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 1.84 by root, Thu Mar 27 06:37:35 2008 UTC vs.
Revision 1.143 by root, Mon Jun 21 01:04:30 2021 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 15#endif
15 16
16// some old perls do not have this, try to make it work, no 17// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 18// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
20#endif 21#endif
21 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 36
24#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
32#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
36#define F_ALLOW_UNKNOWN 0x00002000UL 49#define F_ALLOW_UNKNOWN 0x00002000UL
37 50#define F_ALLOW_TAGS 0x00004000UL
38#define F_MAXDEPTH 0xf8000000UL
39#define S_MAXDEPTH 27
40#define F_MAXSIZE 0x01f00000UL
41#define S_MAXSIZE 20
42#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
43 52
44#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
45#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
46
47#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
48#define F_DEFAULT (9UL << S_MAXDEPTH)
49 54
50#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
51#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
52 57
53#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)
54 61
55#define SB do { 62#define SB do {
56#define SE } while (0) 63#define SE } while (0)
57 64
58#if __GNUC__ >= 3 65#if __GNUC__ >= 3
68 75
69#define IN_RANGE_INC(type,val,beg,end) \ 76#define IN_RANGE_INC(type,val,beg,end) \
70 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 77 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
71 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
72 79
80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
81
73#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
74# define JSON_SLOW 1
75# 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))
76#else 86#else
77# define JSON_SLOW 0
78# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
79#endif 90#endif
80 91
81static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
82static 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;
83 98
84enum { 99enum {
85 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
86 INCR_M_STR, // inside string 103 INCR_M_STR, // inside string
87 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
88 INCR_M_JSON // outside anything, count nesting 107 INCR_M_JSON // outside anything, count nesting
89}; 108};
90 109
91#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)
92 111
93typedef struct { 112typedef struct {
94 U32 flags; 113 U32 flags;
114 U32 max_depth;
115 STRLEN max_size;
116
95 SV *cb_object; 117 SV *cb_object;
96 HV *cb_sk_object; 118 HV *cb_sk_object;
97 119
98 // for the incremental parser 120 // for the incremental parser
99 SV *incr_text; // the source text so far 121 SV *incr_text; // the source text so far
100 STRLEN incr_pos; // the current offset into the text 122 STRLEN incr_pos; // the current offset into the text
101 int incr_nest; // {[]}-nesting level 123 int incr_nest; // {[]}-nesting level
102 int incr_mode; 124 unsigned char incr_mode;
125
126 SV *v_false, *v_true;
103} 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}
104 136
105///////////////////////////////////////////////////////////////////////////// 137/////////////////////////////////////////////////////////////////////////////
106// utility functions 138// utility functions
107 139
108INLINE SV * 140INLINE SV *
118 150
119INLINE void 151INLINE void
120shrink (SV *sv) 152shrink (SV *sv)
121{ 153{
122 sv_utf8_downgrade (sv, 1); 154 sv_utf8_downgrade (sv, 1);
155
123 if (SvLEN (sv) > SvCUR (sv) + 1) 156 if (SvLEN (sv) > SvCUR (sv) + 1)
124 { 157 {
125#ifdef SvPV_shrink_to_cur 158#ifdef SvPV_shrink_to_cur
126 SvPV_shrink_to_cur (sv); 159 SvPV_shrink_to_cur (sv);
127#elif defined (SvPV_renew) 160#elif defined (SvPV_renew)
128 SvPV_renew (sv, SvCUR (sv) + 1); 161 SvPV_renew (sv, SvCUR (sv) + 1);
129#endif 162#endif
130 } 163 }
131} 164}
132 165
166/* adds two STRLENs together, slow, and with paranoia */
167static STRLEN
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
133// 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
134// case of an error. 192// case of an error.
135// we special-case "safe" characters from U+80 .. U+7FF, 193// we special-case "safe" characters from U+80 .. U+7FF,
136// but use the very good perl function to parse anything else. 194// but use the very good perl function to parse anything else.
137// note that we never call this function for a ascii codepoints 195// note that we never call this function for a ascii codepoints
138INLINE UV 196INLINE UV
171 *s++ = 0x80 | ( ch & 0x3f); 229 *s++ = 0x80 | ( ch & 0x3f);
172 230
173 return s; 231 return s;
174} 232}
175 233
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
176///////////////////////////////////////////////////////////////////////////// 381/////////////////////////////////////////////////////////////////////////////
177// encoder 382// encoder
178 383
179// structure used for encoding JSON 384// structure used for encoding JSON
180typedef struct 385typedef struct
182 char *cur; // SvPVX (sv) + current output position 387 char *cur; // SvPVX (sv) + current output position
183 char *end; // SvEND (sv) 388 char *end; // SvEND (sv)
184 SV *sv; // result scalar 389 SV *sv; // result scalar
185 JSON json; 390 JSON json;
186 U32 indent; // indentation level 391 U32 indent; // indentation level
187 U32 maxdepth; // max. indentation/recursion level
188 UV limit; // escape character values >= this value when encoding 392 UV limit; // escape character values >= this value when encoding
189} enc_t; 393} enc_t;
190 394
191INLINE void 395INLINE void
192need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
193{ 397{
194 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
195 { 399 {
196 STRLEN cur = enc->cur - SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
197 SvGROW (enc->sv, cur + len + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
198 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
199 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
200 } 404 }
201} 405}
202 406
203INLINE void 407INLINE void
204encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
220 424
221 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
222 { 426 {
223 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
224 { 428 {
225 need (enc, len += 1); 429 need (enc, len + 1);
226 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
227 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
228 } 432 }
229 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
230 { 434 {
231 need (enc, len += 1); 435 need (enc, len + 1);
232 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
233 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
234 } 438 }
235 else 439 else
236 *enc->cur++ = ch; 440 *enc->cur++ = ch;
239 } 443 }
240 else 444 else
241 { 445 {
242 switch (ch) 446 switch (ch)
243 { 447 {
244 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;
245 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;
246 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;
247 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;
248 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;
249 453
250 default: 454 default:
251 { 455 {
252 STRLEN clen; 456 STRLEN clen;
253 UV uch; 457 UV uch;
269 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
270 { 474 {
271 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
272 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);
273 477
274 need (enc, len += 11); 478 need (enc, len + 11);
275 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
276 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
277 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
278 enc->cur += 12; 482 enc->cur += 12;
279 } 483 }
280 else 484 else
281 { 485 {
282 static char hexdigit [16] = "0123456789abcdef";
283 need (enc, len += 5); 486 need (enc, len + 5);
284 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
285 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
286 *enc->cur++ = hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
287 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
288 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
289 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 492 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
290 } 493 }
291 494
292 str += clen; 495 str += clen;
293 } 496 }
294 else if (enc->json.flags & F_LATIN1) 497 else if (enc->json.flags & F_LATIN1)
296 *enc->cur++ = uch; 499 *enc->cur++ = uch;
297 str += clen; 500 str += clen;
298 } 501 }
299 else if (is_utf8) 502 else if (is_utf8)
300 { 503 {
301 need (enc, len += clen); 504 need (enc, len + clen);
302 do 505 do
303 { 506 {
304 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
305 } 508 }
306 while (--clen); 509 while (--clen);
307 } 510 }
308 else 511 else
309 { 512 {
310 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
311 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
312 ++str; 515 ++str;
313 } 516 }
314 } 517 }
315 } 518 }
365static void 568static void
366encode_av (enc_t *enc, AV *av) 569encode_av (enc_t *enc, AV *av)
367{ 570{
368 int i, len = av_len (av); 571 int i, len = av_len (av);
369 572
370 if (enc->indent >= enc->maxdepth) 573 if (enc->indent >= enc->json.max_depth)
371 croak ("data structure too deep (hit recursion limit)"); 574 croak (ERR_NESTING_EXCEEDED);
372 575
373 encode_ch (enc, '['); 576 encode_ch (enc, '[');
374 577
375 if (len >= 0) 578 if (len >= 0)
376 { 579 {
377 encode_nl (enc); ++enc->indent; 580 encode_nl (enc); ++enc->indent;
378 581
379 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
391 encode_comma (enc); 594 encode_comma (enc);
392 } 595 }
393 596
394 encode_nl (enc); --enc->indent; encode_indent (enc); 597 encode_nl (enc); --enc->indent; encode_indent (enc);
395 } 598 }
396 599
397 encode_ch (enc, ']'); 600 encode_ch (enc, ']');
398} 601}
399 602
400static void 603static void
401encode_hk (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
405 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
406 { 609 {
407 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
408 STRLEN len; 611 STRLEN len;
409 char *str; 612 char *str;
410 613
411 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
412 str = SvPV (sv, len); 615 str = SvPV (sv, len);
413 616
414 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
415 } 618 }
451static void 654static void
452encode_hv (enc_t *enc, HV *hv) 655encode_hv (enc_t *enc, HV *hv)
453{ 656{
454 HE *he; 657 HE *he;
455 658
456 if (enc->indent >= enc->maxdepth) 659 if (enc->indent >= enc->json.max_depth)
457 croak ("data structure too deep (hit recursion limit)"); 660 croak (ERR_NESTING_EXCEEDED);
458 661
459 encode_ch (enc, '{'); 662 encode_ch (enc, '{');
460 663
461 // for canonical output we have to sort by keys first 664 // for canonical output we have to sort by keys first
462 // actually, this is mostly due to the stupid so-called 665 // actually, this is mostly due to the stupid so-called
463 // security workaround added somewhere in 5.8.x. 666 // security workaround added somewhere in 5.8.x
464 // that randomises hash orderings 667 // that randomises hash orderings
465 if (enc->json.flags & F_CANONICAL) 668 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
466 { 669 {
467 int count = hv_iterinit (hv); 670 int count = hv_iterinit (hv);
468 671
469 if (SvMAGICAL (hv)) 672 if (SvMAGICAL (hv))
470 { 673 {
481 } 684 }
482 685
483 if (count) 686 if (count)
484 { 687 {
485 int i, fast = 1; 688 int i, fast = 1;
486#if defined(__BORLANDC__) || defined(_MSC_VER) 689 HE *hes_stack [STACK_HES];
487 HE **hes = _alloca (count * sizeof (HE)); 690 HE **hes = hes_stack;
488#else 691
489 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 692 // allocate larger arrays on the heap
490#endif 693 if (count > STACK_HES)
694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
491 698
492 i = 0; 699 i = 0;
493 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
494 { 701 {
495 hes [i++] = he; 702 hes [i++] = he;
564// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
565static void 772static void
566encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
567{ 774{
568 svtype svt; 775 svtype svt;
776 GV *method;
569 777
570 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
571 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
572 780
573 if (expect_false (SvOBJECT (sv))) 781 if (expect_false (SvOBJECT (sv)))
574 { 782 {
575 HV *stash = !JSON_SLOW || json_boolean_stash 783 HV *stash = SvSTASH (sv);
576 ? json_boolean_stash
577 : gv_stashpv ("JSON::XS::Boolean", 1);
578 784
579 if (SvSTASH (sv) == stash) 785 if (stash == bool_stash)
580 { 786 {
581 if (SvIV (sv))
582 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
583 else
584 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
585 } 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 PUTBACK;
832
833 encode_ch (enc, ']');
834
835 FREETMPS; LEAVE;
836 }
837 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
838 {
839 dSP;
840
841 ENTER; SAVETMPS;
842 PUSHMARK (SP);
843 // we re-bless the reference to get overload and other niceties right
844 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
845
846 // calling with G_SCALAR ensures that we always get a 1 return value
847 PUTBACK;
848 call_sv ((SV *)GvCV (method), G_SCALAR);
849 SPAGAIN;
850
851 // catch this surprisingly common error
852 if (SvROK (TOPs) && SvRV (TOPs) == sv)
853 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
854
855 sv = POPs;
856 PUTBACK;
857
858 encode_sv (enc, sv);
859
860 FREETMPS; LEAVE;
861 }
862 else if (enc->json.flags & F_ALLOW_BLESSED)
863 encode_str (enc, "null", 4, 0);
586 else 864 else
587 {
588#if 0
589 if (0 && sv_derived_from (rv, "JSON::Literal"))
590 {
591 // not yet
592 }
593#endif
594 if (enc->json.flags & F_CONV_BLESSED)
595 {
596 // we re-bless the reference to get overload and other niceties right
597 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
598
599 if (to_json)
600 {
601 dSP;
602
603 ENTER; SAVETMPS; PUSHMARK (SP);
604 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
605
606 // calling with G_SCALAR ensures that we always get a 1 return value
607 PUTBACK;
608 call_sv ((SV *)GvCV (to_json), G_SCALAR);
609 SPAGAIN;
610
611 // catch this surprisingly common error
612 if (SvROK (TOPs) && SvRV (TOPs) == sv)
613 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
614
615 sv = POPs;
616 PUTBACK;
617
618 encode_sv (enc, sv);
619
620 FREETMPS; LEAVE;
621 }
622 else if (enc->json.flags & F_ALLOW_BLESSED)
623 encode_str (enc, "null", 4, 0);
624 else
625 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
626 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
627 }
628 else if (enc->json.flags & F_ALLOW_BLESSED)
629 encode_str (enc, "null", 4, 0);
630 else
631 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 865 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
632 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 866 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
633 }
634 } 867 }
635 else if (svt == SVt_PVHV) 868 else if (svt == SVt_PVHV)
636 encode_hv (enc, (HV *)sv); 869 encode_hv (enc, (HV *)sv);
637 else if (svt == SVt_PVAV) 870 else if (svt == SVt_PVAV)
638 encode_av (enc, (AV *)sv); 871 encode_av (enc, (AV *)sv);
639 else if (svt < SVt_PVAV) 872 else if (svt < SVt_PVAV)
640 { 873 {
641 STRLEN len = 0; 874 int bool_type = ref_bool_type (sv);
642 char *pv = svt ? SvPV (sv, len) : 0;
643 875
644 if (len == 1 && *pv == '1') 876 if (bool_type == 1)
645 encode_str (enc, "true", 4, 0); 877 encode_str (enc, "true", 4, 0);
646 else if (len == 1 && *pv == '0') 878 else if (bool_type == 0)
647 encode_str (enc, "false", 5, 0); 879 encode_str (enc, "false", 5, 0);
648 else if (enc->json.flags & F_ALLOW_UNKNOWN) 880 else if (enc->json.flags & F_ALLOW_UNKNOWN)
649 encode_str (enc, "null", 4, 0); 881 encode_str (enc, "null", 4, 0);
650 else 882 else
651 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 883 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
686 if (SvIsUV (sv) ? SvUVX (sv) <= 59000 918 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
687 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) 919 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
688 { 920 {
689 // optimise the "small number case" 921 // optimise the "small number case"
690 // code will likely be branchless and use only a single multiplication 922 // code will likely be branchless and use only a single multiplication
691 // works for numbers up to 59074 923 // 4.28 works for numbers up to 59074
692 I32 i = SvIVX (sv); 924 I32 i = SvIVX (sv);
693 U32 u; 925 U32 u;
694 char digit, nz = 0; 926 char digit, nz = 0;
695 927
696 need (enc, 6); 928 need (enc, 6);
714 } 946 }
715 else 947 else
716 { 948 {
717 // large integer, use the (rather slow) snprintf way. 949 // large integer, use the (rather slow) snprintf way.
718 need (enc, IVUV_MAXCHARS); 950 need (enc, IVUV_MAXCHARS);
719 enc->cur += 951 enc->cur +=
720 SvIsUV(sv) 952 SvIsUV(sv)
721 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 953 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
722 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 954 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
723 } 955 }
724 } 956 }
725 else if (SvROK (sv)) 957 else if (SvROK (sv))
726 encode_rv (enc, SvRV (sv)); 958 encode_rv (enc, SvRV (sv));
727 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 959 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
728 encode_str (enc, "null", 4, 0); 960 encode_str (enc, "null", 4, 0);
729 else 961 else
730 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 962 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
731 SvPV_nolen (sv), SvFLAGS (sv)); 963 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
732} 964}
733 965
734static SV * 966static SV *
735encode_json (SV *scalar, JSON *json) 967encode_json (SV *scalar, JSON *json)
736{ 968{
737 enc_t enc; 969 enc_t enc;
738 970
739 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 971 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
740 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 972 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
741 973
742 enc.json = *json; 974 enc.json = *json;
743 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 975 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
744 enc.cur = SvPVX (enc.sv); 976 enc.cur = SvPVX (enc.sv);
745 enc.end = SvEND (enc.sv); 977 enc.end = SvEND (enc.sv);
746 enc.indent = 0; 978 enc.indent = 0;
747 enc.maxdepth = DEC_DEPTH (enc.json.flags);
748 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 979 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
749 : enc.json.flags & F_LATIN1 ? 0x000100UL 980 : enc.json.flags & F_LATIN1 ? 0x000100UL
750 : 0x110000UL; 981 : 0x110000UL;
751 982
752 SvPOK_only (enc.sv); 983 SvPOK_only (enc.sv);
753 encode_sv (&enc, scalar); 984 encode_sv (&enc, scalar);
985 encode_nl (&enc);
754 986
755 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 987 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
756 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 988 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
757 989
758 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 990 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
806 else 1038 else
807 break; 1039 break;
808 } 1040 }
809 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) 1041 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
810 break; // parse error, but let higher level handle it, gives better error messages 1042 break; // parse error, but let higher level handle it, gives better error messages
811 1043 else
812 ++dec->cur; 1044 ++dec->cur;
813 } 1045 }
814} 1046}
815 1047
816#define ERR(reason) SB dec->err = reason; goto fail; SE 1048#define ERR(reason) SB dec->err = reason; goto fail; SE
817 1049
819 if (*dec->cur != ch) \ 1051 if (*dec->cur != ch) \
820 ERR (# ch " expected"); \ 1052 ERR (# ch " expected"); \
821 ++dec->cur; \ 1053 ++dec->cur; \
822 SE 1054 SE
823 1055
824#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1056#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
825#define DEC_DEC_DEPTH --dec->depth 1057#define DEC_DEC_DEPTH --dec->depth
826 1058
827static SV *decode_sv (dec_t *dec); 1059static SV *decode_sv (dec_t *dec);
828 1060
829static signed char decode_hexdigit[256]; 1061static signed char decode_hexdigit[256];
938 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1170 else if (expect_true (ch >= 0x20 && ch < 0x80))
939 *cur++ = ch; 1171 *cur++ = ch;
940 else if (ch >= 0x80) 1172 else if (ch >= 0x80)
941 { 1173 {
942 STRLEN clen; 1174 STRLEN clen;
943 UV uch;
944 1175
945 --dec_cur; 1176 --dec_cur;
946 1177
947 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1178 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
948 if (clen == (STRLEN)-1) 1179 if (clen == (STRLEN)-1)
949 ERR ("malformed UTF-8 character in JSON string"); 1180 ERR ("malformed UTF-8 character in JSON string");
950 1181
951 do 1182 do
952 *cur++ = *dec_cur++; 1183 *cur++ = *dec_cur++;
953 while (--clen); 1184 while (--clen);
954 1185
955 utf8 = 1; 1186 utf8 = 1;
956 } 1187 }
1188 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1189 *cur++ = ch;
957 else 1190 else
958 { 1191 {
959 --dec_cur; 1192 --dec_cur;
960 1193
961 if (!ch) 1194 if (!ch)
969 { 1202 {
970 STRLEN len = cur - buf; 1203 STRLEN len = cur - buf;
971 1204
972 if (sv) 1205 if (sv)
973 { 1206 {
974 SvGROW (sv, SvCUR (sv) + len + 1); 1207 STRLEN cur = SvCUR (sv);
1208
1209 if (SvLEN (sv) - cur <= len)
1210 json_sv_grow (sv, cur, len);
1211
975 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1212 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
976 SvCUR_set (sv, SvCUR (sv) + len); 1213 SvCUR_set (sv, SvCUR (sv) + len);
977 } 1214 }
978 else 1215 else
979 sv = newSVpvn (buf, len); 1216 sv = newSVpvn (buf, len);
1070 1307
1071 // special case the rather common 1..5-digit-int case 1308 // special case the rather common 1..5-digit-int case
1072 if (*start == '-') 1309 if (*start == '-')
1073 switch (len) 1310 switch (len)
1074 { 1311 {
1075 case 2: return newSViv (-( start [1] - '0' * 1)); 1312 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1076 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1313 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1077 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1314 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1078 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1315 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1079 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1316 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1080 } 1317 }
1081 else 1318 else
1082 switch (len) 1319 switch (len)
1083 { 1320 {
1084 case 1: return newSViv ( start [0] - '0' * 1); 1321 case 1: return newSViv ( start [0] - '0' * 1);
1085 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1322 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1086 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1323 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1087 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1324 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1088 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1325 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1089 } 1326 }
1090 1327
1091 { 1328 {
1092 UV uv; 1329 UV uv;
1093 int numtype = grok_number (start, len, &uv); 1330 int numtype = grok_number (start, len, &uv);
1102 } 1339 }
1103 1340
1104 len -= *start == '-' ? 1 : 0; 1341 len -= *start == '-' ? 1 : 0;
1105 1342
1106 // does not fit into IV or UV, try NV 1343 // does not fit into IV or UV, try NV
1107 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1344 if (len <= NV_DIG)
1108 #if defined (LDBL_DIG)
1109 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1110 #endif
1111 )
1112 // fits into NV without loss of precision 1345 // fits into NV without loss of precision
1113 return newSVnv (Atof (start)); 1346 return newSVnv (json_atof (start));
1114 1347
1115 // everything else fails, convert it to a string 1348 // everything else fails, convert it to a string
1116 return newSVpvn (start, dec->cur - start); 1349 return newSVpvn (start, dec->cur - start);
1117 } 1350 }
1118 1351
1119 // loss of precision here 1352 // loss of precision here
1120 return newSVnv (Atof (start)); 1353 return newSVnv (json_atof (start));
1121 1354
1122fail: 1355fail:
1123 return 0; 1356 return 0;
1124} 1357}
1125 1358
1149 if (*dec->cur == ']') 1382 if (*dec->cur == ']')
1150 { 1383 {
1151 ++dec->cur; 1384 ++dec->cur;
1152 break; 1385 break;
1153 } 1386 }
1154 1387
1155 if (*dec->cur != ',') 1388 if (*dec->cur != ',')
1156 ERR (", or ] expected while parsing array"); 1389 ERR (", or ] expected while parsing array");
1157 1390
1158 ++dec->cur; 1391 ++dec->cur;
1159 1392
1273 1506
1274 DEC_DEC_DEPTH; 1507 DEC_DEC_DEPTH;
1275 sv = newRV_noinc ((SV *)hv); 1508 sv = newRV_noinc ((SV *)hv);
1276 1509
1277 // check filter callbacks 1510 // check filter callbacks
1278 if (dec->json.flags & F_HOOK) 1511 if (expect_false (dec->json.flags & F_HOOK))
1279 { 1512 {
1280 if (dec->json.cb_sk_object && HvKEYS (hv) == 1) 1513 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1281 { 1514 {
1282 HE *cb, *he; 1515 HE *cb, *he;
1283 1516
1284 hv_iterinit (hv); 1517 hv_iterinit (hv);
1285 he = hv_iternext (hv); 1518 he = hv_iternext (hv);
1286 hv_iterinit (hv); 1519 hv_iterinit (hv);
1287 1520
1288 // the next line creates a mortal sv each time its called. 1521 // the next line creates a mortal sv each time it's called.
1289 // might want to optimise this for common cases. 1522 // might want to optimise this for common cases.
1290 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1523 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1291 1524
1292 if (cb) 1525 if (cb)
1293 { 1526 {
1294 dSP; 1527 dSP;
1295 int count; 1528 int count;
1296 1529
1297 ENTER; SAVETMPS; PUSHMARK (SP); 1530 ENTER; SAVETMPS;
1531 PUSHMARK (SP);
1298 XPUSHs (HeVAL (he)); 1532 XPUSHs (HeVAL (he));
1533 sv_2mortal (sv);
1299 1534
1300 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1535 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1301 1536
1302 if (count == 1) 1537 if (count == 1)
1303 { 1538 {
1304 sv = newSVsv (POPs); 1539 sv = newSVsv (POPs);
1540 PUTBACK;
1305 FREETMPS; LEAVE; 1541 FREETMPS; LEAVE;
1306 return sv; 1542 return sv;
1307 } 1543 }
1544 else if (count)
1545 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1546
1547 PUTBACK;
1548
1549 SvREFCNT_inc (sv);
1308 1550
1309 FREETMPS; LEAVE; 1551 FREETMPS; LEAVE;
1310 } 1552 }
1311 } 1553 }
1312 1554
1313 if (dec->json.cb_object) 1555 if (dec->json.cb_object)
1314 { 1556 {
1315 dSP; 1557 dSP;
1316 int count; 1558 int count;
1317 1559
1318 ENTER; SAVETMPS; PUSHMARK (SP); 1560 ENTER; SAVETMPS;
1561 PUSHMARK (SP);
1319 XPUSHs (sv_2mortal (sv)); 1562 XPUSHs (sv_2mortal (sv));
1320 1563
1321 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1564 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1322 1565
1323 if (count == 1) 1566 if (count == 1)
1324 {
1325 sv = newSVsv (POPs); 1567 sv = newSVsv (POPs);
1326 FREETMPS; LEAVE; 1568 else if (count == 0)
1327 return sv;
1328 }
1329
1330 SvREFCNT_inc (sv); 1569 SvREFCNT_inc (sv);
1570 else
1571 croak ("filter_json_object callbacks must not return more than one scalar");
1572
1573 PUTBACK;
1574
1331 FREETMPS; LEAVE; 1575 FREETMPS; LEAVE;
1332 } 1576 }
1333 } 1577 }
1334 1578
1335 return sv; 1579 return sv;
1339 DEC_DEC_DEPTH; 1583 DEC_DEC_DEPTH;
1340 return 0; 1584 return 0;
1341} 1585}
1342 1586
1343static SV * 1587static SV *
1588decode_tag (dec_t *dec)
1589{
1590 SV *tag = 0;
1591 SV *val = 0;
1592
1593 if (!(dec->json.flags & F_ALLOW_TAGS))
1594 ERR ("malformed JSON string, neither array, object, number, string or atom");
1595
1596 ++dec->cur;
1597
1598 decode_ws (dec);
1599
1600 tag = decode_sv (dec);
1601 if (!tag)
1602 goto fail;
1603
1604 if (!SvPOK (tag))
1605 ERR ("malformed JSON string, (tag) must be a string");
1606
1607 decode_ws (dec);
1608
1609 if (*dec->cur != ')')
1610 ERR (") expected after tag");
1611
1612 ++dec->cur;
1613
1614 decode_ws (dec);
1615
1616 val = decode_sv (dec);
1617 if (!val)
1618 goto fail;
1619
1620 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1621 ERR ("malformed JSON string, tag value must be an array");
1622
1623 {
1624 AV *av = (AV *)SvRV (val);
1625 int i, len = av_len (av) + 1;
1626 HV *stash = gv_stashsv (tag, 0);
1627 SV *sv;
1628
1629 if (!stash)
1630 ERR ("cannot decode perl-object (package does not exist)");
1631
1632 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1633
1634 if (!method)
1635 ERR ("cannot decode perl-object (package does not have a THAW method)");
1636
1637 dSP;
1638
1639 ENTER; SAVETMPS;
1640 PUSHMARK (SP);
1641 EXTEND (SP, len + 2);
1642 // we re-bless the reference to get overload and other niceties right
1643 PUSHs (tag);
1644 PUSHs (sv_json);
1645
1646 for (i = 0; i < len; ++i)
1647 PUSHs (*av_fetch (av, i, 1));
1648
1649 PUTBACK;
1650 call_sv ((SV *)GvCV (method), G_SCALAR);
1651 SPAGAIN;
1652
1653 SvREFCNT_dec (tag);
1654 SvREFCNT_dec (val);
1655 sv = SvREFCNT_inc (POPs);
1656
1657 PUTBACK;
1658
1659 FREETMPS; LEAVE;
1660
1661 return sv;
1662 }
1663
1664fail:
1665 SvREFCNT_dec (tag);
1666 SvREFCNT_dec (val);
1667 return 0;
1668}
1669
1670static SV *
1344decode_sv (dec_t *dec) 1671decode_sv (dec_t *dec)
1345{ 1672{
1346 // the beauty of JSON: you need exactly one character lookahead 1673 // the beauty of JSON: you need exactly one character lookahead
1347 // to parse everything. 1674 // to parse everything.
1348 switch (*dec->cur) 1675 switch (*dec->cur)
1349 { 1676 {
1350 case '"': ++dec->cur; return decode_str (dec); 1677 case '"': ++dec->cur; return decode_str (dec);
1351 case '[': ++dec->cur; return decode_av (dec); 1678 case '[': ++dec->cur; return decode_av (dec);
1352 case '{': ++dec->cur; return decode_hv (dec); 1679 case '{': ++dec->cur; return decode_hv (dec);
1680 case '(': return decode_tag (dec);
1353 1681
1354 case '-': 1682 case '-':
1355 case '0': case '1': case '2': case '3': case '4': 1683 case '0': case '1': case '2': case '3': case '4':
1356 case '5': case '6': case '7': case '8': case '9': 1684 case '5': case '6': case '7': case '8': case '9':
1357 return decode_num (dec); 1685 return decode_num (dec);
1358 1686
1687 case 'f':
1688 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1689 {
1690 dec->cur += 5;
1691
1692 if (expect_false (!dec->json.v_false))
1693 dec->json.v_false = GET_BOOL (false);
1694
1695 return newSVsv (dec->json.v_false);
1696 }
1697 else
1698 ERR ("'false' expected");
1699
1700 break;
1701
1359 case 't': 1702 case 't':
1360 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1703 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1361 { 1704 {
1362 dec->cur += 4; 1705 dec->cur += 4;
1363#if JSON_SLOW 1706
1364 json_true = get_bool ("JSON::XS::true"); 1707 if (expect_false (!dec->json.v_true))
1365#endif 1708 dec->json.v_true = GET_BOOL (true);
1709
1366 return newSVsv (json_true); 1710 return newSVsv (dec->json.v_true);
1367 } 1711 }
1368 else 1712 else
1369 ERR ("'true' expected"); 1713 ERR ("'true' expected");
1370
1371 break;
1372
1373 case 'f':
1374 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1375 {
1376 dec->cur += 5;
1377#if JSON_SLOW
1378 json_false = get_bool ("JSON::XS::false");
1379#endif
1380 return newSVsv (json_false);
1381 }
1382 else
1383 ERR ("'false' expected");
1384 1714
1385 break; 1715 break;
1386 1716
1387 case 'n': 1717 case 'n':
1388 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1718 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1394 ERR ("'null' expected"); 1724 ERR ("'null' expected");
1395 1725
1396 break; 1726 break;
1397 1727
1398 default: 1728 default:
1399 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1729 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1400 break; 1730 break;
1401 } 1731 }
1402 1732
1403fail: 1733fail:
1404 return 0; 1734 return 0;
1406 1736
1407static SV * 1737static SV *
1408decode_json (SV *string, JSON *json, STRLEN *offset_return) 1738decode_json (SV *string, JSON *json, STRLEN *offset_return)
1409{ 1739{
1410 dec_t dec; 1740 dec_t dec;
1411 STRLEN offset;
1412 SV *sv; 1741 SV *sv;
1413 1742
1743 /* work around bugs in 5.10 where manipulating magic values
1744 * makes perl ignore the magic in subsequent accesses.
1745 * also make a copy of non-PV values, to get them into a clean
1746 * state (SvPV should do that, but it's buggy, see below).
1747 *
1748 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1749 * as reported by Reini Urban.
1750 */
1414 SvGETMAGIC (string); 1751 /*SvGETMAGIC (string);*/
1752 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1753 string = sv_2mortal (newSVsv (string));
1754
1415 SvUPGRADE (string, SVt_PV); 1755 SvUPGRADE (string, SVt_PV);
1416 1756
1417 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1757 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1758 * assertion with -DDEBUGGING, although SvCUR is documented to
1759 * return the xpv_cur field which certainly exists after upgrading.
1760 * according to nicholas clark, calling SvPOK fixes this.
1761 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1762 * and hope for the best.
1763 * Damnit, SvPV_nolen still trips over yet another assertion. This
1764 * assertion business is seriously broken, try yet another workaround
1765 * for the broken -DDEBUGGING.
1766 */
1767 {
1768#ifdef DEBUGGING
1769 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1770#else
1771 STRLEN offset = SvCUR (string);
1772#endif
1773
1774 if (offset > json->max_size && json->max_size)
1418 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1775 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1419 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1776 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1777 }
1420 1778
1421 if (json->flags & F_UTF8) 1779 if (DECODE_WANTS_OCTETS (json))
1422 sv_utf8_downgrade (string, 0); 1780 sv_utf8_downgrade (string, 0);
1423 else 1781 else
1424 sv_utf8_upgrade (string); 1782 sv_utf8_upgrade (string);
1425 1783
1426 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1784 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1427 1785
1428 dec.json = *json; 1786 dec.json = *json;
1429 dec.cur = SvPVX (string); 1787 dec.cur = SvPVX (string);
1430 dec.end = SvEND (string); 1788 dec.end = SvEND (string);
1431 dec.err = 0; 1789 dec.err = 0;
1432 dec.depth = 0; 1790 dec.depth = 0;
1433 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1434 1791
1435 if (dec.json.cb_object || dec.json.cb_sk_object) 1792 if (dec.json.cb_object || dec.json.cb_sk_object)
1436 dec.json.flags |= F_HOOK; 1793 dec.json.flags |= F_HOOK;
1437 1794
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1795 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439 1796
1440 decode_ws (&dec); 1797 decode_ws (&dec);
1441 sv = decode_sv (&dec); 1798 sv = decode_sv (&dec);
1442 1799
1443 if (!(offset_return || !sv)) 1800 if (offset_return)
1801 *offset_return = dec.cur - SvPVX (string);
1802 else if (sv)
1444 { 1803 {
1445 // check for trailing garbage 1804 // check for trailing garbage
1446 decode_ws (&dec); 1805 decode_ws (&dec);
1447 1806
1448 if (*dec.cur) 1807 if (dec.cur != dec.end)
1449 { 1808 {
1450 dec.err = "garbage after JSON object"; 1809 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv); 1810 SvREFCNT_dec (sv);
1452 sv = 0; 1811 sv = 0;
1453 } 1812 }
1454 }
1455
1456 if (offset_return || !sv)
1457 {
1458 offset = dec.json.flags & F_UTF8
1459 ? dec.cur - SvPVX (string)
1460 : utf8_distance (dec.cur, SvPVX (string));
1461
1462 if (offset_return)
1463 *offset_return = offset;
1464 } 1813 }
1465 1814
1466 if (!sv) 1815 if (!sv)
1467 { 1816 {
1468 SV *uni = sv_newmortal (); 1817 SV *uni = sv_newmortal ();
1474 SAVEVPTR (PL_curcop); 1823 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop; 1824 PL_curcop = &cop;
1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1825 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE; 1826 LEAVE;
1478 1827
1479 croak ("%s, at character offset %d [\"%s\"]", 1828 croak ("%s, at character offset %d (before \"%s\")",
1480 dec.err, 1829 dec.err,
1481 (int)offset, 1830 (int)ptr_to_index (string, dec.cur),
1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1831 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1483 } 1832 }
1484 1833
1485 sv = sv_2mortal (sv); 1834 sv = sv_2mortal (sv);
1486 1835
1487 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1836 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1488 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1837 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1489 1838
1490 return sv; 1839 return sv;
1491} 1840}
1492 1841
1496static void 1845static void
1497incr_parse (JSON *self) 1846incr_parse (JSON *self)
1498{ 1847{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1848 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1500 1849
1850 // the state machine here is a bit convoluted and could be simplified a lot
1851 // but this would make it slower, so...
1852
1501 for (;;) 1853 for (;;)
1502 { 1854 {
1503 //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
1504 switch (self->incr_mode) 1855 switch (self->incr_mode)
1505 { 1856 {
1857 // reached end of a scalar, see if we are inside a nested structure or not
1858 end_of_scalar:
1859 self->incr_mode = INCR_M_JSON;
1860
1861 if (self->incr_nest) // end of a scalar inside array, object or tag
1862 goto incr_m_json;
1863 else // end of scalar outside structure, json text ends here
1864 goto interrupt;
1865
1506 // only used for intiial whitespace skipping 1866 // only used for initial whitespace skipping
1507 case INCR_M_WS: 1867 case INCR_M_WS:
1508 for (;;) 1868 for (;;)
1509 { 1869 {
1510 if (*p > 0x20) 1870 if (*p > 0x20)
1511 { 1871 {
1872 if (*p == '#')
1873 {
1874 self->incr_mode = INCR_M_C0;
1875 goto incr_m_c;
1876 }
1877 else
1878 {
1512 self->incr_mode = INCR_M_JSON; 1879 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json; 1880 goto incr_m_json;
1881 }
1514 } 1882 }
1515 else if (!*p) 1883 else if (!*p)
1516 goto interrupt; 1884 goto interrupt;
1517 1885
1518 ++p; 1886 ++p;
1524 goto interrupt; 1892 goto interrupt;
1525 1893
1526 ++p; 1894 ++p;
1527 self->incr_mode = INCR_M_STR; 1895 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str; 1896 goto incr_m_str;
1897
1898 // inside #-style comments
1899 case INCR_M_C0:
1900 case INCR_M_C1:
1901 incr_m_c:
1902 for (;;)
1903 {
1904 if (*p == '\n')
1905 {
1906 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1907 break;
1908 }
1909 else if (!*p)
1910 goto interrupt;
1911
1912 ++p;
1913 }
1914
1915 break;
1916
1917 // inside true/false/null
1918 case INCR_M_TFN:
1919 incr_m_tfn:
1920 for (;;)
1921 switch (*p++)
1922 {
1923 case 'r': case 'u': case 'e': // tRUE, falsE, nUll
1924 case 'a': case 'l': case 's': // fALSe, nuLL
1925 // allowed
1926 break;
1927
1928 default:
1929 --p;
1930 goto end_of_scalar;
1931 }
1932
1933 // inside a number
1934 case INCR_M_NUM:
1935 incr_m_num:
1936 for (;;)
1937 switch (*p++)
1938 {
1939 case 'e': case 'E': case '.': case '+':
1940 case '-':
1941 case '0': case '1': case '2': case '3': case '4':
1942 case '5': case '6': case '7': case '8': case '9':
1943 // allowed
1944 break;
1945
1946 default:
1947 --p;
1948 goto end_of_scalar;
1949 }
1529 1950
1530 // inside a string 1951 // inside a string
1531 case INCR_M_STR: 1952 case INCR_M_STR:
1532 incr_m_str: 1953 incr_m_str:
1533 for (;;) 1954 for (;;)
1534 { 1955 {
1535 if (*p == '"') 1956 if (*p == '"')
1536 { 1957 {
1537 ++p; 1958 ++p;
1538 self->incr_mode = INCR_M_JSON;
1539
1540 if (!self->incr_nest)
1541 goto interrupt;
1542
1543 goto incr_m_json; 1959 goto end_of_scalar;
1544 } 1960 }
1545 else if (*p == '\\') 1961 else if (*p == '\\')
1546 { 1962 {
1547 ++p; // "virtually" consumes character after \ 1963 ++p; // "virtually" consumes character after \
1548 1964
1578 --p; // do not eat the whitespace, let the next round do it 1994 --p; // do not eat the whitespace, let the next round do it
1579 goto interrupt; 1995 goto interrupt;
1580 } 1996 }
1581 break; 1997 break;
1582 1998
1999 // the following three blocks handle scalars. this makes the parser
2000 // more strict than required inside arrays or objects, and could
2001 // be moved to a special case on the toplevel (except strings)
2002 case 't':
2003 case 'f':
2004 case 'n':
2005 self->incr_mode = INCR_M_TFN;
2006 goto incr_m_tfn;
2007
2008 case '-':
2009 case '0': case '1': case '2': case '3': case '4':
2010 case '5': case '6': case '7': case '8': case '9':
2011 self->incr_mode = INCR_M_NUM;
2012 goto incr_m_num;
2013
1583 case '"': 2014 case '"':
1584 self->incr_mode = INCR_M_STR; 2015 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str; 2016 goto incr_m_str;
1586 2017
1587 case '[': 2018 case '[':
1588 case '{': 2019 case '{':
2020 case '(':
1589 ++self->incr_nest; 2021 if (++self->incr_nest > self->max_depth)
2022 croak (ERR_NESTING_EXCEEDED);
1590 break; 2023 break;
1591 2024
1592 case ']': 2025 case ']':
1593 case '}': 2026 case '}':
1594 if (!--self->incr_nest) 2027 if (--self->incr_nest <= 0)
1595 goto interrupt; 2028 goto interrupt;
2029 break;
2030
2031 case ')':
2032 --self->incr_nest;
2033 break;
2034
2035 case '#':
2036 self->incr_mode = INCR_M_C1;
2037 goto incr_m_c;
1596 } 2038 }
1597 } 2039 }
1598 } 2040 }
1599 2041
1600 modechange: 2042 modechange:
1601 ; 2043 ;
1602 } 2044 }
1603 2045
1604interrupt: 2046interrupt:
1605 self->incr_pos = p - SvPVX (self->incr_text); 2047 self->incr_pos = p - SvPVX (self->incr_text);
2048 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1606 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 2049 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1607} 2050}
1608 2051
1609///////////////////////////////////////////////////////////////////////////// 2052/////////////////////////////////////////////////////////////////////////////
1610// XS interface functions 2053// XS interface functions
1620 i >= '0' && i <= '9' ? i - '0' 2063 i >= '0' && i <= '9' ? i - '0'
1621 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2064 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1622 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2065 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1623 : -1; 2066 : -1;
1624 2067
1625 json_stash = gv_stashpv ("JSON::XS" , 1); 2068 json_stash = gv_stashpv ("JSON::XS" , 1);
1626 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2069 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1627
1628 json_true = get_bool ("JSON::XS::true");
1629 json_false = get_bool ("JSON::XS::false"); 2070 bool_false = get_bool ("Types::Serialiser::false");
2071 bool_true = get_bool ("Types::Serialiser::true");
2072
2073 sv_json = newSVpv ("JSON", 0);
2074 SvREADONLY_on (sv_json);
2075
2076 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1630} 2077}
1631 2078
1632PROTOTYPES: DISABLE 2079PROTOTYPES: DISABLE
1633 2080
1634void CLONE (...) 2081void CLONE (...)
1635 CODE: 2082 CODE:
2083 // as long as these writes are atomic, the race should not matter
2084 // as existing threads either already use 0, or use the old value,
2085 // which is sitll correct for the initial thread.
1636 json_stash = 0; 2086 json_stash = 0;
1637 json_boolean_stash = 0; 2087 bool_stash = 0;
2088 bool_false = 0;
2089 bool_true = 0;
1638 2090
1639void new (char *klass) 2091void new (char *klass)
1640 PPCODE: 2092 PPCODE:
1641{ 2093{
1642 SV *pv = NEWSV (0, sizeof (JSON)); 2094 SV *pv = NEWSV (0, sizeof (JSON));
1643 SvPOK_only (pv); 2095 SvPOK_only (pv);
1644 Zero (SvPVX (pv), 1, JSON); 2096 json_init ((JSON *)SvPVX (pv));
1645 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1646 XPUSHs (sv_2mortal (sv_bless ( 2097 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv), 2098 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2099 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1649 ))); 2100 )));
1650} 2101}
2102
2103void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2104 PPCODE:
2105 self->v_false = newSVsv (v_false);
2106 self->v_true = newSVsv (v_true);
2107 XPUSHs (ST (0));
2108
2109void get_boolean_values (JSON *self)
2110 PPCODE:
2111 if (self->v_false && self->v_true)
2112 {
2113 EXTEND (SP, 2);
2114 PUSHs (self->v_false);
2115 PUSHs (self->v_true);
2116 }
1651 2117
1652void ascii (JSON *self, int enable = 1) 2118void ascii (JSON *self, int enable = 1)
1653 ALIAS: 2119 ALIAS:
1654 ascii = F_ASCII 2120 ascii = F_ASCII
1655 latin1 = F_LATIN1 2121 latin1 = F_LATIN1
1663 shrink = F_SHRINK 2129 shrink = F_SHRINK
1664 allow_blessed = F_ALLOW_BLESSED 2130 allow_blessed = F_ALLOW_BLESSED
1665 convert_blessed = F_CONV_BLESSED 2131 convert_blessed = F_CONV_BLESSED
1666 relaxed = F_RELAXED 2132 relaxed = F_RELAXED
1667 allow_unknown = F_ALLOW_UNKNOWN 2133 allow_unknown = F_ALLOW_UNKNOWN
2134 allow_tags = F_ALLOW_TAGS
1668 PPCODE: 2135 PPCODE:
1669{ 2136{
1670 if (enable) 2137 if (enable)
1671 self->flags |= ix; 2138 self->flags |= ix;
1672 else 2139 else
1688 get_shrink = F_SHRINK 2155 get_shrink = F_SHRINK
1689 get_allow_blessed = F_ALLOW_BLESSED 2156 get_allow_blessed = F_ALLOW_BLESSED
1690 get_convert_blessed = F_CONV_BLESSED 2157 get_convert_blessed = F_CONV_BLESSED
1691 get_relaxed = F_RELAXED 2158 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN 2159 get_allow_unknown = F_ALLOW_UNKNOWN
2160 get_allow_tags = F_ALLOW_TAGS
1693 PPCODE: 2161 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix)); 2162 XPUSHs (boolSV (self->flags & ix));
1695 2163
1696void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2164void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1697 PPCODE: 2165 PPCODE:
1698{ 2166 self->max_depth = max_depth;
1699 UV log2 = 0;
1700
1701 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1702
1703 while ((1UL << log2) < max_depth)
1704 ++log2;
1705
1706 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1707
1708 XPUSHs (ST (0)); 2167 XPUSHs (ST (0));
1709}
1710 2168
1711U32 get_max_depth (JSON *self) 2169U32 get_max_depth (JSON *self)
1712 CODE: 2170 CODE:
1713 RETVAL = DEC_DEPTH (self->flags); 2171 RETVAL = self->max_depth;
1714 OUTPUT: 2172 OUTPUT:
1715 RETVAL 2173 RETVAL
1716 2174
1717void max_size (JSON *self, UV max_size = 0) 2175void max_size (JSON *self, U32 max_size = 0)
1718 PPCODE: 2176 PPCODE:
1719{ 2177 self->max_size = max_size;
1720 UV log2 = 0;
1721
1722 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1723 if (max_size == 1) max_size = 2;
1724
1725 while ((1UL << log2) < max_size)
1726 ++log2;
1727
1728 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1729
1730 XPUSHs (ST (0)); 2178 XPUSHs (ST (0));
1731}
1732 2179
1733int get_max_size (JSON *self) 2180int get_max_size (JSON *self)
1734 CODE: 2181 CODE:
1735 RETVAL = DEC_SIZE (self->flags); 2182 RETVAL = self->max_size;
1736 OUTPUT: 2183 OUTPUT:
1737 RETVAL 2184 RETVAL
1738 2185
1739void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2186void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1740 PPCODE: 2187 PPCODE:
1746} 2193}
1747 2194
1748void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2195void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1749 PPCODE: 2196 PPCODE:
1750{ 2197{
1751 if (!self->cb_sk_object) 2198 if (!self->cb_sk_object)
1752 self->cb_sk_object = newHV (); 2199 self->cb_sk_object = newHV ();
1753 2200
1754 if (SvOK (cb)) 2201 if (SvOK (cb))
1755 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2202 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1756 else 2203 else
1767 XPUSHs (ST (0)); 2214 XPUSHs (ST (0));
1768} 2215}
1769 2216
1770void encode (JSON *self, SV *scalar) 2217void encode (JSON *self, SV *scalar)
1771 PPCODE: 2218 PPCODE:
1772 XPUSHs (encode_json (scalar, self)); 2219 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2220 XPUSHs (scalar);
1773 2221
1774void decode (JSON *self, SV *jsonstr) 2222void decode (JSON *self, SV *jsonstr)
1775 PPCODE: 2223 PPCODE:
1776 XPUSHs (decode_json (jsonstr, self, 0)); 2224 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2225 XPUSHs (jsonstr);
1777 2226
1778void decode_prefix (JSON *self, SV *jsonstr) 2227void decode_prefix (JSON *self, SV *jsonstr)
1779 PPCODE: 2228 PPCODE:
1780{ 2229{
2230 SV *sv;
1781 STRLEN offset; 2231 STRLEN offset;
2232 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1782 EXTEND (SP, 2); 2233 EXTEND (SP, 2);
1783 PUSHs (decode_json (jsonstr, self, &offset)); 2234 PUSHs (sv);
1784 PUSHs (sv_2mortal (newSVuv (offset))); 2235 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1785} 2236}
1786 2237
1787void incr_parse (JSON *self, SV *jsonstr = 0) 2238void incr_parse (JSON *self, SV *jsonstr = 0)
1788 PPCODE: 2239 PPCODE:
1789{ 2240{
1790 if (!self->incr_text) 2241 if (!self->incr_text)
1791 self->incr_text = newSVpvn ("", 0); 2242 self->incr_text = newSVpvn ("", 0);
2243
2244 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2245 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2246 if (DECODE_WANTS_OCTETS (self))
2247 {
2248 if (self->incr_pos)
2249 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2250 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2251
2252 sv_utf8_downgrade (self->incr_text, 0);
2253 }
2254 else
2255 {
2256 sv_utf8_upgrade (self->incr_text);
2257
2258 if (self->incr_pos)
2259 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2260 - (U8 *)SvPVX (self->incr_text);
2261 }
1792 2262
1793 // append data, if any 2263 // append data, if any
1794 if (jsonstr) 2264 if (jsonstr)
1795 { 2265 {
2266 /* make sure both strings have same encoding */
1796 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2267 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2268 if (SvUTF8 (jsonstr))
2269 sv_utf8_downgrade (jsonstr, 0);
1797 { 2270 else
1798 /* utf-8-ness differs, need to upgrade */
1799 sv_utf8_upgrade (self->incr_text); 2271 sv_utf8_upgrade (jsonstr);
1800 2272
1801 if (self->incr_pos) 2273 /* and then just blindly append */
1802 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1803 - (U8 *)SvPVX (self->incr_text);
1804 }
1805
1806 { 2274 {
1807 STRLEN len; 2275 STRLEN len;
1808 const char *str = SvPV (jsonstr, len); 2276 const char *str = SvPV (jsonstr, len);
1809 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2277 STRLEN cur = SvCUR (self->incr_text);
2278
2279 if (SvLEN (self->incr_text) - cur <= len)
2280 json_sv_grow (self->incr_text, cur, len);
2281
1810 Move (str, SvEND (self->incr_text), len, char); 2282 Move (str, SvEND (self->incr_text), len, char);
1811 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2283 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1812 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2284 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1813 } 2285 }
1814 } 2286 }
1815 2287
1816 if (GIMME_V != G_VOID) 2288 if (GIMME_V != G_VOID)
1817 do 2289 do
1818 { 2290 {
2291 SV *sv;
1819 STRLEN offset; 2292 STRLEN offset;
1820 2293
1821 if (!INCR_DONE (self)) 2294 if (!INCR_DONE (self))
1822 { 2295 {
1823 incr_parse (self); 2296 incr_parse (self);
2297
2298 if (self->incr_pos > self->max_size && self->max_size)
2299 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2300 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2301
1824 if (!INCR_DONE (self)) 2302 if (!INCR_DONE (self))
2303 {
2304 // as an optimisation, do not accumulate white space in the incr buffer
2305 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2306 {
2307 self->incr_pos = 0;
2308 SvCUR_set (self->incr_text, 0);
2309 }
2310
1825 break; 2311 break;
2312 }
1826 } 2313 }
1827 2314
1828 XPUSHs (decode_json (self->incr_text, self, &offset)); 2315 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2316 XPUSHs (sv);
1829 2317
1830 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1831 self->incr_pos -= offset; 2318 self->incr_pos -= offset;
1832 self->incr_nest = 0; 2319 self->incr_nest = 0;
1833 self->incr_mode = 0; 2320 self->incr_mode = 0;
2321
2322 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1834 } 2323 }
1835 while (GIMME_V == G_ARRAY); 2324 while (GIMME_V == G_ARRAY);
1836} 2325}
1837 2326
1838SV *incr_text (JSON *self) 2327SV *incr_text (JSON *self)
1857 self->incr_nest = 0; 2346 self->incr_nest = 0;
1858 self->incr_mode = 0; 2347 self->incr_mode = 0;
1859 } 2348 }
1860} 2349}
1861 2350
2351void incr_reset (JSON *self)
2352 CODE:
2353{
2354 SvREFCNT_dec (self->incr_text);
2355 self->incr_text = 0;
2356 self->incr_pos = 0;
2357 self->incr_nest = 0;
2358 self->incr_mode = 0;
2359}
2360
1862void DESTROY (JSON *self) 2361void DESTROY (JSON *self)
1863 CODE: 2362 CODE:
2363 SvREFCNT_dec (self->v_false);
2364 SvREFCNT_dec (self->v_true);
1864 SvREFCNT_dec (self->cb_sk_object); 2365 SvREFCNT_dec (self->cb_sk_object);
1865 SvREFCNT_dec (self->cb_object); 2366 SvREFCNT_dec (self->cb_object);
1866 SvREFCNT_dec (self->incr_text); 2367 SvREFCNT_dec (self->incr_text);
1867 2368
1868PROTOTYPES: ENABLE 2369PROTOTYPES: ENABLE
1869 2370
1870void encode_json (SV *scalar) 2371void encode_json (SV *scalar)
1871 ALIAS:
1872 to_json_ = 0
1873 encode_json = F_UTF8
1874 PPCODE: 2372 PPCODE:
1875{ 2373{
1876 JSON json = { F_DEFAULT | ix }; 2374 JSON json;
1877 XPUSHs (encode_json (scalar, &json)); 2375 json_init (&json);
2376 json.flags |= F_UTF8;
2377 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2378 XPUSHs (scalar);
1878} 2379}
1879 2380
1880void decode_json (SV *jsonstr) 2381void decode_json (SV *jsonstr)
1881 ALIAS:
1882 from_json_ = 0
1883 decode_json = F_UTF8
1884 PPCODE: 2382 PPCODE:
1885{ 2383{
1886 JSON json = { F_DEFAULT | ix }; 2384 JSON json;
1887 XPUSHs (decode_json (jsonstr, &json, 0)); 2385 json_init (&json);
2386 json.flags |= F_UTF8;
2387 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2388 XPUSHs (jsonstr);
1888} 2389}
1889 2390
1890

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines