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 1.48 by root, Sun Jul 1 22:20:00 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
64typedef struct {
65 U32 flags;
66} JSON;
67
68/////////////////////////////////////////////////////////////////////////////
69// utility functions
70
71inline void
72shrink (SV *sv)
73{
74 sv_utf8_downgrade (sv, 1);
75 if (SvLEN (sv) > SvCUR (sv) + 1)
76 {
77#ifdef SvPV_shrink_to_cur
78 SvPV_shrink_to_cur (sv);
79#elif defined (SvPV_renew)
80 SvPV_renew (sv, SvCUR (sv) + 1);
81#endif
82 }
83}
84
85// decode an utf-8 character and return it, or (UV)-1 in
86// case of an error.
87// we special-case "safe" characters from U+80 .. U+7FF,
88// but use the very good perl function to parse anything else.
89// note that we never call this function for a ascii codepoints
90inline UV
91decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
92{
93 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
94 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
95 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
96 {
97 *clen = 2;
98 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
99 }
100 else
101 {
102 *clen = (STRLEN)-1;
103 return (UV)-1;
104 }
105}
106
107/////////////////////////////////////////////////////////////////////////////
108// encoder
27 109
28// structure used for encoding JSON 110// structure used for encoding JSON
29typedef struct 111typedef struct
30{ 112{
31 char *cur; 113 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 114 char *end; // SvEND (sv)
34 SV *sv; 115 SV *sv; // result scalar
35 UV flags; 116 JSON json;
36 int max_recurse; 117 U32 indent; // indentation level
37 int indent; 118 U32 maxdepth; // max. indentation/recursion level
38} enc_t; 119} enc_t;
39 120
40// structure used for decoding JSON 121inline 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) 122need (enc_t *enc, STRLEN len)
71{ 123{
72 if (enc->cur + len >= enc->end) 124 if (expect_false (enc->cur + len >= enc->end))
73 { 125 {
74 STRLEN cur = enc->cur - SvPVX (enc->sv); 126 STRLEN cur = enc->cur - SvPVX (enc->sv);
75 SvGROW (enc->sv, cur + len + 1); 127 SvGROW (enc->sv, cur + len + 1);
76 enc->cur = SvPVX (enc->sv) + cur; 128 enc->cur = SvPVX (enc->sv) + cur;
77 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 129 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78 } 130 }
79} 131}
80 132
81static void 133inline void
82encode_ch (enc_t *enc, char ch) 134encode_ch (enc_t *enc, char ch)
83{ 135{
84 need (enc, 1); 136 need (enc, 1);
85 *enc->cur++ = ch; 137 *enc->cur++ = ch;
86} 138}
… …
94 146
95 while (str < end) 147 while (str < end)
96 { 148 {
97 unsigned char ch = *(unsigned char *)str; 149 unsigned char ch = *(unsigned char *)str;
98 150
99 if (ch >= 0x20 && ch < 0x80) // most common case 151 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100 { 152 {
101 if (ch == '"') // but with slow exceptions 153 if (expect_false (ch == '"')) // but with slow exceptions
102 { 154 {
103 need (enc, len += 1); 155 need (enc, len += 1);
104 *enc->cur++ = '\\'; 156 *enc->cur++ = '\\';
105 *enc->cur++ = '"'; 157 *enc->cur++ = '"';
106 } 158 }
107 else if (ch == '\\') 159 else if (expect_false (ch == '\\'))
108 { 160 {
109 need (enc, len += 1); 161 need (enc, len += 1);
110 *enc->cur++ = '\\'; 162 *enc->cur++ = '\\';
111 *enc->cur++ = '\\'; 163 *enc->cur++ = '\\';
112 } 164 }
… …
130 STRLEN clen; 182 STRLEN clen;
131 UV uch; 183 UV uch;
132 184
133 if (is_utf8) 185 if (is_utf8)
134 { 186 {
135 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 187 uch = decode_utf8 (str, end - str, &clen);
136 if (clen == (STRLEN)-1) 188 if (clen == (STRLEN)-1)
137 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); 189 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138 } 190 }
139 else 191 else
140 { 192 {
… …
143 } 195 }
144 196
145 if (uch > 0x10FFFFUL) 197 if (uch > 0x10FFFFUL)
146 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 198 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147 199
148 if (uch < 0x80 || enc->flags & F_ASCII) 200 if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
149 { 201 {
150 if (uch > 0xFFFFUL) 202 if (uch > 0xFFFFUL)
151 { 203 {
152 need (enc, len += 11); 204 need (enc, len += 11);
153 sprintf (enc->cur, "\\u%04x\\u%04x", 205 sprintf (enc->cur, "\\u%04x\\u%04x",
… …
167 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 219 *enc->cur++ = hexdigit [(uch >> 0) & 15];
168 } 220 }
169 221
170 str += clen; 222 str += clen;
171 } 223 }
224 else if (enc->json.flags & F_LATIN1)
225 {
226 *enc->cur++ = uch;
227 str += clen;
228 }
172 else if (is_utf8) 229 else if (is_utf8)
173 { 230 {
174 need (enc, len += clen); 231 need (enc, len += clen);
175 do 232 do
176 { 233 {
… …
178 } 235 }
179 while (--clen); 236 while (--clen);
180 } 237 }
181 else 238 else
182 { 239 {
183 need (enc, len += 10); // never more than 11 bytes needed 240 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 241 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185 ++str; 242 ++str;
186 } 243 }
187 } 244 }
188 } 245 }
… …
190 247
191 --len; 248 --len;
192 } 249 }
193} 250}
194 251
195#define INDENT SB \ 252inline void
253encode_indent (enc_t *enc)
254{
196 if (enc->flags & F_INDENT) \ 255 if (enc->json.flags & F_INDENT)
197 { \ 256 {
198 int i_; \ 257 int spaces = enc->indent * INDENT_STEP;
199 need (enc, enc->indent); \ 258
200 for (i_ = enc->indent * 3; i_--; )\ 259 need (enc, spaces);
260 memset (enc->cur, ' ', spaces);
261 enc->cur += spaces;
262 }
263}
264
265inline void
266encode_space (enc_t *enc)
267{
268 need (enc, 1);
269 encode_ch (enc, ' ');
270}
271
272inline void
273encode_nl (enc_t *enc)
274{
275 if (enc->json.flags & F_INDENT)
276 {
277 need (enc, 1);
201 encode_ch (enc, ' '); \ 278 encode_ch (enc, '\n');
202 } \ 279 }
203 SE 280}
204 281
205#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 282inline void
206#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 283encode_comma (enc_t *enc)
207#define COMMA SB \ 284{
208 encode_ch (enc, ','); \ 285 encode_ch (enc, ',');
286
209 if (enc->flags & F_INDENT) \ 287 if (enc->json.flags & F_INDENT)
210 NL; \ 288 encode_nl (enc);
211 else if (enc->flags & F_SPACE_AFTER) \ 289 else if (enc->json.flags & F_SPACE_AFTER)
212 SPACE; \ 290 encode_space (enc);
213 SE 291}
214 292
215static void encode_sv (enc_t *enc, SV *sv); 293static void encode_sv (enc_t *enc, SV *sv);
216 294
217static void 295static void
218encode_av (enc_t *enc, AV *av) 296encode_av (enc_t *enc, AV *av)
219{ 297{
220 int i, len = av_len (av); 298 int i, len = av_len (av);
221 299
222 encode_ch (enc, '['); NL; 300 if (enc->indent >= enc->maxdepth)
301 croak ("data structure too deep (hit recursion limit)");
302
303 encode_ch (enc, '['); encode_nl (enc);
223 ++enc->indent; 304 ++enc->indent;
224 305
225 for (i = 0; i <= len; ++i) 306 for (i = 0; i <= len; ++i)
226 { 307 {
227 INDENT; 308 encode_indent (enc);
228 encode_sv (enc, *av_fetch (av, i, 0)); 309 encode_sv (enc, *av_fetch (av, i, 0));
229 310
230 if (i < len) 311 if (i < len)
231 COMMA; 312 encode_comma (enc);
232 } 313 }
233 314
234 NL; 315 encode_nl (enc);
235 316
236 --enc->indent; 317 --enc->indent;
237 INDENT; encode_ch (enc, ']'); 318 encode_indent (enc); encode_ch (enc, ']');
238} 319}
239 320
240static void 321static void
241encode_he (enc_t *enc, HE *he) 322encode_he (enc_t *enc, HE *he)
242{ 323{
… …
256 else 337 else
257 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 338 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258 339
259 encode_ch (enc, '"'); 340 encode_ch (enc, '"');
260 341
261 if (enc->flags & F_SPACE_BEFORE) SPACE; 342 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262 encode_ch (enc, ':'); 343 encode_ch (enc, ':');
263 if (enc->flags & F_SPACE_AFTER ) SPACE; 344 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264 encode_sv (enc, HeVAL (he)); 345 encode_sv (enc, HeVAL (he));
265} 346}
266 347
267// compare hash entries, used when all keys are bytestrings 348// compare hash entries, used when all keys are bytestrings
268static int 349static int
… …
274 HE *b = *(HE **)b_; 355 HE *b = *(HE **)b_;
275 356
276 STRLEN la = HeKLEN (a); 357 STRLEN la = HeKLEN (a);
277 STRLEN lb = HeKLEN (b); 358 STRLEN lb = HeKLEN (b);
278 359
279 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 360 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
280 cmp = la < lb ? -1 : la == lb ? 0 : 1; 361 cmp = la - lb;
281 362
282 return cmp; 363 return cmp;
283} 364}
284 365
285// compare hash entries, used when some keys are sv's or utf-x 366// compare hash entries, used when some keys are sv's or utf-x
… …
292static void 373static void
293encode_hv (enc_t *enc, HV *hv) 374encode_hv (enc_t *enc, HV *hv)
294{ 375{
295 int count, i; 376 int count, i;
296 377
378 if (enc->indent >= enc->maxdepth)
379 croak ("data structure too deep (hit recursion limit)");
380
297 encode_ch (enc, '{'); NL; ++enc->indent; 381 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
298 382
299 if ((count = hv_iterinit (hv))) 383 if ((count = hv_iterinit (hv)))
300 { 384 {
301 // for canonical output we have to sort by keys first 385 // for canonical output we have to sort by keys first
302 // actually, this is mostly due to the stupid so-called 386 // actually, this is mostly due to the stupid so-called
303 // security workaround added somewhere in 5.8.x. 387 // security workaround added somewhere in 5.8.x.
304 // that randomises hash orderings 388 // that randomises hash orderings
305 if (enc->flags & F_CANONICAL) 389 if (enc->json.flags & F_CANONICAL)
306 { 390 {
307 HE *he, *hes [count];
308 int fast = 1; 391 int fast = 1;
392 HE *he;
393#if defined(__BORLANDC__) || defined(_MSC_VER)
394 HE **hes = _alloca (count * sizeof (HE));
395#else
396 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
397#endif
309 398
310 i = 0; 399 i = 0;
311 while ((he = hv_iternext (hv))) 400 while ((he = hv_iternext (hv)))
312 { 401 {
313 hes [i++] = he; 402 hes [i++] = he;
… …
337 LEAVE; 426 LEAVE;
338 } 427 }
339 428
340 for (i = 0; i < count; ++i) 429 for (i = 0; i < count; ++i)
341 { 430 {
342 INDENT; 431 encode_indent (enc);
343 encode_he (enc, hes [i]); 432 encode_he (enc, hes [i]);
344 433
345 if (i < count - 1) 434 if (i < count - 1)
346 COMMA; 435 encode_comma (enc);
347 }
348
349 NL; 436 }
437
438 encode_nl (enc);
350 } 439 }
351 else 440 else
352 { 441 {
353 SV *sv;
354 HE *he = hv_iternext (hv); 442 HE *he = hv_iternext (hv);
355 443
356 for (;;) 444 for (;;)
357 { 445 {
358 INDENT; 446 encode_indent (enc);
359 encode_he (enc, he); 447 encode_he (enc, he);
360 448
361 if (!(he = hv_iternext (hv))) 449 if (!(he = hv_iternext (hv)))
362 break; 450 break;
363 451
364 COMMA; 452 encode_comma (enc);
365 }
366
367 NL; 453 }
454
455 encode_nl (enc);
368 } 456 }
369 } 457 }
370 458
371 --enc->indent; INDENT; encode_ch (enc, '}'); 459 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
460}
461
462// encode objects, arrays and special \0=false and \1=true values.
463static void
464encode_rv (enc_t *enc, SV *sv)
465{
466 svtype svt;
467
468 SvGETMAGIC (sv);
469 svt = SvTYPE (sv);
470
471 if (expect_false (SvOBJECT (sv)))
472 {
473 if (SvSTASH (sv) == json_boolean_stash)
474 {
475 if (SvIV (sv) == 0)
476 encode_str (enc, "false", 5, 0);
477 else
478 encode_str (enc, "true", 4, 0);
479 }
480 else
481 {
482#if 0
483 if (0 && sv_derived_from (rv, "JSON::Literal"))
484 {
485 // not yet
486 }
487#endif
488 if (enc->json.flags & F_CONV_BLESSED)
489 {
490 // we re-bless the reference to get overload and other niceties right
491 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
492
493 if (to_json)
494 {
495 dSP;
496 ENTER;
497 SAVETMPS;
498 PUSHMARK (SP);
499 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
500
501 // calling with G_SCALAR ensures that we always get a 1 reutrn value
502 // check anyways.
503 PUTBACK;
504 assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
505 SPAGAIN;
506
507 encode_sv (enc, POPs);
508
509 FREETMPS;
510 LEAVE;
511 }
512 else if (enc->json.flags & F_ALLOW_BLESSED)
513 encode_str (enc, "null", 4, 0);
514 else
515 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
516 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
517 }
518 else if (enc->json.flags & F_ALLOW_BLESSED)
519 encode_str (enc, "null", 4, 0);
520 else
521 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
522 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
523 }
524 }
525 else if (svt == SVt_PVHV)
526 encode_hv (enc, (HV *)sv);
527 else if (svt == SVt_PVAV)
528 encode_av (enc, (AV *)sv);
529 else if (svt < SVt_PVAV)
530 {
531 if (SvNIOK (sv) && SvIV (sv) == 0)
532 encode_str (enc, "false", 5, 0);
533 else if (SvNIOK (sv) && SvIV (sv) == 1)
534 encode_str (enc, "true", 4, 0);
535 else
536 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
537 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
538 }
539 else
540 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
541 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372} 542}
373 543
374static void 544static void
375encode_sv (enc_t *enc, SV *sv) 545encode_sv (enc_t *enc, SV *sv)
376{ 546{
… …
384 encode_str (enc, str, len, SvUTF8 (sv)); 554 encode_str (enc, str, len, SvUTF8 (sv));
385 encode_ch (enc, '"'); 555 encode_ch (enc, '"');
386 } 556 }
387 else if (SvNOKp (sv)) 557 else if (SvNOKp (sv))
388 { 558 {
559 // trust that perl will do the right thing w.r.t. JSON syntax.
389 need (enc, NV_DIG + 32); 560 need (enc, NV_DIG + 32);
390 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 561 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391 enc->cur += strlen (enc->cur); 562 enc->cur += strlen (enc->cur);
392 } 563 }
393 else if (SvIOKp (sv)) 564 else if (SvIOKp (sv))
394 { 565 {
395 need (enc, 64); 566 // we assume we can always read an IV as a UV
567 if (SvUV (sv) & ~(UV)0x7fff)
568 {
569 // large integer, use the (rather slow) snprintf way.
570 need (enc, sizeof (UV) * 3);
396 enc->cur += 571 enc->cur +=
397 SvIsUV(sv) 572 SvIsUV(sv)
398 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 573 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 574 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
575 }
576 else
577 {
578 // optimise the "small number case"
579 // code will likely be branchless and use only a single multiplication
580 I32 i = SvIV (sv);
581 U32 u;
582 char digit, nz = 0;
583
584 need (enc, 6);
585
586 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
587 u = i < 0 ? -i : i;
588
589 // convert to 4.28 fixed-point representation
590 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
591
592 // now output digit by digit, each time masking out the integer part
593 // and multiplying by 5 while moving the decimal point one to the right,
594 // resulting in a net multiplication by 10.
595 // we always write the digit to memory but conditionally increment
596 // the pointer, to ease the usage of conditional move instructions.
597 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
598 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
599 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
600 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
601 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
602 }
400 } 603 }
401 else if (SvROK (sv)) 604 else if (SvROK (sv))
402 { 605 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)) 606 else if (!SvOK (sv))
419 encode_str (enc, "null", 4, 0); 607 encode_str (enc, "null", 4, 0);
420 else 608 else
421 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 609 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422 SvPV_nolen (sv), SvFLAGS (sv)); 610 SvPV_nolen (sv), SvFLAGS (sv));
423} 611}
424 612
425static SV * 613static SV *
426encode_json (SV *scalar, UV flags) 614encode_json (SV *scalar, JSON *json)
427{ 615{
616 enc_t enc;
617
428 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) 618 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 619 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430 620
431 enc_t enc; 621 enc.json = *json;
432 enc.flags = flags;
433 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 622 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434 enc.cur = SvPVX (enc.sv); 623 enc.cur = SvPVX (enc.sv);
435 enc.end = SvEND (enc.sv); 624 enc.end = SvEND (enc.sv);
436 enc.max_recurse = 0;
437 enc.indent = 0; 625 enc.indent = 0;
626 enc.maxdepth = DEC_DEPTH (enc.json.flags);
438 627
439 SvPOK_only (enc.sv); 628 SvPOK_only (enc.sv);
440 encode_sv (&enc, scalar); 629 encode_sv (&enc, scalar);
441 630
631 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
632 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
633
442 if (!(flags & (F_ASCII | F_UTF8))) 634 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443 SvUTF8_on (enc.sv); 635 SvUTF8_on (enc.sv);
444 636
445 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447 if (enc.flags & F_SHRINK) 637 if (enc.json.flags & F_SHRINK)
448 shrink (enc.sv); 638 shrink (enc.sv);
449 639
450 return enc.sv; 640 return enc.sv;
451} 641}
452 642
453///////////////////////////////////////////////////////////////////////////// 643/////////////////////////////////////////////////////////////////////////////
644// decoder
454 645
455#define WS \ 646// structure used for decoding JSON
647typedef struct
648{
649 char *cur; // current parser pointer
650 char *end; // end of input string
651 const char *err; // parse error, if != 0
652 JSON json;
653 U32 depth; // recursion depth
654 U32 maxdepth; // recursion depth limit
655} dec_t;
656
657inline void
658decode_ws (dec_t *dec)
659{
456 for (;;) \ 660 for (;;)
457 { \ 661 {
458 char ch = *dec->cur; \ 662 char ch = *dec->cur;
663
459 if (ch > 0x20 \ 664 if (ch > 0x20
460 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 665 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
461 break; \ 666 break;
667
462 ++dec->cur; \ 668 ++dec->cur;
463 } 669 }
670}
464 671
465#define ERR(reason) SB dec->err = reason; goto fail; SE 672#define ERR(reason) SB dec->err = reason; goto fail; SE
673
466#define EXPECT_CH(ch) SB \ 674#define EXPECT_CH(ch) SB \
467 if (*dec->cur != ch) \ 675 if (*dec->cur != ch) \
468 ERR (# ch " expected"); \ 676 ERR (# ch " expected"); \
469 ++dec->cur; \ 677 ++dec->cur; \
470 SE 678 SE
471 679
680#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
681#define DEC_DEC_DEPTH --dec->depth
682
472static SV *decode_sv (dec_t *dec); 683static SV *decode_sv (dec_t *dec);
473 684
474static signed char decode_hexdigit[256]; 685static signed char decode_hexdigit[256];
475 686
476static UV 687static UV
477decode_4hex (dec_t *dec) 688decode_4hex (dec_t *dec)
478{ 689{
479 signed char d1, d2, d3, d4; 690 signed char d1, d2, d3, d4;
691 unsigned char *cur = (unsigned char *)dec->cur;
480 692
481 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 693 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"); 694 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]]; 695 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"); 696 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");
489 697
490 dec->cur += 4; 698 dec->cur += 4;
491 699
492 return ((UV)d1) << 12 700 return ((UV)d1) << 12
493 | ((UV)d2) << 8 701 | ((UV)d2) << 8
… …
496 704
497fail: 705fail:
498 return (UV)-1; 706 return (UV)-1;
499} 707}
500 708
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 * 709static SV *
517decode_str (dec_t *dec) 710decode_str (dec_t *dec)
518{ 711{
519 SV *sv = NEWSV (0,2); 712 SV *sv = 0;
520 int utf8 = 0; 713 int utf8 = 0;
521 char *cur = SvPVX (sv); 714 char *dec_cur = dec->cur;
522 char *end = SvEND (sv);
523 715
524 for (;;) 716 do
525 { 717 {
526 unsigned char ch = *(unsigned char *)dec->cur; 718 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
719 char *cur = buf;
527 720
528 if (ch == '"') 721 do
529 break;
530 else if (ch == '\\')
531 { 722 {
532 switch (*++dec->cur) 723 unsigned char ch = *(unsigned char *)dec_cur++;
724
725 if (expect_false (ch == '"'))
726 {
727 --dec_cur;
728 break;
533 { 729 }
534 case '\\': 730 else if (expect_false (ch == '\\'))
535 case '/': 731 {
536 case '"': APPEND_CH (*dec->cur++); break; 732 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 { 733 {
546 UV lo, hi; 734 case '\\':
547 ++dec->cur; 735 case '/':
736 case '"': *cur++ = *dec_cur++; break;
548 737
549 hi = decode_4hex (dec); 738 case 'b': ++dec_cur; *cur++ = '\010'; break;
550 if (hi == (UV)-1) 739 case 't': ++dec_cur; *cur++ = '\011'; break;
551 goto fail; 740 case 'n': ++dec_cur; *cur++ = '\012'; break;
741 case 'f': ++dec_cur; *cur++ = '\014'; break;
742 case 'r': ++dec_cur; *cur++ = '\015'; break;
552 743
553 // possibly a surrogate pair 744 case 'u':
554 if (hi >= 0xd800 && hi < 0xdc00)
555 { 745 {
556 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 746 UV lo, hi;
557 ERR ("missing low surrogate character in surrogate pair"); 747 ++dec_cur;
558 748
559 dec->cur += 2; 749 dec->cur = dec_cur;
560
561 lo = decode_4hex (dec); 750 hi = decode_4hex (dec);
751 dec_cur = dec->cur;
562 if (lo == (UV)-1) 752 if (hi == (UV)-1)
563 goto fail; 753 goto fail;
564 754
755 // possibly a surrogate pair
756 if (hi >= 0xd800)
757 if (hi < 0xdc00)
758 {
759 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
760 ERR ("missing low surrogate character in surrogate pair");
761
762 dec_cur += 2;
763
764 dec->cur = dec_cur;
765 lo = decode_4hex (dec);
766 dec_cur = dec->cur;
767 if (lo == (UV)-1)
768 goto fail;
769
565 if (lo < 0xdc00 || lo >= 0xe000) 770 if (lo < 0xdc00 || lo >= 0xe000)
566 ERR ("surrogate pair expected"); 771 ERR ("surrogate pair expected");
567 772
568 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 773 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
774 }
775 else if (hi < 0xe000)
776 ERR ("missing high surrogate character in surrogate pair");
777
778 if (hi >= 0x80)
779 {
780 utf8 = 1;
781
782 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
783 }
784 else
785 *cur++ = hi;
569 } 786 }
570 else if (hi >= 0xdc00 && hi < 0xe000)
571 ERR ("missing high surrogate character in surrogate pair");
572
573 if (hi >= 0x80)
574 { 787 break;
575 utf8 = 1;
576 788
577 APPEND_GROW (4); // at most 4 bytes for 21 bits
578 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579 }
580 else 789 default:
581 APPEND_CH (hi); 790 --dec_cur;
791 ERR ("illegal backslash escape sequence in string");
582 } 792 }
583 break; 793 }
794 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
795 *cur++ = ch;
796 else if (ch >= 0x80)
797 {
798 STRLEN clen;
799 UV uch;
584 800
585 default:
586 --dec->cur; 801 --dec_cur;
587 ERR ("illegal backslash escape sequence in string"); 802
803 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
804 if (clen == (STRLEN)-1)
805 ERR ("malformed UTF-8 character in JSON string");
806
807 do
808 *cur++ = *dec_cur++;
809 while (--clen);
810
811 utf8 = 1;
812 }
813 else
814 {
815 --dec_cur;
816
817 if (!ch)
818 ERR ("unexpected end of string while parsing JSON string");
819 else
820 ERR ("invalid character encountered while parsing JSON string");
588 } 821 }
589 } 822 }
590 else if (ch >= 0x20 && ch <= 0x7f) 823 while (cur < buf + SHORT_STRING_LEN);
591 APPEND_CH (*dec->cur++); 824
592 else if (ch >= 0x80)
593 { 825 {
594 STRLEN clen; 826 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");
598 827
599 APPEND_GROW (clen); 828 if (sv)
600 do
601 { 829 {
602 *cur++ = *dec->cur++; 830 SvGROW (sv, SvCUR (sv) + len + 1);
831 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
832 SvCUR_set (sv, SvCUR (sv) + len);
603 } 833 }
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 834 else
611 ERR ("invalid character encountered"); 835 sv = newSVpvn (buf, len);
612 } 836 }
837 }
838 while (*dec_cur != '"');
613 839
614 ++dec->cur; 840 ++dec_cur;
615 841
616 SvCUR_set (sv, cur - SvPVX (sv)); 842 if (sv)
617 843 {
618 SvPOK_only (sv); 844 SvPOK_only (sv);
619 *SvEND (sv) = 0; 845 *SvEND (sv) = 0;
620 846
621 if (utf8) 847 if (utf8)
622 SvUTF8_on (sv); 848 SvUTF8_on (sv);
849 }
850 else
851 sv = newSVpvn ("", 0);
623 852
624 if (dec->flags & F_SHRINK) 853 dec->cur = dec_cur;
625 shrink (sv);
626
627 return sv; 854 return sv;
628 855
629fail: 856fail:
630 SvREFCNT_dec (sv); 857 dec->cur = dec_cur;
631 return 0; 858 return 0;
632} 859}
633 860
634static SV * 861static SV *
635decode_num (dec_t *dec) 862decode_num (dec_t *dec)
… …
693 is_nv = 1; 920 is_nv = 1;
694 } 921 }
695 922
696 if (!is_nv) 923 if (!is_nv)
697 { 924 {
698 UV uv; 925 // 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); 926 if (*start == '-')
700 if (numtype & IS_NUMBER_IN_UV) 927 switch (dec->cur - start)
701 if (numtype & IS_NUMBER_NEG)
702 { 928 {
703 if (uv < (UV)IV_MIN) 929 case 2: return newSViv (-( start [1] - '0' * 1));
704 return newSViv (-(IV)uv); 930 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
931 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
932 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
705 } 933 }
934 else
935 switch (dec->cur - start)
936 {
937 case 1: return newSViv ( start [0] - '0' * 1);
938 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
939 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
940 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
941 }
942
943 {
944 UV uv;
945 int numtype = grok_number (start, dec->cur - start, &uv);
946 if (numtype & IS_NUMBER_IN_UV)
947 if (numtype & IS_NUMBER_NEG)
948 {
949 if (uv < (UV)IV_MIN)
950 return newSViv (-(IV)uv);
951 }
706 else 952 else
707 return newSVuv (uv); 953 return newSVuv (uv);
954
955 // here would likely be the place for bigint support
708 } 956 }
957 }
709 958
959 // if we ever support bigint or bigfloat, this is the place for bigfloat
710 return newSVnv (Atof (start)); 960 return newSVnv (Atof (start));
711 961
712fail: 962fail:
713 return 0; 963 return 0;
714} 964}
… …
716static SV * 966static SV *
717decode_av (dec_t *dec) 967decode_av (dec_t *dec)
718{ 968{
719 AV *av = newAV (); 969 AV *av = newAV ();
720 970
721 WS; 971 DEC_INC_DEPTH;
972 decode_ws (dec);
973
722 if (*dec->cur == ']') 974 if (*dec->cur == ']')
723 ++dec->cur; 975 ++dec->cur;
724 else 976 else
725 for (;;) 977 for (;;)
726 { 978 {
… …
730 if (!value) 982 if (!value)
731 goto fail; 983 goto fail;
732 984
733 av_push (av, value); 985 av_push (av, value);
734 986
735 WS; 987 decode_ws (dec);
736 988
737 if (*dec->cur == ']') 989 if (*dec->cur == ']')
738 { 990 {
739 ++dec->cur; 991 ++dec->cur;
740 break; 992 break;
… …
744 ERR (", or ] expected while parsing array"); 996 ERR (", or ] expected while parsing array");
745 997
746 ++dec->cur; 998 ++dec->cur;
747 } 999 }
748 1000
1001 DEC_DEC_DEPTH;
749 return newRV_noinc ((SV *)av); 1002 return newRV_noinc ((SV *)av);
750 1003
751fail: 1004fail:
752 SvREFCNT_dec (av); 1005 SvREFCNT_dec (av);
1006 DEC_DEC_DEPTH;
753 return 0; 1007 return 0;
754} 1008}
755 1009
756static SV * 1010static SV *
757decode_hv (dec_t *dec) 1011decode_hv (dec_t *dec)
758{ 1012{
759 HV *hv = newHV (); 1013 HV *hv = newHV ();
760 1014
761 WS; 1015 DEC_INC_DEPTH;
1016 decode_ws (dec);
1017
762 if (*dec->cur == '}') 1018 if (*dec->cur == '}')
763 ++dec->cur; 1019 ++dec->cur;
764 else 1020 else
765 for (;;) 1021 for (;;)
766 { 1022 {
1023 decode_ws (dec); EXPECT_CH ('"');
1024
1025 // heuristic: assume that
1026 // a) decode_str + hv_store_ent are abysmally slow.
1027 // b) most hash keys are short, simple ascii text.
1028 // => try to "fast-match" such strings to avoid
1029 // the overhead of decode_str + hv_store_ent.
1030 {
767 SV *key, *value; 1031 SV *value;
1032 char *p = dec->cur;
1033 char *e = p + 24; // only try up to 24 bytes
768 1034
769 WS; EXPECT_CH ('"'); 1035 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 { 1036 {
1037 // the >= 0x80 is true on most architectures
1038 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1039 {
1040 // slow path, back up and use decode_str
1041 SV *key = decode_str (dec);
1042 if (!key)
1043 goto fail;
1044
1045 decode_ws (dec); EXPECT_CH (':');
1046
1047 value = decode_sv (dec);
1048 if (!value)
1049 {
1050 SvREFCNT_dec (key);
1051 goto fail;
1052 }
1053
1054 hv_store_ent (hv, key, value, 0);
780 SvREFCNT_dec (key); 1055 SvREFCNT_dec (key);
1056
1057 break;
1058 }
1059 else if (*p == '"')
1060 {
1061 // fast path, got a simple key
1062 char *key = dec->cur;
1063 int len = p - key;
1064 dec->cur = p + 1;
1065
1066 decode_ws (dec); EXPECT_CH (':');
1067
1068 value = decode_sv (dec);
1069 if (!value)
781 goto fail; 1070 goto fail;
1071
1072 hv_store (hv, key, len, value, 0);
1073
1074 break;
1075 }
1076
1077 ++p;
782 } 1078 }
783
784 //TODO: optimise
785 hv_store_ent (hv, key, value, 0);
786
787 WS; 1079 }
1080
1081 decode_ws (dec);
788 1082
789 if (*dec->cur == '}') 1083 if (*dec->cur == '}')
790 { 1084 {
791 ++dec->cur; 1085 ++dec->cur;
792 break; 1086 break;
… …
796 ERR (", or } expected while parsing object/hash"); 1090 ERR (", or } expected while parsing object/hash");
797 1091
798 ++dec->cur; 1092 ++dec->cur;
799 } 1093 }
800 1094
1095 DEC_DEC_DEPTH;
801 return newRV_noinc ((SV *)hv); 1096 return newRV_noinc ((SV *)hv);
802 1097
803fail: 1098fail:
804 SvREFCNT_dec (hv); 1099 SvREFCNT_dec (hv);
1100 DEC_DEC_DEPTH;
805 return 0; 1101 return 0;
806} 1102}
807 1103
808static SV * 1104static SV *
809decode_sv (dec_t *dec) 1105decode_sv (dec_t *dec)
810{ 1106{
811 WS; 1107 decode_ws (dec);
1108
1109 // the beauty of JSON: you need exactly one character lookahead
1110 // to parse anything.
812 switch (*dec->cur) 1111 switch (*dec->cur)
813 { 1112 {
814 case '"': ++dec->cur; return decode_str (dec); 1113 case '"': ++dec->cur; return decode_str (dec);
815 case '[': ++dec->cur; return decode_av (dec); 1114 case '[': ++dec->cur; return decode_av (dec);
816 case '{': ++dec->cur; return decode_hv (dec); 1115 case '{': ++dec->cur; return decode_hv (dec);
… …
822 1121
823 case 't': 1122 case 't':
824 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1123 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825 { 1124 {
826 dec->cur += 4; 1125 dec->cur += 4;
827 return newSViv (1); 1126 return SvREFCNT_inc (json_true);
828 } 1127 }
829 else 1128 else
830 ERR ("'true' expected"); 1129 ERR ("'true' expected");
831 1130
832 break; 1131 break;
833 1132
834 case 'f': 1133 case 'f':
835 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1134 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836 { 1135 {
837 dec->cur += 5; 1136 dec->cur += 5;
838 return newSViv (0); 1137 return SvREFCNT_inc (json_false);
839 } 1138 }
840 else 1139 else
841 ERR ("'false' expected"); 1140 ERR ("'false' expected");
842 1141
843 break; 1142 break;
… …
852 ERR ("'null' expected"); 1151 ERR ("'null' expected");
853 1152
854 break; 1153 break;
855 1154
856 default: 1155 default:
857 ERR ("malformed json string, neither array, object, number, string or atom"); 1156 ERR ("malformed JSON string, neither array, object, number, string or atom");
858 break; 1157 break;
859 } 1158 }
860 1159
861fail: 1160fail:
862 return 0; 1161 return 0;
863} 1162}
864 1163
865static SV * 1164static SV *
866decode_json (SV *string, UV flags) 1165decode_json (SV *string, JSON *json, UV *offset_return)
867{ 1166{
1167 dec_t dec;
1168 UV offset;
868 SV *sv; 1169 SV *sv;
869 1170
1171 SvGETMAGIC (string);
1172 SvUPGRADE (string, SVt_PV);
1173
1174 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1175 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1176 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1177
870 if (flags & F_UTF8) 1178 if (json->flags & F_UTF8)
871 sv_utf8_downgrade (string, 0); 1179 sv_utf8_downgrade (string, 0);
872 else 1180 else
873 sv_utf8_upgrade (string); 1181 sv_utf8_upgrade (string);
874 1182
875 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1183 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876 1184
877 dec_t dec; 1185 dec.json = *json;
878 dec.flags = flags;
879 dec.cur = SvPVX (string); 1186 dec.cur = SvPVX (string);
880 dec.end = SvEND (string); 1187 dec.end = SvEND (string);
881 dec.err = 0; 1188 dec.err = 0;
1189 dec.depth = 0;
1190 dec.maxdepth = DEC_DEPTH (dec.json.flags);
882 1191
1192 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
883 sv = decode_sv (&dec); 1193 sv = decode_sv (&dec);
884 1194
1195 if (!(offset_return || !sv))
1196 {
1197 // check for trailing garbage
1198 decode_ws (&dec);
1199
1200 if (*dec.cur)
1201 {
1202 dec.err = "garbage after JSON object";
1203 SvREFCNT_dec (sv);
1204 sv = 0;
1205 }
1206 }
1207
1208 if (offset_return || !sv)
1209 {
1210 offset = dec.json.flags & F_UTF8
1211 ? dec.cur - SvPVX (string)
1212 : utf8_distance (dec.cur, SvPVX (string));
1213
1214 if (offset_return)
1215 *offset_return = offset;
1216 }
1217
885 if (!sv) 1218 if (!sv)
886 { 1219 {
887 IV offset = dec.flags & F_UTF8
888 ? dec.cur - SvPVX (string)
889 : utf8_distance (dec.cur, SvPVX (string));
890 SV *uni = sv_newmortal (); 1220 SV *uni = sv_newmortal ();
891 1221
892 // horrible hack to silence warning inside pv_uni_display 1222 // horrible hack to silence warning inside pv_uni_display
893 COP cop = *PL_curcop; 1223 COP cop = *PL_curcop;
894 cop.cop_warnings = pWARN_NONE; 1224 cop.cop_warnings = pWARN_NONE;
… …
896 SAVEVPTR (PL_curcop); 1226 SAVEVPTR (PL_curcop);
897 PL_curcop = &cop; 1227 PL_curcop = &cop;
898 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1228 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899 LEAVE; 1229 LEAVE;
900 1230
901 croak ("%s, at character offset %d (%s)", 1231 croak ("%s, at character offset %d [\"%s\"]",
902 dec.err, 1232 dec.err,
903 (int)offset, 1233 (int)offset,
904 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1234 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905 } 1235 }
906 1236
907 sv = sv_2mortal (sv); 1237 sv = sv_2mortal (sv);
908 1238
909 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1239 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1240 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911 1241
912 return sv; 1242 return sv;
913} 1243}
914 1244
1245/////////////////////////////////////////////////////////////////////////////
1246// XS interface functions
1247
915MODULE = JSON::XS PACKAGE = JSON::XS 1248MODULE = JSON::XS PACKAGE = JSON::XS
916 1249
917BOOT: 1250BOOT:
918{ 1251{
919 int i; 1252 int i;
920 1253
921 memset (decode_hexdigit, 0xff, 256);
922 for (i = 10; i--; ) 1254 for (i = 0; i < 256; ++i)
923 decode_hexdigit ['0' + i] = i; 1255 decode_hexdigit [i] =
1256 i >= '0' && i <= '9' ? i - '0'
1257 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1258 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1259 : -1;
924 1260
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); 1261 json_stash = gv_stashpv ("JSON::XS" , 1);
1262 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1263
1264 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1265 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
932} 1266}
933 1267
934PROTOTYPES: DISABLE 1268PROTOTYPES: DISABLE
935 1269
936SV *new (char *dummy) 1270void new (char *klass)
937 CODE: 1271 PPCODE:
938 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1272{
939 OUTPUT: 1273 SV *pv = NEWSV (0, sizeof (JSON));
940 RETVAL 1274 SvPOK_only (pv);
1275 Zero (SvPVX (pv), 1, sizeof (JSON));
1276 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1277 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
1278}
941 1279
942SV *ascii (SV *self, int enable = 1) 1280void ascii (JSON *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 PPCODE:
954{ 1295{
955 UV *uv = SvJSON (self);
956 if (enable) 1296 if (enable)
957 *uv |= ix; 1297 self->flags |= ix;
958 else 1298 else
959 *uv &= ~ix; 1299 self->flags &= ~ix;
960 1300
961 RETVAL = newSVsv (self); 1301 XPUSHs (ST (0));
962} 1302}
963 OUTPUT:
964 RETVAL
965 1303
966void encode (SV *self, SV *scalar) 1304void max_depth (JSON *self, UV max_depth = 0x80000000UL)
967 PPCODE: 1305 PPCODE:
968 XPUSHs (encode_json (scalar, *SvJSON (self))); 1306{
1307 UV log2 = 0;
969 1308
970void decode (SV *self, SV *jsonstr) 1309 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1310
1311 while ((1UL << log2) < max_depth)
1312 ++log2;
1313
1314 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1315
1316 XPUSHs (ST (0));
1317}
1318
1319void max_size (JSON *self, UV max_size = 0)
971 PPCODE: 1320 PPCODE:
1321{
1322 UV log2 = 0;
1323
1324 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1325 if (max_size == 1) max_size = 2;
1326
1327 while ((1UL << log2) < max_size)
1328 ++log2;
1329
1330 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1331
1332 XPUSHs (ST (0));
1333}
1334
1335void encode (JSON *self, SV *scalar)
1336 PPCODE:
1337 XPUSHs (encode_json (scalar, self));
1338
1339void decode (JSON *self, SV *jsonstr)
1340 PPCODE:
972 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1341 XPUSHs (decode_json (jsonstr, self, 0));
1342
1343void decode_prefix (JSON *self, SV *jsonstr)
1344 PPCODE:
1345{
1346 UV offset;
1347 EXTEND (SP, 2);
1348 PUSHs (decode_json (jsonstr, self, &offset));
1349 PUSHs (sv_2mortal (newSVuv (offset)));
1350}
973 1351
974PROTOTYPES: ENABLE 1352PROTOTYPES: ENABLE
975 1353
976void to_json (SV *scalar) 1354void to_json (SV *scalar)
977 PPCODE: 1355 PPCODE:
1356{
1357 JSON json = { F_DEFAULT | F_UTF8 };
978 XPUSHs (encode_json (scalar, F_UTF8)); 1358 XPUSHs (encode_json (scalar, &json));
1359}
979 1360
980void from_json (SV *jsonstr) 1361void from_json (SV *jsonstr)
981 PPCODE: 1362 PPCODE:
1363{
1364 JSON json = { F_DEFAULT | F_UTF8 };
982 XPUSHs (decode_json (jsonstr, F_UTF8)); 1365 XPUSHs (decode_json (jsonstr, &json, 0));
1366}
983 1367

Diff Legend

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