ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/JSON-XS/XS.xs
(Generate patch)

Comparing JSON-XS/XS.xs (file contents):
Revision 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs.
Revision 1.109 by root, Tue Mar 8 10:18:50 2011 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// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13
20#endif
21
22// three extra for rounding, sign, and end of string
23#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
24
9#define F_ASCII 0x00000001 25#define F_ASCII 0x00000001UL
26#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 27#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 28#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 29#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 30#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 31#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040
16#define F_ALLOW_NONREF 0x00000080 32#define F_ALLOW_NONREF 0x00000100UL
33#define F_SHRINK 0x00000200UL
34#define F_ALLOW_BLESSED 0x00000400UL
35#define F_CONV_BLESSED 0x00000800UL
36#define F_RELAXED 0x00001000UL
37#define F_ALLOW_UNKNOWN 0x00002000UL
38#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17 39
18#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 40#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19#define F_DEFAULT 0
20 41
21#define INIT_SIZE 32 // initial scalar size to be allocated 42#define INIT_SIZE 32 // initial scalar size to be allocated
43#define INDENT_STEP 3 // spaces per indentation level
44
45#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
46
47#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
22 48
23#define SB do { 49#define SB do {
24#define SE } while (0) 50#define SE } while (0)
25 51
26static HV *json_stash; 52#if __GNUC__ >= 3
53# define expect(expr,value) __builtin_expect ((expr), (value))
54# define INLINE static inline
55#else
56# define expect(expr,value) (expr)
57# define INLINE static
58#endif
59
60#define expect_false(expr) expect ((expr) != 0, 0)
61#define expect_true(expr) expect ((expr) != 0, 1)
62
63#define IN_RANGE_INC(type,val,beg,end) \
64 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
65 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
66
67#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
68
69#ifdef USE_ITHREADS
70# define JSON_SLOW 1
71# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
72#else
73# define JSON_SLOW 0
74# define JSON_STASH json_stash
75#endif
76
77static HV *json_stash, *json_boolean_stash; // JSON::XS::
78static SV *json_true, *json_false;
79
80enum {
81 INCR_M_WS = 0, // initial whitespace skipping, must be 0
82 INCR_M_STR, // inside string
83 INCR_M_BS, // inside backslash
84 INCR_M_C0, // inside comment in initial whitespace sequence
85 INCR_M_C1, // inside comment in other places
86 INCR_M_JSON // outside anything, count nesting
87};
88
89#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
90
91typedef struct {
92 U32 flags;
93 U32 max_depth;
94 STRLEN max_size;
95
96 SV *cb_object;
97 HV *cb_sk_object;
98
99 // for the incremental parser
100 SV *incr_text; // the source text so far
101 STRLEN incr_pos; // the current offset into the text
102 int incr_nest; // {[]}-nesting level
103 unsigned char incr_mode;
104} JSON;
105
106INLINE void
107json_init (JSON *json)
108{
109 Zero (json, 1, JSON);
110 json->max_depth = 512;
111}
112
113/////////////////////////////////////////////////////////////////////////////
114// utility functions
115
116INLINE SV *
117get_bool (const char *name)
118{
119 SV *sv = get_sv (name, 1);
120
121 SvREADONLY_on (sv);
122 SvREADONLY_on (SvRV (sv));
123
124 return sv;
125}
126
127INLINE void
128shrink (SV *sv)
129{
130 sv_utf8_downgrade (sv, 1);
131
132 if (SvLEN (sv) > SvCUR (sv) + 1)
133 {
134#ifdef SvPV_shrink_to_cur
135 SvPV_shrink_to_cur (sv);
136#elif defined (SvPV_renew)
137 SvPV_renew (sv, SvCUR (sv) + 1);
138#endif
139 }
140}
141
142// decode an utf-8 character and return it, or (UV)-1 in
143// case of an error.
144// we special-case "safe" characters from U+80 .. U+7FF,
145// but use the very good perl function to parse anything else.
146// note that we never call this function for a ascii codepoints
147INLINE UV
148decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
149{
150 if (expect_true (len >= 2
151 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
152 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
153 {
154 *clen = 2;
155 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
156 }
157 else
158 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
159}
160
161// likewise for encoding, also never called for ascii codepoints
162// this function takes advantage of this fact, although current gccs
163// seem to optimise the check for >= 0x80 away anyways
164INLINE unsigned char *
165encode_utf8 (unsigned char *s, UV ch)
166{
167 if (expect_false (ch < 0x000080))
168 *s++ = ch;
169 else if (expect_true (ch < 0x000800))
170 *s++ = 0xc0 | ( ch >> 6),
171 *s++ = 0x80 | ( ch & 0x3f);
172 else if ( ch < 0x010000)
173 *s++ = 0xe0 | ( ch >> 12),
174 *s++ = 0x80 | ((ch >> 6) & 0x3f),
175 *s++ = 0x80 | ( ch & 0x3f);
176 else if ( ch < 0x110000)
177 *s++ = 0xf0 | ( ch >> 18),
178 *s++ = 0x80 | ((ch >> 12) & 0x3f),
179 *s++ = 0x80 | ((ch >> 6) & 0x3f),
180 *s++ = 0x80 | ( ch & 0x3f);
181
182 return s;
183}
184
185// convert offset pointer to character index, sv must be string
186static STRLEN
187ptr_to_index (SV *sv, char *offset)
188{
189 return SvUTF8 (sv)
190 ? utf8_distance (offset, SvPVX (sv))
191 : offset - SvPVX (sv);
192}
193
194/////////////////////////////////////////////////////////////////////////////
195// fp hell
196
197// scan a group of digits, and a trailing exponent
198static void
199json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
200{
201 UV uaccum = 0;
202 int eaccum = 0;
203
204 // if we recurse too deep, skip all remaining digits
205 // to avoid a stack overflow attack
206 if (expect_false (--maxdepth <= 0))
207 while (((U8)*s - '0') < 10)
208 ++s;
209
210 for (;;)
211 {
212 U8 dig = (U8)*s - '0';
213
214 if (expect_false (dig >= 10))
215 {
216 if (dig == (U8)((U8)'.' - (U8)'0'))
217 {
218 ++s;
219 json_atof_scan1 (s, accum, expo, 1, maxdepth);
220 }
221 else if ((dig | ' ') == 'e' - '0')
222 {
223 int exp2 = 0;
224 int neg = 0;
225
226 ++s;
227
228 if (*s == '-')
229 {
230 ++s;
231 neg = 1;
232 }
233 else if (*s == '+')
234 ++s;
235
236 while ((dig = (U8)*s - '0') < 10)
237 exp2 = exp2 * 10 + *s++ - '0';
238
239 *expo += neg ? -exp2 : exp2;
240 }
241
242 break;
243 }
244
245 ++s;
246
247 uaccum = uaccum * 10 + dig;
248 ++eaccum;
249
250 // if we have too many digits, then recurse for more
251 // we actually do this for rather few digits
252 if (uaccum >= (UV_MAX - 9) / 10)
253 {
254 if (postdp) *expo -= eaccum;
255 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
256 if (postdp) *expo += eaccum;
257
258 break;
259 }
260 }
261
262 // this relies greatly on the quality of the pow ()
263 // implementation of the platform, but a good
264 // implementation is hard to beat.
265 if (postdp) *expo -= eaccum;
266 *accum += uaccum * Perl_pow (10., *expo);
267 *expo += eaccum;
268}
269
270static NV
271json_atof (const char *s)
272{
273 NV accum = 0.;
274 int expo = 0;
275 int neg = 0;
276
277 if (*s == '-')
278 {
279 ++s;
280 neg = 1;
281 }
282
283 // a recursion depth of ten gives us >>500 bits
284 json_atof_scan1 (s, &accum, &expo, 0, 10);
285
286 return neg ? -accum : accum;
287}
288/////////////////////////////////////////////////////////////////////////////
289// encoder
27 290
28// structure used for encoding JSON 291// structure used for encoding JSON
29typedef struct 292typedef struct
30{ 293{
31 char *cur; 294 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 295 char *end; // SvEND (sv)
34 SV *sv; 296 SV *sv; // result scalar
35 UV flags; 297 JSON json;
36 int max_recurse; 298 U32 indent; // indentation level
37 int indent; 299 UV limit; // escape character values >= this value when encoding
38} enc_t; 300} enc_t;
39 301
40// structure used for decoding JSON 302INLINE void
41typedef struct
42{
43 char *cur;
44 char *end;
45 const char *err;
46 UV flags;
47} dec_t;
48
49static UV *
50SvJSON (SV *sv)
51{
52 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53 croak ("object is not of type JSON::XS");
54
55 return &SvUVX (SvRV (sv));
56}
57
58/////////////////////////////////////////////////////////////////////////////
59
60static void
61need (enc_t *enc, STRLEN len) 303need (enc_t *enc, STRLEN len)
62{ 304{
63 if (enc->cur + len >= enc->end) 305 if (expect_false (enc->cur + len >= enc->end))
64 { 306 {
65 STRLEN cur = enc->cur - SvPVX (enc->sv); 307 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66 SvGROW (enc->sv, cur + len + 1); 308 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
67 enc->cur = SvPVX (enc->sv) + cur; 309 enc->cur = SvPVX (enc->sv) + cur;
68 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 310 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69 } 311 }
70} 312}
71 313
72static void 314INLINE void
73encode_ch (enc_t *enc, char ch) 315encode_ch (enc_t *enc, char ch)
74{ 316{
75 need (enc, 1); 317 need (enc, 1);
76 *enc->cur++ = ch; 318 *enc->cur++ = ch;
77} 319}
85 327
86 while (str < end) 328 while (str < end)
87 { 329 {
88 unsigned char ch = *(unsigned char *)str; 330 unsigned char ch = *(unsigned char *)str;
89 331
90 if (ch == '"') 332 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91 { 333 {
334 if (expect_false (ch == '"')) // but with slow exceptions
335 {
92 need (enc, len += 1); 336 need (enc, len += 1);
93 *enc->cur++ = '\\'; 337 *enc->cur++ = '\\';
94 *enc->cur++ = '"'; 338 *enc->cur++ = '"';
95 ++str;
96 } 339 }
97 else if (ch == '\\') 340 else if (expect_false (ch == '\\'))
98 { 341 {
99 need (enc, len += 1); 342 need (enc, len += 1);
100 *enc->cur++ = '\\'; 343 *enc->cur++ = '\\';
101 *enc->cur++ = '\\'; 344 *enc->cur++ = '\\';
102 ++str;
103 } 345 }
104 else if (ch >= 0x20 && ch < 0x80) // most common case 346 else
105 {
106 *enc->cur++ = ch; 347 *enc->cur++ = ch;
107 ++str; 348
108 }
109 else if (ch == '\015')
110 {
111 need (enc, len += 1);
112 *enc->cur++ = '\\';
113 *enc->cur++ = 'r';
114 ++str;
115 }
116 else if (ch == '\012')
117 {
118 need (enc, len += 1);
119 *enc->cur++ = '\\';
120 *enc->cur++ = 'n';
121 ++str; 349 ++str;
122 } 350 }
123 else 351 else
124 { 352 {
125 STRLEN clen; 353 switch (ch)
126 UV uch;
127
128 if (is_utf8)
129 {
130 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
131 if (clen < 0)
132 croak ("malformed UTF-8 character in string, cannot convert to JSON");
133 } 354 {
134 else 355 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
135 { 356 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
136 uch = ch; 357 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
137 clen = 1; 358 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
138 } 359 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
139 360
140 if (uch < 0x80 || enc->flags & F_ASCII) 361 default:
141 {
142 if (uch > 0xFFFFUL)
143 { 362 {
363 STRLEN clen;
364 UV uch;
365
366 if (is_utf8)
367 {
368 uch = decode_utf8 (str, end - str, &clen);
369 if (clen == (STRLEN)-1)
370 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
371 }
372 else
373 {
374 uch = ch;
375 clen = 1;
376 }
377
378 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
379 {
380 if (uch >= 0x10000UL)
381 {
382 if (uch >= 0x110000UL)
383 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
384
144 need (enc, len += 11); 385 need (enc, len += 11);
145 sprintf (enc->cur, "\\u%04x\\u%04x", 386 sprintf (enc->cur, "\\u%04x\\u%04x",
146 (uch - 0x10000) / 0x400 + 0xD800, 387 (int)((uch - 0x10000) / 0x400 + 0xD800),
147 (uch - 0x10000) % 0x400 + 0xDC00); 388 (int)((uch - 0x10000) % 0x400 + 0xDC00));
148 enc->cur += 12; 389 enc->cur += 12;
390 }
391 else
392 {
393 need (enc, len += 5);
394 *enc->cur++ = '\\';
395 *enc->cur++ = 'u';
396 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
397 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
398 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
399 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
400 }
401
402 str += clen;
403 }
404 else if (enc->json.flags & F_LATIN1)
405 {
406 *enc->cur++ = uch;
407 str += clen;
408 }
409 else if (is_utf8)
410 {
411 need (enc, len += clen);
412 do
413 {
414 *enc->cur++ = *str++;
415 }
416 while (--clen);
417 }
418 else
419 {
420 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
421 enc->cur = encode_utf8 (enc->cur, uch);
422 ++str;
423 }
149 } 424 }
150 else
151 {
152 static char hexdigit [16] = "0123456789abcdef";
153 need (enc, len += 5);
154 *enc->cur++ = '\\';
155 *enc->cur++ = 'u';
156 *enc->cur++ = hexdigit [ uch >> 12 ];
157 *enc->cur++ = hexdigit [(uch >> 8) & 15];
158 *enc->cur++ = hexdigit [(uch >> 4) & 15];
159 *enc->cur++ = hexdigit [(uch >> 0) & 15];
160 }
161
162 str += clen;
163 }
164 else if (is_utf8)
165 {
166 need (enc, len += clen);
167 while (clen--)
168 *enc->cur++ = *str++;
169 }
170 else
171 {
172 need (enc, 10); // never more than 11 bytes needed
173 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174 ++str;
175 } 425 }
176 } 426 }
177 427
178 --len; 428 --len;
179 } 429 }
180} 430}
181 431
182#define INDENT SB \ 432INLINE void
433encode_indent (enc_t *enc)
434{
183 if (enc->flags & F_INDENT) \ 435 if (enc->json.flags & F_INDENT)
184 { \ 436 {
185 int i_; \ 437 int spaces = enc->indent * INDENT_STEP;
186 need (enc, enc->indent); \ 438
187 for (i_ = enc->indent * 3; i_--; )\ 439 need (enc, spaces);
440 memset (enc->cur, ' ', spaces);
441 enc->cur += spaces;
442 }
443}
444
445INLINE void
446encode_space (enc_t *enc)
447{
448 need (enc, 1);
449 encode_ch (enc, ' ');
450}
451
452INLINE void
453encode_nl (enc_t *enc)
454{
455 if (enc->json.flags & F_INDENT)
456 {
457 need (enc, 1);
188 encode_ch (enc, ' '); \ 458 encode_ch (enc, '\n');
189 } \ 459 }
190 SE 460}
191 461
192#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 462INLINE void
193#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 463encode_comma (enc_t *enc)
194#define COMMA SB \ 464{
195 encode_ch (enc, ','); \ 465 encode_ch (enc, ',');
466
196 if (enc->flags & F_INDENT) \ 467 if (enc->json.flags & F_INDENT)
197 NL; \ 468 encode_nl (enc);
198 else if (enc->flags & F_SPACE_AFTER) \ 469 else if (enc->json.flags & F_SPACE_AFTER)
199 SPACE; \ 470 encode_space (enc);
200 SE 471}
201 472
202static void encode_sv (enc_t *enc, SV *sv); 473static void encode_sv (enc_t *enc, SV *sv);
203 474
204static void 475static void
205encode_av (enc_t *enc, AV *av) 476encode_av (enc_t *enc, AV *av)
206{ 477{
207 int i, len = av_len (av); 478 int i, len = av_len (av);
208 479
480 if (enc->indent >= enc->json.max_depth)
481 croak (ERR_NESTING_EXCEEDED);
482
209 encode_ch (enc, '['); NL; 483 encode_ch (enc, '[');
210 ++enc->indent; 484
485 if (len >= 0)
486 {
487 encode_nl (enc); ++enc->indent;
211 488
212 for (i = 0; i <= len; ++i) 489 for (i = 0; i <= len; ++i)
213 { 490 {
214 INDENT; 491 SV **svp = av_fetch (av, i, 0);
215 encode_sv (enc, *av_fetch (av, i, 0));
216 492
493 encode_indent (enc);
494
495 if (svp)
496 encode_sv (enc, *svp);
497 else
498 encode_str (enc, "null", 4, 0);
499
217 if (i < len) 500 if (i < len)
218 COMMA; 501 encode_comma (enc);
219 } 502 }
220 503
221 NL; 504 encode_nl (enc); --enc->indent; encode_indent (enc);
222 505 }
223 --enc->indent; 506
224 INDENT; encode_ch (enc, ']'); 507 encode_ch (enc, ']');
225} 508}
226 509
227static void 510static void
228encode_he (enc_t *enc, HE *he) 511encode_hk (enc_t *enc, HE *he)
229{ 512{
230 encode_ch (enc, '"'); 513 encode_ch (enc, '"');
231 514
232 if (HeKLEN (he) == HEf_SVKEY) 515 if (HeKLEN (he) == HEf_SVKEY)
233 { 516 {
243 else 526 else
244 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 527 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245 528
246 encode_ch (enc, '"'); 529 encode_ch (enc, '"');
247 530
248 if (enc->flags & F_SPACE_BEFORE) SPACE; 531 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249 encode_ch (enc, ':'); 532 encode_ch (enc, ':');
250 if (enc->flags & F_SPACE_AFTER ) SPACE; 533 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251 encode_sv (enc, HeVAL (he));
252} 534}
253 535
254// compare hash entries, used when all keys are bytestrings 536// compare hash entries, used when all keys are bytestrings
255static int 537static int
256he_cmp_fast (const void *a_, const void *b_) 538he_cmp_fast (const void *a_, const void *b_)
261 HE *b = *(HE **)b_; 543 HE *b = *(HE **)b_;
262 544
263 STRLEN la = HeKLEN (a); 545 STRLEN la = HeKLEN (a);
264 STRLEN lb = HeKLEN (b); 546 STRLEN lb = HeKLEN (b);
265 547
266 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 548 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267 cmp = la < lb ? -1 : la == lb ? 0 : 1; 549 cmp = lb - la;
268 550
269 return cmp; 551 return cmp;
270} 552}
271 553
272// compare hash entries, used when some keys are sv's or utf-x 554// compare hash entries, used when some keys are sv's or utf-x
273static int 555static int
274he_cmp_slow (const void *a, const void *b) 556he_cmp_slow (const void *a, const void *b)
275{ 557{
276 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 558 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
277} 559}
278 560
279static void 561static void
280encode_hv (enc_t *enc, HV *hv) 562encode_hv (enc_t *enc, HV *hv)
281{ 563{
282 int count, i; 564 HE *he;
283 565
284 encode_ch (enc, '{'); NL; ++enc->indent; 566 if (enc->indent >= enc->json.max_depth)
567 croak (ERR_NESTING_EXCEEDED);
285 568
286 if ((count = hv_iterinit (hv))) 569 encode_ch (enc, '{');
287 { 570
288 // for canonical output we have to sort by keys first 571 // for canonical output we have to sort by keys first
289 // actually, this is mostly due to the stupid so-called 572 // actually, this is mostly due to the stupid so-called
290 // security workaround added somewhere in 5.8.x. 573 // security workaround added somewhere in 5.8.x
291 // that randomises hash orderings 574 // that randomises hash orderings
292 if (enc->flags & F_CANONICAL) 575 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
576 {
577 int count = hv_iterinit (hv);
578
579 if (SvMAGICAL (hv))
293 { 580 {
294 HE *he, *hes [count]; 581 // need to count by iterating. could improve by dynamically building the vector below
582 // but I don't care for the speed of this special case.
583 // note also that we will run into undefined behaviour when the two iterations
584 // do not result in the same count, something I might care for in some later release.
585
586 count = 0;
587 while (hv_iternext (hv))
588 ++count;
589
590 hv_iterinit (hv);
591 }
592
593 if (count)
594 {
295 int fast = 1; 595 int i, fast = 1;
596#if defined(__BORLANDC__) || defined(_MSC_VER)
597 HE **hes = _alloca (count * sizeof (HE));
598#else
599 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
600#endif
296 601
297 i = 0; 602 i = 0;
298 while ((he = hv_iternext (hv))) 603 while ((he = hv_iternext (hv)))
299 { 604 {
300 hes [i++] = he; 605 hes [i++] = he;
306 611
307 if (fast) 612 if (fast)
308 qsort (hes, count, sizeof (HE *), he_cmp_fast); 613 qsort (hes, count, sizeof (HE *), he_cmp_fast);
309 else 614 else
310 { 615 {
311 // hack to disable "use bytes" 616 // hack to forcefully disable "use bytes"
312 COP *oldcop = PL_curcop, cop; 617 COP cop = *PL_curcop;
313 cop.op_private = 0; 618 cop.op_private = 0;
619
620 ENTER;
621 SAVETMPS;
622
623 SAVEVPTR (PL_curcop);
314 PL_curcop = &cop; 624 PL_curcop = &cop;
315 625
316 SAVETMPS;
317 qsort (hes, count, sizeof (HE *), he_cmp_slow); 626 qsort (hes, count, sizeof (HE *), he_cmp_slow);
627
318 FREETMPS; 628 FREETMPS;
629 LEAVE;
630 }
319 631
320 PL_curcop = oldcop; 632 encode_nl (enc); ++enc->indent;
633
634 while (count--)
321 } 635 {
636 encode_indent (enc);
637 he = hes [count];
638 encode_hk (enc, he);
639 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
322 640
323 for (i = 0; i < count; ++i) 641 if (count)
642 encode_comma (enc);
643 }
644
645 encode_nl (enc); --enc->indent; encode_indent (enc);
646 }
647 }
648 else
649 {
650 if (hv_iterinit (hv) || SvMAGICAL (hv))
651 if ((he = hv_iternext (hv)))
652 {
653 encode_nl (enc); ++enc->indent;
654
655 for (;;)
324 { 656 {
325 INDENT; 657 encode_indent (enc);
326 encode_he (enc, hes [i]); 658 encode_hk (enc, he);
659 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
327 660
328 if (i < count - 1) 661 if (!(he = hv_iternext (hv)))
329 COMMA; 662 break;
663
664 encode_comma (enc);
330 } 665 }
331 666
667 encode_nl (enc); --enc->indent; encode_indent (enc);
332 NL; 668 }
669 }
670
671 encode_ch (enc, '}');
672}
673
674// encode objects, arrays and special \0=false and \1=true values.
675static void
676encode_rv (enc_t *enc, SV *sv)
677{
678 svtype svt;
679
680 SvGETMAGIC (sv);
681 svt = SvTYPE (sv);
682
683 if (expect_false (SvOBJECT (sv)))
684 {
685 HV *stash = !JSON_SLOW || json_boolean_stash
686 ? json_boolean_stash
687 : gv_stashpv ("JSON::XS::Boolean", 1);
688
689 if (SvSTASH (sv) == stash)
690 {
691 if (SvIV (sv))
692 encode_str (enc, "true", 4, 0);
693 else
694 encode_str (enc, "false", 5, 0);
333 } 695 }
334 else 696 else
335 { 697 {
336 SV *sv; 698#if 0
337 HE *he = hv_iternext (hv); 699 if (0 && sv_derived_from (rv, "JSON::Literal"))
338
339 for (;;)
340 {
341 INDENT;
342 encode_he (enc, he);
343
344 if (!(he = hv_iternext (hv)))
345 break;
346
347 COMMA;
348 } 700 {
349 701 // not yet
350 NL; 702 }
703#endif
704 if (enc->json.flags & F_CONV_BLESSED)
705 {
706 // we re-bless the reference to get overload and other niceties right
707 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
708
709 if (to_json)
710 {
711 dSP;
712
713 ENTER; SAVETMPS; PUSHMARK (SP);
714 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
715
716 // calling with G_SCALAR ensures that we always get a 1 return value
717 PUTBACK;
718 call_sv ((SV *)GvCV (to_json), G_SCALAR);
719 SPAGAIN;
720
721 // catch this surprisingly common error
722 if (SvROK (TOPs) && SvRV (TOPs) == sv)
723 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
724
725 sv = POPs;
726 PUTBACK;
727
728 encode_sv (enc, sv);
729
730 FREETMPS; LEAVE;
731 }
732 else if (enc->json.flags & F_ALLOW_BLESSED)
733 encode_str (enc, "null", 4, 0);
734 else
735 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
736 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
737 }
738 else if (enc->json.flags & F_ALLOW_BLESSED)
739 encode_str (enc, "null", 4, 0);
740 else
741 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
742 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
351 } 743 }
352 } 744 }
745 else if (svt == SVt_PVHV)
746 encode_hv (enc, (HV *)sv);
747 else if (svt == SVt_PVAV)
748 encode_av (enc, (AV *)sv);
749 else if (svt < SVt_PVAV)
750 {
751 STRLEN len = 0;
752 char *pv = svt ? SvPV (sv, len) : 0;
353 753
354 --enc->indent; INDENT; encode_ch (enc, '}'); 754 if (len == 1 && *pv == '1')
755 encode_str (enc, "true", 4, 0);
756 else if (len == 1 && *pv == '0')
757 encode_str (enc, "false", 5, 0);
758 else if (enc->json.flags & F_ALLOW_UNKNOWN)
759 encode_str (enc, "null", 4, 0);
760 else
761 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
762 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
763 }
764 else if (enc->json.flags & F_ALLOW_UNKNOWN)
765 encode_str (enc, "null", 4, 0);
766 else
767 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
768 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355} 769}
356 770
357static void 771static void
358encode_sv (enc_t *enc, SV *sv) 772encode_sv (enc_t *enc, SV *sv)
359{ 773{
367 encode_str (enc, str, len, SvUTF8 (sv)); 781 encode_str (enc, str, len, SvUTF8 (sv));
368 encode_ch (enc, '"'); 782 encode_ch (enc, '"');
369 } 783 }
370 else if (SvNOKp (sv)) 784 else if (SvNOKp (sv))
371 { 785 {
786 // trust that perl will do the right thing w.r.t. JSON syntax.
372 need (enc, NV_DIG + 32); 787 need (enc, NV_DIG + 32);
373 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 788 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374 enc->cur += strlen (enc->cur); 789 enc->cur += strlen (enc->cur);
375 } 790 }
376 else if (SvIOKp (sv)) 791 else if (SvIOKp (sv))
377 { 792 {
793 // we assume we can always read an IV as a UV and vice versa
794 // we assume two's complement
795 // we assume no aliasing issues in the union
796 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
797 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
798 {
799 // optimise the "small number case"
800 // code will likely be branchless and use only a single multiplication
801 // works for numbers up to 59074
802 I32 i = SvIVX (sv);
803 U32 u;
804 char digit, nz = 0;
805
378 need (enc, 64); 806 need (enc, 6);
807
808 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
809 u = i < 0 ? -i : i;
810
811 // convert to 4.28 fixed-point representation
812 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
813
814 // now output digit by digit, each time masking out the integer part
815 // and multiplying by 5 while moving the decimal point one to the right,
816 // resulting in a net multiplication by 10.
817 // we always write the digit to memory but conditionally increment
818 // the pointer, to enable the use of conditional move instructions.
819 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
820 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
821 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
822 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
823 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
824 }
825 else
826 {
827 // large integer, use the (rather slow) snprintf way.
828 need (enc, IVUV_MAXCHARS);
379 enc->cur += 829 enc->cur +=
380 SvIsUV(sv) 830 SvIsUV(sv)
381 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 831 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 832 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
833 }
383 } 834 }
384 else if (SvROK (sv)) 835 else if (SvROK (sv))
385 { 836 encode_rv (enc, SvRV (sv));
386 if (!--enc->max_recurse) 837 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
387 croak ("data structure too deep (hit recursion limit)");
388
389 sv = SvRV (sv);
390
391 switch (SvTYPE (sv))
392 {
393 case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394 case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396 default:
397 croak ("JSON can only represent references to arrays or hashes");
398 }
399 }
400 else if (!SvOK (sv))
401 encode_str (enc, "null", 4, 0); 838 encode_str (enc, "null", 4, 0);
402 else 839 else
403 croak ("encountered perl type that JSON cannot handle"); 840 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
841 SvPV_nolen (sv), SvFLAGS (sv));
404} 842}
405 843
406static SV * 844static SV *
407encode_json (SV *scalar, UV flags) 845encode_json (SV *scalar, JSON *json)
408{ 846{
409 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410 croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
411
412 enc_t enc; 847 enc_t enc;
413 enc.flags = flags; 848
849 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
850 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
851
852 enc.json = *json;
414 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 853 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415 enc.cur = SvPVX (enc.sv); 854 enc.cur = SvPVX (enc.sv);
416 enc.end = SvEND (enc.sv); 855 enc.end = SvEND (enc.sv);
417 enc.max_recurse = 0;
418 enc.indent = 0; 856 enc.indent = 0;
857 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
858 : enc.json.flags & F_LATIN1 ? 0x000100UL
859 : 0x110000UL;
419 860
420 SvPOK_only (enc.sv); 861 SvPOK_only (enc.sv);
421 encode_sv (&enc, scalar); 862 encode_sv (&enc, scalar);
863 encode_nl (&enc);
422 864
865 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
866 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
867
423 if (!(flags & (F_ASCII | F_UTF8))) 868 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424 SvUTF8_on (enc.sv); 869 SvUTF8_on (enc.sv);
425 870
426 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 871 if (enc.json.flags & F_SHRINK)
872 shrink (enc.sv);
873
427 return enc.sv; 874 return enc.sv;
428} 875}
429 876
430///////////////////////////////////////////////////////////////////////////// 877/////////////////////////////////////////////////////////////////////////////
878// decoder
431 879
432#define WS \ 880// structure used for decoding JSON
881typedef struct
882{
883 char *cur; // current parser pointer
884 char *end; // end of input string
885 const char *err; // parse error, if != 0
886 JSON json;
887 U32 depth; // recursion depth
888 U32 maxdepth; // recursion depth limit
889} dec_t;
890
891INLINE void
892decode_comment (dec_t *dec)
893{
894 // only '#'-style comments allowed a.t.m.
895
896 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
897 ++dec->cur;
898}
899
900INLINE void
901decode_ws (dec_t *dec)
902{
433 for (;;) \ 903 for (;;)
434 { \ 904 {
435 char ch = *dec->cur; \ 905 char ch = *dec->cur;
906
436 if (ch > 0x20 \ 907 if (ch > 0x20)
908 {
909 if (expect_false (ch == '#'))
910 {
911 if (dec->json.flags & F_RELAXED)
912 decode_comment (dec);
913 else
914 break;
915 }
916 else
917 break;
918 }
437 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 919 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438 break; \ 920 break; // parse error, but let higher level handle it, gives better error messages
921
439 ++dec->cur; \ 922 ++dec->cur;
440 } 923 }
924}
441 925
442#define ERR(reason) SB dec->err = reason; goto fail; SE 926#define ERR(reason) SB dec->err = reason; goto fail; SE
927
443#define EXPECT_CH(ch) SB \ 928#define EXPECT_CH(ch) SB \
444 if (*dec->cur != ch) \ 929 if (*dec->cur != ch) \
445 ERR (# ch " expected"); \ 930 ERR (# ch " expected"); \
446 ++dec->cur; \ 931 ++dec->cur; \
447 SE 932 SE
448 933
934#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
935#define DEC_DEC_DEPTH --dec->depth
936
449static SV *decode_sv (dec_t *dec); 937static SV *decode_sv (dec_t *dec);
450 938
451static signed char decode_hexdigit[256]; 939static signed char decode_hexdigit[256];
452 940
453static UV 941static UV
454decode_4hex (dec_t *dec) 942decode_4hex (dec_t *dec)
455{ 943{
456 signed char d1, d2, d3, d4; 944 signed char d1, d2, d3, d4;
945 unsigned char *cur = (unsigned char *)dec->cur;
457 946
458 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 947 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
459 if (d1 < 0) ERR ("four hexadecimal digits expected"); 948 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
460 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 949 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
461 if (d2 < 0) ERR ("four hexadecimal digits expected"); 950 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
462 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463 if (d3 < 0) ERR ("four hexadecimal digits expected");
464 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465 if (d4 < 0) ERR ("four hexadecimal digits expected");
466 951
467 dec->cur += 4; 952 dec->cur += 4;
468 953
469 return ((UV)d1) << 12 954 return ((UV)d1) << 12
470 | ((UV)d2) << 8 955 | ((UV)d2) << 8
473 958
474fail: 959fail:
475 return (UV)-1; 960 return (UV)-1;
476} 961}
477 962
478#define APPEND_GROW(n) SB \
479 if (cur + (n) >= end) \
480 { \
481 STRLEN ofs = cur - SvPVX (sv); \
482 SvGROW (sv, ofs + (n) + 1); \
483 cur = SvPVX (sv) + ofs; \
484 end = SvEND (sv); \
485 } \
486 SE
487
488#define APPEND_CH(ch) SB \
489 APPEND_GROW (1); \
490 *cur++ = (ch); \
491 SE
492
493static SV * 963static SV *
494decode_str (dec_t *dec) 964decode_str (dec_t *dec)
495{ 965{
496 SV *sv = NEWSV (0,2); 966 SV *sv = 0;
497 int utf8 = 0; 967 int utf8 = 0;
498 char *cur = SvPVX (sv); 968 char *dec_cur = dec->cur;
499 char *end = SvEND (sv);
500 969
501 for (;;) 970 do
502 { 971 {
503 unsigned char ch = *(unsigned char *)dec->cur; 972 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
973 char *cur = buf;
504 974
505 if (ch == '"') 975 do
506 break;
507 else if (ch == '\\')
508 { 976 {
509 switch (*++dec->cur) 977 unsigned char ch = *(unsigned char *)dec_cur++;
978
979 if (expect_false (ch == '"'))
980 {
981 --dec_cur;
982 break;
510 { 983 }
511 case '\\': 984 else if (expect_false (ch == '\\'))
512 case '/': 985 {
513 case '"': APPEND_CH (*dec->cur++); break; 986 switch (*dec_cur)
514
515 case 'b': APPEND_CH ('\010'); ++dec->cur; break;
516 case 't': APPEND_CH ('\011'); ++dec->cur; break;
517 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521 case 'u':
522 { 987 {
523 UV lo, hi; 988 case '\\':
524 ++dec->cur; 989 case '/':
990 case '"': *cur++ = *dec_cur++; break;
525 991
526 hi = decode_4hex (dec); 992 case 'b': ++dec_cur; *cur++ = '\010'; break;
527 if (hi == (UV)-1) 993 case 't': ++dec_cur; *cur++ = '\011'; break;
528 goto fail; 994 case 'n': ++dec_cur; *cur++ = '\012'; break;
995 case 'f': ++dec_cur; *cur++ = '\014'; break;
996 case 'r': ++dec_cur; *cur++ = '\015'; break;
529 997
530 // possibly a surrogate pair 998 case 'u':
531 if (hi >= 0xd800 && hi < 0xdc00)
532 { 999 {
533 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 1000 UV lo, hi;
534 ERR ("illegal surrogate character"); 1001 ++dec_cur;
535 1002
536 dec->cur += 2; 1003 dec->cur = dec_cur;
537
538 lo = decode_4hex (dec); 1004 hi = decode_4hex (dec);
1005 dec_cur = dec->cur;
539 if (lo == (UV)-1) 1006 if (hi == (UV)-1)
540 goto fail; 1007 goto fail;
541 1008
1009 // possibly a surrogate pair
1010 if (hi >= 0xd800)
1011 if (hi < 0xdc00)
1012 {
1013 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
1014 ERR ("missing low surrogate character in surrogate pair");
1015
1016 dec_cur += 2;
1017
1018 dec->cur = dec_cur;
1019 lo = decode_4hex (dec);
1020 dec_cur = dec->cur;
1021 if (lo == (UV)-1)
1022 goto fail;
1023
542 if (lo < 0xdc00 || lo >= 0xe000) 1024 if (lo < 0xdc00 || lo >= 0xe000)
543 ERR ("surrogate pair expected"); 1025 ERR ("surrogate pair expected");
544 1026
545 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 1027 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
1028 }
1029 else if (hi < 0xe000)
1030 ERR ("missing high surrogate character in surrogate pair");
1031
1032 if (hi >= 0x80)
1033 {
1034 utf8 = 1;
1035
1036 cur = encode_utf8 (cur, hi);
1037 }
1038 else
1039 *cur++ = hi;
546 } 1040 }
547 else if (lo >= 0xdc00 && lo < 0xe000)
548 ERR ("illegal surrogate character");
549
550 if (hi >= 0x80)
551 { 1041 break;
552 utf8 = 1;
553 1042
554 APPEND_GROW (4); // at most 4 bytes for 21 bits
555 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556 }
557 else 1043 default:
558 APPEND_CH (hi); 1044 --dec_cur;
1045 ERR ("illegal backslash escape sequence in string");
559 } 1046 }
1047 }
1048 else if (expect_true (ch >= 0x20 && ch < 0x80))
1049 *cur++ = ch;
1050 else if (ch >= 0x80)
1051 {
1052 STRLEN clen;
1053
1054 --dec_cur;
1055
1056 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
1057 if (clen == (STRLEN)-1)
1058 ERR ("malformed UTF-8 character in JSON string");
1059
1060 do
1061 *cur++ = *dec_cur++;
1062 while (--clen);
1063
1064 utf8 = 1;
1065 }
1066 else
1067 {
1068 --dec_cur;
1069
1070 if (!ch)
1071 ERR ("unexpected end of string while parsing JSON string");
560 break; 1072 else
1073 ERR ("invalid character encountered while parsing JSON string");
561 } 1074 }
562 } 1075 }
563 else if (ch >= 0x20 && ch <= 0x7f) 1076 while (cur < buf + SHORT_STRING_LEN);
564 APPEND_CH (*dec->cur++); 1077
565 else if (ch >= 0x80) 1078 {
1079 STRLEN len = cur - buf;
1080
1081 if (sv)
566 { 1082 {
567 STRLEN clen; 1083 STRLEN cur = SvCUR (sv);
568 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
569 if (clen < 0)
570 ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571 1084
572 APPEND_GROW (clen); 1085 if (SvLEN (sv) <= cur + len)
573 memcpy (cur, dec->cur, clen); 1086 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
574 cur += clen; 1087
575 dec->cur += clen; 1088 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
1089 SvCUR_set (sv, SvCUR (sv) + len);
576 } 1090 }
577 else 1091 else
578 ERR ("invalid character encountered"); 1092 sv = newSVpvn (buf, len);
579 } 1093 }
1094 }
1095 while (*dec_cur != '"');
580 1096
581 ++dec->cur; 1097 ++dec_cur;
582 1098
583 SvCUR_set (sv, cur - SvPVX (sv)); 1099 if (sv)
584 1100 {
585 SvPOK_only (sv); 1101 SvPOK_only (sv);
586 *SvEND (sv) = 0; 1102 *SvEND (sv) = 0;
587 1103
588 if (utf8) 1104 if (utf8)
589 SvUTF8_on (sv); 1105 SvUTF8_on (sv);
1106 }
1107 else
1108 sv = newSVpvn ("", 0);
590 1109
1110 dec->cur = dec_cur;
591 return sv; 1111 return sv;
592 1112
593fail: 1113fail:
594 SvREFCNT_dec (sv); 1114 dec->cur = dec_cur;
595 return 0; 1115 return 0;
596} 1116}
597 1117
598static SV * 1118static SV *
599decode_num (dec_t *dec) 1119decode_num (dec_t *dec)
609 { 1129 {
610 ++dec->cur; 1130 ++dec->cur;
611 if (*dec->cur >= '0' && *dec->cur <= '9') 1131 if (*dec->cur >= '0' && *dec->cur <= '9')
612 ERR ("malformed number (leading zero must not be followed by another digit)"); 1132 ERR ("malformed number (leading zero must not be followed by another digit)");
613 } 1133 }
614 1134 else if (*dec->cur < '0' || *dec->cur > '9')
615 // int 1135 ERR ("malformed number (no digits after initial minus)");
1136 else
1137 do
1138 {
1139 ++dec->cur;
1140 }
616 while (*dec->cur >= '0' && *dec->cur <= '9') 1141 while (*dec->cur >= '0' && *dec->cur <= '9');
617 ++dec->cur;
618 1142
619 // [frac] 1143 // [frac]
620 if (*dec->cur == '.') 1144 if (*dec->cur == '.')
621 { 1145 {
622 is_nv = 1; 1146 ++dec->cur;
1147
1148 if (*dec->cur < '0' || *dec->cur > '9')
1149 ERR ("malformed number (no digits after decimal point)");
623 1150
624 do 1151 do
625 { 1152 {
626 ++dec->cur; 1153 ++dec->cur;
627 } 1154 }
628 while (*dec->cur >= '0' && *dec->cur <= '9'); 1155 while (*dec->cur >= '0' && *dec->cur <= '9');
1156
1157 is_nv = 1;
629 } 1158 }
630 1159
631 // [exp] 1160 // [exp]
632 if (*dec->cur == 'e' || *dec->cur == 'E') 1161 if (*dec->cur == 'e' || *dec->cur == 'E')
633 { 1162 {
634 is_nv = 1;
635
636 ++dec->cur; 1163 ++dec->cur;
1164
637 if (*dec->cur == '-' || *dec->cur == '+') 1165 if (*dec->cur == '-' || *dec->cur == '+')
638 ++dec->cur; 1166 ++dec->cur;
639 1167
1168 if (*dec->cur < '0' || *dec->cur > '9')
1169 ERR ("malformed number (no digits after exp sign)");
1170
1171 do
1172 {
1173 ++dec->cur;
1174 }
640 while (*dec->cur >= '0' && *dec->cur <= '9') 1175 while (*dec->cur >= '0' && *dec->cur <= '9');
641 ++dec->cur; 1176
1177 is_nv = 1;
642 } 1178 }
643 1179
644 if (!is_nv) 1180 if (!is_nv)
645 { 1181 {
646 UV uv; 1182 int len = dec->cur - start;
647 int numtype = grok_number (start, dec->cur - start, &uv); 1183
648 if (numtype & IS_NUMBER_IN_UV) 1184 // special case the rather common 1..5-digit-int case
649 if (numtype & IS_NUMBER_NEG) 1185 if (*start == '-')
1186 switch (len)
650 { 1187 {
651 if (uv < (UV)IV_MIN) 1188 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
652 return newSViv (-(IV)uv); 1189 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1190 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1191 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1192 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
653 } 1193 }
1194 else
1195 switch (len)
1196 {
1197 case 1: return newSViv ( start [0] - '0' * 1);
1198 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1199 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1200 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1201 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1202 }
1203
1204 {
1205 UV uv;
1206 int numtype = grok_number (start, len, &uv);
1207 if (numtype & IS_NUMBER_IN_UV)
1208 if (numtype & IS_NUMBER_NEG)
1209 {
1210 if (uv < (UV)IV_MIN)
1211 return newSViv (-(IV)uv);
1212 }
654 else 1213 else
655 return newSVuv (uv); 1214 return newSVuv (uv);
656 } 1215 }
657 1216
1217 len -= *start == '-' ? 1 : 0;
1218
1219 // does not fit into IV or UV, try NV
1220 if (len <= NV_DIG)
1221 // fits into NV without loss of precision
1222 return newSVnv (json_atof (start));
1223
1224 // everything else fails, convert it to a string
1225 return newSVpvn (start, dec->cur - start);
1226 }
1227
1228 // loss of precision here
658 return newSVnv (Atof (start)); 1229 return newSVnv (json_atof (start));
659 1230
660fail: 1231fail:
661 return 0; 1232 return 0;
662} 1233}
663 1234
664static SV * 1235static SV *
665decode_av (dec_t *dec) 1236decode_av (dec_t *dec)
666{ 1237{
667 AV *av = newAV (); 1238 AV *av = newAV ();
668 1239
1240 DEC_INC_DEPTH;
1241 decode_ws (dec);
1242
1243 if (*dec->cur == ']')
1244 ++dec->cur;
1245 else
669 for (;;) 1246 for (;;)
670 { 1247 {
671 SV *value; 1248 SV *value;
672 1249
673 value = decode_sv (dec); 1250 value = decode_sv (dec);
674 if (!value) 1251 if (!value)
675 goto fail; 1252 goto fail;
676 1253
677 av_push (av, value); 1254 av_push (av, value);
678 1255
679 WS; 1256 decode_ws (dec);
680 1257
681 if (*dec->cur == ']') 1258 if (*dec->cur == ']')
682 { 1259 {
683 ++dec->cur; 1260 ++dec->cur;
684 break; 1261 break;
1262 }
685 } 1263
686
687 if (*dec->cur != ',') 1264 if (*dec->cur != ',')
688 ERR (", or ] expected while parsing array"); 1265 ERR (", or ] expected while parsing array");
689 1266
690 ++dec->cur; 1267 ++dec->cur;
1268
1269 decode_ws (dec);
1270
1271 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1272 {
1273 ++dec->cur;
1274 break;
1275 }
691 } 1276 }
692 1277
1278 DEC_DEC_DEPTH;
693 return newRV_noinc ((SV *)av); 1279 return newRV_noinc ((SV *)av);
694 1280
695fail: 1281fail:
696 SvREFCNT_dec (av); 1282 SvREFCNT_dec (av);
1283 DEC_DEC_DEPTH;
697 return 0; 1284 return 0;
698} 1285}
699 1286
700static SV * 1287static SV *
701decode_hv (dec_t *dec) 1288decode_hv (dec_t *dec)
702{ 1289{
1290 SV *sv;
703 HV *hv = newHV (); 1291 HV *hv = newHV ();
704 1292
1293 DEC_INC_DEPTH;
1294 decode_ws (dec);
1295
1296 if (*dec->cur == '}')
1297 ++dec->cur;
1298 else
705 for (;;) 1299 for (;;)
706 { 1300 {
707 SV *key, *value;
708
709 WS; EXPECT_CH ('"'); 1301 EXPECT_CH ('"');
710 1302
711 key = decode_str (dec); 1303 // heuristic: assume that
712 if (!key) 1304 // a) decode_str + hv_store_ent are abysmally slow.
713 goto fail; 1305 // b) most hash keys are short, simple ascii text.
714 1306 // => try to "fast-match" such strings to avoid
715 WS; EXPECT_CH (':'); 1307 // the overhead of decode_str + hv_store_ent.
716
717 value = decode_sv (dec);
718 if (!value)
719 { 1308 {
1309 SV *value;
1310 char *p = dec->cur;
1311 char *e = p + 24; // only try up to 24 bytes
1312
1313 for (;;)
1314 {
1315 // the >= 0x80 is false on most architectures
1316 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1317 {
1318 // slow path, back up and use decode_str
1319 SV *key = decode_str (dec);
1320 if (!key)
1321 goto fail;
1322
1323 decode_ws (dec); EXPECT_CH (':');
1324
1325 decode_ws (dec);
1326 value = decode_sv (dec);
1327 if (!value)
1328 {
1329 SvREFCNT_dec (key);
1330 goto fail;
1331 }
1332
1333 hv_store_ent (hv, key, value, 0);
720 SvREFCNT_dec (key); 1334 SvREFCNT_dec (key);
1335
1336 break;
1337 }
1338 else if (*p == '"')
1339 {
1340 // fast path, got a simple key
1341 char *key = dec->cur;
1342 int len = p - key;
1343 dec->cur = p + 1;
1344
1345 decode_ws (dec); EXPECT_CH (':');
1346
1347 decode_ws (dec);
1348 value = decode_sv (dec);
1349 if (!value)
721 goto fail; 1350 goto fail;
1351
1352 hv_store (hv, key, len, value, 0);
1353
1354 break;
1355 }
1356
1357 ++p;
1358 }
722 } 1359 }
723 1360
724 //TODO: optimise 1361 decode_ws (dec);
725 hv_store_ent (hv, key, value, 0);
726 1362
727 WS;
728
729 if (*dec->cur == '}') 1363 if (*dec->cur == '}')
1364 {
1365 ++dec->cur;
1366 break;
1367 }
1368
1369 if (*dec->cur != ',')
1370 ERR (", or } expected while parsing object/hash");
1371
1372 ++dec->cur;
1373
1374 decode_ws (dec);
1375
1376 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1377 {
1378 ++dec->cur;
1379 break;
1380 }
1381 }
1382
1383 DEC_DEC_DEPTH;
1384 sv = newRV_noinc ((SV *)hv);
1385
1386 // check filter callbacks
1387 if (dec->json.flags & F_HOOK)
1388 {
1389 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
730 { 1390 {
731 ++dec->cur; 1391 HE *cb, *he;
732 break; 1392
1393 hv_iterinit (hv);
1394 he = hv_iternext (hv);
1395 hv_iterinit (hv);
1396
1397 // the next line creates a mortal sv each time its called.
1398 // might want to optimise this for common cases.
1399 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1400
1401 if (cb)
1402 {
1403 dSP;
1404 int count;
1405
1406 ENTER; SAVETMPS; PUSHMARK (SP);
1407 XPUSHs (HeVAL (he));
1408 sv_2mortal (sv);
1409
1410 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1411
1412 if (count == 1)
1413 {
1414 sv = newSVsv (POPs);
1415 FREETMPS; LEAVE;
1416 return sv;
1417 }
1418
1419 SvREFCNT_inc (sv);
1420 FREETMPS; LEAVE;
1421 }
733 } 1422 }
734 1423
735 if (*dec->cur != ',') 1424 if (dec->json.cb_object)
736 ERR (", or } expected while parsing object/hash"); 1425 {
1426 dSP;
1427 int count;
737 1428
738 ++dec->cur; 1429 ENTER; SAVETMPS; PUSHMARK (SP);
739 } 1430 XPUSHs (sv_2mortal (sv));
740 1431
741 return newRV_noinc ((SV *)hv); 1432 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1433
1434 if (count == 1)
1435 {
1436 sv = newSVsv (POPs);
1437 FREETMPS; LEAVE;
1438 return sv;
1439 }
1440
1441 SvREFCNT_inc (sv);
1442 FREETMPS; LEAVE;
1443 }
1444 }
1445
1446 return sv;
742 1447
743fail: 1448fail:
744 SvREFCNT_dec (hv); 1449 SvREFCNT_dec (hv);
1450 DEC_DEC_DEPTH;
745 return 0; 1451 return 0;
746} 1452}
747 1453
748static SV * 1454static SV *
749decode_sv (dec_t *dec) 1455decode_sv (dec_t *dec)
750{ 1456{
751 WS; 1457 // the beauty of JSON: you need exactly one character lookahead
1458 // to parse everything.
752 switch (*dec->cur) 1459 switch (*dec->cur)
753 { 1460 {
754 case '"': ++dec->cur; return decode_str (dec); 1461 case '"': ++dec->cur; return decode_str (dec);
755 case '[': ++dec->cur; return decode_av (dec); 1462 case '[': ++dec->cur; return decode_av (dec);
756 case '{': ++dec->cur; return decode_hv (dec); 1463 case '{': ++dec->cur; return decode_hv (dec);
757 1464
758 case '-': 1465 case '-':
759 case '0': case '1': case '2': case '3': case '4': 1466 case '0': case '1': case '2': case '3': case '4':
760 case '5': case '6': case '7': case '8': case '9': 1467 case '5': case '6': case '7': case '8': case '9':
761 return decode_num (dec); 1468 return decode_num (dec);
762 1469
763 case 't': 1470 case 't':
764 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1471 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765 { 1472 {
766 dec->cur += 4; 1473 dec->cur += 4;
1474#if JSON_SLOW
1475 json_true = get_bool ("JSON::XS::true");
1476#endif
767 return newSViv (1); 1477 return newSVsv (json_true);
768 } 1478 }
769 else 1479 else
770 ERR ("'true' expected"); 1480 ERR ("'true' expected");
771 1481
772 break; 1482 break;
773 1483
774 case 'f': 1484 case 'f':
775 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1485 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776 { 1486 {
777 dec->cur += 5; 1487 dec->cur += 5;
1488#if JSON_SLOW
1489 json_false = get_bool ("JSON::XS::false");
1490#endif
778 return newSViv (0); 1491 return newSVsv (json_false);
779 } 1492 }
780 else 1493 else
781 ERR ("'false' expected"); 1494 ERR ("'false' expected");
782 1495
783 break; 1496 break;
784 1497
785 case 'n': 1498 case 'n':
786 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1499 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787 { 1500 {
788 dec->cur += 4; 1501 dec->cur += 4;
789 return newSViv (1); 1502 return newSVsv (&PL_sv_undef);
790 } 1503 }
791 else 1504 else
792 ERR ("'null' expected"); 1505 ERR ("'null' expected");
793 1506
794 break; 1507 break;
795 1508
796 default: 1509 default:
797 ERR ("malformed json string"); 1510 ERR ("malformed JSON string, neither array, object, number, string or atom");
798 break; 1511 break;
799 } 1512 }
800 1513
801fail: 1514fail:
802 return 0; 1515 return 0;
803} 1516}
804 1517
805static SV * 1518static SV *
806decode_json (SV *string, UV flags) 1519decode_json (SV *string, JSON *json, char **offset_return)
807{ 1520{
1521 dec_t dec;
808 SV *sv; 1522 SV *sv;
809 1523
810 if (!(flags & F_UTF8)) 1524 /* work around bugs in 5.10 where manipulating magic values
1525 * will perl ignore the magic in subsequent accesses.
1526 * also make a copy of non-PV values, to get them into a clean
1527 * state (SvPV should do that, but it's buggy, see below).
1528 */
1529 /*SvGETMAGIC (string);*/
1530 if (SvMAGICAL (string) || !SvPOK (string))
1531 string = sv_2mortal (newSVsv (string));
1532
1533 SvUPGRADE (string, SVt_PV);
1534
1535 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1536 * assertion with -DDEBUGGING, although SvCUR is documented to
1537 * return the xpv_cur field which certainly exists after upgrading.
1538 * according to nicholas clark, calling SvPOK fixes this.
1539 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1540 * and hope for the best.
1541 * Damnit, SvPV_nolen still trips over yet another assertion. This
1542 * assertion business is seriously broken, try yet another workaround
1543 * for the broken -DDEBUGGING.
1544 */
1545 {
1546#ifdef DEBUGGING
1547 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1548#else
1549 STRLEN offset = SvCUR (string);
1550#endif
1551
1552 if (offset > json->max_size && json->max_size)
1553 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1554 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1555 }
1556
1557 if (DECODE_WANTS_OCTETS (json))
1558 sv_utf8_downgrade (string, 0);
1559 else
811 sv_utf8_upgrade (string); 1560 sv_utf8_upgrade (string);
812 1561
813 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1562 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814 1563
815 dec_t dec; 1564 dec.json = *json;
816 dec.flags = flags;
817 dec.cur = SvPVX (string); 1565 dec.cur = SvPVX (string);
818 dec.end = SvEND (string); 1566 dec.end = SvEND (string);
819 dec.err = 0; 1567 dec.err = 0;
1568 dec.depth = 0;
820 1569
821 *dec.end = 1; // invalid anywhere 1570 if (dec.json.cb_object || dec.json.cb_sk_object)
1571 dec.json.flags |= F_HOOK;
1572
1573 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1574
1575 decode_ws (&dec);
822 sv = decode_sv (&dec); 1576 sv = decode_sv (&dec);
823 *dec.end = 0; 1577
1578 if (offset_return)
1579 *offset_return = dec.cur;
1580
1581 if (!(offset_return || !sv))
1582 {
1583 // check for trailing garbage
1584 decode_ws (&dec);
1585
1586 if (*dec.cur)
1587 {
1588 dec.err = "garbage after JSON object";
1589 SvREFCNT_dec (sv);
1590 sv = 0;
1591 }
1592 }
824 1593
825 if (!sv) 1594 if (!sv)
826 { 1595 {
827 IV offset = utf8_distance (dec.cur, SvPVX (string));
828 SV *uni = sv_newmortal (); 1596 SV *uni = sv_newmortal ();
829 1597
1598 // horrible hack to silence warning inside pv_uni_display
1599 COP cop = *PL_curcop;
1600 cop.cop_warnings = pWARN_NONE;
1601 ENTER;
1602 SAVEVPTR (PL_curcop);
1603 PL_curcop = &cop;
830 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1604 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1605 LEAVE;
1606
831 croak ("%s, at character %d (%s)", 1607 croak ("%s, at character offset %d (before \"%s\")",
832 dec.err, 1608 dec.err,
833 (int)offset, 1609 ptr_to_index (string, dec.cur),
834 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1610 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835 } 1611 }
836 1612
837 sv = sv_2mortal (sv); 1613 sv = sv_2mortal (sv);
838 1614
839 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1615 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
840 croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); 1616 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841 1617
842 return sv; 1618 return sv;
843} 1619}
844 1620
1621/////////////////////////////////////////////////////////////////////////////
1622// incremental parser
1623
1624static void
1625incr_parse (JSON *self)
1626{
1627 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1628
1629 // the state machine here is a bit convoluted and could be simplified a lot
1630 // but this would make it slower, so...
1631
1632 for (;;)
1633 {
1634 //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
1635 switch (self->incr_mode)
1636 {
1637 // only used for initial whitespace skipping
1638 case INCR_M_WS:
1639 for (;;)
1640 {
1641 if (*p > 0x20)
1642 {
1643 if (*p == '#')
1644 {
1645 self->incr_mode = INCR_M_C0;
1646 goto incr_m_c;
1647 }
1648 else
1649 {
1650 self->incr_mode = INCR_M_JSON;
1651 goto incr_m_json;
1652 }
1653 }
1654 else if (!*p)
1655 goto interrupt;
1656
1657 ++p;
1658 }
1659
1660 // skip a single char inside a string (for \\-processing)
1661 case INCR_M_BS:
1662 if (!*p)
1663 goto interrupt;
1664
1665 ++p;
1666 self->incr_mode = INCR_M_STR;
1667 goto incr_m_str;
1668
1669 // inside #-style comments
1670 case INCR_M_C0:
1671 case INCR_M_C1:
1672 incr_m_c:
1673 for (;;)
1674 {
1675 if (*p == '\n')
1676 {
1677 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1678 break;
1679 }
1680 else if (!*p)
1681 goto interrupt;
1682
1683 ++p;
1684 }
1685
1686 break;
1687
1688 // inside a string
1689 case INCR_M_STR:
1690 incr_m_str:
1691 for (;;)
1692 {
1693 if (*p == '"')
1694 {
1695 ++p;
1696 self->incr_mode = INCR_M_JSON;
1697
1698 if (!self->incr_nest)
1699 goto interrupt;
1700
1701 goto incr_m_json;
1702 }
1703 else if (*p == '\\')
1704 {
1705 ++p; // "virtually" consumes character after \
1706
1707 if (!*p) // if at end of string we have to switch modes
1708 {
1709 self->incr_mode = INCR_M_BS;
1710 goto interrupt;
1711 }
1712 }
1713 else if (!*p)
1714 goto interrupt;
1715
1716 ++p;
1717 }
1718
1719 // after initial ws, outside string
1720 case INCR_M_JSON:
1721 incr_m_json:
1722 for (;;)
1723 {
1724 switch (*p++)
1725 {
1726 case 0:
1727 --p;
1728 goto interrupt;
1729
1730 case 0x09:
1731 case 0x0a:
1732 case 0x0d:
1733 case 0x20:
1734 if (!self->incr_nest)
1735 {
1736 --p; // do not eat the whitespace, let the next round do it
1737 goto interrupt;
1738 }
1739 break;
1740
1741 case '"':
1742 self->incr_mode = INCR_M_STR;
1743 goto incr_m_str;
1744
1745 case '[':
1746 case '{':
1747 if (++self->incr_nest > self->max_depth)
1748 croak (ERR_NESTING_EXCEEDED);
1749 break;
1750
1751 case ']':
1752 case '}':
1753 if (--self->incr_nest <= 0)
1754 goto interrupt;
1755 break;
1756
1757 case '#':
1758 self->incr_mode = INCR_M_C1;
1759 goto incr_m_c;
1760 }
1761 }
1762 }
1763
1764 modechange:
1765 ;
1766 }
1767
1768interrupt:
1769 self->incr_pos = p - SvPVX (self->incr_text);
1770 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1771 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1772}
1773
1774/////////////////////////////////////////////////////////////////////////////
1775// XS interface functions
1776
845MODULE = JSON::XS PACKAGE = JSON::XS 1777MODULE = JSON::XS PACKAGE = JSON::XS
846 1778
847BOOT: 1779BOOT:
848{ 1780{
849 int i; 1781 int i;
850 1782
851 memset (decode_hexdigit, 0xff, 256);
852 for (i = 10; i--; ) 1783 for (i = 0; i < 256; ++i)
853 decode_hexdigit ['0' + i] = i; 1784 decode_hexdigit [i] =
1785 i >= '0' && i <= '9' ? i - '0'
1786 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1787 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1788 : -1;
854 1789
855 for (i = 7; i--; )
856 {
857 decode_hexdigit ['a' + i] = 10 + i;
858 decode_hexdigit ['A' + i] = 10 + i;
859 }
860
861 json_stash = gv_stashpv ("JSON::XS", 1); 1790 json_stash = gv_stashpv ("JSON::XS" , 1);
1791 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1792
1793 json_true = get_bool ("JSON::XS::true");
1794 json_false = get_bool ("JSON::XS::false");
1795
1796 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
862} 1797}
863 1798
864PROTOTYPES: DISABLE 1799PROTOTYPES: DISABLE
865 1800
866SV *new (char *dummy) 1801void CLONE (...)
867 CODE: 1802 CODE:
868 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1803 json_stash = 0;
1804 json_boolean_stash = 0;
1805
1806void new (char *klass)
1807 PPCODE:
1808{
1809 SV *pv = NEWSV (0, sizeof (JSON));
1810 SvPOK_only (pv);
1811 json_init ((JSON *)SvPVX (pv));
1812 XPUSHs (sv_2mortal (sv_bless (
1813 newRV_noinc (pv),
1814 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1815 )));
1816}
1817
1818void ascii (JSON *self, int enable = 1)
1819 ALIAS:
1820 ascii = F_ASCII
1821 latin1 = F_LATIN1
1822 utf8 = F_UTF8
1823 indent = F_INDENT
1824 canonical = F_CANONICAL
1825 space_before = F_SPACE_BEFORE
1826 space_after = F_SPACE_AFTER
1827 pretty = F_PRETTY
1828 allow_nonref = F_ALLOW_NONREF
1829 shrink = F_SHRINK
1830 allow_blessed = F_ALLOW_BLESSED
1831 convert_blessed = F_CONV_BLESSED
1832 relaxed = F_RELAXED
1833 allow_unknown = F_ALLOW_UNKNOWN
1834 PPCODE:
1835{
1836 if (enable)
1837 self->flags |= ix;
1838 else
1839 self->flags &= ~ix;
1840
1841 XPUSHs (ST (0));
1842}
1843
1844void get_ascii (JSON *self)
1845 ALIAS:
1846 get_ascii = F_ASCII
1847 get_latin1 = F_LATIN1
1848 get_utf8 = F_UTF8
1849 get_indent = F_INDENT
1850 get_canonical = F_CANONICAL
1851 get_space_before = F_SPACE_BEFORE
1852 get_space_after = F_SPACE_AFTER
1853 get_allow_nonref = F_ALLOW_NONREF
1854 get_shrink = F_SHRINK
1855 get_allow_blessed = F_ALLOW_BLESSED
1856 get_convert_blessed = F_CONV_BLESSED
1857 get_relaxed = F_RELAXED
1858 get_allow_unknown = F_ALLOW_UNKNOWN
1859 PPCODE:
1860 XPUSHs (boolSV (self->flags & ix));
1861
1862void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1863 PPCODE:
1864 self->max_depth = max_depth;
1865 XPUSHs (ST (0));
1866
1867U32 get_max_depth (JSON *self)
1868 CODE:
1869 RETVAL = self->max_depth;
869 OUTPUT: 1870 OUTPUT:
870 RETVAL 1871 RETVAL
871 1872
872SV *ascii (SV *self, int enable) 1873void max_size (JSON *self, U32 max_size = 0)
873 ALIAS: 1874 PPCODE:
874 ascii = F_ASCII 1875 self->max_size = max_size;
875 utf8 = F_UTF8 1876 XPUSHs (ST (0));
876 indent = F_INDENT 1877
877 canonical = F_CANONICAL 1878int get_max_size (JSON *self)
878 space_before = F_SPACE_BEFORE
879 space_after = F_SPACE_AFTER
880 json_rpc = F_JSON_RPC
881 pretty = F_PRETTY
882 allow_nonref = F_ALLOW_NONREF
883 CODE: 1879 CODE:
884{ 1880 RETVAL = self->max_size;
885 UV *uv = SvJSON (self);
886 if (enable)
887 *uv |= ix;
888 else
889 *uv &= ~ix;
890
891 RETVAL = newSVsv (self);
892}
893 OUTPUT: 1881 OUTPUT:
894 RETVAL 1882 RETVAL
895 1883
896void encode (SV *self, SV *scalar) 1884void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
897 PPCODE: 1885 PPCODE:
898 XPUSHs (encode_json (scalar, *SvJSON (self))); 1886{
1887 SvREFCNT_dec (self->cb_object);
1888 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899 1889
900void decode (SV *self, SV *jsonstr) 1890 XPUSHs (ST (0));
1891}
1892
1893void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
901 PPCODE: 1894 PPCODE:
1895{
1896 if (!self->cb_sk_object)
1897 self->cb_sk_object = newHV ();
1898
1899 if (SvOK (cb))
1900 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1901 else
1902 {
1903 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1904
1905 if (!HvKEYS (self->cb_sk_object))
1906 {
1907 SvREFCNT_dec (self->cb_sk_object);
1908 self->cb_sk_object = 0;
1909 }
1910 }
1911
1912 XPUSHs (ST (0));
1913}
1914
1915void encode (JSON *self, SV *scalar)
1916 PPCODE:
1917 XPUSHs (encode_json (scalar, self));
1918
1919void decode (JSON *self, SV *jsonstr)
1920 PPCODE:
902 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1921 XPUSHs (decode_json (jsonstr, self, 0));
1922
1923void decode_prefix (JSON *self, SV *jsonstr)
1924 PPCODE:
1925{
1926 char *offset;
1927 EXTEND (SP, 2);
1928 PUSHs (decode_json (jsonstr, self, &offset));
1929 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
1930}
1931
1932void incr_parse (JSON *self, SV *jsonstr = 0)
1933 PPCODE:
1934{
1935 if (!self->incr_text)
1936 self->incr_text = newSVpvn ("", 0);
1937
1938 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
1939 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
1940 if (DECODE_WANTS_OCTETS (self))
1941 {
1942 if (self->incr_pos)
1943 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
1944 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
1945
1946 sv_utf8_downgrade (self->incr_text, 0);
1947 }
1948 else
1949 {
1950 sv_utf8_upgrade (self->incr_text);
1951
1952 if (self->incr_pos)
1953 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1954 - (U8 *)SvPVX (self->incr_text);
1955 }
1956
1957 // append data, if any
1958 if (jsonstr)
1959 {
1960 /* make sure both strings have same encoding */
1961 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
1962 if (SvUTF8 (jsonstr))
1963 sv_utf8_downgrade (jsonstr, 0);
1964 else
1965 sv_utf8_upgrade (jsonstr);
1966
1967 /* and then just blindly append */
1968 {
1969 STRLEN len;
1970 const char *str = SvPV (jsonstr, len);
1971 STRLEN cur = SvCUR (self->incr_text);
1972
1973 if (SvLEN (self->incr_text) <= cur + len)
1974 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1975
1976 Move (str, SvEND (self->incr_text), len, char);
1977 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1978 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1979 }
1980 }
1981
1982 if (GIMME_V != G_VOID)
1983 do
1984 {
1985 char *offset;
1986
1987 if (!INCR_DONE (self))
1988 {
1989 incr_parse (self);
1990
1991 if (self->incr_pos > self->max_size && self->max_size)
1992 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1993 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
1994
1995 if (!INCR_DONE (self))
1996 break;
1997 }
1998
1999 XPUSHs (decode_json (self->incr_text, self, &offset));
2000
2001 self->incr_pos -= offset - SvPVX (self->incr_text);
2002 self->incr_nest = 0;
2003 self->incr_mode = 0;
2004
2005 sv_chop (self->incr_text, offset);
2006 }
2007 while (GIMME_V == G_ARRAY);
2008}
2009
2010SV *incr_text (JSON *self)
2011 ATTRS: lvalue
2012 CODE:
2013{
2014 if (self->incr_pos)
2015 croak ("incr_text can not be called when the incremental parser already started parsing");
2016
2017 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2018}
2019 OUTPUT:
2020 RETVAL
2021
2022void incr_skip (JSON *self)
2023 CODE:
2024{
2025 if (self->incr_pos)
2026 {
2027 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2028 self->incr_pos = 0;
2029 self->incr_nest = 0;
2030 self->incr_mode = 0;
2031 }
2032}
2033
2034void incr_reset (JSON *self)
2035 CODE:
2036{
2037 SvREFCNT_dec (self->incr_text);
2038 self->incr_text = 0;
2039 self->incr_pos = 0;
2040 self->incr_nest = 0;
2041 self->incr_mode = 0;
2042}
2043
2044void DESTROY (JSON *self)
2045 CODE:
2046 SvREFCNT_dec (self->cb_sk_object);
2047 SvREFCNT_dec (self->cb_object);
2048 SvREFCNT_dec (self->incr_text);
903 2049
904PROTOTYPES: ENABLE 2050PROTOTYPES: ENABLE
905 2051
906void to_json (SV *scalar) 2052void encode_json (SV *scalar)
2053 ALIAS:
2054 to_json_ = 0
2055 encode_json = F_UTF8
907 PPCODE: 2056 PPCODE:
2057{
2058 JSON json;
2059 json_init (&json);
2060 json.flags |= ix;
908 XPUSHs (encode_json (scalar, F_UTF8)); 2061 XPUSHs (encode_json (scalar, &json));
2062}
909 2063
910void from_json (SV *jsonstr) 2064void decode_json (SV *jsonstr)
2065 ALIAS:
2066 from_json_ = 0
2067 decode_json = F_UTF8
911 PPCODE: 2068 PPCODE:
2069{
2070 JSON json;
2071 json_init (&json);
2072 json.flags |= ix;
912 XPUSHs (decode_json (jsonstr, F_UTF8)); 2073 XPUSHs (decode_json (jsonstr, &json, 0));
2074}
913 2075

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines