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 1.3 by root, Thu Mar 22 18:10:29 2007 UTC vs.
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 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>
9#include <float.h>
8 10
11#if defined(__BORLANDC__) || defined(_MSC_VER)
12# define snprintf _snprintf // C compilers have this in stdio.h
13#endif
14
15// some old perls do not have this, try to make it work, no
16// guarentees, though. if it breaks, you get to keep the pieces.
17#ifndef UTF8_MAXBYTES
18# define UTF8_MAXBYTES 13
19#endif
20
9#define F_ASCII 0x00000001 21#define F_ASCII 0x00000001UL
22#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 23#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 24#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 25#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 26#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 27#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040
16#define F_ALLOW_NONREF 0x00000080 28#define F_ALLOW_NONREF 0x00000100UL
29#define F_SHRINK 0x00000200UL
30#define F_ALLOW_BLESSED 0x00000400UL
31#define F_CONV_BLESSED 0x00000800UL
32#define F_MAXDEPTH 0xf8000000UL
33#define S_MAXDEPTH 27
34#define F_MAXSIZE 0x01f00000UL
35#define S_MAXSIZE 20
36#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
37
38#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
39#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17 40
18#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 41#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19#define F_DEFAULT 0 42#define F_DEFAULT (9UL << S_MAXDEPTH)
20 43
21#define INIT_SIZE 32 // initial scalar size to be allocated 44#define INIT_SIZE 32 // initial scalar size to be allocated
45#define INDENT_STEP 3 // spaces per indentation level
46
47#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22 48
23#define SB do { 49#define SB do {
24#define SE } while (0) 50#define SE } while (0)
25 51
26static HV *json_stash; 52#if __GNUC__ >= 3
53# define expect(expr,value) __builtin_expect ((expr),(value))
54# define inline inline
55#else
56# define expect(expr,value) (expr)
57# define inline static
58#endif
59
60#define expect_false(expr) expect ((expr) != 0, 0)
61#define expect_true(expr) expect ((expr) != 0, 1)
62
63#ifdef USE_ITHREADS
64# define JSON_SLOW 1
65#else
66# define JSON_SLOW 0
67#endif
68
69static HV *json_stash, *json_boolean_stash; // JSON::XS::
70static SV *json_true, *json_false;
71
72typedef struct {
73 U32 flags;
74 SV *cb_object;
75 HV *cb_sk_object;
76} JSON;
77
78/////////////////////////////////////////////////////////////////////////////
79// utility functions
80
81inline void
82shrink (SV *sv)
83{
84 sv_utf8_downgrade (sv, 1);
85 if (SvLEN (sv) > SvCUR (sv) + 1)
86 {
87#ifdef SvPV_shrink_to_cur
88 SvPV_shrink_to_cur (sv);
89#elif defined (SvPV_renew)
90 SvPV_renew (sv, SvCUR (sv) + 1);
91#endif
92 }
93}
94
95// decode an utf-8 character and return it, or (UV)-1 in
96// case of an error.
97// we special-case "safe" characters from U+80 .. U+7FF,
98// but use the very good perl function to parse anything else.
99// note that we never call this function for a ascii codepoints
100inline UV
101decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
102{
103 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
104 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
105 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
106 {
107 *clen = 2;
108 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
109 }
110 else
111 {
112 *clen = (STRLEN)-1;
113 return (UV)-1;
114 }
115}
116
117/////////////////////////////////////////////////////////////////////////////
118// encoder
27 119
28// structure used for encoding JSON 120// structure used for encoding JSON
29typedef struct 121typedef struct
30{ 122{
31 char *cur; 123 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 124 char *end; // SvEND (sv)
34 SV *sv; 125 SV *sv; // result scalar
35 UV flags; 126 JSON json;
36 int max_recurse; 127 U32 indent; // indentation level
37 int indent; 128 U32 maxdepth; // max. indentation/recursion level
38} enc_t; 129} enc_t;
39 130
40// structure used for decoding JSON 131inline void
41typedef struct
42{
43 char *cur;
44 char *end;
45 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
58/////////////////////////////////////////////////////////////////////////////
59
60static void
61need (enc_t *enc, STRLEN len) 132need (enc_t *enc, STRLEN len)
62{ 133{
63 if (enc->cur + len >= enc->end) 134 if (expect_false (enc->cur + len >= enc->end))
64 { 135 {
65 STRLEN cur = enc->cur - SvPVX (enc->sv); 136 STRLEN cur = enc->cur - SvPVX (enc->sv);
66 SvGROW (enc->sv, cur + len + 1); 137 SvGROW (enc->sv, cur + len + 1);
67 enc->cur = SvPVX (enc->sv) + cur; 138 enc->cur = SvPVX (enc->sv) + cur;
68 enc->end = SvEND (enc->sv); 139 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69 } 140 }
70} 141}
71 142
72static void 143inline void
73encode_ch (enc_t *enc, char ch) 144encode_ch (enc_t *enc, char ch)
74{ 145{
75 need (enc, 1); 146 need (enc, 1);
76 *enc->cur++ = ch; 147 *enc->cur++ = ch;
77} 148}
79static void 150static void
80encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) 151encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
81{ 152{
82 char *end = str + len; 153 char *end = str + len;
83 154
155 need (enc, len);
156
84 while (str < end) 157 while (str < end)
85 { 158 {
86 unsigned char ch = *(unsigned char *)str; 159 unsigned char ch = *(unsigned char *)str;
160
87 if (ch >= 0x20 && ch < 0x80) // most common case 161 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88 { 162 {
163 if (expect_false (ch == '"')) // but with slow exceptions
164 {
165 need (enc, len += 1);
166 *enc->cur++ = '\\';
167 *enc->cur++ = '"';
168 }
169 else if (expect_false (ch == '\\'))
170 {
171 need (enc, len += 1);
172 *enc->cur++ = '\\';
173 *enc->cur++ = '\\';
174 }
175 else
89 *enc->cur++ = ch; 176 *enc->cur++ = ch;
177
90 str++; 178 ++str;
91 } 179 }
92 else 180 else
93 { 181 {
94 STRLEN clen; 182 switch (ch)
95 UV uch;
96
97 if (is_utf8)
98 {
99 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
100 if (clen < 0)
101 croak ("malformed UTF-8 character in string, cannot convert to JSON");
102 } 183 {
103 else 184 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
104 { 185 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
105 uch = ch; 186 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
106 clen = 1; 187 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
107 } 188 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
108 189
109 need (enc, len += 6); 190 default:
110
111 if (uch < 0xa0 || enc->flags & F_ASCII)
112 {
113 if (uch > 0xFFFFUL)
114 { 191 {
192 STRLEN clen;
193 UV uch;
194
195 if (is_utf8)
196 {
197 uch = decode_utf8 (str, end - str, &clen);
198 if (clen == (STRLEN)-1)
199 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
200 }
201 else
202 {
203 uch = ch;
115 len += 6; 204 clen = 1;
205 }
206
207 if (uch > 0x10FFFFUL)
208 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
209
210 if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
211 {
212 if (uch > 0xFFFFUL)
213 {
116 need (enc, len += 6); 214 need (enc, len += 11);
117 sprintf (enc->cur, "\\u%04x\\u%04x", 215 sprintf (enc->cur, "\\u%04x\\u%04x",
118 (uch - 0x10000) / 0x400 + 0xD800, 216 (int)((uch - 0x10000) / 0x400 + 0xD800),
119 (uch - 0x10000) % 0x400 + 0xDC00); 217 (int)((uch - 0x10000) % 0x400 + 0xDC00));
120 enc->cur += 12; 218 enc->cur += 12;
219 }
220 else
221 {
222 static char hexdigit [16] = "0123456789abcdef";
223 need (enc, len += 5);
224 *enc->cur++ = '\\';
225 *enc->cur++ = 'u';
226 *enc->cur++ = hexdigit [ uch >> 12 ];
227 *enc->cur++ = hexdigit [(uch >> 8) & 15];
228 *enc->cur++ = hexdigit [(uch >> 4) & 15];
229 *enc->cur++ = hexdigit [(uch >> 0) & 15];
230 }
231
232 str += clen;
233 }
234 else if (enc->json.flags & F_LATIN1)
235 {
236 *enc->cur++ = uch;
237 str += clen;
238 }
239 else if (is_utf8)
240 {
241 need (enc, len += clen);
242 do
243 {
244 *enc->cur++ = *str++;
245 }
246 while (--clen);
247 }
248 else
249 {
250 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
251 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
252 ++str;
253 }
121 } 254 }
122 else
123 {
124 sprintf (enc->cur, "\\u%04x", uch);
125 enc->cur += 6;
126 }
127 }
128 else if (is_utf8)
129 { 255 }
130 memcpy (enc->cur, str, clen);
131 enc->cur += clen;
132 }
133 else
134 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
135
136 str += clen;
137 } 256 }
138 257
139 --len; 258 --len;
140 } 259 }
141} 260}
142 261
143#define INDENT SB \ 262inline void
263encode_indent (enc_t *enc)
264{
144 if (enc->flags & F_INDENT) \ 265 if (enc->json.flags & F_INDENT)
145 { \ 266 {
146 int i_; \ 267 int spaces = enc->indent * INDENT_STEP;
147 need (enc, enc->indent); \ 268
148 for (i_ = enc->indent * 3; i_--; )\ 269 need (enc, spaces);
270 memset (enc->cur, ' ', spaces);
271 enc->cur += spaces;
272 }
273}
274
275inline void
276encode_space (enc_t *enc)
277{
278 need (enc, 1);
279 encode_ch (enc, ' ');
280}
281
282inline void
283encode_nl (enc_t *enc)
284{
285 if (enc->json.flags & F_INDENT)
286 {
287 need (enc, 1);
149 encode_ch (enc, ' '); \ 288 encode_ch (enc, '\n');
150 } \ 289 }
151 SE 290}
152 291
153#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 292inline void
154#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 293encode_comma (enc_t *enc)
155#define COMMA SB \ 294{
156 encode_ch (enc, ','); \ 295 encode_ch (enc, ',');
296
157 if (enc->flags & F_INDENT) \ 297 if (enc->json.flags & F_INDENT)
158 NL; \ 298 encode_nl (enc);
159 else if (enc->flags & F_SPACE_AFTER) \ 299 else if (enc->json.flags & F_SPACE_AFTER)
160 SPACE; \ 300 encode_space (enc);
161 SE 301}
162 302
163static void encode_sv (enc_t *enc, SV *sv); 303static void encode_sv (enc_t *enc, SV *sv);
164 304
165static void 305static void
166encode_av (enc_t *enc, AV *av) 306encode_av (enc_t *enc, AV *av)
167{ 307{
168 int i, len = av_len (av); 308 int i, len = av_len (av);
169 309
170 encode_ch (enc, '['); NL; 310 if (enc->indent >= enc->maxdepth)
311 croak ("data structure too deep (hit recursion limit)");
312
313 encode_ch (enc, '['); encode_nl (enc);
171 ++enc->indent; 314 ++enc->indent;
172 315
173 for (i = 0; i <= len; ++i) 316 for (i = 0; i <= len; ++i)
174 { 317 {
175 INDENT; 318 SV **svp = av_fetch (av, i, 0);
176 encode_sv (enc, *av_fetch (av, i, 0)); 319
320 encode_indent (enc);
321
322 if (svp)
323 encode_sv (enc, *svp);
324 else
325 encode_str (enc, "null", 4, 0);
177 326
178 if (i < len) 327 if (i < len)
179 COMMA; 328 encode_comma (enc);
180 } 329 }
181 330
182 NL; 331 encode_nl (enc);
183 332
184 --enc->indent; 333 --enc->indent;
185 INDENT; encode_ch (enc, ']'); 334 encode_indent (enc); encode_ch (enc, ']');
186} 335}
187 336
188static void 337static void
189encode_he (enc_t *enc, HE *he) 338encode_he (enc_t *enc, HE *he)
190{ 339{
192 341
193 if (HeKLEN (he) == HEf_SVKEY) 342 if (HeKLEN (he) == HEf_SVKEY)
194 { 343 {
195 SV *sv = HeSVKEY (he); 344 SV *sv = HeSVKEY (he);
196 STRLEN len; 345 STRLEN len;
346 char *str;
347
348 SvGETMAGIC (sv);
197 char *str = SvPV (sv, len); 349 str = SvPV (sv, len);
198 350
199 encode_str (enc, str, len, SvUTF8 (sv)); 351 encode_str (enc, str, len, SvUTF8 (sv));
200 } 352 }
201 else 353 else
202 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 354 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203 355
204 encode_ch (enc, '"'); 356 encode_ch (enc, '"');
205 357
206 if (enc->flags & F_SPACE_BEFORE) SPACE; 358 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
207 encode_ch (enc, ':'); 359 encode_ch (enc, ':');
208 if (enc->flags & F_SPACE_AFTER ) SPACE; 360 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
209 encode_sv (enc, HeVAL (he)); 361 encode_sv (enc, HeVAL (he));
210} 362}
211 363
212// compare hash entries, used when all keys are bytestrings 364// compare hash entries, used when all keys are bytestrings
213static int 365static int
219 HE *b = *(HE **)b_; 371 HE *b = *(HE **)b_;
220 372
221 STRLEN la = HeKLEN (a); 373 STRLEN la = HeKLEN (a);
222 STRLEN lb = HeKLEN (b); 374 STRLEN lb = HeKLEN (b);
223 375
224 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 376 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
225 cmp = la < lb ? -1 : la == lb ? 0 : 1; 377 cmp = la - lb;
226 378
227 return cmp; 379 return cmp;
228} 380}
229 381
230// compare hash entries, used when some keys are sv's or utf-x 382// compare hash entries, used when some keys are sv's or utf-x
237static void 389static void
238encode_hv (enc_t *enc, HV *hv) 390encode_hv (enc_t *enc, HV *hv)
239{ 391{
240 int count, i; 392 int count, i;
241 393
394 if (enc->indent >= enc->maxdepth)
395 croak ("data structure too deep (hit recursion limit)");
396
242 encode_ch (enc, '{'); NL; ++enc->indent; 397 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
243 398
244 if ((count = hv_iterinit (hv))) 399 if ((count = hv_iterinit (hv)))
245 { 400 {
246 // for canonical output we have to sort by keys first 401 // for canonical output we have to sort by keys first
247 // actually, this is mostly due to the stupid so-called 402 // actually, this is mostly due to the stupid so-called
248 // security workaround added somewhere in 5.8.x. 403 // security workaround added somewhere in 5.8.x.
249 // that randomises hash orderings 404 // that randomises hash orderings
250 if (enc->flags & F_CANONICAL) 405 if (enc->json.flags & F_CANONICAL)
251 { 406 {
252 HE *he, *hes [count];
253 int fast = 1; 407 int fast = 1;
408 HE *he;
409#if defined(__BORLANDC__) || defined(_MSC_VER)
410 HE **hes = _alloca (count * sizeof (HE));
411#else
412 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
413#endif
254 414
255 i = 0; 415 i = 0;
256 while ((he = hv_iternext (hv))) 416 while ((he = hv_iternext (hv)))
257 { 417 {
258 hes [i++] = he; 418 hes [i++] = he;
264 424
265 if (fast) 425 if (fast)
266 qsort (hes, count, sizeof (HE *), he_cmp_fast); 426 qsort (hes, count, sizeof (HE *), he_cmp_fast);
267 else 427 else
268 { 428 {
269 // hack to disable "use bytes" 429 // hack to forcefully disable "use bytes"
270 COP *oldcop = PL_curcop, cop; 430 COP cop = *PL_curcop;
271 cop.op_private = 0; 431 cop.op_private = 0;
432
433 ENTER;
434 SAVETMPS;
435
436 SAVEVPTR (PL_curcop);
272 PL_curcop = &cop; 437 PL_curcop = &cop;
273 438
274 SAVETMPS;
275 qsort (hes, count, sizeof (HE *), he_cmp_slow); 439 qsort (hes, count, sizeof (HE *), he_cmp_slow);
440
276 FREETMPS; 441 FREETMPS;
277 442 LEAVE;
278 PL_curcop = oldcop;
279 } 443 }
280 444
281 for (i = 0; i < count; ++i) 445 for (i = 0; i < count; ++i)
282 { 446 {
283 INDENT; 447 encode_indent (enc);
284 encode_he (enc, hes [i]); 448 encode_he (enc, hes [i]);
285 449
286 if (i < count - 1) 450 if (i < count - 1)
287 COMMA; 451 encode_comma (enc);
288 }
289
290 NL; 452 }
453
454 encode_nl (enc);
291 } 455 }
292 else 456 else
293 { 457 {
294 SV *sv;
295 HE *he = hv_iternext (hv); 458 HE *he = hv_iternext (hv);
296 459
297 for (;;) 460 for (;;)
298 { 461 {
299 INDENT; 462 encode_indent (enc);
300 encode_he (enc, he); 463 encode_he (enc, he);
301 464
302 if (!(he = hv_iternext (hv))) 465 if (!(he = hv_iternext (hv)))
303 break; 466 break;
304 467
305 COMMA; 468 encode_comma (enc);
306 }
307
308 NL; 469 }
470
471 encode_nl (enc);
309 } 472 }
310 } 473 }
311 474
312 --enc->indent; INDENT; encode_ch (enc, '}'); 475 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
476}
477
478// encode objects, arrays and special \0=false and \1=true values.
479static void
480encode_rv (enc_t *enc, SV *sv)
481{
482 svtype svt;
483
484 SvGETMAGIC (sv);
485 svt = SvTYPE (sv);
486
487 if (expect_false (SvOBJECT (sv)))
488 {
489 HV *stash = !JSON_SLOW || json_boolean_stash
490 ? json_boolean_stash
491 : gv_stashpv ("JSON::XS::Boolean", 1);
492
493 if (SvSTASH (sv) == stash)
494 {
495 if (SvIV (sv))
496 encode_str (enc, "true", 4, 0);
497 else
498 encode_str (enc, "false", 5, 0);
499 }
500 else
501 {
502#if 0
503 if (0 && sv_derived_from (rv, "JSON::Literal"))
504 {
505 // not yet
506 }
507#endif
508 if (enc->json.flags & F_CONV_BLESSED)
509 {
510 // we re-bless the reference to get overload and other niceties right
511 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
512
513 if (to_json)
514 {
515 dSP;
516
517 ENTER; SAVETMPS; PUSHMARK (SP);
518 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
519
520 // calling with G_SCALAR ensures that we always get a 1 return value
521 PUTBACK;
522 call_sv ((SV *)GvCV (to_json), G_SCALAR);
523 SPAGAIN;
524
525 // catch this surprisingly common error
526 if (SvROK (TOPs) && SvRV (TOPs) == sv)
527 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
528
529 sv = POPs;
530 PUTBACK;
531
532 encode_sv (enc, sv);
533
534 FREETMPS; LEAVE;
535 }
536 else if (enc->json.flags & F_ALLOW_BLESSED)
537 encode_str (enc, "null", 4, 0);
538 else
539 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
540 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
541 }
542 else if (enc->json.flags & F_ALLOW_BLESSED)
543 encode_str (enc, "null", 4, 0);
544 else
545 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
546 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
547 }
548 }
549 else if (svt == SVt_PVHV)
550 encode_hv (enc, (HV *)sv);
551 else if (svt == SVt_PVAV)
552 encode_av (enc, (AV *)sv);
553 else if (svt < SVt_PVAV)
554 {
555 STRLEN len = 0;
556 char *pv = svt ? SvPV (sv, len) : 0;
557
558 if (len == 1 && *pv == '1')
559 encode_str (enc, "true", 4, 0);
560 else if (len == 1 && *pv == '0')
561 encode_str (enc, "false", 5, 0);
562 else
563 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
564 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
565 }
566 else
567 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
568 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313} 569}
314 570
315static void 571static void
316encode_sv (enc_t *enc, SV *sv) 572encode_sv (enc_t *enc, SV *sv)
317{ 573{
574 SvGETMAGIC (sv);
575
318 if (SvPOKp (sv)) 576 if (SvPOKp (sv))
319 { 577 {
320 STRLEN len; 578 STRLEN len;
321 char *str = SvPV (sv, len); 579 char *str = SvPV (sv, len);
322 encode_ch (enc, '"'); 580 encode_ch (enc, '"');
323 encode_str (enc, str, len, SvUTF8 (sv)); 581 encode_str (enc, str, len, SvUTF8 (sv));
324 encode_ch (enc, '"'); 582 encode_ch (enc, '"');
325 } 583 }
326 else if (SvNOKp (sv)) 584 else if (SvNOKp (sv))
327 { 585 {
586 // trust that perl will do the right thing w.r.t. JSON syntax.
328 need (enc, NV_DIG + 32); 587 need (enc, NV_DIG + 32);
329 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 588 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330 enc->cur += strlen (enc->cur); 589 enc->cur += strlen (enc->cur);
331 } 590 }
332 else if (SvIOKp (sv)) 591 else if (SvIOKp (sv))
333 { 592 {
334 need (enc, 64); 593 // we assume we can always read an IV as a UV
594 if (SvUV (sv) & ~(UV)0x7fff)
595 {
596 // large integer, use the (rather slow) snprintf way.
597 need (enc, sizeof (UV) * 3);
335 enc->cur += 598 enc->cur +=
336 SvIsUV(sv) 599 SvIsUV(sv)
337 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 600 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
338 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 601 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
602 }
603 else
604 {
605 // optimise the "small number case"
606 // code will likely be branchless and use only a single multiplication
607 I32 i = SvIV (sv);
608 U32 u;
609 char digit, nz = 0;
610
611 need (enc, 6);
612
613 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
614 u = i < 0 ? -i : i;
615
616 // convert to 4.28 fixed-point representation
617 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
618
619 // now output digit by digit, each time masking out the integer part
620 // and multiplying by 5 while moving the decimal point one to the right,
621 // resulting in a net multiplication by 10.
622 // we always write the digit to memory but conditionally increment
623 // the pointer, to ease the usage of conditional move instructions.
624 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
625 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
626 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
627 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
628 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
629 }
339 } 630 }
340 else if (SvROK (sv)) 631 else if (SvROK (sv))
341 { 632 encode_rv (enc, SvRV (sv));
342 if (!--enc->max_recurse)
343 croak ("data structure too deep (hit recursion limit)");
344
345 sv = SvRV (sv);
346
347 switch (SvTYPE (sv))
348 {
349 case SVt_PVAV: encode_av (enc, (AV *)sv); break;
350 case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
351
352 default:
353 croak ("JSON can only represent references to arrays or hashes");
354 }
355 }
356 else if (!SvOK (sv)) 633 else if (!SvOK (sv))
357 encode_str (enc, "null", 4, 0); 634 encode_str (enc, "null", 4, 0);
358 else 635 else
359 croak ("encountered perl type that JSON cannot handle"); 636 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
637 SvPV_nolen (sv), SvFLAGS (sv));
360} 638}
361 639
362static SV * 640static SV *
363encode_json (SV *scalar, UV flags) 641encode_json (SV *scalar, JSON *json)
364{ 642{
365 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
366 croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
367
368 enc_t enc; 643 enc_t enc;
369 enc.flags = flags; 644
645 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
646 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
647
648 enc.json = *json;
370 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 649 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371 enc.cur = SvPVX (enc.sv); 650 enc.cur = SvPVX (enc.sv);
372 enc.end = SvEND (enc.sv); 651 enc.end = SvEND (enc.sv);
373 enc.max_recurse = 0;
374 enc.indent = 0; 652 enc.indent = 0;
653 enc.maxdepth = DEC_DEPTH (enc.json.flags);
375 654
376 SvPOK_only (enc.sv); 655 SvPOK_only (enc.sv);
377 encode_sv (&enc, scalar); 656 encode_sv (&enc, scalar);
378 657
658 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
659 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
660
379 if (!(flags & (F_ASCII | F_UTF8))) 661 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380 SvUTF8_on (enc.sv); 662 SvUTF8_on (enc.sv);
381 663
382 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 664 if (enc.json.flags & F_SHRINK)
665 shrink (enc.sv);
666
383 return enc.sv; 667 return enc.sv;
384} 668}
385 669
386///////////////////////////////////////////////////////////////////////////// 670/////////////////////////////////////////////////////////////////////////////
671// decoder
387 672
388#define WS \ 673// structure used for decoding JSON
674typedef struct
675{
676 char *cur; // current parser pointer
677 char *end; // end of input string
678 const char *err; // parse error, if != 0
679 JSON json;
680 U32 depth; // recursion depth
681 U32 maxdepth; // recursion depth limit
682} dec_t;
683
684inline void
685decode_ws (dec_t *dec)
686{
389 for (;;) \ 687 for (;;)
390 { \ 688 {
391 char ch = *dec->cur; \ 689 char ch = *dec->cur;
690
392 if (ch > 0x20 \ 691 if (ch > 0x20
393 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 692 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
394 break; \ 693 break;
694
395 ++dec->cur; \ 695 ++dec->cur;
396 } 696 }
697}
397 698
398#define ERR(reason) SB dec->err = reason; goto fail; SE 699#define ERR(reason) SB dec->err = reason; goto fail; SE
700
399#define EXPECT_CH(ch) SB \ 701#define EXPECT_CH(ch) SB \
400 if (*dec->cur != ch) \ 702 if (*dec->cur != ch) \
401 ERR (# ch " expected"); \ 703 ERR (# ch " expected"); \
402 ++dec->cur; \ 704 ++dec->cur; \
403 SE 705 SE
404 706
707#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
708#define DEC_DEC_DEPTH --dec->depth
709
405static SV *decode_sv (dec_t *dec); 710static SV *decode_sv (dec_t *dec);
406
407#define APPEND_CH(ch) SB \
408 SvGROW (sv, cur + 1 + 1); \
409 SvPVX (sv)[cur++] = (ch); \
410 SE
411 711
412static signed char decode_hexdigit[256]; 712static signed char decode_hexdigit[256];
413 713
414static UV 714static UV
415decode_4hex (dec_t *dec) 715decode_4hex (dec_t *dec)
416{ 716{
417 signed char d1, d2, d3, d4; 717 signed char d1, d2, d3, d4;
718 unsigned char *cur = (unsigned char *)dec->cur;
418 719
419 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 720 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
420 if (d1 < 0) ERR ("four hexadecimal digits expected"); 721 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
421 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 722 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
422 if (d2 < 0) ERR ("four hexadecimal digits expected"); 723 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
423 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
424 if (d3 < 0) ERR ("four hexadecimal digits expected");
425 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
426 if (d4 < 0) ERR ("four hexadecimal digits expected");
427 724
428 dec->cur += 4; 725 dec->cur += 4;
429 726
430 return ((UV)d1) << 12 727 return ((UV)d1) << 12
431 | ((UV)d2) << 8 728 | ((UV)d2) << 8
437} 734}
438 735
439static SV * 736static SV *
440decode_str (dec_t *dec) 737decode_str (dec_t *dec)
441{ 738{
442 SV *sv = NEWSV (0,2); 739 SV *sv = 0;
443 STRLEN cur = 0;
444 int utf8 = 0; 740 int utf8 = 0;
741 char *dec_cur = dec->cur;
445 742
446 for (;;) 743 do
447 { 744 {
448 unsigned char ch = *(unsigned char *)dec->cur; 745 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
746 char *cur = buf;
449 747
450 if (ch == '"') 748 do
451 break;
452 else if (ch == '\\')
453 { 749 {
454 switch (*++dec->cur) 750 unsigned char ch = *(unsigned char *)dec_cur++;
751
752 if (expect_false (ch == '"'))
753 {
754 --dec_cur;
755 break;
455 { 756 }
456 case '\\': 757 else if (expect_false (ch == '\\'))
457 case '/': 758 {
458 case '"': APPEND_CH (*dec->cur++); break; 759 switch (*dec_cur)
459
460 case 'b': APPEND_CH ('\010'); ++dec->cur; break;
461 case 't': APPEND_CH ('\011'); ++dec->cur; break;
462 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
463 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
464 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
465
466 case 'u':
467 { 760 {
468 UV lo, hi; 761 case '\\':
469 ++dec->cur; 762 case '/':
763 case '"': *cur++ = *dec_cur++; break;
470 764
471 hi = decode_4hex (dec); 765 case 'b': ++dec_cur; *cur++ = '\010'; break;
472 if (hi == (UV)-1) 766 case 't': ++dec_cur; *cur++ = '\011'; break;
473 goto fail; 767 case 'n': ++dec_cur; *cur++ = '\012'; break;
768 case 'f': ++dec_cur; *cur++ = '\014'; break;
769 case 'r': ++dec_cur; *cur++ = '\015'; break;
474 770
475 // possibly a surrogate pair 771 case 'u':
476 if (hi >= 0xd800 && hi < 0xdc00)
477 { 772 {
478 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 773 UV lo, hi;
479 ERR ("illegal surrogate character"); 774 ++dec_cur;
480 775
481 dec->cur += 2; 776 dec->cur = dec_cur;
482
483 lo = decode_4hex (dec); 777 hi = decode_4hex (dec);
778 dec_cur = dec->cur;
484 if (lo == (UV)-1) 779 if (hi == (UV)-1)
485 goto fail; 780 goto fail;
486 781
782 // possibly a surrogate pair
783 if (hi >= 0xd800)
784 if (hi < 0xdc00)
785 {
786 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
787 ERR ("missing low surrogate character in surrogate pair");
788
789 dec_cur += 2;
790
791 dec->cur = dec_cur;
792 lo = decode_4hex (dec);
793 dec_cur = dec->cur;
794 if (lo == (UV)-1)
795 goto fail;
796
487 if (lo < 0xdc00 || lo >= 0xe000) 797 if (lo < 0xdc00 || lo >= 0xe000)
488 ERR ("surrogate pair expected"); 798 ERR ("surrogate pair expected");
489 799
490 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 800 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
801 }
802 else if (hi < 0xe000)
803 ERR ("missing high surrogate character in surrogate pair");
804
805 if (hi >= 0x80)
806 {
807 utf8 = 1;
808
809 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
810 }
811 else
812 *cur++ = hi;
491 } 813 }
492 else if (lo >= 0xdc00 && lo < 0xe000)
493 ERR ("illegal surrogate character");
494
495 if (hi >= 0x80)
496 { 814 break;
497 utf8 = 1;
498 815
499 SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
500 cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
501 }
502 else 816 default:
503 APPEND_CH (hi); 817 --dec_cur;
818 ERR ("illegal backslash escape sequence in string");
504 } 819 }
820 }
821 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
822 *cur++ = ch;
823 else if (ch >= 0x80)
824 {
825 STRLEN clen;
826 UV uch;
827
828 --dec_cur;
829
830 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
831 if (clen == (STRLEN)-1)
832 ERR ("malformed UTF-8 character in JSON string");
833
834 do
835 *cur++ = *dec_cur++;
836 while (--clen);
837
838 utf8 = 1;
839 }
840 else
841 {
842 --dec_cur;
843
844 if (!ch)
845 ERR ("unexpected end of string while parsing JSON string");
505 break; 846 else
847 ERR ("invalid character encountered while parsing JSON string");
506 } 848 }
507 } 849 }
508 else if (ch >= 0x20 && ch <= 0x7f) 850 while (cur < buf + SHORT_STRING_LEN);
509 APPEND_CH (*dec->cur++); 851
510 else if (ch >= 0x80) 852 {
853 STRLEN len = cur - buf;
854
855 if (sv)
511 { 856 {
512 STRLEN clen; 857 SvGROW (sv, SvCUR (sv) + len + 1);
513 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 858 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
514 if (clen < 0) 859 SvCUR_set (sv, SvCUR (sv) + len);
515 ERR ("malformed UTF-8 character in string, cannot convert to JSON");
516
517 SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
518 memcpy (SvPVX (sv) + cur, dec->cur, clen);
519 dec->cur += clen;
520 } 860 }
521 else 861 else
522 ERR ("invalid character encountered"); 862 sv = newSVpvn (buf, len);
523 } 863 }
864 }
865 while (*dec_cur != '"');
524 866
525 ++dec->cur; 867 ++dec_cur;
526 868
869 if (sv)
870 {
527 SvPOK_only (sv); 871 SvPOK_only (sv);
528
529 SvCUR_set (sv, cur);
530 *SvEND (sv) = 0; 872 *SvEND (sv) = 0;
531 873
532 if (utf8) 874 if (utf8)
533 SvUTF8_on (sv); 875 SvUTF8_on (sv);
876 }
877 else
878 sv = newSVpvn ("", 0);
534 879
880 dec->cur = dec_cur;
535 return sv; 881 return sv;
536 882
537fail: 883fail:
538 SvREFCNT_dec (sv); 884 dec->cur = dec_cur;
539 return 0; 885 return 0;
540} 886}
541 887
542static SV * 888static SV *
543decode_num (dec_t *dec) 889decode_num (dec_t *dec)
553 { 899 {
554 ++dec->cur; 900 ++dec->cur;
555 if (*dec->cur >= '0' && *dec->cur <= '9') 901 if (*dec->cur >= '0' && *dec->cur <= '9')
556 ERR ("malformed number (leading zero must not be followed by another digit)"); 902 ERR ("malformed number (leading zero must not be followed by another digit)");
557 } 903 }
558 904 else if (*dec->cur < '0' || *dec->cur > '9')
559 // int 905 ERR ("malformed number (no digits after initial minus)");
906 else
907 do
908 {
909 ++dec->cur;
910 }
560 while (*dec->cur >= '0' && *dec->cur <= '9') 911 while (*dec->cur >= '0' && *dec->cur <= '9');
561 ++dec->cur;
562 912
563 // [frac] 913 // [frac]
564 if (*dec->cur == '.') 914 if (*dec->cur == '.')
565 { 915 {
566 is_nv = 1; 916 ++dec->cur;
917
918 if (*dec->cur < '0' || *dec->cur > '9')
919 ERR ("malformed number (no digits after decimal point)");
567 920
568 do 921 do
569 { 922 {
570 ++dec->cur; 923 ++dec->cur;
571 } 924 }
572 while (*dec->cur >= '0' && *dec->cur <= '9'); 925 while (*dec->cur >= '0' && *dec->cur <= '9');
926
927 is_nv = 1;
573 } 928 }
574 929
575 // [exp] 930 // [exp]
576 if (*dec->cur == 'e' || *dec->cur == 'E') 931 if (*dec->cur == 'e' || *dec->cur == 'E')
577 { 932 {
578 is_nv = 1;
579
580 ++dec->cur; 933 ++dec->cur;
934
581 if (*dec->cur == '-' || *dec->cur == '+') 935 if (*dec->cur == '-' || *dec->cur == '+')
582 ++dec->cur; 936 ++dec->cur;
583 937
938 if (*dec->cur < '0' || *dec->cur > '9')
939 ERR ("malformed number (no digits after exp sign)");
940
941 do
942 {
943 ++dec->cur;
944 }
584 while (*dec->cur >= '0' && *dec->cur <= '9') 945 while (*dec->cur >= '0' && *dec->cur <= '9');
585 ++dec->cur; 946
947 is_nv = 1;
586 } 948 }
587 949
588 if (!is_nv) 950 if (!is_nv)
589 { 951 {
590 UV uv; 952 int len = dec->cur - start;
591 int numtype = grok_number (start, dec->cur - start, &uv); 953
592 if (numtype & IS_NUMBER_IN_UV) 954 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
593 if (numtype & IS_NUMBER_NEG) 955 if (*start == '-')
956 switch (len)
594 { 957 {
595 if (uv < (UV)IV_MIN) 958 case 2: return newSViv (-( start [1] - '0' * 1));
596 return newSViv (-(IV)uv); 959 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
960 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
961 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
597 } 962 }
963 else
964 switch (len)
965 {
966 case 1: return newSViv ( start [0] - '0' * 1);
967 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
968 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
969 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
970 }
971
972 {
973 UV uv;
974 int numtype = grok_number (start, len, &uv);
975 if (numtype & IS_NUMBER_IN_UV)
976 if (numtype & IS_NUMBER_NEG)
977 {
978 if (uv < (UV)IV_MIN)
979 return newSViv (-(IV)uv);
980 }
598 else 981 else
599 return newSVuv (uv); 982 return newSVuv (uv);
600 } 983 }
601 984
985 len -= *start == '-' ? 1 : 0;
986
987 // does not fit into IV or UV, try NV
988 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
989 #if defined (LDBL_DIG)
990 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
991 #endif
992 )
993 // fits into NV without loss of precision
994 return newSVnv (Atof (start));
995
996 // everything else fails, convert it to a string
997 return newSVpvn (start, dec->cur - start);
998 }
999
1000 // loss of precision here
602 return newSVnv (Atof (start)); 1001 return newSVnv (Atof (start));
603 1002
604fail: 1003fail:
605 return 0; 1004 return 0;
606} 1005}
608static SV * 1007static SV *
609decode_av (dec_t *dec) 1008decode_av (dec_t *dec)
610{ 1009{
611 AV *av = newAV (); 1010 AV *av = newAV ();
612 1011
1012 DEC_INC_DEPTH;
1013 decode_ws (dec);
1014
1015 if (*dec->cur == ']')
1016 ++dec->cur;
1017 else
613 for (;;) 1018 for (;;)
614 { 1019 {
615 SV *value; 1020 SV *value;
616 1021
617 value = decode_sv (dec); 1022 value = decode_sv (dec);
618 if (!value) 1023 if (!value)
619 goto fail; 1024 goto fail;
620 1025
621 av_push (av, value); 1026 av_push (av, value);
622 1027
623 WS; 1028 decode_ws (dec);
624 1029
625 if (*dec->cur == ']') 1030 if (*dec->cur == ']')
626 { 1031 {
627 ++dec->cur; 1032 ++dec->cur;
628 break; 1033 break;
1034 }
629 } 1035
630
631 if (*dec->cur != ',') 1036 if (*dec->cur != ',')
632 ERR (", or ] expected while parsing array"); 1037 ERR (", or ] expected while parsing array");
633 1038
634 ++dec->cur; 1039 ++dec->cur;
635 } 1040 }
636 1041
1042 DEC_DEC_DEPTH;
637 return newRV_noinc ((SV *)av); 1043 return newRV_noinc ((SV *)av);
638 1044
639fail: 1045fail:
640 SvREFCNT_dec (av); 1046 SvREFCNT_dec (av);
1047 DEC_DEC_DEPTH;
641 return 0; 1048 return 0;
642} 1049}
643 1050
644static SV * 1051static SV *
645decode_hv (dec_t *dec) 1052decode_hv (dec_t *dec)
646{ 1053{
1054 SV *sv;
647 HV *hv = newHV (); 1055 HV *hv = newHV ();
648 1056
1057 DEC_INC_DEPTH;
1058 decode_ws (dec);
1059
1060 if (*dec->cur == '}')
1061 ++dec->cur;
1062 else
649 for (;;) 1063 for (;;)
650 { 1064 {
651 SV *key, *value; 1065 decode_ws (dec); EXPECT_CH ('"');
652 1066
653 WS; EXPECT_CH ('"'); 1067 // heuristic: assume that
654 1068 // a) decode_str + hv_store_ent are abysmally slow.
655 key = decode_str (dec); 1069 // b) most hash keys are short, simple ascii text.
656 if (!key) 1070 // => try to "fast-match" such strings to avoid
657 goto fail; 1071 // the overhead of decode_str + hv_store_ent.
658
659 WS; EXPECT_CH (':');
660
661 value = decode_sv (dec);
662 if (!value)
663 { 1072 {
1073 SV *value;
1074 char *p = dec->cur;
1075 char *e = p + 24; // only try up to 24 bytes
1076
1077 for (;;)
1078 {
1079 // the >= 0x80 is true on most architectures
1080 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1081 {
1082 // slow path, back up and use decode_str
1083 SV *key = decode_str (dec);
1084 if (!key)
1085 goto fail;
1086
1087 decode_ws (dec); EXPECT_CH (':');
1088
1089 value = decode_sv (dec);
1090 if (!value)
1091 {
1092 SvREFCNT_dec (key);
1093 goto fail;
1094 }
1095
1096 hv_store_ent (hv, key, value, 0);
664 SvREFCNT_dec (key); 1097 SvREFCNT_dec (key);
1098
1099 break;
1100 }
1101 else if (*p == '"')
1102 {
1103 // fast path, got a simple key
1104 char *key = dec->cur;
1105 int len = p - key;
1106 dec->cur = p + 1;
1107
1108 decode_ws (dec); EXPECT_CH (':');
1109
1110 value = decode_sv (dec);
1111 if (!value)
665 goto fail; 1112 goto fail;
1113
1114 hv_store (hv, key, len, value, 0);
1115
1116 break;
1117 }
1118
1119 ++p;
1120 }
666 } 1121 }
667 1122
668 //TODO: optimise 1123 decode_ws (dec);
669 hv_store_ent (hv, key, value, 0);
670 1124
671 WS;
672
673 if (*dec->cur == '}') 1125 if (*dec->cur == '}')
1126 {
1127 ++dec->cur;
1128 break;
1129 }
1130
1131 if (*dec->cur != ',')
1132 ERR (", or } expected while parsing object/hash");
1133
1134 ++dec->cur;
1135 }
1136
1137 DEC_DEC_DEPTH;
1138 sv = newRV_noinc ((SV *)hv);
1139
1140 // check filter callbacks
1141 if (dec->json.flags & F_HOOK)
1142 {
1143 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
674 { 1144 {
675 ++dec->cur; 1145 HE *cb, *he;
676 break; 1146
1147 hv_iterinit (hv);
1148 he = hv_iternext (hv);
1149 hv_iterinit (hv);
1150
1151 // the next line creates a mortal sv each time its called.
1152 // might want to optimise this for common cases.
1153 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1154
1155 if (cb)
1156 {
1157 dSP;
1158 int count;
1159
1160 ENTER; SAVETMPS; PUSHMARK (SP);
1161 XPUSHs (HeVAL (he));
1162
1163 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1164
1165 if (count == 1)
1166 {
1167 sv = newSVsv (POPs);
1168 FREETMPS; LEAVE;
1169 return sv;
1170 }
1171
1172 FREETMPS; LEAVE;
1173 }
677 } 1174 }
678 1175
679 if (*dec->cur != ',') 1176 if (dec->json.cb_object)
680 ERR (", or } expected while parsing object/hash"); 1177 {
1178 dSP;
1179 int count;
681 1180
682 ++dec->cur; 1181 ENTER; SAVETMPS; PUSHMARK (SP);
683 } 1182 XPUSHs (sv_2mortal (sv));
684 1183
685 return newRV_noinc ((SV *)hv); 1184 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1185
1186 if (count == 1)
1187 {
1188 sv = newSVsv (POPs);
1189 FREETMPS; LEAVE;
1190 return sv;
1191 }
1192
1193 SvREFCNT_inc (sv);
1194 FREETMPS; LEAVE;
1195 }
1196 }
1197
1198 return sv;
686 1199
687fail: 1200fail:
688 SvREFCNT_dec (hv); 1201 SvREFCNT_dec (hv);
1202 DEC_DEC_DEPTH;
689 return 0; 1203 return 0;
690} 1204}
691 1205
692static SV * 1206static SV *
693decode_sv (dec_t *dec) 1207decode_sv (dec_t *dec)
694{ 1208{
695 WS; 1209 decode_ws (dec);
1210
1211 // the beauty of JSON: you need exactly one character lookahead
1212 // to parse anything.
696 switch (*dec->cur) 1213 switch (*dec->cur)
697 { 1214 {
698 case '"': ++dec->cur; return decode_str (dec); 1215 case '"': ++dec->cur; return decode_str (dec);
699 case '[': ++dec->cur; return decode_av (dec); 1216 case '[': ++dec->cur; return decode_av (dec);
700 case '{': ++dec->cur; return decode_hv (dec); 1217 case '{': ++dec->cur; return decode_hv (dec);
706 1223
707 case 't': 1224 case 't':
708 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1225 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709 { 1226 {
710 dec->cur += 4; 1227 dec->cur += 4;
711 return newSViv (1); 1228 return SvREFCNT_inc (json_true);
712 } 1229 }
713 else 1230 else
714 ERR ("'true' expected"); 1231 ERR ("'true' expected");
715 1232
716 break; 1233 break;
717 1234
718 case 'f': 1235 case 'f':
719 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1236 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720 { 1237 {
721 dec->cur += 5; 1238 dec->cur += 5;
722 return newSViv (0); 1239 return SvREFCNT_inc (json_false);
723 } 1240 }
724 else 1241 else
725 ERR ("'false' expected"); 1242 ERR ("'false' expected");
726 1243
727 break; 1244 break;
728 1245
729 case 'n': 1246 case 'n':
730 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1247 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731 { 1248 {
732 dec->cur += 4; 1249 dec->cur += 4;
733 return newSViv (1); 1250 return newSVsv (&PL_sv_undef);
734 } 1251 }
735 else 1252 else
736 ERR ("'null' expected"); 1253 ERR ("'null' expected");
737 1254
738 break; 1255 break;
739 1256
740 default: 1257 default:
741 ERR ("malformed json string"); 1258 ERR ("malformed JSON string, neither array, object, number, string or atom");
742 break; 1259 break;
743 } 1260 }
744 1261
745fail: 1262fail:
746 return 0; 1263 return 0;
747} 1264}
748 1265
749static SV * 1266static SV *
750decode_json (SV *string, UV flags) 1267decode_json (SV *string, JSON *json, UV *offset_return)
751{ 1268{
1269 dec_t dec;
1270 UV offset;
752 SV *sv; 1271 SV *sv;
753 1272
1273 SvGETMAGIC (string);
1274 SvUPGRADE (string, SVt_PV);
1275
1276 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1277 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1278 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1279
754 if (!(flags & F_UTF8)) 1280 if (json->flags & F_UTF8)
1281 sv_utf8_downgrade (string, 0);
1282 else
755 sv_utf8_upgrade (string); 1283 sv_utf8_upgrade (string);
756 1284
757 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1285 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758 1286
759 dec_t dec; 1287 dec.json = *json;
760 dec.flags = flags;
761 dec.cur = SvPVX (string); 1288 dec.cur = SvPVX (string);
762 dec.end = SvEND (string); 1289 dec.end = SvEND (string);
763 dec.err = 0; 1290 dec.err = 0;
1291 dec.depth = 0;
1292 dec.maxdepth = DEC_DEPTH (dec.json.flags);
764 1293
765 *dec.end = 1; // invalid anywhere 1294 if (dec.json.cb_object || dec.json.cb_sk_object)
1295 dec.json.flags |= F_HOOK;
1296
1297 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
766 sv = decode_sv (&dec); 1298 sv = decode_sv (&dec);
767 *dec.end = 0; 1299
1300 if (!(offset_return || !sv))
1301 {
1302 // check for trailing garbage
1303 decode_ws (&dec);
1304
1305 if (*dec.cur)
1306 {
1307 dec.err = "garbage after JSON object";
1308 SvREFCNT_dec (sv);
1309 sv = 0;
1310 }
1311 }
1312
1313 if (offset_return || !sv)
1314 {
1315 offset = dec.json.flags & F_UTF8
1316 ? dec.cur - SvPVX (string)
1317 : utf8_distance (dec.cur, SvPVX (string));
1318
1319 if (offset_return)
1320 *offset_return = offset;
1321 }
768 1322
769 if (!sv) 1323 if (!sv)
770 { 1324 {
771 IV offset = utf8_distance (dec.cur, SvPVX (string));
772 SV *uni = sv_newmortal (); 1325 SV *uni = sv_newmortal ();
773 1326
1327 // horrible hack to silence warning inside pv_uni_display
1328 COP cop = *PL_curcop;
1329 cop.cop_warnings = pWARN_NONE;
1330 ENTER;
1331 SAVEVPTR (PL_curcop);
1332 PL_curcop = &cop;
774 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1333 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1334 LEAVE;
1335
775 croak ("%s, at character %d (%s)", 1336 croak ("%s, at character offset %d [\"%s\"]",
776 dec.err, 1337 dec.err,
777 (int)offset, 1338 (int)offset,
778 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1339 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779 } 1340 }
780 1341
781 sv = sv_2mortal (sv); 1342 sv = sv_2mortal (sv);
782 1343
783 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1344 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
784 croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); 1345 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785 1346
786 return sv; 1347 return sv;
787} 1348}
788 1349
1350/////////////////////////////////////////////////////////////////////////////
1351// XS interface functions
1352
789MODULE = JSON::XS PACKAGE = JSON::XS 1353MODULE = JSON::XS PACKAGE = JSON::XS
790 1354
791BOOT: 1355BOOT:
792{ 1356{
793 int i; 1357 int i;
794 1358
795 memset (decode_hexdigit, 0xff, 256);
796 for (i = 10; i--; ) 1359 for (i = 0; i < 256; ++i)
797 decode_hexdigit ['0' + i] = i; 1360 decode_hexdigit [i] =
1361 i >= '0' && i <= '9' ? i - '0'
1362 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1363 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1364 : -1;
798 1365
799 for (i = 6; --i; ) 1366 json_stash = gv_stashpv ("JSON::XS" , 1);
1367 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1368
1369 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1370 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1371}
1372
1373PROTOTYPES: DISABLE
1374
1375void CLONE (...)
1376 CODE:
1377 json_stash = 0;
1378 json_boolean_stash = 0;
1379
1380void new (char *klass)
1381 PPCODE:
1382{
1383 HV *stash = !JSON_SLOW || json_stash
1384 ? json_stash
1385 : gv_stashpv ("JSON::XS", 1);
1386 SV *pv = NEWSV (0, sizeof (JSON));
1387 SvPOK_only (pv);
1388 Zero (SvPVX (pv), 1, JSON);
1389 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1390 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));
1391}
1392
1393void ascii (JSON *self, int enable = 1)
1394 ALIAS:
1395 ascii = F_ASCII
1396 latin1 = F_LATIN1
1397 utf8 = F_UTF8
1398 indent = F_INDENT
1399 canonical = F_CANONICAL
1400 space_before = F_SPACE_BEFORE
1401 space_after = F_SPACE_AFTER
1402 pretty = F_PRETTY
1403 allow_nonref = F_ALLOW_NONREF
1404 shrink = F_SHRINK
1405 allow_blessed = F_ALLOW_BLESSED
1406 convert_blessed = F_CONV_BLESSED
1407 PPCODE:
1408{
1409 if (enable)
1410 self->flags |= ix;
1411 else
1412 self->flags &= ~ix;
1413
1414 XPUSHs (ST (0));
1415}
1416
1417void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1418 PPCODE:
1419{
1420 UV log2 = 0;
1421
1422 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1423
1424 while ((1UL << log2) < max_depth)
1425 ++log2;
1426
1427 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1428
1429 XPUSHs (ST (0));
1430}
1431
1432void max_size (JSON *self, UV max_size = 0)
1433 PPCODE:
1434{
1435 UV log2 = 0;
1436
1437 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1438 if (max_size == 1) max_size = 2;
1439
1440 while ((1UL << log2) < max_size)
1441 ++log2;
1442
1443 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1444
1445 XPUSHs (ST (0));
1446}
1447
1448void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1449 PPCODE:
1450{
1451 SvREFCNT_dec (self->cb_object);
1452 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1453
1454 XPUSHs (ST (0));
1455}
1456
1457void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1458 PPCODE:
1459{
1460 if (!self->cb_sk_object)
1461 self->cb_sk_object = newHV ();
1462
1463 if (SvOK (cb))
1464 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1465 else
800 { 1466 {
801 decode_hexdigit ['a' + i] = 10 + i; 1467 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
802 decode_hexdigit ['A' + i] = 10 + i; 1468
1469 if (!HvKEYS (self->cb_sk_object))
1470 {
1471 SvREFCNT_dec (self->cb_sk_object);
1472 self->cb_sk_object = 0;
1473 }
803 } 1474 }
804 1475
805 json_stash = gv_stashpv ("JSON::XS", 1); 1476 XPUSHs (ST (0));
806} 1477}
807 1478
808SV *new (char *dummy) 1479void encode (JSON *self, SV *scalar)
1480 PPCODE:
1481 XPUSHs (encode_json (scalar, self));
1482
1483void decode (JSON *self, SV *jsonstr)
1484 PPCODE:
1485 XPUSHs (decode_json (jsonstr, self, 0));
1486
1487void decode_prefix (JSON *self, SV *jsonstr)
1488 PPCODE:
1489{
1490 UV offset;
1491 EXTEND (SP, 2);
1492 PUSHs (decode_json (jsonstr, self, &offset));
1493 PUSHs (sv_2mortal (newSVuv (offset)));
1494}
1495
1496void DESTROY (JSON *self)
809 CODE: 1497 CODE:
810 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1498 SvREFCNT_dec (self->cb_sk_object);
811 OUTPUT: 1499 SvREFCNT_dec (self->cb_object);
812 RETVAL
813 1500
814SV *ascii (SV *self, int enable) 1501PROTOTYPES: ENABLE
815 ALIAS:
816 ascii = F_ASCII
817 utf8 = F_UTF8
818 indent = F_INDENT
819 canonical = F_CANONICAL
820 space_before = F_SPACE_BEFORE
821 space_after = F_SPACE_AFTER
822 json_rpc = F_JSON_RPC
823 pretty = F_PRETTY
824 allow_nonref = F_ALLOW_NONREF
825 CODE:
826{
827 UV *uv = SvJSON (self);
828 if (enable)
829 *uv |= ix;
830 else
831 *uv &= ~ix;
832
833 RETVAL = newSVsv (self);
834}
835 OUTPUT:
836 RETVAL
837
838void encode (SV *self, SV *scalar)
839 PPCODE:
840 XPUSHs (encode_json (scalar, *SvJSON (self)));
841
842void decode (SV *self, SV *jsonstr)
843 PPCODE:
844 XPUSHs (decode_json (jsonstr, *SvJSON (self)));
845 1502
846void to_json (SV *scalar) 1503void to_json (SV *scalar)
847 PPCODE: 1504 PPCODE:
1505{
1506 JSON json = { F_DEFAULT | F_UTF8 };
848 XPUSHs (encode_json (scalar, F_UTF8)); 1507 XPUSHs (encode_json (scalar, &json));
1508}
849 1509
850void from_json (SV *jsonstr) 1510void from_json (SV *jsonstr)
851 PPCODE: 1511 PPCODE:
1512{
1513 JSON json = { F_DEFAULT | F_UTF8 };
852 XPUSHs (decode_json (jsonstr, F_UTF8)); 1514 XPUSHs (decode_json (jsonstr, &json, 0));
1515}
853 1516

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines