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 1.103 by root, Wed Jan 6 08:01:39 2010 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
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// 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
21#endif
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
20#endif 32#endif
21 33
22// three extra for rounding, sign, and end of string 34// three extra for rounding, sign, and end of string
23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
24 36
33#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
34#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
35#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
36#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
37#define F_ALLOW_UNKNOWN 0x00002000UL 49#define F_ALLOW_UNKNOWN 0x00002000UL
50#define F_ALLOW_TAGS 0x00004000UL
38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39 52
40#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
41 54
42#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
43#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
44 57
45#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)
46 61
47#define SB do { 62#define SB do {
48#define SE } while (0) 63#define SE } while (0)
49 64
50#if __GNUC__ >= 3 65#if __GNUC__ >= 3
63 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
64 79
65#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?)"
66 81
67#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
68# define JSON_SLOW 1
69# 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))
70#else 86#else
71# define JSON_SLOW 0
72# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
73#endif 90#endif
74 91
75static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
76static 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;
77 98
78enum { 99enum {
79 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
80 INCR_M_STR, // inside string 103 INCR_M_STR, // inside string
81 INCR_M_BS, // inside backslash 104 INCR_M_BS, // inside backslash
82 INCR_M_C0, // inside comment in initial whitespace sequence 105 INCR_M_C0, // inside comment in initial whitespace sequence
83 INCR_M_C1, // inside comment in other places 106 INCR_M_C1, // inside comment in other places
84 INCR_M_JSON // outside anything, count nesting 107 INCR_M_JSON // outside anything, count nesting
97 // for the incremental parser 120 // for the incremental parser
98 SV *incr_text; // the source text so far 121 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text 122 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level 123 int incr_nest; // {[]}-nesting level
101 unsigned char incr_mode; 124 unsigned char incr_mode;
125
126 SV *v_false, *v_true;
102} JSON; 127} JSON;
103 128
104INLINE void 129INLINE void
105json_init (JSON *json) 130json_init (JSON *json)
106{ 131{
107 Zero (json, 1, JSON); 132 static const JSON init = { F_ALLOW_NONREF, 512 };
108 json->max_depth = 512; 133
134 *json = init;
109} 135}
110 136
111///////////////////////////////////////////////////////////////////////////// 137/////////////////////////////////////////////////////////////////////////////
112// utility functions 138// utility functions
113 139
135 SvPV_renew (sv, SvCUR (sv) + 1); 161 SvPV_renew (sv, SvCUR (sv) + 1);
136#endif 162#endif
137 } 163 }
138} 164}
139 165
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
140// 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
141// case of an error. 192// case of an error.
142// we special-case "safe" characters from U+80 .. U+7FF, 193// we special-case "safe" characters from U+80 .. U+7FF,
143// but use the very good perl function to parse anything else. 194// but use the very good perl function to parse anything else.
144// note that we never call this function for a ascii codepoints 195// note that we never call this function for a ascii codepoints
145INLINE UV 196INLINE UV
188 ? utf8_distance (offset, SvPVX (sv)) 239 ? utf8_distance (offset, SvPVX (sv))
189 : offset - SvPVX (sv); 240 : offset - SvPVX (sv);
190} 241}
191 242
192///////////////////////////////////////////////////////////////////////////// 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/////////////////////////////////////////////////////////////////////////////
193// encoder 382// encoder
194 383
195// structure used for encoding JSON 384// structure used for encoding JSON
196typedef struct 385typedef struct
197{ 386{
204} enc_t; 393} enc_t;
205 394
206INLINE void 395INLINE void
207need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
208{ 397{
209 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
210 { 399 {
211 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
212 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
213 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
214 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
215 } 404 }
216} 405}
217 406
218INLINE void 407INLINE void
219encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
235 424
236 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
237 { 426 {
238 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
239 { 428 {
240 need (enc, len += 1); 429 need (enc, len + 1);
241 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
242 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
243 } 432 }
244 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
245 { 434 {
246 need (enc, len += 1); 435 need (enc, len + 1);
247 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
248 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
249 } 438 }
250 else 439 else
251 *enc->cur++ = ch; 440 *enc->cur++ = ch;
254 } 443 }
255 else 444 else
256 { 445 {
257 switch (ch) 446 switch (ch)
258 { 447 {
259 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;
260 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;
261 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;
262 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;
263 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;
264 453
265 default: 454 default:
266 { 455 {
267 STRLEN clen; 456 STRLEN clen;
268 UV uch; 457 UV uch;
284 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
285 { 474 {
286 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
287 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);
288 477
289 need (enc, len += 11); 478 need (enc, len + 11);
290 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
291 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
292 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
293 enc->cur += 12; 482 enc->cur += 12;
294 } 483 }
295 else 484 else
296 { 485 {
297 need (enc, len += 5); 486 need (enc, len + 5);
298 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
299 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
300 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
301 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
302 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
310 *enc->cur++ = uch; 499 *enc->cur++ = uch;
311 str += clen; 500 str += clen;
312 } 501 }
313 else if (is_utf8) 502 else if (is_utf8)
314 { 503 {
315 need (enc, len += clen); 504 need (enc, len + clen);
316 do 505 do
317 { 506 {
318 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
319 } 508 }
320 while (--clen); 509 while (--clen);
321 } 510 }
322 else 511 else
323 { 512 {
324 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
325 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
326 ++str; 515 ++str;
327 } 516 }
328 } 517 }
329 } 518 }
383 572
384 if (enc->indent >= enc->json.max_depth) 573 if (enc->indent >= enc->json.max_depth)
385 croak (ERR_NESTING_EXCEEDED); 574 croak (ERR_NESTING_EXCEEDED);
386 575
387 encode_ch (enc, '['); 576 encode_ch (enc, '[');
388 577
389 if (len >= 0) 578 if (len >= 0)
390 { 579 {
391 encode_nl (enc); ++enc->indent; 580 encode_nl (enc); ++enc->indent;
392 581
393 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
405 encode_comma (enc); 594 encode_comma (enc);
406 } 595 }
407 596
408 encode_nl (enc); --enc->indent; encode_indent (enc); 597 encode_nl (enc); --enc->indent; encode_indent (enc);
409 } 598 }
410 599
411 encode_ch (enc, ']'); 600 encode_ch (enc, ']');
412} 601}
413 602
414static void 603static void
415encode_hk (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
419 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
420 { 609 {
421 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
422 STRLEN len; 611 STRLEN len;
423 char *str; 612 char *str;
424 613
425 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
426 str = SvPV (sv, len); 615 str = SvPV (sv, len);
427 616
428 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
429 } 618 }
495 } 684 }
496 685
497 if (count) 686 if (count)
498 { 687 {
499 int i, fast = 1; 688 int i, fast = 1;
500#if defined(__BORLANDC__) || defined(_MSC_VER) 689 HE *hes_stack [STACK_HES];
501 HE **hes = _alloca (count * sizeof (HE)); 690 HE **hes = hes_stack;
502#else 691
503 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 692 // allocate larger arrays on the heap
504#endif 693 if (count > STACK_HES)
694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
505 698
506 i = 0; 699 i = 0;
507 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
508 { 701 {
509 hes [i++] = he; 702 hes [i++] = he;
578// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
579static void 772static void
580encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
581{ 774{
582 svtype svt; 775 svtype svt;
776 GV *method;
583 777
584 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
585 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
586 780
587 if (expect_false (SvOBJECT (sv))) 781 if (expect_false (SvOBJECT (sv)))
588 { 782 {
589 HV *stash = !JSON_SLOW || json_boolean_stash 783 HV *stash = SvSTASH (sv);
590 ? json_boolean_stash
591 : gv_stashpv ("JSON::XS::Boolean", 1);
592 784
593 if (SvSTASH (sv) == stash) 785 if (stash == bool_stash)
594 { 786 {
595 if (SvIV (sv))
596 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
597 else
598 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
599 } 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);
600 else 862 else
601 {
602#if 0
603 if (0 && sv_derived_from (rv, "JSON::Literal"))
604 {
605 // not yet
606 }
607#endif
608 if (enc->json.flags & F_CONV_BLESSED)
609 {
610 // we re-bless the reference to get overload and other niceties right
611 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
612
613 if (to_json)
614 {
615 dSP;
616
617 ENTER; SAVETMPS; PUSHMARK (SP);
618 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
619
620 // calling with G_SCALAR ensures that we always get a 1 return value
621 PUTBACK;
622 call_sv ((SV *)GvCV (to_json), G_SCALAR);
623 SPAGAIN;
624
625 // catch this surprisingly common error
626 if (SvROK (TOPs) && SvRV (TOPs) == sv)
627 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
628
629 sv = POPs;
630 PUTBACK;
631
632 encode_sv (enc, sv);
633
634 FREETMPS; LEAVE;
635 }
636 else if (enc->json.flags & F_ALLOW_BLESSED)
637 encode_str (enc, "null", 4, 0);
638 else
639 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
640 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
641 }
642 else if (enc->json.flags & F_ALLOW_BLESSED)
643 encode_str (enc, "null", 4, 0);
644 else
645 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)",
646 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 864 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
647 }
648 } 865 }
649 else if (svt == SVt_PVHV) 866 else if (svt == SVt_PVHV)
650 encode_hv (enc, (HV *)sv); 867 encode_hv (enc, (HV *)sv);
651 else if (svt == SVt_PVAV) 868 else if (svt == SVt_PVAV)
652 encode_av (enc, (AV *)sv); 869 encode_av (enc, (AV *)sv);
653 else if (svt < SVt_PVAV) 870 else if (svt < SVt_PVAV)
654 { 871 {
655 STRLEN len = 0; 872 int bool_type = ref_bool_type (sv);
656 char *pv = svt ? SvPV (sv, len) : 0;
657 873
658 if (len == 1 && *pv == '1') 874 if (bool_type == 1)
659 encode_str (enc, "true", 4, 0); 875 encode_str (enc, "true", 4, 0);
660 else if (len == 1 && *pv == '0') 876 else if (bool_type == 0)
661 encode_str (enc, "false", 5, 0); 877 encode_str (enc, "false", 5, 0);
662 else if (enc->json.flags & F_ALLOW_UNKNOWN) 878 else if (enc->json.flags & F_ALLOW_UNKNOWN)
663 encode_str (enc, "null", 4, 0); 879 encode_str (enc, "null", 4, 0);
664 else 880 else
665 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",
728 } 944 }
729 else 945 else
730 { 946 {
731 // large integer, use the (rather slow) snprintf way. 947 // large integer, use the (rather slow) snprintf way.
732 need (enc, IVUV_MAXCHARS); 948 need (enc, IVUV_MAXCHARS);
733 enc->cur += 949 enc->cur +=
734 SvIsUV(sv) 950 SvIsUV(sv)
735 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 951 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
736 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 952 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
737 } 953 }
738 } 954 }
739 else if (SvROK (sv)) 955 else if (SvROK (sv))
740 encode_rv (enc, SvRV (sv)); 956 encode_rv (enc, SvRV (sv));
741 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 957 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
742 encode_str (enc, "null", 4, 0); 958 encode_str (enc, "null", 4, 0);
743 else 959 else
744 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",
745 SvPV_nolen (sv), SvFLAGS (sv)); 961 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
746} 962}
747 963
748static SV * 964static SV *
749encode_json (SV *scalar, JSON *json) 965encode_json (SV *scalar, JSON *json)
750{ 966{
751 enc_t enc; 967 enc_t enc;
752 968
753 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 969 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
754 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)");
755 971
756 enc.json = *json; 972 enc.json = *json;
757 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 973 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
758 enc.cur = SvPVX (enc.sv); 974 enc.cur = SvPVX (enc.sv);
965 *cur++ = *dec_cur++; 1181 *cur++ = *dec_cur++;
966 while (--clen); 1182 while (--clen);
967 1183
968 utf8 = 1; 1184 utf8 = 1;
969 } 1185 }
1186 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1187 *cur++ = ch;
970 else 1188 else
971 { 1189 {
972 --dec_cur; 1190 --dec_cur;
973 1191
974 if (!ch) 1192 if (!ch)
984 1202
985 if (sv) 1203 if (sv)
986 { 1204 {
987 STRLEN cur = SvCUR (sv); 1205 STRLEN cur = SvCUR (sv);
988 1206
989 if (SvLEN (sv) <= cur + len) 1207 if (SvLEN (sv) - cur <= len)
990 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1208 json_sv_grow (sv, cur, len);
991 1209
992 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1210 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
993 SvCUR_set (sv, SvCUR (sv) + len); 1211 SvCUR_set (sv, SvCUR (sv) + len);
994 } 1212 }
995 else 1213 else
1119 } 1337 }
1120 1338
1121 len -= *start == '-' ? 1 : 0; 1339 len -= *start == '-' ? 1 : 0;
1122 1340
1123 // does not fit into IV or UV, try NV 1341 // does not fit into IV or UV, try NV
1124 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1342 if (len <= NV_DIG)
1125 #if defined (LDBL_DIG)
1126 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1127 #endif
1128 )
1129 // fits into NV without loss of precision 1343 // fits into NV without loss of precision
1130 return newSVnv (Atof (start)); 1344 return newSVnv (json_atof (start));
1131 1345
1132 // everything else fails, convert it to a string 1346 // everything else fails, convert it to a string
1133 return newSVpvn (start, dec->cur - start); 1347 return newSVpvn (start, dec->cur - start);
1134 } 1348 }
1135 1349
1136 // loss of precision here 1350 // loss of precision here
1137 return newSVnv (Atof (start)); 1351 return newSVnv (json_atof (start));
1138 1352
1139fail: 1353fail:
1140 return 0; 1354 return 0;
1141} 1355}
1142 1356
1166 if (*dec->cur == ']') 1380 if (*dec->cur == ']')
1167 { 1381 {
1168 ++dec->cur; 1382 ++dec->cur;
1169 break; 1383 break;
1170 } 1384 }
1171 1385
1172 if (*dec->cur != ',') 1386 if (*dec->cur != ',')
1173 ERR (", or ] expected while parsing array"); 1387 ERR (", or ] expected while parsing array");
1174 1388
1175 ++dec->cur; 1389 ++dec->cur;
1176 1390
1290 1504
1291 DEC_DEC_DEPTH; 1505 DEC_DEC_DEPTH;
1292 sv = newRV_noinc ((SV *)hv); 1506 sv = newRV_noinc ((SV *)hv);
1293 1507
1294 // check filter callbacks 1508 // check filter callbacks
1295 if (dec->json.flags & F_HOOK) 1509 if (expect_false (dec->json.flags & F_HOOK))
1296 { 1510 {
1297 if (dec->json.cb_sk_object && HvKEYS (hv) == 1) 1511 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1298 { 1512 {
1299 HE *cb, *he; 1513 HE *cb, *he;
1300 1514
1301 hv_iterinit (hv); 1515 hv_iterinit (hv);
1302 he = hv_iternext (hv); 1516 he = hv_iternext (hv);
1303 hv_iterinit (hv); 1517 hv_iterinit (hv);
1304 1518
1305 // the next line creates a mortal sv each time its called. 1519 // the next line creates a mortal sv each time it's called.
1306 // might want to optimise this for common cases. 1520 // might want to optimise this for common cases.
1307 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);
1308 1522
1309 if (cb) 1523 if (cb)
1310 { 1524 {
1311 dSP; 1525 dSP;
1312 int count; 1526 int count;
1313 1527
1314 ENTER; SAVETMPS; PUSHMARK (SP); 1528 ENTER; SAVETMPS;
1529 PUSHMARK (SP);
1315 XPUSHs (HeVAL (he)); 1530 XPUSHs (HeVAL (he));
1531 sv_2mortal (sv);
1316 1532
1317 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1533 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1318 1534
1319 if (count == 1) 1535 if (count == 1)
1320 { 1536 {
1321 sv = newSVsv (POPs); 1537 sv = newSVsv (POPs);
1322 FREETMPS; LEAVE; 1538 FREETMPS; LEAVE;
1323 return sv; 1539 return sv;
1324 } 1540 }
1541 else if (count)
1542 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1325 1543
1544 SvREFCNT_inc (sv);
1326 FREETMPS; LEAVE; 1545 FREETMPS; LEAVE;
1327 } 1546 }
1328 } 1547 }
1329 1548
1330 if (dec->json.cb_object) 1549 if (dec->json.cb_object)
1331 { 1550 {
1332 dSP; 1551 dSP;
1333 int count; 1552 int count;
1334 1553
1335 ENTER; SAVETMPS; PUSHMARK (SP); 1554 ENTER; SAVETMPS;
1555 PUSHMARK (SP);
1336 XPUSHs (sv_2mortal (sv)); 1556 XPUSHs (sv_2mortal (sv));
1337 1557
1338 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1558 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1339 1559
1340 if (count == 1) 1560 if (count == 1)
1341 {
1342 sv = newSVsv (POPs); 1561 sv = newSVsv (POPs);
1343 FREETMPS; LEAVE; 1562 else if (count == 0)
1344 return sv;
1345 }
1346
1347 SvREFCNT_inc (sv); 1563 SvREFCNT_inc (sv);
1564 else
1565 croak ("filter_json_object callbacks must not return more than one scalar");
1566
1348 FREETMPS; LEAVE; 1567 FREETMPS; LEAVE;
1349 } 1568 }
1350 } 1569 }
1351 1570
1352 return sv; 1571 return sv;
1356 DEC_DEC_DEPTH; 1575 DEC_DEC_DEPTH;
1357 return 0; 1576 return 0;
1358} 1577}
1359 1578
1360static 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 *
1361decode_sv (dec_t *dec) 1663decode_sv (dec_t *dec)
1362{ 1664{
1363 // the beauty of JSON: you need exactly one character lookahead 1665 // the beauty of JSON: you need exactly one character lookahead
1364 // to parse everything. 1666 // to parse everything.
1365 switch (*dec->cur) 1667 switch (*dec->cur)
1366 { 1668 {
1367 case '"': ++dec->cur; return decode_str (dec); 1669 case '"': ++dec->cur; return decode_str (dec);
1368 case '[': ++dec->cur; return decode_av (dec); 1670 case '[': ++dec->cur; return decode_av (dec);
1369 case '{': ++dec->cur; return decode_hv (dec); 1671 case '{': ++dec->cur; return decode_hv (dec);
1672 case '(': return decode_tag (dec);
1370 1673
1371 case '-': 1674 case '-':
1372 case '0': case '1': case '2': case '3': case '4': 1675 case '0': case '1': case '2': case '3': case '4':
1373 case '5': case '6': case '7': case '8': case '9': 1676 case '5': case '6': case '7': case '8': case '9':
1374 return decode_num (dec); 1677 return decode_num (dec);
1375 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
1376 case 't': 1694 case 't':
1377 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1695 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1378 { 1696 {
1379 dec->cur += 4; 1697 dec->cur += 4;
1380#if JSON_SLOW 1698
1381 json_true = get_bool ("JSON::XS::true"); 1699 if (expect_false (!dec->json.v_true))
1382#endif 1700 dec->json.v_true = GET_BOOL (true);
1701
1383 return newSVsv (json_true); 1702 return newSVsv (dec->json.v_true);
1384 } 1703 }
1385 else 1704 else
1386 ERR ("'true' expected"); 1705 ERR ("'true' expected");
1387
1388 break;
1389
1390 case 'f':
1391 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1392 {
1393 dec->cur += 5;
1394#if JSON_SLOW
1395 json_false = get_bool ("JSON::XS::false");
1396#endif
1397 return newSVsv (json_false);
1398 }
1399 else
1400 ERR ("'false' expected");
1401 1706
1402 break; 1707 break;
1403 1708
1404 case 'n': 1709 case 'n':
1405 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1710 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1411 ERR ("'null' expected"); 1716 ERR ("'null' expected");
1412 1717
1413 break; 1718 break;
1414 1719
1415 default: 1720 default:
1416 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1721 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1417 break; 1722 break;
1418 } 1723 }
1419 1724
1420fail: 1725fail:
1421 return 0; 1726 return 0;
1422} 1727}
1423 1728
1424static SV * 1729static SV *
1425decode_json (SV *string, JSON *json, char **offset_return) 1730decode_json (SV *string, JSON *json, STRLEN *offset_return)
1426{ 1731{
1427 dec_t dec; 1732 dec_t dec;
1428 SV *sv; 1733 SV *sv;
1429 1734
1430 /* work around bugs in 5.10 where manipulating magic values 1735 /* work around bugs in 5.10 where manipulating magic values
1431 * 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.
1432 */ 1742 */
1433 /*SvGETMAGIC (string);*/ 1743 /*SvGETMAGIC (string);*/
1434 if (SvMAGICAL (string)) 1744 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1435 string = sv_2mortal (newSVsv (string)); 1745 string = sv_2mortal (newSVsv (string));
1436 1746
1437 SvUPGRADE (string, SVt_PV); 1747 SvUPGRADE (string, SVt_PV);
1438 1748
1439 /* 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
1456 if (offset > json->max_size && json->max_size) 1766 if (offset > json->max_size && json->max_size)
1457 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",
1458 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1768 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1459 } 1769 }
1460 1770
1461 if (json->flags & F_UTF8) 1771 if (DECODE_WANTS_OCTETS (json))
1462 sv_utf8_downgrade (string, 0); 1772 sv_utf8_downgrade (string, 0);
1463 else 1773 else
1464 sv_utf8_upgrade (string); 1774 sv_utf8_upgrade (string);
1465 1775
1466 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1776 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1478 1788
1479 decode_ws (&dec); 1789 decode_ws (&dec);
1480 sv = decode_sv (&dec); 1790 sv = decode_sv (&dec);
1481 1791
1482 if (offset_return) 1792 if (offset_return)
1483 *offset_return = dec.cur; 1793 *offset_return = dec.cur - SvPVX (string);
1484 1794 else if (sv)
1485 if (!(offset_return || !sv))
1486 { 1795 {
1487 // check for trailing garbage 1796 // check for trailing garbage
1488 decode_ws (&dec); 1797 decode_ws (&dec);
1489 1798
1490 if (*dec.cur) 1799 if (dec.cur != dec.end)
1491 { 1800 {
1492 dec.err = "garbage after JSON object"; 1801 dec.err = "garbage after JSON object";
1493 SvREFCNT_dec (sv); 1802 SvREFCNT_dec (sv);
1494 sv = 0; 1803 sv = 0;
1495 } 1804 }
1508 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);
1509 LEAVE; 1818 LEAVE;
1510 1819
1511 croak ("%s, at character offset %d (before \"%s\")", 1820 croak ("%s, at character offset %d (before \"%s\")",
1512 dec.err, 1821 dec.err,
1513 ptr_to_index (string, dec.cur), 1822 (int)ptr_to_index (string, dec.cur),
1514 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1823 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1515 } 1824 }
1516 1825
1517 sv = sv_2mortal (sv); 1826 sv = sv_2mortal (sv);
1518 1827
1519 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1828 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1520 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)");
1521 1830
1522 return sv; 1831 return sv;
1523} 1832}
1524 1833
1533 // the state machine here is a bit convoluted and could be simplified a lot 1842 // the state machine here is a bit convoluted and could be simplified a lot
1534 // but this would make it slower, so... 1843 // but this would make it slower, so...
1535 1844
1536 for (;;) 1845 for (;;)
1537 { 1846 {
1538 //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
1539 switch (self->incr_mode) 1847 switch (self->incr_mode)
1540 { 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
1541 // only used for initial whitespace skipping 1858 // only used for initial whitespace skipping
1542 case INCR_M_WS: 1859 case INCR_M_WS:
1543 for (;;) 1860 for (;;)
1544 { 1861 {
1545 if (*p > 0x20) 1862 if (*p > 0x20)
1587 ++p; 1904 ++p;
1588 } 1905 }
1589 1906
1590 break; 1907 break;
1591 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 }
1942
1592 // inside a string 1943 // inside a string
1593 case INCR_M_STR: 1944 case INCR_M_STR:
1594 incr_m_str: 1945 incr_m_str:
1595 for (;;) 1946 for (;;)
1596 { 1947 {
1597 if (*p == '"') 1948 if (*p == '"')
1598 { 1949 {
1599 ++p; 1950 ++p;
1600 self->incr_mode = INCR_M_JSON;
1601
1602 if (!self->incr_nest)
1603 goto interrupt;
1604
1605 goto incr_m_json; 1951 goto end_of_scalar;
1606 } 1952 }
1607 else if (*p == '\\') 1953 else if (*p == '\\')
1608 { 1954 {
1609 ++p; // "virtually" consumes character after \ 1955 ++p; // "virtually" consumes character after \
1610 1956
1640 --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
1641 goto interrupt; 1987 goto interrupt;
1642 } 1988 }
1643 break; 1989 break;
1644 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
1645 case '"': 2006 case '"':
1646 self->incr_mode = INCR_M_STR; 2007 self->incr_mode = INCR_M_STR;
1647 goto incr_m_str; 2008 goto incr_m_str;
1648 2009
1649 case '[': 2010 case '[':
1650 case '{': 2011 case '{':
2012 case '(':
1651 if (++self->incr_nest > self->max_depth) 2013 if (++self->incr_nest > self->max_depth)
1652 croak (ERR_NESTING_EXCEEDED); 2014 croak (ERR_NESTING_EXCEEDED);
1653 break; 2015 break;
1654 2016
1655 case ']': 2017 case ']':
1656 case '}': 2018 case '}':
1657 if (--self->incr_nest <= 0) 2019 if (--self->incr_nest <= 0)
1658 goto interrupt; 2020 goto interrupt;
2021 break;
2022
2023 case ')':
2024 --self->incr_nest;
1659 break; 2025 break;
1660 2026
1661 case '#': 2027 case '#':
1662 self->incr_mode = INCR_M_C1; 2028 self->incr_mode = INCR_M_C1;
1663 goto incr_m_c; 2029 goto incr_m_c;
1689 i >= '0' && i <= '9' ? i - '0' 2055 i >= '0' && i <= '9' ? i - '0'
1690 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2056 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1691 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2057 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1692 : -1; 2058 : -1;
1693 2059
1694 json_stash = gv_stashpv ("JSON::XS" , 1); 2060 json_stash = gv_stashpv ("JSON::XS" , 1);
1695 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2061 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1696
1697 json_true = get_bool ("JSON::XS::true");
1698 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);
1699 2067
1700 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 */
1701} 2069}
1702 2070
1703PROTOTYPES: DISABLE 2071PROTOTYPES: DISABLE
1704 2072
1705void CLONE (...) 2073void CLONE (...)
1706 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.
1707 json_stash = 0; 2078 json_stash = 0;
1708 json_boolean_stash = 0; 2079 bool_stash = 0;
2080 bool_false = 0;
2081 bool_true = 0;
1709 2082
1710void new (char *klass) 2083void new (char *klass)
1711 PPCODE: 2084 PPCODE:
1712{ 2085{
1713 SV *pv = NEWSV (0, sizeof (JSON)); 2086 SV *pv = NEWSV (0, sizeof (JSON));
1714 SvPOK_only (pv); 2087 SvPOK_only (pv);
1715 json_init ((JSON *)SvPVX (pv)); 2088 json_init ((JSON *)SvPVX (pv));
1716 XPUSHs (sv_2mortal (sv_bless ( 2089 XPUSHs (sv_2mortal (sv_bless (
1717 newRV_noinc (pv), 2090 newRV_noinc (pv),
1718 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2091 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1719 ))); 2092 )));
1720} 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 }
1721 2109
1722void ascii (JSON *self, int enable = 1) 2110void ascii (JSON *self, int enable = 1)
1723 ALIAS: 2111 ALIAS:
1724 ascii = F_ASCII 2112 ascii = F_ASCII
1725 latin1 = F_LATIN1 2113 latin1 = F_LATIN1
1733 shrink = F_SHRINK 2121 shrink = F_SHRINK
1734 allow_blessed = F_ALLOW_BLESSED 2122 allow_blessed = F_ALLOW_BLESSED
1735 convert_blessed = F_CONV_BLESSED 2123 convert_blessed = F_CONV_BLESSED
1736 relaxed = F_RELAXED 2124 relaxed = F_RELAXED
1737 allow_unknown = F_ALLOW_UNKNOWN 2125 allow_unknown = F_ALLOW_UNKNOWN
2126 allow_tags = F_ALLOW_TAGS
1738 PPCODE: 2127 PPCODE:
1739{ 2128{
1740 if (enable) 2129 if (enable)
1741 self->flags |= ix; 2130 self->flags |= ix;
1742 else 2131 else
1758 get_shrink = F_SHRINK 2147 get_shrink = F_SHRINK
1759 get_allow_blessed = F_ALLOW_BLESSED 2148 get_allow_blessed = F_ALLOW_BLESSED
1760 get_convert_blessed = F_CONV_BLESSED 2149 get_convert_blessed = F_CONV_BLESSED
1761 get_relaxed = F_RELAXED 2150 get_relaxed = F_RELAXED
1762 get_allow_unknown = F_ALLOW_UNKNOWN 2151 get_allow_unknown = F_ALLOW_UNKNOWN
2152 get_allow_tags = F_ALLOW_TAGS
1763 PPCODE: 2153 PPCODE:
1764 XPUSHs (boolSV (self->flags & ix)); 2154 XPUSHs (boolSV (self->flags & ix));
1765 2155
1766void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2156void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1767 PPCODE: 2157 PPCODE:
1795} 2185}
1796 2186
1797void 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)
1798 PPCODE: 2188 PPCODE:
1799{ 2189{
1800 if (!self->cb_sk_object) 2190 if (!self->cb_sk_object)
1801 self->cb_sk_object = newHV (); 2191 self->cb_sk_object = newHV ();
1802 2192
1803 if (SvOK (cb)) 2193 if (SvOK (cb))
1804 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2194 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1805 else 2195 else
1816 XPUSHs (ST (0)); 2206 XPUSHs (ST (0));
1817} 2207}
1818 2208
1819void encode (JSON *self, SV *scalar) 2209void encode (JSON *self, SV *scalar)
1820 PPCODE: 2210 PPCODE:
1821 XPUSHs (encode_json (scalar, self)); 2211 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2212 XPUSHs (scalar);
1822 2213
1823void decode (JSON *self, SV *jsonstr) 2214void decode (JSON *self, SV *jsonstr)
1824 PPCODE: 2215 PPCODE:
1825 XPUSHs (decode_json (jsonstr, self, 0)); 2216 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2217 XPUSHs (jsonstr);
1826 2218
1827void decode_prefix (JSON *self, SV *jsonstr) 2219void decode_prefix (JSON *self, SV *jsonstr)
1828 PPCODE: 2220 PPCODE:
1829{ 2221{
1830 char *offset; 2222 SV *sv;
2223 STRLEN offset;
2224 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1831 EXTEND (SP, 2); 2225 EXTEND (SP, 2);
1832 PUSHs (decode_json (jsonstr, self, &offset)); 2226 PUSHs (sv);
1833 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2227 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1834} 2228}
1835 2229
1836void incr_parse (JSON *self, SV *jsonstr = 0) 2230void incr_parse (JSON *self, SV *jsonstr = 0)
1837 PPCODE: 2231 PPCODE:
1838{ 2232{
1839 if (!self->incr_text) 2233 if (!self->incr_text)
1840 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 }
1841 2254
1842 // append data, if any 2255 // append data, if any
1843 if (jsonstr) 2256 if (jsonstr)
1844 { 2257 {
2258 /* make sure both strings have same encoding */
2259 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1845 if (SvUTF8 (jsonstr)) 2260 if (SvUTF8 (jsonstr))
2261 sv_utf8_downgrade (jsonstr, 0);
1846 { 2262 else
1847 if (!SvUTF8 (self->incr_text))
1848 {
1849 /* utf-8-ness differs, need to upgrade */
1850 sv_utf8_upgrade (self->incr_text);
1851
1852 if (self->incr_pos)
1853 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1854 - (U8 *)SvPVX (self->incr_text);
1855 }
1856 }
1857 else if (SvUTF8 (self->incr_text))
1858 sv_utf8_upgrade (jsonstr); 2263 sv_utf8_upgrade (jsonstr);
1859 2264
2265 /* and then just blindly append */
1860 { 2266 {
1861 STRLEN len; 2267 STRLEN len;
1862 const char *str = SvPV (jsonstr, len); 2268 const char *str = SvPV (jsonstr, len);
1863 STRLEN cur = SvCUR (self->incr_text); 2269 STRLEN cur = SvCUR (self->incr_text);
1864 2270
1865 if (SvLEN (self->incr_text) <= cur + len) 2271 if (SvLEN (self->incr_text) - cur <= len)
1866 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2272 json_sv_grow (self->incr_text, cur, len);
1867 2273
1868 Move (str, SvEND (self->incr_text), len, char); 2274 Move (str, SvEND (self->incr_text), len, char);
1869 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2275 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1870 *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
1871 } 2277 }
1872 } 2278 }
1873 2279
1874 if (GIMME_V != G_VOID) 2280 if (GIMME_V != G_VOID)
1875 do 2281 do
1876 { 2282 {
2283 SV *sv;
1877 char *offset; 2284 STRLEN offset;
1878 2285
1879 if (!INCR_DONE (self)) 2286 if (!INCR_DONE (self))
1880 { 2287 {
1881 incr_parse (self); 2288 incr_parse (self);
1882 2289
1883 if (self->incr_pos > self->max_size && self->max_size) 2290 if (self->incr_pos > self->max_size && self->max_size)
1884 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",
1885 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2292 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1886 2293
1887 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
1888 break; 2303 break;
2304 }
1889 } 2305 }
1890 2306
1891 XPUSHs (decode_json (self->incr_text, self, &offset)); 2307 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2308 XPUSHs (sv);
1892 2309
1893 self->incr_pos -= offset - SvPVX (self->incr_text); 2310 self->incr_pos -= offset;
1894 self->incr_nest = 0; 2311 self->incr_nest = 0;
1895 self->incr_mode = 0; 2312 self->incr_mode = 0;
1896 2313
1897 sv_chop (self->incr_text, offset); 2314 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1898 } 2315 }
1899 while (GIMME_V == G_ARRAY); 2316 while (GIMME_V == G_ARRAY);
1900} 2317}
1901 2318
1902SV *incr_text (JSON *self) 2319SV *incr_text (JSON *self)
1933 self->incr_mode = 0; 2350 self->incr_mode = 0;
1934} 2351}
1935 2352
1936void DESTROY (JSON *self) 2353void DESTROY (JSON *self)
1937 CODE: 2354 CODE:
2355 SvREFCNT_dec (self->v_false);
2356 SvREFCNT_dec (self->v_true);
1938 SvREFCNT_dec (self->cb_sk_object); 2357 SvREFCNT_dec (self->cb_sk_object);
1939 SvREFCNT_dec (self->cb_object); 2358 SvREFCNT_dec (self->cb_object);
1940 SvREFCNT_dec (self->incr_text); 2359 SvREFCNT_dec (self->incr_text);
1941 2360
1942PROTOTYPES: ENABLE 2361PROTOTYPES: ENABLE
1943 2362
1944void encode_json (SV *scalar) 2363void encode_json (SV *scalar)
1945 ALIAS:
1946 to_json_ = 0
1947 encode_json = F_UTF8
1948 PPCODE: 2364 PPCODE:
1949{ 2365{
1950 JSON json; 2366 JSON json;
1951 json_init (&json); 2367 json_init (&json);
1952 json.flags |= ix; 2368 json.flags |= F_UTF8;
1953 XPUSHs (encode_json (scalar, &json)); 2369 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2370 XPUSHs (scalar);
1954} 2371}
1955 2372
1956void decode_json (SV *jsonstr) 2373void decode_json (SV *jsonstr)
1957 ALIAS:
1958 from_json_ = 0
1959 decode_json = F_UTF8
1960 PPCODE: 2374 PPCODE:
1961{ 2375{
1962 JSON json; 2376 JSON json;
1963 json_init (&json); 2377 json_init (&json);
1964 json.flags |= ix; 2378 json.flags |= F_UTF8;
1965 XPUSHs (decode_json (jsonstr, &json, 0)); 2379 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2380 XPUSHs (jsonstr);
1966} 2381}
1967 2382
1968

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines