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 1.84 by root, Thu Mar 27 06:37:35 2008 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 <limits.h>
10#include <float.h>
8 11
12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif
15
16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13
20#endif
21
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
23
9#define F_ASCII 0x00000001 24#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 26#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 27#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 28#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 29#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 30#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040 31#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL
36#define F_ALLOW_UNKNOWN 0x00002000UL
16 37
17#define F_DEFAULT 0 38#define F_MAXDEPTH 0xf8000000UL
39#define S_MAXDEPTH 27
40#define F_MAXSIZE 0x01f00000UL
41#define S_MAXSIZE 20
42#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
43
44#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
45#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
46
47#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
48#define F_DEFAULT (9UL << S_MAXDEPTH)
18 49
19#define INIT_SIZE 32 // initial scalar size to be allocated 50#define INIT_SIZE 32 // initial scalar size to be allocated
51#define INDENT_STEP 3 // spaces per indentation level
52
53#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
20 54
21#define SB do { 55#define SB do {
22#define SE } while (0) 56#define SE } while (0)
23 57
24static HV *json_stash; 58#if __GNUC__ >= 3
59# define expect(expr,value) __builtin_expect ((expr), (value))
60# define INLINE static inline
61#else
62# define expect(expr,value) (expr)
63# define INLINE static
64#endif
65
66#define expect_false(expr) expect ((expr) != 0, 0)
67#define expect_true(expr) expect ((expr) != 0, 1)
68
69#define IN_RANGE_INC(type,val,beg,end) \
70 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
71 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
72
73#ifdef USE_ITHREADS
74# define JSON_SLOW 1
75# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
76#else
77# define JSON_SLOW 0
78# define JSON_STASH json_stash
79#endif
80
81static HV *json_stash, *json_boolean_stash; // JSON::XS::
82static SV *json_true, *json_false;
83
84enum {
85 INCR_M_WS = 0, // initial whitespace skipping, must be 0
86 INCR_M_STR, // inside string
87 INCR_M_BS, // inside backslash
88 INCR_M_JSON // outside anything, count nesting
89};
90
91#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
92
93typedef struct {
94 U32 flags;
95 SV *cb_object;
96 HV *cb_sk_object;
97
98 // for the incremental parser
99 SV *incr_text; // the source text so far
100 STRLEN incr_pos; // the current offset into the text
101 int incr_nest; // {[]}-nesting level
102 int incr_mode;
103} JSON;
104
105/////////////////////////////////////////////////////////////////////////////
106// utility functions
107
108INLINE SV *
109get_bool (const char *name)
110{
111 SV *sv = get_sv (name, 1);
112
113 SvREADONLY_on (sv);
114 SvREADONLY_on (SvRV (sv));
115
116 return sv;
117}
118
119INLINE void
120shrink (SV *sv)
121{
122 sv_utf8_downgrade (sv, 1);
123 if (SvLEN (sv) > SvCUR (sv) + 1)
124 {
125#ifdef SvPV_shrink_to_cur
126 SvPV_shrink_to_cur (sv);
127#elif defined (SvPV_renew)
128 SvPV_renew (sv, SvCUR (sv) + 1);
129#endif
130 }
131}
132
133// decode an utf-8 character and return it, or (UV)-1 in
134// case of an error.
135// we special-case "safe" characters from U+80 .. U+7FF,
136// but use the very good perl function to parse anything else.
137// note that we never call this function for a ascii codepoints
138INLINE UV
139decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
140{
141 if (expect_true (len >= 2
142 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
143 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
144 {
145 *clen = 2;
146 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
147 }
148 else
149 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
150}
151
152// likewise for encoding, also never called for ascii codepoints
153// this function takes advantage of this fact, although current gccs
154// seem to optimise the check for >= 0x80 away anyways
155INLINE unsigned char *
156encode_utf8 (unsigned char *s, UV ch)
157{
158 if (expect_false (ch < 0x000080))
159 *s++ = ch;
160 else if (expect_true (ch < 0x000800))
161 *s++ = 0xc0 | ( ch >> 6),
162 *s++ = 0x80 | ( ch & 0x3f);
163 else if ( ch < 0x010000)
164 *s++ = 0xe0 | ( ch >> 12),
165 *s++ = 0x80 | ((ch >> 6) & 0x3f),
166 *s++ = 0x80 | ( ch & 0x3f);
167 else if ( ch < 0x110000)
168 *s++ = 0xf0 | ( ch >> 18),
169 *s++ = 0x80 | ((ch >> 12) & 0x3f),
170 *s++ = 0x80 | ((ch >> 6) & 0x3f),
171 *s++ = 0x80 | ( ch & 0x3f);
172
173 return s;
174}
175
176/////////////////////////////////////////////////////////////////////////////
177// encoder
25 178
26// structure used for encoding JSON 179// structure used for encoding JSON
27typedef struct 180typedef struct
28{ 181{
29 char *cur; 182 char *cur; // SvPVX (sv) + current output position
30 STRLEN len; // SvLEN (sv)
31 char *end; // SvEND (sv) 183 char *end; // SvEND (sv)
32 SV *sv; 184 SV *sv; // result scalar
33 UV flags; 185 JSON json;
34 int max_recurse; 186 U32 indent; // indentation level
35 int indent; 187 U32 maxdepth; // max. indentation/recursion level
188 UV limit; // escape character values >= this value when encoding
36} enc_t; 189} enc_t;
37 190
38// structure used for decoding JSON 191INLINE 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) 192need (enc_t *enc, STRLEN len)
60{ 193{
61 if (enc->cur + len >= enc->end) 194 if (expect_false (enc->cur + len >= enc->end))
62 { 195 {
63 STRLEN cur = enc->cur - SvPVX (enc->sv); 196 STRLEN cur = enc->cur - SvPVX (enc->sv);
64 SvGROW (enc->sv, cur + len + 1); 197 SvGROW (enc->sv, cur + len + 1);
65 enc->cur = SvPVX (enc->sv) + cur; 198 enc->cur = SvPVX (enc->sv) + cur;
66 enc->end = SvEND (enc->sv); 199 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
67 } 200 }
68} 201}
69 202
70static void 203INLINE void
71encode_ch (enc_t *enc, char ch) 204encode_ch (enc_t *enc, char ch)
72{ 205{
73 need (enc, 1); 206 need (enc, 1);
74 *enc->cur++ = ch; 207 *enc->cur++ = ch;
75} 208}
… …
77static void 210static void
78encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) 211encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
79{ 212{
80 char *end = str + len; 213 char *end = str + len;
81 214
215 need (enc, len);
216
82 while (str < end) 217 while (str < end)
83 { 218 {
84 unsigned char ch = *(unsigned char *)str; 219 unsigned char ch = *(unsigned char *)str;
220
85 if (ch >= 0x20 && ch < 0x80) // most common case 221 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
86 { 222 {
223 if (expect_false (ch == '"')) // but with slow exceptions
224 {
225 need (enc, len += 1);
226 *enc->cur++ = '\\';
227 *enc->cur++ = '"';
228 }
229 else if (expect_false (ch == '\\'))
230 {
231 need (enc, len += 1);
232 *enc->cur++ = '\\';
233 *enc->cur++ = '\\';
234 }
235 else
87 *enc->cur++ = ch; 236 *enc->cur++ = ch;
237
88 str++; 238 ++str;
89 } 239 }
90 else 240 else
91 { 241 {
92 STRLEN clen; 242 switch (ch)
93 UV uch;
94
95 if (is_utf8)
96 { 243 {
97 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 244 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
98 if (clen < 0) 245 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
99 croak ("malformed UTF-8 character in string, cannot convert to JSON"); 246 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
100 } 247 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
101 else 248 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
102 {
103 uch = ch;
104 clen = 1;
105 }
106 249
107 need (enc, len += 6); 250 default:
108
109 if (uch < 0xa0 || enc->flags & F_ASCII)
110 {
111 if (uch > 0xFFFFUL)
112 { 251 {
252 STRLEN clen;
253 UV uch;
254
255 if (is_utf8)
256 {
257 uch = decode_utf8 (str, end - str, &clen);
258 if (clen == (STRLEN)-1)
259 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
260 }
261 else
262 {
263 uch = ch;
113 len += 6; 264 clen = 1;
265 }
266
267 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
268 {
269 if (uch >= 0x10000UL)
270 {
271 if (uch >= 0x110000UL)
272 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
273
114 need (enc, len += 6); 274 need (enc, len += 11);
115 sprintf (enc->cur, "\\u%04x\\u%04x", 275 sprintf (enc->cur, "\\u%04x\\u%04x",
116 (uch - 0x10000) / 0x400 + 0xD800, 276 (int)((uch - 0x10000) / 0x400 + 0xD800),
117 (uch - 0x10000) % 0x400 + 0xDC00); 277 (int)((uch - 0x10000) % 0x400 + 0xDC00));
118 enc->cur += 12; 278 enc->cur += 12;
279 }
280 else
281 {
282 static char hexdigit [16] = "0123456789abcdef";
283 need (enc, len += 5);
284 *enc->cur++ = '\\';
285 *enc->cur++ = 'u';
286 *enc->cur++ = hexdigit [ uch >> 12 ];
287 *enc->cur++ = hexdigit [(uch >> 8) & 15];
288 *enc->cur++ = hexdigit [(uch >> 4) & 15];
289 *enc->cur++ = hexdigit [(uch >> 0) & 15];
290 }
291
292 str += clen;
119 } 293 }
294 else if (enc->json.flags & F_LATIN1)
295 {
296 *enc->cur++ = uch;
297 str += clen;
298 }
299 else if (is_utf8)
300 {
301 need (enc, len += clen);
302 do
303 {
304 *enc->cur++ = *str++;
305 }
306 while (--clen);
307 }
120 else 308 else
121 { 309 {
122 sprintf (enc->cur, "\\u%04x", uch); 310 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
123 enc->cur += 6; 311 enc->cur = encode_utf8 (enc->cur, uch);
312 ++str;
313 }
124 } 314 }
125 } 315 }
126 else if (is_utf8)
127 {
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 } 316 }
136 317
137 --len; 318 --len;
138 } 319 }
139} 320}
140 321
141#define INDENT SB \ 322INLINE void
323encode_indent (enc_t *enc)
324{
142 if (enc->flags & F_INDENT) \ 325 if (enc->json.flags & F_INDENT)
143 { \ 326 {
144 int i_; \ 327 int spaces = enc->indent * INDENT_STEP;
145 need (enc, enc->indent); \ 328
146 for (i_ = enc->indent * 3; i_--; )\ 329 need (enc, spaces);
330 memset (enc->cur, ' ', spaces);
331 enc->cur += spaces;
332 }
333}
334
335INLINE void
336encode_space (enc_t *enc)
337{
338 need (enc, 1);
339 encode_ch (enc, ' ');
340}
341
342INLINE void
343encode_nl (enc_t *enc)
344{
345 if (enc->json.flags & F_INDENT)
346 {
347 need (enc, 1);
147 encode_ch (enc, ' '); \ 348 encode_ch (enc, '\n');
148 } \ 349 }
149 SE 350}
150 351
151#define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE 352INLINE void
152#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 353encode_comma (enc_t *enc)
153#define COMMA SB \ 354{
154 encode_ch (enc, ','); \ 355 encode_ch (enc, ',');
356
155 if (enc->flags & F_INDENT) \ 357 if (enc->json.flags & F_INDENT)
156 NL; \ 358 encode_nl (enc);
157 else if (enc->flags & F_SPACE_AFTER) \ 359 else if (enc->json.flags & F_SPACE_AFTER)
158 SPACE; \ 360 encode_space (enc);
159 SE 361}
160 362
161static void encode_sv (enc_t *enc, SV *sv); 363static void encode_sv (enc_t *enc, SV *sv);
162 364
163static void 365static void
164encode_av (enc_t *enc, AV *av) 366encode_av (enc_t *enc, AV *av)
165{ 367{
166 int i, len = av_len (av); 368 int i, len = av_len (av);
167 369
370 if (enc->indent >= enc->maxdepth)
371 croak ("data structure too deep (hit recursion limit)");
372
168 encode_ch (enc, '['); NL; 373 encode_ch (enc, '[');
169 ++enc->indent; 374
375 if (len >= 0)
376 {
377 encode_nl (enc); ++enc->indent;
170 378
171 for (i = 0; i <= len; ++i) 379 for (i = 0; i <= len; ++i)
172 { 380 {
173 INDENT; 381 SV **svp = av_fetch (av, i, 0);
174 encode_sv (enc, *av_fetch (av, i, 0));
175 382
383 encode_indent (enc);
384
385 if (svp)
386 encode_sv (enc, *svp);
387 else
388 encode_str (enc, "null", 4, 0);
389
176 if (i < len) 390 if (i < len)
177 COMMA; 391 encode_comma (enc);
178 } 392 }
179 393
180 NL; 394 encode_nl (enc); --enc->indent; encode_indent (enc);
181 395 }
182 --enc->indent; 396
183 INDENT; encode_ch (enc, ']'); 397 encode_ch (enc, ']');
184} 398}
185 399
186static void 400static void
187encode_he (enc_t *enc, HE *he) 401encode_hk (enc_t *enc, HE *he)
188{ 402{
189 encode_ch (enc, '"'); 403 encode_ch (enc, '"');
190 404
191 if (HeKLEN (he) == HEf_SVKEY) 405 if (HeKLEN (he) == HEf_SVKEY)
192 { 406 {
193 SV *sv = HeSVKEY (he); 407 SV *sv = HeSVKEY (he);
194 STRLEN len; 408 STRLEN len;
409 char *str;
410
411 SvGETMAGIC (sv);
195 char *str = SvPV (sv, len); 412 str = SvPV (sv, len);
196 413
197 encode_str (enc, str, len, SvUTF8 (sv)); 414 encode_str (enc, str, len, SvUTF8 (sv));
198 } 415 }
199 else 416 else
200 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 417 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
201 418
202 encode_ch (enc, '"'); 419 encode_ch (enc, '"');
203 420
204 if (enc->flags & F_SPACE_BEFORE) SPACE; 421 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
205 encode_ch (enc, ':'); 422 encode_ch (enc, ':');
206 if (enc->flags & F_SPACE_AFTER ) SPACE; 423 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
207 encode_sv (enc, HeVAL (he));
208} 424}
209 425
210// compare hash entries, used when all keys are bytestrings 426// compare hash entries, used when all keys are bytestrings
211static int 427static int
212he_cmp_fast (const void *a_, const void *b_) 428he_cmp_fast (const void *a_, const void *b_)
… …
217 HE *b = *(HE **)b_; 433 HE *b = *(HE **)b_;
218 434
219 STRLEN la = HeKLEN (a); 435 STRLEN la = HeKLEN (a);
220 STRLEN lb = HeKLEN (b); 436 STRLEN lb = HeKLEN (b);
221 437
222 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 438 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
223 cmp = la < lb ? -1 : la == lb ? 0 : 1; 439 cmp = lb - la;
224 440
225 return cmp; 441 return cmp;
226} 442}
227 443
228// compare hash entries, used when some keys are sv's or utf-x 444// compare hash entries, used when some keys are sv's or utf-x
229static int 445static int
230he_cmp_slow (const void *a, const void *b) 446he_cmp_slow (const void *a, const void *b)
231{ 447{
232 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 448 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
233} 449}
234 450
235static void 451static void
236encode_hv (enc_t *enc, HV *hv) 452encode_hv (enc_t *enc, HV *hv)
237{ 453{
238 int count, i; 454 HE *he;
239 455
240 encode_ch (enc, '{'); NL; ++enc->indent; 456 if (enc->indent >= enc->maxdepth)
457 croak ("data structure too deep (hit recursion limit)");
241 458
242 if ((count = hv_iterinit (hv))) 459 encode_ch (enc, '{');
243 { 460
244 // for canonical output we have to sort by keys first 461 // for canonical output we have to sort by keys first
245 // actually, this is mostly due to the stupid so-called 462 // actually, this is mostly due to the stupid so-called
246 // security workaround added somewhere in 5.8.x. 463 // security workaround added somewhere in 5.8.x.
247 // that randomises hash orderings 464 // that randomises hash orderings
248 if (enc->flags & F_CANONICAL) 465 if (enc->json.flags & F_CANONICAL)
466 {
467 int count = hv_iterinit (hv);
468
469 if (SvMAGICAL (hv))
249 { 470 {
250 HE *he, *hes [count]; 471 // need to count by iterating. could improve by dynamically building the vector below
472 // but I don't care for the speed of this special case.
473 // note also that we will run into undefined behaviour when the two iterations
474 // do not result in the same count, something I might care for in some later release.
475
476 count = 0;
477 while (hv_iternext (hv))
478 ++count;
479
480 hv_iterinit (hv);
481 }
482
483 if (count)
484 {
251 int fast = 1; 485 int i, fast = 1;
486#if defined(__BORLANDC__) || defined(_MSC_VER)
487 HE **hes = _alloca (count * sizeof (HE));
488#else
489 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
490#endif
252 491
253 i = 0; 492 i = 0;
254 while ((he = hv_iternext (hv))) 493 while ((he = hv_iternext (hv)))
255 { 494 {
256 hes [i++] = he; 495 hes [i++] = he;
… …
262 501
263 if (fast) 502 if (fast)
264 qsort (hes, count, sizeof (HE *), he_cmp_fast); 503 qsort (hes, count, sizeof (HE *), he_cmp_fast);
265 else 504 else
266 { 505 {
267 // hack to disable "use bytes" 506 // hack to forcefully disable "use bytes"
268 COP *oldcop = PL_curcop, cop; 507 COP cop = *PL_curcop;
269 cop.op_private = 0; 508 cop.op_private = 0;
509
510 ENTER;
511 SAVETMPS;
512
513 SAVEVPTR (PL_curcop);
270 PL_curcop = &cop; 514 PL_curcop = &cop;
271 515
272 SAVETMPS;
273 qsort (hes, count, sizeof (HE *), he_cmp_slow); 516 qsort (hes, count, sizeof (HE *), he_cmp_slow);
517
274 FREETMPS; 518 FREETMPS;
275 519 LEAVE;
276 PL_curcop = oldcop;
277 } 520 }
278 521
279 for (i = 0; i < count; ++i) 522 encode_nl (enc); ++enc->indent;
523
524 while (count--)
280 { 525 {
281 INDENT; 526 encode_indent (enc);
527 he = hes [count];
282 encode_he (enc, hes [i]); 528 encode_hk (enc, he);
529 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
283 530
284 if (i < count - 1) 531 if (count)
285 COMMA; 532 encode_comma (enc);
286 } 533 }
287 534
535 encode_nl (enc); --enc->indent; encode_indent (enc);
536 }
537 }
538 else
539 {
540 if (hv_iterinit (hv) || SvMAGICAL (hv))
541 if ((he = hv_iternext (hv)))
288 NL; 542 {
543 encode_nl (enc); ++enc->indent;
544
545 for (;;)
546 {
547 encode_indent (enc);
548 encode_hk (enc, he);
549 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
550
551 if (!(he = hv_iternext (hv)))
552 break;
553
554 encode_comma (enc);
555 }
556
557 encode_nl (enc); --enc->indent; encode_indent (enc);
558 }
559 }
560
561 encode_ch (enc, '}');
562}
563
564// encode objects, arrays and special \0=false and \1=true values.
565static void
566encode_rv (enc_t *enc, SV *sv)
567{
568 svtype svt;
569
570 SvGETMAGIC (sv);
571 svt = SvTYPE (sv);
572
573 if (expect_false (SvOBJECT (sv)))
574 {
575 HV *stash = !JSON_SLOW || json_boolean_stash
576 ? json_boolean_stash
577 : gv_stashpv ("JSON::XS::Boolean", 1);
578
579 if (SvSTASH (sv) == stash)
580 {
581 if (SvIV (sv))
582 encode_str (enc, "true", 4, 0);
583 else
584 encode_str (enc, "false", 5, 0);
289 } 585 }
290 else 586 else
291 { 587 {
292 SV *sv; 588#if 0
293 HE *he = hv_iternext (hv); 589 if (0 && sv_derived_from (rv, "JSON::Literal"))
294
295 for (;;)
296 { 590 {
297 INDENT; 591 // not yet
298 encode_he (enc, he);
299
300 if (!(he = hv_iternext (hv)))
301 break;
302
303 COMMA;
304 } 592 }
305 593#endif
594 if (enc->json.flags & F_CONV_BLESSED)
306 NL; 595 {
596 // we re-bless the reference to get overload and other niceties right
597 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
598
599 if (to_json)
600 {
601 dSP;
602
603 ENTER; SAVETMPS; PUSHMARK (SP);
604 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
605
606 // calling with G_SCALAR ensures that we always get a 1 return value
607 PUTBACK;
608 call_sv ((SV *)GvCV (to_json), G_SCALAR);
609 SPAGAIN;
610
611 // catch this surprisingly common error
612 if (SvROK (TOPs) && SvRV (TOPs) == sv)
613 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
614
615 sv = POPs;
616 PUTBACK;
617
618 encode_sv (enc, sv);
619
620 FREETMPS; LEAVE;
621 }
622 else if (enc->json.flags & F_ALLOW_BLESSED)
623 encode_str (enc, "null", 4, 0);
624 else
625 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
626 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
627 }
628 else if (enc->json.flags & F_ALLOW_BLESSED)
629 encode_str (enc, "null", 4, 0);
630 else
631 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
632 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
307 } 633 }
308 } 634 }
635 else if (svt == SVt_PVHV)
636 encode_hv (enc, (HV *)sv);
637 else if (svt == SVt_PVAV)
638 encode_av (enc, (AV *)sv);
639 else if (svt < SVt_PVAV)
640 {
641 STRLEN len = 0;
642 char *pv = svt ? SvPV (sv, len) : 0;
309 643
310 --enc->indent; INDENT; encode_ch (enc, '}'); 644 if (len == 1 && *pv == '1')
645 encode_str (enc, "true", 4, 0);
646 else if (len == 1 && *pv == '0')
647 encode_str (enc, "false", 5, 0);
648 else if (enc->json.flags & F_ALLOW_UNKNOWN)
649 encode_str (enc, "null", 4, 0);
650 else
651 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
652 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
653 }
654 else if (enc->json.flags & F_ALLOW_UNKNOWN)
655 encode_str (enc, "null", 4, 0);
656 else
657 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
658 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
311} 659}
312 660
313static void 661static void
314encode_sv (enc_t *enc, SV *sv) 662encode_sv (enc_t *enc, SV *sv)
315{ 663{
664 SvGETMAGIC (sv);
665
316 if (SvPOKp (sv)) 666 if (SvPOKp (sv))
317 { 667 {
318 STRLEN len; 668 STRLEN len;
319 char *str = SvPV (sv, len); 669 char *str = SvPV (sv, len);
320 encode_ch (enc, '"'); 670 encode_ch (enc, '"');
321 encode_str (enc, str, len, SvUTF8 (sv)); 671 encode_str (enc, str, len, SvUTF8 (sv));
322 encode_ch (enc, '"'); 672 encode_ch (enc, '"');
323 } 673 }
324 else if (SvNOKp (sv)) 674 else if (SvNOKp (sv))
325 { 675 {
676 // trust that perl will do the right thing w.r.t. JSON syntax.
326 need (enc, NV_DIG + 32); 677 need (enc, NV_DIG + 32);
327 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 678 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
328 enc->cur += strlen (enc->cur); 679 enc->cur += strlen (enc->cur);
329 } 680 }
330 else if (SvIOKp (sv)) 681 else if (SvIOKp (sv))
331 { 682 {
683 // we assume we can always read an IV as a UV and vice versa
684 // we assume two's complement
685 // we assume no aliasing issues in the union
686 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
687 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
688 {
689 // optimise the "small number case"
690 // code will likely be branchless and use only a single multiplication
691 // works for numbers up to 59074
692 I32 i = SvIVX (sv);
693 U32 u;
694 char digit, nz = 0;
695
332 need (enc, 64); 696 need (enc, 6);
697
698 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
699 u = i < 0 ? -i : i;
700
701 // convert to 4.28 fixed-point representation
702 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
703
704 // now output digit by digit, each time masking out the integer part
705 // and multiplying by 5 while moving the decimal point one to the right,
706 // resulting in a net multiplication by 10.
707 // we always write the digit to memory but conditionally increment
708 // the pointer, to enable the use of conditional move instructions.
709 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
710 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
711 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
712 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
713 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
714 }
715 else
716 {
717 // large integer, use the (rather slow) snprintf way.
718 need (enc, IVUV_MAXCHARS);
333 enc->cur += 719 enc->cur +=
334 SvIsUV(sv) 720 SvIsUV(sv)
335 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 721 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
336 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 722 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
723 }
337 } 724 }
338 else if (SvROK (sv)) 725 else if (SvROK (sv))
339 { 726 encode_rv (enc, SvRV (sv));
340 if (!--enc->max_recurse) 727 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
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))
355 encode_str (enc, "null", 4, 0); 728 encode_str (enc, "null", 4, 0);
356 else 729 else
357 croak ("encountered perl type that JSON cannot handle"); 730 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
731 SvPV_nolen (sv), SvFLAGS (sv));
358} 732}
359 733
360static SV * 734static SV *
361encode_json (SV *scalar, UV flags) 735encode_json (SV *scalar, JSON *json)
362{ 736{
363 enc_t enc; 737 enc_t enc;
364 enc.flags = flags; 738
739 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
740 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
741
742 enc.json = *json;
365 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 743 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
366 enc.cur = SvPVX (enc.sv); 744 enc.cur = SvPVX (enc.sv);
367 enc.end = SvEND (enc.sv); 745 enc.end = SvEND (enc.sv);
368 enc.max_recurse = 0;
369 enc.indent = 0; 746 enc.indent = 0;
747 enc.maxdepth = DEC_DEPTH (enc.json.flags);
748 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
749 : enc.json.flags & F_LATIN1 ? 0x000100UL
750 : 0x110000UL;
370 751
371 SvPOK_only (enc.sv); 752 SvPOK_only (enc.sv);
372 encode_sv (&enc, scalar); 753 encode_sv (&enc, scalar);
373 754
755 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
756 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
757
374 if (!(flags & (F_ASCII | F_UTF8))) 758 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
375 SvUTF8_on (enc.sv); 759 SvUTF8_on (enc.sv);
376 760
377 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 761 if (enc.json.flags & F_SHRINK)
762 shrink (enc.sv);
763
378 return enc.sv; 764 return enc.sv;
379} 765}
380 766
381///////////////////////////////////////////////////////////////////////////// 767/////////////////////////////////////////////////////////////////////////////
768// decoder
382 769
383#define WS \ 770// structure used for decoding JSON
771typedef struct
772{
773 char *cur; // current parser pointer
774 char *end; // end of input string
775 const char *err; // parse error, if != 0
776 JSON json;
777 U32 depth; // recursion depth
778 U32 maxdepth; // recursion depth limit
779} dec_t;
780
781INLINE void
782decode_comment (dec_t *dec)
783{
784 // only '#'-style comments allowed a.t.m.
785
786 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
787 ++dec->cur;
788}
789
790INLINE void
791decode_ws (dec_t *dec)
792{
384 for (;;) \ 793 for (;;)
385 { \ 794 {
386 char ch = *dec->cur; \ 795 char ch = *dec->cur;
796
387 if (ch > 0x20 \ 797 if (ch > 0x20)
798 {
799 if (expect_false (ch == '#'))
800 {
801 if (dec->json.flags & F_RELAXED)
802 decode_comment (dec);
803 else
804 break;
805 }
806 else
807 break;
808 }
388 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 809 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
389 break; \ 810 break; // parse error, but let higher level handle it, gives better error messages
811
390 ++dec->cur; \ 812 ++dec->cur;
391 } 813 }
814}
392 815
393#define ERR(reason) SB dec->err = reason; goto fail; SE 816#define ERR(reason) SB dec->err = reason; goto fail; SE
817
394#define EXPECT_CH(ch) SB \ 818#define EXPECT_CH(ch) SB \
395 if (*dec->cur != ch) \ 819 if (*dec->cur != ch) \
396 ERR (# ch " expected"); \ 820 ERR (# ch " expected"); \
397 ++dec->cur; \ 821 ++dec->cur; \
398 SE 822 SE
399 823
824#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
825#define DEC_DEC_DEPTH --dec->depth
826
400static SV *decode_sv (dec_t *dec); 827static 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
406 828
407static signed char decode_hexdigit[256]; 829static signed char decode_hexdigit[256];
408 830
409static UV 831static UV
410decode_4hex (dec_t *dec) 832decode_4hex (dec_t *dec)
411{ 833{
412 signed char d1, d2, d3, d4; 834 signed char d1, d2, d3, d4;
835 unsigned char *cur = (unsigned char *)dec->cur;
413 836
414 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 837 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"); 838 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]]; 839 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"); 840 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");
422 841
423 dec->cur += 4; 842 dec->cur += 4;
424 843
425 return ((UV)d1) << 12 844 return ((UV)d1) << 12
426 | ((UV)d2) << 8 845 | ((UV)d2) << 8
… …
432} 851}
433 852
434static SV * 853static SV *
435decode_str (dec_t *dec) 854decode_str (dec_t *dec)
436{ 855{
437 SV *sv = NEWSV (0,2); 856 SV *sv = 0;
438 STRLEN cur = 0;
439 int utf8 = 0; 857 int utf8 = 0;
858 char *dec_cur = dec->cur;
440 859
441 for (;;) 860 do
442 { 861 {
443 unsigned char ch = *(unsigned char *)dec->cur; 862 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
863 char *cur = buf;
444 864
445 if (ch == '"') 865 do
446 break;
447 else if (ch == '\\')
448 { 866 {
449 switch (*++dec->cur) 867 unsigned char ch = *(unsigned char *)dec_cur++;
868
869 if (expect_false (ch == '"'))
450 { 870 {
451 case '\\': 871 --dec_cur;
452 case '/': 872 break;
453 case '"': APPEND_CH (*dec->cur++); break; 873 }
454 874 else if (expect_false (ch == '\\'))
455 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 875 {
456 case 't': APPEND_CH ('\011'); ++dec->cur; break; 876 switch (*dec_cur)
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 { 877 {
463 UV lo, hi; 878 case '\\':
464 ++dec->cur; 879 case '/':
880 case '"': *cur++ = *dec_cur++; break;
465 881
466 hi = decode_4hex (dec); 882 case 'b': ++dec_cur; *cur++ = '\010'; break;
467 if (hi == (UV)-1) 883 case 't': ++dec_cur; *cur++ = '\011'; break;
468 goto fail; 884 case 'n': ++dec_cur; *cur++ = '\012'; break;
885 case 'f': ++dec_cur; *cur++ = '\014'; break;
886 case 'r': ++dec_cur; *cur++ = '\015'; break;
469 887
470 // possibly a surrogate pair 888 case 'u':
471 if (hi >= 0xd800 && hi < 0xdc00)
472 { 889 {
473 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 890 UV lo, hi;
474 ERR ("illegal surrogate character"); 891 ++dec_cur;
475 892
476 dec->cur += 2; 893 dec->cur = dec_cur;
477
478 lo = decode_4hex (dec); 894 hi = decode_4hex (dec);
895 dec_cur = dec->cur;
479 if (lo == (UV)-1) 896 if (hi == (UV)-1)
480 goto fail; 897 goto fail;
481 898
899 // possibly a surrogate pair
900 if (hi >= 0xd800)
901 if (hi < 0xdc00)
902 {
903 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
904 ERR ("missing low surrogate character in surrogate pair");
905
906 dec_cur += 2;
907
908 dec->cur = dec_cur;
909 lo = decode_4hex (dec);
910 dec_cur = dec->cur;
911 if (lo == (UV)-1)
912 goto fail;
913
482 if (lo < 0xdc00 || lo >= 0xe000) 914 if (lo < 0xdc00 || lo >= 0xe000)
483 ERR ("surrogate pair expected"); 915 ERR ("surrogate pair expected");
484 916
485 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 917 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
918 }
919 else if (hi < 0xe000)
920 ERR ("missing high surrogate character in surrogate pair");
921
922 if (hi >= 0x80)
923 {
924 utf8 = 1;
925
926 cur = encode_utf8 (cur, hi);
927 }
928 else
929 *cur++ = hi;
486 } 930 }
487 else if (lo >= 0xdc00 && lo < 0xe000)
488 ERR ("illegal surrogate character");
489
490 if (hi >= 0x80)
491 { 931 break;
492 utf8 = 1;
493 932
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 933 default:
498 APPEND_CH (hi); 934 --dec_cur;
935 ERR ("illegal backslash escape sequence in string");
499 } 936 }
937 }
938 else if (expect_true (ch >= 0x20 && ch < 0x80))
939 *cur++ = ch;
940 else if (ch >= 0x80)
941 {
942 STRLEN clen;
943 UV uch;
944
945 --dec_cur;
946
947 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
948 if (clen == (STRLEN)-1)
949 ERR ("malformed UTF-8 character in JSON string");
950
951 do
952 *cur++ = *dec_cur++;
953 while (--clen);
954
955 utf8 = 1;
956 }
957 else
958 {
959 --dec_cur;
960
961 if (!ch)
962 ERR ("unexpected end of string while parsing JSON string");
500 break; 963 else
964 ERR ("invalid character encountered while parsing JSON string");
501 } 965 }
502 } 966 }
503 else if (ch >= 0x20 && ch <= 0x7f) 967 while (cur < buf + SHORT_STRING_LEN);
504 APPEND_CH (*dec->cur++); 968
505 else if (ch >= 0x80) 969 {
970 STRLEN len = cur - buf;
971
972 if (sv)
506 { 973 {
507 STRLEN clen; 974 SvGROW (sv, SvCUR (sv) + len + 1);
508 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 975 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
509 if (clen < 0) 976 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 } 977 }
516 else 978 else
517 ERR ("invalid character encountered"); 979 sv = newSVpvn (buf, len);
518 } 980 }
981 }
982 while (*dec_cur != '"');
519 983
520 ++dec->cur; 984 ++dec_cur;
521 985
986 if (sv)
987 {
522 SvPOK_only (sv); 988 SvPOK_only (sv);
523
524 SvCUR_set (sv, cur);
525 *SvEND (sv) = 0; 989 *SvEND (sv) = 0;
526 990
527 if (utf8) 991 if (utf8)
528 SvUTF8_on (sv); 992 SvUTF8_on (sv);
993 }
994 else
995 sv = newSVpvn ("", 0);
529 996
997 dec->cur = dec_cur;
530 return sv; 998 return sv;
531 999
532fail: 1000fail:
533 SvREFCNT_dec (sv); 1001 dec->cur = dec_cur;
534 return 0; 1002 return 0;
535} 1003}
536 1004
537static SV * 1005static SV *
538decode_num (dec_t *dec) 1006decode_num (dec_t *dec)
… …
548 { 1016 {
549 ++dec->cur; 1017 ++dec->cur;
550 if (*dec->cur >= '0' && *dec->cur <= '9') 1018 if (*dec->cur >= '0' && *dec->cur <= '9')
551 ERR ("malformed number (leading zero must not be followed by another digit)"); 1019 ERR ("malformed number (leading zero must not be followed by another digit)");
552 } 1020 }
553 1021 else if (*dec->cur < '0' || *dec->cur > '9')
554 // int 1022 ERR ("malformed number (no digits after initial minus)");
1023 else
1024 do
1025 {
1026 ++dec->cur;
1027 }
555 while (*dec->cur >= '0' && *dec->cur <= '9') 1028 while (*dec->cur >= '0' && *dec->cur <= '9');
556 ++dec->cur;
557 1029
558 // [frac] 1030 // [frac]
559 if (*dec->cur == '.') 1031 if (*dec->cur == '.')
560 { 1032 {
561 is_nv = 1; 1033 ++dec->cur;
1034
1035 if (*dec->cur < '0' || *dec->cur > '9')
1036 ERR ("malformed number (no digits after decimal point)");
562 1037
563 do 1038 do
564 { 1039 {
565 ++dec->cur; 1040 ++dec->cur;
566 } 1041 }
567 while (*dec->cur >= '0' && *dec->cur <= '9'); 1042 while (*dec->cur >= '0' && *dec->cur <= '9');
1043
1044 is_nv = 1;
568 } 1045 }
569 1046
570 // [exp] 1047 // [exp]
571 if (*dec->cur == 'e' || *dec->cur == 'E') 1048 if (*dec->cur == 'e' || *dec->cur == 'E')
572 { 1049 {
573 is_nv = 1;
574
575 ++dec->cur; 1050 ++dec->cur;
1051
576 if (*dec->cur == '-' || *dec->cur == '+') 1052 if (*dec->cur == '-' || *dec->cur == '+')
577 ++dec->cur; 1053 ++dec->cur;
578 1054
1055 if (*dec->cur < '0' || *dec->cur > '9')
1056 ERR ("malformed number (no digits after exp sign)");
1057
1058 do
1059 {
1060 ++dec->cur;
1061 }
579 while (*dec->cur >= '0' && *dec->cur <= '9') 1062 while (*dec->cur >= '0' && *dec->cur <= '9');
580 ++dec->cur; 1063
1064 is_nv = 1;
581 } 1065 }
582 1066
583 if (!is_nv) 1067 if (!is_nv)
584 { 1068 {
585 UV uv; 1069 int len = dec->cur - start;
586 int numtype = grok_number (start, dec->cur - start, &uv); 1070
587 if (numtype & IS_NUMBER_IN_UV) 1071 // special case the rather common 1..5-digit-int case
588 if (numtype & IS_NUMBER_NEG) 1072 if (*start == '-')
1073 switch (len)
589 { 1074 {
590 if (uv < (UV)IV_MIN) 1075 case 2: return newSViv (-( start [1] - '0' * 1));
591 return newSViv (-(IV)uv); 1076 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
1077 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1078 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1079 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
592 } 1080 }
1081 else
1082 switch (len)
1083 {
1084 case 1: return newSViv ( start [0] - '0' * 1);
1085 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1086 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1087 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1088 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1089 }
1090
1091 {
1092 UV uv;
1093 int numtype = grok_number (start, len, &uv);
1094 if (numtype & IS_NUMBER_IN_UV)
1095 if (numtype & IS_NUMBER_NEG)
1096 {
1097 if (uv < (UV)IV_MIN)
1098 return newSViv (-(IV)uv);
1099 }
593 else 1100 else
594 return newSVuv (uv); 1101 return newSVuv (uv);
595 } 1102 }
596 1103
1104 len -= *start == '-' ? 1 : 0;
1105
1106 // does not fit into IV or UV, try NV
1107 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
1108 #if defined (LDBL_DIG)
1109 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1110 #endif
1111 )
1112 // fits into NV without loss of precision
1113 return newSVnv (Atof (start));
1114
1115 // everything else fails, convert it to a string
1116 return newSVpvn (start, dec->cur - start);
1117 }
1118
1119 // loss of precision here
597 return newSVnv (Atof (start)); 1120 return newSVnv (Atof (start));
598 1121
599fail: 1122fail:
600 return 0; 1123 return 0;
601} 1124}
… …
603static SV * 1126static SV *
604decode_av (dec_t *dec) 1127decode_av (dec_t *dec)
605{ 1128{
606 AV *av = newAV (); 1129 AV *av = newAV ();
607 1130
1131 DEC_INC_DEPTH;
1132 decode_ws (dec);
1133
1134 if (*dec->cur == ']')
1135 ++dec->cur;
1136 else
608 for (;;) 1137 for (;;)
609 { 1138 {
610 SV *value; 1139 SV *value;
611 1140
612 value = decode_sv (dec); 1141 value = decode_sv (dec);
613 if (!value) 1142 if (!value)
614 goto fail; 1143 goto fail;
615 1144
616 av_push (av, value); 1145 av_push (av, value);
617 1146
618 WS; 1147 decode_ws (dec);
619 1148
620 if (*dec->cur == ']') 1149 if (*dec->cur == ']')
621 { 1150 {
622 ++dec->cur; 1151 ++dec->cur;
623 break; 1152 break;
1153 }
624 } 1154
625
626 if (*dec->cur != ',') 1155 if (*dec->cur != ',')
627 ERR (", or ] expected while parsing array"); 1156 ERR (", or ] expected while parsing array");
628 1157
629 ++dec->cur; 1158 ++dec->cur;
1159
1160 decode_ws (dec);
1161
1162 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1163 {
1164 ++dec->cur;
1165 break;
1166 }
630 } 1167 }
631 1168
1169 DEC_DEC_DEPTH;
632 return newRV_noinc ((SV *)av); 1170 return newRV_noinc ((SV *)av);
633 1171
634fail: 1172fail:
635 SvREFCNT_dec (av); 1173 SvREFCNT_dec (av);
1174 DEC_DEC_DEPTH;
636 return 0; 1175 return 0;
637} 1176}
638 1177
639static SV * 1178static SV *
640decode_hv (dec_t *dec) 1179decode_hv (dec_t *dec)
641{ 1180{
1181 SV *sv;
642 HV *hv = newHV (); 1182 HV *hv = newHV ();
643 1183
1184 DEC_INC_DEPTH;
1185 decode_ws (dec);
1186
1187 if (*dec->cur == '}')
1188 ++dec->cur;
1189 else
644 for (;;) 1190 for (;;)
645 { 1191 {
646 SV *key, *value;
647
648 WS; EXPECT_CH ('"'); 1192 EXPECT_CH ('"');
649 1193
650 key = decode_str (dec); 1194 // heuristic: assume that
651 if (!key) 1195 // a) decode_str + hv_store_ent are abysmally slow.
652 goto fail; 1196 // b) most hash keys are short, simple ascii text.
653 1197 // => try to "fast-match" such strings to avoid
654 WS; EXPECT_CH (':'); 1198 // the overhead of decode_str + hv_store_ent.
655
656 value = decode_sv (dec);
657 if (!value)
658 { 1199 {
1200 SV *value;
1201 char *p = dec->cur;
1202 char *e = p + 24; // only try up to 24 bytes
1203
1204 for (;;)
1205 {
1206 // the >= 0x80 is false on most architectures
1207 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1208 {
1209 // slow path, back up and use decode_str
1210 SV *key = decode_str (dec);
1211 if (!key)
1212 goto fail;
1213
1214 decode_ws (dec); EXPECT_CH (':');
1215
1216 decode_ws (dec);
1217 value = decode_sv (dec);
1218 if (!value)
1219 {
1220 SvREFCNT_dec (key);
1221 goto fail;
1222 }
1223
1224 hv_store_ent (hv, key, value, 0);
659 SvREFCNT_dec (key); 1225 SvREFCNT_dec (key);
1226
1227 break;
1228 }
1229 else if (*p == '"')
1230 {
1231 // fast path, got a simple key
1232 char *key = dec->cur;
1233 int len = p - key;
1234 dec->cur = p + 1;
1235
1236 decode_ws (dec); EXPECT_CH (':');
1237
1238 decode_ws (dec);
1239 value = decode_sv (dec);
1240 if (!value)
660 goto fail; 1241 goto fail;
1242
1243 hv_store (hv, key, len, value, 0);
1244
1245 break;
1246 }
1247
1248 ++p;
1249 }
661 } 1250 }
662 1251
663 //TODO: optimise 1252 decode_ws (dec);
664 hv_store_ent (hv, key, value, 0);
665 1253
666 WS;
667
668 if (*dec->cur == '}') 1254 if (*dec->cur == '}')
1255 {
1256 ++dec->cur;
1257 break;
1258 }
1259
1260 if (*dec->cur != ',')
1261 ERR (", or } expected while parsing object/hash");
1262
1263 ++dec->cur;
1264
1265 decode_ws (dec);
1266
1267 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1268 {
1269 ++dec->cur;
1270 break;
1271 }
1272 }
1273
1274 DEC_DEC_DEPTH;
1275 sv = newRV_noinc ((SV *)hv);
1276
1277 // check filter callbacks
1278 if (dec->json.flags & F_HOOK)
1279 {
1280 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
669 { 1281 {
670 ++dec->cur; 1282 HE *cb, *he;
671 break; 1283
1284 hv_iterinit (hv);
1285 he = hv_iternext (hv);
1286 hv_iterinit (hv);
1287
1288 // the next line creates a mortal sv each time its called.
1289 // might want to optimise this for common cases.
1290 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1291
1292 if (cb)
1293 {
1294 dSP;
1295 int count;
1296
1297 ENTER; SAVETMPS; PUSHMARK (SP);
1298 XPUSHs (HeVAL (he));
1299
1300 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1301
1302 if (count == 1)
1303 {
1304 sv = newSVsv (POPs);
1305 FREETMPS; LEAVE;
1306 return sv;
1307 }
1308
1309 FREETMPS; LEAVE;
1310 }
672 } 1311 }
673 1312
674 if (*dec->cur != ',') 1313 if (dec->json.cb_object)
675 ERR (", or } expected while parsing object/hash"); 1314 {
1315 dSP;
1316 int count;
676 1317
677 ++dec->cur; 1318 ENTER; SAVETMPS; PUSHMARK (SP);
678 } 1319 XPUSHs (sv_2mortal (sv));
679 1320
680 return newRV_noinc ((SV *)hv); 1321 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1322
1323 if (count == 1)
1324 {
1325 sv = newSVsv (POPs);
1326 FREETMPS; LEAVE;
1327 return sv;
1328 }
1329
1330 SvREFCNT_inc (sv);
1331 FREETMPS; LEAVE;
1332 }
1333 }
1334
1335 return sv;
681 1336
682fail: 1337fail:
683 SvREFCNT_dec (hv); 1338 SvREFCNT_dec (hv);
1339 DEC_DEC_DEPTH;
684 return 0; 1340 return 0;
685} 1341}
686 1342
687static SV * 1343static SV *
688decode_sv (dec_t *dec) 1344decode_sv (dec_t *dec)
689{ 1345{
690 WS; 1346 // the beauty of JSON: you need exactly one character lookahead
1347 // to parse everything.
691 switch (*dec->cur) 1348 switch (*dec->cur)
692 { 1349 {
693 case '"': ++dec->cur; return decode_str (dec); 1350 case '"': ++dec->cur; return decode_str (dec);
694 case '[': ++dec->cur; return decode_av (dec); 1351 case '[': ++dec->cur; return decode_av (dec);
695 case '{': ++dec->cur; return decode_hv (dec); 1352 case '{': ++dec->cur; return decode_hv (dec);
696 1353
697 case '-': 1354 case '-':
698 case '0': case '1': case '2': case '3': case '4': 1355 case '0': case '1': case '2': case '3': case '4':
699 case '5': case '6': case '7': case '8': case '9': 1356 case '5': case '6': case '7': case '8': case '9':
700 return decode_num (dec); 1357 return decode_num (dec);
701 1358
702 case 't': 1359 case 't':
703 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1360 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
704 { 1361 {
705 dec->cur += 4; 1362 dec->cur += 4;
1363#if JSON_SLOW
1364 json_true = get_bool ("JSON::XS::true");
1365#endif
706 return newSViv (1); 1366 return newSVsv (json_true);
707 } 1367 }
708 else 1368 else
709 ERR ("'true' expected"); 1369 ERR ("'true' expected");
710 1370
711 break; 1371 break;
712 1372
713 case 'f': 1373 case 'f':
714 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1374 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
715 { 1375 {
716 dec->cur += 5; 1376 dec->cur += 5;
1377#if JSON_SLOW
1378 json_false = get_bool ("JSON::XS::false");
1379#endif
717 return newSViv (0); 1380 return newSVsv (json_false);
718 } 1381 }
719 else 1382 else
720 ERR ("'false' expected"); 1383 ERR ("'false' expected");
721 1384
722 break; 1385 break;
723 1386
724 case 'n': 1387 case 'n':
725 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1388 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
726 { 1389 {
727 dec->cur += 4; 1390 dec->cur += 4;
728 return newSViv (1); 1391 return newSVsv (&PL_sv_undef);
729 } 1392 }
730 else 1393 else
731 ERR ("'null' expected"); 1394 ERR ("'null' expected");
732 1395
733 break; 1396 break;
734 1397
735 default: 1398 default:
736 ERR ("malformed json string"); 1399 ERR ("malformed JSON string, neither array, object, number, string or atom");
737 break; 1400 break;
738 } 1401 }
739 1402
740fail: 1403fail:
741 return 0; 1404 return 0;
742} 1405}
743 1406
744static SV * 1407static SV *
745decode_json (SV *string, UV flags) 1408decode_json (SV *string, JSON *json, STRLEN *offset_return)
746{ 1409{
1410 dec_t dec;
1411 STRLEN offset;
747 SV *sv; 1412 SV *sv;
748 1413
1414 SvGETMAGIC (string);
1415 SvUPGRADE (string, SVt_PV);
1416
1417 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1418 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1419 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1420
749 if (!(flags & F_UTF8)) 1421 if (json->flags & F_UTF8)
1422 sv_utf8_downgrade (string, 0);
1423 else
750 sv_utf8_upgrade (string); 1424 sv_utf8_upgrade (string);
751 1425
752 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1426 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
753 1427
754 dec_t dec; 1428 dec.json = *json;
755 dec.flags = flags;
756 dec.cur = SvPVX (string); 1429 dec.cur = SvPVX (string);
757 dec.end = SvEND (string); 1430 dec.end = SvEND (string);
758 dec.err = 0; 1431 dec.err = 0;
1432 dec.depth = 0;
1433 dec.maxdepth = DEC_DEPTH (dec.json.flags);
759 1434
760 *dec.end = 1; // invalid anywhere 1435 if (dec.json.cb_object || dec.json.cb_sk_object)
1436 dec.json.flags |= F_HOOK;
1437
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439
1440 decode_ws (&dec);
761 sv = decode_sv (&dec); 1441 sv = decode_sv (&dec);
762 *dec.end = 0; 1442
1443 if (!(offset_return || !sv))
1444 {
1445 // check for trailing garbage
1446 decode_ws (&dec);
1447
1448 if (*dec.cur)
1449 {
1450 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv);
1452 sv = 0;
1453 }
1454 }
1455
1456 if (offset_return || !sv)
1457 {
1458 offset = dec.json.flags & F_UTF8
1459 ? dec.cur - SvPVX (string)
1460 : utf8_distance (dec.cur, SvPVX (string));
1461
1462 if (offset_return)
1463 *offset_return = offset;
1464 }
763 1465
764 if (!sv) 1466 if (!sv)
765 { 1467 {
766 IV offset = utf8_distance (dec.cur, SvPVX (string));
767 SV *uni = sv_newmortal (); 1468 SV *uni = sv_newmortal ();
768 1469
1470 // horrible hack to silence warning inside pv_uni_display
1471 COP cop = *PL_curcop;
1472 cop.cop_warnings = pWARN_NONE;
1473 ENTER;
1474 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop;
769 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE;
1478
770 croak ("%s, at character %d (%s)", 1479 croak ("%s, at character offset %d [\"%s\"]",
771 dec.err, 1480 dec.err,
772 (int)offset, 1481 (int)offset,
773 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
774 } 1483 }
775 1484
776 sv_dump (sv);//D
777 return sv_2mortal (sv); 1485 sv = sv_2mortal (sv);
1486
1487 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1488 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1489
1490 return sv;
778} 1491}
1492
1493/////////////////////////////////////////////////////////////////////////////
1494// incremental parser
1495
1496static void
1497incr_parse (JSON *self)
1498{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1500
1501 for (;;)
1502 {
1503 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1504 switch (self->incr_mode)
1505 {
1506 // only used for intiial whitespace skipping
1507 case INCR_M_WS:
1508 for (;;)
1509 {
1510 if (*p > 0x20)
1511 {
1512 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json;
1514 }
1515 else if (!*p)
1516 goto interrupt;
1517
1518 ++p;
1519 }
1520
1521 // skip a single char inside a string (for \\-processing)
1522 case INCR_M_BS:
1523 if (!*p)
1524 goto interrupt;
1525
1526 ++p;
1527 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str;
1529
1530 // inside a string
1531 case INCR_M_STR:
1532 incr_m_str:
1533 for (;;)
1534 {
1535 if (*p == '"')
1536 {
1537 ++p;
1538 self->incr_mode = INCR_M_JSON;
1539
1540 if (!self->incr_nest)
1541 goto interrupt;
1542
1543 goto incr_m_json;
1544 }
1545 else if (*p == '\\')
1546 {
1547 ++p; // "virtually" consumes character after \
1548
1549 if (!*p) // if at end of string we have to switch modes
1550 {
1551 self->incr_mode = INCR_M_BS;
1552 goto interrupt;
1553 }
1554 }
1555 else if (!*p)
1556 goto interrupt;
1557
1558 ++p;
1559 }
1560
1561 // after initial ws, outside string
1562 case INCR_M_JSON:
1563 incr_m_json:
1564 for (;;)
1565 {
1566 switch (*p++)
1567 {
1568 case 0:
1569 --p;
1570 goto interrupt;
1571
1572 case 0x09:
1573 case 0x0a:
1574 case 0x0d:
1575 case 0x20:
1576 if (!self->incr_nest)
1577 {
1578 --p; // do not eat the whitespace, let the next round do it
1579 goto interrupt;
1580 }
1581 break;
1582
1583 case '"':
1584 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str;
1586
1587 case '[':
1588 case '{':
1589 ++self->incr_nest;
1590 break;
1591
1592 case ']':
1593 case '}':
1594 if (!--self->incr_nest)
1595 goto interrupt;
1596 }
1597 }
1598 }
1599
1600 modechange:
1601 ;
1602 }
1603
1604interrupt:
1605 self->incr_pos = p - SvPVX (self->incr_text);
1606 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1607}
1608
1609/////////////////////////////////////////////////////////////////////////////
1610// XS interface functions
779 1611
780MODULE = JSON::XS PACKAGE = JSON::XS 1612MODULE = JSON::XS PACKAGE = JSON::XS
781 1613
782BOOT: 1614BOOT:
783{ 1615{
784 int i; 1616 int i;
785 1617
786 memset (decode_hexdigit, 0xff, 256);
787 for (i = 10; i--; ) 1618 for (i = 0; i < 256; ++i)
788 decode_hexdigit ['0' + i] = i; 1619 decode_hexdigit [i] =
1620 i >= '0' && i <= '9' ? i - '0'
1621 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1622 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1623 : -1;
789 1624
790 for (i = 6; --i; )
791 {
792 decode_hexdigit ['a' + i] = 10 + i;
793 decode_hexdigit ['A' + i] = 10 + i;
794 }
795
796 json_stash = gv_stashpv ("JSON::XS", 1); 1625 json_stash = gv_stashpv ("JSON::XS" , 1);
797} 1626 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
798 1627
799SV *new (char *dummy) 1628 json_true = get_bool ("JSON::XS::true");
1629 json_false = get_bool ("JSON::XS::false");
1630}
1631
1632PROTOTYPES: DISABLE
1633
1634void CLONE (...)
800 CODE: 1635 CODE:
801 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1636 json_stash = 0;
1637 json_boolean_stash = 0;
1638
1639void new (char *klass)
1640 PPCODE:
1641{
1642 SV *pv = NEWSV (0, sizeof (JSON));
1643 SvPOK_only (pv);
1644 Zero (SvPVX (pv), 1, JSON);
1645 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1646 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1649 )));
1650}
1651
1652void ascii (JSON *self, int enable = 1)
1653 ALIAS:
1654 ascii = F_ASCII
1655 latin1 = F_LATIN1
1656 utf8 = F_UTF8
1657 indent = F_INDENT
1658 canonical = F_CANONICAL
1659 space_before = F_SPACE_BEFORE
1660 space_after = F_SPACE_AFTER
1661 pretty = F_PRETTY
1662 allow_nonref = F_ALLOW_NONREF
1663 shrink = F_SHRINK
1664 allow_blessed = F_ALLOW_BLESSED
1665 convert_blessed = F_CONV_BLESSED
1666 relaxed = F_RELAXED
1667 allow_unknown = F_ALLOW_UNKNOWN
1668 PPCODE:
1669{
1670 if (enable)
1671 self->flags |= ix;
1672 else
1673 self->flags &= ~ix;
1674
1675 XPUSHs (ST (0));
1676}
1677
1678void get_ascii (JSON *self)
1679 ALIAS:
1680 get_ascii = F_ASCII
1681 get_latin1 = F_LATIN1
1682 get_utf8 = F_UTF8
1683 get_indent = F_INDENT
1684 get_canonical = F_CANONICAL
1685 get_space_before = F_SPACE_BEFORE
1686 get_space_after = F_SPACE_AFTER
1687 get_allow_nonref = F_ALLOW_NONREF
1688 get_shrink = F_SHRINK
1689 get_allow_blessed = F_ALLOW_BLESSED
1690 get_convert_blessed = F_CONV_BLESSED
1691 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN
1693 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix));
1695
1696void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1697 PPCODE:
1698{
1699 UV log2 = 0;
1700
1701 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1702
1703 while ((1UL << log2) < max_depth)
1704 ++log2;
1705
1706 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1707
1708 XPUSHs (ST (0));
1709}
1710
1711U32 get_max_depth (JSON *self)
1712 CODE:
1713 RETVAL = DEC_DEPTH (self->flags);
802 OUTPUT: 1714 OUTPUT:
803 RETVAL 1715 RETVAL
804 1716
805SV *ascii (SV *self, int enable) 1717void max_size (JSON *self, UV max_size = 0)
806 ALIAS: 1718 PPCODE:
807 ascii = F_ASCII 1719{
808 utf8 = F_UTF8 1720 UV log2 = 0;
809 indent = F_INDENT 1721
810 canonical = F_CANONICAL 1722 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
811 space_before = F_SPACE_BEFORE 1723 if (max_size == 1) max_size = 2;
812 space_after = F_SPACE_AFTER 1724
813 json_rpc = F_JSON_RPC 1725 while ((1UL << log2) < max_size)
1726 ++log2;
1727
1728 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1729
1730 XPUSHs (ST (0));
1731}
1732
1733int get_max_size (JSON *self)
814 CODE: 1734 CODE:
815{ 1735 RETVAL = DEC_SIZE (self->flags);
816 UV *uv = SvJSON (self);
817 if (enable)
818 *uv |= ix;
819 else
820 *uv &= ~ix;
821
822 RETVAL = newSVsv (self);
823}
824 OUTPUT: 1736 OUTPUT:
825 RETVAL 1737 RETVAL
826 1738
827void encode (SV *self, SV *scalar) 1739void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
828 PPCODE: 1740 PPCODE:
829 XPUSHs (encode_json (scalar, *SvJSON (self))); 1741{
1742 SvREFCNT_dec (self->cb_object);
1743 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
830 1744
831void decode (SV *self, SV *jsondata) 1745 XPUSHs (ST (0));
1746}
1747
1748void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
832 PPCODE: 1749 PPCODE:
1750{
1751 if (!self->cb_sk_object)
1752 self->cb_sk_object = newHV ();
1753
1754 if (SvOK (cb))
1755 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1756 else
1757 {
1758 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1759
1760 if (!HvKEYS (self->cb_sk_object))
1761 {
1762 SvREFCNT_dec (self->cb_sk_object);
1763 self->cb_sk_object = 0;
1764 }
1765 }
1766
1767 XPUSHs (ST (0));
1768}
1769
1770void encode (JSON *self, SV *scalar)
1771 PPCODE:
1772 XPUSHs (encode_json (scalar, self));
1773
1774void decode (JSON *self, SV *jsonstr)
1775 PPCODE:
833 XPUSHs (decode_json (jsondata, *SvJSON (self))); 1776 XPUSHs (decode_json (jsonstr, self, 0));
834 1777
1778void decode_prefix (JSON *self, SV *jsonstr)
1779 PPCODE:
1780{
1781 STRLEN offset;
1782 EXTEND (SP, 2);
1783 PUSHs (decode_json (jsonstr, self, &offset));
1784 PUSHs (sv_2mortal (newSVuv (offset)));
1785}
1786
1787void incr_parse (JSON *self, SV *jsonstr = 0)
1788 PPCODE:
1789{
1790 if (!self->incr_text)
1791 self->incr_text = newSVpvn ("", 0);
1792
1793 // append data, if any
1794 if (jsonstr)
1795 {
1796 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
1797 {
1798 /* utf-8-ness differs, need to upgrade */
1799 sv_utf8_upgrade (self->incr_text);
1800
1801 if (self->incr_pos)
1802 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1803 - (U8 *)SvPVX (self->incr_text);
1804 }
1805
1806 {
1807 STRLEN len;
1808 const char *str = SvPV (jsonstr, len);
1809 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
1810 Move (str, SvEND (self->incr_text), len, char);
1811 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1812 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1813 }
1814 }
1815
1816 if (GIMME_V != G_VOID)
1817 do
1818 {
1819 STRLEN offset;
1820
1821 if (!INCR_DONE (self))
1822 {
1823 incr_parse (self);
1824 if (!INCR_DONE (self))
1825 break;
1826 }
1827
1828 XPUSHs (decode_json (self->incr_text, self, &offset));
1829
1830 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1831 self->incr_pos -= offset;
1832 self->incr_nest = 0;
1833 self->incr_mode = 0;
1834 }
1835 while (GIMME_V == G_ARRAY);
1836}
1837
1838SV *incr_text (JSON *self)
1839 ATTRS: lvalue
1840 CODE:
1841{
1842 if (self->incr_pos)
1843 croak ("incr_text can not be called when the incremental parser already started parsing");
1844
1845 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1846}
1847 OUTPUT:
1848 RETVAL
1849
1850void incr_skip (JSON *self)
1851 CODE:
1852{
1853 if (self->incr_pos)
1854 {
1855 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
1856 self->incr_pos = 0;
1857 self->incr_nest = 0;
1858 self->incr_mode = 0;
1859 }
1860}
1861
1862void DESTROY (JSON *self)
1863 CODE:
1864 SvREFCNT_dec (self->cb_sk_object);
1865 SvREFCNT_dec (self->cb_object);
1866 SvREFCNT_dec (self->incr_text);
1867
1868PROTOTYPES: ENABLE
1869
1870void encode_json (SV *scalar)
1871 ALIAS:
1872 to_json_ = 0
1873 encode_json = F_UTF8
1874 PPCODE:
1875{
1876 JSON json = { F_DEFAULT | ix };
1877 XPUSHs (encode_json (scalar, &json));
1878}
1879
1880void decode_json (SV *jsonstr)
1881 ALIAS:
1882 from_json_ = 0
1883 decode_json = F_UTF8
1884 PPCODE:
1885{
1886 JSON json = { F_DEFAULT | ix };
1887 XPUSHs (decode_json (jsonstr, &json, 0));
1888}
1889
1890

Diff Legend

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