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 1.104 by root, Tue Jan 19 00:31:13 2010 UTC vs.
Revision 1.136 by root, Thu Nov 15 22:35:35 2018 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 15#endif
15 16
16// some old perls do not have this, try to make it work, no 17// some old perls do not have this, try to make it work, no
17// 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
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)
254 if (postdp) *expo -= eaccum; 315 if (postdp) *expo -= eaccum;
255 *accum += uaccum * pow (10., *expo); 316 *accum += uaccum * Perl_pow (10., *expo);
256 *expo += eaccum; 317 *expo += eaccum;
257} 318}
258 319
259static NV 320static NV
260json_atof (const char *s) 321json_atof (const char *s)
267 { 328 {
268 ++s; 329 ++s;
269 neg = 1; 330 neg = 1;
270 } 331 }
271 332
333 // a recursion depth of ten gives us >>500 bits
272 json_atof_scan1 (s, &accum, &expo, 0); 334 json_atof_scan1 (s, &accum, &expo, 0, 10);
273 335
274 return neg ? -accum : accum; 336 return neg ? -accum : accum;
275} 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
276///////////////////////////////////////////////////////////////////////////// 381/////////////////////////////////////////////////////////////////////////////
277// encoder 382// encoder
278 383
279// structure used for encoding JSON 384// structure used for encoding JSON
280typedef struct 385typedef struct
288} enc_t; 393} enc_t;
289 394
290INLINE void 395INLINE void
291need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
292{ 397{
293 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
294 { 399 {
295 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
296 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
297 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
298 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
299 } 404 }
300} 405}
301 406
302INLINE void 407INLINE void
303encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
319 424
320 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
321 { 426 {
322 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
323 { 428 {
324 need (enc, len += 1); 429 need (enc, len + 1);
325 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
326 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
327 } 432 }
328 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
329 { 434 {
330 need (enc, len += 1); 435 need (enc, len + 1);
331 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
332 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
333 } 438 }
334 else 439 else
335 *enc->cur++ = ch; 440 *enc->cur++ = ch;
338 } 443 }
339 else 444 else
340 { 445 {
341 switch (ch) 446 switch (ch)
342 { 447 {
343 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;
344 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;
345 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;
346 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;
347 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;
348 453
349 default: 454 default:
350 { 455 {
351 STRLEN clen; 456 STRLEN clen;
352 UV uch; 457 UV uch;
368 if (uch >= 0x10000UL) 473 if (uch >= 0x10000UL)
369 { 474 {
370 if (uch >= 0x110000UL) 475 if (uch >= 0x110000UL)
371 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);
372 477
373 need (enc, len += 11); 478 need (enc, len + 11);
374 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
375 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
376 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
377 enc->cur += 12; 482 enc->cur += 12;
378 } 483 }
379 else 484 else
380 { 485 {
381 need (enc, len += 5); 486 need (enc, len + 5);
382 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
383 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
384 *enc->cur++ = PL_hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
385 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
386 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
394 *enc->cur++ = uch; 499 *enc->cur++ = uch;
395 str += clen; 500 str += clen;
396 } 501 }
397 else if (is_utf8) 502 else if (is_utf8)
398 { 503 {
399 need (enc, len += clen); 504 need (enc, len + clen);
400 do 505 do
401 { 506 {
402 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
403 } 508 }
404 while (--clen); 509 while (--clen);
405 } 510 }
406 else 511 else
407 { 512 {
408 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
409 enc->cur = encode_utf8 (enc->cur, uch); 514 enc->cur = encode_utf8 (enc->cur, uch);
410 ++str; 515 ++str;
411 } 516 }
412 } 517 }
413 } 518 }
467 572
468 if (enc->indent >= enc->json.max_depth) 573 if (enc->indent >= enc->json.max_depth)
469 croak (ERR_NESTING_EXCEEDED); 574 croak (ERR_NESTING_EXCEEDED);
470 575
471 encode_ch (enc, '['); 576 encode_ch (enc, '[');
472 577
473 if (len >= 0) 578 if (len >= 0)
474 { 579 {
475 encode_nl (enc); ++enc->indent; 580 encode_nl (enc); ++enc->indent;
476 581
477 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
489 encode_comma (enc); 594 encode_comma (enc);
490 } 595 }
491 596
492 encode_nl (enc); --enc->indent; encode_indent (enc); 597 encode_nl (enc); --enc->indent; encode_indent (enc);
493 } 598 }
494 599
495 encode_ch (enc, ']'); 600 encode_ch (enc, ']');
496} 601}
497 602
498static void 603static void
499encode_hk (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
503 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
504 { 609 {
505 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
506 STRLEN len; 611 STRLEN len;
507 char *str; 612 char *str;
508 613
509 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
510 str = SvPV (sv, len); 615 str = SvPV (sv, len);
511 616
512 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
513 } 618 }
579 } 684 }
580 685
581 if (count) 686 if (count)
582 { 687 {
583 int i, fast = 1; 688 int i, fast = 1;
584#if defined(__BORLANDC__) || defined(_MSC_VER) 689 HE *hes_stack [STACK_HES];
585 HE **hes = _alloca (count * sizeof (HE)); 690 HE **hes = hes_stack;
586#else 691
587 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 692 // allocate larger arrays on the heap
588#endif 693 if (count > STACK_HES)
694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
589 698
590 i = 0; 699 i = 0;
591 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
592 { 701 {
593 hes [i++] = he; 702 hes [i++] = he;
662// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
663static void 772static void
664encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
665{ 774{
666 svtype svt; 775 svtype svt;
776 GV *method;
667 777
668 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
669 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
670 780
671 if (expect_false (SvOBJECT (sv))) 781 if (expect_false (SvOBJECT (sv)))
672 { 782 {
673 HV *stash = !JSON_SLOW || json_boolean_stash 783 HV *stash = SvSTASH (sv);
674 ? json_boolean_stash
675 : gv_stashpv ("JSON::XS::Boolean", 1);
676 784
677 if (SvSTASH (sv) == stash) 785 if (stash == bool_stash)
678 { 786 {
679 if (SvIV (sv))
680 encode_str (enc, "true", 4, 0); 787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
681 else
682 encode_str (enc, "false", 5, 0); 788 else encode_str (enc, "false", 5, 0);
683 } 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);
684 else 862 else
685 {
686#if 0
687 if (0 && sv_derived_from (rv, "JSON::Literal"))
688 {
689 // not yet
690 }
691#endif
692 if (enc->json.flags & F_CONV_BLESSED)
693 {
694 // we re-bless the reference to get overload and other niceties right
695 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
696
697 if (to_json)
698 {
699 dSP;
700
701 ENTER; SAVETMPS; PUSHMARK (SP);
702 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
703
704 // calling with G_SCALAR ensures that we always get a 1 return value
705 PUTBACK;
706 call_sv ((SV *)GvCV (to_json), G_SCALAR);
707 SPAGAIN;
708
709 // catch this surprisingly common error
710 if (SvROK (TOPs) && SvRV (TOPs) == sv)
711 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
712
713 sv = POPs;
714 PUTBACK;
715
716 encode_sv (enc, sv);
717
718 FREETMPS; LEAVE;
719 }
720 else if (enc->json.flags & F_ALLOW_BLESSED)
721 encode_str (enc, "null", 4, 0);
722 else
723 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
724 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
725 }
726 else if (enc->json.flags & F_ALLOW_BLESSED)
727 encode_str (enc, "null", 4, 0);
728 else
729 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)",
730 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 864 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
731 }
732 } 865 }
733 else if (svt == SVt_PVHV) 866 else if (svt == SVt_PVHV)
734 encode_hv (enc, (HV *)sv); 867 encode_hv (enc, (HV *)sv);
735 else if (svt == SVt_PVAV) 868 else if (svt == SVt_PVAV)
736 encode_av (enc, (AV *)sv); 869 encode_av (enc, (AV *)sv);
737 else if (svt < SVt_PVAV) 870 else if (svt < SVt_PVAV)
738 { 871 {
739 STRLEN len = 0; 872 int bool_type = ref_bool_type (sv);
740 char *pv = svt ? SvPV (sv, len) : 0;
741 873
742 if (len == 1 && *pv == '1') 874 if (bool_type == 1)
743 encode_str (enc, "true", 4, 0); 875 encode_str (enc, "true", 4, 0);
744 else if (len == 1 && *pv == '0') 876 else if (bool_type == 0)
745 encode_str (enc, "false", 5, 0); 877 encode_str (enc, "false", 5, 0);
746 else if (enc->json.flags & F_ALLOW_UNKNOWN) 878 else if (enc->json.flags & F_ALLOW_UNKNOWN)
747 encode_str (enc, "null", 4, 0); 879 encode_str (enc, "null", 4, 0);
748 else 880 else
749 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",
812 } 944 }
813 else 945 else
814 { 946 {
815 // large integer, use the (rather slow) snprintf way. 947 // large integer, use the (rather slow) snprintf way.
816 need (enc, IVUV_MAXCHARS); 948 need (enc, IVUV_MAXCHARS);
817 enc->cur += 949 enc->cur +=
818 SvIsUV(sv) 950 SvIsUV(sv)
819 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 951 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
820 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 952 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
821 } 953 }
822 } 954 }
823 else if (SvROK (sv)) 955 else if (SvROK (sv))
824 encode_rv (enc, SvRV (sv)); 956 encode_rv (enc, SvRV (sv));
825 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 957 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
826 encode_str (enc, "null", 4, 0); 958 encode_str (enc, "null", 4, 0);
827 else 959 else
828 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",
829 SvPV_nolen (sv), SvFLAGS (sv)); 961 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
830} 962}
831 963
832static SV * 964static SV *
833encode_json (SV *scalar, JSON *json) 965encode_json (SV *scalar, JSON *json)
834{ 966{
835 enc_t enc; 967 enc_t enc;
836 968
837 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 969 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
838 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)");
839 971
840 enc.json = *json; 972 enc.json = *json;
841 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 973 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
842 enc.cur = SvPVX (enc.sv); 974 enc.cur = SvPVX (enc.sv);
1049 *cur++ = *dec_cur++; 1181 *cur++ = *dec_cur++;
1050 while (--clen); 1182 while (--clen);
1051 1183
1052 utf8 = 1; 1184 utf8 = 1;
1053 } 1185 }
1186 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1187 *cur++ = ch;
1054 else 1188 else
1055 { 1189 {
1056 --dec_cur; 1190 --dec_cur;
1057 1191
1058 if (!ch) 1192 if (!ch)
1068 1202
1069 if (sv) 1203 if (sv)
1070 { 1204 {
1071 STRLEN cur = SvCUR (sv); 1205 STRLEN cur = SvCUR (sv);
1072 1206
1073 if (SvLEN (sv) <= cur + len) 1207 if (SvLEN (sv) - cur <= len)
1074 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 1208 json_sv_grow (sv, cur, len);
1075 1209
1076 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1210 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
1077 SvCUR_set (sv, SvCUR (sv) + len); 1211 SvCUR_set (sv, SvCUR (sv) + len);
1078 } 1212 }
1079 else 1213 else
1246 if (*dec->cur == ']') 1380 if (*dec->cur == ']')
1247 { 1381 {
1248 ++dec->cur; 1382 ++dec->cur;
1249 break; 1383 break;
1250 } 1384 }
1251 1385
1252 if (*dec->cur != ',') 1386 if (*dec->cur != ',')
1253 ERR (", or ] expected while parsing array"); 1387 ERR (", or ] expected while parsing array");
1254 1388
1255 ++dec->cur; 1389 ++dec->cur;
1256 1390
1380 1514
1381 hv_iterinit (hv); 1515 hv_iterinit (hv);
1382 he = hv_iternext (hv); 1516 he = hv_iternext (hv);
1383 hv_iterinit (hv); 1517 hv_iterinit (hv);
1384 1518
1385 // the next line creates a mortal sv each time its called. 1519 // the next line creates a mortal sv each time it's called.
1386 // might want to optimise this for common cases. 1520 // might want to optimise this for common cases.
1387 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);
1388 1522
1389 if (cb) 1523 if (cb)
1390 { 1524 {
1391 dSP; 1525 dSP;
1392 int count; 1526 int count;
1393 1527
1394 ENTER; SAVETMPS; PUSHMARK (SP); 1528 ENTER; SAVETMPS;
1529 PUSHMARK (SP);
1395 XPUSHs (HeVAL (he)); 1530 XPUSHs (HeVAL (he));
1531 sv_2mortal (sv);
1396 1532
1397 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1533 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1398 1534
1399 if (count == 1) 1535 if (count == 1)
1400 { 1536 {
1401 sv = newSVsv (POPs); 1537 sv = newSVsv (POPs);
1402 FREETMPS; LEAVE; 1538 FREETMPS; LEAVE;
1403 return sv; 1539 return sv;
1404 } 1540 }
1541 else if (count)
1542 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1405 1543
1544 SvREFCNT_inc (sv);
1406 FREETMPS; LEAVE; 1545 FREETMPS; LEAVE;
1407 } 1546 }
1408 } 1547 }
1409 1548
1410 if (dec->json.cb_object) 1549 if (dec->json.cb_object)
1411 { 1550 {
1412 dSP; 1551 dSP;
1413 int count; 1552 int count;
1414 1553
1415 ENTER; SAVETMPS; PUSHMARK (SP); 1554 ENTER; SAVETMPS;
1555 PUSHMARK (SP);
1416 XPUSHs (sv_2mortal (sv)); 1556 XPUSHs (sv_2mortal (sv));
1417 1557
1418 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1558 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1419 1559
1420 if (count == 1) 1560 if (count == 1)
1421 { 1561 {
1422 sv = newSVsv (POPs); 1562 sv = newSVsv (POPs);
1423 FREETMPS; LEAVE; 1563 FREETMPS; LEAVE;
1424 return sv; 1564 return sv;
1425 } 1565 }
1566 else if (count)
1567 croak ("filter_json_object callbacks must not return more than one scalar");
1426 1568
1427 SvREFCNT_inc (sv); 1569 SvREFCNT_inc (sv);
1428 FREETMPS; LEAVE; 1570 FREETMPS; LEAVE;
1429 } 1571 }
1430 } 1572 }
1436 DEC_DEC_DEPTH; 1578 DEC_DEC_DEPTH;
1437 return 0; 1579 return 0;
1438} 1580}
1439 1581
1440static SV * 1582static SV *
1583decode_tag (dec_t *dec)
1584{
1585 SV *tag = 0;
1586 SV *val = 0;
1587
1588 if (!(dec->json.flags & F_ALLOW_TAGS))
1589 ERR ("malformed JSON string, neither array, object, number, string or atom");
1590
1591 ++dec->cur;
1592
1593 decode_ws (dec);
1594
1595 tag = decode_sv (dec);
1596 if (!tag)
1597 goto fail;
1598
1599 if (!SvPOK (tag))
1600 ERR ("malformed JSON string, (tag) must be a string");
1601
1602 decode_ws (dec);
1603
1604 if (*dec->cur != ')')
1605 ERR (") expected after tag");
1606
1607 ++dec->cur;
1608
1609 decode_ws (dec);
1610
1611 val = decode_sv (dec);
1612 if (!val)
1613 goto fail;
1614
1615 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1616 ERR ("malformed JSON string, tag value must be an array");
1617
1618 {
1619 AV *av = (AV *)SvRV (val);
1620 int i, len = av_len (av) + 1;
1621 HV *stash = gv_stashsv (tag, 0);
1622 SV *sv;
1623
1624 if (!stash)
1625 ERR ("cannot decode perl-object (package does not exist)");
1626
1627 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1628
1629 if (!method)
1630 ERR ("cannot decode perl-object (package does not have a THAW method)");
1631
1632 dSP;
1633
1634 ENTER; SAVETMPS;
1635 PUSHMARK (SP);
1636 EXTEND (SP, len + 2);
1637 // we re-bless the reference to get overload and other niceties right
1638 PUSHs (tag);
1639 PUSHs (sv_json);
1640
1641 for (i = 0; i < len; ++i)
1642 PUSHs (*av_fetch (av, i, 1));
1643
1644 PUTBACK;
1645 call_sv ((SV *)GvCV (method), G_SCALAR);
1646 SPAGAIN;
1647
1648 SvREFCNT_dec (tag);
1649 SvREFCNT_dec (val);
1650 sv = SvREFCNT_inc (POPs);
1651
1652 PUTBACK;
1653
1654 FREETMPS; LEAVE;
1655
1656 return sv;
1657 }
1658
1659fail:
1660 SvREFCNT_dec (tag);
1661 SvREFCNT_dec (val);
1662 return 0;
1663}
1664
1665static SV *
1441decode_sv (dec_t *dec) 1666decode_sv (dec_t *dec)
1442{ 1667{
1443 // the beauty of JSON: you need exactly one character lookahead 1668 // the beauty of JSON: you need exactly one character lookahead
1444 // to parse everything. 1669 // to parse everything.
1445 switch (*dec->cur) 1670 switch (*dec->cur)
1446 { 1671 {
1447 case '"': ++dec->cur; return decode_str (dec); 1672 case '"': ++dec->cur; return decode_str (dec);
1448 case '[': ++dec->cur; return decode_av (dec); 1673 case '[': ++dec->cur; return decode_av (dec);
1449 case '{': ++dec->cur; return decode_hv (dec); 1674 case '{': ++dec->cur; return decode_hv (dec);
1675 case '(': return decode_tag (dec);
1450 1676
1451 case '-': 1677 case '-':
1452 case '0': case '1': case '2': case '3': case '4': 1678 case '0': case '1': case '2': case '3': case '4':
1453 case '5': case '6': case '7': case '8': case '9': 1679 case '5': case '6': case '7': case '8': case '9':
1454 return decode_num (dec); 1680 return decode_num (dec);
1455 1681
1682 case 'f':
1683 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1684 {
1685 dec->cur += 5;
1686
1687 if (expect_false (!dec->json.v_false))
1688 dec->json.v_false = GET_BOOL (false);
1689
1690 return newSVsv (dec->json.v_false);
1691 }
1692 else
1693 ERR ("'false' expected");
1694
1695 break;
1696
1456 case 't': 1697 case 't':
1457 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1698 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1458 { 1699 {
1459 dec->cur += 4; 1700 dec->cur += 4;
1460#if JSON_SLOW 1701
1461 json_true = get_bool ("JSON::XS::true"); 1702 if (expect_false (!dec->json.v_true))
1462#endif 1703 dec->json.v_true = GET_BOOL (true);
1704
1463 return newSVsv (json_true); 1705 return newSVsv (dec->json.v_true);
1464 } 1706 }
1465 else 1707 else
1466 ERR ("'true' expected"); 1708 ERR ("'true' expected");
1467
1468 break;
1469
1470 case 'f':
1471 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1472 {
1473 dec->cur += 5;
1474#if JSON_SLOW
1475 json_false = get_bool ("JSON::XS::false");
1476#endif
1477 return newSVsv (json_false);
1478 }
1479 else
1480 ERR ("'false' expected");
1481 1709
1482 break; 1710 break;
1483 1711
1484 case 'n': 1712 case 'n':
1485 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1713 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1491 ERR ("'null' expected"); 1719 ERR ("'null' expected");
1492 1720
1493 break; 1721 break;
1494 1722
1495 default: 1723 default:
1496 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1724 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1497 break; 1725 break;
1498 } 1726 }
1499 1727
1500fail: 1728fail:
1501 return 0; 1729 return 0;
1502} 1730}
1503 1731
1504static SV * 1732static SV *
1505decode_json (SV *string, JSON *json, char **offset_return) 1733decode_json (SV *string, JSON *json, STRLEN *offset_return)
1506{ 1734{
1507 dec_t dec; 1735 dec_t dec;
1508 SV *sv; 1736 SV *sv;
1509 1737
1510 /* work around bugs in 5.10 where manipulating magic values 1738 /* work around bugs in 5.10 where manipulating magic values
1511 * will perl ignore the magic in subsequent accesses 1739 * makes perl ignore the magic in subsequent accesses.
1740 * also make a copy of non-PV values, to get them into a clean
1741 * state (SvPV should do that, but it's buggy, see below).
1742 *
1743 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1744 * as reported by Reini Urban.
1512 */ 1745 */
1513 /*SvGETMAGIC (string);*/ 1746 /*SvGETMAGIC (string);*/
1514 if (SvMAGICAL (string)) 1747 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1515 string = sv_2mortal (newSVsv (string)); 1748 string = sv_2mortal (newSVsv (string));
1516 1749
1517 SvUPGRADE (string, SVt_PV); 1750 SvUPGRADE (string, SVt_PV);
1518 1751
1519 /* work around a bug in perl 5.10, which causes SvCUR to fail an 1752 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1536 if (offset > json->max_size && json->max_size) 1769 if (offset > json->max_size && json->max_size)
1537 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1770 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1538 (unsigned long)SvCUR (string), (unsigned long)json->max_size); 1771 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1539 } 1772 }
1540 1773
1541 if (json->flags & F_UTF8) 1774 if (DECODE_WANTS_OCTETS (json))
1542 sv_utf8_downgrade (string, 0); 1775 sv_utf8_downgrade (string, 0);
1543 else 1776 else
1544 sv_utf8_upgrade (string); 1777 sv_utf8_upgrade (string);
1545 1778
1546 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1779 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1558 1791
1559 decode_ws (&dec); 1792 decode_ws (&dec);
1560 sv = decode_sv (&dec); 1793 sv = decode_sv (&dec);
1561 1794
1562 if (offset_return) 1795 if (offset_return)
1563 *offset_return = dec.cur; 1796 *offset_return = dec.cur - SvPVX (string);
1564 1797 else if (sv)
1565 if (!(offset_return || !sv))
1566 { 1798 {
1567 // check for trailing garbage 1799 // check for trailing garbage
1568 decode_ws (&dec); 1800 decode_ws (&dec);
1569 1801
1570 if (*dec.cur) 1802 if (dec.cur != dec.end)
1571 { 1803 {
1572 dec.err = "garbage after JSON object"; 1804 dec.err = "garbage after JSON object";
1573 SvREFCNT_dec (sv); 1805 SvREFCNT_dec (sv);
1574 sv = 0; 1806 sv = 0;
1575 } 1807 }
1588 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1820 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1589 LEAVE; 1821 LEAVE;
1590 1822
1591 croak ("%s, at character offset %d (before \"%s\")", 1823 croak ("%s, at character offset %d (before \"%s\")",
1592 dec.err, 1824 dec.err,
1593 ptr_to_index (string, dec.cur), 1825 (int)ptr_to_index (string, dec.cur),
1594 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1826 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1595 } 1827 }
1596 1828
1597 sv = sv_2mortal (sv); 1829 sv = sv_2mortal (sv);
1598 1830
1599 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1831 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1600 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1832 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1601 1833
1602 return sv; 1834 return sv;
1603} 1835}
1604 1836
1613 // the state machine here is a bit convoluted and could be simplified a lot 1845 // the state machine here is a bit convoluted and could be simplified a lot
1614 // but this would make it slower, so... 1846 // but this would make it slower, so...
1615 1847
1616 for (;;) 1848 for (;;)
1617 { 1849 {
1618 //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
1619 switch (self->incr_mode) 1850 switch (self->incr_mode)
1620 { 1851 {
1852 // reached end of a scalar, see if we are inside a nested structure or not
1853 end_of_scalar:
1854 self->incr_mode = INCR_M_JSON;
1855
1856 if (self->incr_nest) // end of a scalar inside array, object or tag
1857 goto incr_m_json;
1858 else // end of scalar outside structure, json text ends here
1859 goto interrupt;
1860
1621 // only used for initial whitespace skipping 1861 // only used for initial whitespace skipping
1622 case INCR_M_WS: 1862 case INCR_M_WS:
1623 for (;;) 1863 for (;;)
1624 { 1864 {
1625 if (*p > 0x20) 1865 if (*p > 0x20)
1667 ++p; 1907 ++p;
1668 } 1908 }
1669 1909
1670 break; 1910 break;
1671 1911
1912 // inside true/false/null
1913 case INCR_M_TFN:
1914 incr_m_tfn:
1915 for (;;)
1916 switch (*p++)
1917 {
1918 case 'r': case 'u': case 'e': // tRUE, falsE, nUll
1919 case 'a': case 'l': case 's': // fALSe, nuLL
1920 // allowed
1921 break;
1922
1923 default:
1924 --p;
1925 goto end_of_scalar;
1926 }
1927
1928 // inside a number
1929 case INCR_M_NUM:
1930 incr_m_num:
1931 for (;;)
1932 switch (*p++)
1933 {
1934 case 'e': case 'E': case '.': case '+':
1935 case '-':
1936 case '0': case '1': case '2': case '3': case '4':
1937 case '5': case '6': case '7': case '8': case '9':
1938 // allowed
1939 break;
1940
1941 default:
1942 --p;
1943 goto end_of_scalar;
1944 }
1945
1672 // inside a string 1946 // inside a string
1673 case INCR_M_STR: 1947 case INCR_M_STR:
1674 incr_m_str: 1948 incr_m_str:
1675 for (;;) 1949 for (;;)
1676 { 1950 {
1677 if (*p == '"') 1951 if (*p == '"')
1678 { 1952 {
1679 ++p; 1953 ++p;
1680 self->incr_mode = INCR_M_JSON;
1681
1682 if (!self->incr_nest)
1683 goto interrupt;
1684
1685 goto incr_m_json; 1954 goto end_of_scalar;
1686 } 1955 }
1687 else if (*p == '\\') 1956 else if (*p == '\\')
1688 { 1957 {
1689 ++p; // "virtually" consumes character after \ 1958 ++p; // "virtually" consumes character after \
1690 1959
1720 --p; // do not eat the whitespace, let the next round do it 1989 --p; // do not eat the whitespace, let the next round do it
1721 goto interrupt; 1990 goto interrupt;
1722 } 1991 }
1723 break; 1992 break;
1724 1993
1994 // the following three blocks handle scalars. this makes the parser
1995 // more strict than required inside arrays or objects, and could
1996 // be moved to a special case on the toplevel (except strings)
1997 case 't':
1998 case 'f':
1999 case 'n':
2000 self->incr_mode = INCR_M_TFN;
2001 goto incr_m_tfn;
2002
2003 case '-':
2004 case '0': case '1': case '2': case '3': case '4':
2005 case '5': case '6': case '7': case '8': case '9':
2006 self->incr_mode = INCR_M_NUM;
2007 goto incr_m_num;
2008
1725 case '"': 2009 case '"':
1726 self->incr_mode = INCR_M_STR; 2010 self->incr_mode = INCR_M_STR;
1727 goto incr_m_str; 2011 goto incr_m_str;
1728 2012
1729 case '[': 2013 case '[':
1730 case '{': 2014 case '{':
2015 case '(':
1731 if (++self->incr_nest > self->max_depth) 2016 if (++self->incr_nest > self->max_depth)
1732 croak (ERR_NESTING_EXCEEDED); 2017 croak (ERR_NESTING_EXCEEDED);
1733 break; 2018 break;
1734 2019
1735 case ']': 2020 case ']':
1736 case '}': 2021 case '}':
1737 if (--self->incr_nest <= 0) 2022 if (--self->incr_nest <= 0)
1738 goto interrupt; 2023 goto interrupt;
2024 break;
2025
2026 case ')':
2027 --self->incr_nest;
1739 break; 2028 break;
1740 2029
1741 case '#': 2030 case '#':
1742 self->incr_mode = INCR_M_C1; 2031 self->incr_mode = INCR_M_C1;
1743 goto incr_m_c; 2032 goto incr_m_c;
1769 i >= '0' && i <= '9' ? i - '0' 2058 i >= '0' && i <= '9' ? i - '0'
1770 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2059 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1771 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2060 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1772 : -1; 2061 : -1;
1773 2062
1774 json_stash = gv_stashpv ("JSON::XS" , 1); 2063 json_stash = gv_stashpv ("JSON::XS" , 1);
1775 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2064 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1776
1777 json_true = get_bool ("JSON::XS::true");
1778 json_false = get_bool ("JSON::XS::false"); 2065 bool_false = get_bool ("Types::Serialiser::false");
2066 bool_true = get_bool ("Types::Serialiser::true");
2067
2068 sv_json = newSVpv ("JSON", 0);
2069 SvREADONLY_on (sv_json);
1779 2070
1780 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ 2071 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1781} 2072}
1782 2073
1783PROTOTYPES: DISABLE 2074PROTOTYPES: DISABLE
1784 2075
1785void CLONE (...) 2076void CLONE (...)
1786 CODE: 2077 CODE:
2078 // as long as these writes are atomic, the race should not matter
2079 // as existing threads either already use 0, or use the old value,
2080 // which is sitll correct for the initial thread.
1787 json_stash = 0; 2081 json_stash = 0;
1788 json_boolean_stash = 0; 2082 bool_stash = 0;
2083 bool_false = 0;
2084 bool_true = 0;
1789 2085
1790void new (char *klass) 2086void new (char *klass)
1791 PPCODE: 2087 PPCODE:
1792{ 2088{
1793 SV *pv = NEWSV (0, sizeof (JSON)); 2089 SV *pv = NEWSV (0, sizeof (JSON));
1794 SvPOK_only (pv); 2090 SvPOK_only (pv);
1795 json_init ((JSON *)SvPVX (pv)); 2091 json_init ((JSON *)SvPVX (pv));
1796 XPUSHs (sv_2mortal (sv_bless ( 2092 XPUSHs (sv_2mortal (sv_bless (
1797 newRV_noinc (pv), 2093 newRV_noinc (pv),
1798 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2094 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1799 ))); 2095 )));
1800} 2096}
2097
2098void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2099 PPCODE:
2100 self->v_false = newSVsv (v_false);
2101 self->v_true = newSVsv (v_true);
2102 XPUSHs (ST (0));
2103
2104void get_boolean_values (JSON *self)
2105 PPCODE:
2106 if (self->v_false && self->v_true)
2107 {
2108 EXTEND (SP, 2);
2109 PUSHs (self->v_false);
2110 PUSHs (self->v_true);
2111 }
1801 2112
1802void ascii (JSON *self, int enable = 1) 2113void ascii (JSON *self, int enable = 1)
1803 ALIAS: 2114 ALIAS:
1804 ascii = F_ASCII 2115 ascii = F_ASCII
1805 latin1 = F_LATIN1 2116 latin1 = F_LATIN1
1813 shrink = F_SHRINK 2124 shrink = F_SHRINK
1814 allow_blessed = F_ALLOW_BLESSED 2125 allow_blessed = F_ALLOW_BLESSED
1815 convert_blessed = F_CONV_BLESSED 2126 convert_blessed = F_CONV_BLESSED
1816 relaxed = F_RELAXED 2127 relaxed = F_RELAXED
1817 allow_unknown = F_ALLOW_UNKNOWN 2128 allow_unknown = F_ALLOW_UNKNOWN
2129 allow_tags = F_ALLOW_TAGS
1818 PPCODE: 2130 PPCODE:
1819{ 2131{
1820 if (enable) 2132 if (enable)
1821 self->flags |= ix; 2133 self->flags |= ix;
1822 else 2134 else
1838 get_shrink = F_SHRINK 2150 get_shrink = F_SHRINK
1839 get_allow_blessed = F_ALLOW_BLESSED 2151 get_allow_blessed = F_ALLOW_BLESSED
1840 get_convert_blessed = F_CONV_BLESSED 2152 get_convert_blessed = F_CONV_BLESSED
1841 get_relaxed = F_RELAXED 2153 get_relaxed = F_RELAXED
1842 get_allow_unknown = F_ALLOW_UNKNOWN 2154 get_allow_unknown = F_ALLOW_UNKNOWN
2155 get_allow_tags = F_ALLOW_TAGS
1843 PPCODE: 2156 PPCODE:
1844 XPUSHs (boolSV (self->flags & ix)); 2157 XPUSHs (boolSV (self->flags & ix));
1845 2158
1846void max_depth (JSON *self, U32 max_depth = 0x80000000UL) 2159void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1847 PPCODE: 2160 PPCODE:
1875} 2188}
1876 2189
1877void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2190void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1878 PPCODE: 2191 PPCODE:
1879{ 2192{
1880 if (!self->cb_sk_object) 2193 if (!self->cb_sk_object)
1881 self->cb_sk_object = newHV (); 2194 self->cb_sk_object = newHV ();
1882 2195
1883 if (SvOK (cb)) 2196 if (SvOK (cb))
1884 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2197 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1885 else 2198 else
1896 XPUSHs (ST (0)); 2209 XPUSHs (ST (0));
1897} 2210}
1898 2211
1899void encode (JSON *self, SV *scalar) 2212void encode (JSON *self, SV *scalar)
1900 PPCODE: 2213 PPCODE:
1901 XPUSHs (encode_json (scalar, self)); 2214 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2215 XPUSHs (scalar);
1902 2216
1903void decode (JSON *self, SV *jsonstr) 2217void decode (JSON *self, SV *jsonstr)
1904 PPCODE: 2218 PPCODE:
1905 XPUSHs (decode_json (jsonstr, self, 0)); 2219 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2220 XPUSHs (jsonstr);
1906 2221
1907void decode_prefix (JSON *self, SV *jsonstr) 2222void decode_prefix (JSON *self, SV *jsonstr)
1908 PPCODE: 2223 PPCODE:
1909{ 2224{
1910 char *offset; 2225 SV *sv;
2226 STRLEN offset;
2227 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1911 EXTEND (SP, 2); 2228 EXTEND (SP, 2);
1912 PUSHs (decode_json (jsonstr, self, &offset)); 2229 PUSHs (sv);
1913 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); 2230 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1914} 2231}
1915 2232
1916void incr_parse (JSON *self, SV *jsonstr = 0) 2233void incr_parse (JSON *self, SV *jsonstr = 0)
1917 PPCODE: 2234 PPCODE:
1918{ 2235{
1919 if (!self->incr_text) 2236 if (!self->incr_text)
1920 self->incr_text = newSVpvn ("", 0); 2237 self->incr_text = newSVpvn ("", 0);
2238
2239 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2240 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2241 if (DECODE_WANTS_OCTETS (self))
2242 {
2243 if (self->incr_pos)
2244 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2245 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2246
2247 sv_utf8_downgrade (self->incr_text, 0);
2248 }
2249 else
2250 {
2251 sv_utf8_upgrade (self->incr_text);
2252
2253 if (self->incr_pos)
2254 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2255 - (U8 *)SvPVX (self->incr_text);
2256 }
1921 2257
1922 // append data, if any 2258 // append data, if any
1923 if (jsonstr) 2259 if (jsonstr)
1924 { 2260 {
2261 /* make sure both strings have same encoding */
2262 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1925 if (SvUTF8 (jsonstr)) 2263 if (SvUTF8 (jsonstr))
2264 sv_utf8_downgrade (jsonstr, 0);
1926 { 2265 else
1927 if (!SvUTF8 (self->incr_text))
1928 {
1929 /* utf-8-ness differs, need to upgrade */
1930 sv_utf8_upgrade (self->incr_text);
1931
1932 if (self->incr_pos)
1933 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1934 - (U8 *)SvPVX (self->incr_text);
1935 }
1936 }
1937 else if (SvUTF8 (self->incr_text))
1938 sv_utf8_upgrade (jsonstr); 2266 sv_utf8_upgrade (jsonstr);
1939 2267
2268 /* and then just blindly append */
1940 { 2269 {
1941 STRLEN len; 2270 STRLEN len;
1942 const char *str = SvPV (jsonstr, len); 2271 const char *str = SvPV (jsonstr, len);
1943 STRLEN cur = SvCUR (self->incr_text); 2272 STRLEN cur = SvCUR (self->incr_text);
1944 2273
1945 if (SvLEN (self->incr_text) <= cur + len) 2274 if (SvLEN (self->incr_text) - cur <= len)
1946 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 2275 json_sv_grow (self->incr_text, cur, len);
1947 2276
1948 Move (str, SvEND (self->incr_text), len, char); 2277 Move (str, SvEND (self->incr_text), len, char);
1949 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2278 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1950 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2279 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1951 } 2280 }
1952 } 2281 }
1953 2282
1954 if (GIMME_V != G_VOID) 2283 if (GIMME_V != G_VOID)
1955 do 2284 do
1956 { 2285 {
2286 SV *sv;
1957 char *offset; 2287 STRLEN offset;
1958 2288
1959 if (!INCR_DONE (self)) 2289 if (!INCR_DONE (self))
1960 { 2290 {
1961 incr_parse (self); 2291 incr_parse (self);
1962 2292
1963 if (self->incr_pos > self->max_size && self->max_size) 2293 if (self->incr_pos > self->max_size && self->max_size)
1964 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 2294 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1965 (unsigned long)self->incr_pos, (unsigned long)self->max_size); 2295 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1966 2296
1967 if (!INCR_DONE (self)) 2297 if (!INCR_DONE (self))
2298 {
2299 // as an optimisation, do not accumulate white space in the incr buffer
2300 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2301 {
2302 self->incr_pos = 0;
2303 SvCUR_set (self->incr_text, 0);
2304 }
2305
1968 break; 2306 break;
2307 }
1969 } 2308 }
1970 2309
1971 XPUSHs (decode_json (self->incr_text, self, &offset)); 2310 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2311 XPUSHs (sv);
1972 2312
1973 self->incr_pos -= offset - SvPVX (self->incr_text); 2313 self->incr_pos -= offset;
1974 self->incr_nest = 0; 2314 self->incr_nest = 0;
1975 self->incr_mode = 0; 2315 self->incr_mode = 0;
1976 2316
1977 sv_chop (self->incr_text, offset); 2317 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1978 } 2318 }
1979 while (GIMME_V == G_ARRAY); 2319 while (GIMME_V == G_ARRAY);
1980} 2320}
1981 2321
1982SV *incr_text (JSON *self) 2322SV *incr_text (JSON *self)
2013 self->incr_mode = 0; 2353 self->incr_mode = 0;
2014} 2354}
2015 2355
2016void DESTROY (JSON *self) 2356void DESTROY (JSON *self)
2017 CODE: 2357 CODE:
2358 SvREFCNT_dec (self->v_false);
2359 SvREFCNT_dec (self->v_true);
2018 SvREFCNT_dec (self->cb_sk_object); 2360 SvREFCNT_dec (self->cb_sk_object);
2019 SvREFCNT_dec (self->cb_object); 2361 SvREFCNT_dec (self->cb_object);
2020 SvREFCNT_dec (self->incr_text); 2362 SvREFCNT_dec (self->incr_text);
2021 2363
2022PROTOTYPES: ENABLE 2364PROTOTYPES: ENABLE
2023 2365
2024void encode_json (SV *scalar) 2366void encode_json (SV *scalar)
2025 ALIAS:
2026 to_json_ = 0
2027 encode_json = F_UTF8
2028 PPCODE: 2367 PPCODE:
2029{ 2368{
2030 JSON json; 2369 JSON json;
2031 json_init (&json); 2370 json_init (&json);
2032 json.flags |= ix; 2371 json.flags |= F_UTF8;
2033 XPUSHs (encode_json (scalar, &json)); 2372 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2373 XPUSHs (scalar);
2034} 2374}
2035 2375
2036void decode_json (SV *jsonstr) 2376void decode_json (SV *jsonstr)
2037 ALIAS:
2038 from_json_ = 0
2039 decode_json = F_UTF8
2040 PPCODE: 2377 PPCODE:
2041{ 2378{
2042 JSON json; 2379 JSON json;
2043 json_init (&json); 2380 json_init (&json);
2044 json.flags |= ix; 2381 json.flags |= F_UTF8;
2045 XPUSHs (decode_json (jsonstr, &json, 0)); 2382 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2383 XPUSHs (jsonstr);
2046} 2384}
2047 2385

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines