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 1.57 by root, Mon Aug 13 16:05:42 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 1
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 int count;
516 dSP;
517
518 ENTER; SAVETMPS; PUSHMARK (SP);
519 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
520
521 // calling with G_SCALAR ensures that we always get a 1 return value
522 PUTBACK;
523 call_sv ((SV *)GvCV (to_json), G_SCALAR);
524 SPAGAIN;
525
526 // catch this surprisingly common error
527 if (SvROK (TOPs) && SvRV (TOPs) == sv)
528 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
529
530 sv = POPs;
531 PUTBACK;
532
533 encode_sv (enc, sv);
534
535 FREETMPS; LEAVE;
536 }
537 else if (enc->json.flags & F_ALLOW_BLESSED)
538 encode_str (enc, "null", 4, 0);
539 else
540 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
541 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
542 }
543 else if (enc->json.flags & F_ALLOW_BLESSED)
544 encode_str (enc, "null", 4, 0);
545 else
546 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
547 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
548 }
549 }
550 else if (svt == SVt_PVHV)
551 encode_hv (enc, (HV *)sv);
552 else if (svt == SVt_PVAV)
553 encode_av (enc, (AV *)sv);
554 else if (svt < SVt_PVAV)
555 {
556 STRLEN len = 0;
557 char *pv = svt ? SvPV (sv, len) : 0;
558
559 if (len == 1 && *pv == '1')
560 encode_str (enc, "true", 4, 0);
561 else if (len == 1 && *pv == '0')
562 encode_str (enc, "false", 5, 0);
563 else
564 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
565 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
566 }
567 else
568 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
569 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313} 570}
314 571
315static void 572static void
316encode_sv (enc_t *enc, SV *sv) 573encode_sv (enc_t *enc, SV *sv)
317{ 574{
575 SvGETMAGIC (sv);
576
318 if (SvPOKp (sv)) 577 if (SvPOKp (sv))
319 { 578 {
320 STRLEN len; 579 STRLEN len;
321 char *str = SvPV (sv, len); 580 char *str = SvPV (sv, len);
322 encode_ch (enc, '"'); 581 encode_ch (enc, '"');
323 encode_str (enc, str, len, SvUTF8 (sv)); 582 encode_str (enc, str, len, SvUTF8 (sv));
324 encode_ch (enc, '"'); 583 encode_ch (enc, '"');
325 } 584 }
326 else if (SvNOKp (sv)) 585 else if (SvNOKp (sv))
327 { 586 {
587 // trust that perl will do the right thing w.r.t. JSON syntax.
328 need (enc, NV_DIG + 32); 588 need (enc, NV_DIG + 32);
329 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 589 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330 enc->cur += strlen (enc->cur); 590 enc->cur += strlen (enc->cur);
331 } 591 }
332 else if (SvIOKp (sv)) 592 else if (SvIOKp (sv))
333 { 593 {
334 need (enc, 64); 594 // we assume we can always read an IV as a UV
595 if (SvUV (sv) & ~(UV)0x7fff)
596 {
597 // large integer, use the (rather slow) snprintf way.
598 need (enc, sizeof (UV) * 3);
335 enc->cur += 599 enc->cur +=
336 SvIsUV(sv) 600 SvIsUV(sv)
337 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 601 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
338 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 602 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
603 }
604 else
605 {
606 // optimise the "small number case"
607 // code will likely be branchless and use only a single multiplication
608 I32 i = SvIV (sv);
609 U32 u;
610 char digit, nz = 0;
611
612 need (enc, 6);
613
614 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
615 u = i < 0 ? -i : i;
616
617 // convert to 4.28 fixed-point representation
618 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
619
620 // now output digit by digit, each time masking out the integer part
621 // and multiplying by 5 while moving the decimal point one to the right,
622 // resulting in a net multiplication by 10.
623 // we always write the digit to memory but conditionally increment
624 // the pointer, to ease the usage of conditional move instructions.
625 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
626 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
627 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
628 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
629 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
630 }
339 } 631 }
340 else if (SvROK (sv)) 632 else if (SvROK (sv))
341 { 633 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)) 634 else if (!SvOK (sv))
357 encode_str (enc, "null", 4, 0); 635 encode_str (enc, "null", 4, 0);
358 else 636 else
359 croak ("encountered perl type that JSON cannot handle"); 637 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
638 SvPV_nolen (sv), SvFLAGS (sv));
360} 639}
361 640
362static SV * 641static SV *
363encode_json (SV *scalar, UV flags) 642encode_json (SV *scalar, JSON *json)
364{ 643{
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; 644 enc_t enc;
369 enc.flags = flags; 645
646 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
647 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
648
649 enc.json = *json;
370 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 650 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371 enc.cur = SvPVX (enc.sv); 651 enc.cur = SvPVX (enc.sv);
372 enc.end = SvEND (enc.sv); 652 enc.end = SvEND (enc.sv);
373 enc.max_recurse = 0;
374 enc.indent = 0; 653 enc.indent = 0;
654 enc.maxdepth = DEC_DEPTH (enc.json.flags);
375 655
376 SvPOK_only (enc.sv); 656 SvPOK_only (enc.sv);
377 encode_sv (&enc, scalar); 657 encode_sv (&enc, scalar);
378 658
659 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
660 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
661
379 if (!(flags & (F_ASCII | F_UTF8))) 662 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380 SvUTF8_on (enc.sv); 663 SvUTF8_on (enc.sv);
381 664
382 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 665 if (enc.json.flags & F_SHRINK)
666 shrink (enc.sv);
667
383 return enc.sv; 668 return enc.sv;
384} 669}
385 670
386///////////////////////////////////////////////////////////////////////////// 671/////////////////////////////////////////////////////////////////////////////
672// decoder
387 673
388#define WS \ 674// structure used for decoding JSON
675typedef struct
676{
677 char *cur; // current parser pointer
678 char *end; // end of input string
679 const char *err; // parse error, if != 0
680 JSON json;
681 U32 depth; // recursion depth
682 U32 maxdepth; // recursion depth limit
683} dec_t;
684
685inline void
686decode_ws (dec_t *dec)
687{
389 for (;;) \ 688 for (;;)
390 { \ 689 {
391 char ch = *dec->cur; \ 690 char ch = *dec->cur;
691
392 if (ch > 0x20 \ 692 if (ch > 0x20
393 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 693 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
394 break; \ 694 break;
695
395 ++dec->cur; \ 696 ++dec->cur;
396 } 697 }
698}
397 699
398#define ERR(reason) SB dec->err = reason; goto fail; SE 700#define ERR(reason) SB dec->err = reason; goto fail; SE
701
399#define EXPECT_CH(ch) SB \ 702#define EXPECT_CH(ch) SB \
400 if (*dec->cur != ch) \ 703 if (*dec->cur != ch) \
401 ERR (# ch " expected"); \ 704 ERR (# ch " expected"); \
402 ++dec->cur; \ 705 ++dec->cur; \
403 SE 706 SE
404 707
708#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
709#define DEC_DEC_DEPTH --dec->depth
710
405static SV *decode_sv (dec_t *dec); 711static 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
411 712
412static signed char decode_hexdigit[256]; 713static signed char decode_hexdigit[256];
413 714
414static UV 715static UV
415decode_4hex (dec_t *dec) 716decode_4hex (dec_t *dec)
416{ 717{
417 signed char d1, d2, d3, d4; 718 signed char d1, d2, d3, d4;
719 unsigned char *cur = (unsigned char *)dec->cur;
418 720
419 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 721 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"); 722 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]]; 723 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"); 724 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");
427 725
428 dec->cur += 4; 726 dec->cur += 4;
429 727
430 return ((UV)d1) << 12 728 return ((UV)d1) << 12
431 | ((UV)d2) << 8 729 | ((UV)d2) << 8
437} 735}
438 736
439static SV * 737static SV *
440decode_str (dec_t *dec) 738decode_str (dec_t *dec)
441{ 739{
442 SV *sv = NEWSV (0,2); 740 SV *sv = 0;
443 STRLEN cur = 0;
444 int utf8 = 0; 741 int utf8 = 0;
742 char *dec_cur = dec->cur;
445 743
446 for (;;) 744 do
447 { 745 {
448 unsigned char ch = *(unsigned char *)dec->cur; 746 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
747 char *cur = buf;
449 748
450 if (ch == '"') 749 do
451 break;
452 else if (ch == '\\')
453 { 750 {
454 switch (*++dec->cur) 751 unsigned char ch = *(unsigned char *)dec_cur++;
752
753 if (expect_false (ch == '"'))
754 {
755 --dec_cur;
756 break;
455 { 757 }
456 case '\\': 758 else if (expect_false (ch == '\\'))
457 case '/': 759 {
458 case '"': APPEND_CH (*dec->cur++); break; 760 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 { 761 {
468 UV lo, hi; 762 case '\\':
469 ++dec->cur; 763 case '/':
764 case '"': *cur++ = *dec_cur++; break;
470 765
471 hi = decode_4hex (dec); 766 case 'b': ++dec_cur; *cur++ = '\010'; break;
472 if (hi == (UV)-1) 767 case 't': ++dec_cur; *cur++ = '\011'; break;
473 goto fail; 768 case 'n': ++dec_cur; *cur++ = '\012'; break;
769 case 'f': ++dec_cur; *cur++ = '\014'; break;
770 case 'r': ++dec_cur; *cur++ = '\015'; break;
474 771
475 // possibly a surrogate pair 772 case 'u':
476 if (hi >= 0xd800 && hi < 0xdc00)
477 { 773 {
478 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 774 UV lo, hi;
479 ERR ("illegal surrogate character"); 775 ++dec_cur;
480 776
481 dec->cur += 2; 777 dec->cur = dec_cur;
482
483 lo = decode_4hex (dec); 778 hi = decode_4hex (dec);
779 dec_cur = dec->cur;
484 if (lo == (UV)-1) 780 if (hi == (UV)-1)
485 goto fail; 781 goto fail;
486 782
783 // possibly a surrogate pair
784 if (hi >= 0xd800)
785 if (hi < 0xdc00)
786 {
787 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
788 ERR ("missing low surrogate character in surrogate pair");
789
790 dec_cur += 2;
791
792 dec->cur = dec_cur;
793 lo = decode_4hex (dec);
794 dec_cur = dec->cur;
795 if (lo == (UV)-1)
796 goto fail;
797
487 if (lo < 0xdc00 || lo >= 0xe000) 798 if (lo < 0xdc00 || lo >= 0xe000)
488 ERR ("surrogate pair expected"); 799 ERR ("surrogate pair expected");
489 800
490 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 801 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
802 }
803 else if (hi < 0xe000)
804 ERR ("missing high surrogate character in surrogate pair");
805
806 if (hi >= 0x80)
807 {
808 utf8 = 1;
809
810 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
811 }
812 else
813 *cur++ = hi;
491 } 814 }
492 else if (lo >= 0xdc00 && lo < 0xe000)
493 ERR ("illegal surrogate character");
494
495 if (hi >= 0x80)
496 { 815 break;
497 utf8 = 1;
498 816
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 817 default:
503 APPEND_CH (hi); 818 --dec_cur;
819 ERR ("illegal backslash escape sequence in string");
504 } 820 }
821 }
822 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
823 *cur++ = ch;
824 else if (ch >= 0x80)
825 {
826 STRLEN clen;
827 UV uch;
828
829 --dec_cur;
830
831 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
832 if (clen == (STRLEN)-1)
833 ERR ("malformed UTF-8 character in JSON string");
834
835 do
836 *cur++ = *dec_cur++;
837 while (--clen);
838
839 utf8 = 1;
840 }
841 else
842 {
843 --dec_cur;
844
845 if (!ch)
846 ERR ("unexpected end of string while parsing JSON string");
505 break; 847 else
848 ERR ("invalid character encountered while parsing JSON string");
506 } 849 }
507 } 850 }
508 else if (ch >= 0x20 && ch <= 0x7f) 851 while (cur < buf + SHORT_STRING_LEN);
509 APPEND_CH (*dec->cur++); 852
510 else if (ch >= 0x80) 853 {
854 STRLEN len = cur - buf;
855
856 if (sv)
511 { 857 {
512 STRLEN clen; 858 SvGROW (sv, SvCUR (sv) + len + 1);
513 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 859 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
514 if (clen < 0) 860 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 } 861 }
521 else 862 else
522 ERR ("invalid character encountered"); 863 sv = newSVpvn (buf, len);
523 } 864 }
865 }
866 while (*dec_cur != '"');
524 867
525 ++dec->cur; 868 ++dec_cur;
526 869
870 if (sv)
871 {
527 SvPOK_only (sv); 872 SvPOK_only (sv);
528
529 SvCUR_set (sv, cur);
530 *SvEND (sv) = 0; 873 *SvEND (sv) = 0;
531 874
532 if (utf8) 875 if (utf8)
533 SvUTF8_on (sv); 876 SvUTF8_on (sv);
877 }
878 else
879 sv = newSVpvn ("", 0);
534 880
881 dec->cur = dec_cur;
535 return sv; 882 return sv;
536 883
537fail: 884fail:
538 SvREFCNT_dec (sv); 885 dec->cur = dec_cur;
539 return 0; 886 return 0;
540} 887}
541 888
542static SV * 889static SV *
543decode_num (dec_t *dec) 890decode_num (dec_t *dec)
553 { 900 {
554 ++dec->cur; 901 ++dec->cur;
555 if (*dec->cur >= '0' && *dec->cur <= '9') 902 if (*dec->cur >= '0' && *dec->cur <= '9')
556 ERR ("malformed number (leading zero must not be followed by another digit)"); 903 ERR ("malformed number (leading zero must not be followed by another digit)");
557 } 904 }
558 905 else if (*dec->cur < '0' || *dec->cur > '9')
559 // int 906 ERR ("malformed number (no digits after initial minus)");
907 else
908 do
909 {
910 ++dec->cur;
911 }
560 while (*dec->cur >= '0' && *dec->cur <= '9') 912 while (*dec->cur >= '0' && *dec->cur <= '9');
561 ++dec->cur;
562 913
563 // [frac] 914 // [frac]
564 if (*dec->cur == '.') 915 if (*dec->cur == '.')
565 { 916 {
566 is_nv = 1; 917 ++dec->cur;
918
919 if (*dec->cur < '0' || *dec->cur > '9')
920 ERR ("malformed number (no digits after decimal point)");
567 921
568 do 922 do
569 { 923 {
570 ++dec->cur; 924 ++dec->cur;
571 } 925 }
572 while (*dec->cur >= '0' && *dec->cur <= '9'); 926 while (*dec->cur >= '0' && *dec->cur <= '9');
927
928 is_nv = 1;
573 } 929 }
574 930
575 // [exp] 931 // [exp]
576 if (*dec->cur == 'e' || *dec->cur == 'E') 932 if (*dec->cur == 'e' || *dec->cur == 'E')
577 { 933 {
578 is_nv = 1;
579
580 ++dec->cur; 934 ++dec->cur;
935
581 if (*dec->cur == '-' || *dec->cur == '+') 936 if (*dec->cur == '-' || *dec->cur == '+')
582 ++dec->cur; 937 ++dec->cur;
583 938
939 if (*dec->cur < '0' || *dec->cur > '9')
940 ERR ("malformed number (no digits after exp sign)");
941
942 do
943 {
944 ++dec->cur;
945 }
584 while (*dec->cur >= '0' && *dec->cur <= '9') 946 while (*dec->cur >= '0' && *dec->cur <= '9');
585 ++dec->cur; 947
948 is_nv = 1;
586 } 949 }
587 950
588 if (!is_nv) 951 if (!is_nv)
589 { 952 {
590 UV uv; 953 int len = dec->cur - start;
591 int numtype = grok_number (start, dec->cur - start, &uv); 954
592 if (numtype & IS_NUMBER_IN_UV) 955 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
593 if (numtype & IS_NUMBER_NEG) 956 if (*start == '-')
957 switch (len)
594 { 958 {
595 if (uv < (UV)IV_MIN) 959 case 2: return newSViv (-( start [1] - '0' * 1));
596 return newSViv (-(IV)uv); 960 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
961 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
962 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
597 } 963 }
964 else
965 switch (len)
966 {
967 case 1: return newSViv ( start [0] - '0' * 1);
968 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
969 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
970 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
971 }
972
973 {
974 UV uv;
975 int numtype = grok_number (start, len, &uv);
976 if (numtype & IS_NUMBER_IN_UV)
977 if (numtype & IS_NUMBER_NEG)
978 {
979 if (uv < (UV)IV_MIN)
980 return newSViv (-(IV)uv);
981 }
598 else 982 else
599 return newSVuv (uv); 983 return newSVuv (uv);
600 } 984 }
601 985
986 len -= *start == '-' ? 1 : 0;
987
988 // does not fit into IV or UV, try NV
989 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
990 #if defined (LDBL_DIG)
991 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
992 #endif
993 )
994 // fits into NV without loss of precision
995 return newSVnv (Atof (start));
996
997 // everything else fails, convert it to a string
998 return newSVpvn (start, dec->cur - start);
999 }
1000
1001 // loss of precision here
602 return newSVnv (Atof (start)); 1002 return newSVnv (Atof (start));
603 1003
604fail: 1004fail:
605 return 0; 1005 return 0;
606} 1006}
608static SV * 1008static SV *
609decode_av (dec_t *dec) 1009decode_av (dec_t *dec)
610{ 1010{
611 AV *av = newAV (); 1011 AV *av = newAV ();
612 1012
1013 DEC_INC_DEPTH;
1014 decode_ws (dec);
1015
1016 if (*dec->cur == ']')
1017 ++dec->cur;
1018 else
613 for (;;) 1019 for (;;)
614 { 1020 {
615 SV *value; 1021 SV *value;
616 1022
617 value = decode_sv (dec); 1023 value = decode_sv (dec);
618 if (!value) 1024 if (!value)
619 goto fail; 1025 goto fail;
620 1026
621 av_push (av, value); 1027 av_push (av, value);
622 1028
623 WS; 1029 decode_ws (dec);
624 1030
625 if (*dec->cur == ']') 1031 if (*dec->cur == ']')
626 { 1032 {
627 ++dec->cur; 1033 ++dec->cur;
628 break; 1034 break;
1035 }
629 } 1036
630
631 if (*dec->cur != ',') 1037 if (*dec->cur != ',')
632 ERR (", or ] expected while parsing array"); 1038 ERR (", or ] expected while parsing array");
633 1039
634 ++dec->cur; 1040 ++dec->cur;
635 } 1041 }
636 1042
1043 DEC_DEC_DEPTH;
637 return newRV_noinc ((SV *)av); 1044 return newRV_noinc ((SV *)av);
638 1045
639fail: 1046fail:
640 SvREFCNT_dec (av); 1047 SvREFCNT_dec (av);
1048 DEC_DEC_DEPTH;
641 return 0; 1049 return 0;
642} 1050}
643 1051
644static SV * 1052static SV *
645decode_hv (dec_t *dec) 1053decode_hv (dec_t *dec)
646{ 1054{
1055 SV *sv;
647 HV *hv = newHV (); 1056 HV *hv = newHV ();
648 1057
1058 DEC_INC_DEPTH;
1059 decode_ws (dec);
1060
1061 if (*dec->cur == '}')
1062 ++dec->cur;
1063 else
649 for (;;) 1064 for (;;)
650 { 1065 {
651 SV *key, *value; 1066 decode_ws (dec); EXPECT_CH ('"');
652 1067
653 WS; EXPECT_CH ('"'); 1068 // heuristic: assume that
654 1069 // a) decode_str + hv_store_ent are abysmally slow.
655 key = decode_str (dec); 1070 // b) most hash keys are short, simple ascii text.
656 if (!key) 1071 // => try to "fast-match" such strings to avoid
657 goto fail; 1072 // the overhead of decode_str + hv_store_ent.
658
659 WS; EXPECT_CH (':');
660
661 value = decode_sv (dec);
662 if (!value)
663 { 1073 {
1074 SV *value;
1075 char *p = dec->cur;
1076 char *e = p + 24; // only try up to 24 bytes
1077
1078 for (;;)
1079 {
1080 // the >= 0x80 is true on most architectures
1081 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1082 {
1083 // slow path, back up and use decode_str
1084 SV *key = decode_str (dec);
1085 if (!key)
1086 goto fail;
1087
1088 decode_ws (dec); EXPECT_CH (':');
1089
1090 value = decode_sv (dec);
1091 if (!value)
1092 {
1093 SvREFCNT_dec (key);
1094 goto fail;
1095 }
1096
1097 hv_store_ent (hv, key, value, 0);
664 SvREFCNT_dec (key); 1098 SvREFCNT_dec (key);
1099
1100 break;
1101 }
1102 else if (*p == '"')
1103 {
1104 // fast path, got a simple key
1105 char *key = dec->cur;
1106 int len = p - key;
1107 dec->cur = p + 1;
1108
1109 decode_ws (dec); EXPECT_CH (':');
1110
1111 value = decode_sv (dec);
1112 if (!value)
665 goto fail; 1113 goto fail;
1114
1115 hv_store (hv, key, len, value, 0);
1116
1117 break;
1118 }
1119
1120 ++p;
1121 }
666 } 1122 }
667 1123
668 //TODO: optimise 1124 decode_ws (dec);
669 hv_store_ent (hv, key, value, 0);
670 1125
671 WS;
672
673 if (*dec->cur == '}') 1126 if (*dec->cur == '}')
1127 {
1128 ++dec->cur;
1129 break;
1130 }
1131
1132 if (*dec->cur != ',')
1133 ERR (", or } expected while parsing object/hash");
1134
1135 ++dec->cur;
1136 }
1137
1138 DEC_DEC_DEPTH;
1139 sv = newRV_noinc ((SV *)hv);
1140
1141 // check filter callbacks
1142 if (dec->json.flags & F_HOOK)
1143 {
1144 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
674 { 1145 {
675 ++dec->cur; 1146 HE *cb, *he;
676 break; 1147
1148 hv_iterinit (hv);
1149 he = hv_iternext (hv);
1150 hv_iterinit (hv);
1151
1152 // the next line creates a mortal sv each time its called.
1153 // might want to optimise this for common cases.
1154 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1155
1156 if (cb)
1157 {
1158 dSP;
1159 int count;
1160
1161 ENTER; SAVETMPS; PUSHMARK (SP);
1162 XPUSHs (HeVAL (he));
1163
1164 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1165
1166 if (count == 1)
1167 {
1168 sv = newSVsv (POPs);
1169 FREETMPS; LEAVE;
1170 return sv;
1171 }
1172
1173 FREETMPS; LEAVE;
1174 }
677 } 1175 }
678 1176
679 if (*dec->cur != ',') 1177 if (dec->json.cb_object)
680 ERR (", or } expected while parsing object/hash"); 1178 {
1179 dSP;
1180 int count;
681 1181
682 ++dec->cur; 1182 ENTER; SAVETMPS; PUSHMARK (SP);
683 } 1183 XPUSHs (sv_2mortal (sv));
684 1184
685 return newRV_noinc ((SV *)hv); 1185 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1186
1187 if (count == 1)
1188 {
1189 sv = newSVsv (POPs);
1190 FREETMPS; LEAVE;
1191 return sv;
1192 }
1193
1194 SvREFCNT_inc (sv);
1195 FREETMPS; LEAVE;
1196 }
1197 }
1198
1199 return sv;
686 1200
687fail: 1201fail:
688 SvREFCNT_dec (hv); 1202 SvREFCNT_dec (hv);
1203 DEC_DEC_DEPTH;
689 return 0; 1204 return 0;
690} 1205}
691 1206
692static SV * 1207static SV *
693decode_sv (dec_t *dec) 1208decode_sv (dec_t *dec)
694{ 1209{
695 WS; 1210 decode_ws (dec);
1211
1212 // the beauty of JSON: you need exactly one character lookahead
1213 // to parse anything.
696 switch (*dec->cur) 1214 switch (*dec->cur)
697 { 1215 {
698 case '"': ++dec->cur; return decode_str (dec); 1216 case '"': ++dec->cur; return decode_str (dec);
699 case '[': ++dec->cur; return decode_av (dec); 1217 case '[': ++dec->cur; return decode_av (dec);
700 case '{': ++dec->cur; return decode_hv (dec); 1218 case '{': ++dec->cur; return decode_hv (dec);
706 1224
707 case 't': 1225 case 't':
708 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1226 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709 { 1227 {
710 dec->cur += 4; 1228 dec->cur += 4;
711 return newSViv (1); 1229 return SvREFCNT_inc (json_true);
712 } 1230 }
713 else 1231 else
714 ERR ("'true' expected"); 1232 ERR ("'true' expected");
715 1233
716 break; 1234 break;
717 1235
718 case 'f': 1236 case 'f':
719 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1237 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720 { 1238 {
721 dec->cur += 5; 1239 dec->cur += 5;
722 return newSViv (0); 1240 return SvREFCNT_inc (json_false);
723 } 1241 }
724 else 1242 else
725 ERR ("'false' expected"); 1243 ERR ("'false' expected");
726 1244
727 break; 1245 break;
728 1246
729 case 'n': 1247 case 'n':
730 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1248 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731 { 1249 {
732 dec->cur += 4; 1250 dec->cur += 4;
733 return newSViv (1); 1251 return newSVsv (&PL_sv_undef);
734 } 1252 }
735 else 1253 else
736 ERR ("'null' expected"); 1254 ERR ("'null' expected");
737 1255
738 break; 1256 break;
739 1257
740 default: 1258 default:
741 ERR ("malformed json string"); 1259 ERR ("malformed JSON string, neither array, object, number, string or atom");
742 break; 1260 break;
743 } 1261 }
744 1262
745fail: 1263fail:
746 return 0; 1264 return 0;
747} 1265}
748 1266
749static SV * 1267static SV *
750decode_json (SV *string, UV flags) 1268decode_json (SV *string, JSON *json, UV *offset_return)
751{ 1269{
1270 dec_t dec;
1271 UV offset;
752 SV *sv; 1272 SV *sv;
753 1273
1274 SvGETMAGIC (string);
1275 SvUPGRADE (string, SVt_PV);
1276
1277 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1278 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1279 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1280
754 if (!(flags & F_UTF8)) 1281 if (json->flags & F_UTF8)
1282 sv_utf8_downgrade (string, 0);
1283 else
755 sv_utf8_upgrade (string); 1284 sv_utf8_upgrade (string);
756 1285
757 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1286 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758 1287
759 dec_t dec; 1288 dec.json = *json;
760 dec.flags = flags;
761 dec.cur = SvPVX (string); 1289 dec.cur = SvPVX (string);
762 dec.end = SvEND (string); 1290 dec.end = SvEND (string);
763 dec.err = 0; 1291 dec.err = 0;
1292 dec.depth = 0;
1293 dec.maxdepth = DEC_DEPTH (dec.json.flags);
764 1294
765 *dec.end = 1; // invalid anywhere 1295 if (dec.json.cb_object || dec.json.cb_sk_object)
1296 dec.json.flags |= F_HOOK;
1297
1298 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
766 sv = decode_sv (&dec); 1299 sv = decode_sv (&dec);
767 *dec.end = 0; 1300
1301 if (!(offset_return || !sv))
1302 {
1303 // check for trailing garbage
1304 decode_ws (&dec);
1305
1306 if (*dec.cur)
1307 {
1308 dec.err = "garbage after JSON object";
1309 SvREFCNT_dec (sv);
1310 sv = 0;
1311 }
1312 }
1313
1314 if (offset_return || !sv)
1315 {
1316 offset = dec.json.flags & F_UTF8
1317 ? dec.cur - SvPVX (string)
1318 : utf8_distance (dec.cur, SvPVX (string));
1319
1320 if (offset_return)
1321 *offset_return = offset;
1322 }
768 1323
769 if (!sv) 1324 if (!sv)
770 { 1325 {
771 IV offset = utf8_distance (dec.cur, SvPVX (string));
772 SV *uni = sv_newmortal (); 1326 SV *uni = sv_newmortal ();
773 1327
1328 // horrible hack to silence warning inside pv_uni_display
1329 COP cop = *PL_curcop;
1330 cop.cop_warnings = pWARN_NONE;
1331 ENTER;
1332 SAVEVPTR (PL_curcop);
1333 PL_curcop = &cop;
774 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1334 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1335 LEAVE;
1336
775 croak ("%s, at character %d (%s)", 1337 croak ("%s, at character offset %d [\"%s\"]",
776 dec.err, 1338 dec.err,
777 (int)offset, 1339 (int)offset,
778 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1340 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779 } 1341 }
780 1342
781 sv = sv_2mortal (sv); 1343 sv = sv_2mortal (sv);
782 1344
783 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1345 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)"); 1346 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785 1347
786 return sv; 1348 return sv;
787} 1349}
788 1350
1351/////////////////////////////////////////////////////////////////////////////
1352// XS interface functions
1353
789MODULE = JSON::XS PACKAGE = JSON::XS 1354MODULE = JSON::XS PACKAGE = JSON::XS
790 1355
791BOOT: 1356BOOT:
792{ 1357{
793 int i; 1358 int i;
794 1359
795 memset (decode_hexdigit, 0xff, 256);
796 for (i = 10; i--; ) 1360 for (i = 0; i < 256; ++i)
797 decode_hexdigit ['0' + i] = i; 1361 decode_hexdigit [i] =
1362 i >= '0' && i <= '9' ? i - '0'
1363 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1364 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1365 : -1;
798 1366
799 for (i = 6; --i; ) 1367 json_stash = gv_stashpv ("JSON::XS" , 1);
1368 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1369
1370 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1371 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1372}
1373
1374PROTOTYPES: DISABLE
1375
1376void CLONE (...)
1377 CODE:
1378 json_stash = 0;
1379 json_boolean_stash = 0;
1380
1381void new (char *klass)
1382 PPCODE:
1383{
1384 HV *stash = !JSON_SLOW || json_stash
1385 ? json_stash
1386 : gv_stashpv ("JSON::XS", 1);
1387 SV *pv = NEWSV (0, sizeof (JSON));
1388 SvPOK_only (pv);
1389 Zero (SvPVX (pv), 1, JSON);
1390 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1391 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));
1392}
1393
1394void ascii (JSON *self, int enable = 1)
1395 ALIAS:
1396 ascii = F_ASCII
1397 latin1 = F_LATIN1
1398 utf8 = F_UTF8
1399 indent = F_INDENT
1400 canonical = F_CANONICAL
1401 space_before = F_SPACE_BEFORE
1402 space_after = F_SPACE_AFTER
1403 pretty = F_PRETTY
1404 allow_nonref = F_ALLOW_NONREF
1405 shrink = F_SHRINK
1406 allow_blessed = F_ALLOW_BLESSED
1407 convert_blessed = F_CONV_BLESSED
1408 PPCODE:
1409{
1410 if (enable)
1411 self->flags |= ix;
1412 else
1413 self->flags &= ~ix;
1414
1415 XPUSHs (ST (0));
1416}
1417
1418void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1419 PPCODE:
1420{
1421 UV log2 = 0;
1422
1423 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1424
1425 while ((1UL << log2) < max_depth)
1426 ++log2;
1427
1428 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1429
1430 XPUSHs (ST (0));
1431}
1432
1433void max_size (JSON *self, UV max_size = 0)
1434 PPCODE:
1435{
1436 UV log2 = 0;
1437
1438 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1439 if (max_size == 1) max_size = 2;
1440
1441 while ((1UL << log2) < max_size)
1442 ++log2;
1443
1444 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1445
1446 XPUSHs (ST (0));
1447}
1448
1449void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1450 PPCODE:
1451{
1452 SvREFCNT_dec (self->cb_object);
1453 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1454
1455 XPUSHs (ST (0));
1456}
1457
1458void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1459 PPCODE:
1460{
1461 if (!self->cb_sk_object)
1462 self->cb_sk_object = newHV ();
1463
1464 if (SvOK (cb))
1465 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1466 else
800 { 1467 {
801 decode_hexdigit ['a' + i] = 10 + i; 1468 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
802 decode_hexdigit ['A' + i] = 10 + i; 1469
1470 if (!HvKEYS (self->cb_sk_object))
1471 {
1472 SvREFCNT_dec (self->cb_sk_object);
1473 self->cb_sk_object = 0;
1474 }
803 } 1475 }
804 1476
805 json_stash = gv_stashpv ("JSON::XS", 1); 1477 XPUSHs (ST (0));
806} 1478}
807 1479
808SV *new (char *dummy) 1480void encode (JSON *self, SV *scalar)
1481 PPCODE:
1482 XPUSHs (encode_json (scalar, self));
1483
1484void decode (JSON *self, SV *jsonstr)
1485 PPCODE:
1486 XPUSHs (decode_json (jsonstr, self, 0));
1487
1488void decode_prefix (JSON *self, SV *jsonstr)
1489 PPCODE:
1490{
1491 UV offset;
1492 EXTEND (SP, 2);
1493 PUSHs (decode_json (jsonstr, self, &offset));
1494 PUSHs (sv_2mortal (newSVuv (offset)));
1495}
1496
1497void DESTROY (JSON *self)
809 CODE: 1498 CODE:
810 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1499 SvREFCNT_dec (self->cb_sk_object);
811 OUTPUT: 1500 SvREFCNT_dec (self->cb_object);
812 RETVAL
813 1501
814SV *ascii (SV *self, int enable) 1502PROTOTYPES: 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 1503
846void to_json (SV *scalar) 1504void to_json (SV *scalar)
847 PPCODE: 1505 PPCODE:
1506{
1507 JSON json = { F_DEFAULT | F_UTF8 };
848 XPUSHs (encode_json (scalar, F_UTF8)); 1508 XPUSHs (encode_json (scalar, &json));
1509}
849 1510
850void from_json (SV *jsonstr) 1511void from_json (SV *jsonstr)
851 PPCODE: 1512 PPCODE:
1513{
1514 JSON json = { F_DEFAULT | F_UTF8 };
852 XPUSHs (decode_json (jsonstr, F_UTF8)); 1515 XPUSHs (decode_json (jsonstr, &json, 0));
1516}
853 1517

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines