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 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs.
Revision 1.49 by root, Sun Jul 1 23:40:07 2007 UTC

… …
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "assert.h" 5#include "assert.h"
6#include "string.h" 6#include "string.h"
7#include "stdlib.h" 7#include "stdlib.h"
8#include "stdio.h"
8 9
10#if defined(__BORLANDC__) || defined(_MSC_VER)
11# define snprintf _snprintf // C compilers have this in stdio.h
12#endif
13
14// some old perls do not have this, try to make it work, no
15// guarentees, though. if it breaks, you get to keep the pieces.
16#ifndef UTF8_MAXBYTES
17# define UTF8_MAXBYTES 13
18#endif
19
9#define F_ASCII 0x00000001 20#define F_ASCII 0x00000001UL
21#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 22#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 23#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 24#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 25#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 26#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040
16#define F_ALLOW_NONREF 0x00000080 27#define F_ALLOW_NONREF 0x00000100UL
28#define F_SHRINK 0x00000200UL
29#define F_ALLOW_BLESSED 0x00000400UL
30#define F_CONV_BLESSED 0x00000800UL
31#define F_MAXDEPTH 0xf8000000UL
32#define S_MAXDEPTH 27
33#define F_MAXSIZE 0x01f00000UL
34#define S_MAXSIZE 20
35#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
36
37#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
38#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17 39
18#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 40#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19#define F_DEFAULT 0 41#define F_DEFAULT (9UL << S_MAXDEPTH)
20 42
21#define INIT_SIZE 32 // initial scalar size to be allocated 43#define INIT_SIZE 32 // initial scalar size to be allocated
44#define INDENT_STEP 3 // spaces per indentation level
45
46#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22 47
23#define SB do { 48#define SB do {
24#define SE } while (0) 49#define SE } while (0)
25 50
26static HV *json_stash; 51#if __GNUC__ >= 3
52# define expect(expr,value) __builtin_expect ((expr),(value))
53# define inline inline
54#else
55# define expect(expr,value) (expr)
56# define inline static
57#endif
58
59#define expect_false(expr) expect ((expr) != 0, 0)
60#define expect_true(expr) expect ((expr) != 0, 1)
61
62static HV *json_stash, *json_boolean_stash; // JSON::XS::
63static SV *json_true, *json_false;
64
65typedef struct {
66 U32 flags;
67 SV *cb_object, *cb_sk_object;
68} JSON;
69
70/////////////////////////////////////////////////////////////////////////////
71// utility functions
72
73inline void
74shrink (SV *sv)
75{
76 sv_utf8_downgrade (sv, 1);
77 if (SvLEN (sv) > SvCUR (sv) + 1)
78 {
79#ifdef SvPV_shrink_to_cur
80 SvPV_shrink_to_cur (sv);
81#elif defined (SvPV_renew)
82 SvPV_renew (sv, SvCUR (sv) + 1);
83#endif
84 }
85}
86
87// decode an utf-8 character and return it, or (UV)-1 in
88// case of an error.
89// we special-case "safe" characters from U+80 .. U+7FF,
90// but use the very good perl function to parse anything else.
91// note that we never call this function for a ascii codepoints
92inline UV
93decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
94{
95 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
96 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
97 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
98 {
99 *clen = 2;
100 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
101 }
102 else
103 {
104 *clen = (STRLEN)-1;
105 return (UV)-1;
106 }
107}
108
109/////////////////////////////////////////////////////////////////////////////
110// encoder
27 111
28// structure used for encoding JSON 112// structure used for encoding JSON
29typedef struct 113typedef struct
30{ 114{
31 char *cur; 115 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 116 char *end; // SvEND (sv)
34 SV *sv; 117 SV *sv; // result scalar
35 UV flags; 118 JSON json;
36 int max_recurse; 119 U32 indent; // indentation level
37 int indent; 120 U32 maxdepth; // max. indentation/recursion level
38} enc_t; 121} enc_t;
39 122
40// structure used for decoding JSON 123inline void
41typedef struct
42{
43 char *cur;
44 char *end;
45 const char *err;
46 UV flags;
47} dec_t;
48
49static UV *
50SvJSON (SV *sv)
51{
52 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53 croak ("object is not of type JSON::XS");
54
55 return &SvUVX (SvRV (sv));
56}
57
58/////////////////////////////////////////////////////////////////////////////
59
60static void
61need (enc_t *enc, STRLEN len) 124need (enc_t *enc, STRLEN len)
62{ 125{
63 if (enc->cur + len >= enc->end) 126 if (expect_false (enc->cur + len >= enc->end))
64 { 127 {
65 STRLEN cur = enc->cur - SvPVX (enc->sv); 128 STRLEN cur = enc->cur - SvPVX (enc->sv);
66 SvGROW (enc->sv, cur + len + 1); 129 SvGROW (enc->sv, cur + len + 1);
67 enc->cur = SvPVX (enc->sv) + cur; 130 enc->cur = SvPVX (enc->sv) + cur;
68 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 131 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69 } 132 }
70} 133}
71 134
72static void 135inline void
73encode_ch (enc_t *enc, char ch) 136encode_ch (enc_t *enc, char ch)
74{ 137{
75 need (enc, 1); 138 need (enc, 1);
76 *enc->cur++ = ch; 139 *enc->cur++ = ch;
77} 140}
… …
85 148
86 while (str < end) 149 while (str < end)
87 { 150 {
88 unsigned char ch = *(unsigned char *)str; 151 unsigned char ch = *(unsigned char *)str;
89 152
90 if (ch == '"') 153 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91 { 154 {
155 if (expect_false (ch == '"')) // but with slow exceptions
156 {
92 need (enc, len += 1); 157 need (enc, len += 1);
93 *enc->cur++ = '\\'; 158 *enc->cur++ = '\\';
94 *enc->cur++ = '"'; 159 *enc->cur++ = '"';
95 ++str;
96 } 160 }
97 else if (ch == '\\') 161 else if (expect_false (ch == '\\'))
98 { 162 {
99 need (enc, len += 1); 163 need (enc, len += 1);
100 *enc->cur++ = '\\'; 164 *enc->cur++ = '\\';
101 *enc->cur++ = '\\'; 165 *enc->cur++ = '\\';
102 ++str;
103 } 166 }
104 else if (ch >= 0x20 && ch < 0x80) // most common case 167 else
105 {
106 *enc->cur++ = ch; 168 *enc->cur++ = ch;
107 ++str; 169
108 }
109 else if (ch == '\015')
110 {
111 need (enc, len += 1);
112 *enc->cur++ = '\\';
113 *enc->cur++ = 'r';
114 ++str;
115 }
116 else if (ch == '\012')
117 {
118 need (enc, len += 1);
119 *enc->cur++ = '\\';
120 *enc->cur++ = 'n';
121 ++str; 170 ++str;
122 } 171 }
123 else 172 else
124 { 173 {
125 STRLEN clen; 174 switch (ch)
126 UV uch;
127
128 if (is_utf8)
129 { 175 {
130 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 176 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131 if (clen < 0) 177 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
132 croak ("malformed UTF-8 character in string, cannot convert to JSON"); 178 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133 } 179 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134 else 180 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135 {
136 uch = ch;
137 clen = 1;
138 }
139 181
140 if (uch < 0x80 || enc->flags & F_ASCII) 182 default:
141 {
142 if (uch > 0xFFFFUL)
143 { 183 {
184 STRLEN clen;
185 UV uch;
186
187 if (is_utf8)
188 {
189 uch = decode_utf8 (str, end - str, &clen);
190 if (clen == (STRLEN)-1)
191 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
192 }
193 else
194 {
195 uch = ch;
196 clen = 1;
197 }
198
199 if (uch > 0x10FFFFUL)
200 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
201
202 if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
203 {
204 if (uch > 0xFFFFUL)
205 {
144 need (enc, len += 11); 206 need (enc, len += 11);
145 sprintf (enc->cur, "\\u%04x\\u%04x", 207 sprintf (enc->cur, "\\u%04x\\u%04x",
146 (uch - 0x10000) / 0x400 + 0xD800, 208 (int)((uch - 0x10000) / 0x400 + 0xD800),
147 (uch - 0x10000) % 0x400 + 0xDC00); 209 (int)((uch - 0x10000) % 0x400 + 0xDC00));
148 enc->cur += 12; 210 enc->cur += 12;
211 }
212 else
213 {
214 static char hexdigit [16] = "0123456789abcdef";
215 need (enc, len += 5);
216 *enc->cur++ = '\\';
217 *enc->cur++ = 'u';
218 *enc->cur++ = hexdigit [ uch >> 12 ];
219 *enc->cur++ = hexdigit [(uch >> 8) & 15];
220 *enc->cur++ = hexdigit [(uch >> 4) & 15];
221 *enc->cur++ = hexdigit [(uch >> 0) & 15];
222 }
223
224 str += clen;
225 }
226 else if (enc->json.flags & F_LATIN1)
227 {
228 *enc->cur++ = uch;
229 str += clen;
230 }
231 else if (is_utf8)
232 {
233 need (enc, len += clen);
234 do
235 {
236 *enc->cur++ = *str++;
237 }
238 while (--clen);
239 }
240 else
241 {
242 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
243 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
244 ++str;
245 }
149 } 246 }
150 else
151 {
152 static char hexdigit [16] = "0123456789abcdef";
153 need (enc, len += 5);
154 *enc->cur++ = '\\';
155 *enc->cur++ = 'u';
156 *enc->cur++ = hexdigit [ uch >> 12 ];
157 *enc->cur++ = hexdigit [(uch >> 8) & 15];
158 *enc->cur++ = hexdigit [(uch >> 4) & 15];
159 *enc->cur++ = hexdigit [(uch >> 0) & 15];
160 }
161
162 str += clen;
163 }
164 else if (is_utf8)
165 {
166 need (enc, len += clen);
167 while (clen--)
168 *enc->cur++ = *str++;
169 }
170 else
171 {
172 need (enc, 10); // never more than 11 bytes needed
173 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174 ++str;
175 } 247 }
176 } 248 }
177 249
178 --len; 250 --len;
179 } 251 }
180} 252}
181 253
182#define INDENT SB \ 254inline void
255encode_indent (enc_t *enc)
256{
183 if (enc->flags & F_INDENT) \ 257 if (enc->json.flags & F_INDENT)
184 { \ 258 {
185 int i_; \ 259 int spaces = enc->indent * INDENT_STEP;
186 need (enc, enc->indent); \ 260
187 for (i_ = enc->indent * 3; i_--; )\ 261 need (enc, spaces);
262 memset (enc->cur, ' ', spaces);
263 enc->cur += spaces;
264 }
265}
266
267inline void
268encode_space (enc_t *enc)
269{
270 need (enc, 1);
271 encode_ch (enc, ' ');
272}
273
274inline void
275encode_nl (enc_t *enc)
276{
277 if (enc->json.flags & F_INDENT)
278 {
279 need (enc, 1);
188 encode_ch (enc, ' '); \ 280 encode_ch (enc, '\n');
189 } \ 281 }
190 SE 282}
191 283
192#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 284inline void
193#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 285encode_comma (enc_t *enc)
194#define COMMA SB \ 286{
195 encode_ch (enc, ','); \ 287 encode_ch (enc, ',');
288
196 if (enc->flags & F_INDENT) \ 289 if (enc->json.flags & F_INDENT)
197 NL; \ 290 encode_nl (enc);
198 else if (enc->flags & F_SPACE_AFTER) \ 291 else if (enc->json.flags & F_SPACE_AFTER)
199 SPACE; \ 292 encode_space (enc);
200 SE 293}
201 294
202static void encode_sv (enc_t *enc, SV *sv); 295static void encode_sv (enc_t *enc, SV *sv);
203 296
204static void 297static void
205encode_av (enc_t *enc, AV *av) 298encode_av (enc_t *enc, AV *av)
206{ 299{
207 int i, len = av_len (av); 300 int i, len = av_len (av);
208 301
209 encode_ch (enc, '['); NL; 302 if (enc->indent >= enc->maxdepth)
303 croak ("data structure too deep (hit recursion limit)");
304
305 encode_ch (enc, '['); encode_nl (enc);
210 ++enc->indent; 306 ++enc->indent;
211 307
212 for (i = 0; i <= len; ++i) 308 for (i = 0; i <= len; ++i)
213 { 309 {
214 INDENT; 310 encode_indent (enc);
215 encode_sv (enc, *av_fetch (av, i, 0)); 311 encode_sv (enc, *av_fetch (av, i, 0));
216 312
217 if (i < len) 313 if (i < len)
218 COMMA; 314 encode_comma (enc);
219 } 315 }
220 316
221 NL; 317 encode_nl (enc);
222 318
223 --enc->indent; 319 --enc->indent;
224 INDENT; encode_ch (enc, ']'); 320 encode_indent (enc); encode_ch (enc, ']');
225} 321}
226 322
227static void 323static void
228encode_he (enc_t *enc, HE *he) 324encode_he (enc_t *enc, HE *he)
229{ 325{
… …
243 else 339 else
244 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 340 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245 341
246 encode_ch (enc, '"'); 342 encode_ch (enc, '"');
247 343
248 if (enc->flags & F_SPACE_BEFORE) SPACE; 344 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249 encode_ch (enc, ':'); 345 encode_ch (enc, ':');
250 if (enc->flags & F_SPACE_AFTER ) SPACE; 346 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251 encode_sv (enc, HeVAL (he)); 347 encode_sv (enc, HeVAL (he));
252} 348}
253 349
254// compare hash entries, used when all keys are bytestrings 350// compare hash entries, used when all keys are bytestrings
255static int 351static int
… …
261 HE *b = *(HE **)b_; 357 HE *b = *(HE **)b_;
262 358
263 STRLEN la = HeKLEN (a); 359 STRLEN la = HeKLEN (a);
264 STRLEN lb = HeKLEN (b); 360 STRLEN lb = HeKLEN (b);
265 361
266 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 362 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
267 cmp = la < lb ? -1 : la == lb ? 0 : 1; 363 cmp = la - lb;
268 364
269 return cmp; 365 return cmp;
270} 366}
271 367
272// compare hash entries, used when some keys are sv's or utf-x 368// compare hash entries, used when some keys are sv's or utf-x
… …
279static void 375static void
280encode_hv (enc_t *enc, HV *hv) 376encode_hv (enc_t *enc, HV *hv)
281{ 377{
282 int count, i; 378 int count, i;
283 379
380 if (enc->indent >= enc->maxdepth)
381 croak ("data structure too deep (hit recursion limit)");
382
284 encode_ch (enc, '{'); NL; ++enc->indent; 383 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
285 384
286 if ((count = hv_iterinit (hv))) 385 if ((count = hv_iterinit (hv)))
287 { 386 {
288 // for canonical output we have to sort by keys first 387 // for canonical output we have to sort by keys first
289 // actually, this is mostly due to the stupid so-called 388 // actually, this is mostly due to the stupid so-called
290 // security workaround added somewhere in 5.8.x. 389 // security workaround added somewhere in 5.8.x.
291 // that randomises hash orderings 390 // that randomises hash orderings
292 if (enc->flags & F_CANONICAL) 391 if (enc->json.flags & F_CANONICAL)
293 { 392 {
294 HE *he, *hes [count];
295 int fast = 1; 393 int fast = 1;
394 HE *he;
395#if defined(__BORLANDC__) || defined(_MSC_VER)
396 HE **hes = _alloca (count * sizeof (HE));
397#else
398 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
399#endif
296 400
297 i = 0; 401 i = 0;
298 while ((he = hv_iternext (hv))) 402 while ((he = hv_iternext (hv)))
299 { 403 {
300 hes [i++] = he; 404 hes [i++] = he;
… …
306 410
307 if (fast) 411 if (fast)
308 qsort (hes, count, sizeof (HE *), he_cmp_fast); 412 qsort (hes, count, sizeof (HE *), he_cmp_fast);
309 else 413 else
310 { 414 {
311 // hack to disable "use bytes" 415 // hack to forcefully disable "use bytes"
312 COP *oldcop = PL_curcop, cop; 416 COP cop = *PL_curcop;
313 cop.op_private = 0; 417 cop.op_private = 0;
418
419 ENTER;
420 SAVETMPS;
421
422 SAVEVPTR (PL_curcop);
314 PL_curcop = &cop; 423 PL_curcop = &cop;
315 424
316 SAVETMPS;
317 qsort (hes, count, sizeof (HE *), he_cmp_slow); 425 qsort (hes, count, sizeof (HE *), he_cmp_slow);
426
318 FREETMPS; 427 FREETMPS;
319 428 LEAVE;
320 PL_curcop = oldcop;
321 } 429 }
322 430
323 for (i = 0; i < count; ++i) 431 for (i = 0; i < count; ++i)
324 { 432 {
325 INDENT; 433 encode_indent (enc);
326 encode_he (enc, hes [i]); 434 encode_he (enc, hes [i]);
327 435
328 if (i < count - 1) 436 if (i < count - 1)
329 COMMA; 437 encode_comma (enc);
330 } 438 }
331 439
332 NL; 440 encode_nl (enc);
333 } 441 }
334 else 442 else
335 { 443 {
336 SV *sv;
337 HE *he = hv_iternext (hv); 444 HE *he = hv_iternext (hv);
338 445
339 for (;;) 446 for (;;)
340 { 447 {
341 INDENT; 448 encode_indent (enc);
342 encode_he (enc, he); 449 encode_he (enc, he);
343 450
344 if (!(he = hv_iternext (hv))) 451 if (!(he = hv_iternext (hv)))
345 break; 452 break;
346 453
347 COMMA; 454 encode_comma (enc);
348 } 455 }
349 456
350 NL; 457 encode_nl (enc);
351 } 458 }
352 } 459 }
353 460
354 --enc->indent; INDENT; encode_ch (enc, '}'); 461 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
462}
463
464// encode objects, arrays and special \0=false and \1=true values.
465static void
466encode_rv (enc_t *enc, SV *sv)
467{
468 svtype svt;
469
470 SvGETMAGIC (sv);
471 svt = SvTYPE (sv);
472
473 if (expect_false (SvOBJECT (sv)))
474 {
475 if (SvSTASH (sv) == json_boolean_stash)
476 {
477 if (SvIV (sv) == 0)
478 encode_str (enc, "false", 5, 0);
479 else
480 encode_str (enc, "true", 4, 0);
481 }
482 else
483 {
484#if 0
485 if (0 && sv_derived_from (rv, "JSON::Literal"))
486 {
487 // not yet
488 }
489#endif
490 if (enc->json.flags & F_CONV_BLESSED)
491 {
492 // we re-bless the reference to get overload and other niceties right
493 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
494
495 if (to_json)
496 {
497 dSP; ENTER; SAVETMPS; PUSHMARK (SP);
498 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
499
500 // calling with G_SCALAR ensures that we always get a 1 reutrn value
501 // check anyways.
502 PUTBACK;
503 assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
504 SPAGAIN;
505
506 encode_sv (enc, POPs);
507
508 FREETMPS; LEAVE;
509 }
510 else if (enc->json.flags & F_ALLOW_BLESSED)
511 encode_str (enc, "null", 4, 0);
512 else
513 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
514 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
515 }
516 else if (enc->json.flags & F_ALLOW_BLESSED)
517 encode_str (enc, "null", 4, 0);
518 else
519 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
520 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
521 }
522 }
523 else if (svt == SVt_PVHV)
524 encode_hv (enc, (HV *)sv);
525 else if (svt == SVt_PVAV)
526 encode_av (enc, (AV *)sv);
527 else if (svt < SVt_PVAV)
528 {
529 if (SvNIOK (sv) && SvIV (sv) == 0)
530 encode_str (enc, "false", 5, 0);
531 else if (SvNIOK (sv) && SvIV (sv) == 1)
532 encode_str (enc, "true", 4, 0);
533 else
534 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
535 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
536 }
537 else
538 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
539 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355} 540}
356 541
357static void 542static void
358encode_sv (enc_t *enc, SV *sv) 543encode_sv (enc_t *enc, SV *sv)
359{ 544{
… …
367 encode_str (enc, str, len, SvUTF8 (sv)); 552 encode_str (enc, str, len, SvUTF8 (sv));
368 encode_ch (enc, '"'); 553 encode_ch (enc, '"');
369 } 554 }
370 else if (SvNOKp (sv)) 555 else if (SvNOKp (sv))
371 { 556 {
557 // trust that perl will do the right thing w.r.t. JSON syntax.
372 need (enc, NV_DIG + 32); 558 need (enc, NV_DIG + 32);
373 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 559 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374 enc->cur += strlen (enc->cur); 560 enc->cur += strlen (enc->cur);
375 } 561 }
376 else if (SvIOKp (sv)) 562 else if (SvIOKp (sv))
377 { 563 {
378 need (enc, 64); 564 // we assume we can always read an IV as a UV
565 if (SvUV (sv) & ~(UV)0x7fff)
566 {
567 // large integer, use the (rather slow) snprintf way.
568 need (enc, sizeof (UV) * 3);
379 enc->cur += 569 enc->cur +=
380 SvIsUV(sv) 570 SvIsUV(sv)
381 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 571 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
382 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 572 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
573 }
574 else
575 {
576 // optimise the "small number case"
577 // code will likely be branchless and use only a single multiplication
578 I32 i = SvIV (sv);
579 U32 u;
580 char digit, nz = 0;
581
582 need (enc, 6);
583
584 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
585 u = i < 0 ? -i : i;
586
587 // convert to 4.28 fixed-point representation
588 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
589
590 // now output digit by digit, each time masking out the integer part
591 // and multiplying by 5 while moving the decimal point one to the right,
592 // resulting in a net multiplication by 10.
593 // we always write the digit to memory but conditionally increment
594 // the pointer, to ease the usage of conditional move instructions.
595 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
596 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
597 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
598 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
599 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
600 }
383 } 601 }
384 else if (SvROK (sv)) 602 else if (SvROK (sv))
385 { 603 encode_rv (enc, SvRV (sv));
386 if (!--enc->max_recurse)
387 croak ("data structure too deep (hit recursion limit)");
388
389 sv = SvRV (sv);
390
391 switch (SvTYPE (sv))
392 {
393 case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394 case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396 default:
397 croak ("JSON can only represent references to arrays or hashes");
398 }
399 }
400 else if (!SvOK (sv)) 604 else if (!SvOK (sv))
401 encode_str (enc, "null", 4, 0); 605 encode_str (enc, "null", 4, 0);
402 else 606 else
403 croak ("encountered perl type that JSON cannot handle"); 607 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
608 SvPV_nolen (sv), SvFLAGS (sv));
404} 609}
405 610
406static SV * 611static SV *
407encode_json (SV *scalar, UV flags) 612encode_json (SV *scalar, JSON *json)
408{ 613{
409 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410 croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
411
412 enc_t enc; 614 enc_t enc;
413 enc.flags = flags; 615
616 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
617 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
618
619 enc.json = *json;
414 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 620 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415 enc.cur = SvPVX (enc.sv); 621 enc.cur = SvPVX (enc.sv);
416 enc.end = SvEND (enc.sv); 622 enc.end = SvEND (enc.sv);
417 enc.max_recurse = 0;
418 enc.indent = 0; 623 enc.indent = 0;
624 enc.maxdepth = DEC_DEPTH (enc.json.flags);
419 625
420 SvPOK_only (enc.sv); 626 SvPOK_only (enc.sv);
421 encode_sv (&enc, scalar); 627 encode_sv (&enc, scalar);
422 628
629 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
630 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
631
423 if (!(flags & (F_ASCII | F_UTF8))) 632 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424 SvUTF8_on (enc.sv); 633 SvUTF8_on (enc.sv);
425 634
426 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 635 if (enc.json.flags & F_SHRINK)
636 shrink (enc.sv);
637
427 return enc.sv; 638 return enc.sv;
428} 639}
429 640
430///////////////////////////////////////////////////////////////////////////// 641/////////////////////////////////////////////////////////////////////////////
642// decoder
431 643
432#define WS \ 644// structure used for decoding JSON
645typedef struct
646{
647 char *cur; // current parser pointer
648 char *end; // end of input string
649 const char *err; // parse error, if != 0
650 JSON json;
651 U32 depth; // recursion depth
652 U32 maxdepth; // recursion depth limit
653} dec_t;
654
655inline void
656decode_ws (dec_t *dec)
657{
433 for (;;) \ 658 for (;;)
434 { \ 659 {
435 char ch = *dec->cur; \ 660 char ch = *dec->cur;
661
436 if (ch > 0x20 \ 662 if (ch > 0x20
437 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 663 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
438 break; \ 664 break;
665
439 ++dec->cur; \ 666 ++dec->cur;
440 } 667 }
668}
441 669
442#define ERR(reason) SB dec->err = reason; goto fail; SE 670#define ERR(reason) SB dec->err = reason; goto fail; SE
671
443#define EXPECT_CH(ch) SB \ 672#define EXPECT_CH(ch) SB \
444 if (*dec->cur != ch) \ 673 if (*dec->cur != ch) \
445 ERR (# ch " expected"); \ 674 ERR (# ch " expected"); \
446 ++dec->cur; \ 675 ++dec->cur; \
447 SE 676 SE
448 677
678#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
679#define DEC_DEC_DEPTH --dec->depth
680
449static SV *decode_sv (dec_t *dec); 681static SV *decode_sv (dec_t *dec);
450 682
451static signed char decode_hexdigit[256]; 683static signed char decode_hexdigit[256];
452 684
453static UV 685static UV
454decode_4hex (dec_t *dec) 686decode_4hex (dec_t *dec)
455{ 687{
456 signed char d1, d2, d3, d4; 688 signed char d1, d2, d3, d4;
689 unsigned char *cur = (unsigned char *)dec->cur;
457 690
458 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 691 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
459 if (d1 < 0) ERR ("four hexadecimal digits expected"); 692 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
460 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 693 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
461 if (d2 < 0) ERR ("four hexadecimal digits expected"); 694 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
462 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463 if (d3 < 0) ERR ("four hexadecimal digits expected");
464 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465 if (d4 < 0) ERR ("four hexadecimal digits expected");
466 695
467 dec->cur += 4; 696 dec->cur += 4;
468 697
469 return ((UV)d1) << 12 698 return ((UV)d1) << 12
470 | ((UV)d2) << 8 699 | ((UV)d2) << 8
… …
473 702
474fail: 703fail:
475 return (UV)-1; 704 return (UV)-1;
476} 705}
477 706
478#define APPEND_GROW(n) SB \
479 if (cur + (n) >= end) \
480 { \
481 STRLEN ofs = cur - SvPVX (sv); \
482 SvGROW (sv, ofs + (n) + 1); \
483 cur = SvPVX (sv) + ofs; \
484 end = SvEND (sv); \
485 } \
486 SE
487
488#define APPEND_CH(ch) SB \
489 APPEND_GROW (1); \
490 *cur++ = (ch); \
491 SE
492
493static SV * 707static SV *
494decode_str (dec_t *dec) 708decode_str (dec_t *dec)
495{ 709{
496 SV *sv = NEWSV (0,2); 710 SV *sv = 0;
497 int utf8 = 0; 711 int utf8 = 0;
498 char *cur = SvPVX (sv); 712 char *dec_cur = dec->cur;
499 char *end = SvEND (sv);
500 713
501 for (;;) 714 do
502 { 715 {
503 unsigned char ch = *(unsigned char *)dec->cur; 716 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
717 char *cur = buf;
504 718
505 if (ch == '"') 719 do
506 break;
507 else if (ch == '\\')
508 { 720 {
509 switch (*++dec->cur) 721 unsigned char ch = *(unsigned char *)dec_cur++;
722
723 if (expect_false (ch == '"'))
510 { 724 {
511 case '\\': 725 --dec_cur;
512 case '/': 726 break;
513 case '"': APPEND_CH (*dec->cur++); break; 727 }
514 728 else if (expect_false (ch == '\\'))
515 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 729 {
516 case 't': APPEND_CH ('\011'); ++dec->cur; break; 730 switch (*dec_cur)
517 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521 case 'u':
522 { 731 {
523 UV lo, hi; 732 case '\\':
524 ++dec->cur; 733 case '/':
734 case '"': *cur++ = *dec_cur++; break;
525 735
526 hi = decode_4hex (dec); 736 case 'b': ++dec_cur; *cur++ = '\010'; break;
527 if (hi == (UV)-1) 737 case 't': ++dec_cur; *cur++ = '\011'; break;
528 goto fail; 738 case 'n': ++dec_cur; *cur++ = '\012'; break;
739 case 'f': ++dec_cur; *cur++ = '\014'; break;
740 case 'r': ++dec_cur; *cur++ = '\015'; break;
529 741
530 // possibly a surrogate pair 742 case 'u':
531 if (hi >= 0xd800 && hi < 0xdc00)
532 { 743 {
533 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 744 UV lo, hi;
534 ERR ("illegal surrogate character"); 745 ++dec_cur;
535 746
536 dec->cur += 2; 747 dec->cur = dec_cur;
537
538 lo = decode_4hex (dec); 748 hi = decode_4hex (dec);
749 dec_cur = dec->cur;
539 if (lo == (UV)-1) 750 if (hi == (UV)-1)
540 goto fail; 751 goto fail;
541 752
753 // possibly a surrogate pair
754 if (hi >= 0xd800)
755 if (hi < 0xdc00)
756 {
757 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
758 ERR ("missing low surrogate character in surrogate pair");
759
760 dec_cur += 2;
761
762 dec->cur = dec_cur;
763 lo = decode_4hex (dec);
764 dec_cur = dec->cur;
765 if (lo == (UV)-1)
766 goto fail;
767
542 if (lo < 0xdc00 || lo >= 0xe000) 768 if (lo < 0xdc00 || lo >= 0xe000)
543 ERR ("surrogate pair expected"); 769 ERR ("surrogate pair expected");
544 770
545 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 771 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
772 }
773 else if (hi < 0xe000)
774 ERR ("missing high surrogate character in surrogate pair");
775
776 if (hi >= 0x80)
777 {
778 utf8 = 1;
779
780 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
781 }
782 else
783 *cur++ = hi;
546 } 784 }
547 else if (lo >= 0xdc00 && lo < 0xe000)
548 ERR ("illegal surrogate character");
549
550 if (hi >= 0x80)
551 { 785 break;
552 utf8 = 1;
553 786
554 APPEND_GROW (4); // at most 4 bytes for 21 bits
555 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556 }
557 else 787 default:
558 APPEND_CH (hi); 788 --dec_cur;
789 ERR ("illegal backslash escape sequence in string");
559 } 790 }
791 }
792 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
793 *cur++ = ch;
794 else if (ch >= 0x80)
795 {
796 STRLEN clen;
797 UV uch;
798
799 --dec_cur;
800
801 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
802 if (clen == (STRLEN)-1)
803 ERR ("malformed UTF-8 character in JSON string");
804
805 do
806 *cur++ = *dec_cur++;
807 while (--clen);
808
809 utf8 = 1;
810 }
811 else
812 {
813 --dec_cur;
814
815 if (!ch)
816 ERR ("unexpected end of string while parsing JSON string");
560 break; 817 else
818 ERR ("invalid character encountered while parsing JSON string");
561 } 819 }
562 } 820 }
563 else if (ch >= 0x20 && ch <= 0x7f) 821 while (cur < buf + SHORT_STRING_LEN);
564 APPEND_CH (*dec->cur++); 822
565 else if (ch >= 0x80) 823 {
824 STRLEN len = cur - buf;
825
826 if (sv)
566 { 827 {
567 STRLEN clen; 828 SvGROW (sv, SvCUR (sv) + len + 1);
568 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 829 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
569 if (clen < 0) 830 SvCUR_set (sv, SvCUR (sv) + len);
570 ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571
572 APPEND_GROW (clen);
573 memcpy (cur, dec->cur, clen);
574 cur += clen;
575 dec->cur += clen;
576 } 831 }
577 else 832 else
578 ERR ("invalid character encountered"); 833 sv = newSVpvn (buf, len);
579 } 834 }
835 }
836 while (*dec_cur != '"');
580 837
581 ++dec->cur; 838 ++dec_cur;
582 839
583 SvCUR_set (sv, cur - SvPVX (sv)); 840 if (sv)
584 841 {
585 SvPOK_only (sv); 842 SvPOK_only (sv);
586 *SvEND (sv) = 0; 843 *SvEND (sv) = 0;
587 844
588 if (utf8) 845 if (utf8)
589 SvUTF8_on (sv); 846 SvUTF8_on (sv);
847 }
848 else
849 sv = newSVpvn ("", 0);
590 850
851 dec->cur = dec_cur;
591 return sv; 852 return sv;
592 853
593fail: 854fail:
594 SvREFCNT_dec (sv); 855 dec->cur = dec_cur;
595 return 0; 856 return 0;
596} 857}
597 858
598static SV * 859static SV *
599decode_num (dec_t *dec) 860decode_num (dec_t *dec)
… …
609 { 870 {
610 ++dec->cur; 871 ++dec->cur;
611 if (*dec->cur >= '0' && *dec->cur <= '9') 872 if (*dec->cur >= '0' && *dec->cur <= '9')
612 ERR ("malformed number (leading zero must not be followed by another digit)"); 873 ERR ("malformed number (leading zero must not be followed by another digit)");
613 } 874 }
614 875 else if (*dec->cur < '0' || *dec->cur > '9')
615 // int 876 ERR ("malformed number (no digits after initial minus)");
877 else
878 do
879 {
880 ++dec->cur;
881 }
616 while (*dec->cur >= '0' && *dec->cur <= '9') 882 while (*dec->cur >= '0' && *dec->cur <= '9');
617 ++dec->cur;
618 883
619 // [frac] 884 // [frac]
620 if (*dec->cur == '.') 885 if (*dec->cur == '.')
621 { 886 {
622 is_nv = 1; 887 ++dec->cur;
888
889 if (*dec->cur < '0' || *dec->cur > '9')
890 ERR ("malformed number (no digits after decimal point)");
623 891
624 do 892 do
625 { 893 {
626 ++dec->cur; 894 ++dec->cur;
627 } 895 }
628 while (*dec->cur >= '0' && *dec->cur <= '9'); 896 while (*dec->cur >= '0' && *dec->cur <= '9');
897
898 is_nv = 1;
629 } 899 }
630 900
631 // [exp] 901 // [exp]
632 if (*dec->cur == 'e' || *dec->cur == 'E') 902 if (*dec->cur == 'e' || *dec->cur == 'E')
633 { 903 {
634 is_nv = 1;
635
636 ++dec->cur; 904 ++dec->cur;
905
637 if (*dec->cur == '-' || *dec->cur == '+') 906 if (*dec->cur == '-' || *dec->cur == '+')
638 ++dec->cur; 907 ++dec->cur;
639 908
909 if (*dec->cur < '0' || *dec->cur > '9')
910 ERR ("malformed number (no digits after exp sign)");
911
912 do
913 {
914 ++dec->cur;
915 }
640 while (*dec->cur >= '0' && *dec->cur <= '9') 916 while (*dec->cur >= '0' && *dec->cur <= '9');
641 ++dec->cur; 917
918 is_nv = 1;
642 } 919 }
643 920
644 if (!is_nv) 921 if (!is_nv)
645 { 922 {
646 UV uv; 923 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
647 int numtype = grok_number (start, dec->cur - start, &uv); 924 if (*start == '-')
648 if (numtype & IS_NUMBER_IN_UV) 925 switch (dec->cur - start)
649 if (numtype & IS_NUMBER_NEG)
650 { 926 {
651 if (uv < (UV)IV_MIN) 927 case 2: return newSViv (-( start [1] - '0' * 1));
652 return newSViv (-(IV)uv); 928 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
929 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
930 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
653 } 931 }
932 else
933 switch (dec->cur - start)
934 {
935 case 1: return newSViv ( start [0] - '0' * 1);
936 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
937 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
938 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
939 }
940
941 {
942 UV uv;
943 int numtype = grok_number (start, dec->cur - start, &uv);
944 if (numtype & IS_NUMBER_IN_UV)
945 if (numtype & IS_NUMBER_NEG)
946 {
947 if (uv < (UV)IV_MIN)
948 return newSViv (-(IV)uv);
949 }
654 else 950 else
655 return newSVuv (uv); 951 return newSVuv (uv);
952
953 // here would likely be the place for bigint support
656 } 954 }
955 }
657 956
957 // if we ever support bigint or bigfloat, this is the place for bigfloat
658 return newSVnv (Atof (start)); 958 return newSVnv (Atof (start));
659 959
660fail: 960fail:
661 return 0; 961 return 0;
662} 962}
… …
664static SV * 964static SV *
665decode_av (dec_t *dec) 965decode_av (dec_t *dec)
666{ 966{
667 AV *av = newAV (); 967 AV *av = newAV ();
668 968
969 DEC_INC_DEPTH;
970 decode_ws (dec);
971
972 if (*dec->cur == ']')
973 ++dec->cur;
974 else
669 for (;;) 975 for (;;)
670 { 976 {
671 SV *value; 977 SV *value;
672 978
673 value = decode_sv (dec); 979 value = decode_sv (dec);
674 if (!value) 980 if (!value)
675 goto fail; 981 goto fail;
676 982
677 av_push (av, value); 983 av_push (av, value);
678 984
679 WS; 985 decode_ws (dec);
680 986
681 if (*dec->cur == ']') 987 if (*dec->cur == ']')
682 { 988 {
683 ++dec->cur; 989 ++dec->cur;
684 break; 990 break;
991 }
685 } 992
686
687 if (*dec->cur != ',') 993 if (*dec->cur != ',')
688 ERR (", or ] expected while parsing array"); 994 ERR (", or ] expected while parsing array");
689 995
690 ++dec->cur; 996 ++dec->cur;
691 } 997 }
692 998
999 DEC_DEC_DEPTH;
693 return newRV_noinc ((SV *)av); 1000 return newRV_noinc ((SV *)av);
694 1001
695fail: 1002fail:
696 SvREFCNT_dec (av); 1003 SvREFCNT_dec (av);
1004 DEC_DEC_DEPTH;
697 return 0; 1005 return 0;
698} 1006}
699 1007
700static SV * 1008static SV *
701decode_hv (dec_t *dec) 1009decode_hv (dec_t *dec)
702{ 1010{
1011 SV *sv;
703 HV *hv = newHV (); 1012 HV *hv = newHV ();
704 1013
1014 DEC_INC_DEPTH;
1015 decode_ws (dec);
1016
1017 if (*dec->cur == '}')
1018 ++dec->cur;
1019 else
705 for (;;) 1020 for (;;)
706 { 1021 {
707 SV *key, *value; 1022 decode_ws (dec); EXPECT_CH ('"');
708 1023
709 WS; EXPECT_CH ('"'); 1024 // heuristic: assume that
710 1025 // a) decode_str + hv_store_ent are abysmally slow.
711 key = decode_str (dec); 1026 // b) most hash keys are short, simple ascii text.
712 if (!key) 1027 // => try to "fast-match" such strings to avoid
713 goto fail; 1028 // the overhead of decode_str + hv_store_ent.
714
715 WS; EXPECT_CH (':');
716
717 value = decode_sv (dec);
718 if (!value)
719 { 1029 {
1030 SV *value;
1031 char *p = dec->cur;
1032 char *e = p + 24; // only try up to 24 bytes
1033
1034 for (;;)
1035 {
1036 // the >= 0x80 is true on most architectures
1037 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1038 {
1039 // slow path, back up and use decode_str
1040 SV *key = decode_str (dec);
1041 if (!key)
1042 goto fail;
1043
1044 decode_ws (dec); EXPECT_CH (':');
1045
1046 value = decode_sv (dec);
1047 if (!value)
1048 {
1049 SvREFCNT_dec (key);
1050 goto fail;
1051 }
1052
1053 hv_store_ent (hv, key, value, 0);
720 SvREFCNT_dec (key); 1054 SvREFCNT_dec (key);
1055
1056 break;
1057 }
1058 else if (*p == '"')
1059 {
1060 // fast path, got a simple key
1061 char *key = dec->cur;
1062 int len = p - key;
1063 dec->cur = p + 1;
1064
1065 decode_ws (dec); EXPECT_CH (':');
1066
1067 value = decode_sv (dec);
1068 if (!value)
721 goto fail; 1069 goto fail;
1070
1071 hv_store (hv, key, len, value, 0);
1072
1073 break;
1074 }
1075
1076 ++p;
1077 }
722 } 1078 }
723 1079
724 //TODO: optimise 1080 decode_ws (dec);
725 hv_store_ent (hv, key, value, 0);
726 1081
727 WS;
728
729 if (*dec->cur == '}') 1082 if (*dec->cur == '}')
1083 {
1084 ++dec->cur;
1085 break;
1086 }
1087
1088 if (*dec->cur != ',')
1089 ERR (", or } expected while parsing object/hash");
1090
1091 ++dec->cur;
1092 }
1093
1094 DEC_DEC_DEPTH;
1095 sv = newRV_noinc ((SV *)hv);
1096
1097 // check filter callbacks
1098 if (dec->json.flags & F_HOOK)
1099 {
1100 ENTER; SAVETMPS;
1101
1102 if (HvKEYS (hv) == 1 && dec->json.cb_sk_object)
730 { 1103 {
731 ++dec->cur; 1104 int count;
732 break; 1105
1106 dSP; PUSHMARK (SP);
1107 XPUSHs (sv_2mortal (sv));
1108
1109 PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN;
1110
1111 if (count == 1)
1112 sv = newSVsv (POPs);
1113 else
1114 SvREFCNT_inc (sv);
733 } 1115 }
734 1116
735 if (*dec->cur != ',') 1117 if (dec->json.cb_object)
736 ERR (", or } expected while parsing object/hash"); 1118 {
1119 int count;
737 1120
738 ++dec->cur; 1121 dSP; ENTER; SAVETMPS; PUSHMARK (SP);
1122 XPUSHs (sv_2mortal (sv));
1123
1124 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1125
1126 if (count == 1)
1127 sv = newSVsv (POPs);
1128 else
1129 SvREFCNT_inc (sv);
1130 }
1131
1132 FREETMPS; LEAVE;
739 } 1133 }
740 1134
741 return newRV_noinc ((SV *)hv); 1135 return newRV_noinc ((SV *)hv);
742 1136
743fail: 1137fail:
744 SvREFCNT_dec (hv); 1138 SvREFCNT_dec (hv);
1139 DEC_DEC_DEPTH;
745 return 0; 1140 return 0;
746} 1141}
747 1142
748static SV * 1143static SV *
749decode_sv (dec_t *dec) 1144decode_sv (dec_t *dec)
750{ 1145{
751 WS; 1146 decode_ws (dec);
1147
1148 // the beauty of JSON: you need exactly one character lookahead
1149 // to parse anything.
752 switch (*dec->cur) 1150 switch (*dec->cur)
753 { 1151 {
754 case '"': ++dec->cur; return decode_str (dec); 1152 case '"': ++dec->cur; return decode_str (dec);
755 case '[': ++dec->cur; return decode_av (dec); 1153 case '[': ++dec->cur; return decode_av (dec);
756 case '{': ++dec->cur; return decode_hv (dec); 1154 case '{': ++dec->cur; return decode_hv (dec);
… …
762 1160
763 case 't': 1161 case 't':
764 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1162 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765 { 1163 {
766 dec->cur += 4; 1164 dec->cur += 4;
767 return newSViv (1); 1165 return SvREFCNT_inc (json_true);
768 } 1166 }
769 else 1167 else
770 ERR ("'true' expected"); 1168 ERR ("'true' expected");
771 1169
772 break; 1170 break;
773 1171
774 case 'f': 1172 case 'f':
775 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1173 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776 { 1174 {
777 dec->cur += 5; 1175 dec->cur += 5;
778 return newSViv (0); 1176 return SvREFCNT_inc (json_false);
779 } 1177 }
780 else 1178 else
781 ERR ("'false' expected"); 1179 ERR ("'false' expected");
782 1180
783 break; 1181 break;
784 1182
785 case 'n': 1183 case 'n':
786 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1184 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787 { 1185 {
788 dec->cur += 4; 1186 dec->cur += 4;
789 return newSViv (1); 1187 return newSVsv (&PL_sv_undef);
790 } 1188 }
791 else 1189 else
792 ERR ("'null' expected"); 1190 ERR ("'null' expected");
793 1191
794 break; 1192 break;
795 1193
796 default: 1194 default:
797 ERR ("malformed json string"); 1195 ERR ("malformed JSON string, neither array, object, number, string or atom");
798 break; 1196 break;
799 } 1197 }
800 1198
801fail: 1199fail:
802 return 0; 1200 return 0;
803} 1201}
804 1202
805static SV * 1203static SV *
806decode_json (SV *string, UV flags) 1204decode_json (SV *string, JSON *json, UV *offset_return)
807{ 1205{
1206 dec_t dec;
1207 UV offset;
808 SV *sv; 1208 SV *sv;
809 1209
1210 SvGETMAGIC (string);
1211 SvUPGRADE (string, SVt_PV);
1212
1213 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1214 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1215 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1216
810 if (!(flags & F_UTF8)) 1217 if (json->flags & F_UTF8)
1218 sv_utf8_downgrade (string, 0);
1219 else
811 sv_utf8_upgrade (string); 1220 sv_utf8_upgrade (string);
812 1221
813 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1222 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814 1223
815 dec_t dec; 1224 dec.json = *json;
816 dec.flags = flags;
817 dec.cur = SvPVX (string); 1225 dec.cur = SvPVX (string);
818 dec.end = SvEND (string); 1226 dec.end = SvEND (string);
819 dec.err = 0; 1227 dec.err = 0;
1228 dec.depth = 0;
1229 dec.maxdepth = DEC_DEPTH (dec.json.flags);
820 1230
821 *dec.end = 1; // invalid anywhere 1231 if (dec.json.cb_object || dec.json.cb_sk_object)
1232 dec.json.flags |= F_HOOK;
1233
1234 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
822 sv = decode_sv (&dec); 1235 sv = decode_sv (&dec);
823 *dec.end = 0; 1236
1237 if (!(offset_return || !sv))
1238 {
1239 // check for trailing garbage
1240 decode_ws (&dec);
1241
1242 if (*dec.cur)
1243 {
1244 dec.err = "garbage after JSON object";
1245 SvREFCNT_dec (sv);
1246 sv = 0;
1247 }
1248 }
1249
1250 if (offset_return || !sv)
1251 {
1252 offset = dec.json.flags & F_UTF8
1253 ? dec.cur - SvPVX (string)
1254 : utf8_distance (dec.cur, SvPVX (string));
1255
1256 if (offset_return)
1257 *offset_return = offset;
1258 }
824 1259
825 if (!sv) 1260 if (!sv)
826 { 1261 {
827 IV offset = utf8_distance (dec.cur, SvPVX (string));
828 SV *uni = sv_newmortal (); 1262 SV *uni = sv_newmortal ();
829 1263
1264 // horrible hack to silence warning inside pv_uni_display
1265 COP cop = *PL_curcop;
1266 cop.cop_warnings = pWARN_NONE;
1267 ENTER;
1268 SAVEVPTR (PL_curcop);
1269 PL_curcop = &cop;
830 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1270 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1271 LEAVE;
1272
831 croak ("%s, at character %d (%s)", 1273 croak ("%s, at character offset %d [\"%s\"]",
832 dec.err, 1274 dec.err,
833 (int)offset, 1275 (int)offset,
834 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1276 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835 } 1277 }
836 1278
837 sv = sv_2mortal (sv); 1279 sv = sv_2mortal (sv);
838 1280
839 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1281 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
840 croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); 1282 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841 1283
842 return sv; 1284 return sv;
843} 1285}
844 1286
1287/////////////////////////////////////////////////////////////////////////////
1288// XS interface functions
1289
845MODULE = JSON::XS PACKAGE = JSON::XS 1290MODULE = JSON::XS PACKAGE = JSON::XS
846 1291
847BOOT: 1292BOOT:
848{ 1293{
849 int i; 1294 int i;
850 1295
851 memset (decode_hexdigit, 0xff, 256);
852 for (i = 10; i--; ) 1296 for (i = 0; i < 256; ++i)
853 decode_hexdigit ['0' + i] = i; 1297 decode_hexdigit [i] =
1298 i >= '0' && i <= '9' ? i - '0'
1299 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1300 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1301 : -1;
854 1302
855 for (i = 7; i--; ) 1303 json_stash = gv_stashpv ("JSON::XS" , 1);
1304 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1305
1306 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1307 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1308}
1309
1310PROTOTYPES: DISABLE
1311
1312void new (char *klass)
1313 PPCODE:
1314{
1315 SV *pv = NEWSV (0, sizeof (JSON));
1316 SvPOK_only (pv);
1317 Zero (SvPVX (pv), 1, JSON);
1318 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1319 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
1320}
1321
1322void ascii (JSON *self, int enable = 1)
1323 ALIAS:
1324 ascii = F_ASCII
1325 latin1 = F_LATIN1
1326 utf8 = F_UTF8
1327 indent = F_INDENT
1328 canonical = F_CANONICAL
1329 space_before = F_SPACE_BEFORE
1330 space_after = F_SPACE_AFTER
1331 pretty = F_PRETTY
1332 allow_nonref = F_ALLOW_NONREF
1333 shrink = F_SHRINK
1334 allow_blessed = F_ALLOW_BLESSED
1335 convert_blessed = F_CONV_BLESSED
1336 PPCODE:
1337{
1338 if (enable)
1339 self->flags |= ix;
1340 else
1341 self->flags &= ~ix;
1342
1343 XPUSHs (ST (0));
1344}
1345
1346void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1347 PPCODE:
1348{
1349 UV log2 = 0;
1350
1351 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1352
1353 while ((1UL << log2) < max_depth)
1354 ++log2;
1355
1356 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1357
1358 XPUSHs (ST (0));
1359}
1360
1361void max_size (JSON *self, UV max_size = 0)
1362 PPCODE:
1363{
1364 UV log2 = 0;
1365
1366 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1367 if (max_size == 1) max_size = 2;
1368
1369 while ((1UL << log2) < max_size)
1370 ++log2;
1371
1372 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1373
1374 XPUSHs (ST (0));
1375}
1376
1377void filter_json_objects (JSON *self, SV *cb = &PL_sv_undef)
1378 ALIAS:
1379 filter_sk_json_objects = 1
1380 PPCODE:
1381{
1382 if (!SvOK (cb))
1383 cb = 0;
1384
1385 switch (ix)
856 { 1386 {
857 decode_hexdigit ['a' + i] = 10 + i; 1387 case 0: self->cb_object = cb; break;
858 decode_hexdigit ['A' + i] = 10 + i; 1388 case 1: self->cb_sk_object = cb; break;
859 } 1389 }
860 1390
861 json_stash = gv_stashpv ("JSON::XS", 1); 1391 XPUSHs (ST (0));
862} 1392}
863 1393
864PROTOTYPES: DISABLE
865
866SV *new (char *dummy)
867 CODE:
868 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
869 OUTPUT:
870 RETVAL
871
872SV *ascii (SV *self, int enable)
873 ALIAS:
874 ascii = F_ASCII
875 utf8 = F_UTF8
876 indent = F_INDENT
877 canonical = F_CANONICAL
878 space_before = F_SPACE_BEFORE
879 space_after = F_SPACE_AFTER
880 json_rpc = F_JSON_RPC
881 pretty = F_PRETTY
882 allow_nonref = F_ALLOW_NONREF
883 CODE:
884{
885 UV *uv = SvJSON (self);
886 if (enable)
887 *uv |= ix;
888 else
889 *uv &= ~ix;
890
891 RETVAL = newSVsv (self);
892}
893 OUTPUT:
894 RETVAL
895
896void encode (SV *self, SV *scalar) 1394void encode (JSON *self, SV *scalar)
897 PPCODE: 1395 PPCODE:
898 XPUSHs (encode_json (scalar, *SvJSON (self))); 1396 XPUSHs (encode_json (scalar, self));
899 1397
900void decode (SV *self, SV *jsonstr) 1398void decode (JSON *self, SV *jsonstr)
901 PPCODE: 1399 PPCODE:
902 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1400 XPUSHs (decode_json (jsonstr, self, 0));
1401
1402void decode_prefix (JSON *self, SV *jsonstr)
1403 PPCODE:
1404{
1405 UV offset;
1406 EXTEND (SP, 2);
1407 PUSHs (decode_json (jsonstr, self, &offset));
1408 PUSHs (sv_2mortal (newSVuv (offset)));
1409}
903 1410
904PROTOTYPES: ENABLE 1411PROTOTYPES: ENABLE
905 1412
906void to_json (SV *scalar) 1413void to_json (SV *scalar)
907 PPCODE: 1414 PPCODE:
1415{
1416 JSON json = { F_DEFAULT | F_UTF8 };
908 XPUSHs (encode_json (scalar, F_UTF8)); 1417 XPUSHs (encode_json (scalar, &json));
1418}
909 1419
910void from_json (SV *jsonstr) 1420void from_json (SV *jsonstr)
911 PPCODE: 1421 PPCODE:
1422{
1423 JSON json = { F_DEFAULT | F_UTF8 };
912 XPUSHs (decode_json (jsonstr, F_UTF8)); 1424 XPUSHs (decode_json (jsonstr, &json, 0));
1425}
913 1426

Diff Legend

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