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 1.1 by root, Thu Mar 22 16:40:16 2007 UTC vs.
Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "assert.h" 5#include <assert.h>
6#include "string.h" 6#include <string.h>
7#include "stdlib.h" 7#include <stdlib.h>
8#include <stdio.h>
9#include <float.h>
8 10
11#if defined(__BORLANDC__) || defined(_MSC_VER)
12# define snprintf _snprintf // C compilers have this in stdio.h
13#endif
14
15// some old perls do not have this, try to make it work, no
16// guarentees, though. if it breaks, you get to keep the pieces.
17#ifndef UTF8_MAXBYTES
18# define UTF8_MAXBYTES 13
19#endif
20
9#define F_ASCII 0x00000001 21#define F_ASCII 0x00000001UL
22#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 23#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 24#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 25#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 26#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 27#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040 28#define F_ALLOW_NONREF 0x00000100UL
29#define F_SHRINK 0x00000200UL
30#define F_ALLOW_BLESSED 0x00000400UL
31#define F_CONV_BLESSED 0x00000800UL
32#define F_RELAXED 0x00001000UL
16 33
17#define F_DEFAULT 0 34#define F_MAXDEPTH 0xf8000000UL
35#define S_MAXDEPTH 27
36#define F_MAXSIZE 0x01f00000UL
37#define S_MAXSIZE 20
38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39
40#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
44#define F_DEFAULT (9UL << S_MAXDEPTH)
18 45
19#define INIT_SIZE 32 // initial scalar size to be allocated 46#define INIT_SIZE 32 // initial scalar size to be allocated
47#define INDENT_STEP 3 // spaces per indentation level
48
49#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
20 50
21#define SB do { 51#define SB do {
22#define SE } while (0) 52#define SE } while (0)
23 53
24static HV *json_stash; 54#if __GNUC__ >= 3
55# define expect(expr,value) __builtin_expect ((expr),(value))
56# define inline inline
57#else
58# define expect(expr,value) (expr)
59# define inline static
60#endif
61
62#define expect_false(expr) expect ((expr) != 0, 0)
63#define expect_true(expr) expect ((expr) != 0, 1)
64
65#ifdef USE_ITHREADS
66# define JSON_SLOW 1
67# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
68#else
69# define JSON_SLOW 0
70# define JSON_STASH json_stash
71#endif
72
73static HV *json_stash, *json_boolean_stash; // JSON::XS::
74static SV *json_true, *json_false;
75
76typedef struct {
77 U32 flags;
78 SV *cb_object;
79 HV *cb_sk_object;
80} JSON;
81
82/////////////////////////////////////////////////////////////////////////////
83// utility functions
84
85inline void
86shrink (SV *sv)
87{
88 sv_utf8_downgrade (sv, 1);
89 if (SvLEN (sv) > SvCUR (sv) + 1)
90 {
91#ifdef SvPV_shrink_to_cur
92 SvPV_shrink_to_cur (sv);
93#elif defined (SvPV_renew)
94 SvPV_renew (sv, SvCUR (sv) + 1);
95#endif
96 }
97}
98
99// decode an utf-8 character and return it, or (UV)-1 in
100// case of an error.
101// we special-case "safe" characters from U+80 .. U+7FF,
102// but use the very good perl function to parse anything else.
103// note that we never call this function for a ascii codepoints
104inline UV
105decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
106{
107 if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
108 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
109 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
110 {
111 *clen = 2;
112 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
113 }
114 else
115 {
116 *clen = (STRLEN)-1;
117 return (UV)-1;
118 }
119}
120
121/////////////////////////////////////////////////////////////////////////////
122// encoder
25 123
26// structure used for encoding JSON 124// structure used for encoding JSON
27typedef struct 125typedef struct
28{ 126{
29 char *cur; 127 char *cur; // SvPVX (sv) + current output position
30 STRLEN len; // SvLEN (sv)
31 char *end; // SvEND (sv) 128 char *end; // SvEND (sv)
32 SV *sv; 129 SV *sv; // result scalar
33 UV flags; 130 JSON json;
34 int max_recurse; 131 U32 indent; // indentation level
35 int indent; 132 U32 maxdepth; // max. indentation/recursion level
36} enc_t; 133} enc_t;
37 134
38// structure used for decoding JSON 135inline void
39typedef struct
40{
41 char *cur;
42 char *end;
43 char *err;
44 UV flags;
45} dec_t;
46
47static UV *
48SvJSON (SV *sv)
49{
50 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
51 croak ("object is not of type JSON::XS");
52
53 return &SvUVX (SvRV (sv));
54}
55
56/////////////////////////////////////////////////////////////////////////////
57
58static void
59need (enc_t *enc, STRLEN len) 136need (enc_t *enc, STRLEN len)
60{ 137{
61 if (enc->cur + len >= enc->end) 138 if (expect_false (enc->cur + len >= enc->end))
62 { 139 {
63 STRLEN cur = enc->cur - SvPVX (enc->sv); 140 STRLEN cur = enc->cur - SvPVX (enc->sv);
64 SvGROW (enc->sv, cur + len + 1); 141 SvGROW (enc->sv, cur + len + 1);
65 enc->cur = SvPVX (enc->sv) + cur; 142 enc->cur = SvPVX (enc->sv) + cur;
66 enc->end = SvEND (enc->sv); 143 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
67 } 144 }
68} 145}
69 146
70static void 147inline void
71encode_ch (enc_t *enc, char ch) 148encode_ch (enc_t *enc, char ch)
72{ 149{
73 need (enc, 1); 150 need (enc, 1);
74 *enc->cur++ = ch; 151 *enc->cur++ = ch;
75} 152}
77static void 154static void
78encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) 155encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
79{ 156{
80 char *end = str + len; 157 char *end = str + len;
81 158
159 need (enc, len);
160
82 while (str < end) 161 while (str < end)
83 { 162 {
84 unsigned char ch = *(unsigned char *)str; 163 unsigned char ch = *(unsigned char *)str;
164
85 if (ch >= 0x20 && ch < 0x80) // most common case 165 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
86 { 166 {
167 if (expect_false (ch == '"')) // but with slow exceptions
168 {
169 need (enc, len += 1);
170 *enc->cur++ = '\\';
171 *enc->cur++ = '"';
172 }
173 else if (expect_false (ch == '\\'))
174 {
175 need (enc, len += 1);
176 *enc->cur++ = '\\';
177 *enc->cur++ = '\\';
178 }
179 else
87 *enc->cur++ = ch; 180 *enc->cur++ = ch;
181
88 str++; 182 ++str;
89 } 183 }
90 else 184 else
91 { 185 {
92 STRLEN clen; 186 switch (ch)
93 UV uch;
94
95 if (is_utf8)
96 {
97 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
98 if (clen < 0)
99 croak ("malformed UTF-8 character in string, cannot convert to JSON");
100 } 187 {
101 else 188 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
102 { 189 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
103 uch = ch; 190 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
104 clen = 1; 191 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
105 } 192 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
106 193
107 need (enc, len += 6); 194 default:
108
109 if (uch < 0xa0 || enc->flags & F_ASCII)
110 {
111 if (uch > 0xFFFFUL)
112 { 195 {
196 STRLEN clen;
197 UV uch;
198
199 if (is_utf8)
200 {
201 uch = decode_utf8 (str, end - str, &clen);
202 if (clen == (STRLEN)-1)
203 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
204 }
205 else
206 {
207 uch = ch;
113 len += 6; 208 clen = 1;
209 }
210
211 if (uch > 0x10FFFFUL)
212 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
213
214 if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
215 {
216 if (uch > 0xFFFFUL)
217 {
114 need (enc, len += 6); 218 need (enc, len += 11);
115 sprintf (enc->cur, "\\u%04x\\u%04x", 219 sprintf (enc->cur, "\\u%04x\\u%04x",
116 (uch - 0x10000) / 0x400 + 0xD800, 220 (int)((uch - 0x10000) / 0x400 + 0xD800),
117 (uch - 0x10000) % 0x400 + 0xDC00); 221 (int)((uch - 0x10000) % 0x400 + 0xDC00));
118 enc->cur += 12; 222 enc->cur += 12;
223 }
224 else
225 {
226 static char hexdigit [16] = "0123456789abcdef";
227 need (enc, len += 5);
228 *enc->cur++ = '\\';
229 *enc->cur++ = 'u';
230 *enc->cur++ = hexdigit [ uch >> 12 ];
231 *enc->cur++ = hexdigit [(uch >> 8) & 15];
232 *enc->cur++ = hexdigit [(uch >> 4) & 15];
233 *enc->cur++ = hexdigit [(uch >> 0) & 15];
234 }
235
236 str += clen;
237 }
238 else if (enc->json.flags & F_LATIN1)
239 {
240 *enc->cur++ = uch;
241 str += clen;
242 }
243 else if (is_utf8)
244 {
245 need (enc, len += clen);
246 do
247 {
248 *enc->cur++ = *str++;
249 }
250 while (--clen);
251 }
252 else
253 {
254 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
256 ++str;
257 }
119 } 258 }
120 else
121 {
122 sprintf (enc->cur, "\\u%04x", uch);
123 enc->cur += 6;
124 }
125 }
126 else if (is_utf8)
127 { 259 }
128 memcpy (enc->cur, str, clen);
129 enc->cur += clen;
130 }
131 else
132 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
133
134 str += clen;
135 } 260 }
136 261
137 --len; 262 --len;
138 } 263 }
139} 264}
140 265
141#define INDENT SB \ 266inline void
267encode_indent (enc_t *enc)
268{
142 if (enc->flags & F_INDENT) \ 269 if (enc->json.flags & F_INDENT)
143 { \ 270 {
144 int i_; \ 271 int spaces = enc->indent * INDENT_STEP;
145 need (enc, enc->indent); \ 272
146 for (i_ = enc->indent * 3; i_--; )\ 273 need (enc, spaces);
274 memset (enc->cur, ' ', spaces);
275 enc->cur += spaces;
276 }
277}
278
279inline void
280encode_space (enc_t *enc)
281{
282 need (enc, 1);
283 encode_ch (enc, ' ');
284}
285
286inline void
287encode_nl (enc_t *enc)
288{
289 if (enc->json.flags & F_INDENT)
290 {
291 need (enc, 1);
147 encode_ch (enc, ' '); \ 292 encode_ch (enc, '\n');
148 } \ 293 }
149 SE 294}
150 295
151#define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE 296inline void
152#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 297encode_comma (enc_t *enc)
153#define COMMA SB \ 298{
154 encode_ch (enc, ','); \ 299 encode_ch (enc, ',');
300
155 if (enc->flags & F_INDENT) \ 301 if (enc->json.flags & F_INDENT)
156 NL; \ 302 encode_nl (enc);
157 else if (enc->flags & F_SPACE_AFTER) \ 303 else if (enc->json.flags & F_SPACE_AFTER)
158 SPACE; \ 304 encode_space (enc);
159 SE 305}
160 306
161static void encode_sv (enc_t *enc, SV *sv); 307static void encode_sv (enc_t *enc, SV *sv);
162 308
163static void 309static void
164encode_av (enc_t *enc, AV *av) 310encode_av (enc_t *enc, AV *av)
165{ 311{
166 int i, len = av_len (av); 312 int i, len = av_len (av);
167 313
168 encode_ch (enc, '['); NL; 314 if (enc->indent >= enc->maxdepth)
315 croak ("data structure too deep (hit recursion limit)");
316
317 encode_ch (enc, '['); encode_nl (enc);
169 ++enc->indent; 318 ++enc->indent;
170 319
171 for (i = 0; i <= len; ++i) 320 for (i = 0; i <= len; ++i)
172 { 321 {
173 INDENT; 322 SV **svp = av_fetch (av, i, 0);
174 encode_sv (enc, *av_fetch (av, i, 0)); 323
324 encode_indent (enc);
325
326 if (svp)
327 encode_sv (enc, *svp);
328 else
329 encode_str (enc, "null", 4, 0);
175 330
176 if (i < len) 331 if (i < len)
177 COMMA; 332 encode_comma (enc);
178 } 333 }
179 334
180 NL; 335 encode_nl (enc);
181 336
182 --enc->indent; 337 --enc->indent;
183 INDENT; encode_ch (enc, ']'); 338 encode_indent (enc); encode_ch (enc, ']');
184} 339}
185 340
186static void 341static void
187encode_he (enc_t *enc, HE *he) 342encode_hk (enc_t *enc, HE *he)
188{ 343{
189 encode_ch (enc, '"'); 344 encode_ch (enc, '"');
190 345
191 if (HeKLEN (he) == HEf_SVKEY) 346 if (HeKLEN (he) == HEf_SVKEY)
192 { 347 {
193 SV *sv = HeSVKEY (he); 348 SV *sv = HeSVKEY (he);
194 STRLEN len; 349 STRLEN len;
350 char *str;
351
352 SvGETMAGIC (sv);
195 char *str = SvPV (sv, len); 353 str = SvPV (sv, len);
196 354
197 encode_str (enc, str, len, SvUTF8 (sv)); 355 encode_str (enc, str, len, SvUTF8 (sv));
198 } 356 }
199 else 357 else
200 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 358 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
201 359
202 encode_ch (enc, '"'); 360 encode_ch (enc, '"');
203 361
204 if (enc->flags & F_SPACE_BEFORE) SPACE; 362 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
205 encode_ch (enc, ':'); 363 encode_ch (enc, ':');
206 if (enc->flags & F_SPACE_AFTER ) SPACE; 364 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
207 encode_sv (enc, HeVAL (he));
208} 365}
209 366
210// compare hash entries, used when all keys are bytestrings 367// compare hash entries, used when all keys are bytestrings
211static int 368static int
212he_cmp_fast (const void *a_, const void *b_) 369he_cmp_fast (const void *a_, const void *b_)
217 HE *b = *(HE **)b_; 374 HE *b = *(HE **)b_;
218 375
219 STRLEN la = HeKLEN (a); 376 STRLEN la = HeKLEN (a);
220 STRLEN lb = HeKLEN (b); 377 STRLEN lb = HeKLEN (b);
221 378
222 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 379 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
223 cmp = la < lb ? -1 : la == lb ? 0 : 1; 380 cmp = lb - la;
224 381
225 return cmp; 382 return cmp;
226} 383}
227 384
228// compare hash entries, used when some keys are sv's or utf-x 385// compare hash entries, used when some keys are sv's or utf-x
229static int 386static int
230he_cmp_slow (const void *a, const void *b) 387he_cmp_slow (const void *a, const void *b)
231{ 388{
232 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 389 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
233} 390}
234 391
235static void 392static void
236encode_hv (enc_t *enc, HV *hv) 393encode_hv (enc_t *enc, HV *hv)
237{ 394{
395 HE *he;
238 int count, i; 396 int count;
239 397
398 if (enc->indent >= enc->maxdepth)
399 croak ("data structure too deep (hit recursion limit)");
400
240 encode_ch (enc, '{'); NL; ++enc->indent; 401 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
241 402
242 if ((count = hv_iterinit (hv)))
243 {
244 // for canonical output we have to sort by keys first 403 // for canonical output we have to sort by keys first
245 // actually, this is mostly due to the stupid so-called 404 // actually, this is mostly due to the stupid so-called
246 // security workaround added somewhere in 5.8.x. 405 // security workaround added somewhere in 5.8.x.
247 // that randomises hash orderings 406 // that randomises hash orderings
248 if (enc->flags & F_CANONICAL) 407 if (enc->json.flags & F_CANONICAL)
408 {
409 int count = hv_iterinit (hv);
410
411 if (SvMAGICAL (hv))
249 { 412 {
250 HE *he, *hes [count]; 413 // need to count by iterating. could improve by dynamically building the vector below
414 // but I don't care for the speed of this special case.
415 // note also that we will run into undefined behaviour when the two iterations
416 // do not result in the same count, something I might care for in some later release.
417
418 count = 0;
419 while (hv_iternext (hv))
420 ++count;
421
422 hv_iterinit (hv);
423 }
424
425 if (count)
426 {
251 int fast = 1; 427 int i, fast = 1;
428#if defined(__BORLANDC__) || defined(_MSC_VER)
429 HE **hes = _alloca (count * sizeof (HE));
430#else
431 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
432#endif
252 433
253 i = 0; 434 i = 0;
254 while ((he = hv_iternext (hv))) 435 while ((he = hv_iternext (hv)))
255 { 436 {
256 hes [i++] = he; 437 hes [i++] = he;
262 443
263 if (fast) 444 if (fast)
264 qsort (hes, count, sizeof (HE *), he_cmp_fast); 445 qsort (hes, count, sizeof (HE *), he_cmp_fast);
265 else 446 else
266 { 447 {
267 // hack to disable "use bytes" 448 // hack to forcefully disable "use bytes"
268 COP *oldcop = PL_curcop, cop; 449 COP cop = *PL_curcop;
269 cop.op_private = 0; 450 cop.op_private = 0;
451
452 ENTER;
453 SAVETMPS;
454
455 SAVEVPTR (PL_curcop);
270 PL_curcop = &cop; 456 PL_curcop = &cop;
271 457
272 SAVETMPS;
273 qsort (hes, count, sizeof (HE *), he_cmp_slow); 458 qsort (hes, count, sizeof (HE *), he_cmp_slow);
459
274 FREETMPS; 460 FREETMPS;
275 461 LEAVE;
276 PL_curcop = oldcop;
277 }
278
279 for (i = 0; i < count; ++i)
280 { 462 }
281 INDENT;
282 encode_he (enc, hes [i]);
283 463
284 if (i < count - 1) 464 while (count--)
285 COMMA;
286 } 465 {
466 encode_indent (enc);
467 he = hes [count];
468 encode_hk (enc, he);
469 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
287 470
471 if (count)
472 encode_comma (enc);
288 NL; 473 }
474 }
475 }
476 else
477 {
478 if (hv_iterinit (hv) || SvMAGICAL (hv))
479 if ((he = hv_iternext (hv)))
480 for (;;)
481 {
482 encode_indent (enc);
483 encode_hk (enc, he);
484 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485
486 if (!(he = hv_iternext (hv)))
487 break;
488
489 encode_comma (enc);
490 }
491 }
492
493 encode_nl (enc);
494
495 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
496}
497
498// encode objects, arrays and special \0=false and \1=true values.
499static void
500encode_rv (enc_t *enc, SV *sv)
501{
502 svtype svt;
503
504 SvGETMAGIC (sv);
505 svt = SvTYPE (sv);
506
507 if (expect_false (SvOBJECT (sv)))
508 {
509 HV *stash = !JSON_SLOW || json_boolean_stash
510 ? json_boolean_stash
511 : gv_stashpv ("JSON::XS::Boolean", 1);
512
513 if (SvSTASH (sv) == stash)
514 {
515 if (SvIV (sv))
516 encode_str (enc, "true", 4, 0);
517 else
518 encode_str (enc, "false", 5, 0);
289 } 519 }
290 else 520 else
291 { 521 {
292 SV *sv; 522#if 0
293 HE *he = hv_iternext (hv); 523 if (0 && sv_derived_from (rv, "JSON::Literal"))
294
295 for (;;)
296 {
297 INDENT;
298 encode_he (enc, he);
299
300 if (!(he = hv_iternext (hv)))
301 break;
302
303 COMMA;
304 } 524 {
305 525 // not yet
306 NL; 526 }
527#endif
528 if (enc->json.flags & F_CONV_BLESSED)
529 {
530 // we re-bless the reference to get overload and other niceties right
531 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
532
533 if (to_json)
534 {
535 dSP;
536
537 ENTER; SAVETMPS; PUSHMARK (SP);
538 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
539
540 // calling with G_SCALAR ensures that we always get a 1 return value
541 PUTBACK;
542 call_sv ((SV *)GvCV (to_json), G_SCALAR);
543 SPAGAIN;
544
545 // catch this surprisingly common error
546 if (SvROK (TOPs) && SvRV (TOPs) == sv)
547 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
548
549 sv = POPs;
550 PUTBACK;
551
552 encode_sv (enc, sv);
553
554 FREETMPS; LEAVE;
555 }
556 else if (enc->json.flags & F_ALLOW_BLESSED)
557 encode_str (enc, "null", 4, 0);
558 else
559 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
560 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
561 }
562 else if (enc->json.flags & F_ALLOW_BLESSED)
563 encode_str (enc, "null", 4, 0);
564 else
565 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
566 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
307 } 567 }
308 } 568 }
569 else if (svt == SVt_PVHV)
570 encode_hv (enc, (HV *)sv);
571 else if (svt == SVt_PVAV)
572 encode_av (enc, (AV *)sv);
573 else if (svt < SVt_PVAV)
574 {
575 STRLEN len = 0;
576 char *pv = svt ? SvPV (sv, len) : 0;
309 577
310 --enc->indent; INDENT; encode_ch (enc, '}'); 578 if (len == 1 && *pv == '1')
579 encode_str (enc, "true", 4, 0);
580 else if (len == 1 && *pv == '0')
581 encode_str (enc, "false", 5, 0);
582 else
583 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585 }
586 else
587 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
311} 589}
312 590
313static void 591static void
314encode_sv (enc_t *enc, SV *sv) 592encode_sv (enc_t *enc, SV *sv)
315{ 593{
594 SvGETMAGIC (sv);
595
316 if (SvPOKp (sv)) 596 if (SvPOKp (sv))
317 { 597 {
318 STRLEN len; 598 STRLEN len;
319 char *str = SvPV (sv, len); 599 char *str = SvPV (sv, len);
320 encode_ch (enc, '"'); 600 encode_ch (enc, '"');
321 encode_str (enc, str, len, SvUTF8 (sv)); 601 encode_str (enc, str, len, SvUTF8 (sv));
322 encode_ch (enc, '"'); 602 encode_ch (enc, '"');
323 } 603 }
324 else if (SvNOKp (sv)) 604 else if (SvNOKp (sv))
325 { 605 {
606 // trust that perl will do the right thing w.r.t. JSON syntax.
326 need (enc, NV_DIG + 32); 607 need (enc, NV_DIG + 32);
327 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 608 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
328 enc->cur += strlen (enc->cur); 609 enc->cur += strlen (enc->cur);
329 } 610 }
330 else if (SvIOKp (sv)) 611 else if (SvIOKp (sv))
331 { 612 {
332 need (enc, 64); 613 // we assume we can always read an IV as a UV
614 if (SvUV (sv) & ~(UV)0x7fff)
615 {
616 // large integer, use the (rather slow) snprintf way.
617 need (enc, sizeof (UV) * 3);
333 enc->cur += 618 enc->cur +=
334 SvIsUV(sv) 619 SvIsUV(sv)
335 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 620 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
336 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 621 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
622 }
623 else
624 {
625 // optimise the "small number case"
626 // code will likely be branchless and use only a single multiplication
627 I32 i = SvIV (sv);
628 U32 u;
629 char digit, nz = 0;
630
631 need (enc, 6);
632
633 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
634 u = i < 0 ? -i : i;
635
636 // convert to 4.28 fixed-point representation
637 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
638
639 // now output digit by digit, each time masking out the integer part
640 // and multiplying by 5 while moving the decimal point one to the right,
641 // resulting in a net multiplication by 10.
642 // we always write the digit to memory but conditionally increment
643 // the pointer, to ease the usage of conditional move instructions.
644 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
645 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
646 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
647 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
648 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649 }
337 } 650 }
338 else if (SvROK (sv)) 651 else if (SvROK (sv))
339 { 652 encode_rv (enc, SvRV (sv));
340 if (!--enc->max_recurse)
341 croak ("data structure too deep (hit recursion limit)");
342
343 sv = SvRV (sv);
344
345 switch (SvTYPE (sv))
346 {
347 case SVt_PVAV: encode_av (enc, (AV *)sv); break;
348 case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
349
350 default:
351 croak ("JSON can only represent references to arrays or hashes");
352 }
353 }
354 else if (!SvOK (sv)) 653 else if (!SvOK (sv))
355 encode_str (enc, "null", 4, 0); 654 encode_str (enc, "null", 4, 0);
356 else 655 else
357 croak ("encountered perl type that JSON cannot handle"); 656 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
657 SvPV_nolen (sv), SvFLAGS (sv));
358} 658}
359 659
360static SV * 660static SV *
361encode_json (SV *scalar, UV flags) 661encode_json (SV *scalar, JSON *json)
362{ 662{
363 enc_t enc; 663 enc_t enc;
364 enc.flags = flags; 664
665 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
666 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
667
668 enc.json = *json;
365 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 669 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
366 enc.cur = SvPVX (enc.sv); 670 enc.cur = SvPVX (enc.sv);
367 enc.end = SvEND (enc.sv); 671 enc.end = SvEND (enc.sv);
368 enc.max_recurse = 0;
369 enc.indent = 0; 672 enc.indent = 0;
673 enc.maxdepth = DEC_DEPTH (enc.json.flags);
370 674
371 SvPOK_only (enc.sv); 675 SvPOK_only (enc.sv);
372 encode_sv (&enc, scalar); 676 encode_sv (&enc, scalar);
373 677
678 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680
374 if (!(flags & (F_ASCII | F_UTF8))) 681 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
375 SvUTF8_on (enc.sv); 682 SvUTF8_on (enc.sv);
376 683
377 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 684 if (enc.json.flags & F_SHRINK)
685 shrink (enc.sv);
686
378 return enc.sv; 687 return enc.sv;
379} 688}
380 689
381///////////////////////////////////////////////////////////////////////////// 690/////////////////////////////////////////////////////////////////////////////
691// decoder
382 692
383#define WS \ 693// structure used for decoding JSON
694typedef struct
695{
696 char *cur; // current parser pointer
697 char *end; // end of input string
698 const char *err; // parse error, if != 0
699 JSON json;
700 U32 depth; // recursion depth
701 U32 maxdepth; // recursion depth limit
702} dec_t;
703
704inline void
705decode_comment (dec_t *dec)
706{
707 // only '#'-style comments allowed a.t.m.
708
709 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
710 ++dec->cur;
711}
712
713inline void
714decode_ws (dec_t *dec)
715{
384 for (;;) \ 716 for (;;)
385 { \ 717 {
386 char ch = *dec->cur; \ 718 char ch = *dec->cur;
719
387 if (ch > 0x20 \ 720 if (ch > 0x20)
721 {
722 if (expect_false (ch == '#'))
723 {
724 if (dec->json.flags & F_RELAXED)
725 decode_comment (dec);
726 else
727 break;
728 }
729 else
730 break;
731 }
388 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 732 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
389 break; \ 733 break; // parse error, but let higher level handle it, gives better error messages
734
390 ++dec->cur; \ 735 ++dec->cur;
391 } 736 }
737}
392 738
393#define ERR(reason) SB dec->err = reason; goto fail; SE 739#define ERR(reason) SB dec->err = reason; goto fail; SE
740
394#define EXPECT_CH(ch) SB \ 741#define EXPECT_CH(ch) SB \
395 if (*dec->cur != ch) \ 742 if (*dec->cur != ch) \
396 ERR (# ch " expected"); \ 743 ERR (# ch " expected"); \
397 ++dec->cur; \ 744 ++dec->cur; \
398 SE 745 SE
399 746
747#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
748#define DEC_DEC_DEPTH --dec->depth
749
400static SV *decode_sv (dec_t *dec); 750static SV *decode_sv (dec_t *dec);
401
402#define APPEND_CH(ch) SB \
403 SvGROW (sv, cur + 1 + 1); \
404 SvPVX (sv)[cur++] = (ch); \
405 SE
406 751
407static signed char decode_hexdigit[256]; 752static signed char decode_hexdigit[256];
408 753
409static UV 754static UV
410decode_4hex (dec_t *dec) 755decode_4hex (dec_t *dec)
411{ 756{
412 signed char d1, d2, d3, d4; 757 signed char d1, d2, d3, d4;
758 unsigned char *cur = (unsigned char *)dec->cur;
413 759
414 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 760 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
415 if (d1 < 0) ERR ("four hexadecimal digits expected"); 761 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
416 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 762 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
417 if (d2 < 0) ERR ("four hexadecimal digits expected"); 763 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
418 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
419 if (d3 < 0) ERR ("four hexadecimal digits expected");
420 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
421 if (d4 < 0) ERR ("four hexadecimal digits expected");
422 764
423 dec->cur += 4; 765 dec->cur += 4;
424 766
425 return ((UV)d1) << 12 767 return ((UV)d1) << 12
426 | ((UV)d2) << 8 768 | ((UV)d2) << 8
432} 774}
433 775
434static SV * 776static SV *
435decode_str (dec_t *dec) 777decode_str (dec_t *dec)
436{ 778{
437 SV *sv = NEWSV (0,2); 779 SV *sv = 0;
438 STRLEN cur = 0;
439 int utf8 = 0; 780 int utf8 = 0;
781 char *dec_cur = dec->cur;
440 782
441 for (;;) 783 do
442 { 784 {
443 unsigned char ch = *(unsigned char *)dec->cur; 785 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
786 char *cur = buf;
444 787
445 if (ch == '"') 788 do
446 break;
447 else if (ch == '\\')
448 { 789 {
449 switch (*++dec->cur) 790 unsigned char ch = *(unsigned char *)dec_cur++;
791
792 if (expect_false (ch == '"'))
793 {
794 --dec_cur;
795 break;
450 { 796 }
451 case '\\': 797 else if (expect_false (ch == '\\'))
452 case '/': 798 {
453 case '"': APPEND_CH (*dec->cur++); break; 799 switch (*dec_cur)
454
455 case 'b': APPEND_CH ('\010'); ++dec->cur; break;
456 case 't': APPEND_CH ('\011'); ++dec->cur; break;
457 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
458 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
459 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
460
461 case 'u':
462 { 800 {
463 UV lo, hi; 801 case '\\':
464 ++dec->cur; 802 case '/':
803 case '"': *cur++ = *dec_cur++; break;
465 804
466 hi = decode_4hex (dec); 805 case 'b': ++dec_cur; *cur++ = '\010'; break;
467 if (hi == (UV)-1) 806 case 't': ++dec_cur; *cur++ = '\011'; break;
468 goto fail; 807 case 'n': ++dec_cur; *cur++ = '\012'; break;
808 case 'f': ++dec_cur; *cur++ = '\014'; break;
809 case 'r': ++dec_cur; *cur++ = '\015'; break;
469 810
470 // possibly a surrogate pair 811 case 'u':
471 if (hi >= 0xd800 && hi < 0xdc00)
472 { 812 {
473 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 813 UV lo, hi;
474 ERR ("illegal surrogate character"); 814 ++dec_cur;
475 815
476 dec->cur += 2; 816 dec->cur = dec_cur;
477
478 lo = decode_4hex (dec); 817 hi = decode_4hex (dec);
818 dec_cur = dec->cur;
479 if (lo == (UV)-1) 819 if (hi == (UV)-1)
480 goto fail; 820 goto fail;
481 821
822 // possibly a surrogate pair
823 if (hi >= 0xd800)
824 if (hi < 0xdc00)
825 {
826 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
827 ERR ("missing low surrogate character in surrogate pair");
828
829 dec_cur += 2;
830
831 dec->cur = dec_cur;
832 lo = decode_4hex (dec);
833 dec_cur = dec->cur;
834 if (lo == (UV)-1)
835 goto fail;
836
482 if (lo < 0xdc00 || lo >= 0xe000) 837 if (lo < 0xdc00 || lo >= 0xe000)
483 ERR ("surrogate pair expected"); 838 ERR ("surrogate pair expected");
484 839
485 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 840 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
841 }
842 else if (hi < 0xe000)
843 ERR ("missing high surrogate character in surrogate pair");
844
845 if (hi >= 0x80)
846 {
847 utf8 = 1;
848
849 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
850 }
851 else
852 *cur++ = hi;
486 } 853 }
487 else if (lo >= 0xdc00 && lo < 0xe000)
488 ERR ("illegal surrogate character");
489
490 if (hi >= 0x80)
491 { 854 break;
492 utf8 = 1;
493 855
494 SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
495 cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
496 }
497 else 856 default:
498 APPEND_CH (hi); 857 --dec_cur;
858 ERR ("illegal backslash escape sequence in string");
499 } 859 }
860 }
861 else if (expect_true (ch >= 0x20 && ch <= 0x7f))
862 *cur++ = ch;
863 else if (ch >= 0x80)
864 {
865 STRLEN clen;
866 UV uch;
867
868 --dec_cur;
869
870 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
871 if (clen == (STRLEN)-1)
872 ERR ("malformed UTF-8 character in JSON string");
873
874 do
875 *cur++ = *dec_cur++;
876 while (--clen);
877
878 utf8 = 1;
879 }
880 else
881 {
882 --dec_cur;
883
884 if (!ch)
885 ERR ("unexpected end of string while parsing JSON string");
500 break; 886 else
887 ERR ("invalid character encountered while parsing JSON string");
501 } 888 }
502 } 889 }
503 else if (ch >= 0x20 && ch <= 0x7f) 890 while (cur < buf + SHORT_STRING_LEN);
504 APPEND_CH (*dec->cur++); 891
505 else if (ch >= 0x80) 892 {
893 STRLEN len = cur - buf;
894
895 if (sv)
506 { 896 {
507 STRLEN clen; 897 SvGROW (sv, SvCUR (sv) + len + 1);
508 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 898 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
509 if (clen < 0) 899 SvCUR_set (sv, SvCUR (sv) + len);
510 ERR ("malformed UTF-8 character in string, cannot convert to JSON");
511
512 SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
513 memcpy (SvPVX (sv) + cur, dec->cur, clen);
514 dec->cur += clen;
515 } 900 }
516 else 901 else
517 ERR ("invalid character encountered"); 902 sv = newSVpvn (buf, len);
518 } 903 }
904 }
905 while (*dec_cur != '"');
519 906
520 ++dec->cur; 907 ++dec_cur;
521 908
909 if (sv)
910 {
522 SvPOK_only (sv); 911 SvPOK_only (sv);
523
524 SvCUR_set (sv, cur);
525 *SvEND (sv) = 0; 912 *SvEND (sv) = 0;
526 913
527 if (utf8) 914 if (utf8)
528 SvUTF8_on (sv); 915 SvUTF8_on (sv);
916 }
917 else
918 sv = newSVpvn ("", 0);
529 919
920 dec->cur = dec_cur;
530 return sv; 921 return sv;
531 922
532fail: 923fail:
533 SvREFCNT_dec (sv); 924 dec->cur = dec_cur;
534 return 0; 925 return 0;
535} 926}
536 927
537static SV * 928static SV *
538decode_num (dec_t *dec) 929decode_num (dec_t *dec)
548 { 939 {
549 ++dec->cur; 940 ++dec->cur;
550 if (*dec->cur >= '0' && *dec->cur <= '9') 941 if (*dec->cur >= '0' && *dec->cur <= '9')
551 ERR ("malformed number (leading zero must not be followed by another digit)"); 942 ERR ("malformed number (leading zero must not be followed by another digit)");
552 } 943 }
553 944 else if (*dec->cur < '0' || *dec->cur > '9')
554 // int 945 ERR ("malformed number (no digits after initial minus)");
946 else
947 do
948 {
949 ++dec->cur;
950 }
555 while (*dec->cur >= '0' && *dec->cur <= '9') 951 while (*dec->cur >= '0' && *dec->cur <= '9');
556 ++dec->cur;
557 952
558 // [frac] 953 // [frac]
559 if (*dec->cur == '.') 954 if (*dec->cur == '.')
560 { 955 {
561 is_nv = 1; 956 ++dec->cur;
957
958 if (*dec->cur < '0' || *dec->cur > '9')
959 ERR ("malformed number (no digits after decimal point)");
562 960
563 do 961 do
564 { 962 {
565 ++dec->cur; 963 ++dec->cur;
566 } 964 }
567 while (*dec->cur >= '0' && *dec->cur <= '9'); 965 while (*dec->cur >= '0' && *dec->cur <= '9');
966
967 is_nv = 1;
568 } 968 }
569 969
570 // [exp] 970 // [exp]
571 if (*dec->cur == 'e' || *dec->cur == 'E') 971 if (*dec->cur == 'e' || *dec->cur == 'E')
572 { 972 {
573 is_nv = 1;
574
575 ++dec->cur; 973 ++dec->cur;
974
576 if (*dec->cur == '-' || *dec->cur == '+') 975 if (*dec->cur == '-' || *dec->cur == '+')
577 ++dec->cur; 976 ++dec->cur;
578 977
978 if (*dec->cur < '0' || *dec->cur > '9')
979 ERR ("malformed number (no digits after exp sign)");
980
981 do
982 {
983 ++dec->cur;
984 }
579 while (*dec->cur >= '0' && *dec->cur <= '9') 985 while (*dec->cur >= '0' && *dec->cur <= '9');
580 ++dec->cur; 986
987 is_nv = 1;
581 } 988 }
582 989
583 if (!is_nv) 990 if (!is_nv)
584 { 991 {
585 UV uv; 992 int len = dec->cur - start;
586 int numtype = grok_number (start, dec->cur - start, &uv); 993
587 if (numtype & IS_NUMBER_IN_UV) 994 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
588 if (numtype & IS_NUMBER_NEG) 995 if (*start == '-')
996 switch (len)
589 { 997 {
590 if (uv < (UV)IV_MIN) 998 case 2: return newSViv (-( start [1] - '0' * 1));
591 return newSViv (-(IV)uv); 999 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
1000 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1001 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
592 } 1002 }
1003 else
1004 switch (len)
1005 {
1006 case 1: return newSViv ( start [0] - '0' * 1);
1007 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1008 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1009 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1010 }
1011
1012 {
1013 UV uv;
1014 int numtype = grok_number (start, len, &uv);
1015 if (numtype & IS_NUMBER_IN_UV)
1016 if (numtype & IS_NUMBER_NEG)
1017 {
1018 if (uv < (UV)IV_MIN)
1019 return newSViv (-(IV)uv);
1020 }
593 else 1021 else
594 return newSVuv (uv); 1022 return newSVuv (uv);
595 } 1023 }
596 1024
1025 len -= *start == '-' ? 1 : 0;
1026
1027 // does not fit into IV or UV, try NV
1028 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
1029 #if defined (LDBL_DIG)
1030 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1031 #endif
1032 )
1033 // fits into NV without loss of precision
1034 return newSVnv (Atof (start));
1035
1036 // everything else fails, convert it to a string
1037 return newSVpvn (start, dec->cur - start);
1038 }
1039
1040 // loss of precision here
597 return newSVnv (Atof (start)); 1041 return newSVnv (Atof (start));
598 1042
599fail: 1043fail:
600 return 0; 1044 return 0;
601} 1045}
603static SV * 1047static SV *
604decode_av (dec_t *dec) 1048decode_av (dec_t *dec)
605{ 1049{
606 AV *av = newAV (); 1050 AV *av = newAV ();
607 1051
1052 DEC_INC_DEPTH;
1053 decode_ws (dec);
1054
1055 if (*dec->cur == ']')
1056 ++dec->cur;
1057 else
608 for (;;) 1058 for (;;)
609 { 1059 {
610 SV *value; 1060 SV *value;
611 1061
612 value = decode_sv (dec); 1062 value = decode_sv (dec);
613 if (!value) 1063 if (!value)
614 goto fail; 1064 goto fail;
615 1065
616 av_push (av, value); 1066 av_push (av, value);
617 1067
618 WS; 1068 decode_ws (dec);
619 1069
620 if (*dec->cur == ']') 1070 if (*dec->cur == ']')
621 { 1071 {
622 ++dec->cur; 1072 ++dec->cur;
623 break; 1073 break;
1074 }
624 } 1075
625
626 if (*dec->cur != ',') 1076 if (*dec->cur != ',')
627 ERR (", or ] expected while parsing array"); 1077 ERR (", or ] expected while parsing array");
628 1078
629 ++dec->cur; 1079 ++dec->cur;
1080
1081 decode_ws (dec);
1082
1083 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1084 {
1085 ++dec->cur;
1086 break;
1087 }
630 } 1088 }
631 1089
1090 DEC_DEC_DEPTH;
632 return newRV_noinc ((SV *)av); 1091 return newRV_noinc ((SV *)av);
633 1092
634fail: 1093fail:
635 SvREFCNT_dec (av); 1094 SvREFCNT_dec (av);
1095 DEC_DEC_DEPTH;
636 return 0; 1096 return 0;
637} 1097}
638 1098
639static SV * 1099static SV *
640decode_hv (dec_t *dec) 1100decode_hv (dec_t *dec)
641{ 1101{
1102 SV *sv;
642 HV *hv = newHV (); 1103 HV *hv = newHV ();
643 1104
1105 DEC_INC_DEPTH;
1106 decode_ws (dec);
1107
1108 if (*dec->cur == '}')
1109 ++dec->cur;
1110 else
644 for (;;) 1111 for (;;)
645 { 1112 {
646 SV *key, *value;
647
648 WS; EXPECT_CH ('"'); 1113 EXPECT_CH ('"');
649 1114
650 key = decode_str (dec); 1115 // heuristic: assume that
651 if (!key) 1116 // a) decode_str + hv_store_ent are abysmally slow.
652 goto fail; 1117 // b) most hash keys are short, simple ascii text.
653 1118 // => try to "fast-match" such strings to avoid
654 WS; EXPECT_CH (':'); 1119 // the overhead of decode_str + hv_store_ent.
655
656 value = decode_sv (dec);
657 if (!value)
658 { 1120 {
1121 SV *value;
1122 char *p = dec->cur;
1123 char *e = p + 24; // only try up to 24 bytes
1124
1125 for (;;)
1126 {
1127 // the >= 0x80 is true on most architectures
1128 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1129 {
1130 // slow path, back up and use decode_str
1131 SV *key = decode_str (dec);
1132 if (!key)
1133 goto fail;
1134
1135 decode_ws (dec); EXPECT_CH (':');
1136
1137 decode_ws (dec);
1138 value = decode_sv (dec);
1139 if (!value)
1140 {
1141 SvREFCNT_dec (key);
1142 goto fail;
1143 }
1144
1145 hv_store_ent (hv, key, value, 0);
659 SvREFCNT_dec (key); 1146 SvREFCNT_dec (key);
1147
1148 break;
1149 }
1150 else if (*p == '"')
1151 {
1152 // fast path, got a simple key
1153 char *key = dec->cur;
1154 int len = p - key;
1155 dec->cur = p + 1;
1156
1157 decode_ws (dec); EXPECT_CH (':');
1158
1159 decode_ws (dec);
1160 value = decode_sv (dec);
1161 if (!value)
660 goto fail; 1162 goto fail;
1163
1164 hv_store (hv, key, len, value, 0);
1165
1166 break;
1167 }
1168
1169 ++p;
1170 }
661 } 1171 }
662 1172
663 //TODO: optimise 1173 decode_ws (dec);
664 hv_store_ent (hv, key, value, 0);
665 1174
666 WS;
667
668 if (*dec->cur == '}') 1175 if (*dec->cur == '}')
1176 {
1177 ++dec->cur;
1178 break;
1179 }
1180
1181 if (*dec->cur != ',')
1182 ERR (", or } expected while parsing object/hash");
1183
1184 ++dec->cur;
1185
1186 decode_ws (dec);
1187
1188 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1189 {
1190 ++dec->cur;
1191 break;
1192 }
1193 }
1194
1195 DEC_DEC_DEPTH;
1196 sv = newRV_noinc ((SV *)hv);
1197
1198 // check filter callbacks
1199 if (dec->json.flags & F_HOOK)
1200 {
1201 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
669 { 1202 {
670 ++dec->cur; 1203 HE *cb, *he;
671 break; 1204
1205 hv_iterinit (hv);
1206 he = hv_iternext (hv);
1207 hv_iterinit (hv);
1208
1209 // the next line creates a mortal sv each time its called.
1210 // might want to optimise this for common cases.
1211 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1212
1213 if (cb)
1214 {
1215 dSP;
1216 int count;
1217
1218 ENTER; SAVETMPS; PUSHMARK (SP);
1219 XPUSHs (HeVAL (he));
1220
1221 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1222
1223 if (count == 1)
1224 {
1225 sv = newSVsv (POPs);
1226 FREETMPS; LEAVE;
1227 return sv;
1228 }
1229
1230 FREETMPS; LEAVE;
1231 }
672 } 1232 }
673 1233
674 if (*dec->cur != ',') 1234 if (dec->json.cb_object)
675 ERR (", or } expected while parsing object/hash"); 1235 {
1236 dSP;
1237 int count;
676 1238
677 ++dec->cur; 1239 ENTER; SAVETMPS; PUSHMARK (SP);
678 } 1240 XPUSHs (sv_2mortal (sv));
679 1241
680 return newRV_noinc ((SV *)hv); 1242 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1243
1244 if (count == 1)
1245 {
1246 sv = newSVsv (POPs);
1247 FREETMPS; LEAVE;
1248 return sv;
1249 }
1250
1251 SvREFCNT_inc (sv);
1252 FREETMPS; LEAVE;
1253 }
1254 }
1255
1256 return sv;
681 1257
682fail: 1258fail:
683 SvREFCNT_dec (hv); 1259 SvREFCNT_dec (hv);
1260 DEC_DEC_DEPTH;
684 return 0; 1261 return 0;
685} 1262}
686 1263
687static SV * 1264static SV *
688decode_sv (dec_t *dec) 1265decode_sv (dec_t *dec)
689{ 1266{
690 WS; 1267 // the beauty of JSON: you need exactly one character lookahead
1268 // to parse anything.
691 switch (*dec->cur) 1269 switch (*dec->cur)
692 { 1270 {
693 case '"': ++dec->cur; return decode_str (dec); 1271 case '"': ++dec->cur; return decode_str (dec);
694 case '[': ++dec->cur; return decode_av (dec); 1272 case '[': ++dec->cur; return decode_av (dec);
695 case '{': ++dec->cur; return decode_hv (dec); 1273 case '{': ++dec->cur; return decode_hv (dec);
701 1279
702 case 't': 1280 case 't':
703 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1281 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
704 { 1282 {
705 dec->cur += 4; 1283 dec->cur += 4;
706 return newSViv (1); 1284#if JSON_SLOW
1285 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
1286#endif
1287 return SvREFCNT_inc (json_true);
707 } 1288 }
708 else 1289 else
709 ERR ("'true' expected"); 1290 ERR ("'true' expected");
710 1291
711 break; 1292 break;
712 1293
713 case 'f': 1294 case 'f':
714 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1295 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
715 { 1296 {
716 dec->cur += 5; 1297 dec->cur += 5;
717 return newSViv (0); 1298#if JSON_SLOW
1299 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1300#endif
1301 return SvREFCNT_inc (json_false);
718 } 1302 }
719 else 1303 else
720 ERR ("'false' expected"); 1304 ERR ("'false' expected");
721 1305
722 break; 1306 break;
723 1307
724 case 'n': 1308 case 'n':
725 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1309 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
726 { 1310 {
727 dec->cur += 4; 1311 dec->cur += 4;
728 return newSViv (1); 1312 return newSVsv (&PL_sv_undef);
729 } 1313 }
730 else 1314 else
731 ERR ("'null' expected"); 1315 ERR ("'null' expected");
732 1316
733 break; 1317 break;
734 1318
735 default: 1319 default:
736 ERR ("malformed json string"); 1320 ERR ("malformed JSON string, neither array, object, number, string or atom");
737 break; 1321 break;
738 } 1322 }
739 1323
740fail: 1324fail:
741 return 0; 1325 return 0;
742} 1326}
743 1327
744static SV * 1328static SV *
745decode_json (SV *string, UV flags) 1329decode_json (SV *string, JSON *json, UV *offset_return)
746{ 1330{
1331 dec_t dec;
1332 UV offset;
747 SV *sv; 1333 SV *sv;
748 1334
1335 SvGETMAGIC (string);
1336 SvUPGRADE (string, SVt_PV);
1337
1338 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1339 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1340 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1341
749 if (!(flags & F_UTF8)) 1342 if (json->flags & F_UTF8)
1343 sv_utf8_downgrade (string, 0);
1344 else
750 sv_utf8_upgrade (string); 1345 sv_utf8_upgrade (string);
751 1346
752 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1347 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
753 1348
754 dec_t dec; 1349 dec.json = *json;
755 dec.flags = flags;
756 dec.cur = SvPVX (string); 1350 dec.cur = SvPVX (string);
757 dec.end = SvEND (string); 1351 dec.end = SvEND (string);
758 dec.err = 0; 1352 dec.err = 0;
1353 dec.depth = 0;
1354 dec.maxdepth = DEC_DEPTH (dec.json.flags);
759 1355
760 *dec.end = 1; // invalid anywhere 1356 if (dec.json.cb_object || dec.json.cb_sk_object)
1357 dec.json.flags |= F_HOOK;
1358
1359 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1360
1361 decode_ws (&dec);
761 sv = decode_sv (&dec); 1362 sv = decode_sv (&dec);
762 *dec.end = 0; 1363
1364 if (!(offset_return || !sv))
1365 {
1366 // check for trailing garbage
1367 decode_ws (&dec);
1368
1369 if (*dec.cur)
1370 {
1371 dec.err = "garbage after JSON object";
1372 SvREFCNT_dec (sv);
1373 sv = 0;
1374 }
1375 }
1376
1377 if (offset_return || !sv)
1378 {
1379 offset = dec.json.flags & F_UTF8
1380 ? dec.cur - SvPVX (string)
1381 : utf8_distance (dec.cur, SvPVX (string));
1382
1383 if (offset_return)
1384 *offset_return = offset;
1385 }
763 1386
764 if (!sv) 1387 if (!sv)
765 { 1388 {
766 IV offset = utf8_distance (dec.cur, SvPVX (string));
767 SV *uni = sv_newmortal (); 1389 SV *uni = sv_newmortal ();
768 1390
1391 // horrible hack to silence warning inside pv_uni_display
1392 COP cop = *PL_curcop;
1393 cop.cop_warnings = pWARN_NONE;
1394 ENTER;
1395 SAVEVPTR (PL_curcop);
1396 PL_curcop = &cop;
769 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1397 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1398 LEAVE;
1399
770 croak ("%s, at character %d (%s)", 1400 croak ("%s, at character offset %d [\"%s\"]",
771 dec.err, 1401 dec.err,
772 (int)offset, 1402 (int)offset,
773 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1403 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
774 } 1404 }
775 1405
776 sv_dump (sv);//D
777 return sv_2mortal (sv); 1406 sv = sv_2mortal (sv);
1407
1408 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1409 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1410
1411 return sv;
778} 1412}
1413
1414/////////////////////////////////////////////////////////////////////////////
1415// XS interface functions
779 1416
780MODULE = JSON::XS PACKAGE = JSON::XS 1417MODULE = JSON::XS PACKAGE = JSON::XS
781 1418
782BOOT: 1419BOOT:
783{ 1420{
784 int i; 1421 int i;
785 1422
786 memset (decode_hexdigit, 0xff, 256);
787 for (i = 10; i--; ) 1423 for (i = 0; i < 256; ++i)
788 decode_hexdigit ['0' + i] = i; 1424 decode_hexdigit [i] =
1425 i >= '0' && i <= '9' ? i - '0'
1426 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1427 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1428 : -1;
789 1429
790 for (i = 6; --i; ) 1430 json_stash = gv_stashpv ("JSON::XS" , 1);
1431 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1432
1433 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
1434 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1435}
1436
1437PROTOTYPES: DISABLE
1438
1439void CLONE (...)
1440 CODE:
1441 json_stash = 0;
1442 json_boolean_stash = 0;
1443
1444void new (char *klass)
1445 PPCODE:
1446{
1447 SV *pv = NEWSV (0, sizeof (JSON));
1448 SvPOK_only (pv);
1449 Zero (SvPVX (pv), 1, JSON);
1450 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1451 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
1452}
1453
1454void ascii (JSON *self, int enable = 1)
1455 ALIAS:
1456 ascii = F_ASCII
1457 latin1 = F_LATIN1
1458 utf8 = F_UTF8
1459 indent = F_INDENT
1460 canonical = F_CANONICAL
1461 space_before = F_SPACE_BEFORE
1462 space_after = F_SPACE_AFTER
1463 pretty = F_PRETTY
1464 allow_nonref = F_ALLOW_NONREF
1465 shrink = F_SHRINK
1466 allow_blessed = F_ALLOW_BLESSED
1467 convert_blessed = F_CONV_BLESSED
1468 relaxed = F_RELAXED
1469 PPCODE:
1470{
1471 if (enable)
1472 self->flags |= ix;
1473 else
1474 self->flags &= ~ix;
1475
1476 XPUSHs (ST (0));
1477}
1478
1479void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1480 PPCODE:
1481{
1482 UV log2 = 0;
1483
1484 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1485
1486 while ((1UL << log2) < max_depth)
1487 ++log2;
1488
1489 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1490
1491 XPUSHs (ST (0));
1492}
1493
1494void max_size (JSON *self, UV max_size = 0)
1495 PPCODE:
1496{
1497 UV log2 = 0;
1498
1499 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1500 if (max_size == 1) max_size = 2;
1501
1502 while ((1UL << log2) < max_size)
1503 ++log2;
1504
1505 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1506
1507 XPUSHs (ST (0));
1508}
1509
1510void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1511 PPCODE:
1512{
1513 SvREFCNT_dec (self->cb_object);
1514 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1515
1516 XPUSHs (ST (0));
1517}
1518
1519void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1520 PPCODE:
1521{
1522 if (!self->cb_sk_object)
1523 self->cb_sk_object = newHV ();
1524
1525 if (SvOK (cb))
1526 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1527 else
791 { 1528 {
792 decode_hexdigit ['a' + i] = 10 + i; 1529 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
793 decode_hexdigit ['A' + i] = 10 + i; 1530
1531 if (!HvKEYS (self->cb_sk_object))
1532 {
1533 SvREFCNT_dec (self->cb_sk_object);
1534 self->cb_sk_object = 0;
1535 }
794 } 1536 }
795 1537
796 json_stash = gv_stashpv ("JSON::XS", 1); 1538 XPUSHs (ST (0));
797} 1539}
798 1540
799SV *new (char *dummy) 1541void encode (JSON *self, SV *scalar)
1542 PPCODE:
1543 XPUSHs (encode_json (scalar, self));
1544
1545void decode (JSON *self, SV *jsonstr)
1546 PPCODE:
1547 XPUSHs (decode_json (jsonstr, self, 0));
1548
1549void decode_prefix (JSON *self, SV *jsonstr)
1550 PPCODE:
1551{
1552 UV offset;
1553 EXTEND (SP, 2);
1554 PUSHs (decode_json (jsonstr, self, &offset));
1555 PUSHs (sv_2mortal (newSVuv (offset)));
1556}
1557
1558void DESTROY (JSON *self)
800 CODE: 1559 CODE:
801 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1560 SvREFCNT_dec (self->cb_sk_object);
802 OUTPUT: 1561 SvREFCNT_dec (self->cb_object);
803 RETVAL
804 1562
805SV *ascii (SV *self, int enable) 1563PROTOTYPES: ENABLE
806 ALIAS:
807 ascii = F_ASCII
808 utf8 = F_UTF8
809 indent = F_INDENT
810 canonical = F_CANONICAL
811 space_before = F_SPACE_BEFORE
812 space_after = F_SPACE_AFTER
813 json_rpc = F_JSON_RPC
814 CODE:
815{
816 UV *uv = SvJSON (self);
817 if (enable)
818 *uv |= ix;
819 else
820 *uv &= ~ix;
821 1564
822 RETVAL = newSVsv (self); 1565void to_json (SV *scalar)
823}
824 OUTPUT:
825 RETVAL
826
827void encode (SV *self, SV *scalar)
828 PPCODE: 1566 PPCODE:
1567{
1568 JSON json = { F_DEFAULT | F_UTF8 };
829 XPUSHs (encode_json (scalar, *SvJSON (self))); 1569 XPUSHs (encode_json (scalar, &json));
1570}
830 1571
831void decode (SV *self, SV *jsondata) 1572void from_json (SV *jsonstr)
832 PPCODE: 1573 PPCODE:
1574{
1575 JSON json = { F_DEFAULT | F_UTF8 };
833 XPUSHs (decode_json (jsondata, *SvJSON (self))); 1576 XPUSHs (decode_json (jsonstr, &json, 0));
1577}
834 1578

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines