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 1.50 by root, Mon Jul 2 00:29:38 2007 UTC vs.
Revision 1.73 by root, Wed Mar 19 13:44:43 2008 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "assert.h" 5#include <assert.h>
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 <float.h>
9 10
10#if defined(__BORLANDC__) || defined(_MSC_VER) 11#if defined(__BORLANDC__) || defined(_MSC_VER)
11# define snprintf _snprintf // C compilers have this in stdio.h 12# define snprintf _snprintf // C compilers have this in stdio.h
12#endif 13#endif
13 14
26#define F_SPACE_AFTER 0x00000040UL 27#define F_SPACE_AFTER 0x00000040UL
27#define F_ALLOW_NONREF 0x00000100UL 28#define F_ALLOW_NONREF 0x00000100UL
28#define F_SHRINK 0x00000200UL 29#define F_SHRINK 0x00000200UL
29#define F_ALLOW_BLESSED 0x00000400UL 30#define F_ALLOW_BLESSED 0x00000400UL
30#define F_CONV_BLESSED 0x00000800UL 31#define F_CONV_BLESSED 0x00000800UL
32#define F_RELAXED 0x00001000UL
33
31#define F_MAXDEPTH 0xf8000000UL 34#define F_MAXDEPTH 0xf8000000UL
32#define S_MAXDEPTH 27 35#define S_MAXDEPTH 27
33#define F_MAXSIZE 0x01f00000UL 36#define F_MAXSIZE 0x01f00000UL
34#define S_MAXSIZE 20 37#define S_MAXSIZE 20
35#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
47 50
48#define SB do { 51#define SB do {
49#define SE } while (0) 52#define SE } while (0)
50 53
51#if __GNUC__ >= 3 54#if __GNUC__ >= 3
52# define expect(expr,value) __builtin_expect ((expr),(value)) 55# define expect(expr,value) __builtin_expect ((expr), (value))
53# define inline inline 56# define INLINE static inline
54#else 57#else
55# define expect(expr,value) (expr) 58# define expect(expr,value) (expr)
56# define inline static 59# define INLINE static
57#endif 60#endif
58 61
59#define expect_false(expr) expect ((expr) != 0, 0) 62#define expect_false(expr) expect ((expr) != 0, 0)
60#define expect_true(expr) expect ((expr) != 0, 1) 63#define expect_true(expr) expect ((expr) != 0, 1)
61 64
65#define IN_RANGE_INC(type,val,beg,end) \
66 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
67 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
68
69#ifdef USE_ITHREADS
70# define JSON_SLOW 1
71# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
72#else
73# define JSON_SLOW 0
74# define JSON_STASH json_stash
75#endif
76
62static HV *json_stash, *json_boolean_stash; // JSON::XS:: 77static HV *json_stash, *json_boolean_stash; // JSON::XS::
63static SV *json_true, *json_false; 78static SV *json_true, *json_false;
64 79
65typedef struct { 80typedef struct {
66 U32 flags; 81 U32 flags;
67 SV *cb_object, *cb_sk_object; 82 SV *cb_object;
83 HV *cb_sk_object;
68} JSON; 84} JSON;
69 85
70///////////////////////////////////////////////////////////////////////////// 86/////////////////////////////////////////////////////////////////////////////
71// utility functions 87// utility functions
72 88
73inline void 89INLINE void
74shrink (SV *sv) 90shrink (SV *sv)
75{ 91{
76 sv_utf8_downgrade (sv, 1); 92 sv_utf8_downgrade (sv, 1);
77 if (SvLEN (sv) > SvCUR (sv) + 1) 93 if (SvLEN (sv) > SvCUR (sv) + 1)
78 { 94 {
87// decode an utf-8 character and return it, or (UV)-1 in 103// decode an utf-8 character and return it, or (UV)-1 in
88// case of an error. 104// case of an error.
89// we special-case "safe" characters from U+80 .. U+7FF, 105// we special-case "safe" characters from U+80 .. U+7FF,
90// but use the very good perl function to parse anything else. 106// but use the very good perl function to parse anything else.
91// note that we never call this function for a ascii codepoints 107// note that we never call this function for a ascii codepoints
92inline UV 108INLINE UV
93decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 109decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
94{ 110{
95 if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) 111 if (expect_true (len >= 2
96 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 112 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
97 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 113 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
98 { 114 {
99 *clen = 2; 115 *clen = 2;
100 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 116 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
101 } 117 }
102 else 118 else
103 { 119 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
104 *clen = (STRLEN)-1; 120}
105 return (UV)-1; 121
122// likewise for encoding, also never called for ascii codepoints
123// this function takes advantage of this fact, although current gccs
124// seem to optimise the check for >= 0x80 away anyways
125INLINE unsigned char *
126encode_utf8 (unsigned char *s, UV ch)
127{
128 if (ch <= 0x7FF)
106 } 129 {
130 *s++ = (ch >> 6) | 0xc0;
131 *s++ = (ch & 0x3f) | 0x80;
132 }
133 else
134 s = uvuni_to_utf8_flags (s, ch, 0);
135
136 return s;
107} 137}
108 138
109///////////////////////////////////////////////////////////////////////////// 139/////////////////////////////////////////////////////////////////////////////
110// encoder 140// encoder
111 141
116 char *end; // SvEND (sv) 146 char *end; // SvEND (sv)
117 SV *sv; // result scalar 147 SV *sv; // result scalar
118 JSON json; 148 JSON json;
119 U32 indent; // indentation level 149 U32 indent; // indentation level
120 U32 maxdepth; // max. indentation/recursion level 150 U32 maxdepth; // max. indentation/recursion level
151 UV limit; // escape character values >= this value when encoding
121} enc_t; 152} enc_t;
122 153
123inline void 154INLINE void
124need (enc_t *enc, STRLEN len) 155need (enc_t *enc, STRLEN len)
125{ 156{
126 if (expect_false (enc->cur + len >= enc->end)) 157 if (expect_false (enc->cur + len >= enc->end))
127 { 158 {
128 STRLEN cur = enc->cur - SvPVX (enc->sv); 159 STRLEN cur = enc->cur - SvPVX (enc->sv);
130 enc->cur = SvPVX (enc->sv) + cur; 161 enc->cur = SvPVX (enc->sv) + cur;
131 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 162 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
132 } 163 }
133} 164}
134 165
135inline void 166INLINE void
136encode_ch (enc_t *enc, char ch) 167encode_ch (enc_t *enc, char ch)
137{ 168{
138 need (enc, 1); 169 need (enc, 1);
139 *enc->cur++ = ch; 170 *enc->cur++ = ch;
140} 171}
194 { 225 {
195 uch = ch; 226 uch = ch;
196 clen = 1; 227 clen = 1;
197 } 228 }
198 229
199 if (uch > 0x10FFFFUL) 230 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
200 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
201
202 if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
203 { 231 {
204 if (uch > 0xFFFFUL) 232 if (uch > 0xFFFFUL)
205 { 233 {
234 if (uch > 0x10FFFFUL)
235 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
236
206 need (enc, len += 11); 237 need (enc, len += 11);
207 sprintf (enc->cur, "\\u%04x\\u%04x", 238 sprintf (enc->cur, "\\u%04x\\u%04x",
208 (int)((uch - 0x10000) / 0x400 + 0xD800), 239 (int)((uch - 0x10000) / 0x400 + 0xD800),
209 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 240 (int)((uch - 0x10000) % 0x400 + 0xDC00));
210 enc->cur += 12; 241 enc->cur += 12;
238 while (--clen); 269 while (--clen);
239 } 270 }
240 else 271 else
241 { 272 {
242 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 273 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
243 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 274 enc->cur = encode_utf8 (enc->cur, uch);
244 ++str; 275 ++str;
245 } 276 }
246 } 277 }
247 } 278 }
248 } 279 }
249 280
250 --len; 281 --len;
251 } 282 }
252} 283}
253 284
254inline void 285INLINE void
255encode_indent (enc_t *enc) 286encode_indent (enc_t *enc)
256{ 287{
257 if (enc->json.flags & F_INDENT) 288 if (enc->json.flags & F_INDENT)
258 { 289 {
259 int spaces = enc->indent * INDENT_STEP; 290 int spaces = enc->indent * INDENT_STEP;
262 memset (enc->cur, ' ', spaces); 293 memset (enc->cur, ' ', spaces);
263 enc->cur += spaces; 294 enc->cur += spaces;
264 } 295 }
265} 296}
266 297
267inline void 298INLINE void
268encode_space (enc_t *enc) 299encode_space (enc_t *enc)
269{ 300{
270 need (enc, 1); 301 need (enc, 1);
271 encode_ch (enc, ' '); 302 encode_ch (enc, ' ');
272} 303}
273 304
274inline void 305INLINE void
275encode_nl (enc_t *enc) 306encode_nl (enc_t *enc)
276{ 307{
277 if (enc->json.flags & F_INDENT) 308 if (enc->json.flags & F_INDENT)
278 { 309 {
279 need (enc, 1); 310 need (enc, 1);
280 encode_ch (enc, '\n'); 311 encode_ch (enc, '\n');
281 } 312 }
282} 313}
283 314
284inline void 315INLINE void
285encode_comma (enc_t *enc) 316encode_comma (enc_t *enc)
286{ 317{
287 encode_ch (enc, ','); 318 encode_ch (enc, ',');
288 319
289 if (enc->json.flags & F_INDENT) 320 if (enc->json.flags & F_INDENT)
300 int i, len = av_len (av); 331 int i, len = av_len (av);
301 332
302 if (enc->indent >= enc->maxdepth) 333 if (enc->indent >= enc->maxdepth)
303 croak ("data structure too deep (hit recursion limit)"); 334 croak ("data structure too deep (hit recursion limit)");
304 335
305 encode_ch (enc, '['); encode_nl (enc); 336 encode_ch (enc, '[');
306 ++enc->indent; 337
338 if (len >= 0)
339 {
340 encode_nl (enc); ++enc->indent;
307 341
308 for (i = 0; i <= len; ++i) 342 for (i = 0; i <= len; ++i)
309 { 343 {
344 SV **svp = av_fetch (av, i, 0);
345
310 encode_indent (enc); 346 encode_indent (enc);
311 encode_sv (enc, *av_fetch (av, i, 0));
312 347
348 if (svp)
349 encode_sv (enc, *svp);
350 else
351 encode_str (enc, "null", 4, 0);
352
313 if (i < len) 353 if (i < len)
314 encode_comma (enc); 354 encode_comma (enc);
315 } 355 }
316 356
357 encode_nl (enc); --enc->indent; encode_indent (enc);
358 }
359
317 encode_nl (enc); 360 encode_ch (enc, ']');
318
319 --enc->indent;
320 encode_indent (enc); encode_ch (enc, ']');
321} 361}
322 362
323static void 363static void
324encode_he (enc_t *enc, HE *he) 364encode_hk (enc_t *enc, HE *he)
325{ 365{
326 encode_ch (enc, '"'); 366 encode_ch (enc, '"');
327 367
328 if (HeKLEN (he) == HEf_SVKEY) 368 if (HeKLEN (he) == HEf_SVKEY)
329 { 369 {
342 encode_ch (enc, '"'); 382 encode_ch (enc, '"');
343 383
344 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); 384 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
345 encode_ch (enc, ':'); 385 encode_ch (enc, ':');
346 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); 386 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
347 encode_sv (enc, HeVAL (he));
348} 387}
349 388
350// compare hash entries, used when all keys are bytestrings 389// compare hash entries, used when all keys are bytestrings
351static int 390static int
352he_cmp_fast (const void *a_, const void *b_) 391he_cmp_fast (const void *a_, const void *b_)
357 HE *b = *(HE **)b_; 396 HE *b = *(HE **)b_;
358 397
359 STRLEN la = HeKLEN (a); 398 STRLEN la = HeKLEN (a);
360 STRLEN lb = HeKLEN (b); 399 STRLEN lb = HeKLEN (b);
361 400
362 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 401 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
363 cmp = la - lb; 402 cmp = lb - la;
364 403
365 return cmp; 404 return cmp;
366} 405}
367 406
368// compare hash entries, used when some keys are sv's or utf-x 407// compare hash entries, used when some keys are sv's or utf-x
369static int 408static int
370he_cmp_slow (const void *a, const void *b) 409he_cmp_slow (const void *a, const void *b)
371{ 410{
372 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 411 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
373} 412}
374 413
375static void 414static void
376encode_hv (enc_t *enc, HV *hv) 415encode_hv (enc_t *enc, HV *hv)
377{ 416{
417 HE *he;
378 int count, i; 418 int count;
379 419
380 if (enc->indent >= enc->maxdepth) 420 if (enc->indent >= enc->maxdepth)
381 croak ("data structure too deep (hit recursion limit)"); 421 croak ("data structure too deep (hit recursion limit)");
382 422
383 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; 423 encode_ch (enc, '{');
384 424
385 if ((count = hv_iterinit (hv)))
386 {
387 // for canonical output we have to sort by keys first 425 // for canonical output we have to sort by keys first
388 // actually, this is mostly due to the stupid so-called 426 // actually, this is mostly due to the stupid so-called
389 // security workaround added somewhere in 5.8.x. 427 // security workaround added somewhere in 5.8.x.
390 // that randomises hash orderings 428 // that randomises hash orderings
391 if (enc->json.flags & F_CANONICAL) 429 if (enc->json.flags & F_CANONICAL)
430 {
431 int count = hv_iterinit (hv);
432
433 if (SvMAGICAL (hv))
392 { 434 {
435 // need to count by iterating. could improve by dynamically building the vector below
436 // but I don't care for the speed of this special case.
437 // note also that we will run into undefined behaviour when the two iterations
438 // do not result in the same count, something I might care for in some later release.
439
440 count = 0;
441 while (hv_iternext (hv))
442 ++count;
443
444 hv_iterinit (hv);
445 }
446
447 if (count)
448 {
393 int fast = 1; 449 int i, fast = 1;
394 HE *he;
395#if defined(__BORLANDC__) || defined(_MSC_VER) 450#if defined(__BORLANDC__) || defined(_MSC_VER)
396 HE **hes = _alloca (count * sizeof (HE)); 451 HE **hes = _alloca (count * sizeof (HE));
397#else 452#else
398 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 453 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
399#endif 454#endif
426 481
427 FREETMPS; 482 FREETMPS;
428 LEAVE; 483 LEAVE;
429 } 484 }
430 485
431 for (i = 0; i < count; ++i) 486 encode_nl (enc); ++enc->indent;
487
488 while (count--)
432 { 489 {
433 encode_indent (enc); 490 encode_indent (enc);
491 he = hes [count];
434 encode_he (enc, hes [i]); 492 encode_hk (enc, he);
493 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
435 494
436 if (i < count - 1) 495 if (count)
437 encode_comma (enc); 496 encode_comma (enc);
438 } 497 }
439 498
440 encode_nl (enc); 499 encode_nl (enc); --enc->indent; encode_indent (enc);
441 } 500 }
501 }
442 else 502 else
503 {
504 if (hv_iterinit (hv) || SvMAGICAL (hv))
505 if ((he = hv_iternext (hv)))
443 { 506 {
444 HE *he = hv_iternext (hv); 507 encode_nl (enc); ++enc->indent;
445 508
446 for (;;) 509 for (;;)
447 { 510 {
448 encode_indent (enc); 511 encode_indent (enc);
449 encode_he (enc, he); 512 encode_hk (enc, he);
513 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
450 514
451 if (!(he = hv_iternext (hv))) 515 if (!(he = hv_iternext (hv)))
452 break; 516 break;
453 517
454 encode_comma (enc); 518 encode_comma (enc);
455 } 519 }
456 520
457 encode_nl (enc); 521 encode_nl (enc); --enc->indent; encode_indent (enc);
458 } 522 }
459 } 523 }
460 524
461 --enc->indent; encode_indent (enc); encode_ch (enc, '}'); 525 encode_ch (enc, '}');
462} 526}
463 527
464// encode objects, arrays and special \0=false and \1=true values. 528// encode objects, arrays and special \0=false and \1=true values.
465static void 529static void
466encode_rv (enc_t *enc, SV *sv) 530encode_rv (enc_t *enc, SV *sv)
470 SvGETMAGIC (sv); 534 SvGETMAGIC (sv);
471 svt = SvTYPE (sv); 535 svt = SvTYPE (sv);
472 536
473 if (expect_false (SvOBJECT (sv))) 537 if (expect_false (SvOBJECT (sv)))
474 { 538 {
539 HV *stash = !JSON_SLOW || json_boolean_stash
540 ? json_boolean_stash
541 : gv_stashpv ("JSON::XS::Boolean", 1);
542
475 if (SvSTASH (sv) == json_boolean_stash) 543 if (SvSTASH (sv) == stash)
476 { 544 {
477 if (SvIV (sv) == 0) 545 if (SvIV (sv))
546 encode_str (enc, "true", 4, 0);
547 else
478 encode_str (enc, "false", 5, 0); 548 encode_str (enc, "false", 5, 0);
479 else
480 encode_str (enc, "true", 4, 0);
481 } 549 }
482 else 550 else
483 { 551 {
484#if 0 552#if 0
485 if (0 && sv_derived_from (rv, "JSON::Literal")) 553 if (0 && sv_derived_from (rv, "JSON::Literal"))
488 } 556 }
489#endif 557#endif
490 if (enc->json.flags & F_CONV_BLESSED) 558 if (enc->json.flags & F_CONV_BLESSED)
491 { 559 {
492 // we re-bless the reference to get overload and other niceties right 560 // we re-bless the reference to get overload and other niceties right
493 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1); 561 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
494 562
495 if (to_json) 563 if (to_json)
496 { 564 {
565 dSP;
566
497 dSP; ENTER; SAVETMPS; PUSHMARK (SP); 567 ENTER; SAVETMPS; PUSHMARK (SP);
498 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); 568 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
499 569
500 // calling with G_SCALAR ensures that we always get a 1 reutrn value 570 // calling with G_SCALAR ensures that we always get a 1 return value
501 // check anyways.
502 PUTBACK; 571 PUTBACK;
503 assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR)); 572 call_sv ((SV *)GvCV (to_json), G_SCALAR);
504 SPAGAIN; 573 SPAGAIN;
505 574
575 // catch this surprisingly common error
576 if (SvROK (TOPs) && SvRV (TOPs) == sv)
577 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
578
579 sv = POPs;
580 PUTBACK;
581
506 encode_sv (enc, POPs); 582 encode_sv (enc, sv);
507 583
508 FREETMPS; LEAVE; 584 FREETMPS; LEAVE;
509 } 585 }
510 else if (enc->json.flags & F_ALLOW_BLESSED) 586 else if (enc->json.flags & F_ALLOW_BLESSED)
511 encode_str (enc, "null", 4, 0); 587 encode_str (enc, "null", 4, 0);
527 else if (svt < SVt_PVAV) 603 else if (svt < SVt_PVAV)
528 { 604 {
529 STRLEN len = 0; 605 STRLEN len = 0;
530 char *pv = svt ? SvPV (sv, len) : 0; 606 char *pv = svt ? SvPV (sv, len) : 0;
531 607
532 if (len == 1 && *pv == '0') 608 if (len == 1 && *pv == '1')
609 encode_str (enc, "true", 4, 0);
610 else if (len == 1 && *pv == '0')
533 encode_str (enc, "false", 5, 0); 611 encode_str (enc, "false", 5, 0);
534 else if (len == 1 && *pv == '1')
535 encode_str (enc, "true", 4, 0);
536 else 612 else
537 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 613 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
538 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 614 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
539 } 615 }
540 else 616 else
623 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 699 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
624 enc.cur = SvPVX (enc.sv); 700 enc.cur = SvPVX (enc.sv);
625 enc.end = SvEND (enc.sv); 701 enc.end = SvEND (enc.sv);
626 enc.indent = 0; 702 enc.indent = 0;
627 enc.maxdepth = DEC_DEPTH (enc.json.flags); 703 enc.maxdepth = DEC_DEPTH (enc.json.flags);
704 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
705 : enc.json.flags & F_LATIN1 ? 0x000100UL
706 : 0x10FFFFUL;
628 707
629 SvPOK_only (enc.sv); 708 SvPOK_only (enc.sv);
630 encode_sv (&enc, scalar); 709 encode_sv (&enc, scalar);
631 710
632 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 711 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
653 JSON json; 732 JSON json;
654 U32 depth; // recursion depth 733 U32 depth; // recursion depth
655 U32 maxdepth; // recursion depth limit 734 U32 maxdepth; // recursion depth limit
656} dec_t; 735} dec_t;
657 736
658inline void 737INLINE void
738decode_comment (dec_t *dec)
739{
740 // only '#'-style comments allowed a.t.m.
741
742 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
743 ++dec->cur;
744}
745
746INLINE void
659decode_ws (dec_t *dec) 747decode_ws (dec_t *dec)
660{ 748{
661 for (;;) 749 for (;;)
662 { 750 {
663 char ch = *dec->cur; 751 char ch = *dec->cur;
664 752
665 if (ch > 0x20 753 if (ch > 0x20)
754 {
755 if (expect_false (ch == '#'))
756 {
757 if (dec->json.flags & F_RELAXED)
758 decode_comment (dec);
759 else
760 break;
761 }
762 else
763 break;
764 }
666 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) 765 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
667 break; 766 break; // parse error, but let higher level handle it, gives better error messages
668 767
669 ++dec->cur; 768 ++dec->cur;
670 } 769 }
671} 770}
672 771
778 877
779 if (hi >= 0x80) 878 if (hi >= 0x80)
780 { 879 {
781 utf8 = 1; 880 utf8 = 1;
782 881
783 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); 882 cur = encode_utf8 (cur, hi);
784 } 883 }
785 else 884 else
786 *cur++ = hi; 885 *cur++ = hi;
787 } 886 }
788 break; 887 break;
921 is_nv = 1; 1020 is_nv = 1;
922 } 1021 }
923 1022
924 if (!is_nv) 1023 if (!is_nv)
925 { 1024 {
1025 int len = dec->cur - start;
1026
926 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so 1027 // special case the rather common 1..5-digit-int case
927 if (*start == '-') 1028 if (*start == '-')
928 switch (dec->cur - start) 1029 switch (len)
929 { 1030 {
930 case 2: return newSViv (-( start [1] - '0' * 1)); 1031 case 2: return newSViv (-( start [1] - '0' * 1));
931 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1032 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
932 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1033 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
933 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1034 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1035 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
934 } 1036 }
935 else 1037 else
936 switch (dec->cur - start) 1038 switch (len)
937 { 1039 {
938 case 1: return newSViv ( start [0] - '0' * 1); 1040 case 1: return newSViv ( start [0] - '0' * 1);
939 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1041 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
940 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1042 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
941 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1043 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1044 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
942 } 1045 }
943 1046
944 { 1047 {
945 UV uv; 1048 UV uv;
946 int numtype = grok_number (start, dec->cur - start, &uv); 1049 int numtype = grok_number (start, len, &uv);
947 if (numtype & IS_NUMBER_IN_UV) 1050 if (numtype & IS_NUMBER_IN_UV)
948 if (numtype & IS_NUMBER_NEG) 1051 if (numtype & IS_NUMBER_NEG)
949 { 1052 {
950 if (uv < (UV)IV_MIN) 1053 if (uv < (UV)IV_MIN)
951 return newSViv (-(IV)uv); 1054 return newSViv (-(IV)uv);
952 } 1055 }
953 else 1056 else
954 return newSVuv (uv); 1057 return newSVuv (uv);
955
956 // here would likely be the place for bigint support
957 } 1058 }
958 }
959 1059
960 // if we ever support bigint or bigfloat, this is the place for bigfloat 1060 len -= *start == '-' ? 1 : 0;
1061
1062 // does not fit into IV or UV, try NV
1063 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
1064 #if defined (LDBL_DIG)
1065 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1066 #endif
1067 )
1068 // fits into NV without loss of precision
1069 return newSVnv (Atof (start));
1070
1071 // everything else fails, convert it to a string
1072 return newSVpvn (start, dec->cur - start);
1073 }
1074
1075 // loss of precision here
961 return newSVnv (Atof (start)); 1076 return newSVnv (Atof (start));
962 1077
963fail: 1078fail:
964 return 0; 1079 return 0;
965} 1080}
995 1110
996 if (*dec->cur != ',') 1111 if (*dec->cur != ',')
997 ERR (", or ] expected while parsing array"); 1112 ERR (", or ] expected while parsing array");
998 1113
999 ++dec->cur; 1114 ++dec->cur;
1115
1116 decode_ws (dec);
1117
1118 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1119 {
1120 ++dec->cur;
1121 break;
1122 }
1000 } 1123 }
1001 1124
1002 DEC_DEC_DEPTH; 1125 DEC_DEC_DEPTH;
1003 return newRV_noinc ((SV *)av); 1126 return newRV_noinc ((SV *)av);
1004 1127
1020 if (*dec->cur == '}') 1143 if (*dec->cur == '}')
1021 ++dec->cur; 1144 ++dec->cur;
1022 else 1145 else
1023 for (;;) 1146 for (;;)
1024 { 1147 {
1025 decode_ws (dec); EXPECT_CH ('"'); 1148 EXPECT_CH ('"');
1026 1149
1027 // heuristic: assume that 1150 // heuristic: assume that
1028 // a) decode_str + hv_store_ent are abysmally slow. 1151 // a) decode_str + hv_store_ent are abysmally slow.
1029 // b) most hash keys are short, simple ascii text. 1152 // b) most hash keys are short, simple ascii text.
1030 // => try to "fast-match" such strings to avoid 1153 // => try to "fast-match" such strings to avoid
1044 if (!key) 1167 if (!key)
1045 goto fail; 1168 goto fail;
1046 1169
1047 decode_ws (dec); EXPECT_CH (':'); 1170 decode_ws (dec); EXPECT_CH (':');
1048 1171
1172 decode_ws (dec);
1049 value = decode_sv (dec); 1173 value = decode_sv (dec);
1050 if (!value) 1174 if (!value)
1051 { 1175 {
1052 SvREFCNT_dec (key); 1176 SvREFCNT_dec (key);
1053 goto fail; 1177 goto fail;
1065 int len = p - key; 1189 int len = p - key;
1066 dec->cur = p + 1; 1190 dec->cur = p + 1;
1067 1191
1068 decode_ws (dec); EXPECT_CH (':'); 1192 decode_ws (dec); EXPECT_CH (':');
1069 1193
1194 decode_ws (dec);
1070 value = decode_sv (dec); 1195 value = decode_sv (dec);
1071 if (!value) 1196 if (!value)
1072 goto fail; 1197 goto fail;
1073 1198
1074 hv_store (hv, key, len, value, 0); 1199 hv_store (hv, key, len, value, 0);
1090 1215
1091 if (*dec->cur != ',') 1216 if (*dec->cur != ',')
1092 ERR (", or } expected while parsing object/hash"); 1217 ERR (", or } expected while parsing object/hash");
1093 1218
1094 ++dec->cur; 1219 ++dec->cur;
1220
1221 decode_ws (dec);
1222
1223 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1224 {
1225 ++dec->cur;
1226 break;
1227 }
1095 } 1228 }
1096 1229
1097 DEC_DEC_DEPTH; 1230 DEC_DEC_DEPTH;
1098 sv = newRV_noinc ((SV *)hv); 1231 sv = newRV_noinc ((SV *)hv);
1099 1232
1100 // check filter callbacks 1233 // check filter callbacks
1101 if (dec->json.flags & F_HOOK) 1234 if (dec->json.flags & F_HOOK)
1102 { 1235 {
1103 ENTER; SAVETMPS; 1236 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1104
1105 if (HvKEYS (hv) == 1 && dec->json.cb_sk_object)
1106 { 1237 {
1238 HE *cb, *he;
1239
1240 hv_iterinit (hv);
1241 he = hv_iternext (hv);
1242 hv_iterinit (hv);
1243
1244 // the next line creates a mortal sv each time its called.
1245 // might want to optimise this for common cases.
1246 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1247
1248 if (cb)
1249 {
1250 dSP;
1107 int count; 1251 int count;
1108 1252
1109 dSP; PUSHMARK (SP); 1253 ENTER; SAVETMPS; PUSHMARK (SP);
1110 XPUSHs (sv_2mortal (sv)); 1254 XPUSHs (HeVAL (he));
1111 1255
1112 PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN; 1256 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1113 1257
1114 if (count == 1) 1258 if (count == 1)
1259 {
1115 sv = newSVsv (POPs); 1260 sv = newSVsv (POPs);
1116 else 1261 FREETMPS; LEAVE;
1117 SvREFCNT_inc (sv); 1262 return sv;
1263 }
1264
1265 FREETMPS; LEAVE;
1266 }
1118 } 1267 }
1119 1268
1120 if (dec->json.cb_object) 1269 if (dec->json.cb_object)
1121 { 1270 {
1271 dSP;
1122 int count; 1272 int count;
1123 1273
1124 dSP; ENTER; SAVETMPS; PUSHMARK (SP); 1274 ENTER; SAVETMPS; PUSHMARK (SP);
1125 XPUSHs (sv_2mortal (sv)); 1275 XPUSHs (sv_2mortal (sv));
1126 1276
1127 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1277 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1128 1278
1129 if (count == 1) 1279 if (count == 1)
1280 {
1130 sv = newSVsv (POPs); 1281 sv = newSVsv (POPs);
1131 else 1282 FREETMPS; LEAVE;
1283 return sv;
1284 }
1285
1132 SvREFCNT_inc (sv); 1286 SvREFCNT_inc (sv);
1287 FREETMPS; LEAVE;
1133 } 1288 }
1134
1135 FREETMPS; LEAVE;
1136 } 1289 }
1137 1290
1138 return newRV_noinc ((SV *)hv); 1291 return sv;
1139 1292
1140fail: 1293fail:
1141 SvREFCNT_dec (hv); 1294 SvREFCNT_dec (hv);
1142 DEC_DEC_DEPTH; 1295 DEC_DEC_DEPTH;
1143 return 0; 1296 return 0;
1144} 1297}
1145 1298
1146static SV * 1299static SV *
1147decode_sv (dec_t *dec) 1300decode_sv (dec_t *dec)
1148{ 1301{
1149 decode_ws (dec);
1150
1151 // the beauty of JSON: you need exactly one character lookahead 1302 // the beauty of JSON: you need exactly one character lookahead
1152 // to parse anything. 1303 // to parse everything.
1153 switch (*dec->cur) 1304 switch (*dec->cur)
1154 { 1305 {
1155 case '"': ++dec->cur; return decode_str (dec); 1306 case '"': ++dec->cur; return decode_str (dec);
1156 case '[': ++dec->cur; return decode_av (dec); 1307 case '[': ++dec->cur; return decode_av (dec);
1157 case '{': ++dec->cur; return decode_hv (dec); 1308 case '{': ++dec->cur; return decode_hv (dec);
1158 1309
1159 case '-': 1310 case '-':
1160 case '0': case '1': case '2': case '3': case '4': 1311 case '0': case '1': case '2': case '3': case '4':
1161 case '5': case '6': case '7': case '8': case '9': 1312 case '5': case '6': case '7': case '8': case '9':
1162 return decode_num (dec); 1313 return decode_num (dec);
1163 1314
1164 case 't': 1315 case 't':
1165 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1316 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1166 { 1317 {
1167 dec->cur += 4; 1318 dec->cur += 4;
1319#if JSON_SLOW
1320 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
1321#endif
1168 return SvREFCNT_inc (json_true); 1322 return SvREFCNT_inc (json_true);
1169 } 1323 }
1170 else 1324 else
1171 ERR ("'true' expected"); 1325 ERR ("'true' expected");
1172 1326
1174 1328
1175 case 'f': 1329 case 'f':
1176 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1330 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1177 { 1331 {
1178 dec->cur += 5; 1332 dec->cur += 5;
1333#if JSON_SLOW
1334 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1335#endif
1179 return SvREFCNT_inc (json_false); 1336 return SvREFCNT_inc (json_false);
1180 } 1337 }
1181 else 1338 else
1182 ERR ("'false' expected"); 1339 ERR ("'false' expected");
1183 1340
1233 1390
1234 if (dec.json.cb_object || dec.json.cb_sk_object) 1391 if (dec.json.cb_object || dec.json.cb_sk_object)
1235 dec.json.flags |= F_HOOK; 1392 dec.json.flags |= F_HOOK;
1236 1393
1237 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1394 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1395
1396 decode_ws (&dec);
1238 sv = decode_sv (&dec); 1397 sv = decode_sv (&dec);
1239 1398
1240 if (!(offset_return || !sv)) 1399 if (!(offset_return || !sv))
1241 { 1400 {
1242 // check for trailing garbage 1401 // check for trailing garbage
1310 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1469 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1311} 1470}
1312 1471
1313PROTOTYPES: DISABLE 1472PROTOTYPES: DISABLE
1314 1473
1474void CLONE (...)
1475 CODE:
1476 json_stash = 0;
1477 json_boolean_stash = 0;
1478
1315void new (char *klass) 1479void new (char *klass)
1316 PPCODE: 1480 PPCODE:
1317{ 1481{
1318 SV *pv = NEWSV (0, sizeof (JSON)); 1482 SV *pv = NEWSV (0, sizeof (JSON));
1319 SvPOK_only (pv); 1483 SvPOK_only (pv);
1320 Zero (SvPVX (pv), 1, JSON); 1484 Zero (SvPVX (pv), 1, JSON);
1321 ((JSON *)SvPVX (pv))->flags = F_DEFAULT; 1485 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1322 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash))); 1486 XPUSHs (sv_2mortal (sv_bless (
1487 newRV_noinc (pv),
1488 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1489 )));
1323} 1490}
1324 1491
1325void ascii (JSON *self, int enable = 1) 1492void ascii (JSON *self, int enable = 1)
1326 ALIAS: 1493 ALIAS:
1327 ascii = F_ASCII 1494 ascii = F_ASCII
1334 pretty = F_PRETTY 1501 pretty = F_PRETTY
1335 allow_nonref = F_ALLOW_NONREF 1502 allow_nonref = F_ALLOW_NONREF
1336 shrink = F_SHRINK 1503 shrink = F_SHRINK
1337 allow_blessed = F_ALLOW_BLESSED 1504 allow_blessed = F_ALLOW_BLESSED
1338 convert_blessed = F_CONV_BLESSED 1505 convert_blessed = F_CONV_BLESSED
1506 relaxed = F_RELAXED
1339 PPCODE: 1507 PPCODE:
1340{ 1508{
1341 if (enable) 1509 if (enable)
1342 self->flags |= ix; 1510 self->flags |= ix;
1343 else 1511 else
1344 self->flags &= ~ix; 1512 self->flags &= ~ix;
1345 1513
1346 XPUSHs (ST (0)); 1514 XPUSHs (ST (0));
1347} 1515}
1348 1516
1517void get_ascii (JSON *self)
1518 ALIAS:
1519 get_ascii = F_ASCII
1520 get_latin1 = F_LATIN1
1521 get_utf8 = F_UTF8
1522 get_indent = F_INDENT
1523 get_canonical = F_CANONICAL
1524 get_space_before = F_SPACE_BEFORE
1525 get_space_after = F_SPACE_AFTER
1526 get_allow_nonref = F_ALLOW_NONREF
1527 get_shrink = F_SHRINK
1528 get_allow_blessed = F_ALLOW_BLESSED
1529 get_convert_blessed = F_CONV_BLESSED
1530 get_relaxed = F_RELAXED
1531 PPCODE:
1532 XPUSHs (boolSV (self->flags & ix));
1533
1349void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1534void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1350 PPCODE: 1535 PPCODE:
1351{ 1536{
1352 UV log2 = 0; 1537 UV log2 = 0;
1353 1538
1359 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); 1544 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1360 1545
1361 XPUSHs (ST (0)); 1546 XPUSHs (ST (0));
1362} 1547}
1363 1548
1549U32 get_max_depth (JSON *self)
1550 CODE:
1551 RETVAL = DEC_DEPTH (self->flags);
1552 OUTPUT:
1553 RETVAL
1554
1364void max_size (JSON *self, UV max_size = 0) 1555void max_size (JSON *self, UV max_size = 0)
1365 PPCODE: 1556 PPCODE:
1366{ 1557{
1367 UV log2 = 0; 1558 UV log2 = 0;
1368 1559
1375 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); 1566 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1376 1567
1377 XPUSHs (ST (0)); 1568 XPUSHs (ST (0));
1378} 1569}
1379 1570
1571int get_max_size (JSON *self)
1572 CODE:
1573 RETVAL = DEC_SIZE (self->flags);
1574 OUTPUT:
1575 RETVAL
1576
1380void filter_json_objects (JSON *self, SV *cb = &PL_sv_undef) 1577void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1381 ALIAS:
1382 filter_sk_json_objects = 1
1383 PPCODE: 1578 PPCODE:
1384{ 1579{
1580 SvREFCNT_dec (self->cb_object);
1581 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1582
1583 XPUSHs (ST (0));
1584}
1585
1586void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1587 PPCODE:
1588{
1589 if (!self->cb_sk_object)
1590 self->cb_sk_object = newHV ();
1591
1385 if (!SvOK (cb)) 1592 if (SvOK (cb))
1386 cb = 0; 1593 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1387 1594 else
1388 switch (ix)
1389 { 1595 {
1390 case 0: self->cb_object = cb; break; 1596 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1597
1598 if (!HvKEYS (self->cb_sk_object))
1599 {
1600 SvREFCNT_dec (self->cb_sk_object);
1391 case 1: self->cb_sk_object = cb; break; 1601 self->cb_sk_object = 0;
1602 }
1392 } 1603 }
1393 1604
1394 XPUSHs (ST (0)); 1605 XPUSHs (ST (0));
1395} 1606}
1396 1607
1409 EXTEND (SP, 2); 1620 EXTEND (SP, 2);
1410 PUSHs (decode_json (jsonstr, self, &offset)); 1621 PUSHs (decode_json (jsonstr, self, &offset));
1411 PUSHs (sv_2mortal (newSVuv (offset))); 1622 PUSHs (sv_2mortal (newSVuv (offset)));
1412} 1623}
1413 1624
1625void DESTROY (JSON *self)
1626 CODE:
1627 SvREFCNT_dec (self->cb_sk_object);
1628 SvREFCNT_dec (self->cb_object);
1629
1414PROTOTYPES: ENABLE 1630PROTOTYPES: ENABLE
1415 1631
1416void to_json (SV *scalar) 1632void encode_json (SV *scalar)
1417 PPCODE: 1633 PPCODE:
1418{ 1634{
1419 JSON json = { F_DEFAULT | F_UTF8 }; 1635 JSON json = { F_DEFAULT | F_UTF8 };
1420 XPUSHs (encode_json (scalar, &json)); 1636 XPUSHs (encode_json (scalar, &json));
1421} 1637}
1422 1638
1423void from_json (SV *jsonstr) 1639void decode_json (SV *jsonstr)
1424 PPCODE: 1640 PPCODE:
1425{ 1641{
1426 JSON json = { F_DEFAULT | F_UTF8 }; 1642 JSON json = { F_DEFAULT | F_UTF8 };
1427 XPUSHs (decode_json (jsonstr, &json, 0)); 1643 XPUSHs (decode_json (jsonstr, &json, 0));
1428} 1644}

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines