");
1390 1718
1391 break; 1719 break;
1392 1720
1393 default: 1721 default:
1394 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1722 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1395 break; 1723 break;
1396 } 1724 }
1397 1725
1398fail: 1726fail:
1399 return 0; 1727 return 0;
1401 1729
1402static SV * 1730static SV *
1403decode_json (SV *string, JSON *json, STRLEN *offset_return) 1731decode_json (SV *string, JSON *json, STRLEN *offset_return)
1404{ 1732{
1405 dec_t dec; 1733 dec_t dec;
1406 STRLEN offset;
1407 SV *sv; 1734 SV *sv;
1408 1735
1736 /* work around bugs in 5.10 where manipulating magic values
1737 * makes perl ignore the magic in subsequent accesses.
1738 * also make a copy of non-PV values, to get them into a clean
1739 * state (SvPV should do that, but it's buggy, see below).
1740 *
1741 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1742 * as reported by Reini Urban.
1743 */
1409 SvGETMAGIC (string); 1744 /*SvGETMAGIC (string);*/
1745 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1746 string = sv_2mortal (newSVsv (string));
1747
1410 SvUPGRADE (string, SVt_PV); 1748 SvUPGRADE (string, SVt_PV);
1411 1749
1412 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1750 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1751 * assertion with -DDEBUGGING, although SvCUR is documented to
1752 * return the xpv_cur field which certainly exists after upgrading.
1753 * according to nicholas clark, calling SvPOK fixes this.
1754 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1755 * and hope for the best.
1756 * Damnit, SvPV_nolen still trips over yet another assertion. This
1757 * assertion business is seriously broken, try yet another workaround
1758 * for the broken -DDEBUGGING.
1759 */
1760 {
1761#ifdef DEBUGGING
1762 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1763#else
1764 STRLEN offset = SvCUR (string);
1765#endif
1766
1767 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", 1768 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)); 1769 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1770 }
1415 1771
1416 if (json->flags & F_UTF8) 1772 if (DECODE_WANTS_OCTETS (json))
1417 sv_utf8_downgrade (string, 0); 1773 sv_utf8_downgrade (string, 0);
1418 else 1774 else
1419 sv_utf8_upgrade (string); 1775 sv_utf8_upgrade (string);
1420 1776
1421 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1777 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1422 1778
1423 dec.json = *json; 1779 dec.json = *json;
1424 dec.cur = SvPVX (string); 1780 dec.cur = SvPVX (string);
1425 dec.end = SvEND (string); 1781 dec.end = SvEND (string);
1426 dec.err = 0; 1782 dec.err = 0;
1427 dec.depth = 0; 1783 dec.depth = 0;
1428 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1429 1784
1430 if (dec.json.cb_object || dec.json.cb_sk_object) 1785 if (dec.json.cb_object || dec.json.cb_sk_object)
1431 dec.json.flags |= F_HOOK; 1786 dec.json.flags |= F_HOOK;
1432 1787
1433 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1788 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1434 1789
1435 decode_ws (&dec); 1790 decode_ws (&dec);
1436 sv = decode_sv (&dec); 1791 sv = decode_sv (&dec);
1437 1792
1438 if (!(offset_return || !sv)) 1793 if (offset_return)
1794 *offset_return = dec.cur - SvPVX (string);
1795 else if (sv)
1439 { 1796 {
1440 // check for trailing garbage 1797 // check for trailing garbage
1441 decode_ws (&dec); 1798 decode_ws (&dec);
1442 1799
1443 if (*dec.cur) 1800 if (dec.cur != dec.end)
1444 { 1801 {
1445 dec.err = "garbage after JSON object"; 1802 dec.err = "garbage after JSON object";
1446 SvREFCNT_dec (sv); 1803 SvREFCNT_dec (sv);
1447 sv = 0; 1804 sv = 0;
1448 } 1805 }
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 } 1806 }
1460 1807
1461 if (!sv) 1808 if (!sv)
1462 { 1809 {
1463 SV *uni = sv_newmortal (); 1810 SV *uni = sv_newmortal ();
1469 SAVEVPTR (PL_curcop); 1816 SAVEVPTR (PL_curcop);
1470 PL_curcop = &cop; 1817 PL_curcop = &cop;
1471 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1818 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1472 LEAVE; 1819 LEAVE;
1473 1820
1474 croak ("%s, at character offset %d [\"%s\"]", 1821 croak ("%s, at character offset %d (before \"%s\")",
1475 dec.err, 1822 dec.err,
1476 (int)offset, 1823 (int)ptr_to_index (string, dec.cur),
1477 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1824 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1478 } 1825 }
1479 1826
1480 sv = sv_2mortal (sv); 1827 sv = sv_2mortal (sv);
1481 1828
1482 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1829 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)"); 1830 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1484 1831
1485 return sv; 1832 return sv;
1486} 1833}
1487 1834
1490 1837
1491static void 1838static void
1492incr_parse (JSON *self) 1839incr_parse (JSON *self)
1493{ 1840{
1494 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1841 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1842
1843 // the state machine here is a bit convoluted and could be simplified a lot
1844 // but this would make it slower, so...
1495 1845
1496 for (;;) 1846 for (;;)
1497 { 1847 {
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 1848 //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) 1849 switch (self->incr_mode)
1500 { 1850 {
1501 // only used for intiial whitespace skipping 1851 // only used for initial whitespace skipping
1502 case INCR_M_WS: 1852 case INCR_M_WS:
1503 for (;;) 1853 for (;;)
1504 { 1854 {
1505 if (*p > 0x20) 1855 if (*p > 0x20)
1506 { 1856 {
1857 if (*p == '#')
1858 {
1859 self->incr_mode = INCR_M_C0;
1860 goto incr_m_c;
1861 }
1862 else
1863 {
1507 self->incr_mode = INCR_M_JSON; 1864 self->incr_mode = INCR_M_JSON;
1508 goto incr_m_json; 1865 goto incr_m_json;
1866 }
1509 } 1867 }
1510 else if (!*p) 1868 else if (!*p)
1511 goto interrupt; 1869 goto interrupt;
1512 1870
1513 ++p; 1871 ++p;
1519 goto interrupt; 1877 goto interrupt;
1520 1878
1521 ++p; 1879 ++p;
1522 self->incr_mode = INCR_M_STR; 1880 self->incr_mode = INCR_M_STR;
1523 goto incr_m_str; 1881 goto incr_m_str;
1882
1883 // inside #-style comments
1884 case INCR_M_C0:
1885 case INCR_M_C1:
1886 incr_m_c:
1887 for (;;)
1888 {
1889 if (*p == '\n')
1890 {
1891 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1892 break;
1893 }
1894 else if (!*p)
1895 goto interrupt;
1896
1897 ++p;
1898 }
1899
1900 break;
1524 1901
1525 // inside a string 1902 // inside a string
1526 case INCR_M_STR: 1903 case INCR_M_STR:
1527 incr_m_str: 1904 incr_m_str:
1528 for (;;) 1905 for (;;)
1579 self->incr_mode = INCR_M_STR; 1956 self->incr_mode = INCR_M_STR;
1580 goto incr_m_str; 1957 goto incr_m_str;
1581 1958
1582 case '[': 1959 case '[':
1583 case '{': 1960 case '{':
1961 case '(':
1584 ++self->incr_nest; 1962 if (++self->incr_nest > self->max_depth)
1963 croak (ERR_NESTING_EXCEEDED);
1585 break; 1964 break;
1586 1965
1587 case ']': 1966 case ']':
1588 case '}': 1967 case '}':
1589 if (!--self->incr_nest) 1968 if (--self->incr_nest <= 0)
1590 goto interrupt; 1969 goto interrupt;
1970 break;
1971
1972 case ')':
1973 --self->incr_nest;
1974 break;
1975
1976 case '#':
1977 self->incr_mode = INCR_M_C1;
1978 goto incr_m_c;
1591 } 1979 }
1592 } 1980 }
1593 } 1981 }
1594 1982
1595 modechange: 1983 modechange:
1596 ; 1984 ;
1597 } 1985 }
1598 1986
1599interrupt: 1987interrupt:
1600 self->incr_pos = p - SvPVX (self->incr_text); 1988 self->incr_pos = p - SvPVX (self->incr_text);
1989 //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 1990 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1602} 1991}
1603 1992
1604///////////////////////////////////////////////////////////////////////////// 1993/////////////////////////////////////////////////////////////////////////////
1605// XS interface functions 1994// XS interface functions
1615 i >= '0' && i <= '9' ? i - '0' 2004 i >= '0' && i <= '9' ? i - '0'
1616 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2005 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1617 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2006 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1618 : -1; 2007 : -1;
1619 2008
1620 json_stash = gv_stashpv ("JSON::XS" , 1); 2009 json_stash = gv_stashpv ("JSON::XS" , 1);
1621 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2010 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1622
1623 json_true = get_bool ("JSON::XS::true");
1624 json_false = get_bool ("JSON::XS::false"); 2011 bool_false = get_bool ("Types::Serialiser::false");
2012 bool_true = get_bool ("Types::Serialiser::true");
2013
2014 sv_json = newSVpv ("JSON", 0);
2015 SvREADONLY_on (sv_json);
2016
2017 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1625} 2018}
1626 2019
1627PROTOTYPES: DISABLE 2020PROTOTYPES: DISABLE
1628 2021
1629void CLONE (...) 2022void CLONE (...)
1630 CODE: 2023 CODE:
2024 // as long as these writes are atomic, the race should not matter
2025 // as existing threads either already use 0, or use the old value,
2026 // which is sitll correct for the initial thread.
1631 json_stash = 0; 2027 json_stash = 0;
1632 json_boolean_stash = 0; 2028 bool_stash = 0;
2029 bool_false = 0;
2030 bool_true = 0;
1633 2031
1634void new (char *klass) 2032void new (char *klass)
1635 PPCODE: 2033 PPCODE:
1636{ 2034{
1637 SV *pv = NEWSV (0, sizeof (JSON)); 2035 SV *pv = NEWSV (0, sizeof (JSON));
1638 SvPOK_only (pv); 2036 SvPOK_only (pv);
1639 Zero (SvPVX (pv), 1, JSON); 2037 json_init ((JSON *)SvPVX (pv));
1640 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1641 XPUSHs (sv_2mortal (sv_bless ( 2038 XPUSHs (sv_2mortal (sv_bless (
1642 newRV_noinc (pv), 2039 newRV_noinc (pv),
1643 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2040 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1644 ))); 2041 )));
1645} 2042}
2043
2044void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2045 PPCODE:
2046 self->v_false = newSVsv (v_false);
2047 self->v_true = newSVsv (v_true);
2048 XPUSHs (ST (0));
2049
2050void get_boolean_values (JSON *self)
2051 PPCODE:
2052 if (self->v_false && self->v_true)
2053 {
2054 EXTEND (SP, 2);
2055 PUSHs (self->v_false);
2056 PUSHs (self->v_true);
2057 }
1646 2058
1647void ascii (JSON *self, int enable = 1) 2059void ascii (JSON *self, int enable = 1)
1648 ALIAS: 2060 ALIAS:
1649 ascii = F_ASCII 2061 ascii = F_ASCII
1650 latin1 = F_LATIN1 2062 latin1 = F_LATIN1
1657 allow_nonref = F_ALLOW_NONREF 2069 allow_nonref = F_ALLOW_NONREF
1658 shrink = F_SHRINK 2070 shrink = F_SHRINK
1659 allow_blessed = F_ALLOW_BLESSED 2071 allow_blessed = F_ALLOW_BLESSED
1660 convert_blessed = F_CONV_BLESSED 2072 convert_blessed = F_CONV_BLESSED
1661 relaxed = F_RELAXED 2073 relaxed = F_RELAXED
2074 allow_unknown = F_ALLOW_UNKNOWN
2075 allow_tags = F_ALLOW_TAGS
1662 PPCODE: 2076 PPCODE:
1663{ 2077{
1664 if (enable) 2078 if (enable)
1665 self->flags |= ix; 2079 self->flags |= ix;
1666 else 2080 else
1681 get_allow_nonref = F_ALLOW_NONREF 2095 get_allow_nonref = F_ALLOW_NONREF
1682 get_shrink = F_SHRINK 2096 get_shrink = F_SHRINK
1683 get_allow_blessed = F_ALLOW_BLESSED 2097 get_allow_blessed = F_ALLOW_BLESSED
1684 get_convert_blessed = F_CONV_BLESSED 2098 get_convert_blessed = F_CONV_BLESSED
1685 get_relaxed = F_RELAXED 2099 get_relaxed = F_RELAXED
2100 get_allow_unknown = F_ALLOW_UNKNOWN
2101 get_allow_tags = F_ALLOW_TAGS
1686 PPCODE: 2102 PPCODE:
1687 XPUSHs (boolSV (self->flags & ix)); 2103 XPUSHs (boolSV (self->flags & ix));
1688 2104
1689void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2105void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1690 PPCODE: 2106 PPCODE:
1691{ 2107 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)); 2108 XPUSHs (ST (0));
1702}
1703 2109
1704U32 get_max_depth (JSON *self) 2110U32 get_max_depth (JSON *self)
1705 CODE: 2111 CODE:
1706 RETVAL = DEC_DEPTH (self->flags); 2112 RETVAL = self->max_depth;
1707 OUTPUT: 2113 OUTPUT:
1708 RETVAL 2114 RETVAL
1709 2115
1710void max_size (JSON *self, UV max_size = 0) 2116void max_size (JSON *self, U32 max_size = 0)
1711 PPCODE: 2117 PPCODE:
1712{ 2118 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)); 2119 XPUSHs (ST (0));
1724}
1725 2120
1726int get_max_size (JSON *self) 2121int get_max_size (JSON *self)
1727 CODE: 2122 CODE:
1728 RETVAL = DEC_SIZE (self->flags); 2123 RETVAL = self->max_size;
1729 OUTPUT: 2124 OUTPUT:
1730 RETVAL 2125 RETVAL
1731 2126
1732void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2127void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1733 PPCODE: 2128 PPCODE:
1739} 2134}
1740 2135
1741void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2136void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1742 PPCODE: 2137 PPCODE:
1743{ 2138{
1744 if (!self->cb_sk_object) 2139 if (!self->cb_sk_object)
1745 self->cb_sk_object = newHV (); 2140 self->cb_sk_object = newHV ();
1746 2141
1747 if (SvOK (cb)) 2142 if (SvOK (cb))
1748 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2143 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1749 else 2144 else
1760 XPUSHs (ST (0)); 2155 XPUSHs (ST (0));
1761} 2156}
1762 2157
1763void encode (JSON *self, SV *scalar) 2158void encode (JSON *self, SV *scalar)
1764 PPCODE: 2159 PPCODE:
1765 XPUSHs (encode_json (scalar, self)); 2160 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2161 XPUSHs (scalar);
1766 2162
1767void decode (JSON *self, SV *jsonstr) 2163void decode (JSON *self, SV *jsonstr)
1768 PPCODE: 2164 PPCODE:
1769 XPUSHs (decode_json (jsonstr, self, 0)); 2165 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2166 XPUSHs (jsonstr);
1770 2167
1771void decode_prefix (JSON *self, SV *jsonstr) 2168void decode_prefix (JSON *self, SV *jsonstr)
1772 PPCODE: 2169 PPCODE:
1773{ 2170{
2171 SV *sv;
1774 STRLEN offset; 2172 STRLEN offset;
2173 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1775 EXTEND (SP, 2); 2174 EXTEND (SP, 2);
1776 PUSHs (decode_json (jsonstr, self, &offset)); 2175 PUSHs (sv);
1777 PUSHs (sv_2mortal (newSVuv (offset))); 2176 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1778} 2177}
1779 2178
1780void incr_parse (JSON *self, SV *jsonstr = 0) 2179void incr_parse (JSON *self, SV *jsonstr = 0)
1781 PPCODE: 2180 PPCODE:
1782{ 2181{
1783 if (!self->incr_text) 2182 if (!self->incr_text)
1784 self->incr_text = newSVpvn ("", 0); 2183 self->incr_text = newSVpvn ("", 0);
2184
2185 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2186 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2187 if (DECODE_WANTS_OCTETS (self))
2188 {
2189 if (self->incr_pos)
2190 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2191 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2192
2193 sv_utf8_downgrade (self->incr_text, 0);
2194 }
2195 else
2196 {
2197 sv_utf8_upgrade (self->incr_text);
2198
2199 if (self->incr_pos)
2200 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2201 - (U8 *)SvPVX (self->incr_text);
2202 }
1785 2203
1786 // append data, if any 2204 // append data, if any
1787 if (jsonstr) 2205 if (jsonstr)
1788 { 2206 {
2207 /* make sure both strings have same encoding */
1789 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2208 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2209 if (SvUTF8 (jsonstr))
2210 sv_utf8_downgrade (jsonstr, 0);
1790 { 2211 else
1791 /* utf-8-ness differs, need to upgrade */
1792 sv_utf8_upgrade (self->incr_text); 2212 sv_utf8_upgrade (jsonstr);
1793 2213
1794 if (self->incr_pos) 2214 /* 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 { 2215 {
1800 STRLEN len; 2216 STRLEN len;
1801 const char *str = SvPV (jsonstr, len); 2217 const char *str = SvPV (jsonstr, len);
1802 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2218 STRLEN cur = SvCUR (self->incr_text);
2219
2220 if (SvLEN (self->incr_text) - cur <= len)
2221 json_sv_grow (self->incr_text, cur, len);
2222
1803 Move (str, SvEND (self->incr_text), len, char); 2223 Move (str, SvEND (self->incr_text), len, char);
1804 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2224 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 2225 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1806 } 2226 }
1807 } 2227 }
1808 2228
1809 if (GIMME_V != G_VOID) 2229 if (GIMME_V != G_VOID)
1810 do 2230 do
1811 { 2231 {
2232 SV *sv;
1812 STRLEN offset; 2233 STRLEN offset;
1813 2234
1814 if (!INCR_DONE (self)) 2235 if (!INCR_DONE (self))
1815 { 2236 {
1816 incr_parse (self); 2237 incr_parse (self);
2238
2239 if (self->incr_pos > self->max_size && self->max_size)
2240 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2241 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2242
1817 if (!INCR_DONE (self)) 2243 if (!INCR_DONE (self))
2244 {
2245 // as an optimisation, do not accumulate white space in the incr buffer
2246 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2247 {
2248 self->incr_pos = 0;
2249 SvCUR_set (self->incr_text, 0);
2250 }
2251
1818 break; 2252 break;
2253 }
1819 } 2254 }
1820 2255
1821 XPUSHs (decode_json (self->incr_text, self, &offset)); 2256 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2257 XPUSHs (sv);
1822 2258
1823 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1824 self->incr_pos -= offset; 2259 self->incr_pos -= offset;
1825 self->incr_nest = 0; 2260 self->incr_nest = 0;
1826 self->incr_mode = 0; 2261 self->incr_mode = 0;
2262
2263 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1827 } 2264 }
1828 while (GIMME_V == G_ARRAY); 2265 while (GIMME_V == G_ARRAY);
1829} 2266}
1830 2267
1831SV *incr_text (JSON *self) 2268SV *incr_text (JSON *self)
1850 self->incr_nest = 0; 2287 self->incr_nest = 0;
1851 self->incr_mode = 0; 2288 self->incr_mode = 0;
1852 } 2289 }
1853} 2290}
1854 2291
2292void incr_reset (JSON *self)
2293 CODE:
2294{
2295 SvREFCNT_dec (self->incr_text);
2296 self->incr_text = 0;
2297 self->incr_pos = 0;
2298 self->incr_nest = 0;
2299 self->incr_mode = 0;
2300}
2301
1855void DESTROY (JSON *self) 2302void DESTROY (JSON *self)
1856 CODE: 2303 CODE:
2304 SvREFCNT_dec (self->v_false);
2305 SvREFCNT_dec (self->v_true);
1857 SvREFCNT_dec (self->cb_sk_object); 2306 SvREFCNT_dec (self->cb_sk_object);
1858 SvREFCNT_dec (self->cb_object); 2307 SvREFCNT_dec (self->cb_object);
1859 SvREFCNT_dec (self->incr_text); 2308 SvREFCNT_dec (self->incr_text);
1860 2309
1861PROTOTYPES: ENABLE 2310PROTOTYPES: ENABLE
1862 2311
1863void encode_json (SV *scalar) 2312void encode_json (SV *scalar)
1864 ALIAS:
1865 to_json_ = 0
1866 encode_json = F_UTF8
1867 PPCODE: 2313 PPCODE:
1868{ 2314{
1869 JSON json = { F_DEFAULT | ix }; 2315 JSON json;
1870 XPUSHs (encode_json (scalar, &json)); 2316 json_init (&json);
2317 json.flags |= F_UTF8;
2318 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2319 XPUSHs (scalar);
1871} 2320}
1872 2321
1873void decode_json (SV *jsonstr) 2322void decode_json (SV *jsonstr)
1874 ALIAS:
1875 from_json_ = 0
1876 decode_json = F_UTF8
1877 PPCODE: 2323 PPCODE:
1878{ 2324{
1879 JSON json = { F_DEFAULT | ix }; 2325 JSON json;
1880 XPUSHs (decode_json (jsonstr, &json, 0)); 2326 json_init (&json);
2327 json.flags |= F_UTF8;
2328 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2329 XPUSHs (jsonstr);
1881} 2330}
1882 2331
1883

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines