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 1.82 by root, Wed Mar 26 02:35:45 2008 UTC vs.
Revision 1.136 by root, Thu Nov 15 22:35:35 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
101 INCR_M_TFN, // inside true/false/null
102 INCR_M_NUM, // inside number
85 INCR_M_STR, // inside string 103 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash 104 INCR_M_BS, // inside backslash
105 INCR_M_C0, // inside comment in initial whitespace sequence
106 INCR_M_C1, // inside comment in other places
87 INCR_M_JSON // outside anything, count nesting 107 INCR_M_JSON // outside anything, count nesting
88}; 108};
89 109
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 110#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
91 111
92typedef struct { 112typedef struct {
93 U32 flags; 113 U32 flags;
114 U32 max_depth;
115 STRLEN max_size;
116
94 SV *cb_object; 117 SV *cb_object;
95 HV *cb_sk_object; 118 HV *cb_sk_object;
96 119
97 // for the incremental parser 120 // for the incremental parser
98 SV *incr_text; // the source text so far 121 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 122 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 123 int incr_nest; // {[]}-nesting level
101 int incr_mode; 124 unsigned char incr_mode;
125
126 SV *v_false, *v_true;
102} JSON; 127} JSON;
128
129INLINE void
130json_init (JSON *json)
131{
132 static const JSON init = { F_ALLOW_NONREF, 512 };
133
134 *json = init;
135}
103 136
104///////////////////////////////////////////////////////////////////////////// 137/////////////////////////////////////////////////////////////////////////////
105// utility functions 138// utility functions
106 139
140INLINE SV *
141get_bool (const char *name)
142{
143 SV *sv = get_sv (name, 1);
144
145 SvREADONLY_on (sv);
146 SvREADONLY_on (SvRV (sv));
147
148 return sv;
149}
150
107INLINE void 151INLINE void
108shrink (SV *sv) 152shrink (SV *sv)
109{ 153{
110 sv_utf8_downgrade (sv, 1); 154 sv_utf8_downgrade (sv, 1);
155
111 if (SvLEN (sv) > SvCUR (sv) + 1) 156 if (SvLEN (sv) > SvCUR (sv) + 1)
112 { 157 {
113#ifdef SvPV_shrink_to_cur 158#ifdef SvPV_shrink_to_cur
114 SvPV_shrink_to_cur (sv); 159 SvPV_shrink_to_cur (sv);
115#elif defined (SvPV_renew) 160#elif defined (SvPV_renew)
116 SvPV_renew (sv, SvCUR (sv) + 1); 161 SvPV_renew (sv, SvCUR (sv) + 1);
117#endif 162#endif
118 } 163 }
119} 164}
120 165
166/* adds two STRLENs together, slow, and with paranoia */
167STRLEN
168strlen_sum (STRLEN l1, STRLEN l2)
169{
170 size_t sum = l1 + l2;
171
172 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
173 croak ("JSON::XS: string size overflow");
174
175 return sum;
176}
177
178/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
179static char *
180json_sv_grow (SV *sv, size_t len1, size_t len2)
181{
182 len1 = strlen_sum (len1, len2);
183 len1 = strlen_sum (len1, len1 >> 1);
184
185 if (len1 > 4096 - 24)
186 len1 = (len1 | 4095) - 24;
187
188 return SvGROW (sv, len1);
189}
190
121// decode an utf-8 character and return it, or (UV)-1 in 191// decode a utf-8 character and return it, or (UV)-1 in
122// case of an error. 192// case of an error.
123// we special-case "safe" characters from U+80 .. U+7FF, 193// we special-case "safe" characters from U+80 .. U+7FF,
124// but use the very good perl function to parse anything else. 194// but use the very good perl function to parse anything else.
125// note that we never call this function for a ascii codepoints 195// note that we never call this function for a ascii codepoints
126INLINE UV 196INLINE UV
159 *s++ = 0x80 | ( ch & 0x3f); 229 *s++ = 0x80 | ( ch & 0x3f);
160 230
161 return s; 231 return s;
162} 232}
163 233
234// convert offset pointer to character index, sv must be string
235static STRLEN
236ptr_to_index (SV *sv, char *offset)
237{
238 return SvUTF8 (sv)
239 ? utf8_distance (offset, SvPVX (sv))
240 : offset - SvPVX (sv);
241}
242
243/////////////////////////////////////////////////////////////////////////////
244// fp hell
245
246// scan a group of digits, and a trailing exponent
247static void
248json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
249{
250 UV uaccum = 0;
251 int eaccum = 0;
252
253 // if we recurse too deep, skip all remaining digits
254 // to avoid a stack overflow attack
255 if (expect_false (--maxdepth <= 0))
256 while (((U8)*s - '0') < 10)
257 ++s;
258
259 for (;;)
260 {
261 U8 dig = (U8)*s - '0';
262
263 if (expect_false (dig >= 10))
264 {
265 if (dig == (U8)((U8)'.' - (U8)'0'))
266 {
267 ++s;
268 json_atof_scan1 (s, accum, expo, 1, maxdepth);
269 }
270 else if ((dig | ' ') == 'e' - '0')
271 {
272 int exp2 = 0;
273 int neg = 0;
274
275 ++s;
276
277 if (*s == '-')
278 {
279 ++s;
280 neg = 1;
281 }
282 else if (*s == '+')
283 ++s;
284
285 while ((dig = (U8)*s - '0') < 10)
286 exp2 = exp2 * 10 + *s++ - '0';
287
288 *expo += neg ? -exp2 : exp2;
289 }
290
291 break;
292 }
293
294 ++s;
295
296 uaccum = uaccum * 10 + dig;
297 ++eaccum;
298
299 // if we have too many digits, then recurse for more
300 // we actually do this for rather few digits
301 if (uaccum >= (UV_MAX - 9) / 10)
302 {
303 if (postdp) *expo -= eaccum;
304 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
305 if (postdp) *expo += eaccum;
306
307 break;
308 }
309 }
310
311 // this relies greatly on the quality of the pow ()
312 // implementation of the platform, but a good
313 // implementation is hard to beat.
314 // (IEEE 754 conformant ones are required to be exact)
315 if (postdp) *expo -= eaccum;
316 *accum += uaccum * Perl_pow (10., *expo);
317 *expo += eaccum;
318}
319
320static NV
321json_atof (const char *s)
322{
323 NV accum = 0.;
324 int expo = 0;
325 int neg = 0;
326
327 if (*s == '-')
328 {
329 ++s;
330 neg = 1;
331 }
332
333 // a recursion depth of ten gives us >>500 bits
334 json_atof_scan1 (s, &accum, &expo, 0, 10);
335
336 return neg ? -accum : accum;
337}
338
339// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
340static int
341ref_bool_type (SV *sv)
342{
343 svtype svt = SvTYPE (sv);
344
345 if (svt < SVt_PVAV)
346 {
347 STRLEN len = 0;
348 char *pv = svt ? SvPV (sv, len) : 0;
349
350 if (len == 1)
351 if (*pv == '1')
352 return 1;
353 else if (*pv == '0')
354 return 0;
355 }
356
357 return -1;
358}
359
360// returns whether scalar is not a reference in the sense of allow_nonref
361static int
362json_nonref (SV *scalar)
363{
364 if (!SvROK (scalar))
365 return 1;
366
367 scalar = SvRV (scalar);
368
369 if (SvTYPE (scalar) >= SVt_PVMG)
370 {
371 if (SvSTASH (scalar) == bool_stash)
372 return 1;
373
374 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
375 return 1;
376 }
377
378 return 0;
379}
380
164///////////////////////////////////////////////////////////////////////////// 381/////////////////////////////////////////////////////////////////////////////
165// encoder 382// encoder
166 383
167// structure used for encoding JSON 384// structure used for encoding JSON
168typedef struct 385typedef struct
170 char *cur; // SvPVX (sv) + current output position 387 char *cur; // SvPVX (sv) + current output position
171 char *end; // SvEND (sv) 388 char *end; // SvEND (sv)
172 SV *sv; // result scalar 389 SV *sv; // result scalar
173 JSON json; 390 JSON json;
174 U32 indent; // indentation level 391 U32 indent; // indentation level
175 U32 maxdepth; // max. indentation/recursion level
176 UV limit; // escape character values >= this value when encoding 392 UV limit; // escape character values >= this value when encoding
177} enc_t; 393} enc_t;
178 394
179INLINE void 395INLINE void
180need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
181{ 397{
182 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
183 { 399 {
184 STRLEN cur = enc->cur - SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
185 SvGROW (enc->sv, cur + len + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
186 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
187 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
188 } 404 }
189} 405}
190 406
191INLINE void 407INLINE void
192encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
208 424
209 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
210 { 426 {
211 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
212 { 428 {
213 need (enc, len += 1); 429 need (enc, len + 1);
214 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
215 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
216 } 432 }
217 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
218 { 434 {
219 need (enc, len += 1); 435 need (enc, len + 1);
220 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
221 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
222 } 438 }
223 else 439 else
224 *enc->cur++ = ch; 440 *enc->cur++ = ch;
227 } 443 }
228 else 444 else
229 { 445 {
230 switch (ch) 446 switch (ch)
231 { 447 {
232 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 448 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
233 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 449 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
234 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 450 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
235 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 451 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
236 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 452 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
237 453
238 default: 454 default:
239 { 455 {
240 STRLEN clen; 456 STRLEN clen;
241 UV uch; 457 UV uch;
257 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
258 { 474 {
259 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
260 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 476 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
261 477
262 need (enc, len += 11); 478 need (enc, len + 11);
263 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
264 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
265 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
266 enc->cur += 12; 482 enc->cur += 12;
267 } 483 }
268 else 484 else
269 { 485 {
270 static char hexdigit [16] = "0123456789abcdef";
271 need (enc, len += 5); 486 need (enc, len + 5);
272 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
273 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
274 *enc->cur++ = hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
275 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
276 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
277 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 492 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
278 } 493 }
279 494
280 str += clen; 495 str += clen;
281 } 496 }
282 else if (enc->json.flags & F_LATIN1) 497 else if (enc->json.flags & F_LATIN1)
284 *enc->cur++ = uch; 499 *enc->cur++ = uch;
285 str += clen; 500 str += clen;
286 } 501 }
287 else if (is_utf8) 502 else if (is_utf8)
288 { 503 {
289 need (enc, len += clen); 504 need (enc, len + clen);
290 do 505 do
291 { 506 {
292 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
293 } 508 }
294 while (--clen); 509 while (--clen);
295 } 510 }
296 else 511 else
297 { 512 {
298 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 513 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
299 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
300 ++str; 515 ++str;
301 } 516 }
302 } 517 }
303 } 518 }
353static void 568static void
354encode_av (enc_t *enc, AV *av) 569encode_av (enc_t *enc, AV *av)
355{ 570{
356 int i, len = av_len (av); 571 int i, len = av_len (av);
357 572
358 if (enc->indent >= enc->maxdepth) 573 if (enc->indent >= enc->json.max_depth)
359 croak ("data structure too deep (hit recursion limit)"); 574 croak (ERR_NESTING_EXCEEDED);
360 575
361 encode_ch (enc, '['); 576 encode_ch (enc, '[');
362 577
363 if (len >= 0) 578 if (len >= 0)
364 { 579 {
365 encode_nl (enc); ++enc->indent; 580 encode_nl (enc); ++enc->indent;
366 581
367 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
379 encode_comma (enc); 594 encode_comma (enc);
380 } 595 }
381 596
382 encode_nl (enc); --enc->indent; encode_indent (enc); 597 encode_nl (enc); --enc->indent; encode_indent (enc);
383 } 598 }
384 599
385 encode_ch (enc, ']'); 600 encode_ch (enc, ']');
386} 601}
387 602
388static void 603static void
389encode_hk (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
393 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
394 { 609 {
395 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
396 STRLEN len; 611 STRLEN len;
397 char *str; 612 char *str;
398 613
399 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
400 str = SvPV (sv, len); 615 str = SvPV (sv, len);
401 616
402 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
403 } 618 }
439static void 654static void
440encode_hv (enc_t *enc, HV *hv) 655encode_hv (enc_t *enc, HV *hv)
441{ 656{
442 HE *he; 657 HE *he;
443 658
444 if (enc->indent >= enc->maxdepth) 659 if (enc->indent >= enc->json.max_depth)
445 croak ("data structure too deep (hit recursion limit)"); 660 croak (ERR_NESTING_EXCEEDED);
446 661
447 encode_ch (enc, '{'); 662 encode_ch (enc, '{');
448 663
449 // for canonical output we have to sort by keys first 664 // for canonical output we have to sort by keys first
450 // actually, this is mostly due to the stupid so-called 665 // actually, this is mostly due to the stupid so-called
451 // security workaround added somewhere in 5.8.x. 666 // security workaround added somewhere in 5.8.x
452 // that randomises hash orderings 667 // that randomises hash orderings
453 if (enc->json.flags & F_CANONICAL) 668 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
454 { 669 {
455 int count = hv_iterinit (hv); 670 int count = hv_iterinit (hv);
456 671
457 if (SvMAGICAL (hv)) 672 if (SvMAGICAL (hv))
458 { 673 {
469 } 684 }
470 685
471 if (count) 686 if (count)
472 { 687 {
473 int i, fast = 1; 688 int i, fast = 1;
474#if defined(__BORLANDC__) || defined(_MSC_VER) 689 HE *hes_stack [STACK_HES];
475 HE **hes = _alloca (count * sizeof (HE)); 690 HE **hes = hes_stack;
476#else 691
477 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 692 // allocate larger arrays on the heap
478#endif 693 if (count > STACK_HES)
694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
479 698
480 i = 0; 699 i = 0;
481 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
482 { 701 {
483 hes [i++] = he; 702 hes [i++] = he;
552// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
553static void 772static void
554encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
555{ 774{
556 svtype svt; 775 svtype svt;
776 GV *method;
557 777
558 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
559 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
560 780
561 if (expect_false (SvOBJECT (sv))) 781 if (expect_false (SvOBJECT (sv)))
562 { 782 {
563 HV *stash = !JSON_SLOW || json_boolean_stash 783 HV *stash = SvSTASH (sv);
564 ? json_boolean_stash
565 : gv_stashpv ("JSON::XS::Boolean", 1);
566 784
567 if (SvSTASH (sv) == stash) 785 if (stash == bool_stash)
568 { 786 {
569 if (SvIV (sv))
570 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
571 else
572 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
573 } 789 }
790 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
791 {
792 int count;
793 dSP;
794
795 ENTER; SAVETMPS;
796 PUSHMARK (SP);
797 EXTEND (SP, 2);
798 // we re-bless the reference to get overload and other niceties right
799 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
800 PUSHs (sv_json);
801
802 PUTBACK;
803 count = call_sv ((SV *)GvCV (method), G_ARRAY);
804 SPAGAIN;
805
806 // catch this surprisingly common error
807 if (SvROK (TOPs) && SvRV (TOPs) == sv)
808 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
809
810 encode_ch (enc, '(');
811 encode_ch (enc, '"');
812 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
813 encode_ch (enc, '"');
814 encode_ch (enc, ')');
815 encode_ch (enc, '[');
816
817 if (count)
818 {
819 int i;
820
821 for (i = 0; i < count - 1; ++i)
822 {
823 encode_sv (enc, SP[i + 1 - count]);
824 encode_ch (enc, ',');
825 }
826
827 encode_sv (enc, TOPs);
828 SP -= count;
829 }
830
831 encode_ch (enc, ']');
832
833 FREETMPS; LEAVE;
834 }
835 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
836 {
837 dSP;
838
839 ENTER; SAVETMPS;
840 PUSHMARK (SP);
841 // we re-bless the reference to get overload and other niceties right
842 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
843
844 // calling with G_SCALAR ensures that we always get a 1 return value
845 PUTBACK;
846 call_sv ((SV *)GvCV (method), G_SCALAR);
847 SPAGAIN;
848
849 // catch this surprisingly common error
850 if (SvROK (TOPs) && SvRV (TOPs) == sv)
851 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
852
853 sv = POPs;
854 PUTBACK;
855
856 encode_sv (enc, sv);
857
858 FREETMPS; LEAVE;
859 }
860 else if (enc->json.flags & F_ALLOW_BLESSED)
861 encode_str (enc, "null", 4, 0);
574 else 862 else
575 {
576#if 0
577 if (0 && sv_derived_from (rv, "JSON::Literal"))
578 {
579 // not yet
580 }
581#endif
582 if (enc->json.flags & F_CONV_BLESSED)
583 {
584 // we re-bless the reference to get overload and other niceties right
585 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
586
587 if (to_json)
588 {
589 dSP;
590
591 ENTER; SAVETMPS; PUSHMARK (SP);
592 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
593
594 // calling with G_SCALAR ensures that we always get a 1 return value
595 PUTBACK;
596 call_sv ((SV *)GvCV (to_json), G_SCALAR);
597 SPAGAIN;
598
599 // catch this surprisingly common error
600 if (SvROK (TOPs) && SvRV (TOPs) == sv)
601 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
602
603 sv = POPs;
604 PUTBACK;
605
606 encode_sv (enc, sv);
607
608 FREETMPS; LEAVE;
609 }
610 else if (enc->json.flags & F_ALLOW_BLESSED)
611 encode_str (enc, "null", 4, 0);
612 else
613 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
614 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
615 }
616 else if (enc->json.flags & F_ALLOW_BLESSED)
617 encode_str (enc, "null", 4, 0);
618 else
619 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 863 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
620 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 864 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
621 }
622 } 865 }
623 else if (svt == SVt_PVHV) 866 else if (svt == SVt_PVHV)
624 encode_hv (enc, (HV *)sv); 867 encode_hv (enc, (HV *)sv);
625 else if (svt == SVt_PVAV) 868 else if (svt == SVt_PVAV)
626 encode_av (enc, (AV *)sv); 869 encode_av (enc, (AV *)sv);
627 else if (svt < SVt_PVAV) 870 else if (svt < SVt_PVAV)
628 { 871 {
629 STRLEN len = 0; 872 int bool_type = ref_bool_type (sv);
630 char *pv = svt ? SvPV (sv, len) : 0;
631 873
632 if (len == 1 && *pv == '1') 874 if (bool_type == 1)
633 encode_str (enc, "true", 4, 0); 875 encode_str (enc, "true", 4, 0);
634 else if (len == 1 && *pv == '0') 876 else if (bool_type == 0)
635 encode_str (enc, "false", 5, 0); 877 encode_str (enc, "false", 5, 0);
878 else if (enc->json.flags & F_ALLOW_UNKNOWN)
879 encode_str (enc, "null", 4, 0);
636 else 880 else
637 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 881 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
638 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 882 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
639 } 883 }
884 else if (enc->json.flags & F_ALLOW_UNKNOWN)
885 encode_str (enc, "null", 4, 0);
640 else 886 else
641 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 887 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
642 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 888 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
643} 889}
644 890
698 } 944 }
699 else 945 else
700 { 946 {
701 // large integer, use the (rather slow) snprintf way. 947 // large integer, use the (rather slow) snprintf way.
702 need (enc, IVUV_MAXCHARS); 948 need (enc, IVUV_MAXCHARS);
703 enc->cur += 949 enc->cur +=
704 SvIsUV(sv) 950 SvIsUV(sv)
705 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 951 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
706 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 952 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
707 } 953 }
708 } 954 }
709 else if (SvROK (sv)) 955 else if (SvROK (sv))
710 encode_rv (enc, SvRV (sv)); 956 encode_rv (enc, SvRV (sv));
711 else if (!SvOK (sv)) 957 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
712 encode_str (enc, "null", 4, 0); 958 encode_str (enc, "null", 4, 0);
713 else 959 else
714 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 960 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
715 SvPV_nolen (sv), SvFLAGS (sv)); 961 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
716} 962}
717 963
718static SV * 964static SV *
719encode_json (SV *scalar, JSON *json) 965encode_json (SV *scalar, JSON *json)
720{ 966{
721 enc_t enc; 967 enc_t enc;
722 968
723 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 969 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
724 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 970 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
725 971
726 enc.json = *json; 972 enc.json = *json;
727 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 973 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
728 enc.cur = SvPVX (enc.sv); 974 enc.cur = SvPVX (enc.sv);
729 enc.end = SvEND (enc.sv); 975 enc.end = SvEND (enc.sv);
730 enc.indent = 0; 976 enc.indent = 0;
731 enc.maxdepth = DEC_DEPTH (enc.json.flags);
732 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 977 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
733 : enc.json.flags & F_LATIN1 ? 0x000100UL 978 : enc.json.flags & F_LATIN1 ? 0x000100UL
734 : 0x110000UL; 979 : 0x110000UL;
735 980
736 SvPOK_only (enc.sv); 981 SvPOK_only (enc.sv);
737 encode_sv (&enc, scalar); 982 encode_sv (&enc, scalar);
983 encode_nl (&enc);
738 984
739 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 985 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
740 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 986 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
741 987
742 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 988 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
803 if (*dec->cur != ch) \ 1049 if (*dec->cur != ch) \
804 ERR (# ch " expected"); \ 1050 ERR (# ch " expected"); \
805 ++dec->cur; \ 1051 ++dec->cur; \
806 SE 1052 SE
807 1053
808#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1054#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
809#define DEC_DEC_DEPTH --dec->depth 1055#define DEC_DEC_DEPTH --dec->depth
810 1056
811static SV *decode_sv (dec_t *dec); 1057static SV *decode_sv (dec_t *dec);
812 1058
813static signed char decode_hexdigit[256]; 1059static signed char decode_hexdigit[256];
922 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1168 else if (expect_true (ch >= 0x20 && ch < 0x80))
923 *cur++ = ch; 1169 *cur++ = ch;
924 else if (ch >= 0x80) 1170 else if (ch >= 0x80)
925 { 1171 {
926 STRLEN clen; 1172 STRLEN clen;
927 UV uch;
928 1173
929 --dec_cur; 1174 --dec_cur;
930 1175
931 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1176 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
932 if (clen == (STRLEN)-1) 1177 if (clen == (STRLEN)-1)
933 ERR ("malformed UTF-8 character in JSON string"); 1178 ERR ("malformed UTF-8 character in JSON string");
934 1179
935 do 1180 do
936 *cur++ = *dec_cur++; 1181 *cur++ = *dec_cur++;
937 while (--clen); 1182 while (--clen);
938 1183
939 utf8 = 1; 1184 utf8 = 1;
940 } 1185 }
1186 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1187 *cur++ = ch;
941 else 1188 else
942 { 1189 {
943 --dec_cur; 1190 --dec_cur;
944 1191
945 if (!ch) 1192 if (!ch)
953 { 1200 {
954 STRLEN len = cur - buf; 1201 STRLEN len = cur - buf;
955 1202
956 if (sv) 1203 if (sv)
957 { 1204 {
958 SvGROW (sv, SvCUR (sv) + len + 1); 1205 STRLEN cur = SvCUR (sv);
1206
1207 if (SvLEN (sv) - cur <= len)
1208 json_sv_grow (sv, cur, len);
1209
959 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1210 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
960 SvCUR_set (sv, SvCUR (sv) + len); 1211 SvCUR_set (sv, SvCUR (sv) + len);
961 } 1212 }
962 else 1213 else
963 sv = newSVpvn (buf, len); 1214 sv = newSVpvn (buf, len);
1054 1305
1055 // special case the rather common 1..5-digit-int case 1306 // special case the rather common 1..5-digit-int case
1056 if (*start == '-') 1307 if (*start == '-')
1057 switch (len) 1308 switch (len)
1058 { 1309 {
1059 case 2: return newSViv (-( start [1] - '0' * 1)); 1310 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1060 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1311 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1061 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1312 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1062 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1313 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1063 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1314 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1064 } 1315 }
1065 else 1316 else
1066 switch (len) 1317 switch (len)
1067 { 1318 {
1068 case 1: return newSViv ( start [0] - '0' * 1); 1319 case 1: return newSViv ( start [0] - '0' * 1);
1069 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1320 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1070 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1321 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1071 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1322 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1072 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1323 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1073 } 1324 }
1074 1325
1075 { 1326 {
1076 UV uv; 1327 UV uv;
1077 int numtype = grok_number (start, len, &uv); 1328 int numtype = grok_number (start, len, &uv);
1086 } 1337 }
1087 1338
1088 len -= *start == '-' ? 1 : 0; 1339 len -= *start == '-' ? 1 : 0;
1089 1340
1090 // does not fit into IV or UV, try NV 1341 // does not fit into IV or UV, try NV
1091 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1342 if (len <= NV_DIG)
1092 #if defined (LDBL_DIG)
1093 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1094 #endif
1095 )
1096 // fits into NV without loss of precision 1343 // fits into NV without loss of precision
1097 return newSVnv (Atof (start)); 1344 return newSVnv (json_atof (start));
1098 1345
1099 // everything else fails, convert it to a string 1346 // everything else fails, convert it to a string
1100 return newSVpvn (start, dec->cur - start); 1347 return newSVpvn (start, dec->cur - start);
1101 } 1348 }
1102 1349
1103 // loss of precision here 1350 // loss of precision here
1104 return newSVnv (Atof (start)); 1351 return newSVnv (json_atof (start));
1105 1352
1106fail: 1353fail:
1107 return 0; 1354 return 0;
1108} 1355}
1109 1356
1133 if (*dec->cur == ']') 1380 if (*dec->cur == ']')
1134 { 1381 {
1135 ++dec->cur; 1382 ++dec->cur;
1136 break; 1383 break;
1137 } 1384 }
1138 1385
1139 if (*dec->cur != ',') 1386 if (*dec->cur != ',')
1140 ERR (", or ] expected while parsing array"); 1387 ERR (", or ] expected while parsing array");
1141 1388
1142 ++dec->cur; 1389 ++dec->cur;
1143 1390
1267 1514
1268 hv_iterinit (hv); 1515 hv_iterinit (hv);
1269 he = hv_iternext (hv); 1516 he = hv_iternext (hv);
1270 hv_iterinit (hv); 1517 hv_iterinit (hv);
1271 1518
1272 // the next line creates a mortal sv each time its called. 1519 // the next line creates a mortal sv each time it's called.
1273 // might want to optimise this for common cases. 1520 // might want to optimise this for common cases.
1274 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1521 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1275 1522
1276 if (cb) 1523 if (cb)
1277 { 1524 {
1278 dSP; 1525 dSP;
1279 int count; 1526 int count;
1280 1527
1281 ENTER; SAVETMPS; PUSHMARK (SP); 1528 ENTER; SAVETMPS;
1529 PUSHMARK (SP);
1282 XPUSHs (HeVAL (he)); 1530 XPUSHs (HeVAL (he));
1531 sv_2mortal (sv);
1283 1532
1284 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1533 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1285 1534
1286 if (count == 1) 1535 if (count == 1)
1287 { 1536 {
1288 sv = newSVsv (POPs); 1537 sv = newSVsv (POPs);
1289 FREETMPS; LEAVE; 1538 FREETMPS; LEAVE;
1290 return sv; 1539 return sv;
1291 } 1540 }
1541 else if (count)
1542 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1292 1543
1544 SvREFCNT_inc (sv);
1293 FREETMPS; LEAVE; 1545 FREETMPS; LEAVE;
1294 } 1546 }
1295 } 1547 }
1296 1548
1297 if (dec->json.cb_object) 1549 if (dec->json.cb_object)
1298 { 1550 {
1299 dSP; 1551 dSP;
1300 int count; 1552 int count;
1301 1553
1302 ENTER; SAVETMPS; PUSHMARK (SP); 1554 ENTER; SAVETMPS;
1555 PUSHMARK (SP);
1303 XPUSHs (sv_2mortal (sv)); 1556 XPUSHs (sv_2mortal (sv));
1304 1557
1305 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1558 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1306 1559
1307 if (count == 1) 1560 if (count == 1)
1308 { 1561 {
1309 sv = newSVsv (POPs); 1562 sv = newSVsv (POPs);
1310 FREETMPS; LEAVE; 1563 FREETMPS; LEAVE;
1311 return sv; 1564 return sv;
1312 } 1565 }
1566 else if (count)
1567 croak ("filter_json_object callbacks must not return more than one scalar");
1313 1568
1314 SvREFCNT_inc (sv); 1569 SvREFCNT_inc (sv);
1315 FREETMPS; LEAVE; 1570 FREETMPS; LEAVE;
1316 } 1571 }
1317 } 1572 }
1323 DEC_DEC_DEPTH; 1578 DEC_DEC_DEPTH;
1324 return 0; 1579 return 0;
1325} 1580}
1326 1581
1327static SV * 1582static SV *
1583decode_tag (dec_t *dec)
1584{
1585 SV *tag = 0;
1586 SV *val = 0;
1587
1588 if (!(dec->json.flags & F_ALLOW_TAGS))
1589 ERR ("malformed JSON string, neither array, object, number, string or atom");
1590
1591 ++dec->cur;
1592
1593 decode_ws (dec);
1594
1595 tag = decode_sv (dec);
1596 if (!tag)
1597 goto fail;
1598
1599 if (!SvPOK (tag))
1600 ERR ("malformed JSON string, (tag) must be a string");
1601
1602 decode_ws (dec);
1603
1604 if (*dec->cur != ')')
1605 ERR (") expected after tag");
1606
1607 ++dec->cur;
1608
1609 decode_ws (dec);
1610
1611 val = decode_sv (dec);
1612 if (!val)
1613 goto fail;
1614
1615 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1616 ERR ("malformed JSON string, tag value must be an array");
1617
1618 {
1619 AV *av = (AV *)SvRV (val);
1620 int i, len = av_len (av) + 1;
1621 HV *stash = gv_stashsv (tag, 0);
1622 SV *sv;
1623
1624 if (!stash)
1625 ERR ("cannot decode perl-object (package does not exist)");
1626
1627 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1628
1629 if (!method)
1630 ERR ("cannot decode perl-object (package does not have a THAW method)");
1631
1632 dSP;
1633
1634 ENTER; SAVETMPS;
1635 PUSHMARK (SP);
1636 EXTEND (SP, len + 2);
1637 // we re-bless the reference to get overload and other niceties right
1638 PUSHs (tag);
1639 PUSHs (sv_json);
1640
1641 for (i = 0; i < len; ++i)
1642 PUSHs (*av_fetch (av, i, 1));
1643
1644 PUTBACK;
1645 call_sv ((SV *)GvCV (method), G_SCALAR);
1646 SPAGAIN;
1647
1648 SvREFCNT_dec (tag);
1649 SvREFCNT_dec (val);
1650 sv = SvREFCNT_inc (POPs);
1651
1652 PUTBACK;
1653
1654 FREETMPS; LEAVE;
1655
1656 return sv;
1657 }
1658
1659fail:
1660 SvREFCNT_dec (tag);
1661 SvREFCNT_dec (val);
1662 return 0;
1663}
1664
1665static SV *
1328decode_sv (dec_t *dec) 1666decode_sv (dec_t *dec)
1329{ 1667{
1330 // the beauty of JSON: you need exactly one character lookahead 1668 // the beauty of JSON: you need exactly one character lookahead
1331 // to parse everything. 1669 // to parse everything.
1332 switch (*dec->cur) 1670 switch (*dec->cur)
1333 { 1671 {
1334 case '"': ++dec->cur; return decode_str (dec); 1672 case '"': ++dec->cur; return decode_str (dec);
1335 case '[': ++dec->cur; return decode_av (dec); 1673 case '[': ++dec->cur; return decode_av (dec);
1336 case '{': ++dec->cur; return decode_hv (dec); 1674 case '{': ++dec->cur; return decode_hv (dec);
1675 case '(': return decode_tag (dec);
1337 1676
1338 case '-': 1677 case '-':
1339 case '0': case '1': case '2': case '3': case '4': 1678 case '0': case '1': case '2': case '3': case '4':
1340 case '5': case '6': case '7': case '8': case '9': 1679 case '5': case '6': case '7': case '8': case '9':
1341 return decode_num (dec); 1680 return decode_num (dec);
1342 1681
1682 case 'f':
1683 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1684 {
1685 dec->cur += 5;
1686
1687 if (expect_false (!dec->json.v_false))
1688 dec->json.v_false = GET_BOOL (false);
1689
1690 return newSVsv (dec->json.v_false);
1691 }
1692 else
1693 ERR ("'false' expected");
1694
1695 break;
1696
1343 case 't': 1697 case 't':
1344 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1698 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1345 { 1699 {
1346 dec->cur += 4; 1700 dec->cur += 4;
1347#if JSON_SLOW 1701
1348 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1702 if (expect_false (!dec->json.v_true))
1349#endif 1703 dec->json.v_true = GET_BOOL (true);
1350 return SvREFCNT_inc (json_true); 1704
1705 return newSVsv (dec->json.v_true);
1351 } 1706 }
1352 else 1707 else
1353 ERR ("'true' expected"); 1708 ERR ("'true' expected");
1354
1355 break;
1356
1357 case 'f':
1358 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1359 {
1360 dec->cur += 5;
1361#if JSON_SLOW
1362 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1363#endif
1364 return SvREFCNT_inc (json_false);
1365 }
1366 else
1367 ERR ("'false' expected");
1368 1709
1369 break; 1710 break;
1370 1711
1371 case 'n': 1712 case 'n':
1372 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1713 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1378 ERR ("'null' expected"); 1719 ERR ("'null' expected");
1379 1720
1380 break; 1721 break;
1381 1722
1382 default: 1723 default:
1383 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1724 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1384 break; 1725 break;
1385 } 1726 }
1386 1727
1387fail: 1728fail:
1388 return 0; 1729 return 0;
1390 1731
1391static SV * 1732static SV *
1392decode_json (SV *string, JSON *json, STRLEN *offset_return) 1733decode_json (SV *string, JSON *json, STRLEN *offset_return)
1393{ 1734{
1394 dec_t dec; 1735 dec_t dec;
1395 STRLEN offset;
1396 SV *sv; 1736 SV *sv;
1397 1737
1738 /* work around bugs in 5.10 where manipulating magic values
1739 * makes perl ignore the magic in subsequent accesses.
1740 * also make a copy of non-PV values, to get them into a clean
1741 * state (SvPV should do that, but it's buggy, see below).
1742 *
1743 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1744 * as reported by Reini Urban.
1745 */
1398 SvGETMAGIC (string); 1746 /*SvGETMAGIC (string);*/
1747 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1748 string = sv_2mortal (newSVsv (string));
1749
1399 SvUPGRADE (string, SVt_PV); 1750 SvUPGRADE (string, SVt_PV);
1400 1751
1401 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1752 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1753 * assertion with -DDEBUGGING, although SvCUR is documented to
1754 * return the xpv_cur field which certainly exists after upgrading.
1755 * according to nicholas clark, calling SvPOK fixes this.
1756 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1757 * and hope for the best.
1758 * Damnit, SvPV_nolen still trips over yet another assertion. This
1759 * assertion business is seriously broken, try yet another workaround
1760 * for the broken -DDEBUGGING.
1761 */
1762 {
1763#ifdef DEBUGGING
1764 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1765#else
1766 STRLEN offset = SvCUR (string);
1767#endif
1768
1769 if (offset > json->max_size && json->max_size)
1402 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1770 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1403 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1771 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1772 }
1404 1773
1405 if (json->flags & F_UTF8) 1774 if (DECODE_WANTS_OCTETS (json))
1406 sv_utf8_downgrade (string, 0); 1775 sv_utf8_downgrade (string, 0);
1407 else 1776 else
1408 sv_utf8_upgrade (string); 1777 sv_utf8_upgrade (string);
1409 1778
1410 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1779 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1411 1780
1412 dec.json = *json; 1781 dec.json = *json;
1413 dec.cur = SvPVX (string); 1782 dec.cur = SvPVX (string);
1414 dec.end = SvEND (string); 1783 dec.end = SvEND (string);
1415 dec.err = 0; 1784 dec.err = 0;
1416 dec.depth = 0; 1785 dec.depth = 0;
1417 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1418 1786
1419 if (dec.json.cb_object || dec.json.cb_sk_object) 1787 if (dec.json.cb_object || dec.json.cb_sk_object)
1420 dec.json.flags |= F_HOOK; 1788 dec.json.flags |= F_HOOK;
1421 1789
1422 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1790 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1423 1791
1424 decode_ws (&dec); 1792 decode_ws (&dec);
1425 sv = decode_sv (&dec); 1793 sv = decode_sv (&dec);
1426 1794
1427 if (!(offset_return || !sv)) 1795 if (offset_return)
1796 *offset_return = dec.cur - SvPVX (string);
1797 else if (sv)
1428 { 1798 {
1429 // check for trailing garbage 1799 // check for trailing garbage
1430 decode_ws (&dec); 1800 decode_ws (&dec);
1431 1801
1432 if (*dec.cur) 1802 if (dec.cur != dec.end)
1433 { 1803 {
1434 dec.err = "garbage after JSON object"; 1804 dec.err = "garbage after JSON object";
1435 SvREFCNT_dec (sv); 1805 SvREFCNT_dec (sv);
1436 sv = 0; 1806 sv = 0;
1437 } 1807 }
1438 }
1439
1440 if (offset_return || !sv)
1441 {
1442 offset = dec.json.flags & F_UTF8
1443 ? dec.cur - SvPVX (string)
1444 : utf8_distance (dec.cur, SvPVX (string));
1445
1446 if (offset_return)
1447 *offset_return = offset;
1448 } 1808 }
1449 1809
1450 if (!sv) 1810 if (!sv)
1451 { 1811 {
1452 SV *uni = sv_newmortal (); 1812 SV *uni = sv_newmortal ();
1458 SAVEVPTR (PL_curcop); 1818 SAVEVPTR (PL_curcop);
1459 PL_curcop = &cop; 1819 PL_curcop = &cop;
1460 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1820 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1461 LEAVE; 1821 LEAVE;
1462 1822
1463 croak ("%s, at character offset %d [\"%s\"]", 1823 croak ("%s, at character offset %d (before \"%s\")",
1464 dec.err, 1824 dec.err,
1465 (int)offset, 1825 (int)ptr_to_index (string, dec.cur),
1466 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1826 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1467 } 1827 }
1468 1828
1469 sv = sv_2mortal (sv); 1829 sv = sv_2mortal (sv);
1470 1830
1471 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1831 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1472 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1832 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1473 1833
1474 return sv; 1834 return sv;
1475} 1835}
1476 1836
1480static void 1840static void
1481incr_parse (JSON *self) 1841incr_parse (JSON *self)
1482{ 1842{
1483 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1843 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1484 1844
1845 // the state machine here is a bit convoluted and could be simplified a lot
1846 // but this would make it slower, so...
1847
1485 for (;;) 1848 for (;;)
1486 { 1849 {
1487 //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
1488 switch (self->incr_mode) 1850 switch (self->incr_mode)
1489 { 1851 {
1852 // reached end of a scalar, see if we are inside a nested structure or not
1853 end_of_scalar:
1854 self->incr_mode = INCR_M_JSON;
1855
1856 if (self->incr_nest) // end of a scalar inside array, object or tag
1857 goto incr_m_json;
1858 else // end of scalar outside structure, json text ends here
1859 goto interrupt;
1860
1490 // only used for intiial whitespace skipping 1861 // only used for initial whitespace skipping
1491 case INCR_M_WS: 1862 case INCR_M_WS:
1492 for (;;) 1863 for (;;)
1493 { 1864 {
1494 if (*p > 0x20) 1865 if (*p > 0x20)
1495 { 1866 {
1867 if (*p == '#')
1868 {
1869 self->incr_mode = INCR_M_C0;
1870 goto incr_m_c;
1871 }
1872 else
1873 {
1496 self->incr_mode = INCR_M_JSON; 1874 self->incr_mode = INCR_M_JSON;
1497 goto incr_m_json; 1875 goto incr_m_json;
1876 }
1498 } 1877 }
1499 else if (!*p) 1878 else if (!*p)
1500 goto interrupt; 1879 goto interrupt;
1501 1880
1502 ++p; 1881 ++p;
1508 goto interrupt; 1887 goto interrupt;
1509 1888
1510 ++p; 1889 ++p;
1511 self->incr_mode = INCR_M_STR; 1890 self->incr_mode = INCR_M_STR;
1512 goto incr_m_str; 1891 goto incr_m_str;
1892
1893 // inside #-style comments
1894 case INCR_M_C0:
1895 case INCR_M_C1:
1896 incr_m_c:
1897 for (;;)
1898 {
1899 if (*p == '\n')
1900 {
1901 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1902 break;
1903 }
1904 else if (!*p)
1905 goto interrupt;
1906
1907 ++p;
1908 }
1909
1910 break;
1911
1912 // inside true/false/null
1913 case INCR_M_TFN:
1914 incr_m_tfn:
1915 for (;;)
1916 switch (*p++)
1917 {
1918 case 'r': case 'u': case 'e': // tRUE, falsE, nUll
1919 case 'a': case 'l': case 's': // fALSe, nuLL
1920 // allowed
1921 break;
1922
1923 default:
1924 --p;
1925 goto end_of_scalar;
1926 }
1927
1928 // inside a number
1929 case INCR_M_NUM:
1930 incr_m_num:
1931 for (;;)
1932 switch (*p++)
1933 {
1934 case 'e': case 'E': case '.': case '+':
1935 case '-':
1936 case '0': case '1': case '2': case '3': case '4':
1937 case '5': case '6': case '7': case '8': case '9':
1938 // allowed
1939 break;
1940
1941 default:
1942 --p;
1943 goto end_of_scalar;
1944 }
1513 1945
1514 // inside a string 1946 // inside a string
1515 case INCR_M_STR: 1947 case INCR_M_STR:
1516 incr_m_str: 1948 incr_m_str:
1517 for (;;) 1949 for (;;)
1518 { 1950 {
1519 if (*p == '"') 1951 if (*p == '"')
1520 { 1952 {
1521 ++p; 1953 ++p;
1522 self->incr_mode = INCR_M_JSON;
1523
1524 if (!self->incr_nest)
1525 goto interrupt;
1526
1527 goto incr_m_json; 1954 goto end_of_scalar;
1528 } 1955 }
1529 else if (*p == '\\') 1956 else if (*p == '\\')
1530 { 1957 {
1531 ++p; // "virtually" consumes character after \ 1958 ++p; // "virtually" consumes character after \
1532 1959
1562 --p; // do not eat the whitespace, let the next round do it 1989 --p; // do not eat the whitespace, let the next round do it
1563 goto interrupt; 1990 goto interrupt;
1564 } 1991 }
1565 break; 1992 break;
1566 1993
1994 // the following three blocks handle scalars. this makes the parser
1995 // more strict than required inside arrays or objects, and could
1996 // be moved to a special case on the toplevel (except strings)
1997 case 't':
1998 case 'f':
1999 case 'n':
2000 self->incr_mode = INCR_M_TFN;
2001 goto incr_m_tfn;
2002
2003 case '-':
2004 case '0': case '1': case '2': case '3': case '4':
2005 case '5': case '6': case '7': case '8': case '9':
2006 self->incr_mode = INCR_M_NUM;
2007 goto incr_m_num;
2008
1567 case '"': 2009 case '"':
1568 self->incr_mode = INCR_M_STR; 2010 self->incr_mode = INCR_M_STR;
1569 goto incr_m_str; 2011 goto incr_m_str;
1570 2012
1571 case '[': 2013 case '[':
1572 case '{': 2014 case '{':
2015 case '(':
1573 ++self->incr_nest; 2016 if (++self->incr_nest > self->max_depth)
2017 croak (ERR_NESTING_EXCEEDED);
1574 break; 2018 break;
1575 2019
1576 case ']': 2020 case ']':
1577 case '}': 2021 case '}':
1578 if (!--self->incr_nest) 2022 if (--self->incr_nest <= 0)
1579 goto interrupt; 2023 goto interrupt;
2024 break;
2025
2026 case ')':
2027 --self->incr_nest;
2028 break;
2029
2030 case '#':
2031 self->incr_mode = INCR_M_C1;
2032 goto incr_m_c;
1580 } 2033 }
1581 } 2034 }
1582 } 2035 }
1583 2036
1584 modechange: 2037 modechange:
1585 ; 2038 ;
1586 } 2039 }
1587 2040
1588interrupt: 2041interrupt:
1589 self->incr_pos = p - SvPVX (self->incr_text); 2042 self->incr_pos = p - SvPVX (self->incr_text);
2043 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1590 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 2044 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1591} 2045}
1592 2046
1593///////////////////////////////////////////////////////////////////////////// 2047/////////////////////////////////////////////////////////////////////////////
1594// XS interface functions 2048// XS interface functions
1604 i >= '0' && i <= '9' ? i - '0' 2058 i >= '0' && i <= '9' ? i - '0'
1605 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2059 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1606 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2060 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1607 : -1; 2061 : -1;
1608 2062
1609 json_stash = gv_stashpv ("JSON::XS" , 1); 2063 json_stash = gv_stashpv ("JSON::XS" , 1);
1610 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2064 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
2065 bool_false = get_bool ("Types::Serialiser::false");
2066 bool_true = get_bool ("Types::Serialiser::true");
1611 2067
1612 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 2068 sv_json = newSVpv ("JSON", 0);
1613 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 2069 SvREADONLY_on (sv_json);
2070
2071 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1614} 2072}
1615 2073
1616PROTOTYPES: DISABLE 2074PROTOTYPES: DISABLE
1617 2075
1618void CLONE (...) 2076void CLONE (...)
1619 CODE: 2077 CODE:
2078 // as long as these writes are atomic, the race should not matter
2079 // as existing threads either already use 0, or use the old value,
2080 // which is sitll correct for the initial thread.
1620 json_stash = 0; 2081 json_stash = 0;
1621 json_boolean_stash = 0; 2082 bool_stash = 0;
2083 bool_false = 0;
2084 bool_true = 0;
1622 2085
1623void new (char *klass) 2086void new (char *klass)
1624 PPCODE: 2087 PPCODE:
1625{ 2088{
1626 SV *pv = NEWSV (0, sizeof (JSON)); 2089 SV *pv = NEWSV (0, sizeof (JSON));
1627 SvPOK_only (pv); 2090 SvPOK_only (pv);
1628 Zero (SvPVX (pv), 1, JSON); 2091 json_init ((JSON *)SvPVX (pv));
1629 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1630 XPUSHs (sv_2mortal (sv_bless ( 2092 XPUSHs (sv_2mortal (sv_bless (
1631 newRV_noinc (pv), 2093 newRV_noinc (pv),
1632 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2094 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1633 ))); 2095 )));
1634} 2096}
2097
2098void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2099 PPCODE:
2100 self->v_false = newSVsv (v_false);
2101 self->v_true = newSVsv (v_true);
2102 XPUSHs (ST (0));
2103
2104void get_boolean_values (JSON *self)
2105 PPCODE:
2106 if (self->v_false && self->v_true)
2107 {
2108 EXTEND (SP, 2);
2109 PUSHs (self->v_false);
2110 PUSHs (self->v_true);
2111 }
1635 2112
1636void ascii (JSON *self, int enable = 1) 2113void ascii (JSON *self, int enable = 1)
1637 ALIAS: 2114 ALIAS:
1638 ascii = F_ASCII 2115 ascii = F_ASCII
1639 latin1 = F_LATIN1 2116 latin1 = F_LATIN1
1646 allow_nonref = F_ALLOW_NONREF 2123 allow_nonref = F_ALLOW_NONREF
1647 shrink = F_SHRINK 2124 shrink = F_SHRINK
1648 allow_blessed = F_ALLOW_BLESSED 2125 allow_blessed = F_ALLOW_BLESSED
1649 convert_blessed = F_CONV_BLESSED 2126 convert_blessed = F_CONV_BLESSED
1650 relaxed = F_RELAXED 2127 relaxed = F_RELAXED
2128 allow_unknown = F_ALLOW_UNKNOWN
2129 allow_tags = F_ALLOW_TAGS
1651 PPCODE: 2130 PPCODE:
1652{ 2131{
1653 if (enable) 2132 if (enable)
1654 self->flags |= ix; 2133 self->flags |= ix;
1655 else 2134 else
1670 get_allow_nonref = F_ALLOW_NONREF 2149 get_allow_nonref = F_ALLOW_NONREF
1671 get_shrink = F_SHRINK 2150 get_shrink = F_SHRINK
1672 get_allow_blessed = F_ALLOW_BLESSED 2151 get_allow_blessed = F_ALLOW_BLESSED
1673 get_convert_blessed = F_CONV_BLESSED 2152 get_convert_blessed = F_CONV_BLESSED
1674 get_relaxed = F_RELAXED 2153 get_relaxed = F_RELAXED
2154 get_allow_unknown = F_ALLOW_UNKNOWN
2155 get_allow_tags = F_ALLOW_TAGS
1675 PPCODE: 2156 PPCODE:
1676 XPUSHs (boolSV (self->flags & ix)); 2157 XPUSHs (boolSV (self->flags & ix));
1677 2158
1678void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2159void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1679 PPCODE: 2160 PPCODE:
1680{ 2161 self->max_depth = max_depth;
1681 UV log2 = 0;
1682
1683 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1684
1685 while ((1UL << log2) < max_depth)
1686 ++log2;
1687
1688 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1689
1690 XPUSHs (ST (0)); 2162 XPUSHs (ST (0));
1691}
1692 2163
1693U32 get_max_depth (JSON *self) 2164U32 get_max_depth (JSON *self)
1694 CODE: 2165 CODE:
1695 RETVAL = DEC_DEPTH (self->flags); 2166 RETVAL = self->max_depth;
1696 OUTPUT: 2167 OUTPUT:
1697 RETVAL 2168 RETVAL
1698 2169
1699void max_size (JSON *self, UV max_size = 0) 2170void max_size (JSON *self, U32 max_size = 0)
1700 PPCODE: 2171 PPCODE:
1701{ 2172 self->max_size = max_size;
1702 UV log2 = 0;
1703
1704 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1705 if (max_size == 1) max_size = 2;
1706
1707 while ((1UL << log2) < max_size)
1708 ++log2;
1709
1710 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1711
1712 XPUSHs (ST (0)); 2173 XPUSHs (ST (0));
1713}
1714 2174
1715int get_max_size (JSON *self) 2175int get_max_size (JSON *self)
1716 CODE: 2176 CODE:
1717 RETVAL = DEC_SIZE (self->flags); 2177 RETVAL = self->max_size;
1718 OUTPUT: 2178 OUTPUT:
1719 RETVAL 2179 RETVAL
1720 2180
1721void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2181void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1722 PPCODE: 2182 PPCODE:
1728} 2188}
1729 2189
1730void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2190void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1731 PPCODE: 2191 PPCODE:
1732{ 2192{
1733 if (!self->cb_sk_object) 2193 if (!self->cb_sk_object)
1734 self->cb_sk_object = newHV (); 2194 self->cb_sk_object = newHV ();
1735 2195
1736 if (SvOK (cb)) 2196 if (SvOK (cb))
1737 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2197 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1738 else 2198 else
1749 XPUSHs (ST (0)); 2209 XPUSHs (ST (0));
1750} 2210}
1751 2211
1752void encode (JSON *self, SV *scalar) 2212void encode (JSON *self, SV *scalar)
1753 PPCODE: 2213 PPCODE:
1754 XPUSHs (encode_json (scalar, self)); 2214 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2215 XPUSHs (scalar);
1755 2216
1756void decode (JSON *self, SV *jsonstr) 2217void decode (JSON *self, SV *jsonstr)
1757 PPCODE: 2218 PPCODE:
1758 XPUSHs (decode_json (jsonstr, self, 0)); 2219 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2220 XPUSHs (jsonstr);
1759 2221
1760void decode_prefix (JSON *self, SV *jsonstr) 2222void decode_prefix (JSON *self, SV *jsonstr)
1761 PPCODE: 2223 PPCODE:
1762{ 2224{
2225 SV *sv;
1763 STRLEN offset; 2226 STRLEN offset;
2227 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1764 EXTEND (SP, 2); 2228 EXTEND (SP, 2);
1765 PUSHs (decode_json (jsonstr, self, &offset)); 2229 PUSHs (sv);
1766 PUSHs (sv_2mortal (newSVuv (offset))); 2230 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1767} 2231}
1768 2232
1769void incr_parse (JSON *self, SV *jsonstr = 0) 2233void incr_parse (JSON *self, SV *jsonstr = 0)
1770 PPCODE: 2234 PPCODE:
1771{ 2235{
1772 if (!self->incr_text) 2236 if (!self->incr_text)
1773 self->incr_text = newSVpvn ("", 0); 2237 self->incr_text = newSVpvn ("", 0);
2238
2239 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2240 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2241 if (DECODE_WANTS_OCTETS (self))
2242 {
2243 if (self->incr_pos)
2244 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2245 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2246
2247 sv_utf8_downgrade (self->incr_text, 0);
2248 }
2249 else
2250 {
2251 sv_utf8_upgrade (self->incr_text);
2252
2253 if (self->incr_pos)
2254 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2255 - (U8 *)SvPVX (self->incr_text);
2256 }
1774 2257
1775 // append data, if any 2258 // append data, if any
1776 if (jsonstr) 2259 if (jsonstr)
1777 { 2260 {
2261 /* make sure both strings have same encoding */
1778 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2262 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2263 if (SvUTF8 (jsonstr))
2264 sv_utf8_downgrade (jsonstr, 0);
1779 { 2265 else
1780 /* utf-8-ness differs, need to upgrade */
1781 sv_utf8_upgrade (self->incr_text); 2266 sv_utf8_upgrade (jsonstr);
1782 2267
1783 if (self->incr_pos) 2268 /* and then just blindly append */
1784 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1785 - (U8 *)SvPVX (self->incr_text);
1786 }
1787
1788 { 2269 {
1789 STRLEN len; 2270 STRLEN len;
1790 const char *str = SvPV (jsonstr, len); 2271 const char *str = SvPV (jsonstr, len);
1791 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2272 STRLEN cur = SvCUR (self->incr_text);
2273
2274 if (SvLEN (self->incr_text) - cur <= len)
2275 json_sv_grow (self->incr_text, cur, len);
2276
1792 Move (str, SvEND (self->incr_text), len, char); 2277 Move (str, SvEND (self->incr_text), len, char);
1793 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2278 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1794 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2279 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1795 } 2280 }
1796 } 2281 }
1797 2282
1798 if (GIMME_V != G_VOID) 2283 if (GIMME_V != G_VOID)
1799 do 2284 do
1800 { 2285 {
2286 SV *sv;
1801 STRLEN offset; 2287 STRLEN offset;
1802 2288
1803 if (!INCR_DONE (self)) 2289 if (!INCR_DONE (self))
1804 { 2290 {
1805 incr_parse (self); 2291 incr_parse (self);
2292
2293 if (self->incr_pos > self->max_size && self->max_size)
2294 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2295 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2296
1806 if (!INCR_DONE (self)) 2297 if (!INCR_DONE (self))
2298 {
2299 // as an optimisation, do not accumulate white space in the incr buffer
2300 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2301 {
2302 self->incr_pos = 0;
2303 SvCUR_set (self->incr_text, 0);
2304 }
2305
1807 break; 2306 break;
2307 }
1808 } 2308 }
1809 2309
1810 XPUSHs (decode_json (self->incr_text, self, &offset)); 2310 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2311 XPUSHs (sv);
1811 2312
1812 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1813 self->incr_pos -= offset; 2313 self->incr_pos -= offset;
1814 self->incr_nest = 0; 2314 self->incr_nest = 0;
1815 self->incr_mode = 0; 2315 self->incr_mode = 0;
2316
2317 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1816 } 2318 }
1817 while (GIMME_V == G_ARRAY); 2319 while (GIMME_V == G_ARRAY);
1818} 2320}
1819 2321
1820SV *incr_text (JSON *self) 2322SV *incr_text (JSON *self)
1839 self->incr_nest = 0; 2341 self->incr_nest = 0;
1840 self->incr_mode = 0; 2342 self->incr_mode = 0;
1841 } 2343 }
1842} 2344}
1843 2345
2346void incr_reset (JSON *self)
2347 CODE:
2348{
2349 SvREFCNT_dec (self->incr_text);
2350 self->incr_text = 0;
2351 self->incr_pos = 0;
2352 self->incr_nest = 0;
2353 self->incr_mode = 0;
2354}
2355
1844void DESTROY (JSON *self) 2356void DESTROY (JSON *self)
1845 CODE: 2357 CODE:
2358 SvREFCNT_dec (self->v_false);
2359 SvREFCNT_dec (self->v_true);
1846 SvREFCNT_dec (self->cb_sk_object); 2360 SvREFCNT_dec (self->cb_sk_object);
1847 SvREFCNT_dec (self->cb_object); 2361 SvREFCNT_dec (self->cb_object);
1848 SvREFCNT_dec (self->incr_text); 2362 SvREFCNT_dec (self->incr_text);
1849 2363
1850PROTOTYPES: ENABLE 2364PROTOTYPES: ENABLE
1851 2365
1852void encode_json (SV *scalar) 2366void encode_json (SV *scalar)
1853 ALIAS:
1854 to_json_ = 0
1855 encode_json = F_UTF8
1856 PPCODE: 2367 PPCODE:
1857{ 2368{
1858 JSON json = { F_DEFAULT | ix }; 2369 JSON json;
1859 XPUSHs (encode_json (scalar, &json)); 2370 json_init (&json);
2371 json.flags |= F_UTF8;
2372 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2373 XPUSHs (scalar);
1860} 2374}
1861 2375
1862void decode_json (SV *jsonstr) 2376void decode_json (SV *jsonstr)
1863 ALIAS:
1864 from_json_ = 0
1865 decode_json = F_UTF8
1866 PPCODE: 2377 PPCODE:
1867{ 2378{
1868 JSON json = { F_DEFAULT | ix }; 2379 JSON json;
1869 XPUSHs (decode_json (jsonstr, &json, 0)); 2380 json_init (&json);
2381 json.flags |= F_UTF8;
2382 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2383 XPUSHs (jsonstr);
1870} 2384}
1871 2385
1872

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines