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 1.99 by root, Sat Aug 8 10:06:02 2009 UTC vs.
Revision 1.138 by root, Wed Mar 6 07:21:17 2019 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
17// guarantees, though. if it breaks, you get to keep the pieces. 18// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
20#endif 21#endif
21 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 36
24#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
32#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
36#define F_ALLOW_UNKNOWN 0x00002000UL 49#define F_ALLOW_UNKNOWN 0x00002000UL
50#define F_ALLOW_TAGS 0x00004000UL
37#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
38 52
39#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
40 54
41#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
42#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
43 57
44#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
59
60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
45 61
46#define SB do { 62#define SB do {
47#define SE } while (0) 63#define SE } while (0)
48 64
49#if __GNUC__ >= 3 65#if __GNUC__ >= 3
62 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
63 79
64#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" 80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
65 81
66#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
67# define JSON_SLOW 1
68# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
84# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
85# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
69#else 86#else
70# define JSON_SLOW 0
71# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
72#endif 90#endif
73 91
74static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
75static SV *json_true, *json_false; 93#define STACK_HES 64
94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_false, *bool_true;
97static SV *sv_json;
76 98
77enum { 99enum {
78 INCR_M_WS = 0, // initial whitespace skipping, must be 0 100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
101 INCR_M_TFN, // inside true/false/null
102 INCR_M_NUM, // inside number
79 INCR_M_STR, // inside string 103 INCR_M_STR, // inside string
80 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
81 INCR_M_JSON // outside anything, count nesting 107 INCR_M_JSON // outside anything, count nesting
82}; 108};
83 109
84#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) 110#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
85 111
94 // for the incremental parser 120 // for the incremental parser
95 SV *incr_text; // the source text so far 121 SV *incr_text; // the source text so far
96 STRLEN incr_pos; // the current offset into the text 122 STRLEN incr_pos; // the current offset into the text
97 int incr_nest; // {[]}-nesting level 123 int incr_nest; // {[]}-nesting level
98 unsigned char incr_mode; 124 unsigned char incr_mode;
125
126 SV *v_false, *v_true;
99} JSON; 127} JSON;
100 128
101INLINE void 129INLINE void
102json_init (JSON *json) 130json_init (JSON *json)
103{ 131{
104 Zero (json, 1, JSON); 132 static const JSON init = { F_ALLOW_NONREF, 512 };
105 json->max_depth = 512; 133
134 *json = init;
106} 135}
107 136
108///////////////////////////////////////////////////////////////////////////// 137/////////////////////////////////////////////////////////////////////////////
109// utility functions 138// utility functions
110 139
132 SvPV_renew (sv, SvCUR (sv) + 1); 161 SvPV_renew (sv, SvCUR (sv) + 1);
133#endif 162#endif
134 } 163 }
135} 164}
136 165
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
137// 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
138// case of an error. 192// case of an error.
139// we special-case "safe" characters from U+80 .. U+7FF, 193// we special-case "safe" characters from U+80 .. U+7FF,
140// but use the very good perl function to parse anything else. 194// but use the very good perl function to parse anything else.
141// note that we never call this function for a ascii codepoints 195// note that we never call this function for a ascii codepoints
142INLINE UV 196INLINE UV
185 ? utf8_distance (offset, SvPVX (sv)) 239 ? utf8_distance (offset, SvPVX (sv))
186 : offset - SvPVX (sv); 240 : offset - SvPVX (sv);
187} 241}
188 242
189///////////////////////////////////////////////////////////////////////////// 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
381/////////////////////////////////////////////////////////////////////////////
190// encoder 382// encoder
191 383
192// structure used for encoding JSON 384// structure used for encoding JSON
193typedef struct 385typedef struct
194{ 386{
201} enc_t; 393} enc_t;
202 394
203INLINE void 395INLINE void
204need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
205{ 397{
206 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
207 { 399 {
208 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
209 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
210 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
211 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
212 } 404 }
213} 405}
214 406
215INLINE void 407INLINE void
216encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
232 424
233 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
234 { 426 {
235 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
236 { 428 {
237 need (enc, len += 1); 429 need (enc, len + 1);
238 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
239 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
240 } 432 }
241 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
242 { 434 {
243 need (enc, len += 1); 435 need (enc, len + 1);
244 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
245 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
246 } 438 }
247 else 439 else
248 *enc->cur++ = ch; 440 *enc->cur++ = ch;
251 } 443 }
252 else 444 else
253 { 445 {
254 switch (ch) 446 switch (ch)
255 { 447 {
256 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;
257 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;
258 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;
259 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;
260 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;
261 453
262 default: 454 default:
263 { 455 {
264 STRLEN clen; 456 STRLEN clen;
265 UV uch; 457 UV uch;
281 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
282 { 474 {
283 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
284 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);
285 477
286 need (enc, len += 11); 478 need (enc, len + 11);
287 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
288 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
289 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
290 enc->cur += 12; 482 enc->cur += 12;
291 } 483 }
292 else 484 else
293 { 485 {
294 need (enc, len += 5); 486 need (enc, len + 5);
295 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
296 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
297 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
298 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
299 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
307 *enc->cur++ = uch; 499 *enc->cur++ = uch;
308 str += clen; 500 str += clen;
309 } 501 }
310 else if (is_utf8) 502 else if (is_utf8)
311 { 503 {
312 need (enc, len += clen); 504 need (enc, len + clen);
313 do 505 do
314 { 506 {
315 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
316 } 508 }
317 while (--clen); 509 while (--clen);
318 } 510 }
319 else 511 else
320 { 512 {
321 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
322 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
323 ++str; 515 ++str;
324 } 516 }
325 } 517 }
326 } 518 }
380 572
381 if (enc->indent >= enc->json.max_depth) 573 if (enc->indent >= enc->json.max_depth)
382 croak (ERR_NESTING_EXCEEDED); 574 croak (ERR_NESTING_EXCEEDED);
383 575
384 encode_ch (enc, '['); 576 encode_ch (enc, '[');
385 577
386 if (len >= 0) 578 if (len >= 0)
387 { 579 {
388 encode_nl (enc); ++enc->indent; 580 encode_nl (enc); ++enc->indent;
389 581
390 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
402 encode_comma (enc); 594 encode_comma (enc);
403 } 595 }
404 596
405 encode_nl (enc); --enc->indent; encode_indent (enc); 597 encode_nl (enc); --enc->indent; encode_indent (enc);
406 } 598 }
407 599
408 encode_ch (enc, ']'); 600 encode_ch (enc, ']');
409} 601}
410 602
411static void 603static void
412encode_hk (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
416 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
417 { 609 {
418 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
419 STRLEN len; 611 STRLEN len;
420 char *str; 612 char *str;
421 613
422 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
423 str = SvPV (sv, len); 615 str = SvPV (sv, len);
424 616
425 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
426 } 618 }
492 } 684 }
493 685
494 if (count) 686 if (count)
495 { 687 {
496 int i, fast = 1; 688 int i, fast = 1;
497#if defined(__BORLANDC__) || defined(_MSC_VER) 689 HE *hes_stack [STACK_HES];
498 HE **hes = _alloca (count * sizeof (HE)); 690 HE **hes = hes_stack;
499#else 691
500 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 692 // allocate larger arrays on the heap
501#endif 693 if (count > STACK_HES)
694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
502 698
503 i = 0; 699 i = 0;
504 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
505 { 701 {
506 hes [i++] = he; 702 hes [i++] = he;
575// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
576static void 772static void
577encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
578{ 774{
579 svtype svt; 775 svtype svt;
776 GV *method;
580 777
581 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
582 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
583 780
584 if (expect_false (SvOBJECT (sv))) 781 if (expect_false (SvOBJECT (sv)))
585 { 782 {
586 HV *stash = !JSON_SLOW || json_boolean_stash 783 HV *stash = SvSTASH (sv);
587 ? json_boolean_stash
588 : gv_stashpv ("JSON::XS::Boolean", 1);
589 784
590 if (SvSTASH (sv) == stash) 785 if (stash == bool_stash)
591 { 786 {
592 if (SvIV (sv))
593 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
594 else
595 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
596 } 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);
597 else 862 else
598 {
599#if 0
600 if (0 && sv_derived_from (rv, "JSON::Literal"))
601 {
602 // not yet
603 }
604#endif
605 if (enc->json.flags & F_CONV_BLESSED)
606 {
607 // we re-bless the reference to get overload and other niceties right
608 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
609
610 if (to_json)
611 {
612 dSP;
613
614 ENTER; SAVETMPS; PUSHMARK (SP);
615 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
616
617 // calling with G_SCALAR ensures that we always get a 1 return value
618 PUTBACK;
619 call_sv ((SV *)GvCV (to_json), G_SCALAR);
620 SPAGAIN;
621
622 // catch this surprisingly common error
623 if (SvROK (TOPs) && SvRV (TOPs) == sv)
624 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
625
626 sv = POPs;
627 PUTBACK;
628
629 encode_sv (enc, sv);
630
631 FREETMPS; LEAVE;
632 }
633 else if (enc->json.flags & F_ALLOW_BLESSED)
634 encode_str (enc, "null", 4, 0);
635 else
636 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
637 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
638 }
639 else if (enc->json.flags & F_ALLOW_BLESSED)
640 encode_str (enc, "null", 4, 0);
641 else
642 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 863 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
643 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 864 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
644 }
645 } 865 }
646 else if (svt == SVt_PVHV) 866 else if (svt == SVt_PVHV)
647 encode_hv (enc, (HV *)sv); 867 encode_hv (enc, (HV *)sv);
648 else if (svt == SVt_PVAV) 868 else if (svt == SVt_PVAV)
649 encode_av (enc, (AV *)sv); 869 encode_av (enc, (AV *)sv);
650 else if (svt < SVt_PVAV) 870 else if (svt < SVt_PVAV)
651 { 871 {
652 STRLEN len = 0; 872 int bool_type = ref_bool_type (sv);
653 char *pv = svt ? SvPV (sv, len) : 0;
654 873
655 if (len == 1 && *pv == '1') 874 if (bool_type == 1)
656 encode_str (enc, "true", 4, 0); 875 encode_str (enc, "true", 4, 0);
657 else if (len == 1 && *pv == '0') 876 else if (bool_type == 0)
658 encode_str (enc, "false", 5, 0); 877 encode_str (enc, "false", 5, 0);
659 else if (enc->json.flags & F_ALLOW_UNKNOWN) 878 else if (enc->json.flags & F_ALLOW_UNKNOWN)
660 encode_str (enc, "null", 4, 0); 879 encode_str (enc, "null", 4, 0);
661 else 880 else
662 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",
725 } 944 }
726 else 945 else
727 { 946 {
728 // large integer, use the (rather slow) snprintf way. 947 // large integer, use the (rather slow) snprintf way.
729 need (enc, IVUV_MAXCHARS); 948 need (enc, IVUV_MAXCHARS);
730 enc->cur += 949 enc->cur +=
731 SvIsUV(sv) 950 SvIsUV(sv)
732 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 951 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
733 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 952 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
734 } 953 }
735 } 954 }
736 else if (SvROK (sv)) 955 else if (SvROK (sv))
737 encode_rv (enc, SvRV (sv)); 956 encode_rv (enc, SvRV (sv));
738 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 957 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
739 encode_str (enc, "null", 4, 0); 958 encode_str (enc, "null", 4, 0);
740 else 959 else
741 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",
742 SvPV_nolen (sv), SvFLAGS (sv)); 961 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
743} 962}
744 963
745static SV * 964static SV *
746encode_json (SV *scalar, JSON *json) 965encode_json (SV *scalar, JSON *json)
747{ 966{
748 enc_t enc; 967 enc_t enc;
749 968
750 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 969 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
751 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 970 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
752 971
753 enc.json = *json; 972 enc.json = *json;
754 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 973 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
755 enc.cur = SvPVX (enc.sv); 974 enc.cur = SvPVX (enc.sv);
759 : enc.json.flags & F_LATIN1 ? 0x000100UL 978 : enc.json.flags & F_LATIN1 ? 0x000100UL
760 : 0x110000UL; 979 : 0x110000UL;
761 980
762 SvPOK_only (enc.sv); 981 SvPOK_only (enc.sv);
763 encode_sv (&enc, scalar); 982 encode_sv (&enc, scalar);
983 encode_nl (&enc);
764 984
765 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 985 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
766 *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
767 987
768 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 988 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
948 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1168 else if (expect_true (ch >= 0x20 && ch < 0x80))
949 *cur++ = ch; 1169 *cur++ = ch;
950 else if (ch >= 0x80) 1170 else if (ch >= 0x80)
951 { 1171 {
952 STRLEN clen; 1172 STRLEN clen;
953 UV uch;
954 1173
955 --dec_cur; 1174 --dec_cur;
956 1175
957 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1176 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
958 if (clen == (STRLEN)-1) 1177 if (clen == (STRLEN)-1)
959 ERR ("malformed UTF-8 character in JSON string"); 1178 ERR ("malformed UTF-8 character in JSON string");
960 1179
961 do 1180 do
962 *cur++ = *dec_cur++; 1181 *cur++ = *dec_cur++;
963 while (--clen); 1182 while (--clen);
964 1183
965 utf8 = 1; 1184 utf8 = 1;
966 } 1185 }
1186 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1187 *cur++ = ch;
967 else 1188 else
968 { 1189 {
969 --dec_cur; 1190 --dec_cur;
970 1191
971 if (!ch) 1192 if (!ch)
981 1202
982 if (sv) 1203 if (sv)
983 { 1204 {
984 STRLEN cur = SvCUR (sv); 1205 STRLEN cur = SvCUR (sv);
985 1206
986 if (SvLEN (sv) <= cur + len) 1207 if (SvLEN (sv) - cur <= len)
987 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1208 json_sv_grow (sv, cur, len);
988 1209
989 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1210 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
990 SvCUR_set (sv, SvCUR (sv) + len); 1211 SvCUR_set (sv, SvCUR (sv) + len);
991 } 1212 }
992 else 1213 else
1116 } 1337 }
1117 1338
1118 len -= *start == '-' ? 1 : 0; 1339 len -= *start == '-' ? 1 : 0;
1119 1340
1120 // does not fit into IV or UV, try NV 1341 // does not fit into IV or UV, try NV
1121 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1342 if (len <= NV_DIG)
1122 #if defined (LDBL_DIG)
1123 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1124 #endif
1125 )
1126 // fits into NV without loss of precision 1343 // fits into NV without loss of precision
1127 return newSVnv (Atof (start)); 1344 return newSVnv (json_atof (start));
1128 1345
1129 // everything else fails, convert it to a string 1346 // everything else fails, convert it to a string
1130 return newSVpvn (start, dec->cur - start); 1347 return newSVpvn (start, dec->cur - start);
1131 } 1348 }
1132 1349
1133 // loss of precision here 1350 // loss of precision here
1134 return newSVnv (Atof (start)); 1351 return newSVnv (json_atof (start));
1135 1352
1136fail: 1353fail:
1137 return 0; 1354 return 0;
1138} 1355}
1139 1356
1163 if (*dec->cur == ']') 1380 if (*dec->cur == ']')
1164 { 1381 {
1165 ++dec->cur; 1382 ++dec->cur;
1166 break; 1383 break;
1167 } 1384 }
1168 1385
1169 if (*dec->cur != ',') 1386 if (*dec->cur != ',')
1170 ERR (", or ] expected while parsing array"); 1387 ERR (", or ] expected while parsing array");
1171 1388
1172 ++dec->cur; 1389 ++dec->cur;
1173 1390
1287 1504
1288 DEC_DEC_DEPTH; 1505 DEC_DEC_DEPTH;
1289 sv = newRV_noinc ((SV *)hv); 1506 sv = newRV_noinc ((SV *)hv);
1290 1507
1291 // check filter callbacks 1508 // check filter callbacks
1292 if (dec->json.flags & F_HOOK) 1509 if (expect_false (dec->json.flags & F_HOOK))
1293 { 1510 {
1294 if (dec->json.cb_sk_object && HvKEYS (hv) == 1) 1511 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1295 { 1512 {
1296 HE *cb, *he; 1513 HE *cb, *he;
1297 1514
1298 hv_iterinit (hv); 1515 hv_iterinit (hv);
1299 he = hv_iternext (hv); 1516 he = hv_iternext (hv);
1300 hv_iterinit (hv); 1517 hv_iterinit (hv);
1301 1518
1302 // the next line creates a mortal sv each time its called. 1519 // the next line creates a mortal sv each time it's called.
1303 // might want to optimise this for common cases. 1520 // might want to optimise this for common cases.
1304 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);
1305 1522
1306 if (cb) 1523 if (cb)
1307 { 1524 {
1308 dSP; 1525 dSP;
1309 int count; 1526 int count;
1310 1527
1311 ENTER; SAVETMPS; PUSHMARK (SP); 1528 ENTER; SAVETMPS;
1529 PUSHMARK (SP);
1312 XPUSHs (HeVAL (he)); 1530 XPUSHs (HeVAL (he));
1531 sv_2mortal (sv);
1313 1532
1314 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1533 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1315 1534
1316 if (count == 1) 1535 if (count == 1)
1317 { 1536 {
1318 sv = newSVsv (POPs); 1537 sv = newSVsv (POPs);
1319 FREETMPS; LEAVE; 1538 FREETMPS; LEAVE;
1320 return sv; 1539 return sv;
1321 } 1540 }
1541 else if (count)
1542 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1322 1543
1544 SvREFCNT_inc (sv);
1323 FREETMPS; LEAVE; 1545 FREETMPS; LEAVE;
1324 } 1546 }
1325 } 1547 }
1326 1548
1327 if (dec->json.cb_object) 1549 if (dec->json.cb_object)
1328 { 1550 {
1329 dSP; 1551 dSP;
1330 int count; 1552 int count;
1331 1553
1332 ENTER; SAVETMPS; PUSHMARK (SP); 1554 ENTER; SAVETMPS;
1555 PUSHMARK (SP);
1333 XPUSHs (sv_2mortal (sv)); 1556 XPUSHs (sv_2mortal (sv));
1334 1557
1335 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1558 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1336 1559
1337 if (count == 1) 1560 if (count == 1)
1338 {
1339 sv = newSVsv (POPs); 1561 sv = newSVsv (POPs);
1340 FREETMPS; LEAVE; 1562 else if (count == 0)
1341 return sv;
1342 }
1343
1344 SvREFCNT_inc (sv); 1563 SvREFCNT_inc (sv);
1564 else
1565 croak ("filter_json_object callbacks must not return more than one scalar");
1566
1345 FREETMPS; LEAVE; 1567 FREETMPS; LEAVE;
1346 } 1568 }
1347 } 1569 }
1348 1570
1349 return sv; 1571 return sv;
1353 DEC_DEC_DEPTH; 1575 DEC_DEC_DEPTH;
1354 return 0; 1576 return 0;
1355} 1577}
1356 1578
1357static SV * 1579static SV *
1580decode_tag (dec_t *dec)
1581{
1582 SV *tag = 0;
1583 SV *val = 0;
1584
1585 if (!(dec->json.flags & F_ALLOW_TAGS))
1586 ERR ("malformed JSON string, neither array, object, number, string or atom");
1587
1588 ++dec->cur;
1589
1590 decode_ws (dec);
1591
1592 tag = decode_sv (dec);
1593 if (!tag)
1594 goto fail;
1595
1596 if (!SvPOK (tag))
1597 ERR ("malformed JSON string, (tag) must be a string");
1598
1599 decode_ws (dec);
1600
1601 if (*dec->cur != ')')
1602 ERR (") expected after tag");
1603
1604 ++dec->cur;
1605
1606 decode_ws (dec);
1607
1608 val = decode_sv (dec);
1609 if (!val)
1610 goto fail;
1611
1612 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1613 ERR ("malformed JSON string, tag value must be an array");
1614
1615 {
1616 AV *av = (AV *)SvRV (val);
1617 int i, len = av_len (av) + 1;
1618 HV *stash = gv_stashsv (tag, 0);
1619 SV *sv;
1620
1621 if (!stash)
1622 ERR ("cannot decode perl-object (package does not exist)");
1623
1624 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1625
1626 if (!method)
1627 ERR ("cannot decode perl-object (package does not have a THAW method)");
1628
1629 dSP;
1630
1631 ENTER; SAVETMPS;
1632 PUSHMARK (SP);
1633 EXTEND (SP, len + 2);
1634 // we re-bless the reference to get overload and other niceties right
1635 PUSHs (tag);
1636 PUSHs (sv_json);
1637
1638 for (i = 0; i < len; ++i)
1639 PUSHs (*av_fetch (av, i, 1));
1640
1641 PUTBACK;
1642 call_sv ((SV *)GvCV (method), G_SCALAR);
1643 SPAGAIN;
1644
1645 SvREFCNT_dec (tag);
1646 SvREFCNT_dec (val);
1647 sv = SvREFCNT_inc (POPs);
1648
1649 PUTBACK;
1650
1651 FREETMPS; LEAVE;
1652
1653 return sv;
1654 }
1655
1656fail:
1657 SvREFCNT_dec (tag);
1658 SvREFCNT_dec (val);
1659 return 0;
1660}
1661
1662static SV *
1358decode_sv (dec_t *dec) 1663decode_sv (dec_t *dec)
1359{ 1664{
1360 // the beauty of JSON: you need exactly one character lookahead 1665 // the beauty of JSON: you need exactly one character lookahead
1361 // to parse everything. 1666 // to parse everything.
1362 switch (*dec->cur) 1667 switch (*dec->cur)
1363 { 1668 {
1364 case '"': ++dec->cur; return decode_str (dec); 1669 case '"': ++dec->cur; return decode_str (dec);
1365 case '[': ++dec->cur; return decode_av (dec); 1670 case '[': ++dec->cur; return decode_av (dec);
1366 case '{': ++dec->cur; return decode_hv (dec); 1671 case '{': ++dec->cur; return decode_hv (dec);
1672 case '(': return decode_tag (dec);
1367 1673
1368 case '-': 1674 case '-':
1369 case '0': case '1': case '2': case '3': case '4': 1675 case '0': case '1': case '2': case '3': case '4':
1370 case '5': case '6': case '7': case '8': case '9': 1676 case '5': case '6': case '7': case '8': case '9':
1371 return decode_num (dec); 1677 return decode_num (dec);
1372 1678
1679 case 'f':
1680 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1681 {
1682 dec->cur += 5;
1683
1684 if (expect_false (!dec->json.v_false))
1685 dec->json.v_false = GET_BOOL (false);
1686
1687 return newSVsv (dec->json.v_false);
1688 }
1689 else
1690 ERR ("'false' expected");
1691
1692 break;
1693
1373 case 't': 1694 case 't':
1374 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1695 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1375 { 1696 {
1376 dec->cur += 4; 1697 dec->cur += 4;
1377#if JSON_SLOW 1698
1378 json_true = get_bool ("JSON::XS::true"); 1699 if (expect_false (!dec->json.v_true))
1379#endif 1700 dec->json.v_true = GET_BOOL (true);
1701
1380 return newSVsv (json_true); 1702 return newSVsv (dec->json.v_true);
1381 } 1703 }
1382 else 1704 else
1383 ERR ("'true' expected"); 1705 ERR ("'true' expected");
1384
1385 break;
1386
1387 case 'f':
1388 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1389 {
1390 dec->cur += 5;
1391#if JSON_SLOW
1392 json_false = get_bool ("JSON::XS::false");
1393#endif
1394 return newSVsv (json_false);
1395 }
1396 else
1397 ERR ("'false' expected");
1398 1706
1399 break; 1707 break;
1400 1708
1401 case 'n': 1709 case 'n':
1402 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1710 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1408 ERR ("'null' expected"); 1716 ERR ("'null' expected");
1409 1717
1410 break; 1718 break;
1411 1719
1412 default: 1720 default:
1413 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1721 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1414 break; 1722 break;
1415 } 1723 }
1416 1724
1417fail: 1725fail:
1418 return 0; 1726 return 0;
1419} 1727}
1420 1728
1421static SV * 1729static SV *
1422decode_json (SV *string, JSON *json, char **offset_return) 1730decode_json (SV *string, JSON *json, STRLEN *offset_return)
1423{ 1731{
1424 dec_t dec; 1732 dec_t dec;
1425 SV *sv; 1733 SV *sv;
1426 1734
1427 /* work around bugs in 5.10 where manipulating magic values 1735 /* work around bugs in 5.10 where manipulating magic values
1428 * will perl ignore the magic in subsequent accesses 1736 * makes perl ignore the magic in subsequent accesses.
1737 * also make a copy of non-PV values, to get them into a clean
1738 * state (SvPV should do that, but it's buggy, see below).
1739 *
1740 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1741 * as reported by Reini Urban.
1429 */ 1742 */
1430 /*SvGETMAGIC (string);*/ 1743 /*SvGETMAGIC (string);*/
1431 if (SvMAGICAL (string)) 1744 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1432 string = sv_2mortal (newSVsv (string)); 1745 string = sv_2mortal (newSVsv (string));
1433 1746
1434 SvUPGRADE (string, SVt_PV); 1747 SvUPGRADE (string, SVt_PV);
1435 1748
1436 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1749 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1453 if (offset > json->max_size && json->max_size) 1766 if (offset > json->max_size && json->max_size)
1454 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1767 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1455 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1768 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1456 } 1769 }
1457 1770
1458 if (json->flags & F_UTF8) 1771 if (DECODE_WANTS_OCTETS (json))
1459 sv_utf8_downgrade (string, 0); 1772 sv_utf8_downgrade (string, 0);
1460 else 1773 else
1461 sv_utf8_upgrade (string); 1774 sv_utf8_upgrade (string);
1462 1775
1463 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1776 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1475 1788
1476 decode_ws (&dec); 1789 decode_ws (&dec);
1477 sv = decode_sv (&dec); 1790 sv = decode_sv (&dec);
1478 1791
1479 if (offset_return) 1792 if (offset_return)
1480 *offset_return = dec.cur; 1793 *offset_return = dec.cur - SvPVX (string);
1481 1794 else if (sv)
1482 if (!(offset_return || !sv))
1483 { 1795 {
1484 // check for trailing garbage 1796 // check for trailing garbage
1485 decode_ws (&dec); 1797 decode_ws (&dec);
1486 1798
1487 if (*dec.cur) 1799 if (dec.cur != dec.end)
1488 { 1800 {
1489 dec.err = "garbage after JSON object"; 1801 dec.err = "garbage after JSON object";
1490 SvREFCNT_dec (sv); 1802 SvREFCNT_dec (sv);
1491 sv = 0; 1803 sv = 0;
1492 } 1804 }
1505 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1817 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1506 LEAVE; 1818 LEAVE;
1507 1819
1508 croak ("%s, at character offset %d (before \"%s\")", 1820 croak ("%s, at character offset %d (before \"%s\")",
1509 dec.err, 1821 dec.err,
1510 ptr_to_index (string, dec.cur), 1822 (int)ptr_to_index (string, dec.cur),
1511 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1823 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1512 } 1824 }
1513 1825
1514 sv = sv_2mortal (sv); 1826 sv = sv_2mortal (sv);
1515 1827
1516 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1828 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1517 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1829 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1518 1830
1519 return sv; 1831 return sv;
1520} 1832}
1521 1833
1525static void 1837static void
1526incr_parse (JSON *self) 1838incr_parse (JSON *self)
1527{ 1839{
1528 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1840 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1529 1841
1842 // the state machine here is a bit convoluted and could be simplified a lot
1843 // but this would make it slower, so...
1844
1530 for (;;) 1845 for (;;)
1531 { 1846 {
1532 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1533 switch (self->incr_mode) 1847 switch (self->incr_mode)
1534 { 1848 {
1849 // reached end of a scalar, see if we are inside a nested structure or not
1850 end_of_scalar:
1851 self->incr_mode = INCR_M_JSON;
1852
1853 if (self->incr_nest) // end of a scalar inside array, object or tag
1854 goto incr_m_json;
1855 else // end of scalar outside structure, json text ends here
1856 goto interrupt;
1857
1535 // only used for intiial whitespace skipping 1858 // only used for initial whitespace skipping
1536 case INCR_M_WS: 1859 case INCR_M_WS:
1537 for (;;) 1860 for (;;)
1538 { 1861 {
1539 if (*p > 0x20) 1862 if (*p > 0x20)
1540 { 1863 {
1864 if (*p == '#')
1865 {
1866 self->incr_mode = INCR_M_C0;
1867 goto incr_m_c;
1868 }
1869 else
1870 {
1541 self->incr_mode = INCR_M_JSON; 1871 self->incr_mode = INCR_M_JSON;
1542 goto incr_m_json; 1872 goto incr_m_json;
1873 }
1543 } 1874 }
1544 else if (!*p) 1875 else if (!*p)
1545 goto interrupt; 1876 goto interrupt;
1546 1877
1547 ++p; 1878 ++p;
1553 goto interrupt; 1884 goto interrupt;
1554 1885
1555 ++p; 1886 ++p;
1556 self->incr_mode = INCR_M_STR; 1887 self->incr_mode = INCR_M_STR;
1557 goto incr_m_str; 1888 goto incr_m_str;
1889
1890 // inside #-style comments
1891 case INCR_M_C0:
1892 case INCR_M_C1:
1893 incr_m_c:
1894 for (;;)
1895 {
1896 if (*p == '\n')
1897 {
1898 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1899 break;
1900 }
1901 else if (!*p)
1902 goto interrupt;
1903
1904 ++p;
1905 }
1906
1907 break;
1908
1909 // inside true/false/null
1910 case INCR_M_TFN:
1911 incr_m_tfn:
1912 for (;;)
1913 switch (*p++)
1914 {
1915 case 'r': case 'u': case 'e': // tRUE, falsE, nUll
1916 case 'a': case 'l': case 's': // fALSe, nuLL
1917 // allowed
1918 break;
1919
1920 default:
1921 --p;
1922 goto end_of_scalar;
1923 }
1924
1925 // inside a number
1926 case INCR_M_NUM:
1927 incr_m_num:
1928 for (;;)
1929 switch (*p++)
1930 {
1931 case 'e': case 'E': case '.': case '+':
1932 case '-':
1933 case '0': case '1': case '2': case '3': case '4':
1934 case '5': case '6': case '7': case '8': case '9':
1935 // allowed
1936 break;
1937
1938 default:
1939 --p;
1940 goto end_of_scalar;
1941 }
1558 1942
1559 // inside a string 1943 // inside a string
1560 case INCR_M_STR: 1944 case INCR_M_STR:
1561 incr_m_str: 1945 incr_m_str:
1562 for (;;) 1946 for (;;)
1563 { 1947 {
1564 if (*p == '"') 1948 if (*p == '"')
1565 { 1949 {
1566 ++p; 1950 ++p;
1567 self->incr_mode = INCR_M_JSON;
1568
1569 if (!self->incr_nest)
1570 goto interrupt;
1571
1572 goto incr_m_json; 1951 goto end_of_scalar;
1573 } 1952 }
1574 else if (*p == '\\') 1953 else if (*p == '\\')
1575 { 1954 {
1576 ++p; // "virtually" consumes character after \ 1955 ++p; // "virtually" consumes character after \
1577 1956
1607 --p; // do not eat the whitespace, let the next round do it 1986 --p; // do not eat the whitespace, let the next round do it
1608 goto interrupt; 1987 goto interrupt;
1609 } 1988 }
1610 break; 1989 break;
1611 1990
1991 // the following three blocks handle scalars. this makes the parser
1992 // more strict than required inside arrays or objects, and could
1993 // be moved to a special case on the toplevel (except strings)
1994 case 't':
1995 case 'f':
1996 case 'n':
1997 self->incr_mode = INCR_M_TFN;
1998 goto incr_m_tfn;
1999
2000 case '-':
2001 case '0': case '1': case '2': case '3': case '4':
2002 case '5': case '6': case '7': case '8': case '9':
2003 self->incr_mode = INCR_M_NUM;
2004 goto incr_m_num;
2005
1612 case '"': 2006 case '"':
1613 self->incr_mode = INCR_M_STR; 2007 self->incr_mode = INCR_M_STR;
1614 goto incr_m_str; 2008 goto incr_m_str;
1615 2009
1616 case '[': 2010 case '[':
1617 case '{': 2011 case '{':
2012 case '(':
1618 if (++self->incr_nest > self->max_depth) 2013 if (++self->incr_nest > self->max_depth)
1619 croak (ERR_NESTING_EXCEEDED); 2014 croak (ERR_NESTING_EXCEEDED);
1620 break; 2015 break;
1621 2016
1622 case ']': 2017 case ']':
1623 case '}': 2018 case '}':
1624 if (--self->incr_nest <= 0) 2019 if (--self->incr_nest <= 0)
1625 goto interrupt; 2020 goto interrupt;
2021 break;
2022
2023 case ')':
2024 --self->incr_nest;
2025 break;
2026
2027 case '#':
2028 self->incr_mode = INCR_M_C1;
2029 goto incr_m_c;
1626 } 2030 }
1627 } 2031 }
1628 } 2032 }
1629 2033
1630 modechange: 2034 modechange:
1631 ; 2035 ;
1632 } 2036 }
1633 2037
1634interrupt: 2038interrupt:
1635 self->incr_pos = p - SvPVX (self->incr_text); 2039 self->incr_pos = p - SvPVX (self->incr_text);
2040 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1636 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 2041 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1637} 2042}
1638 2043
1639///////////////////////////////////////////////////////////////////////////// 2044/////////////////////////////////////////////////////////////////////////////
1640// XS interface functions 2045// XS interface functions
1650 i >= '0' && i <= '9' ? i - '0' 2055 i >= '0' && i <= '9' ? i - '0'
1651 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2056 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1652 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2057 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1653 : -1; 2058 : -1;
1654 2059
1655 json_stash = gv_stashpv ("JSON::XS" , 1); 2060 json_stash = gv_stashpv ("JSON::XS" , 1);
1656 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2061 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1657
1658 json_true = get_bool ("JSON::XS::true");
1659 json_false = get_bool ("JSON::XS::false"); 2062 bool_false = get_bool ("Types::Serialiser::false");
2063 bool_true = get_bool ("Types::Serialiser::true");
2064
2065 sv_json = newSVpv ("JSON", 0);
2066 SvREADONLY_on (sv_json);
1660 2067
1661 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ 2068 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1662} 2069}
1663 2070
1664PROTOTYPES: DISABLE 2071PROTOTYPES: DISABLE
1665 2072
1666void CLONE (...) 2073void CLONE (...)
1667 CODE: 2074 CODE:
2075 // as long as these writes are atomic, the race should not matter
2076 // as existing threads either already use 0, or use the old value,
2077 // which is sitll correct for the initial thread.
1668 json_stash = 0; 2078 json_stash = 0;
1669 json_boolean_stash = 0; 2079 bool_stash = 0;
2080 bool_false = 0;
2081 bool_true = 0;
1670 2082
1671void new (char *klass) 2083void new (char *klass)
1672 PPCODE: 2084 PPCODE:
1673{ 2085{
1674 SV *pv = NEWSV (0, sizeof (JSON)); 2086 SV *pv = NEWSV (0, sizeof (JSON));
1675 SvPOK_only (pv); 2087 SvPOK_only (pv);
1676 json_init ((JSON *)SvPVX (pv)); 2088 json_init ((JSON *)SvPVX (pv));
1677 XPUSHs (sv_2mortal (sv_bless ( 2089 XPUSHs (sv_2mortal (sv_bless (
1678 newRV_noinc (pv), 2090 newRV_noinc (pv),
1679 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2091 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1680 ))); 2092 )));
1681} 2093}
2094
2095void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2096 PPCODE:
2097 self->v_false = newSVsv (v_false);
2098 self->v_true = newSVsv (v_true);
2099 XPUSHs (ST (0));
2100
2101void get_boolean_values (JSON *self)
2102 PPCODE:
2103 if (self->v_false && self->v_true)
2104 {
2105 EXTEND (SP, 2);
2106 PUSHs (self->v_false);
2107 PUSHs (self->v_true);
2108 }
1682 2109
1683void ascii (JSON *self, int enable = 1) 2110void ascii (JSON *self, int enable = 1)
1684 ALIAS: 2111 ALIAS:
1685 ascii = F_ASCII 2112 ascii = F_ASCII
1686 latin1 = F_LATIN1 2113 latin1 = F_LATIN1
1694 shrink = F_SHRINK 2121 shrink = F_SHRINK
1695 allow_blessed = F_ALLOW_BLESSED 2122 allow_blessed = F_ALLOW_BLESSED
1696 convert_blessed = F_CONV_BLESSED 2123 convert_blessed = F_CONV_BLESSED
1697 relaxed = F_RELAXED 2124 relaxed = F_RELAXED
1698 allow_unknown = F_ALLOW_UNKNOWN 2125 allow_unknown = F_ALLOW_UNKNOWN
2126 allow_tags = F_ALLOW_TAGS
1699 PPCODE: 2127 PPCODE:
1700{ 2128{
1701 if (enable) 2129 if (enable)
1702 self->flags |= ix; 2130 self->flags |= ix;
1703 else 2131 else
1719 get_shrink = F_SHRINK 2147 get_shrink = F_SHRINK
1720 get_allow_blessed = F_ALLOW_BLESSED 2148 get_allow_blessed = F_ALLOW_BLESSED
1721 get_convert_blessed = F_CONV_BLESSED 2149 get_convert_blessed = F_CONV_BLESSED
1722 get_relaxed = F_RELAXED 2150 get_relaxed = F_RELAXED
1723 get_allow_unknown = F_ALLOW_UNKNOWN 2151 get_allow_unknown = F_ALLOW_UNKNOWN
2152 get_allow_tags = F_ALLOW_TAGS
1724 PPCODE: 2153 PPCODE:
1725 XPUSHs (boolSV (self->flags & ix)); 2154 XPUSHs (boolSV (self->flags & ix));
1726 2155
1727void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2156void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1728 PPCODE: 2157 PPCODE:
1756} 2185}
1757 2186
1758void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2187void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1759 PPCODE: 2188 PPCODE:
1760{ 2189{
1761 if (!self->cb_sk_object) 2190 if (!self->cb_sk_object)
1762 self->cb_sk_object = newHV (); 2191 self->cb_sk_object = newHV ();
1763 2192
1764 if (SvOK (cb)) 2193 if (SvOK (cb))
1765 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2194 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1766 else 2195 else
1777 XPUSHs (ST (0)); 2206 XPUSHs (ST (0));
1778} 2207}
1779 2208
1780void encode (JSON *self, SV *scalar) 2209void encode (JSON *self, SV *scalar)
1781 PPCODE: 2210 PPCODE:
1782 XPUSHs (encode_json (scalar, self)); 2211 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2212 XPUSHs (scalar);
1783 2213
1784void decode (JSON *self, SV *jsonstr) 2214void decode (JSON *self, SV *jsonstr)
1785 PPCODE: 2215 PPCODE:
1786 XPUSHs (decode_json (jsonstr, self, 0)); 2216 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2217 XPUSHs (jsonstr);
1787 2218
1788void decode_prefix (JSON *self, SV *jsonstr) 2219void decode_prefix (JSON *self, SV *jsonstr)
1789 PPCODE: 2220 PPCODE:
1790{ 2221{
1791 char *offset; 2222 SV *sv;
2223 STRLEN offset;
2224 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1792 EXTEND (SP, 2); 2225 EXTEND (SP, 2);
1793 PUSHs (decode_json (jsonstr, self, &offset)); 2226 PUSHs (sv);
1794 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2227 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1795} 2228}
1796 2229
1797void incr_parse (JSON *self, SV *jsonstr = 0) 2230void incr_parse (JSON *self, SV *jsonstr = 0)
1798 PPCODE: 2231 PPCODE:
1799{ 2232{
1800 if (!self->incr_text) 2233 if (!self->incr_text)
1801 self->incr_text = newSVpvn ("", 0); 2234 self->incr_text = newSVpvn ("", 0);
2235
2236 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2237 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2238 if (DECODE_WANTS_OCTETS (self))
2239 {
2240 if (self->incr_pos)
2241 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2242 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2243
2244 sv_utf8_downgrade (self->incr_text, 0);
2245 }
2246 else
2247 {
2248 sv_utf8_upgrade (self->incr_text);
2249
2250 if (self->incr_pos)
2251 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2252 - (U8 *)SvPVX (self->incr_text);
2253 }
1802 2254
1803 // append data, if any 2255 // append data, if any
1804 if (jsonstr) 2256 if (jsonstr)
1805 { 2257 {
2258 /* make sure both strings have same encoding */
2259 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1806 if (SvUTF8 (jsonstr)) 2260 if (SvUTF8 (jsonstr))
2261 sv_utf8_downgrade (jsonstr, 0);
1807 { 2262 else
1808 if (!SvUTF8 (self->incr_text))
1809 {
1810 /* utf-8-ness differs, need to upgrade */
1811 sv_utf8_upgrade (self->incr_text);
1812
1813 if (self->incr_pos)
1814 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1815 - (U8 *)SvPVX (self->incr_text);
1816 }
1817 }
1818 else if (SvUTF8 (self->incr_text))
1819 sv_utf8_upgrade (jsonstr); 2263 sv_utf8_upgrade (jsonstr);
1820 2264
2265 /* and then just blindly append */
1821 { 2266 {
1822 STRLEN len; 2267 STRLEN len;
1823 const char *str = SvPV (jsonstr, len); 2268 const char *str = SvPV (jsonstr, len);
1824 STRLEN cur = SvCUR (self->incr_text); 2269 STRLEN cur = SvCUR (self->incr_text);
1825 2270
1826 if (SvLEN (self->incr_text) <= cur + len) 2271 if (SvLEN (self->incr_text) - cur <= len)
1827 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2272 json_sv_grow (self->incr_text, cur, len);
1828 2273
1829 Move (str, SvEND (self->incr_text), len, char); 2274 Move (str, SvEND (self->incr_text), len, char);
1830 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2275 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1831 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2276 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1832 } 2277 }
1833 } 2278 }
1834 2279
1835 if (GIMME_V != G_VOID) 2280 if (GIMME_V != G_VOID)
1836 do 2281 do
1837 { 2282 {
2283 SV *sv;
1838 char *offset; 2284 STRLEN offset;
1839 2285
1840 if (!INCR_DONE (self)) 2286 if (!INCR_DONE (self))
1841 { 2287 {
1842 incr_parse (self); 2288 incr_parse (self);
1843 2289
1844 if (self->incr_pos > self->max_size && self->max_size) 2290 if (self->incr_pos > self->max_size && self->max_size)
1845 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2291 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1846 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2292 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1847 2293
1848 if (!INCR_DONE (self)) 2294 if (!INCR_DONE (self))
2295 {
2296 // as an optimisation, do not accumulate white space in the incr buffer
2297 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2298 {
2299 self->incr_pos = 0;
2300 SvCUR_set (self->incr_text, 0);
2301 }
2302
1849 break; 2303 break;
2304 }
1850 } 2305 }
1851 2306
1852 XPUSHs (decode_json (self->incr_text, self, &offset)); 2307 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2308 XPUSHs (sv);
1853 2309
1854 self->incr_pos -= offset - SvPVX (self->incr_text); 2310 self->incr_pos -= offset;
1855 self->incr_nest = 0; 2311 self->incr_nest = 0;
1856 self->incr_mode = 0; 2312 self->incr_mode = 0;
1857 2313
1858 sv_chop (self->incr_text, offset); 2314 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1859 } 2315 }
1860 while (GIMME_V == G_ARRAY); 2316 while (GIMME_V == G_ARRAY);
1861} 2317}
1862 2318
1863SV *incr_text (JSON *self) 2319SV *incr_text (JSON *self)
1894 self->incr_mode = 0; 2350 self->incr_mode = 0;
1895} 2351}
1896 2352
1897void DESTROY (JSON *self) 2353void DESTROY (JSON *self)
1898 CODE: 2354 CODE:
2355 SvREFCNT_dec (self->v_false);
2356 SvREFCNT_dec (self->v_true);
1899 SvREFCNT_dec (self->cb_sk_object); 2357 SvREFCNT_dec (self->cb_sk_object);
1900 SvREFCNT_dec (self->cb_object); 2358 SvREFCNT_dec (self->cb_object);
1901 SvREFCNT_dec (self->incr_text); 2359 SvREFCNT_dec (self->incr_text);
1902 2360
1903PROTOTYPES: ENABLE 2361PROTOTYPES: ENABLE
1904 2362
1905void encode_json (SV *scalar) 2363void encode_json (SV *scalar)
1906 ALIAS:
1907 to_json_ = 0
1908 encode_json = F_UTF8
1909 PPCODE: 2364 PPCODE:
1910{ 2365{
1911 JSON json; 2366 JSON json;
1912 json_init (&json); 2367 json_init (&json);
1913 json.flags |= ix; 2368 json.flags |= F_UTF8;
1914 XPUSHs (encode_json (scalar, &json)); 2369 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2370 XPUSHs (scalar);
1915} 2371}
1916 2372
1917void decode_json (SV *jsonstr) 2373void decode_json (SV *jsonstr)
1918 ALIAS:
1919 from_json_ = 0
1920 decode_json = F_UTF8
1921 PPCODE: 2374 PPCODE:
1922{ 2375{
1923 JSON json; 2376 JSON json;
1924 json_init (&json); 2377 json_init (&json);
1925 json.flags |= ix; 2378 json.flags |= F_UTF8;
1926 XPUSHs (decode_json (jsonstr, &json, 0)); 2379 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2380 XPUSHs (jsonstr);
1927} 2381}
1928 2382
1929

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines