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 1.43 by root, Sat Jun 23 23:49:29 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::
60static SV *json_true, *json_false;
61
62/////////////////////////////////////////////////////////////////////////////
63// utility functions
64
65static UV *
66SvJSON (SV *sv)
67{
68 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
69 croak ("object is not of type JSON::XS");
70
71 return &SvUVX (SvRV (sv));
72}
73
74static void
75shrink (SV *sv)
76{
77 sv_utf8_downgrade (sv, 1);
78 if (SvLEN (sv) > SvCUR (sv) + 1)
79 {
80#ifdef SvPV_shrink_to_cur
81 SvPV_shrink_to_cur (sv);
82#elif defined (SvPV_renew)
83 SvPV_renew (sv, SvCUR (sv) + 1);
84#endif
85 }
86}
87
88// decode an utf-8 character and return it, or (UV)-1 in
89// case of an error.
90// we special-case "safe" characters from U+80 .. U+7FF,
91// but use the very good perl function to parse anything else.
92// note that we never call this function for a ascii codepoints
93inline UV
94decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
95{
96 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
97 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
98 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
99 {
100 *clen = 2;
101 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
102 }
103 else
104 {
105 *clen = (STRLEN)-1;
106 return (UV)-1;
107 }
108}
109
110/////////////////////////////////////////////////////////////////////////////
111// encoder
27 112
28// structure used for encoding JSON 113// structure used for encoding JSON
29typedef struct 114typedef struct
30{ 115{
31 char *cur; 116 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 117 char *end; // SvEND (sv)
34 SV *sv; 118 SV *sv; // result scalar
35 UV flags; 119 U32 flags; // F_*
36 int max_recurse; 120 U32 indent; // indentation level
37 int indent; 121 U32 maxdepth; // max. indentation/recursion level
38} enc_t; 122} enc_t;
39 123
40// structure used for decoding JSON 124inline 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) 125need (enc_t *enc, STRLEN len)
62{ 126{
63 if (enc->cur + len >= enc->end) 127 if (expect_false (enc->cur + len >= enc->end))
64 { 128 {
65 STRLEN cur = enc->cur - SvPVX (enc->sv); 129 STRLEN cur = enc->cur - SvPVX (enc->sv);
66 SvGROW (enc->sv, cur + len + 1); 130 SvGROW (enc->sv, cur + len + 1);
67 enc->cur = SvPVX (enc->sv) + cur; 131 enc->cur = SvPVX (enc->sv) + cur;
68 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 132 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69 } 133 }
70} 134}
71 135
72static void 136inline void
73encode_ch (enc_t *enc, char ch) 137encode_ch (enc_t *enc, char ch)
74{ 138{
75 need (enc, 1); 139 need (enc, 1);
76 *enc->cur++ = ch; 140 *enc->cur++ = ch;
77} 141}
… …
85 149
86 while (str < end) 150 while (str < end)
87 { 151 {
88 unsigned char ch = *(unsigned char *)str; 152 unsigned char ch = *(unsigned char *)str;
89 153
90 if (ch == '"') 154 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91 { 155 {
156 if (expect_false (ch == '"')) // but with slow exceptions
157 {
92 need (enc, len += 1); 158 need (enc, len += 1);
93 *enc->cur++ = '\\'; 159 *enc->cur++ = '\\';
94 *enc->cur++ = '"'; 160 *enc->cur++ = '"';
95 ++str;
96 } 161 }
97 else if (ch == '\\') 162 else if (expect_false (ch == '\\'))
98 { 163 {
99 need (enc, len += 1); 164 need (enc, len += 1);
100 *enc->cur++ = '\\'; 165 *enc->cur++ = '\\';
101 *enc->cur++ = '\\'; 166 *enc->cur++ = '\\';
102 ++str;
103 } 167 }
104 else if (ch >= 0x20 && ch < 0x80) // most common case 168 else
105 {
106 *enc->cur++ = ch; 169 *enc->cur++ = ch;
107 ++str; 170
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; 171 ++str;
122 } 172 }
123 else 173 else
124 { 174 {
125 STRLEN clen; 175 switch (ch)
126 UV uch;
127
128 if (is_utf8)
129 { 176 {
130 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 177 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131 if (clen < 0) 178 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"); 179 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133 } 180 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134 else 181 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135 {
136 uch = ch;
137 clen = 1;
138 }
139 182
140 if (uch < 0x80 || enc->flags & F_ASCII) 183 default:
141 {
142 if (uch > 0xFFFFUL)
143 { 184 {
185 STRLEN clen;
186 UV uch;
187
188 if (is_utf8)
189 {
190 uch = decode_utf8 (str, end - str, &clen);
191 if (clen == (STRLEN)-1)
192 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
193 }
194 else
195 {
196 uch = ch;
197 clen = 1;
198 }
199
200 if (uch > 0x10FFFFUL)
201 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
202
203 if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
204 {
205 if (uch > 0xFFFFUL)
206 {
144 need (enc, len += 11); 207 need (enc, len += 11);
145 sprintf (enc->cur, "\\u%04x\\u%04x", 208 sprintf (enc->cur, "\\u%04x\\u%04x",
146 (uch - 0x10000) / 0x400 + 0xD800, 209 (int)((uch - 0x10000) / 0x400 + 0xD800),
147 (uch - 0x10000) % 0x400 + 0xDC00); 210 (int)((uch - 0x10000) % 0x400 + 0xDC00));
148 enc->cur += 12; 211 enc->cur += 12;
212 }
213 else
214 {
215 static char hexdigit [16] = "0123456789abcdef";
216 need (enc, len += 5);
217 *enc->cur++ = '\\';
218 *enc->cur++ = 'u';
219 *enc->cur++ = hexdigit [ uch >> 12 ];
220 *enc->cur++ = hexdigit [(uch >> 8) & 15];
221 *enc->cur++ = hexdigit [(uch >> 4) & 15];
222 *enc->cur++ = hexdigit [(uch >> 0) & 15];
223 }
224
225 str += clen;
226 }
227 else if (enc->flags & F_LATIN1)
228 {
229 *enc->cur++ = uch;
230 str += clen;
231 }
232 else if (is_utf8)
233 {
234 need (enc, len += clen);
235 do
236 {
237 *enc->cur++ = *str++;
238 }
239 while (--clen);
240 }
241 else
242 {
243 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
244 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
245 ++str;
246 }
149 } 247 }
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 } 248 }
176 } 249 }
177 250
178 --len; 251 --len;
179 } 252 }
180} 253}
181 254
182#define INDENT SB \ 255inline void
256encode_indent (enc_t *enc)
257{
183 if (enc->flags & F_INDENT) \ 258 if (enc->flags & F_INDENT)
184 { \ 259 {
185 int i_; \ 260 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
191 261
192#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 262 need (enc, spaces);
193#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 263 memset (enc->cur, ' ', spaces);
194#define COMMA SB \ 264 enc->cur += spaces;
265 }
266}
267
268inline void
269encode_space (enc_t *enc)
270{
271 need (enc, 1);
195 encode_ch (enc, ','); \ 272 encode_ch (enc, ' ');
273}
274
275inline void
276encode_nl (enc_t *enc)
277{
196 if (enc->flags & F_INDENT) \ 278 if (enc->flags & F_INDENT)
197 NL; \ 279 {
280 need (enc, 1);
281 encode_ch (enc, '\n');
282 }
283}
284
285inline void
286encode_comma (enc_t *enc)
287{
288 encode_ch (enc, ',');
289
290 if (enc->flags & F_INDENT)
291 encode_nl (enc);
198 else if (enc->flags & F_SPACE_AFTER) \ 292 else if (enc->flags & F_SPACE_AFTER)
199 SPACE; \ 293 encode_space (enc);
200 SE 294}
201 295
202static void encode_sv (enc_t *enc, SV *sv); 296static void encode_sv (enc_t *enc, SV *sv);
203 297
204static void 298static void
205encode_av (enc_t *enc, AV *av) 299encode_av (enc_t *enc, AV *av)
206{ 300{
207 int i, len = av_len (av); 301 int i, len = av_len (av);
208 302
209 encode_ch (enc, '['); NL; 303 if (enc->indent >= enc->maxdepth)
304 croak ("data structure too deep (hit recursion limit)");
305
306 encode_ch (enc, '['); encode_nl (enc);
210 ++enc->indent; 307 ++enc->indent;
211 308
212 for (i = 0; i <= len; ++i) 309 for (i = 0; i <= len; ++i)
213 { 310 {
214 INDENT; 311 encode_indent (enc);
215 encode_sv (enc, *av_fetch (av, i, 0)); 312 encode_sv (enc, *av_fetch (av, i, 0));
216 313
217 if (i < len) 314 if (i < len)
218 COMMA; 315 encode_comma (enc);
219 } 316 }
220 317
221 NL; 318 encode_nl (enc);
222 319
223 --enc->indent; 320 --enc->indent;
224 INDENT; encode_ch (enc, ']'); 321 encode_indent (enc); encode_ch (enc, ']');
225} 322}
226 323
227static void 324static void
228encode_he (enc_t *enc, HE *he) 325encode_he (enc_t *enc, HE *he)
229{ 326{
… …
243 else 340 else
244 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 341 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245 342
246 encode_ch (enc, '"'); 343 encode_ch (enc, '"');
247 344
248 if (enc->flags & F_SPACE_BEFORE) SPACE; 345 if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
249 encode_ch (enc, ':'); 346 encode_ch (enc, ':');
250 if (enc->flags & F_SPACE_AFTER ) SPACE; 347 if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
251 encode_sv (enc, HeVAL (he)); 348 encode_sv (enc, HeVAL (he));
252} 349}
253 350
254// compare hash entries, used when all keys are bytestrings 351// compare hash entries, used when all keys are bytestrings
255static int 352static int
… …
261 HE *b = *(HE **)b_; 358 HE *b = *(HE **)b_;
262 359
263 STRLEN la = HeKLEN (a); 360 STRLEN la = HeKLEN (a);
264 STRLEN lb = HeKLEN (b); 361 STRLEN lb = HeKLEN (b);
265 362
266 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 363 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
267 cmp = la < lb ? -1 : la == lb ? 0 : 1; 364 cmp = la - lb;
268 365
269 return cmp; 366 return cmp;
270} 367}
271 368
272// compare hash entries, used when some keys are sv's or utf-x 369// compare hash entries, used when some keys are sv's or utf-x
… …
279static void 376static void
280encode_hv (enc_t *enc, HV *hv) 377encode_hv (enc_t *enc, HV *hv)
281{ 378{
282 int count, i; 379 int count, i;
283 380
381 if (enc->indent >= enc->maxdepth)
382 croak ("data structure too deep (hit recursion limit)");
383
284 encode_ch (enc, '{'); NL; ++enc->indent; 384 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
285 385
286 if ((count = hv_iterinit (hv))) 386 if ((count = hv_iterinit (hv)))
287 { 387 {
288 // for canonical output we have to sort by keys first 388 // for canonical output we have to sort by keys first
289 // actually, this is mostly due to the stupid so-called 389 // actually, this is mostly due to the stupid so-called
290 // security workaround added somewhere in 5.8.x. 390 // security workaround added somewhere in 5.8.x.
291 // that randomises hash orderings 391 // that randomises hash orderings
292 if (enc->flags & F_CANONICAL) 392 if (enc->flags & F_CANONICAL)
293 { 393 {
294 HE *he, *hes [count];
295 int fast = 1; 394 int fast = 1;
395 HE *he;
396#if defined(__BORLANDC__) || defined(_MSC_VER)
397 HE **hes = _alloca (count * sizeof (HE));
398#else
399 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
400#endif
296 401
297 i = 0; 402 i = 0;
298 while ((he = hv_iternext (hv))) 403 while ((he = hv_iternext (hv)))
299 { 404 {
300 hes [i++] = he; 405 hes [i++] = he;
… …
306 411
307 if (fast) 412 if (fast)
308 qsort (hes, count, sizeof (HE *), he_cmp_fast); 413 qsort (hes, count, sizeof (HE *), he_cmp_fast);
309 else 414 else
310 { 415 {
311 // hack to disable "use bytes" 416 // hack to forcefully disable "use bytes"
312 COP *oldcop = PL_curcop, cop; 417 COP cop = *PL_curcop;
313 cop.op_private = 0; 418 cop.op_private = 0;
419
420 ENTER;
421 SAVETMPS;
422
423 SAVEVPTR (PL_curcop);
314 PL_curcop = &cop; 424 PL_curcop = &cop;
315 425
316 SAVETMPS;
317 qsort (hes, count, sizeof (HE *), he_cmp_slow); 426 qsort (hes, count, sizeof (HE *), he_cmp_slow);
427
318 FREETMPS; 428 FREETMPS;
319 429 LEAVE;
320 PL_curcop = oldcop;
321 } 430 }
322 431
323 for (i = 0; i < count; ++i) 432 for (i = 0; i < count; ++i)
324 { 433 {
325 INDENT; 434 encode_indent (enc);
326 encode_he (enc, hes [i]); 435 encode_he (enc, hes [i]);
327 436
328 if (i < count - 1) 437 if (i < count - 1)
329 COMMA; 438 encode_comma (enc);
330 } 439 }
331 440
332 NL; 441 encode_nl (enc);
333 } 442 }
334 else 443 else
335 { 444 {
336 SV *sv;
337 HE *he = hv_iternext (hv); 445 HE *he = hv_iternext (hv);
338 446
339 for (;;) 447 for (;;)
340 { 448 {
341 INDENT; 449 encode_indent (enc);
342 encode_he (enc, he); 450 encode_he (enc, he);
343 451
344 if (!(he = hv_iternext (hv))) 452 if (!(he = hv_iternext (hv)))
345 break; 453 break;
346 454
347 COMMA; 455 encode_comma (enc);
348 } 456 }
349 457
350 NL; 458 encode_nl (enc);
351 } 459 }
352 } 460 }
353 461
354 --enc->indent; INDENT; encode_ch (enc, '}'); 462 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
463}
464
465// encode objects, arrays and special \0=false and \1=true values.
466static void
467encode_rv (enc_t *enc, SV *sv)
468{
469 svtype svt;
470
471 SvGETMAGIC (sv);
472 svt = SvTYPE (sv);
473
474 if (svt == SVt_PVHV)
475 encode_hv (enc, (HV *)sv);
476 else if (svt == SVt_PVAV)
477 encode_av (enc, (AV *)sv);
478 else if (svt < SVt_PVAV)
479 {
480 if (SvNIOK (sv) && SvIV (sv) == 0)
481 encode_str (enc, "false", 5, 0);
482 else if (SvNIOK (sv) && SvIV (sv) == 1)
483 encode_str (enc, "true", 4, 0);
484 else
485 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
486 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
487 }
488 else
489 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
490 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355} 491}
356 492
357static void 493static void
358encode_sv (enc_t *enc, SV *sv) 494encode_sv (enc_t *enc, SV *sv)
359{ 495{
… …
367 encode_str (enc, str, len, SvUTF8 (sv)); 503 encode_str (enc, str, len, SvUTF8 (sv));
368 encode_ch (enc, '"'); 504 encode_ch (enc, '"');
369 } 505 }
370 else if (SvNOKp (sv)) 506 else if (SvNOKp (sv))
371 { 507 {
508 // trust that perl will do the right thing w.r.t. JSON syntax.
372 need (enc, NV_DIG + 32); 509 need (enc, NV_DIG + 32);
373 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 510 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374 enc->cur += strlen (enc->cur); 511 enc->cur += strlen (enc->cur);
375 } 512 }
376 else if (SvIOKp (sv)) 513 else if (SvIOKp (sv))
377 { 514 {
378 need (enc, 64); 515 // we assume we can always read an IV as a UV
516 if (SvUV (sv) & ~(UV)0x7fff)
517 {
518 // large integer, use the (rather slow) snprintf way.
519 need (enc, sizeof (UV) * 3);
379 enc->cur += 520 enc->cur +=
380 SvIsUV(sv) 521 SvIsUV(sv)
381 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 522 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
382 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 523 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
524 }
525 else
526 {
527 // optimise the "small number case"
528 // code will likely be branchless and use only a single multiplication
529 I32 i = SvIV (sv);
530 U32 u;
531 char digit, nz = 0;
532
533 need (enc, 6);
534
535 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
536 u = i < 0 ? -i : i;
537
538 // convert to 4.28 fixed-point representation
539 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
540
541 // now output digit by digit, each time masking out the integer part
542 // and multiplying by 5 while moving the decimal point one to the right,
543 // resulting in a net multiplication by 10.
544 // we always write the digit to memory but conditionally increment
545 // the pointer, to ease the usage of conditional move instructions.
546 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
547 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
548 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
549 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
550 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
551 }
383 } 552 }
384 else if (SvROK (sv)) 553 else if (SvROK (sv))
385 { 554 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)) 555 else if (!SvOK (sv))
401 encode_str (enc, "null", 4, 0); 556 encode_str (enc, "null", 4, 0);
402 else 557 else
403 croak ("encountered perl type that JSON cannot handle"); 558 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
559 SvPV_nolen (sv), SvFLAGS (sv));
404} 560}
405 561
406static SV * 562static SV *
407encode_json (SV *scalar, UV flags) 563encode_json (SV *scalar, U32 flags)
408{ 564{
565 enc_t enc;
566
409 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) 567 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410 croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)"); 568 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
411 569
412 enc_t enc;
413 enc.flags = flags; 570 enc.flags = flags;
414 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 571 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415 enc.cur = SvPVX (enc.sv); 572 enc.cur = SvPVX (enc.sv);
416 enc.end = SvEND (enc.sv); 573 enc.end = SvEND (enc.sv);
417 enc.max_recurse = 0;
418 enc.indent = 0; 574 enc.indent = 0;
575 enc.maxdepth = DEC_DEPTH (flags);
419 576
420 SvPOK_only (enc.sv); 577 SvPOK_only (enc.sv);
421 encode_sv (&enc, scalar); 578 encode_sv (&enc, scalar);
422 579
580 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
581 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
582
423 if (!(flags & (F_ASCII | F_UTF8))) 583 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424 SvUTF8_on (enc.sv); 584 SvUTF8_on (enc.sv);
425 585
426 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 586 if (enc.flags & F_SHRINK)
587 shrink (enc.sv);
588
427 return enc.sv; 589 return enc.sv;
428} 590}
429 591
430///////////////////////////////////////////////////////////////////////////// 592/////////////////////////////////////////////////////////////////////////////
593// decoder
431 594
432#define WS \ 595// structure used for decoding JSON
596typedef struct
597{
598 char *cur; // current parser pointer
599 char *end; // end of input string
600 const char *err; // parse error, if != 0
601 U32 flags; // F_*
602 U32 depth; // recursion depth
603 U32 maxdepth; // recursion depth limit
604} dec_t;
605
606inline void
607decode_ws (dec_t *dec)
608{
433 for (;;) \ 609 for (;;)
434 { \ 610 {
435 char ch = *dec->cur; \ 611 char ch = *dec->cur;
612
436 if (ch > 0x20 \ 613 if (ch > 0x20
437 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 614 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
438 break; \ 615 break;
616
439 ++dec->cur; \ 617 ++dec->cur;
440 } 618 }
619}
441 620
442#define ERR(reason) SB dec->err = reason; goto fail; SE 621#define ERR(reason) SB dec->err = reason; goto fail; SE
622
443#define EXPECT_CH(ch) SB \ 623#define EXPECT_CH(ch) SB \
444 if (*dec->cur != ch) \ 624 if (*dec->cur != ch) \
445 ERR (# ch " expected"); \ 625 ERR (# ch " expected"); \
446 ++dec->cur; \ 626 ++dec->cur; \
447 SE 627 SE
448 628
629#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
630#define DEC_DEC_DEPTH --dec->depth
631
449static SV *decode_sv (dec_t *dec); 632static SV *decode_sv (dec_t *dec);
450 633
451static signed char decode_hexdigit[256]; 634static signed char decode_hexdigit[256];
452 635
453static UV 636static UV
454decode_4hex (dec_t *dec) 637decode_4hex (dec_t *dec)
455{ 638{
456 signed char d1, d2, d3, d4; 639 signed char d1, d2, d3, d4;
640 unsigned char *cur = (unsigned char *)dec->cur;
457 641
458 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 642 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"); 643 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]]; 644 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"); 645 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");
466 646
467 dec->cur += 4; 647 dec->cur += 4;
468 648
469 return ((UV)d1) << 12 649 return ((UV)d1) << 12
470 | ((UV)d2) << 8 650 | ((UV)d2) << 8
… …
473 653
474fail: 654fail:
475 return (UV)-1; 655 return (UV)-1;
476} 656}
477 657
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 * 658static SV *
494decode_str (dec_t *dec) 659decode_str (dec_t *dec)
495{ 660{
496 SV *sv = NEWSV (0,2); 661 SV *sv = 0;
497 int utf8 = 0; 662 int utf8 = 0;
498 char *cur = SvPVX (sv); 663 char *dec_cur = dec->cur;
499 char *end = SvEND (sv);
500 664
501 for (;;) 665 do
502 { 666 {
503 unsigned char ch = *(unsigned char *)dec->cur; 667 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
668 char *cur = buf;
504 669
505 if (ch == '"') 670 do
506 break;
507 else if (ch == '\\')
508 { 671 {
509 switch (*++dec->cur) 672 unsigned char ch = *(unsigned char *)dec_cur++;
673
674 if (expect_false (ch == '"'))
510 { 675 {
511 case '\\': 676 --dec_cur;
512 case '/': 677 break;
513 case '"': APPEND_CH (*dec->cur++); break; 678 }
514 679 else if (expect_false (ch == '\\'))
515 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 680 {
516 case 't': APPEND_CH ('\011'); ++dec->cur; break; 681 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 { 682 {
523 UV lo, hi; 683 case '\\':
524 ++dec->cur; 684 case '/':
685 case '"': *cur++ = *dec_cur++; break;
525 686
526 hi = decode_4hex (dec); 687 case 'b': ++dec_cur; *cur++ = '\010'; break;
527 if (hi == (UV)-1) 688 case 't': ++dec_cur; *cur++ = '\011'; break;
528 goto fail; 689 case 'n': ++dec_cur; *cur++ = '\012'; break;
690 case 'f': ++dec_cur; *cur++ = '\014'; break;
691 case 'r': ++dec_cur; *cur++ = '\015'; break;
529 692
530 // possibly a surrogate pair 693 case 'u':
531 if (hi >= 0xd800 && hi < 0xdc00)
532 { 694 {
533 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 695 UV lo, hi;
534 ERR ("illegal surrogate character"); 696 ++dec_cur;
535 697
536 dec->cur += 2; 698 dec->cur = dec_cur;
537
538 lo = decode_4hex (dec); 699 hi = decode_4hex (dec);
700 dec_cur = dec->cur;
539 if (lo == (UV)-1) 701 if (hi == (UV)-1)
540 goto fail; 702 goto fail;
541 703
704 // possibly a surrogate pair
705 if (hi >= 0xd800)
706 if (hi < 0xdc00)
707 {
708 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
709 ERR ("missing low surrogate character in surrogate pair");
710
711 dec_cur += 2;
712
713 dec->cur = dec_cur;
714 lo = decode_4hex (dec);
715 dec_cur = dec->cur;
716 if (lo == (UV)-1)
717 goto fail;
718
542 if (lo < 0xdc00 || lo >= 0xe000) 719 if (lo < 0xdc00 || lo >= 0xe000)
543 ERR ("surrogate pair expected"); 720 ERR ("surrogate pair expected");
544 721
545 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 722 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
723 }
724 else if (hi < 0xe000)
725 ERR ("missing high surrogate character in surrogate pair");
726
727 if (hi >= 0x80)
728 {
729 utf8 = 1;
730
731 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
732 }
733 else
734 *cur++ = hi;
546 } 735 }
547 else if (lo >= 0xdc00 && lo < 0xe000)
548 ERR ("illegal surrogate character");
549
550 if (hi >= 0x80)
551 { 736 break;
552 utf8 = 1;
553 737
554 APPEND_GROW (4); // at most 4 bytes for 21 bits
555 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556 }
557 else 738 default:
558 APPEND_CH (hi); 739 --dec_cur;
740 ERR ("illegal backslash escape sequence in string");
559 } 741 }
742 }
743 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
744 *cur++ = ch;
745 else if (ch >= 0x80)
746 {
747 STRLEN clen;
748 UV uch;
749
750 --dec_cur;
751
752 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
753 if (clen == (STRLEN)-1)
754 ERR ("malformed UTF-8 character in JSON string");
755
756 do
757 *cur++ = *dec_cur++;
758 while (--clen);
759
760 utf8 = 1;
761 }
762 else
763 {
764 --dec_cur;
765
766 if (!ch)
767 ERR ("unexpected end of string while parsing JSON string");
560 break; 768 else
769 ERR ("invalid character encountered while parsing JSON string");
561 } 770 }
562 } 771 }
563 else if (ch >= 0x20 && ch <= 0x7f) 772 while (cur < buf + SHORT_STRING_LEN);
564 APPEND_CH (*dec->cur++); 773
565 else if (ch >= 0x80) 774 {
775 STRLEN len = cur - buf;
776
777 if (sv)
566 { 778 {
567 STRLEN clen; 779 SvGROW (sv, SvCUR (sv) + len + 1);
568 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 780 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
569 if (clen < 0) 781 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 } 782 }
577 else 783 else
578 ERR ("invalid character encountered"); 784 sv = newSVpvn (buf, len);
579 } 785 }
786 }
787 while (*dec_cur != '"');
580 788
581 ++dec->cur; 789 ++dec_cur;
582 790
583 SvCUR_set (sv, cur - SvPVX (sv)); 791 if (sv)
584 792 {
585 SvPOK_only (sv); 793 SvPOK_only (sv);
586 *SvEND (sv) = 0; 794 *SvEND (sv) = 0;
587 795
588 if (utf8) 796 if (utf8)
589 SvUTF8_on (sv); 797 SvUTF8_on (sv);
798 }
799 else
800 sv = newSVpvn ("", 0);
590 801
802 dec->cur = dec_cur;
591 return sv; 803 return sv;
592 804
593fail: 805fail:
594 SvREFCNT_dec (sv); 806 dec->cur = dec_cur;
595 return 0; 807 return 0;
596} 808}
597 809
598static SV * 810static SV *
599decode_num (dec_t *dec) 811decode_num (dec_t *dec)
… …
609 { 821 {
610 ++dec->cur; 822 ++dec->cur;
611 if (*dec->cur >= '0' && *dec->cur <= '9') 823 if (*dec->cur >= '0' && *dec->cur <= '9')
612 ERR ("malformed number (leading zero must not be followed by another digit)"); 824 ERR ("malformed number (leading zero must not be followed by another digit)");
613 } 825 }
614 826 else if (*dec->cur < '0' || *dec->cur > '9')
615 // int 827 ERR ("malformed number (no digits after initial minus)");
828 else
829 do
830 {
831 ++dec->cur;
832 }
616 while (*dec->cur >= '0' && *dec->cur <= '9') 833 while (*dec->cur >= '0' && *dec->cur <= '9');
617 ++dec->cur;
618 834
619 // [frac] 835 // [frac]
620 if (*dec->cur == '.') 836 if (*dec->cur == '.')
621 { 837 {
622 is_nv = 1; 838 ++dec->cur;
839
840 if (*dec->cur < '0' || *dec->cur > '9')
841 ERR ("malformed number (no digits after decimal point)");
623 842
624 do 843 do
625 { 844 {
626 ++dec->cur; 845 ++dec->cur;
627 } 846 }
628 while (*dec->cur >= '0' && *dec->cur <= '9'); 847 while (*dec->cur >= '0' && *dec->cur <= '9');
848
849 is_nv = 1;
629 } 850 }
630 851
631 // [exp] 852 // [exp]
632 if (*dec->cur == 'e' || *dec->cur == 'E') 853 if (*dec->cur == 'e' || *dec->cur == 'E')
633 { 854 {
634 is_nv = 1;
635
636 ++dec->cur; 855 ++dec->cur;
856
637 if (*dec->cur == '-' || *dec->cur == '+') 857 if (*dec->cur == '-' || *dec->cur == '+')
638 ++dec->cur; 858 ++dec->cur;
639 859
860 if (*dec->cur < '0' || *dec->cur > '9')
861 ERR ("malformed number (no digits after exp sign)");
862
863 do
864 {
865 ++dec->cur;
866 }
640 while (*dec->cur >= '0' && *dec->cur <= '9') 867 while (*dec->cur >= '0' && *dec->cur <= '9');
641 ++dec->cur; 868
869 is_nv = 1;
642 } 870 }
643 871
644 if (!is_nv) 872 if (!is_nv)
645 { 873 {
646 UV uv; 874 // 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); 875 if (*start == '-')
648 if (numtype & IS_NUMBER_IN_UV) 876 switch (dec->cur - start)
649 if (numtype & IS_NUMBER_NEG)
650 { 877 {
651 if (uv < (UV)IV_MIN) 878 case 2: return newSViv (-( start [1] - '0' * 1));
652 return newSViv (-(IV)uv); 879 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
880 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
881 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
653 } 882 }
883 else
884 switch (dec->cur - start)
885 {
886 case 1: return newSViv ( start [0] - '0' * 1);
887 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
888 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
889 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
890 }
891
892 {
893 UV uv;
894 int numtype = grok_number (start, dec->cur - start, &uv);
895 if (numtype & IS_NUMBER_IN_UV)
896 if (numtype & IS_NUMBER_NEG)
897 {
898 if (uv < (UV)IV_MIN)
899 return newSViv (-(IV)uv);
900 }
654 else 901 else
655 return newSVuv (uv); 902 return newSVuv (uv);
903
904 // here would likely be the place for bigint support
656 } 905 }
906 }
657 907
908 // if we ever support bigint or bigfloat, this is the place for bigfloat
658 return newSVnv (Atof (start)); 909 return newSVnv (Atof (start));
659 910
660fail: 911fail:
661 return 0; 912 return 0;
662} 913}
… …
664static SV * 915static SV *
665decode_av (dec_t *dec) 916decode_av (dec_t *dec)
666{ 917{
667 AV *av = newAV (); 918 AV *av = newAV ();
668 919
920 DEC_INC_DEPTH;
921 decode_ws (dec);
922
923 if (*dec->cur == ']')
924 ++dec->cur;
925 else
669 for (;;) 926 for (;;)
670 { 927 {
671 SV *value; 928 SV *value;
672 929
673 value = decode_sv (dec); 930 value = decode_sv (dec);
674 if (!value) 931 if (!value)
675 goto fail; 932 goto fail;
676 933
677 av_push (av, value); 934 av_push (av, value);
678 935
679 WS; 936 decode_ws (dec);
680 937
681 if (*dec->cur == ']') 938 if (*dec->cur == ']')
682 { 939 {
683 ++dec->cur; 940 ++dec->cur;
684 break; 941 break;
942 }
685 } 943
686
687 if (*dec->cur != ',') 944 if (*dec->cur != ',')
688 ERR (", or ] expected while parsing array"); 945 ERR (", or ] expected while parsing array");
689 946
690 ++dec->cur; 947 ++dec->cur;
691 } 948 }
692 949
950 DEC_DEC_DEPTH;
693 return newRV_noinc ((SV *)av); 951 return newRV_noinc ((SV *)av);
694 952
695fail: 953fail:
696 SvREFCNT_dec (av); 954 SvREFCNT_dec (av);
955 DEC_DEC_DEPTH;
697 return 0; 956 return 0;
698} 957}
699 958
700static SV * 959static SV *
701decode_hv (dec_t *dec) 960decode_hv (dec_t *dec)
702{ 961{
703 HV *hv = newHV (); 962 HV *hv = newHV ();
704 963
964 DEC_INC_DEPTH;
965 decode_ws (dec);
966
967 if (*dec->cur == '}')
968 ++dec->cur;
969 else
705 for (;;) 970 for (;;)
706 { 971 {
707 SV *key, *value; 972 SV *key, *value;
708 973
709 WS; EXPECT_CH ('"'); 974 decode_ws (dec); EXPECT_CH ('"');
710 975
711 key = decode_str (dec); 976 key = decode_str (dec);
712 if (!key) 977 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; 978 goto fail;
979
980 decode_ws (dec); EXPECT_CH (':');
981
982 value = decode_sv (dec);
983 if (!value)
984 {
985 SvREFCNT_dec (key);
986 goto fail;
722 } 987 }
723 988
724 //TODO: optimise
725 hv_store_ent (hv, key, value, 0); 989 hv_store_ent (hv, key, value, 0);
990 SvREFCNT_dec (key);
726 991
727 WS; 992 decode_ws (dec);
728 993
729 if (*dec->cur == '}') 994 if (*dec->cur == '}')
730 { 995 {
731 ++dec->cur; 996 ++dec->cur;
732 break; 997 break;
733 } 998 }
734 999
735 if (*dec->cur != ',') 1000 if (*dec->cur != ',')
736 ERR (", or } expected while parsing object/hash"); 1001 ERR (", or } expected while parsing object/hash");
737 1002
738 ++dec->cur; 1003 ++dec->cur;
739 } 1004 }
740 1005
1006 DEC_DEC_DEPTH;
741 return newRV_noinc ((SV *)hv); 1007 return newRV_noinc ((SV *)hv);
742 1008
743fail: 1009fail:
744 SvREFCNT_dec (hv); 1010 SvREFCNT_dec (hv);
1011 DEC_DEC_DEPTH;
745 return 0; 1012 return 0;
746} 1013}
747 1014
748static SV * 1015static SV *
749decode_sv (dec_t *dec) 1016decode_sv (dec_t *dec)
750{ 1017{
751 WS; 1018 decode_ws (dec);
1019
1020 // the beauty of JSON: you need exactly one character lookahead
1021 // to parse anything.
752 switch (*dec->cur) 1022 switch (*dec->cur)
753 { 1023 {
754 case '"': ++dec->cur; return decode_str (dec); 1024 case '"': ++dec->cur; return decode_str (dec);
755 case '[': ++dec->cur; return decode_av (dec); 1025 case '[': ++dec->cur; return decode_av (dec);
756 case '{': ++dec->cur; return decode_hv (dec); 1026 case '{': ++dec->cur; return decode_hv (dec);
… …
762 1032
763 case 't': 1033 case 't':
764 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1034 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765 { 1035 {
766 dec->cur += 4; 1036 dec->cur += 4;
767 return newSViv (1); 1037 return SvREFCNT_inc (json_true);
768 } 1038 }
769 else 1039 else
770 ERR ("'true' expected"); 1040 ERR ("'true' expected");
771 1041
772 break; 1042 break;
773 1043
774 case 'f': 1044 case 'f':
775 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1045 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776 { 1046 {
777 dec->cur += 5; 1047 dec->cur += 5;
778 return newSViv (0); 1048 return SvREFCNT_inc (json_false);
779 } 1049 }
780 else 1050 else
781 ERR ("'false' expected"); 1051 ERR ("'false' expected");
782 1052
783 break; 1053 break;
784 1054
785 case 'n': 1055 case 'n':
786 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1056 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787 { 1057 {
788 dec->cur += 4; 1058 dec->cur += 4;
789 return newSViv (1); 1059 return newSVsv (&PL_sv_undef);
790 } 1060 }
791 else 1061 else
792 ERR ("'null' expected"); 1062 ERR ("'null' expected");
793 1063
794 break; 1064 break;
795 1065
796 default: 1066 default:
797 ERR ("malformed json string"); 1067 ERR ("malformed JSON string, neither array, object, number, string or atom");
798 break; 1068 break;
799 } 1069 }
800 1070
801fail: 1071fail:
802 return 0; 1072 return 0;
803} 1073}
804 1074
805static SV * 1075static SV *
806decode_json (SV *string, UV flags) 1076decode_json (SV *string, U32 flags, UV *offset_return)
807{ 1077{
1078 dec_t dec;
1079 UV offset;
808 SV *sv; 1080 SV *sv;
809 1081
1082 SvGETMAGIC (string);
1083 SvUPGRADE (string, SVt_PV);
1084
810 if (!(flags & F_UTF8)) 1085 if (flags & F_UTF8)
1086 sv_utf8_downgrade (string, 0);
1087 else
811 sv_utf8_upgrade (string); 1088 sv_utf8_upgrade (string);
812 1089
813 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1090 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814 1091
815 dec_t dec;
816 dec.flags = flags; 1092 dec.flags = flags;
817 dec.cur = SvPVX (string); 1093 dec.cur = SvPVX (string);
818 dec.end = SvEND (string); 1094 dec.end = SvEND (string);
819 dec.err = 0; 1095 dec.err = 0;
1096 dec.depth = 0;
1097 dec.maxdepth = DEC_DEPTH (dec.flags);
820 1098
821 *dec.end = 1; // invalid anywhere 1099 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
822 sv = decode_sv (&dec); 1100 sv = decode_sv (&dec);
823 *dec.end = 0; 1101
1102 if (!(offset_return || !sv))
1103 {
1104 // check for trailing garbage
1105 decode_ws (&dec);
1106
1107 if (*dec.cur)
1108 {
1109 dec.err = "garbage after JSON object";
1110 SvREFCNT_dec (sv);
1111 sv = 0;
1112 }
1113 }
1114
1115 if (offset_return || !sv)
1116 {
1117 offset = dec.flags & F_UTF8
1118 ? dec.cur - SvPVX (string)
1119 : utf8_distance (dec.cur, SvPVX (string));
1120
1121 if (offset_return)
1122 *offset_return = offset;
1123 }
824 1124
825 if (!sv) 1125 if (!sv)
826 { 1126 {
827 IV offset = utf8_distance (dec.cur, SvPVX (string));
828 SV *uni = sv_newmortal (); 1127 SV *uni = sv_newmortal ();
829 1128
1129 // horrible hack to silence warning inside pv_uni_display
1130 COP cop = *PL_curcop;
1131 cop.cop_warnings = pWARN_NONE;
1132 ENTER;
1133 SAVEVPTR (PL_curcop);
1134 PL_curcop = &cop;
830 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1135 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1136 LEAVE;
1137
831 croak ("%s, at character %d (%s)", 1138 croak ("%s, at character offset %d [\"%s\"]",
832 dec.err, 1139 dec.err,
833 (int)offset, 1140 (int)offset,
834 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1141 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835 } 1142 }
836 1143
837 sv = sv_2mortal (sv); 1144 sv = sv_2mortal (sv);
838 1145
839 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1146 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)"); 1147 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841 1148
842 return sv; 1149 return sv;
843} 1150}
844 1151
1152/////////////////////////////////////////////////////////////////////////////
1153// XS interface functions
1154
845MODULE = JSON::XS PACKAGE = JSON::XS 1155MODULE = JSON::XS PACKAGE = JSON::XS
846 1156
847BOOT: 1157BOOT:
848{ 1158{
849 int i; 1159 int i;
850 1160
851 memset (decode_hexdigit, 0xff, 256);
852 for (i = 10; i--; ) 1161 for (i = 0; i < 256; ++i)
853 decode_hexdigit ['0' + i] = i; 1162 decode_hexdigit [i] =
854 1163 i >= '0' && i <= '9' ? i - '0'
855 for (i = 7; i--; ) 1164 : i >= 'a' && i <= 'f' ? i - 'a' + 10
856 { 1165 : i >= 'A' && i <= 'F' ? i - 'A' + 10
857 decode_hexdigit ['a' + i] = 10 + i; 1166 : -1;
858 decode_hexdigit ['A' + i] = 10 + i;
859 }
860 1167
861 json_stash = gv_stashpv ("JSON::XS", 1); 1168 json_stash = gv_stashpv ("JSON::XS", 1);
1169
1170 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1171 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
862} 1172}
863 1173
864PROTOTYPES: DISABLE 1174PROTOTYPES: DISABLE
865 1175
866SV *new (char *dummy) 1176SV *new (char *dummy)
867 CODE: 1177 CODE:
868 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1178 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
869 OUTPUT: 1179 OUTPUT:
870 RETVAL 1180 RETVAL
871 1181
872SV *ascii (SV *self, int enable) 1182SV *ascii (SV *self, int enable = 1)
873 ALIAS: 1183 ALIAS:
874 ascii = F_ASCII 1184 ascii = F_ASCII
1185 latin1 = F_LATIN1
875 utf8 = F_UTF8 1186 utf8 = F_UTF8
876 indent = F_INDENT 1187 indent = F_INDENT
877 canonical = F_CANONICAL 1188 canonical = F_CANONICAL
878 space_before = F_SPACE_BEFORE 1189 space_before = F_SPACE_BEFORE
879 space_after = F_SPACE_AFTER 1190 space_after = F_SPACE_AFTER
880 json_rpc = F_JSON_RPC
881 pretty = F_PRETTY 1191 pretty = F_PRETTY
882 allow_nonref = F_ALLOW_NONREF 1192 allow_nonref = F_ALLOW_NONREF
1193 shrink = F_SHRINK
883 CODE: 1194 CODE:
884{ 1195{
885 UV *uv = SvJSON (self); 1196 UV *uv = SvJSON (self);
886 if (enable) 1197 if (enable)
887 *uv |= ix; 1198 *uv |= ix;
… …
891 RETVAL = newSVsv (self); 1202 RETVAL = newSVsv (self);
892} 1203}
893 OUTPUT: 1204 OUTPUT:
894 RETVAL 1205 RETVAL
895 1206
1207SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
1208 CODE:
1209{
1210 UV *uv = SvJSON (self);
1211 UV log2 = 0;
1212
1213 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1214
1215 while ((1UL << log2) < max_depth)
1216 ++log2;
1217
1218 *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1219
1220 RETVAL = newSVsv (self);
1221}
1222 OUTPUT:
1223 RETVAL
1224
896void encode (SV *self, SV *scalar) 1225void encode (SV *self, SV *scalar)
897 PPCODE: 1226 PPCODE:
898 XPUSHs (encode_json (scalar, *SvJSON (self))); 1227 XPUSHs (encode_json (scalar, *SvJSON (self)));
899 1228
900void decode (SV *self, SV *jsonstr) 1229void decode (SV *self, SV *jsonstr)
901 PPCODE: 1230 PPCODE:
902 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1231 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1232
1233void decode_prefix (SV *self, SV *jsonstr)
1234 PPCODE:
1235{
1236 UV offset;
1237 EXTEND (SP, 2);
1238 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1239 PUSHs (sv_2mortal (newSVuv (offset)));
1240}
903 1241
904PROTOTYPES: ENABLE 1242PROTOTYPES: ENABLE
905 1243
906void to_json (SV *scalar) 1244void to_json (SV *scalar)
1245 ALIAS:
1246 objToJson = 0
907 PPCODE: 1247 PPCODE:
908 XPUSHs (encode_json (scalar, F_UTF8)); 1248 XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
909 1249
910void from_json (SV *jsonstr) 1250void from_json (SV *jsonstr)
1251 ALIAS:
1252 jsonToObj = 0
911 PPCODE: 1253 PPCODE:
912 XPUSHs (decode_json (jsonstr, F_UTF8)); 1254 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
913 1255

Diff Legend

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