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 1.105 by root, Tue Jan 19 01:07:27 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
192///////////////////////////////////////////////////////////////////////////// 243/////////////////////////////////////////////////////////////////////////////
193// fp hell 244// fp hell
194 245
195// scan a group of digits, and a trailing exponent 246// scan a group of digits, and a trailing exponent
196static void 247static void
197json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp) 248json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
198{ 249{
199 UV uaccum = 0; 250 UV uaccum = 0;
200 int eaccum = 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;
201 258
202 for (;;) 259 for (;;)
203 { 260 {
204 U8 dig = (U8)*s - '0'; 261 U8 dig = (U8)*s - '0';
205 262
206 if (expect_false (dig >= 10)) 263 if (expect_false (dig >= 10))
207 { 264 {
208 if (dig == (U8)((U8)'.' - (U8)'0')) 265 if (dig == (U8)((U8)'.' - (U8)'0'))
209 { 266 {
210 ++s; 267 ++s;
211 json_atof_scan1 (s, accum, expo, 1); 268 json_atof_scan1 (s, accum, expo, 1, maxdepth);
212 } 269 }
213 else if ((dig | ' ') == 'e' - '0') 270 else if ((dig | ' ') == 'e' - '0')
214 { 271 {
215 int exp2 = 0; 272 int exp2 = 0;
216 int neg = 0; 273 int neg = 0;
242 // if we have too many digits, then recurse for more 299 // if we have too many digits, then recurse for more
243 // we actually do this for rather few digits 300 // we actually do this for rather few digits
244 if (uaccum >= (UV_MAX - 9) / 10) 301 if (uaccum >= (UV_MAX - 9) / 10)
245 { 302 {
246 if (postdp) *expo -= eaccum; 303 if (postdp) *expo -= eaccum;
247 json_atof_scan1 (s, accum, expo, postdp); 304 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
248 if (postdp) *expo += eaccum; 305 if (postdp) *expo += eaccum;
249 306
250 break; 307 break;
251 } 308 }
252 } 309 }
253 310
254 // this relies greatly on the quality of the pow () 311 // this relies greatly on the quality of the pow ()
255 // implementation of the platform, but a good 312 // implementation of the platform, but a good
256 // implementation is hard to beat. 313 // implementation is hard to beat.
314 // (IEEE 754 conformant ones are required to be exact)
257 if (postdp) *expo -= eaccum; 315 if (postdp) *expo -= eaccum;
258 *accum += uaccum * Perl_pow (10., *expo); 316 *accum += uaccum * Perl_pow (10., *expo);
259 *expo += eaccum; 317 *expo += eaccum;
260} 318}
261 319
270 { 328 {
271 ++s; 329 ++s;
272 neg = 1; 330 neg = 1;
273 } 331 }
274 332
333 // a recursion depth of ten gives us >>500 bits
275 json_atof_scan1 (s, &accum, &expo, 0); 334 json_atof_scan1 (s, &accum, &expo, 0, 10);
276 335
277 return neg ? -accum : accum; 336 return neg ? -accum : accum;
278} 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
279///////////////////////////////////////////////////////////////////////////// 381/////////////////////////////////////////////////////////////////////////////
280// encoder 382// encoder
281 383
282// structure used for encoding JSON 384// structure used for encoding JSON
283typedef struct 385typedef struct
291} enc_t; 393} enc_t;
292 394
293INLINE void 395INLINE void
294need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
295{ 397{
296 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
297 { 399 {
298 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
299 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
300 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
301 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
302 } 404 }
303} 405}
304 406
305INLINE void 407INLINE void
306encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
322 424
323 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
324 { 426 {
325 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
326 { 428 {
327 need (enc, len += 1); 429 need (enc, len + 1);
328 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
329 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
330 } 432 }
331 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
332 { 434 {
333 need (enc, len += 1); 435 need (enc, len + 1);
334 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
335 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
336 } 438 }
337 else 439 else
338 *enc->cur++ = ch; 440 *enc->cur++ = ch;
341 } 443 }
342 else 444 else
343 { 445 {
344 switch (ch) 446 switch (ch)
345 { 447 {
346 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;
347 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;
348 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;
349 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;
350 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;
351 453
352 default: 454 default:
353 { 455 {
354 STRLEN clen; 456 STRLEN clen;
355 UV uch; 457 UV uch;
371 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
372 { 474 {
373 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
374 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);
375 477
376 need (enc, len += 11); 478 need (enc, len + 11);
377 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
378 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
379 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
380 enc->cur += 12; 482 enc->cur += 12;
381 } 483 }
382 else 484 else
383 { 485 {
384 need (enc, len += 5); 486 need (enc, len + 5);
385 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
386 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
387 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
388 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
389 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
397 *enc->cur++ = uch; 499 *enc->cur++ = uch;
398 str += clen; 500 str += clen;
399 } 501 }
400 else if (is_utf8) 502 else if (is_utf8)
401 { 503 {
402 need (enc, len += clen); 504 need (enc, len + clen);
403 do 505 do
404 { 506 {
405 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
406 } 508 }
407 while (--clen); 509 while (--clen);
408 } 510 }
409 else 511 else
410 { 512 {
411 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
412 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
413 ++str; 515 ++str;
414 } 516 }
415 } 517 }
416 } 518 }
470 572
471 if (enc->indent >= enc->json.max_depth) 573 if (enc->indent >= enc->json.max_depth)
472 croak (ERR_NESTING_EXCEEDED); 574 croak (ERR_NESTING_EXCEEDED);
473 575
474 encode_ch (enc, '['); 576 encode_ch (enc, '[');
475 577
476 if (len >= 0) 578 if (len >= 0)
477 { 579 {
478 encode_nl (enc); ++enc->indent; 580 encode_nl (enc); ++enc->indent;
479 581
480 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
492 encode_comma (enc); 594 encode_comma (enc);
493 } 595 }
494 596
495 encode_nl (enc); --enc->indent; encode_indent (enc); 597 encode_nl (enc); --enc->indent; encode_indent (enc);
496 } 598 }
497 599
498 encode_ch (enc, ']'); 600 encode_ch (enc, ']');
499} 601}
500 602
501static void 603static void
502encode_hk (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
506 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
507 { 609 {
508 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
509 STRLEN len; 611 STRLEN len;
510 char *str; 612 char *str;
511 613
512 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
513 str = SvPV (sv, len); 615 str = SvPV (sv, len);
514 616
515 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
516 } 618 }
582 } 684 }
583 685
584 if (count) 686 if (count)
585 { 687 {
586 int i, fast = 1; 688 int i, fast = 1;
587#if defined(__BORLANDC__) || defined(_MSC_VER) 689 HE *hes_stack [STACK_HES];
588 HE **hes = _alloca (count * sizeof (HE)); 690 HE **hes = hes_stack;
589#else 691
590 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 692 // allocate larger arrays on the heap
591#endif 693 if (count > STACK_HES)
694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
592 698
593 i = 0; 699 i = 0;
594 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
595 { 701 {
596 hes [i++] = he; 702 hes [i++] = he;
665// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
666static void 772static void
667encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
668{ 774{
669 svtype svt; 775 svtype svt;
776 GV *method;
670 777
671 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
672 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
673 780
674 if (expect_false (SvOBJECT (sv))) 781 if (expect_false (SvOBJECT (sv)))
675 { 782 {
676 HV *stash = !JSON_SLOW || json_boolean_stash 783 HV *stash = SvSTASH (sv);
677 ? json_boolean_stash
678 : gv_stashpv ("JSON::XS::Boolean", 1);
679 784
680 if (SvSTASH (sv) == stash) 785 if (stash == bool_stash)
681 { 786 {
682 if (SvIV (sv))
683 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
684 else
685 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
686 } 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);
687 else 862 else
688 {
689#if 0
690 if (0 && sv_derived_from (rv, "JSON::Literal"))
691 {
692 // not yet
693 }
694#endif
695 if (enc->json.flags & F_CONV_BLESSED)
696 {
697 // we re-bless the reference to get overload and other niceties right
698 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
699
700 if (to_json)
701 {
702 dSP;
703
704 ENTER; SAVETMPS; PUSHMARK (SP);
705 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
706
707 // calling with G_SCALAR ensures that we always get a 1 return value
708 PUTBACK;
709 call_sv ((SV *)GvCV (to_json), G_SCALAR);
710 SPAGAIN;
711
712 // catch this surprisingly common error
713 if (SvROK (TOPs) && SvRV (TOPs) == sv)
714 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
715
716 sv = POPs;
717 PUTBACK;
718
719 encode_sv (enc, sv);
720
721 FREETMPS; LEAVE;
722 }
723 else if (enc->json.flags & F_ALLOW_BLESSED)
724 encode_str (enc, "null", 4, 0);
725 else
726 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
727 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
728 }
729 else if (enc->json.flags & F_ALLOW_BLESSED)
730 encode_str (enc, "null", 4, 0);
731 else
732 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)",
733 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 864 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
734 }
735 } 865 }
736 else if (svt == SVt_PVHV) 866 else if (svt == SVt_PVHV)
737 encode_hv (enc, (HV *)sv); 867 encode_hv (enc, (HV *)sv);
738 else if (svt == SVt_PVAV) 868 else if (svt == SVt_PVAV)
739 encode_av (enc, (AV *)sv); 869 encode_av (enc, (AV *)sv);
740 else if (svt < SVt_PVAV) 870 else if (svt < SVt_PVAV)
741 { 871 {
742 STRLEN len = 0; 872 int bool_type = ref_bool_type (sv);
743 char *pv = svt ? SvPV (sv, len) : 0;
744 873
745 if (len == 1 && *pv == '1') 874 if (bool_type == 1)
746 encode_str (enc, "true", 4, 0); 875 encode_str (enc, "true", 4, 0);
747 else if (len == 1 && *pv == '0') 876 else if (bool_type == 0)
748 encode_str (enc, "false", 5, 0); 877 encode_str (enc, "false", 5, 0);
749 else if (enc->json.flags & F_ALLOW_UNKNOWN) 878 else if (enc->json.flags & F_ALLOW_UNKNOWN)
750 encode_str (enc, "null", 4, 0); 879 encode_str (enc, "null", 4, 0);
751 else 880 else
752 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",
815 } 944 }
816 else 945 else
817 { 946 {
818 // large integer, use the (rather slow) snprintf way. 947 // large integer, use the (rather slow) snprintf way.
819 need (enc, IVUV_MAXCHARS); 948 need (enc, IVUV_MAXCHARS);
820 enc->cur += 949 enc->cur +=
821 SvIsUV(sv) 950 SvIsUV(sv)
822 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 951 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
823 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 952 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
824 } 953 }
825 } 954 }
826 else if (SvROK (sv)) 955 else if (SvROK (sv))
827 encode_rv (enc, SvRV (sv)); 956 encode_rv (enc, SvRV (sv));
828 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 957 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
829 encode_str (enc, "null", 4, 0); 958 encode_str (enc, "null", 4, 0);
830 else 959 else
831 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",
832 SvPV_nolen (sv), SvFLAGS (sv)); 961 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
833} 962}
834 963
835static SV * 964static SV *
836encode_json (SV *scalar, JSON *json) 965encode_json (SV *scalar, JSON *json)
837{ 966{
838 enc_t enc; 967 enc_t enc;
839 968
840 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 969 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
841 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)");
842 971
843 enc.json = *json; 972 enc.json = *json;
844 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 973 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
845 enc.cur = SvPVX (enc.sv); 974 enc.cur = SvPVX (enc.sv);
1052 *cur++ = *dec_cur++; 1181 *cur++ = *dec_cur++;
1053 while (--clen); 1182 while (--clen);
1054 1183
1055 utf8 = 1; 1184 utf8 = 1;
1056 } 1185 }
1186 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1187 *cur++ = ch;
1057 else 1188 else
1058 { 1189 {
1059 --dec_cur; 1190 --dec_cur;
1060 1191
1061 if (!ch) 1192 if (!ch)
1071 1202
1072 if (sv) 1203 if (sv)
1073 { 1204 {
1074 STRLEN cur = SvCUR (sv); 1205 STRLEN cur = SvCUR (sv);
1075 1206
1076 if (SvLEN (sv) <= cur + len) 1207 if (SvLEN (sv) - cur <= len)
1077 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1208 json_sv_grow (sv, cur, len);
1078 1209
1079 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1210 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
1080 SvCUR_set (sv, SvCUR (sv) + len); 1211 SvCUR_set (sv, SvCUR (sv) + len);
1081 } 1212 }
1082 else 1213 else
1249 if (*dec->cur == ']') 1380 if (*dec->cur == ']')
1250 { 1381 {
1251 ++dec->cur; 1382 ++dec->cur;
1252 break; 1383 break;
1253 } 1384 }
1254 1385
1255 if (*dec->cur != ',') 1386 if (*dec->cur != ',')
1256 ERR (", or ] expected while parsing array"); 1387 ERR (", or ] expected while parsing array");
1257 1388
1258 ++dec->cur; 1389 ++dec->cur;
1259 1390
1373 1504
1374 DEC_DEC_DEPTH; 1505 DEC_DEC_DEPTH;
1375 sv = newRV_noinc ((SV *)hv); 1506 sv = newRV_noinc ((SV *)hv);
1376 1507
1377 // check filter callbacks 1508 // check filter callbacks
1378 if (dec->json.flags & F_HOOK) 1509 if (expect_false (dec->json.flags & F_HOOK))
1379 { 1510 {
1380 if (dec->json.cb_sk_object && HvKEYS (hv) == 1) 1511 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1381 { 1512 {
1382 HE *cb, *he; 1513 HE *cb, *he;
1383 1514
1384 hv_iterinit (hv); 1515 hv_iterinit (hv);
1385 he = hv_iternext (hv); 1516 he = hv_iternext (hv);
1386 hv_iterinit (hv); 1517 hv_iterinit (hv);
1387 1518
1388 // the next line creates a mortal sv each time its called. 1519 // the next line creates a mortal sv each time it's called.
1389 // might want to optimise this for common cases. 1520 // might want to optimise this for common cases.
1390 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);
1391 1522
1392 if (cb) 1523 if (cb)
1393 { 1524 {
1394 dSP; 1525 dSP;
1395 int count; 1526 int count;
1396 1527
1397 ENTER; SAVETMPS; PUSHMARK (SP); 1528 ENTER; SAVETMPS;
1529 PUSHMARK (SP);
1398 XPUSHs (HeVAL (he)); 1530 XPUSHs (HeVAL (he));
1531 sv_2mortal (sv);
1399 1532
1400 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1533 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1401 1534
1402 if (count == 1) 1535 if (count == 1)
1403 { 1536 {
1404 sv = newSVsv (POPs); 1537 sv = newSVsv (POPs);
1405 FREETMPS; LEAVE; 1538 FREETMPS; LEAVE;
1406 return sv; 1539 return sv;
1407 } 1540 }
1541 else if (count)
1542 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1408 1543
1544 SvREFCNT_inc (sv);
1409 FREETMPS; LEAVE; 1545 FREETMPS; LEAVE;
1410 } 1546 }
1411 } 1547 }
1412 1548
1413 if (dec->json.cb_object) 1549 if (dec->json.cb_object)
1414 { 1550 {
1415 dSP; 1551 dSP;
1416 int count; 1552 int count;
1417 1553
1418 ENTER; SAVETMPS; PUSHMARK (SP); 1554 ENTER; SAVETMPS;
1555 PUSHMARK (SP);
1419 XPUSHs (sv_2mortal (sv)); 1556 XPUSHs (sv_2mortal (sv));
1420 1557
1421 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1558 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1422 1559
1423 if (count == 1) 1560 if (count == 1)
1424 {
1425 sv = newSVsv (POPs); 1561 sv = newSVsv (POPs);
1426 FREETMPS; LEAVE; 1562 else if (count == 0)
1427 return sv;
1428 }
1429
1430 SvREFCNT_inc (sv); 1563 SvREFCNT_inc (sv);
1564 else
1565 croak ("filter_json_object callbacks must not return more than one scalar");
1566
1431 FREETMPS; LEAVE; 1567 FREETMPS; LEAVE;
1432 } 1568 }
1433 } 1569 }
1434 1570
1435 return sv; 1571 return sv;
1439 DEC_DEC_DEPTH; 1575 DEC_DEC_DEPTH;
1440 return 0; 1576 return 0;
1441} 1577}
1442 1578
1443static 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 *
1444decode_sv (dec_t *dec) 1663decode_sv (dec_t *dec)
1445{ 1664{
1446 // the beauty of JSON: you need exactly one character lookahead 1665 // the beauty of JSON: you need exactly one character lookahead
1447 // to parse everything. 1666 // to parse everything.
1448 switch (*dec->cur) 1667 switch (*dec->cur)
1449 { 1668 {
1450 case '"': ++dec->cur; return decode_str (dec); 1669 case '"': ++dec->cur; return decode_str (dec);
1451 case '[': ++dec->cur; return decode_av (dec); 1670 case '[': ++dec->cur; return decode_av (dec);
1452 case '{': ++dec->cur; return decode_hv (dec); 1671 case '{': ++dec->cur; return decode_hv (dec);
1672 case '(': return decode_tag (dec);
1453 1673
1454 case '-': 1674 case '-':
1455 case '0': case '1': case '2': case '3': case '4': 1675 case '0': case '1': case '2': case '3': case '4':
1456 case '5': case '6': case '7': case '8': case '9': 1676 case '5': case '6': case '7': case '8': case '9':
1457 return decode_num (dec); 1677 return decode_num (dec);
1458 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
1459 case 't': 1694 case 't':
1460 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1695 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1461 { 1696 {
1462 dec->cur += 4; 1697 dec->cur += 4;
1463#if JSON_SLOW 1698
1464 json_true = get_bool ("JSON::XS::true"); 1699 if (expect_false (!dec->json.v_true))
1465#endif 1700 dec->json.v_true = GET_BOOL (true);
1701
1466 return newSVsv (json_true); 1702 return newSVsv (dec->json.v_true);
1467 } 1703 }
1468 else 1704 else
1469 ERR ("'true' expected"); 1705 ERR ("'true' expected");
1470
1471 break;
1472
1473 case 'f':
1474 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1475 {
1476 dec->cur += 5;
1477#if JSON_SLOW
1478 json_false = get_bool ("JSON::XS::false");
1479#endif
1480 return newSVsv (json_false);
1481 }
1482 else
1483 ERR ("'false' expected");
1484 1706
1485 break; 1707 break;
1486 1708
1487 case 'n': 1709 case 'n':
1488 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1710 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1494 ERR ("'null' expected"); 1716 ERR ("'null' expected");
1495 1717
1496 break; 1718 break;
1497 1719
1498 default: 1720 default:
1499 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1721 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1500 break; 1722 break;
1501 } 1723 }
1502 1724
1503fail: 1725fail:
1504 return 0; 1726 return 0;
1505} 1727}
1506 1728
1507static SV * 1729static SV *
1508decode_json (SV *string, JSON *json, char **offset_return) 1730decode_json (SV *string, JSON *json, STRLEN *offset_return)
1509{ 1731{
1510 dec_t dec; 1732 dec_t dec;
1511 SV *sv; 1733 SV *sv;
1512 1734
1513 /* work around bugs in 5.10 where manipulating magic values 1735 /* work around bugs in 5.10 where manipulating magic values
1514 * 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.
1515 */ 1742 */
1516 /*SvGETMAGIC (string);*/ 1743 /*SvGETMAGIC (string);*/
1517 if (SvMAGICAL (string)) 1744 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1518 string = sv_2mortal (newSVsv (string)); 1745 string = sv_2mortal (newSVsv (string));
1519 1746
1520 SvUPGRADE (string, SVt_PV); 1747 SvUPGRADE (string, SVt_PV);
1521 1748
1522 /* 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
1539 if (offset > json->max_size && json->max_size) 1766 if (offset > json->max_size && json->max_size)
1540 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",
1541 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1768 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1542 } 1769 }
1543 1770
1544 if (json->flags & F_UTF8) 1771 if (DECODE_WANTS_OCTETS (json))
1545 sv_utf8_downgrade (string, 0); 1772 sv_utf8_downgrade (string, 0);
1546 else 1773 else
1547 sv_utf8_upgrade (string); 1774 sv_utf8_upgrade (string);
1548 1775
1549 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1776 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1561 1788
1562 decode_ws (&dec); 1789 decode_ws (&dec);
1563 sv = decode_sv (&dec); 1790 sv = decode_sv (&dec);
1564 1791
1565 if (offset_return) 1792 if (offset_return)
1566 *offset_return = dec.cur; 1793 *offset_return = dec.cur - SvPVX (string);
1567 1794 else if (sv)
1568 if (!(offset_return || !sv))
1569 { 1795 {
1570 // check for trailing garbage 1796 // check for trailing garbage
1571 decode_ws (&dec); 1797 decode_ws (&dec);
1572 1798
1573 if (*dec.cur) 1799 if (dec.cur != dec.end)
1574 { 1800 {
1575 dec.err = "garbage after JSON object"; 1801 dec.err = "garbage after JSON object";
1576 SvREFCNT_dec (sv); 1802 SvREFCNT_dec (sv);
1577 sv = 0; 1803 sv = 0;
1578 } 1804 }
1591 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);
1592 LEAVE; 1818 LEAVE;
1593 1819
1594 croak ("%s, at character offset %d (before \"%s\")", 1820 croak ("%s, at character offset %d (before \"%s\")",
1595 dec.err, 1821 dec.err,
1596 ptr_to_index (string, dec.cur), 1822 (int)ptr_to_index (string, dec.cur),
1597 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1823 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1598 } 1824 }
1599 1825
1600 sv = sv_2mortal (sv); 1826 sv = sv_2mortal (sv);
1601 1827
1602 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1828 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1603 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)");
1604 1830
1605 return sv; 1831 return sv;
1606} 1832}
1607 1833
1616 // 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
1617 // but this would make it slower, so... 1843 // but this would make it slower, so...
1618 1844
1619 for (;;) 1845 for (;;)
1620 { 1846 {
1621 //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
1622 switch (self->incr_mode) 1847 switch (self->incr_mode)
1623 { 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
1624 // only used for initial whitespace skipping 1858 // only used for initial whitespace skipping
1625 case INCR_M_WS: 1859 case INCR_M_WS:
1626 for (;;) 1860 for (;;)
1627 { 1861 {
1628 if (*p > 0x20) 1862 if (*p > 0x20)
1670 ++p; 1904 ++p;
1671 } 1905 }
1672 1906
1673 break; 1907 break;
1674 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
1675 // inside a string 1943 // inside a string
1676 case INCR_M_STR: 1944 case INCR_M_STR:
1677 incr_m_str: 1945 incr_m_str:
1678 for (;;) 1946 for (;;)
1679 { 1947 {
1680 if (*p == '"') 1948 if (*p == '"')
1681 { 1949 {
1682 ++p; 1950 ++p;
1683 self->incr_mode = INCR_M_JSON;
1684
1685 if (!self->incr_nest)
1686 goto interrupt;
1687
1688 goto incr_m_json; 1951 goto end_of_scalar;
1689 } 1952 }
1690 else if (*p == '\\') 1953 else if (*p == '\\')
1691 { 1954 {
1692 ++p; // "virtually" consumes character after \ 1955 ++p; // "virtually" consumes character after \
1693 1956
1723 --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
1724 goto interrupt; 1987 goto interrupt;
1725 } 1988 }
1726 break; 1989 break;
1727 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
1728 case '"': 2006 case '"':
1729 self->incr_mode = INCR_M_STR; 2007 self->incr_mode = INCR_M_STR;
1730 goto incr_m_str; 2008 goto incr_m_str;
1731 2009
1732 case '[': 2010 case '[':
1733 case '{': 2011 case '{':
2012 case '(':
1734 if (++self->incr_nest > self->max_depth) 2013 if (++self->incr_nest > self->max_depth)
1735 croak (ERR_NESTING_EXCEEDED); 2014 croak (ERR_NESTING_EXCEEDED);
1736 break; 2015 break;
1737 2016
1738 case ']': 2017 case ']':
1739 case '}': 2018 case '}':
1740 if (--self->incr_nest <= 0) 2019 if (--self->incr_nest <= 0)
1741 goto interrupt; 2020 goto interrupt;
2021 break;
2022
2023 case ')':
2024 --self->incr_nest;
1742 break; 2025 break;
1743 2026
1744 case '#': 2027 case '#':
1745 self->incr_mode = INCR_M_C1; 2028 self->incr_mode = INCR_M_C1;
1746 goto incr_m_c; 2029 goto incr_m_c;
1772 i >= '0' && i <= '9' ? i - '0' 2055 i >= '0' && i <= '9' ? i - '0'
1773 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2056 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1774 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2057 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1775 : -1; 2058 : -1;
1776 2059
1777 json_stash = gv_stashpv ("JSON::XS" , 1); 2060 json_stash = gv_stashpv ("JSON::XS" , 1);
1778 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2061 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1779
1780 json_true = get_bool ("JSON::XS::true");
1781 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);
1782 2067
1783 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 */
1784} 2069}
1785 2070
1786PROTOTYPES: DISABLE 2071PROTOTYPES: DISABLE
1787 2072
1788void CLONE (...) 2073void CLONE (...)
1789 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.
1790 json_stash = 0; 2078 json_stash = 0;
1791 json_boolean_stash = 0; 2079 bool_stash = 0;
2080 bool_false = 0;
2081 bool_true = 0;
1792 2082
1793void new (char *klass) 2083void new (char *klass)
1794 PPCODE: 2084 PPCODE:
1795{ 2085{
1796 SV *pv = NEWSV (0, sizeof (JSON)); 2086 SV *pv = NEWSV (0, sizeof (JSON));
1797 SvPOK_only (pv); 2087 SvPOK_only (pv);
1798 json_init ((JSON *)SvPVX (pv)); 2088 json_init ((JSON *)SvPVX (pv));
1799 XPUSHs (sv_2mortal (sv_bless ( 2089 XPUSHs (sv_2mortal (sv_bless (
1800 newRV_noinc (pv), 2090 newRV_noinc (pv),
1801 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2091 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1802 ))); 2092 )));
1803} 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 }
1804 2109
1805void ascii (JSON *self, int enable = 1) 2110void ascii (JSON *self, int enable = 1)
1806 ALIAS: 2111 ALIAS:
1807 ascii = F_ASCII 2112 ascii = F_ASCII
1808 latin1 = F_LATIN1 2113 latin1 = F_LATIN1
1816 shrink = F_SHRINK 2121 shrink = F_SHRINK
1817 allow_blessed = F_ALLOW_BLESSED 2122 allow_blessed = F_ALLOW_BLESSED
1818 convert_blessed = F_CONV_BLESSED 2123 convert_blessed = F_CONV_BLESSED
1819 relaxed = F_RELAXED 2124 relaxed = F_RELAXED
1820 allow_unknown = F_ALLOW_UNKNOWN 2125 allow_unknown = F_ALLOW_UNKNOWN
2126 allow_tags = F_ALLOW_TAGS
1821 PPCODE: 2127 PPCODE:
1822{ 2128{
1823 if (enable) 2129 if (enable)
1824 self->flags |= ix; 2130 self->flags |= ix;
1825 else 2131 else
1841 get_shrink = F_SHRINK 2147 get_shrink = F_SHRINK
1842 get_allow_blessed = F_ALLOW_BLESSED 2148 get_allow_blessed = F_ALLOW_BLESSED
1843 get_convert_blessed = F_CONV_BLESSED 2149 get_convert_blessed = F_CONV_BLESSED
1844 get_relaxed = F_RELAXED 2150 get_relaxed = F_RELAXED
1845 get_allow_unknown = F_ALLOW_UNKNOWN 2151 get_allow_unknown = F_ALLOW_UNKNOWN
2152 get_allow_tags = F_ALLOW_TAGS
1846 PPCODE: 2153 PPCODE:
1847 XPUSHs (boolSV (self->flags & ix)); 2154 XPUSHs (boolSV (self->flags & ix));
1848 2155
1849void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2156void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1850 PPCODE: 2157 PPCODE:
1878} 2185}
1879 2186
1880void 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)
1881 PPCODE: 2188 PPCODE:
1882{ 2189{
1883 if (!self->cb_sk_object) 2190 if (!self->cb_sk_object)
1884 self->cb_sk_object = newHV (); 2191 self->cb_sk_object = newHV ();
1885 2192
1886 if (SvOK (cb)) 2193 if (SvOK (cb))
1887 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2194 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1888 else 2195 else
1899 XPUSHs (ST (0)); 2206 XPUSHs (ST (0));
1900} 2207}
1901 2208
1902void encode (JSON *self, SV *scalar) 2209void encode (JSON *self, SV *scalar)
1903 PPCODE: 2210 PPCODE:
1904 XPUSHs (encode_json (scalar, self)); 2211 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2212 XPUSHs (scalar);
1905 2213
1906void decode (JSON *self, SV *jsonstr) 2214void decode (JSON *self, SV *jsonstr)
1907 PPCODE: 2215 PPCODE:
1908 XPUSHs (decode_json (jsonstr, self, 0)); 2216 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2217 XPUSHs (jsonstr);
1909 2218
1910void decode_prefix (JSON *self, SV *jsonstr) 2219void decode_prefix (JSON *self, SV *jsonstr)
1911 PPCODE: 2220 PPCODE:
1912{ 2221{
1913 char *offset; 2222 SV *sv;
2223 STRLEN offset;
2224 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1914 EXTEND (SP, 2); 2225 EXTEND (SP, 2);
1915 PUSHs (decode_json (jsonstr, self, &offset)); 2226 PUSHs (sv);
1916 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2227 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1917} 2228}
1918 2229
1919void incr_parse (JSON *self, SV *jsonstr = 0) 2230void incr_parse (JSON *self, SV *jsonstr = 0)
1920 PPCODE: 2231 PPCODE:
1921{ 2232{
1922 if (!self->incr_text) 2233 if (!self->incr_text)
1923 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 }
1924 2254
1925 // append data, if any 2255 // append data, if any
1926 if (jsonstr) 2256 if (jsonstr)
1927 { 2257 {
2258 /* make sure both strings have same encoding */
2259 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1928 if (SvUTF8 (jsonstr)) 2260 if (SvUTF8 (jsonstr))
2261 sv_utf8_downgrade (jsonstr, 0);
1929 { 2262 else
1930 if (!SvUTF8 (self->incr_text))
1931 {
1932 /* utf-8-ness differs, need to upgrade */
1933 sv_utf8_upgrade (self->incr_text);
1934
1935 if (self->incr_pos)
1936 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1937 - (U8 *)SvPVX (self->incr_text);
1938 }
1939 }
1940 else if (SvUTF8 (self->incr_text))
1941 sv_utf8_upgrade (jsonstr); 2263 sv_utf8_upgrade (jsonstr);
1942 2264
2265 /* and then just blindly append */
1943 { 2266 {
1944 STRLEN len; 2267 STRLEN len;
1945 const char *str = SvPV (jsonstr, len); 2268 const char *str = SvPV (jsonstr, len);
1946 STRLEN cur = SvCUR (self->incr_text); 2269 STRLEN cur = SvCUR (self->incr_text);
1947 2270
1948 if (SvLEN (self->incr_text) <= cur + len) 2271 if (SvLEN (self->incr_text) - cur <= len)
1949 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2272 json_sv_grow (self->incr_text, cur, len);
1950 2273
1951 Move (str, SvEND (self->incr_text), len, char); 2274 Move (str, SvEND (self->incr_text), len, char);
1952 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2275 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1953 *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
1954 } 2277 }
1955 } 2278 }
1956 2279
1957 if (GIMME_V != G_VOID) 2280 if (GIMME_V != G_VOID)
1958 do 2281 do
1959 { 2282 {
2283 SV *sv;
1960 char *offset; 2284 STRLEN offset;
1961 2285
1962 if (!INCR_DONE (self)) 2286 if (!INCR_DONE (self))
1963 { 2287 {
1964 incr_parse (self); 2288 incr_parse (self);
1965 2289
1966 if (self->incr_pos > self->max_size && self->max_size) 2290 if (self->incr_pos > self->max_size && self->max_size)
1967 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",
1968 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2292 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1969 2293
1970 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
1971 break; 2303 break;
2304 }
1972 } 2305 }
1973 2306
1974 XPUSHs (decode_json (self->incr_text, self, &offset)); 2307 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2308 XPUSHs (sv);
1975 2309
1976 self->incr_pos -= offset - SvPVX (self->incr_text); 2310 self->incr_pos -= offset;
1977 self->incr_nest = 0; 2311 self->incr_nest = 0;
1978 self->incr_mode = 0; 2312 self->incr_mode = 0;
1979 2313
1980 sv_chop (self->incr_text, offset); 2314 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1981 } 2315 }
1982 while (GIMME_V == G_ARRAY); 2316 while (GIMME_V == G_ARRAY);
1983} 2317}
1984 2318
1985SV *incr_text (JSON *self) 2319SV *incr_text (JSON *self)
2016 self->incr_mode = 0; 2350 self->incr_mode = 0;
2017} 2351}
2018 2352
2019void DESTROY (JSON *self) 2353void DESTROY (JSON *self)
2020 CODE: 2354 CODE:
2355 SvREFCNT_dec (self->v_false);
2356 SvREFCNT_dec (self->v_true);
2021 SvREFCNT_dec (self->cb_sk_object); 2357 SvREFCNT_dec (self->cb_sk_object);
2022 SvREFCNT_dec (self->cb_object); 2358 SvREFCNT_dec (self->cb_object);
2023 SvREFCNT_dec (self->incr_text); 2359 SvREFCNT_dec (self->incr_text);
2024 2360
2025PROTOTYPES: ENABLE 2361PROTOTYPES: ENABLE
2026 2362
2027void encode_json (SV *scalar) 2363void encode_json (SV *scalar)
2028 ALIAS:
2029 to_json_ = 0
2030 encode_json = F_UTF8
2031 PPCODE: 2364 PPCODE:
2032{ 2365{
2033 JSON json; 2366 JSON json;
2034 json_init (&json); 2367 json_init (&json);
2035 json.flags |= ix; 2368 json.flags |= F_UTF8;
2036 XPUSHs (encode_json (scalar, &json)); 2369 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2370 XPUSHs (scalar);
2037} 2371}
2038 2372
2039void decode_json (SV *jsonstr) 2373void decode_json (SV *jsonstr)
2040 ALIAS:
2041 from_json_ = 0
2042 decode_json = F_UTF8
2043 PPCODE: 2374 PPCODE:
2044{ 2375{
2045 JSON json; 2376 JSON json;
2046 json_init (&json); 2377 json_init (&json);
2047 json.flags |= ix; 2378 json.flags |= F_UTF8;
2048 XPUSHs (decode_json (jsonstr, &json, 0)); 2379 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2380 XPUSHs (jsonstr);
2049} 2381}
2050 2382

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines