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 1.42 by root, Sat Jun 23 22:56:08 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_MAXDEPTH 0xf8000000UL
30#define S_MAXDEPTH 27
31
32#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
33
34// F_SELFCONVERT? <=> to_json/toJson
35// F_BLESSED? <=> { $__class__$ => }
17 36
18#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 37#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19#define F_DEFAULT 0 38#define F_DEFAULT (9UL << S_MAXDEPTH)
20 39
21#define INIT_SIZE 32 // initial scalar size to be allocated 40#define INIT_SIZE 32 // initial scalar size to be allocated
41#define INDENT_STEP 3 // spaces per indentation level
42
43#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22 44
23#define SB do { 45#define SB do {
24#define SE } while (0) 46#define SE } while (0)
25 47
48#if __GNUC__ >= 3
49# define expect(expr,value) __builtin_expect ((expr),(value))
50# define inline inline
51#else
52# define expect(expr,value) (expr)
53# define inline static
54#endif
55
56#define expect_false(expr) expect ((expr) != 0, 0)
57#define expect_true(expr) expect ((expr) != 0, 1)
58
26static HV *json_stash; 59static HV *json_stash; // JSON::XS::
60
61/////////////////////////////////////////////////////////////////////////////
62// utility functions
63
64static UV *
65SvJSON (SV *sv)
66{
67 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
68 croak ("object is not of type JSON::XS");
69
70 return &SvUVX (SvRV (sv));
71}
72
73static 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 U32 flags; // F_*
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->flags & F_ASCII || (enc->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->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->flags & F_INDENT)
184 { \ 258 {
185 int i_; \ 259 int spaces = enc->indent * INDENT_STEP;
186 need (enc, enc->indent); \
187 for (i_ = enc->indent * 3; i_--; )\
188 encode_ch (enc, ' '); \
189 } \
190 SE
191 260
192#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 261 need (enc, spaces);
193#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 262 memset (enc->cur, ' ', spaces);
194#define COMMA SB \ 263 enc->cur += spaces;
264 }
265}
266
267inline void
268encode_space (enc_t *enc)
269{
270 need (enc, 1);
195 encode_ch (enc, ','); \ 271 encode_ch (enc, ' ');
272}
273
274inline void
275encode_nl (enc_t *enc)
276{
196 if (enc->flags & F_INDENT) \ 277 if (enc->flags & F_INDENT)
197 NL; \ 278 {
279 need (enc, 1);
280 encode_ch (enc, '\n');
281 }
282}
283
284inline void
285encode_comma (enc_t *enc)
286{
287 encode_ch (enc, ',');
288
289 if (enc->flags & F_INDENT)
290 encode_nl (enc);
198 else if (enc->flags & F_SPACE_AFTER) \ 291 else if (enc->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->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->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->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 (svt == SVt_PVHV)
474 encode_hv (enc, (HV *)sv);
475 else if (svt == SVt_PVAV)
476 encode_av (enc, (AV *)sv);
477 else if (svt < SVt_PVAV)
478 {
479 if (SvNIOK (sv) && SvIV (sv) == 0)
480 encode_str (enc, "false", 5, 0);
481 else if (SvNIOK (sv) && SvIV (sv) == 1)
482 encode_str (enc, "true", 4, 0);
483 else
484 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
485 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
486 }
487 else
488 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
489 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355} 490}
356 491
357static void 492static void
358encode_sv (enc_t *enc, SV *sv) 493encode_sv (enc_t *enc, SV *sv)
359{ 494{
… …
367 encode_str (enc, str, len, SvUTF8 (sv)); 502 encode_str (enc, str, len, SvUTF8 (sv));
368 encode_ch (enc, '"'); 503 encode_ch (enc, '"');
369 } 504 }
370 else if (SvNOKp (sv)) 505 else if (SvNOKp (sv))
371 { 506 {
507 // trust that perl will do the right thing w.r.t. JSON syntax.
372 need (enc, NV_DIG + 32); 508 need (enc, NV_DIG + 32);
373 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 509 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374 enc->cur += strlen (enc->cur); 510 enc->cur += strlen (enc->cur);
375 } 511 }
376 else if (SvIOKp (sv)) 512 else if (SvIOKp (sv))
377 { 513 {
378 need (enc, 64); 514 // we assume we can always read an IV as a UV
515 if (SvUV (sv) & ~(UV)0x7fff)
516 {
517 // large integer, use the (rather slow) snprintf way.
518 need (enc, sizeof (UV) * 3);
379 enc->cur += 519 enc->cur +=
380 SvIsUV(sv) 520 SvIsUV(sv)
381 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 521 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
382 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 522 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
523 }
524 else
525 {
526 // optimise the "small number case"
527 // code will likely be branchless and use only a single multiplication
528 I32 i = SvIV (sv);
529 U32 u;
530 char digit, nz = 0;
531
532 need (enc, 6);
533
534 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
535 u = i < 0 ? -i : i;
536
537 // convert to 4.28 fixed-point representation
538 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
539
540 // now output digit by digit, each time masking out the integer part
541 // and multiplying by 5 while moving the decimal point one to the right,
542 // resulting in a net multiplication by 10.
543 // we always write the digit to memory but conditionally increment
544 // the pointer, to ease the usage of conditional move instructions.
545 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
546 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
547 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
548 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
549 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
550 }
383 } 551 }
384 else if (SvROK (sv)) 552 else if (SvROK (sv))
385 { 553 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)) 554 else if (!SvOK (sv))
401 encode_str (enc, "null", 4, 0); 555 encode_str (enc, "null", 4, 0);
402 else 556 else
403 croak ("encountered perl type that JSON cannot handle"); 557 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
558 SvPV_nolen (sv), SvFLAGS (sv));
404} 559}
405 560
406static SV * 561static SV *
407encode_json (SV *scalar, UV flags) 562encode_json (SV *scalar, U32 flags)
408{ 563{
564 enc_t enc;
565
409 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) 566 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410 croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)"); 567 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
411 568
412 enc_t enc;
413 enc.flags = flags; 569 enc.flags = flags;
414 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 570 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415 enc.cur = SvPVX (enc.sv); 571 enc.cur = SvPVX (enc.sv);
416 enc.end = SvEND (enc.sv); 572 enc.end = SvEND (enc.sv);
417 enc.max_recurse = 0;
418 enc.indent = 0; 573 enc.indent = 0;
574 enc.maxdepth = DEC_DEPTH (flags);
419 575
420 SvPOK_only (enc.sv); 576 SvPOK_only (enc.sv);
421 encode_sv (&enc, scalar); 577 encode_sv (&enc, scalar);
422 578
579 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
580 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
581
423 if (!(flags & (F_ASCII | F_UTF8))) 582 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424 SvUTF8_on (enc.sv); 583 SvUTF8_on (enc.sv);
425 584
426 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 585 if (enc.flags & F_SHRINK)
586 shrink (enc.sv);
587
427 return enc.sv; 588 return enc.sv;
428} 589}
429 590
430///////////////////////////////////////////////////////////////////////////// 591/////////////////////////////////////////////////////////////////////////////
592// decoder
431 593
432#define WS \ 594// structure used for decoding JSON
595typedef struct
596{
597 char *cur; // current parser pointer
598 char *end; // end of input string
599 const char *err; // parse error, if != 0
600 U32 flags; // F_*
601 U32 depth; // recursion depth
602 U32 maxdepth; // recursion depth limit
603} dec_t;
604
605inline void
606decode_ws (dec_t *dec)
607{
433 for (;;) \ 608 for (;;)
434 { \ 609 {
435 char ch = *dec->cur; \ 610 char ch = *dec->cur;
611
436 if (ch > 0x20 \ 612 if (ch > 0x20
437 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 613 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
438 break; \ 614 break;
615
439 ++dec->cur; \ 616 ++dec->cur;
440 } 617 }
618}
441 619
442#define ERR(reason) SB dec->err = reason; goto fail; SE 620#define ERR(reason) SB dec->err = reason; goto fail; SE
621
443#define EXPECT_CH(ch) SB \ 622#define EXPECT_CH(ch) SB \
444 if (*dec->cur != ch) \ 623 if (*dec->cur != ch) \
445 ERR (# ch " expected"); \ 624 ERR (# ch " expected"); \
446 ++dec->cur; \ 625 ++dec->cur; \
447 SE 626 SE
448 627
628#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
629#define DEC_DEC_DEPTH --dec->depth
630
449static SV *decode_sv (dec_t *dec); 631static SV *decode_sv (dec_t *dec);
450 632
451static signed char decode_hexdigit[256]; 633static signed char decode_hexdigit[256];
452 634
453static UV 635static UV
454decode_4hex (dec_t *dec) 636decode_4hex (dec_t *dec)
455{ 637{
456 signed char d1, d2, d3, d4; 638 signed char d1, d2, d3, d4;
639 unsigned char *cur = (unsigned char *)dec->cur;
457 640
458 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 641 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"); 642 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]]; 643 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"); 644 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");
466 645
467 dec->cur += 4; 646 dec->cur += 4;
468 647
469 return ((UV)d1) << 12 648 return ((UV)d1) << 12
470 | ((UV)d2) << 8 649 | ((UV)d2) << 8
… …
473 652
474fail: 653fail:
475 return (UV)-1; 654 return (UV)-1;
476} 655}
477 656
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 * 657static SV *
494decode_str (dec_t *dec) 658decode_str (dec_t *dec)
495{ 659{
496 SV *sv = NEWSV (0,2); 660 SV *sv = 0;
497 int utf8 = 0; 661 int utf8 = 0;
498 char *cur = SvPVX (sv); 662 char *dec_cur = dec->cur;
499 char *end = SvEND (sv);
500 663
501 for (;;) 664 do
502 { 665 {
503 unsigned char ch = *(unsigned char *)dec->cur; 666 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
667 char *cur = buf;
504 668
505 if (ch == '"') 669 do
506 break;
507 else if (ch == '\\')
508 { 670 {
509 switch (*++dec->cur) 671 unsigned char ch = *(unsigned char *)dec_cur++;
672
673 if (expect_false (ch == '"'))
510 { 674 {
511 case '\\': 675 --dec_cur;
512 case '/': 676 break;
513 case '"': APPEND_CH (*dec->cur++); break; 677 }
514 678 else if (expect_false (ch == '\\'))
515 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 679 {
516 case 't': APPEND_CH ('\011'); ++dec->cur; break; 680 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 { 681 {
523 UV lo, hi; 682 case '\\':
524 ++dec->cur; 683 case '/':
684 case '"': *cur++ = *dec_cur++; break;
525 685
526 hi = decode_4hex (dec); 686 case 'b': ++dec_cur; *cur++ = '\010'; break;
527 if (hi == (UV)-1) 687 case 't': ++dec_cur; *cur++ = '\011'; break;
528 goto fail; 688 case 'n': ++dec_cur; *cur++ = '\012'; break;
689 case 'f': ++dec_cur; *cur++ = '\014'; break;
690 case 'r': ++dec_cur; *cur++ = '\015'; break;
529 691
530 // possibly a surrogate pair 692 case 'u':
531 if (hi >= 0xd800 && hi < 0xdc00)
532 { 693 {
533 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 694 UV lo, hi;
534 ERR ("illegal surrogate character"); 695 ++dec_cur;
535 696
536 dec->cur += 2; 697 dec->cur = dec_cur;
537
538 lo = decode_4hex (dec); 698 hi = decode_4hex (dec);
699 dec_cur = dec->cur;
539 if (lo == (UV)-1) 700 if (hi == (UV)-1)
540 goto fail; 701 goto fail;
541 702
703 // possibly a surrogate pair
704 if (hi >= 0xd800)
705 if (hi < 0xdc00)
706 {
707 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
708 ERR ("missing low surrogate character in surrogate pair");
709
710 dec_cur += 2;
711
712 dec->cur = dec_cur;
713 lo = decode_4hex (dec);
714 dec_cur = dec->cur;
715 if (lo == (UV)-1)
716 goto fail;
717
542 if (lo < 0xdc00 || lo >= 0xe000) 718 if (lo < 0xdc00 || lo >= 0xe000)
543 ERR ("surrogate pair expected"); 719 ERR ("surrogate pair expected");
544 720
545 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 721 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
722 }
723 else if (hi < 0xe000)
724 ERR ("missing high surrogate character in surrogate pair");
725
726 if (hi >= 0x80)
727 {
728 utf8 = 1;
729
730 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
731 }
732 else
733 *cur++ = hi;
546 } 734 }
547 else if (lo >= 0xdc00 && lo < 0xe000)
548 ERR ("illegal surrogate character");
549
550 if (hi >= 0x80)
551 { 735 break;
552 utf8 = 1;
553 736
554 APPEND_GROW (4); // at most 4 bytes for 21 bits
555 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556 }
557 else 737 default:
558 APPEND_CH (hi); 738 --dec_cur;
739 ERR ("illegal backslash escape sequence in string");
559 } 740 }
741 }
742 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
743 *cur++ = ch;
744 else if (ch >= 0x80)
745 {
746 STRLEN clen;
747 UV uch;
748
749 --dec_cur;
750
751 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
752 if (clen == (STRLEN)-1)
753 ERR ("malformed UTF-8 character in JSON string");
754
755 do
756 *cur++ = *dec_cur++;
757 while (--clen);
758
759 utf8 = 1;
760 }
761 else
762 {
763 --dec_cur;
764
765 if (!ch)
766 ERR ("unexpected end of string while parsing JSON string");
560 break; 767 else
768 ERR ("invalid character encountered while parsing JSON string");
561 } 769 }
562 } 770 }
563 else if (ch >= 0x20 && ch <= 0x7f) 771 while (cur < buf + SHORT_STRING_LEN);
564 APPEND_CH (*dec->cur++); 772
565 else if (ch >= 0x80) 773 {
774 STRLEN len = cur - buf;
775
776 if (sv)
566 { 777 {
567 STRLEN clen; 778 SvGROW (sv, SvCUR (sv) + len + 1);
568 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 779 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
569 if (clen < 0) 780 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 } 781 }
577 else 782 else
578 ERR ("invalid character encountered"); 783 sv = newSVpvn (buf, len);
579 } 784 }
785 }
786 while (*dec_cur != '"');
580 787
581 ++dec->cur; 788 ++dec_cur;
582 789
583 SvCUR_set (sv, cur - SvPVX (sv)); 790 if (sv)
584 791 {
585 SvPOK_only (sv); 792 SvPOK_only (sv);
586 *SvEND (sv) = 0; 793 *SvEND (sv) = 0;
587 794
588 if (utf8) 795 if (utf8)
589 SvUTF8_on (sv); 796 SvUTF8_on (sv);
797 }
798 else
799 sv = newSVpvn ("", 0);
590 800
801 dec->cur = dec_cur;
591 return sv; 802 return sv;
592 803
593fail: 804fail:
594 SvREFCNT_dec (sv); 805 dec->cur = dec_cur;
595 return 0; 806 return 0;
596} 807}
597 808
598static SV * 809static SV *
599decode_num (dec_t *dec) 810decode_num (dec_t *dec)
… …
609 { 820 {
610 ++dec->cur; 821 ++dec->cur;
611 if (*dec->cur >= '0' && *dec->cur <= '9') 822 if (*dec->cur >= '0' && *dec->cur <= '9')
612 ERR ("malformed number (leading zero must not be followed by another digit)"); 823 ERR ("malformed number (leading zero must not be followed by another digit)");
613 } 824 }
614 825 else if (*dec->cur < '0' || *dec->cur > '9')
615 // int 826 ERR ("malformed number (no digits after initial minus)");
827 else
828 do
829 {
830 ++dec->cur;
831 }
616 while (*dec->cur >= '0' && *dec->cur <= '9') 832 while (*dec->cur >= '0' && *dec->cur <= '9');
617 ++dec->cur;
618 833
619 // [frac] 834 // [frac]
620 if (*dec->cur == '.') 835 if (*dec->cur == '.')
621 { 836 {
622 is_nv = 1; 837 ++dec->cur;
838
839 if (*dec->cur < '0' || *dec->cur > '9')
840 ERR ("malformed number (no digits after decimal point)");
623 841
624 do 842 do
625 { 843 {
626 ++dec->cur; 844 ++dec->cur;
627 } 845 }
628 while (*dec->cur >= '0' && *dec->cur <= '9'); 846 while (*dec->cur >= '0' && *dec->cur <= '9');
847
848 is_nv = 1;
629 } 849 }
630 850
631 // [exp] 851 // [exp]
632 if (*dec->cur == 'e' || *dec->cur == 'E') 852 if (*dec->cur == 'e' || *dec->cur == 'E')
633 { 853 {
634 is_nv = 1;
635
636 ++dec->cur; 854 ++dec->cur;
855
637 if (*dec->cur == '-' || *dec->cur == '+') 856 if (*dec->cur == '-' || *dec->cur == '+')
638 ++dec->cur; 857 ++dec->cur;
639 858
859 if (*dec->cur < '0' || *dec->cur > '9')
860 ERR ("malformed number (no digits after exp sign)");
861
862 do
863 {
864 ++dec->cur;
865 }
640 while (*dec->cur >= '0' && *dec->cur <= '9') 866 while (*dec->cur >= '0' && *dec->cur <= '9');
641 ++dec->cur; 867
868 is_nv = 1;
642 } 869 }
643 870
644 if (!is_nv) 871 if (!is_nv)
645 { 872 {
646 UV uv; 873 // 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); 874 if (*start == '-')
648 if (numtype & IS_NUMBER_IN_UV) 875 switch (dec->cur - start)
649 if (numtype & IS_NUMBER_NEG)
650 { 876 {
651 if (uv < (UV)IV_MIN) 877 case 2: return newSViv (-( start [1] - '0' * 1));
652 return newSViv (-(IV)uv); 878 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
879 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
880 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
653 } 881 }
882 else
883 switch (dec->cur - start)
884 {
885 case 1: return newSViv ( start [0] - '0' * 1);
886 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
887 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
888 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
889 }
890
891 {
892 UV uv;
893 int numtype = grok_number (start, dec->cur - start, &uv);
894 if (numtype & IS_NUMBER_IN_UV)
895 if (numtype & IS_NUMBER_NEG)
896 {
897 if (uv < (UV)IV_MIN)
898 return newSViv (-(IV)uv);
899 }
654 else 900 else
655 return newSVuv (uv); 901 return newSVuv (uv);
902
903 // here would likely be the place for bigint support
656 } 904 }
905 }
657 906
907 // if we ever support bigint or bigfloat, this is the place for bigfloat
658 return newSVnv (Atof (start)); 908 return newSVnv (Atof (start));
659 909
660fail: 910fail:
661 return 0; 911 return 0;
662} 912}
… …
664static SV * 914static SV *
665decode_av (dec_t *dec) 915decode_av (dec_t *dec)
666{ 916{
667 AV *av = newAV (); 917 AV *av = newAV ();
668 918
919 DEC_INC_DEPTH;
920 decode_ws (dec);
921
922 if (*dec->cur == ']')
923 ++dec->cur;
924 else
669 for (;;) 925 for (;;)
670 { 926 {
671 SV *value; 927 SV *value;
672 928
673 value = decode_sv (dec); 929 value = decode_sv (dec);
674 if (!value) 930 if (!value)
675 goto fail; 931 goto fail;
676 932
677 av_push (av, value); 933 av_push (av, value);
678 934
679 WS; 935 decode_ws (dec);
680 936
681 if (*dec->cur == ']') 937 if (*dec->cur == ']')
682 { 938 {
683 ++dec->cur; 939 ++dec->cur;
684 break; 940 break;
941 }
685 } 942
686
687 if (*dec->cur != ',') 943 if (*dec->cur != ',')
688 ERR (", or ] expected while parsing array"); 944 ERR (", or ] expected while parsing array");
689 945
690 ++dec->cur; 946 ++dec->cur;
691 } 947 }
692 948
949 DEC_DEC_DEPTH;
693 return newRV_noinc ((SV *)av); 950 return newRV_noinc ((SV *)av);
694 951
695fail: 952fail:
696 SvREFCNT_dec (av); 953 SvREFCNT_dec (av);
954 DEC_DEC_DEPTH;
697 return 0; 955 return 0;
698} 956}
699 957
700static SV * 958static SV *
701decode_hv (dec_t *dec) 959decode_hv (dec_t *dec)
702{ 960{
703 HV *hv = newHV (); 961 HV *hv = newHV ();
704 962
963 DEC_INC_DEPTH;
964 decode_ws (dec);
965
966 if (*dec->cur == '}')
967 ++dec->cur;
968 else
705 for (;;) 969 for (;;)
706 { 970 {
707 SV *key, *value; 971 SV *key, *value;
708 972
709 WS; EXPECT_CH ('"'); 973 decode_ws (dec); EXPECT_CH ('"');
710 974
711 key = decode_str (dec); 975 key = decode_str (dec);
712 if (!key) 976 if (!key)
713 goto fail;
714
715 WS; EXPECT_CH (':');
716
717 value = decode_sv (dec);
718 if (!value)
719 {
720 SvREFCNT_dec (key);
721 goto fail; 977 goto fail;
978
979 decode_ws (dec); EXPECT_CH (':');
980
981 value = decode_sv (dec);
982 if (!value)
983 {
984 SvREFCNT_dec (key);
985 goto fail;
722 } 986 }
723 987
724 //TODO: optimise
725 hv_store_ent (hv, key, value, 0); 988 hv_store_ent (hv, key, value, 0);
989 SvREFCNT_dec (key);
726 990
727 WS; 991 decode_ws (dec);
728 992
729 if (*dec->cur == '}') 993 if (*dec->cur == '}')
730 { 994 {
731 ++dec->cur; 995 ++dec->cur;
732 break; 996 break;
733 } 997 }
734 998
735 if (*dec->cur != ',') 999 if (*dec->cur != ',')
736 ERR (", or } expected while parsing object/hash"); 1000 ERR (", or } expected while parsing object/hash");
737 1001
738 ++dec->cur; 1002 ++dec->cur;
739 } 1003 }
740 1004
1005 DEC_DEC_DEPTH;
741 return newRV_noinc ((SV *)hv); 1006 return newRV_noinc ((SV *)hv);
742 1007
743fail: 1008fail:
744 SvREFCNT_dec (hv); 1009 SvREFCNT_dec (hv);
1010 DEC_DEC_DEPTH;
745 return 0; 1011 return 0;
746} 1012}
747 1013
748static SV * 1014static SV *
749decode_sv (dec_t *dec) 1015decode_sv (dec_t *dec)
750{ 1016{
751 WS; 1017 decode_ws (dec);
1018
1019 // the beauty of JSON: you need exactly one character lookahead
1020 // to parse anything.
752 switch (*dec->cur) 1021 switch (*dec->cur)
753 { 1022 {
754 case '"': ++dec->cur; return decode_str (dec); 1023 case '"': ++dec->cur; return decode_str (dec);
755 case '[': ++dec->cur; return decode_av (dec); 1024 case '[': ++dec->cur; return decode_av (dec);
756 case '{': ++dec->cur; return decode_hv (dec); 1025 case '{': ++dec->cur; return decode_hv (dec);
… …
784 1053
785 case 'n': 1054 case 'n':
786 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1055 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787 { 1056 {
788 dec->cur += 4; 1057 dec->cur += 4;
789 return newSViv (1); 1058 return newSVsv (&PL_sv_undef);
790 } 1059 }
791 else 1060 else
792 ERR ("'null' expected"); 1061 ERR ("'null' expected");
793 1062
794 break; 1063 break;
795 1064
796 default: 1065 default:
797 ERR ("malformed json string"); 1066 ERR ("malformed JSON string, neither array, object, number, string or atom");
798 break; 1067 break;
799 } 1068 }
800 1069
801fail: 1070fail:
802 return 0; 1071 return 0;
803} 1072}
804 1073
805static SV * 1074static SV *
806decode_json (SV *string, UV flags) 1075decode_json (SV *string, U32 flags, UV *offset_return)
807{ 1076{
1077 dec_t dec;
1078 UV offset;
808 SV *sv; 1079 SV *sv;
809 1080
1081 SvGETMAGIC (string);
1082 SvUPGRADE (string, SVt_PV);
1083
810 if (!(flags & F_UTF8)) 1084 if (flags & F_UTF8)
1085 sv_utf8_downgrade (string, 0);
1086 else
811 sv_utf8_upgrade (string); 1087 sv_utf8_upgrade (string);
812 1088
813 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1089 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814 1090
815 dec_t dec;
816 dec.flags = flags; 1091 dec.flags = flags;
817 dec.cur = SvPVX (string); 1092 dec.cur = SvPVX (string);
818 dec.end = SvEND (string); 1093 dec.end = SvEND (string);
819 dec.err = 0; 1094 dec.err = 0;
1095 dec.depth = 0;
1096 dec.maxdepth = DEC_DEPTH (dec.flags);
820 1097
821 *dec.end = 1; // invalid anywhere 1098 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
822 sv = decode_sv (&dec); 1099 sv = decode_sv (&dec);
823 *dec.end = 0; 1100
1101 if (!(offset_return || !sv))
1102 {
1103 // check for trailing garbage
1104 decode_ws (&dec);
1105
1106 if (*dec.cur)
1107 {
1108 dec.err = "garbage after JSON object";
1109 SvREFCNT_dec (sv);
1110 sv = 0;
1111 }
1112 }
1113
1114 if (offset_return || !sv)
1115 {
1116 offset = dec.flags & F_UTF8
1117 ? dec.cur - SvPVX (string)
1118 : utf8_distance (dec.cur, SvPVX (string));
1119
1120 if (offset_return)
1121 *offset_return = offset;
1122 }
824 1123
825 if (!sv) 1124 if (!sv)
826 { 1125 {
827 IV offset = utf8_distance (dec.cur, SvPVX (string));
828 SV *uni = sv_newmortal (); 1126 SV *uni = sv_newmortal ();
829 1127
1128 // horrible hack to silence warning inside pv_uni_display
1129 COP cop = *PL_curcop;
1130 cop.cop_warnings = pWARN_NONE;
1131 ENTER;
1132 SAVEVPTR (PL_curcop);
1133 PL_curcop = &cop;
830 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1134 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1135 LEAVE;
1136
831 croak ("%s, at character %d (%s)", 1137 croak ("%s, at character offset %d [\"%s\"]",
832 dec.err, 1138 dec.err,
833 (int)offset, 1139 (int)offset,
834 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1140 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835 } 1141 }
836 1142
837 sv = sv_2mortal (sv); 1143 sv = sv_2mortal (sv);
838 1144
839 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1145 if (!(dec.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)"); 1146 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841 1147
842 return sv; 1148 return sv;
843} 1149}
844 1150
1151/////////////////////////////////////////////////////////////////////////////
1152// XS interface functions
1153
845MODULE = JSON::XS PACKAGE = JSON::XS 1154MODULE = JSON::XS PACKAGE = JSON::XS
846 1155
847BOOT: 1156BOOT:
848{ 1157{
849 int i; 1158 int i;
850 1159
851 memset (decode_hexdigit, 0xff, 256);
852 for (i = 10; i--; ) 1160 for (i = 0; i < 256; ++i)
853 decode_hexdigit ['0' + i] = i; 1161 decode_hexdigit [i] =
854 1162 i >= '0' && i <= '9' ? i - '0'
855 for (i = 7; i--; ) 1163 : i >= 'a' && i <= 'f' ? i - 'a' + 10
856 { 1164 : i >= 'A' && i <= 'F' ? i - 'A' + 10
857 decode_hexdigit ['a' + i] = 10 + i; 1165 : -1;
858 decode_hexdigit ['A' + i] = 10 + i;
859 }
860 1166
861 json_stash = gv_stashpv ("JSON::XS", 1); 1167 json_stash = gv_stashpv ("JSON::XS", 1);
862} 1168}
863 1169
864PROTOTYPES: DISABLE 1170PROTOTYPES: DISABLE
… …
867 CODE: 1173 CODE:
868 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1174 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
869 OUTPUT: 1175 OUTPUT:
870 RETVAL 1176 RETVAL
871 1177
872SV *ascii (SV *self, int enable) 1178SV *ascii (SV *self, int enable = 1)
873 ALIAS: 1179 ALIAS:
874 ascii = F_ASCII 1180 ascii = F_ASCII
1181 latin1 = F_LATIN1
875 utf8 = F_UTF8 1182 utf8 = F_UTF8
876 indent = F_INDENT 1183 indent = F_INDENT
877 canonical = F_CANONICAL 1184 canonical = F_CANONICAL
878 space_before = F_SPACE_BEFORE 1185 space_before = F_SPACE_BEFORE
879 space_after = F_SPACE_AFTER 1186 space_after = F_SPACE_AFTER
880 json_rpc = F_JSON_RPC
881 pretty = F_PRETTY 1187 pretty = F_PRETTY
882 allow_nonref = F_ALLOW_NONREF 1188 allow_nonref = F_ALLOW_NONREF
1189 shrink = F_SHRINK
883 CODE: 1190 CODE:
884{ 1191{
885 UV *uv = SvJSON (self); 1192 UV *uv = SvJSON (self);
886 if (enable) 1193 if (enable)
887 *uv |= ix; 1194 *uv |= ix;
… …
891 RETVAL = newSVsv (self); 1198 RETVAL = newSVsv (self);
892} 1199}
893 OUTPUT: 1200 OUTPUT:
894 RETVAL 1201 RETVAL
895 1202
1203SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
1204 CODE:
1205{
1206 UV *uv = SvJSON (self);
1207 UV log2 = 0;
1208
1209 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1210
1211 while ((1UL << log2) < max_depth)
1212 ++log2;
1213
1214 *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1215
1216 RETVAL = newSVsv (self);
1217}
1218 OUTPUT:
1219 RETVAL
1220
896void encode (SV *self, SV *scalar) 1221void encode (SV *self, SV *scalar)
897 PPCODE: 1222 PPCODE:
898 XPUSHs (encode_json (scalar, *SvJSON (self))); 1223 XPUSHs (encode_json (scalar, *SvJSON (self)));
899 1224
900void decode (SV *self, SV *jsonstr) 1225void decode (SV *self, SV *jsonstr)
901 PPCODE: 1226 PPCODE:
902 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1227 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1228
1229void decode_prefix (SV *self, SV *jsonstr)
1230 PPCODE:
1231{
1232 UV offset;
1233 EXTEND (SP, 2);
1234 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1235 PUSHs (sv_2mortal (newSVuv (offset)));
1236}
903 1237
904PROTOTYPES: ENABLE 1238PROTOTYPES: ENABLE
905 1239
906void to_json (SV *scalar) 1240void to_json (SV *scalar)
1241 ALIAS:
1242 objToJson = 0
907 PPCODE: 1243 PPCODE:
908 XPUSHs (encode_json (scalar, F_UTF8)); 1244 XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
909 1245
910void from_json (SV *jsonstr) 1246void from_json (SV *jsonstr)
1247 ALIAS:
1248 jsonToObj = 0
911 PPCODE: 1249 PPCODE:
912 XPUSHs (decode_json (jsonstr, F_UTF8)); 1250 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
913 1251

Diff Legend

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