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 1.10 by root, Sat Mar 24 01:15:22 2007 UTC vs.
Revision 1.46 by root, Mon Jun 25 22:11:39 2007 UTC

… …
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 9
10#if defined(__BORLANDC__) || defined(_MSC_VER)
11# define snprintf _snprintf // C compilers have this in stdio.h
12#endif
13
14// some old perls do not have this, try to make it work, no
15// guarentees, though. if it breaks, you get to keep the pieces.
16#ifndef UTF8_MAXBYTES
17# define UTF8_MAXBYTES 13
18#endif
19
9#define F_ASCII 0x00000001 20#define F_ASCII 0x00000001UL
21#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 22#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 23#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 24#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 25#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 26#define F_SPACE_AFTER 0x00000040UL
15#define F_ALLOW_NONREF 0x00000080 27#define F_ALLOW_NONREF 0x00000100UL
16#define F_SHRINK 0x00000100 28#define F_SHRINK 0x00000200UL
29#define F_ALLOW_BLESSED 0x00000400UL
30#define F_CONV_BLESSED 0x00000800UL // NYI
31#define F_MAXDEPTH 0xf8000000UL
32#define S_MAXDEPTH 27
33#define F_MAXSIZE 0x01f00000UL
34#define S_MAXSIZE 20
35
36#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
37#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17 38
18#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 39#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19#define F_DEFAULT 0 40#define F_DEFAULT (9UL << S_MAXDEPTH)
20 41
21#define INIT_SIZE 32 // initial scalar size to be allocated 42#define INIT_SIZE 32 // initial scalar size to be allocated
43#define INDENT_STEP 3 // spaces per indentation level
44
45#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22 46
23#define SB do { 47#define SB do {
24#define SE } while (0) 48#define SE } while (0)
25 49
26static HV *json_stash; 50#if __GNUC__ >= 3
51# define expect(expr,value) __builtin_expect ((expr),(value))
52# define inline inline
53#else
54# define expect(expr,value) (expr)
55# define inline static
56#endif
57
58#define expect_false(expr) expect ((expr) != 0, 0)
59#define expect_true(expr) expect ((expr) != 0, 1)
60
61static HV *json_stash, *json_boolean_stash; // JSON::XS::
62static SV *json_true, *json_false;
63
64/////////////////////////////////////////////////////////////////////////////
65// utility functions
66
67static UV *
68SvJSON (SV *sv)
69{
70 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
71 croak ("object is not of type JSON::XS");
72
73 return &SvUVX (SvRV (sv));
74}
75
76static void
77shrink (SV *sv)
78{
79 sv_utf8_downgrade (sv, 1);
80 if (SvLEN (sv) > SvCUR (sv) + 1)
81 {
82#ifdef SvPV_shrink_to_cur
83 SvPV_shrink_to_cur (sv);
84#elif defined (SvPV_renew)
85 SvPV_renew (sv, SvCUR (sv) + 1);
86#endif
87 }
88}
89
90// decode an utf-8 character and return it, or (UV)-1 in
91// case of an error.
92// we special-case "safe" characters from U+80 .. U+7FF,
93// but use the very good perl function to parse anything else.
94// note that we never call this function for a ascii codepoints
95inline UV
96decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
97{
98 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
99 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
100 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
101 {
102 *clen = 2;
103 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
104 }
105 else
106 {
107 *clen = (STRLEN)-1;
108 return (UV)-1;
109 }
110}
111
112/////////////////////////////////////////////////////////////////////////////
113// encoder
27 114
28// structure used for encoding JSON 115// structure used for encoding JSON
29typedef struct 116typedef struct
30{ 117{
31 char *cur; 118 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 119 char *end; // SvEND (sv)
34 SV *sv; 120 SV *sv; // result scalar
35 UV flags; 121 U32 flags; // F_*
36 int max_recurse; 122 U32 indent; // indentation level
37 int indent; 123 U32 maxdepth; // max. indentation/recursion level
38} enc_t; 124} enc_t;
39 125
40// structure used for decoding JSON 126inline void
41typedef struct
42{
43 char *cur;
44 char *end;
45 const char *err;
46 UV flags;
47} dec_t;
48
49static UV *
50SvJSON (SV *sv)
51{
52 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53 croak ("object is not of type JSON::XS");
54
55 return &SvUVX (SvRV (sv));
56}
57
58static void
59shrink (SV *sv)
60{
61 sv_utf8_downgrade (sv, 1);
62#ifdef SvPV_shrink_to_cur
63 SvPV_shrink_to_cur (sv);
64#endif
65}
66
67/////////////////////////////////////////////////////////////////////////////
68
69static void
70need (enc_t *enc, STRLEN len) 127need (enc_t *enc, STRLEN len)
71{ 128{
72 if (enc->cur + len >= enc->end) 129 if (expect_false (enc->cur + len >= enc->end))
73 { 130 {
74 STRLEN cur = enc->cur - SvPVX (enc->sv); 131 STRLEN cur = enc->cur - SvPVX (enc->sv);
75 SvGROW (enc->sv, cur + len + 1); 132 SvGROW (enc->sv, cur + len + 1);
76 enc->cur = SvPVX (enc->sv) + cur; 133 enc->cur = SvPVX (enc->sv) + cur;
77 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 134 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78 } 135 }
79} 136}
80 137
81static void 138inline void
82encode_ch (enc_t *enc, char ch) 139encode_ch (enc_t *enc, char ch)
83{ 140{
84 need (enc, 1); 141 need (enc, 1);
85 *enc->cur++ = ch; 142 *enc->cur++ = ch;
86} 143}
… …
94 151
95 while (str < end) 152 while (str < end)
96 { 153 {
97 unsigned char ch = *(unsigned char *)str; 154 unsigned char ch = *(unsigned char *)str;
98 155
99 if (ch >= 0x20 && ch < 0x80) // most common case 156 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100 { 157 {
101 if (ch == '"') // but with slow exceptions 158 if (expect_false (ch == '"')) // but with slow exceptions
102 { 159 {
103 need (enc, len += 1); 160 need (enc, len += 1);
104 *enc->cur++ = '\\'; 161 *enc->cur++ = '\\';
105 *enc->cur++ = '"'; 162 *enc->cur++ = '"';
106 } 163 }
107 else if (ch == '\\') 164 else if (expect_false (ch == '\\'))
108 { 165 {
109 need (enc, len += 1); 166 need (enc, len += 1);
110 *enc->cur++ = '\\'; 167 *enc->cur++ = '\\';
111 *enc->cur++ = '\\'; 168 *enc->cur++ = '\\';
112 } 169 }
… …
130 STRLEN clen; 187 STRLEN clen;
131 UV uch; 188 UV uch;
132 189
133 if (is_utf8) 190 if (is_utf8)
134 { 191 {
135 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 192 uch = decode_utf8 (str, end - str, &clen);
136 if (clen == (STRLEN)-1) 193 if (clen == (STRLEN)-1)
137 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); 194 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138 } 195 }
139 else 196 else
140 { 197 {
… …
143 } 200 }
144 201
145 if (uch > 0x10FFFFUL) 202 if (uch > 0x10FFFFUL)
146 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 203 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147 204
148 if (uch < 0x80 || enc->flags & F_ASCII) 205 if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
149 { 206 {
150 if (uch > 0xFFFFUL) 207 if (uch > 0xFFFFUL)
151 { 208 {
152 need (enc, len += 11); 209 need (enc, len += 11);
153 sprintf (enc->cur, "\\u%04x\\u%04x", 210 sprintf (enc->cur, "\\u%04x\\u%04x",
… …
167 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 224 *enc->cur++ = hexdigit [(uch >> 0) & 15];
168 } 225 }
169 226
170 str += clen; 227 str += clen;
171 } 228 }
229 else if (enc->flags & F_LATIN1)
230 {
231 *enc->cur++ = uch;
232 str += clen;
233 }
172 else if (is_utf8) 234 else if (is_utf8)
173 { 235 {
174 need (enc, len += clen); 236 need (enc, len += clen);
175 do 237 do
176 { 238 {
… …
178 } 240 }
179 while (--clen); 241 while (--clen);
180 } 242 }
181 else 243 else
182 { 244 {
183 need (enc, len += 10); // never more than 11 bytes needed 245 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 246 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185 ++str; 247 ++str;
186 } 248 }
187 } 249 }
188 } 250 }
… …
190 252
191 --len; 253 --len;
192 } 254 }
193} 255}
194 256
195#define INDENT SB \ 257inline void
258encode_indent (enc_t *enc)
259{
196 if (enc->flags & F_INDENT) \ 260 if (enc->flags & F_INDENT)
197 { \ 261 {
198 int i_; \ 262 int spaces = enc->indent * INDENT_STEP;
199 need (enc, enc->indent); \
200 for (i_ = enc->indent * 3; i_--; )\
201 encode_ch (enc, ' '); \
202 } \
203 SE
204 263
205#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 264 need (enc, spaces);
206#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 265 memset (enc->cur, ' ', spaces);
207#define COMMA SB \ 266 enc->cur += spaces;
267 }
268}
269
270inline void
271encode_space (enc_t *enc)
272{
273 need (enc, 1);
208 encode_ch (enc, ','); \ 274 encode_ch (enc, ' ');
275}
276
277inline void
278encode_nl (enc_t *enc)
279{
209 if (enc->flags & F_INDENT) \ 280 if (enc->flags & F_INDENT)
210 NL; \ 281 {
282 need (enc, 1);
283 encode_ch (enc, '\n');
284 }
285}
286
287inline void
288encode_comma (enc_t *enc)
289{
290 encode_ch (enc, ',');
291
292 if (enc->flags & F_INDENT)
293 encode_nl (enc);
211 else if (enc->flags & F_SPACE_AFTER) \ 294 else if (enc->flags & F_SPACE_AFTER)
212 SPACE; \ 295 encode_space (enc);
213 SE 296}
214 297
215static void encode_sv (enc_t *enc, SV *sv); 298static void encode_sv (enc_t *enc, SV *sv);
216 299
217static void 300static void
218encode_av (enc_t *enc, AV *av) 301encode_av (enc_t *enc, AV *av)
219{ 302{
220 int i, len = av_len (av); 303 int i, len = av_len (av);
221 304
222 encode_ch (enc, '['); NL; 305 if (enc->indent >= enc->maxdepth)
306 croak ("data structure too deep (hit recursion limit)");
307
308 encode_ch (enc, '['); encode_nl (enc);
223 ++enc->indent; 309 ++enc->indent;
224 310
225 for (i = 0; i <= len; ++i) 311 for (i = 0; i <= len; ++i)
226 { 312 {
227 INDENT; 313 encode_indent (enc);
228 encode_sv (enc, *av_fetch (av, i, 0)); 314 encode_sv (enc, *av_fetch (av, i, 0));
229 315
230 if (i < len) 316 if (i < len)
231 COMMA; 317 encode_comma (enc);
232 } 318 }
233 319
234 NL; 320 encode_nl (enc);
235 321
236 --enc->indent; 322 --enc->indent;
237 INDENT; encode_ch (enc, ']'); 323 encode_indent (enc); encode_ch (enc, ']');
238} 324}
239 325
240static void 326static void
241encode_he (enc_t *enc, HE *he) 327encode_he (enc_t *enc, HE *he)
242{ 328{
… …
256 else 342 else
257 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 343 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258 344
259 encode_ch (enc, '"'); 345 encode_ch (enc, '"');
260 346
261 if (enc->flags & F_SPACE_BEFORE) SPACE; 347 if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
262 encode_ch (enc, ':'); 348 encode_ch (enc, ':');
263 if (enc->flags & F_SPACE_AFTER ) SPACE; 349 if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
264 encode_sv (enc, HeVAL (he)); 350 encode_sv (enc, HeVAL (he));
265} 351}
266 352
267// compare hash entries, used when all keys are bytestrings 353// compare hash entries, used when all keys are bytestrings
268static int 354static int
… …
274 HE *b = *(HE **)b_; 360 HE *b = *(HE **)b_;
275 361
276 STRLEN la = HeKLEN (a); 362 STRLEN la = HeKLEN (a);
277 STRLEN lb = HeKLEN (b); 363 STRLEN lb = HeKLEN (b);
278 364
279 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 365 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
280 cmp = la < lb ? -1 : la == lb ? 0 : 1; 366 cmp = la - lb;
281 367
282 return cmp; 368 return cmp;
283} 369}
284 370
285// compare hash entries, used when some keys are sv's or utf-x 371// compare hash entries, used when some keys are sv's or utf-x
… …
292static void 378static void
293encode_hv (enc_t *enc, HV *hv) 379encode_hv (enc_t *enc, HV *hv)
294{ 380{
295 int count, i; 381 int count, i;
296 382
383 if (enc->indent >= enc->maxdepth)
384 croak ("data structure too deep (hit recursion limit)");
385
297 encode_ch (enc, '{'); NL; ++enc->indent; 386 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
298 387
299 if ((count = hv_iterinit (hv))) 388 if ((count = hv_iterinit (hv)))
300 { 389 {
301 // for canonical output we have to sort by keys first 390 // for canonical output we have to sort by keys first
302 // actually, this is mostly due to the stupid so-called 391 // actually, this is mostly due to the stupid so-called
303 // security workaround added somewhere in 5.8.x. 392 // security workaround added somewhere in 5.8.x.
304 // that randomises hash orderings 393 // that randomises hash orderings
305 if (enc->flags & F_CANONICAL) 394 if (enc->flags & F_CANONICAL)
306 { 395 {
307 HE *he, *hes [count];
308 int fast = 1; 396 int fast = 1;
397 HE *he;
398#if defined(__BORLANDC__) || defined(_MSC_VER)
399 HE **hes = _alloca (count * sizeof (HE));
400#else
401 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
402#endif
309 403
310 i = 0; 404 i = 0;
311 while ((he = hv_iternext (hv))) 405 while ((he = hv_iternext (hv)))
312 { 406 {
313 hes [i++] = he; 407 hes [i++] = he;
… …
337 LEAVE; 431 LEAVE;
338 } 432 }
339 433
340 for (i = 0; i < count; ++i) 434 for (i = 0; i < count; ++i)
341 { 435 {
342 INDENT; 436 encode_indent (enc);
343 encode_he (enc, hes [i]); 437 encode_he (enc, hes [i]);
344 438
345 if (i < count - 1) 439 if (i < count - 1)
346 COMMA; 440 encode_comma (enc);
347 }
348
349 NL; 441 }
442
443 encode_nl (enc);
350 } 444 }
351 else 445 else
352 { 446 {
353 SV *sv;
354 HE *he = hv_iternext (hv); 447 HE *he = hv_iternext (hv);
355 448
356 for (;;) 449 for (;;)
357 { 450 {
358 INDENT; 451 encode_indent (enc);
359 encode_he (enc, he); 452 encode_he (enc, he);
360 453
361 if (!(he = hv_iternext (hv))) 454 if (!(he = hv_iternext (hv)))
362 break; 455 break;
363 456
364 COMMA; 457 encode_comma (enc);
365 }
366
367 NL; 458 }
459
460 encode_nl (enc);
368 } 461 }
369 } 462 }
370 463
371 --enc->indent; INDENT; encode_ch (enc, '}'); 464 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
465}
466
467// encode objects, arrays and special \0=false and \1=true values.
468static void
469encode_rv (enc_t *enc, SV *sv)
470{
471 svtype svt;
472
473 SvGETMAGIC (sv);
474 svt = SvTYPE (sv);
475
476 if (expect_false (SvOBJECT (sv)))
477 {
478 if (SvSTASH (sv) == json_boolean_stash)
479 {
480 if (SvIV (sv) == 0)
481 encode_str (enc, "false", 5, 0);
482 else
483 encode_str (enc, "true", 4, 0);
484 }
485 else
486 {
487#if 0
488 if (0 && sv_derived_from (rv, "JSON::Literal"))
489 {
490 // not yet
491 }
492#endif
493 if (enc->flags & F_CONV_BLESSED)
494 {
495 // we re-bless the reference to get overload and other niceties right
496 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
497
498 if (to_json)
499 {
500 dSP;
501 ENTER;
502 SAVETMPS;
503 PUSHMARK (SP);
504 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
505
506 // calling with G_SCALAR ensures that we always get a 1 reutrn value
507 // check anyways.
508 PUTBACK;
509 assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
510 SPAGAIN;
511
512 encode_sv (enc, POPs);
513
514 FREETMPS;
515 LEAVE;
516 }
517 else if (enc->flags & F_ALLOW_BLESSED)
518 encode_str (enc, "null", 4, 0);
519 else
520 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
521 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
522 }
523 else if (enc->flags & F_ALLOW_BLESSED)
524 encode_str (enc, "null", 4, 0);
525 else
526 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
527 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
528 }
529 }
530 else if (svt == SVt_PVHV)
531 encode_hv (enc, (HV *)sv);
532 else if (svt == SVt_PVAV)
533 encode_av (enc, (AV *)sv);
534 else if (svt < SVt_PVAV)
535 {
536 if (SvNIOK (sv) && SvIV (sv) == 0)
537 encode_str (enc, "false", 5, 0);
538 else if (SvNIOK (sv) && SvIV (sv) == 1)
539 encode_str (enc, "true", 4, 0);
540 else
541 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
542 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
543 }
544 else
545 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
546 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372} 547}
373 548
374static void 549static void
375encode_sv (enc_t *enc, SV *sv) 550encode_sv (enc_t *enc, SV *sv)
376{ 551{
… …
384 encode_str (enc, str, len, SvUTF8 (sv)); 559 encode_str (enc, str, len, SvUTF8 (sv));
385 encode_ch (enc, '"'); 560 encode_ch (enc, '"');
386 } 561 }
387 else if (SvNOKp (sv)) 562 else if (SvNOKp (sv))
388 { 563 {
564 // trust that perl will do the right thing w.r.t. JSON syntax.
389 need (enc, NV_DIG + 32); 565 need (enc, NV_DIG + 32);
390 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 566 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391 enc->cur += strlen (enc->cur); 567 enc->cur += strlen (enc->cur);
392 } 568 }
393 else if (SvIOKp (sv)) 569 else if (SvIOKp (sv))
394 { 570 {
395 need (enc, 64); 571 // we assume we can always read an IV as a UV
572 if (SvUV (sv) & ~(UV)0x7fff)
573 {
574 // large integer, use the (rather slow) snprintf way.
575 need (enc, sizeof (UV) * 3);
396 enc->cur += 576 enc->cur +=
397 SvIsUV(sv) 577 SvIsUV(sv)
398 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 578 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 579 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
580 }
581 else
582 {
583 // optimise the "small number case"
584 // code will likely be branchless and use only a single multiplication
585 I32 i = SvIV (sv);
586 U32 u;
587 char digit, nz = 0;
588
589 need (enc, 6);
590
591 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
592 u = i < 0 ? -i : i;
593
594 // convert to 4.28 fixed-point representation
595 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
596
597 // now output digit by digit, each time masking out the integer part
598 // and multiplying by 5 while moving the decimal point one to the right,
599 // resulting in a net multiplication by 10.
600 // we always write the digit to memory but conditionally increment
601 // the pointer, to ease the usage of conditional move instructions.
602 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
603 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
604 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
605 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
606 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
607 }
400 } 608 }
401 else if (SvROK (sv)) 609 else if (SvROK (sv))
402 { 610 encode_rv (enc, SvRV (sv));
403 SV *rv = SvRV (sv);
404
405 if (!--enc->max_recurse)
406 croak ("data structure too deep (hit recursion limit)");
407
408 switch (SvTYPE (rv))
409 {
410 case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411 case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413 default:
414 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415 SvPV_nolen (sv));
416 }
417 }
418 else if (!SvOK (sv)) 611 else if (!SvOK (sv))
419 encode_str (enc, "null", 4, 0); 612 encode_str (enc, "null", 4, 0);
420 else 613 else
421 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 614 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422 SvPV_nolen (sv), SvFLAGS (sv)); 615 SvPV_nolen (sv), SvFLAGS (sv));
423} 616}
424 617
425static SV * 618static SV *
426encode_json (SV *scalar, UV flags) 619encode_json (SV *scalar, U32 flags)
427{ 620{
621 enc_t enc;
622
428 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) 623 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
429 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 624 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430 625
431 enc_t enc;
432 enc.flags = flags; 626 enc.flags = flags;
433 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 627 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434 enc.cur = SvPVX (enc.sv); 628 enc.cur = SvPVX (enc.sv);
435 enc.end = SvEND (enc.sv); 629 enc.end = SvEND (enc.sv);
436 enc.max_recurse = 0;
437 enc.indent = 0; 630 enc.indent = 0;
631 enc.maxdepth = DEC_DEPTH (flags);
438 632
439 SvPOK_only (enc.sv); 633 SvPOK_only (enc.sv);
440 encode_sv (&enc, scalar); 634 encode_sv (&enc, scalar);
441 635
636 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
637 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
638
442 if (!(flags & (F_ASCII | F_UTF8))) 639 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443 SvUTF8_on (enc.sv); 640 SvUTF8_on (enc.sv);
444
445 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446 641
447 if (enc.flags & F_SHRINK) 642 if (enc.flags & F_SHRINK)
448 shrink (enc.sv); 643 shrink (enc.sv);
449 644
450 return enc.sv; 645 return enc.sv;
451} 646}
452 647
453///////////////////////////////////////////////////////////////////////////// 648/////////////////////////////////////////////////////////////////////////////
649// decoder
454 650
455#define WS \ 651// structure used for decoding JSON
652typedef struct
653{
654 char *cur; // current parser pointer
655 char *end; // end of input string
656 const char *err; // parse error, if != 0
657 U32 flags; // F_*
658 U32 depth; // recursion depth
659 U32 maxdepth; // recursion depth limit
660} dec_t;
661
662inline void
663decode_ws (dec_t *dec)
664{
456 for (;;) \ 665 for (;;)
457 { \ 666 {
458 char ch = *dec->cur; \ 667 char ch = *dec->cur;
668
459 if (ch > 0x20 \ 669 if (ch > 0x20
460 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 670 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
461 break; \ 671 break;
672
462 ++dec->cur; \ 673 ++dec->cur;
463 } 674 }
675}
464 676
465#define ERR(reason) SB dec->err = reason; goto fail; SE 677#define ERR(reason) SB dec->err = reason; goto fail; SE
678
466#define EXPECT_CH(ch) SB \ 679#define EXPECT_CH(ch) SB \
467 if (*dec->cur != ch) \ 680 if (*dec->cur != ch) \
468 ERR (# ch " expected"); \ 681 ERR (# ch " expected"); \
469 ++dec->cur; \ 682 ++dec->cur; \
470 SE 683 SE
471 684
685#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
686#define DEC_DEC_DEPTH --dec->depth
687
472static SV *decode_sv (dec_t *dec); 688static SV *decode_sv (dec_t *dec);
473 689
474static signed char decode_hexdigit[256]; 690static signed char decode_hexdigit[256];
475 691
476static UV 692static UV
477decode_4hex (dec_t *dec) 693decode_4hex (dec_t *dec)
478{ 694{
479 signed char d1, d2, d3, d4; 695 signed char d1, d2, d3, d4;
696 unsigned char *cur = (unsigned char *)dec->cur;
480 697
481 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 698 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482 if (d1 < 0) ERR ("four hexadecimal digits expected"); 699 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 700 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484 if (d2 < 0) ERR ("four hexadecimal digits expected"); 701 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486 if (d3 < 0) ERR ("four hexadecimal digits expected");
487 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488 if (d4 < 0) ERR ("four hexadecimal digits expected");
489 702
490 dec->cur += 4; 703 dec->cur += 4;
491 704
492 return ((UV)d1) << 12 705 return ((UV)d1) << 12
493 | ((UV)d2) << 8 706 | ((UV)d2) << 8
… …
496 709
497fail: 710fail:
498 return (UV)-1; 711 return (UV)-1;
499} 712}
500 713
501#define APPEND_GROW(n) SB \
502 if (cur + (n) >= end) \
503 { \
504 STRLEN ofs = cur - SvPVX (sv); \
505 SvGROW (sv, ofs + (n) + 1); \
506 cur = SvPVX (sv) + ofs; \
507 end = SvEND (sv); \
508 } \
509 SE
510
511#define APPEND_CH(ch) SB \
512 APPEND_GROW (1); \
513 *cur++ = (ch); \
514 SE
515
516static SV * 714static SV *
517decode_str (dec_t *dec) 715decode_str (dec_t *dec)
518{ 716{
519 SV *sv = NEWSV (0,2); 717 SV *sv = 0;
520 int utf8 = 0; 718 int utf8 = 0;
521 char *cur = SvPVX (sv); 719 char *dec_cur = dec->cur;
522 char *end = SvEND (sv);
523 720
524 for (;;) 721 do
525 { 722 {
526 unsigned char ch = *(unsigned char *)dec->cur; 723 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
724 char *cur = buf;
527 725
528 if (ch == '"') 726 do
529 break;
530 else if (ch == '\\')
531 { 727 {
532 switch (*++dec->cur) 728 unsigned char ch = *(unsigned char *)dec_cur++;
729
730 if (expect_false (ch == '"'))
731 {
732 --dec_cur;
733 break;
533 { 734 }
534 case '\\': 735 else if (expect_false (ch == '\\'))
535 case '/': 736 {
536 case '"': APPEND_CH (*dec->cur++); break; 737 switch (*dec_cur)
537
538 case 'b': APPEND_CH ('\010'); ++dec->cur; break;
539 case 't': APPEND_CH ('\011'); ++dec->cur; break;
540 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544 case 'u':
545 { 738 {
546 UV lo, hi; 739 case '\\':
547 ++dec->cur; 740 case '/':
741 case '"': *cur++ = *dec_cur++; break;
548 742
549 hi = decode_4hex (dec); 743 case 'b': ++dec_cur; *cur++ = '\010'; break;
550 if (hi == (UV)-1) 744 case 't': ++dec_cur; *cur++ = '\011'; break;
551 goto fail; 745 case 'n': ++dec_cur; *cur++ = '\012'; break;
746 case 'f': ++dec_cur; *cur++ = '\014'; break;
747 case 'r': ++dec_cur; *cur++ = '\015'; break;
552 748
553 // possibly a surrogate pair 749 case 'u':
554 if (hi >= 0xd800 && hi < 0xdc00)
555 { 750 {
556 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 751 UV lo, hi;
557 ERR ("missing low surrogate character in surrogate pair"); 752 ++dec_cur;
558 753
559 dec->cur += 2; 754 dec->cur = dec_cur;
560
561 lo = decode_4hex (dec); 755 hi = decode_4hex (dec);
756 dec_cur = dec->cur;
562 if (lo == (UV)-1) 757 if (hi == (UV)-1)
563 goto fail; 758 goto fail;
564 759
760 // possibly a surrogate pair
761 if (hi >= 0xd800)
762 if (hi < 0xdc00)
763 {
764 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
765 ERR ("missing low surrogate character in surrogate pair");
766
767 dec_cur += 2;
768
769 dec->cur = dec_cur;
770 lo = decode_4hex (dec);
771 dec_cur = dec->cur;
772 if (lo == (UV)-1)
773 goto fail;
774
565 if (lo < 0xdc00 || lo >= 0xe000) 775 if (lo < 0xdc00 || lo >= 0xe000)
566 ERR ("surrogate pair expected"); 776 ERR ("surrogate pair expected");
567 777
568 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 778 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
779 }
780 else if (hi < 0xe000)
781 ERR ("missing high surrogate character in surrogate pair");
782
783 if (hi >= 0x80)
784 {
785 utf8 = 1;
786
787 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
788 }
789 else
790 *cur++ = hi;
569 } 791 }
570 else if (hi >= 0xdc00 && hi < 0xe000)
571 ERR ("missing high surrogate character in surrogate pair");
572
573 if (hi >= 0x80)
574 { 792 break;
575 utf8 = 1;
576 793
577 APPEND_GROW (4); // at most 4 bytes for 21 bits
578 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579 }
580 else 794 default:
581 APPEND_CH (hi); 795 --dec_cur;
796 ERR ("illegal backslash escape sequence in string");
582 } 797 }
583 break; 798 }
799 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
800 *cur++ = ch;
801 else if (ch >= 0x80)
802 {
803 STRLEN clen;
804 UV uch;
584 805
585 default:
586 --dec->cur; 806 --dec_cur;
587 ERR ("illegal backslash escape sequence in string"); 807
808 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
809 if (clen == (STRLEN)-1)
810 ERR ("malformed UTF-8 character in JSON string");
811
812 do
813 *cur++ = *dec_cur++;
814 while (--clen);
815
816 utf8 = 1;
817 }
818 else
819 {
820 --dec_cur;
821
822 if (!ch)
823 ERR ("unexpected end of string while parsing JSON string");
824 else
825 ERR ("invalid character encountered while parsing JSON string");
588 } 826 }
589 } 827 }
590 else if (ch >= 0x20 && ch <= 0x7f) 828 while (cur < buf + SHORT_STRING_LEN);
591 APPEND_CH (*dec->cur++); 829
592 else if (ch >= 0x80)
593 { 830 {
594 STRLEN clen; 831 STRLEN len = cur - buf;
595 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596 if (clen == (STRLEN)-1)
597 ERR ("malformed UTF-8 character in JSON string");
598 832
599 APPEND_GROW (clen); 833 if (sv)
600 do
601 { 834 {
602 *cur++ = *dec->cur++; 835 SvGROW (sv, SvCUR (sv) + len + 1);
836 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
837 SvCUR_set (sv, SvCUR (sv) + len);
603 } 838 }
604 while (--clen);
605
606 utf8 = 1;
607 }
608 else if (dec->cur == dec->end)
609 ERR ("unexpected end of string while parsing json string");
610 else 839 else
611 ERR ("invalid character encountered"); 840 sv = newSVpvn (buf, len);
612 } 841 }
842 }
843 while (*dec_cur != '"');
613 844
614 ++dec->cur; 845 ++dec_cur;
615 846
616 SvCUR_set (sv, cur - SvPVX (sv)); 847 if (sv)
617 848 {
618 SvPOK_only (sv); 849 SvPOK_only (sv);
619 *SvEND (sv) = 0; 850 *SvEND (sv) = 0;
620 851
621 if (utf8) 852 if (utf8)
622 SvUTF8_on (sv); 853 SvUTF8_on (sv);
854 }
855 else
856 sv = newSVpvn ("", 0);
623 857
624 if (dec->flags & F_SHRINK) 858 dec->cur = dec_cur;
625 shrink (sv);
626
627 return sv; 859 return sv;
628 860
629fail: 861fail:
630 SvREFCNT_dec (sv); 862 dec->cur = dec_cur;
631 return 0; 863 return 0;
632} 864}
633 865
634static SV * 866static SV *
635decode_num (dec_t *dec) 867decode_num (dec_t *dec)
… …
693 is_nv = 1; 925 is_nv = 1;
694 } 926 }
695 927
696 if (!is_nv) 928 if (!is_nv)
697 { 929 {
698 UV uv; 930 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
699 int numtype = grok_number (start, dec->cur - start, &uv); 931 if (*start == '-')
700 if (numtype & IS_NUMBER_IN_UV) 932 switch (dec->cur - start)
701 if (numtype & IS_NUMBER_NEG)
702 { 933 {
703 if (uv < (UV)IV_MIN) 934 case 2: return newSViv (-( start [1] - '0' * 1));
704 return newSViv (-(IV)uv); 935 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
936 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
937 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
705 } 938 }
939 else
940 switch (dec->cur - start)
941 {
942 case 1: return newSViv ( start [0] - '0' * 1);
943 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
944 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
945 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
946 }
947
948 {
949 UV uv;
950 int numtype = grok_number (start, dec->cur - start, &uv);
951 if (numtype & IS_NUMBER_IN_UV)
952 if (numtype & IS_NUMBER_NEG)
953 {
954 if (uv < (UV)IV_MIN)
955 return newSViv (-(IV)uv);
956 }
706 else 957 else
707 return newSVuv (uv); 958 return newSVuv (uv);
959
960 // here would likely be the place for bigint support
708 } 961 }
962 }
709 963
964 // if we ever support bigint or bigfloat, this is the place for bigfloat
710 return newSVnv (Atof (start)); 965 return newSVnv (Atof (start));
711 966
712fail: 967fail:
713 return 0; 968 return 0;
714} 969}
… …
716static SV * 971static SV *
717decode_av (dec_t *dec) 972decode_av (dec_t *dec)
718{ 973{
719 AV *av = newAV (); 974 AV *av = newAV ();
720 975
721 WS; 976 DEC_INC_DEPTH;
977 decode_ws (dec);
978
722 if (*dec->cur == ']') 979 if (*dec->cur == ']')
723 ++dec->cur; 980 ++dec->cur;
724 else 981 else
725 for (;;) 982 for (;;)
726 { 983 {
… …
730 if (!value) 987 if (!value)
731 goto fail; 988 goto fail;
732 989
733 av_push (av, value); 990 av_push (av, value);
734 991
735 WS; 992 decode_ws (dec);
736 993
737 if (*dec->cur == ']') 994 if (*dec->cur == ']')
738 { 995 {
739 ++dec->cur; 996 ++dec->cur;
740 break; 997 break;
… …
744 ERR (", or ] expected while parsing array"); 1001 ERR (", or ] expected while parsing array");
745 1002
746 ++dec->cur; 1003 ++dec->cur;
747 } 1004 }
748 1005
1006 DEC_DEC_DEPTH;
749 return newRV_noinc ((SV *)av); 1007 return newRV_noinc ((SV *)av);
750 1008
751fail: 1009fail:
752 SvREFCNT_dec (av); 1010 SvREFCNT_dec (av);
1011 DEC_DEC_DEPTH;
753 return 0; 1012 return 0;
754} 1013}
755 1014
756static SV * 1015static SV *
757decode_hv (dec_t *dec) 1016decode_hv (dec_t *dec)
758{ 1017{
759 HV *hv = newHV (); 1018 HV *hv = newHV ();
760 1019
761 WS; 1020 DEC_INC_DEPTH;
1021 decode_ws (dec);
1022
762 if (*dec->cur == '}') 1023 if (*dec->cur == '}')
763 ++dec->cur; 1024 ++dec->cur;
764 else 1025 else
765 for (;;) 1026 for (;;)
766 { 1027 {
1028 decode_ws (dec); EXPECT_CH ('"');
1029
1030 // heuristic: assume that
1031 // a) decode_str + hv_store_ent are abysmally slow
1032 // b) most hash keys are short, simple ascii text
1033 // so try to "fast-match" such strings to avoid
1034 // the overhead of hv_store_ent.
1035 {
767 SV *key, *value; 1036 SV *value;
1037 char *p = dec->cur;
1038 char *e = p + 24; // only try up to 24 bytes
768 1039
769 WS; EXPECT_CH ('"'); 1040 for (;;)
770
771 key = decode_str (dec);
772 if (!key)
773 goto fail;
774
775 WS; EXPECT_CH (':');
776
777 value = decode_sv (dec);
778 if (!value)
779 { 1041 {
1042 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1043 {
1044 // slow path, back up and use decode_str
1045 SV *key = decode_str (dec);
1046 if (!key)
1047 goto fail;
1048
1049 decode_ws (dec); EXPECT_CH (':');
1050
1051 value = decode_sv (dec);
1052 if (!value)
1053 {
1054 SvREFCNT_dec (key);
1055 goto fail;
1056 }
1057
1058 hv_store_ent (hv, key, value, 0);
780 SvREFCNT_dec (key); 1059 SvREFCNT_dec (key);
1060
1061 break;
1062 }
1063 else if (*p == '"')
1064 {
1065 // fast path, got a simple key
1066 char *key = dec->cur;
1067 int len = p - key;
1068 dec->cur = p + 1;
1069
1070 decode_ws (dec); EXPECT_CH (':');
1071
1072 value = decode_sv (dec);
1073 if (!value)
781 goto fail; 1074 goto fail;
1075
1076 hv_store (hv, key, len, value, 0);
1077
1078 break;
1079 }
1080
1081 ++p;
782 } 1082 }
783
784 //TODO: optimise
785 hv_store_ent (hv, key, value, 0);
786
787 WS; 1083 }
1084
1085 decode_ws (dec);
788 1086
789 if (*dec->cur == '}') 1087 if (*dec->cur == '}')
790 { 1088 {
791 ++dec->cur; 1089 ++dec->cur;
792 break; 1090 break;
… …
796 ERR (", or } expected while parsing object/hash"); 1094 ERR (", or } expected while parsing object/hash");
797 1095
798 ++dec->cur; 1096 ++dec->cur;
799 } 1097 }
800 1098
1099 DEC_DEC_DEPTH;
801 return newRV_noinc ((SV *)hv); 1100 return newRV_noinc ((SV *)hv);
802 1101
803fail: 1102fail:
804 SvREFCNT_dec (hv); 1103 SvREFCNT_dec (hv);
1104 DEC_DEC_DEPTH;
805 return 0; 1105 return 0;
806} 1106}
807 1107
808static SV * 1108static SV *
809decode_sv (dec_t *dec) 1109decode_sv (dec_t *dec)
810{ 1110{
811 WS; 1111 decode_ws (dec);
1112
1113 // the beauty of JSON: you need exactly one character lookahead
1114 // to parse anything.
812 switch (*dec->cur) 1115 switch (*dec->cur)
813 { 1116 {
814 case '"': ++dec->cur; return decode_str (dec); 1117 case '"': ++dec->cur; return decode_str (dec);
815 case '[': ++dec->cur; return decode_av (dec); 1118 case '[': ++dec->cur; return decode_av (dec);
816 case '{': ++dec->cur; return decode_hv (dec); 1119 case '{': ++dec->cur; return decode_hv (dec);
… …
822 1125
823 case 't': 1126 case 't':
824 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1127 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825 { 1128 {
826 dec->cur += 4; 1129 dec->cur += 4;
827 return newSViv (1); 1130 return SvREFCNT_inc (json_true);
828 } 1131 }
829 else 1132 else
830 ERR ("'true' expected"); 1133 ERR ("'true' expected");
831 1134
832 break; 1135 break;
833 1136
834 case 'f': 1137 case 'f':
835 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1138 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836 { 1139 {
837 dec->cur += 5; 1140 dec->cur += 5;
838 return newSViv (0); 1141 return SvREFCNT_inc (json_false);
839 } 1142 }
840 else 1143 else
841 ERR ("'false' expected"); 1144 ERR ("'false' expected");
842 1145
843 break; 1146 break;
… …
852 ERR ("'null' expected"); 1155 ERR ("'null' expected");
853 1156
854 break; 1157 break;
855 1158
856 default: 1159 default:
857 ERR ("malformed json string, neither array, object, number, string or atom"); 1160 ERR ("malformed JSON string, neither array, object, number, string or atom");
858 break; 1161 break;
859 } 1162 }
860 1163
861fail: 1164fail:
862 return 0; 1165 return 0;
863} 1166}
864 1167
865static SV * 1168static SV *
866decode_json (SV *string, UV flags) 1169decode_json (SV *string, U32 flags, UV *offset_return)
867{ 1170{
1171 dec_t dec;
1172 UV offset;
868 SV *sv; 1173 SV *sv;
1174
1175 SvGETMAGIC (string);
1176 SvUPGRADE (string, SVt_PV);
1177
1178 if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
1179 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1180 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
869 1181
870 if (flags & F_UTF8) 1182 if (flags & F_UTF8)
871 sv_utf8_downgrade (string, 0); 1183 sv_utf8_downgrade (string, 0);
872 else 1184 else
873 sv_utf8_upgrade (string); 1185 sv_utf8_upgrade (string);
874 1186
875 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1187 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876 1188
877 dec_t dec;
878 dec.flags = flags; 1189 dec.flags = flags;
879 dec.cur = SvPVX (string); 1190 dec.cur = SvPVX (string);
880 dec.end = SvEND (string); 1191 dec.end = SvEND (string);
881 dec.err = 0; 1192 dec.err = 0;
1193 dec.depth = 0;
1194 dec.maxdepth = DEC_DEPTH (dec.flags);
882 1195
1196 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
883 sv = decode_sv (&dec); 1197 sv = decode_sv (&dec);
884 1198
1199 if (!(offset_return || !sv))
1200 {
1201 // check for trailing garbage
1202 decode_ws (&dec);
1203
1204 if (*dec.cur)
1205 {
1206 dec.err = "garbage after JSON object";
1207 SvREFCNT_dec (sv);
1208 sv = 0;
1209 }
1210 }
1211
1212 if (offset_return || !sv)
1213 {
1214 offset = dec.flags & F_UTF8
1215 ? dec.cur - SvPVX (string)
1216 : utf8_distance (dec.cur, SvPVX (string));
1217
1218 if (offset_return)
1219 *offset_return = offset;
1220 }
1221
885 if (!sv) 1222 if (!sv)
886 { 1223 {
887 IV offset = dec.flags & F_UTF8
888 ? dec.cur - SvPVX (string)
889 : utf8_distance (dec.cur, SvPVX (string));
890 SV *uni = sv_newmortal (); 1224 SV *uni = sv_newmortal ();
891 1225
892 // horrible hack to silence warning inside pv_uni_display 1226 // horrible hack to silence warning inside pv_uni_display
893 COP cop = *PL_curcop; 1227 COP cop = *PL_curcop;
894 cop.cop_warnings = pWARN_NONE; 1228 cop.cop_warnings = pWARN_NONE;
… …
896 SAVEVPTR (PL_curcop); 1230 SAVEVPTR (PL_curcop);
897 PL_curcop = &cop; 1231 PL_curcop = &cop;
898 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1232 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899 LEAVE; 1233 LEAVE;
900 1234
901 croak ("%s, at character offset %d (%s)", 1235 croak ("%s, at character offset %d [\"%s\"]",
902 dec.err, 1236 dec.err,
903 (int)offset, 1237 (int)offset,
904 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1238 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905 } 1239 }
906 1240
… …
910 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1244 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911 1245
912 return sv; 1246 return sv;
913} 1247}
914 1248
1249/////////////////////////////////////////////////////////////////////////////
1250// XS interface functions
1251
915MODULE = JSON::XS PACKAGE = JSON::XS 1252MODULE = JSON::XS PACKAGE = JSON::XS
916 1253
917BOOT: 1254BOOT:
918{ 1255{
919 int i; 1256 int i;
920 1257
921 memset (decode_hexdigit, 0xff, 256);
922 for (i = 10; i--; ) 1258 for (i = 0; i < 256; ++i)
923 decode_hexdigit ['0' + i] = i; 1259 decode_hexdigit [i] =
1260 i >= '0' && i <= '9' ? i - '0'
1261 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1262 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1263 : -1;
924 1264
925 for (i = 7; i--; )
926 {
927 decode_hexdigit ['a' + i] = 10 + i;
928 decode_hexdigit ['A' + i] = 10 + i;
929 }
930
931 json_stash = gv_stashpv ("JSON::XS", 1); 1265 json_stash = gv_stashpv ("JSON::XS" , 1);
1266 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1267
1268 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1269 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
932} 1270}
933 1271
934PROTOTYPES: DISABLE 1272PROTOTYPES: DISABLE
935 1273
936SV *new (char *dummy) 1274SV *new (char *dummy)
… …
939 OUTPUT: 1277 OUTPUT:
940 RETVAL 1278 RETVAL
941 1279
942SV *ascii (SV *self, int enable = 1) 1280SV *ascii (SV *self, int enable = 1)
943 ALIAS: 1281 ALIAS:
944 ascii = F_ASCII 1282 ascii = F_ASCII
1283 latin1 = F_LATIN1
945 utf8 = F_UTF8 1284 utf8 = F_UTF8
946 indent = F_INDENT 1285 indent = F_INDENT
947 canonical = F_CANONICAL 1286 canonical = F_CANONICAL
948 space_before = F_SPACE_BEFORE 1287 space_before = F_SPACE_BEFORE
949 space_after = F_SPACE_AFTER 1288 space_after = F_SPACE_AFTER
950 pretty = F_PRETTY 1289 pretty = F_PRETTY
951 allow_nonref = F_ALLOW_NONREF 1290 allow_nonref = F_ALLOW_NONREF
952 shrink = F_SHRINK 1291 shrink = F_SHRINK
1292 allow_blessed = F_ALLOW_BLESSED
1293 convert_blessed = F_CONV_BLESSED
953 CODE: 1294 CODE:
954{ 1295{
955 UV *uv = SvJSON (self); 1296 UV *uv = SvJSON (self);
956 if (enable) 1297 if (enable)
957 *uv |= ix; 1298 *uv |= ix;
… …
961 RETVAL = newSVsv (self); 1302 RETVAL = newSVsv (self);
962} 1303}
963 OUTPUT: 1304 OUTPUT:
964 RETVAL 1305 RETVAL
965 1306
1307SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
1308 CODE:
1309{
1310 UV *uv = SvJSON (self);
1311 UV log2 = 0;
1312
1313 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1314
1315 while ((1UL << log2) < max_depth)
1316 ++log2;
1317
1318 *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1319
1320 RETVAL = newSVsv (self);
1321}
1322 OUTPUT:
1323 RETVAL
1324
1325SV *max_size (SV *self, UV max_size = 0)
1326 CODE:
1327{
1328 UV *uv = SvJSON (self);
1329 UV log2 = 0;
1330
1331 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1332 if (max_size == 1) max_size = 2;
1333
1334 while ((1UL << log2) < max_size)
1335 ++log2;
1336
1337 *uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1338
1339 RETVAL = newSVsv (self);
1340}
1341 OUTPUT:
1342 RETVAL
1343
966void encode (SV *self, SV *scalar) 1344void encode (SV *self, SV *scalar)
967 PPCODE: 1345 PPCODE:
968 XPUSHs (encode_json (scalar, *SvJSON (self))); 1346 XPUSHs (encode_json (scalar, *SvJSON (self)));
969 1347
970void decode (SV *self, SV *jsonstr) 1348void decode (SV *self, SV *jsonstr)
971 PPCODE: 1349 PPCODE:
972 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1350 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1351
1352void decode_prefix (SV *self, SV *jsonstr)
1353 PPCODE:
1354{
1355 UV offset;
1356 EXTEND (SP, 2);
1357 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1358 PUSHs (sv_2mortal (newSVuv (offset)));
1359}
973 1360
974PROTOTYPES: ENABLE 1361PROTOTYPES: ENABLE
975 1362
976void to_json (SV *scalar) 1363void to_json (SV *scalar)
1364 ALIAS:
1365 objToJson = 0
977 PPCODE: 1366 PPCODE:
978 XPUSHs (encode_json (scalar, F_UTF8)); 1367 XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
979 1368
980void from_json (SV *jsonstr) 1369void from_json (SV *jsonstr)
1370 ALIAS:
1371 jsonToObj = 0
981 PPCODE: 1372 PPCODE:
982 XPUSHs (decode_json (jsonstr, F_UTF8)); 1373 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
983 1374

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines