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 1.5 by root, Thu Mar 22 23:24:18 2007 UTC vs.
Revision 1.123 by root, Sun Dec 8 20:44:23 2013 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// compatibility with perl <5.18
23#ifndef HvNAMELEN_get
24# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
25#endif
26#ifndef HvNAMELEN
27# define HvNAMELEN(hv) HvNAMELEN_get (hv)
28#endif
29#ifndef HvNAMEUTF8
30# define HvNAMEUTF8(hv) 0
31#endif
32
33// three extra for rounding, sign, and end of string
34#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
35
9#define F_ASCII 0x00000001 36#define F_ASCII 0x00000001UL
37#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 38#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 39#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 40#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 41#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 42#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040
16#define F_ALLOW_NONREF 0x00000080 43#define F_ALLOW_NONREF 0x00000100UL
44#define F_SHRINK 0x00000200UL
45#define F_ALLOW_BLESSED 0x00000400UL
46#define F_CONV_BLESSED 0x00000800UL
47#define F_RELAXED 0x00001000UL
48#define F_ALLOW_UNKNOWN 0x00002000UL
49#define F_ALLOW_TAGS 0x00004000UL
50#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17 51
18#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 52#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19#define F_DEFAULT 0
20 53
21#define INIT_SIZE 32 // initial scalar size to be allocated 54#define INIT_SIZE 32 // initial scalar size to be allocated
55#define INDENT_STEP 3 // spaces per indentation level
56
57#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
58
59#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
22 60
23#define SB do { 61#define SB do {
24#define SE } while (0) 62#define SE } while (0)
25 63
26static HV *json_stash; 64#if __GNUC__ >= 3
65# define expect(expr,value) __builtin_expect ((expr), (value))
66# define INLINE static inline
67#else
68# define expect(expr,value) (expr)
69# define INLINE static
70#endif
71
72#define expect_false(expr) expect ((expr) != 0, 0)
73#define expect_true(expr) expect ((expr) != 0, 1)
74
75#define IN_RANGE_INC(type,val,beg,end) \
76 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
77 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
78
79#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
80
81#ifdef USE_ITHREADS
82# define JSON_SLOW 1
83# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
84#else
85# define JSON_SLOW 0
86# define JSON_STASH json_stash
87#endif
88
89// the amount of HEs to allocate on the stack, when sorting keys
90#define STACK_HES 64
91
92static HV *json_stash, *types_boolean_stash; // JSON::XS::
93static SV *types_true, *types_false, *sv_json;
94
95enum {
96 INCR_M_WS = 0, // initial whitespace skipping, must be 0
97 INCR_M_STR, // inside string
98 INCR_M_BS, // inside backslash
99 INCR_M_C0, // inside comment in initial whitespace sequence
100 INCR_M_C1, // inside comment in other places
101 INCR_M_JSON // outside anything, count nesting
102};
103
104#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
105
106typedef struct {
107 U32 flags;
108 U32 max_depth;
109 STRLEN max_size;
110
111 SV *cb_object;
112 HV *cb_sk_object;
113
114 // for the incremental parser
115 SV *incr_text; // the source text so far
116 STRLEN incr_pos; // the current offset into the text
117 int incr_nest; // {[]}-nesting level
118 unsigned char incr_mode;
119} JSON;
120
121INLINE void
122json_init (JSON *json)
123{
124 Zero (json, 1, JSON);
125 json->max_depth = 512;
126}
127
128/////////////////////////////////////////////////////////////////////////////
129// utility functions
130
131INLINE SV *
132get_bool (const char *name)
133{
134 SV *sv = get_sv (name, 1);
135
136 SvREADONLY_on (sv);
137 SvREADONLY_on (SvRV (sv));
138
139 return sv;
140}
141
142INLINE void
143shrink (SV *sv)
144{
145 sv_utf8_downgrade (sv, 1);
146
147 if (SvLEN (sv) > SvCUR (sv) + 1)
148 {
149#ifdef SvPV_shrink_to_cur
150 SvPV_shrink_to_cur (sv);
151#elif defined (SvPV_renew)
152 SvPV_renew (sv, SvCUR (sv) + 1);
153#endif
154 }
155}
156
157// decode an utf-8 character and return it, or (UV)-1 in
158// case of an error.
159// we special-case "safe" characters from U+80 .. U+7FF,
160// but use the very good perl function to parse anything else.
161// note that we never call this function for a ascii codepoints
162INLINE UV
163decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
164{
165 if (expect_true (len >= 2
166 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
167 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
168 {
169 *clen = 2;
170 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
171 }
172 else
173 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
174}
175
176// likewise for encoding, also never called for ascii codepoints
177// this function takes advantage of this fact, although current gccs
178// seem to optimise the check for >= 0x80 away anyways
179INLINE unsigned char *
180encode_utf8 (unsigned char *s, UV ch)
181{
182 if (expect_false (ch < 0x000080))
183 *s++ = ch;
184 else if (expect_true (ch < 0x000800))
185 *s++ = 0xc0 | ( ch >> 6),
186 *s++ = 0x80 | ( ch & 0x3f);
187 else if ( ch < 0x010000)
188 *s++ = 0xe0 | ( ch >> 12),
189 *s++ = 0x80 | ((ch >> 6) & 0x3f),
190 *s++ = 0x80 | ( ch & 0x3f);
191 else if ( ch < 0x110000)
192 *s++ = 0xf0 | ( ch >> 18),
193 *s++ = 0x80 | ((ch >> 12) & 0x3f),
194 *s++ = 0x80 | ((ch >> 6) & 0x3f),
195 *s++ = 0x80 | ( ch & 0x3f);
196
197 return s;
198}
199
200// convert offset pointer to character index, sv must be string
201static STRLEN
202ptr_to_index (SV *sv, char *offset)
203{
204 return SvUTF8 (sv)
205 ? utf8_distance (offset, SvPVX (sv))
206 : offset - SvPVX (sv);
207}
208
209/////////////////////////////////////////////////////////////////////////////
210// fp hell
211
212// scan a group of digits, and a trailing exponent
213static void
214json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
215{
216 UV uaccum = 0;
217 int eaccum = 0;
218
219 // if we recurse too deep, skip all remaining digits
220 // to avoid a stack overflow attack
221 if (expect_false (--maxdepth <= 0))
222 while (((U8)*s - '0') < 10)
223 ++s;
224
225 for (;;)
226 {
227 U8 dig = (U8)*s - '0';
228
229 if (expect_false (dig >= 10))
230 {
231 if (dig == (U8)((U8)'.' - (U8)'0'))
232 {
233 ++s;
234 json_atof_scan1 (s, accum, expo, 1, maxdepth);
235 }
236 else if ((dig | ' ') == 'e' - '0')
237 {
238 int exp2 = 0;
239 int neg = 0;
240
241 ++s;
242
243 if (*s == '-')
244 {
245 ++s;
246 neg = 1;
247 }
248 else if (*s == '+')
249 ++s;
250
251 while ((dig = (U8)*s - '0') < 10)
252 exp2 = exp2 * 10 + *s++ - '0';
253
254 *expo += neg ? -exp2 : exp2;
255 }
256
257 break;
258 }
259
260 ++s;
261
262 uaccum = uaccum * 10 + dig;
263 ++eaccum;
264
265 // if we have too many digits, then recurse for more
266 // we actually do this for rather few digits
267 if (uaccum >= (UV_MAX - 9) / 10)
268 {
269 if (postdp) *expo -= eaccum;
270 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
271 if (postdp) *expo += eaccum;
272
273 break;
274 }
275 }
276
277 // this relies greatly on the quality of the pow ()
278 // implementation of the platform, but a good
279 // implementation is hard to beat.
280 // (IEEE 754 conformant ones are required to be exact)
281 if (postdp) *expo -= eaccum;
282 *accum += uaccum * Perl_pow (10., *expo);
283 *expo += eaccum;
284}
285
286static NV
287json_atof (const char *s)
288{
289 NV accum = 0.;
290 int expo = 0;
291 int neg = 0;
292
293 if (*s == '-')
294 {
295 ++s;
296 neg = 1;
297 }
298
299 // a recursion depth of ten gives us >>500 bits
300 json_atof_scan1 (s, &accum, &expo, 0, 10);
301
302 return neg ? -accum : accum;
303}
304/////////////////////////////////////////////////////////////////////////////
305// encoder
27 306
28// structure used for encoding JSON 307// structure used for encoding JSON
29typedef struct 308typedef struct
30{ 309{
31 char *cur; 310 char *cur; // SvPVX (sv) + current output position
32 STRLEN len; // SvLEN (sv)
33 char *end; // SvEND (sv) 311 char *end; // SvEND (sv)
34 SV *sv; 312 SV *sv; // result scalar
35 UV flags; 313 JSON json;
36 int max_recurse; 314 U32 indent; // indentation level
37 int indent; 315 UV limit; // escape character values >= this value when encoding
38} enc_t; 316} enc_t;
39 317
40// structure used for decoding JSON 318INLINE 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) 319need (enc_t *enc, STRLEN len)
62{ 320{
63 if (enc->cur + len >= enc->end) 321 if (expect_false (enc->cur + len >= enc->end))
64 { 322 {
65 STRLEN cur = enc->cur - SvPVX (enc->sv); 323 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66 SvGROW (enc->sv, cur + len + 1); 324 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
67 enc->cur = SvPVX (enc->sv) + cur; 325 enc->cur = SvPVX (enc->sv) + cur;
68 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); 326 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69 } 327 }
70} 328}
71 329
72static void 330INLINE void
73encode_ch (enc_t *enc, char ch) 331encode_ch (enc_t *enc, char ch)
74{ 332{
75 need (enc, 1); 333 need (enc, 1);
76 *enc->cur++ = ch; 334 *enc->cur++ = ch;
77} 335}
85 343
86 while (str < end) 344 while (str < end)
87 { 345 {
88 unsigned char ch = *(unsigned char *)str; 346 unsigned char ch = *(unsigned char *)str;
89 347
90 if (ch == '"') 348 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91 { 349 {
350 if (expect_false (ch == '"')) // but with slow exceptions
351 {
92 need (enc, len += 1); 352 need (enc, len += 1);
93 *enc->cur++ = '\\'; 353 *enc->cur++ = '\\';
94 *enc->cur++ = '"'; 354 *enc->cur++ = '"';
95 ++str;
96 } 355 }
97 else if (ch == '\\') 356 else if (expect_false (ch == '\\'))
98 { 357 {
99 need (enc, len += 1); 358 need (enc, len += 1);
100 *enc->cur++ = '\\'; 359 *enc->cur++ = '\\';
101 *enc->cur++ = '\\'; 360 *enc->cur++ = '\\';
102 ++str;
103 } 361 }
104 else if (ch >= 0x20 && ch < 0x80) // most common case 362 else
105 {
106 *enc->cur++ = ch; 363 *enc->cur++ = ch;
107 ++str; 364
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; 365 ++str;
122 } 366 }
123 else 367 else
124 { 368 {
125 STRLEN clen; 369 switch (ch)
126 UV uch;
127
128 if (is_utf8)
129 { 370 {
130 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 371 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131 if (clen == (STRLEN)-1) 372 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
132 croak ("malformed UTF-8 character in string, cannot convert to JSON"); 373 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133 } 374 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134 else 375 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135 {
136 uch = ch;
137 clen = 1;
138 }
139 376
140 if (uch < 0x80 || enc->flags & F_ASCII) 377 default:
141 {
142 if (uch > 0xFFFFUL)
143 { 378 {
379 STRLEN clen;
380 UV uch;
381
382 if (is_utf8)
383 {
384 uch = decode_utf8 (str, end - str, &clen);
385 if (clen == (STRLEN)-1)
386 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
387 }
388 else
389 {
390 uch = ch;
391 clen = 1;
392 }
393
394 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
395 {
396 if (uch >= 0x10000UL)
397 {
398 if (uch >= 0x110000UL)
399 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
400
144 need (enc, len += 11); 401 need (enc, len += 11);
145 sprintf (enc->cur, "\\u%04x\\u%04x", 402 sprintf (enc->cur, "\\u%04x\\u%04x",
146 (uch - 0x10000) / 0x400 + 0xD800, 403 (int)((uch - 0x10000) / 0x400 + 0xD800),
147 (uch - 0x10000) % 0x400 + 0xDC00); 404 (int)((uch - 0x10000) % 0x400 + 0xDC00));
148 enc->cur += 12; 405 enc->cur += 12;
406 }
407 else
408 {
409 need (enc, len += 5);
410 *enc->cur++ = '\\';
411 *enc->cur++ = 'u';
412 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
413 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
414 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
415 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
416 }
417
418 str += clen;
419 }
420 else if (enc->json.flags & F_LATIN1)
421 {
422 *enc->cur++ = uch;
423 str += clen;
424 }
425 else if (is_utf8)
426 {
427 need (enc, len += clen);
428 do
429 {
430 *enc->cur++ = *str++;
431 }
432 while (--clen);
433 }
434 else
435 {
436 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
437 enc->cur = encode_utf8 (enc->cur, uch);
438 ++str;
439 }
149 } 440 }
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 do
168 {
169 *enc->cur++ = *str++;
170 }
171 while (--clen);
172 }
173 else
174 {
175 need (enc, 10); // never more than 11 bytes needed
176 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
177 ++str;
178 } 441 }
179 } 442 }
180 443
181 --len; 444 --len;
182 } 445 }
183} 446}
184 447
185#define INDENT SB \ 448INLINE void
449encode_indent (enc_t *enc)
450{
186 if (enc->flags & F_INDENT) \ 451 if (enc->json.flags & F_INDENT)
187 { \ 452 {
188 int i_; \ 453 int spaces = enc->indent * INDENT_STEP;
189 need (enc, enc->indent); \ 454
190 for (i_ = enc->indent * 3; i_--; )\ 455 need (enc, spaces);
456 memset (enc->cur, ' ', spaces);
457 enc->cur += spaces;
458 }
459}
460
461INLINE void
462encode_space (enc_t *enc)
463{
464 need (enc, 1);
465 encode_ch (enc, ' ');
466}
467
468INLINE void
469encode_nl (enc_t *enc)
470{
471 if (enc->json.flags & F_INDENT)
472 {
473 need (enc, 1);
191 encode_ch (enc, ' '); \ 474 encode_ch (enc, '\n');
192 } \ 475 }
193 SE 476}
194 477
195#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 478INLINE void
196#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 479encode_comma (enc_t *enc)
197#define COMMA SB \ 480{
198 encode_ch (enc, ','); \ 481 encode_ch (enc, ',');
482
199 if (enc->flags & F_INDENT) \ 483 if (enc->json.flags & F_INDENT)
200 NL; \ 484 encode_nl (enc);
201 else if (enc->flags & F_SPACE_AFTER) \ 485 else if (enc->json.flags & F_SPACE_AFTER)
202 SPACE; \ 486 encode_space (enc);
203 SE 487}
204 488
205static void encode_sv (enc_t *enc, SV *sv); 489static void encode_sv (enc_t *enc, SV *sv);
206 490
207static void 491static void
208encode_av (enc_t *enc, AV *av) 492encode_av (enc_t *enc, AV *av)
209{ 493{
210 int i, len = av_len (av); 494 int i, len = av_len (av);
211 495
496 if (enc->indent >= enc->json.max_depth)
497 croak (ERR_NESTING_EXCEEDED);
498
212 encode_ch (enc, '['); NL; 499 encode_ch (enc, '[');
213 ++enc->indent;
214 500
501 if (len >= 0)
502 {
503 encode_nl (enc); ++enc->indent;
504
215 for (i = 0; i <= len; ++i) 505 for (i = 0; i <= len; ++i)
216 { 506 {
217 INDENT; 507 SV **svp = av_fetch (av, i, 0);
218 encode_sv (enc, *av_fetch (av, i, 0));
219 508
509 encode_indent (enc);
510
511 if (svp)
512 encode_sv (enc, *svp);
513 else
514 encode_str (enc, "null", 4, 0);
515
220 if (i < len) 516 if (i < len)
221 COMMA; 517 encode_comma (enc);
222 } 518 }
223 519
224 NL; 520 encode_nl (enc); --enc->indent; encode_indent (enc);
521 }
225 522
226 --enc->indent;
227 INDENT; encode_ch (enc, ']'); 523 encode_ch (enc, ']');
228} 524}
229 525
230static void 526static void
231encode_he (enc_t *enc, HE *he) 527encode_hk (enc_t *enc, HE *he)
232{ 528{
233 encode_ch (enc, '"'); 529 encode_ch (enc, '"');
234 530
235 if (HeKLEN (he) == HEf_SVKEY) 531 if (HeKLEN (he) == HEf_SVKEY)
236 { 532 {
237 SV *sv = HeSVKEY (he); 533 SV *sv = HeSVKEY (he);
238 STRLEN len; 534 STRLEN len;
239 char *str; 535 char *str;
240 536
241 SvGETMAGIC (sv); 537 SvGETMAGIC (sv);
242 str = SvPV (sv, len); 538 str = SvPV (sv, len);
243 539
244 encode_str (enc, str, len, SvUTF8 (sv)); 540 encode_str (enc, str, len, SvUTF8 (sv));
245 } 541 }
246 else 542 else
247 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 543 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
248 544
249 encode_ch (enc, '"'); 545 encode_ch (enc, '"');
250 546
251 if (enc->flags & F_SPACE_BEFORE) SPACE; 547 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
252 encode_ch (enc, ':'); 548 encode_ch (enc, ':');
253 if (enc->flags & F_SPACE_AFTER ) SPACE; 549 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
254 encode_sv (enc, HeVAL (he));
255} 550}
256 551
257// compare hash entries, used when all keys are bytestrings 552// compare hash entries, used when all keys are bytestrings
258static int 553static int
259he_cmp_fast (const void *a_, const void *b_) 554he_cmp_fast (const void *a_, const void *b_)
264 HE *b = *(HE **)b_; 559 HE *b = *(HE **)b_;
265 560
266 STRLEN la = HeKLEN (a); 561 STRLEN la = HeKLEN (a);
267 STRLEN lb = HeKLEN (b); 562 STRLEN lb = HeKLEN (b);
268 563
269 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 564 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
270 cmp = la < lb ? -1 : la == lb ? 0 : 1; 565 cmp = lb - la;
271 566
272 return cmp; 567 return cmp;
273} 568}
274 569
275// compare hash entries, used when some keys are sv's or utf-x 570// compare hash entries, used when some keys are sv's or utf-x
276static int 571static int
277he_cmp_slow (const void *a, const void *b) 572he_cmp_slow (const void *a, const void *b)
278{ 573{
279 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 574 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
280} 575}
281 576
282static void 577static void
283encode_hv (enc_t *enc, HV *hv) 578encode_hv (enc_t *enc, HV *hv)
284{ 579{
285 int count, i; 580 HE *he;
286 581
287 encode_ch (enc, '{'); NL; ++enc->indent; 582 if (enc->indent >= enc->json.max_depth)
583 croak (ERR_NESTING_EXCEEDED);
288 584
289 if ((count = hv_iterinit (hv))) 585 encode_ch (enc, '{');
290 { 586
291 // for canonical output we have to sort by keys first 587 // for canonical output we have to sort by keys first
292 // actually, this is mostly due to the stupid so-called 588 // actually, this is mostly due to the stupid so-called
293 // security workaround added somewhere in 5.8.x. 589 // security workaround added somewhere in 5.8.x
294 // that randomises hash orderings 590 // that randomises hash orderings
295 if (enc->flags & F_CANONICAL) 591 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
592 {
593 int count = hv_iterinit (hv);
594
595 if (SvMAGICAL (hv))
296 { 596 {
297 HE *he, *hes [count]; 597 // need to count by iterating. could improve by dynamically building the vector below
598 // but I don't care for the speed of this special case.
599 // note also that we will run into undefined behaviour when the two iterations
600 // do not result in the same count, something I might care for in some later release.
601
602 count = 0;
603 while (hv_iternext (hv))
604 ++count;
605
606 hv_iterinit (hv);
607 }
608
609 if (count)
610 {
298 int fast = 1; 611 int i, fast = 1;
612 HE *hes_stack [STACK_HES];
613 HE **hes = hes_stack;
614
615 // allocate larger arrays on the heap
616 if (count > STACK_HES)
617 {
618 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
619 hes = (HE **)SvPVX (sv);
620 }
299 621
300 i = 0; 622 i = 0;
301 while ((he = hv_iternext (hv))) 623 while ((he = hv_iternext (hv)))
302 { 624 {
303 hes [i++] = he; 625 hes [i++] = he;
309 631
310 if (fast) 632 if (fast)
311 qsort (hes, count, sizeof (HE *), he_cmp_fast); 633 qsort (hes, count, sizeof (HE *), he_cmp_fast);
312 else 634 else
313 { 635 {
314 // hack to disable "use bytes" 636 // hack to forcefully disable "use bytes"
315 COP *oldcop = PL_curcop, cop; 637 COP cop = *PL_curcop;
316 cop.op_private = 0; 638 cop.op_private = 0;
639
640 ENTER;
641 SAVETMPS;
642
643 SAVEVPTR (PL_curcop);
317 PL_curcop = &cop; 644 PL_curcop = &cop;
318 645
319 SAVETMPS;
320 qsort (hes, count, sizeof (HE *), he_cmp_slow); 646 qsort (hes, count, sizeof (HE *), he_cmp_slow);
647
321 FREETMPS; 648 FREETMPS;
322 649 LEAVE;
323 PL_curcop = oldcop;
324 } 650 }
325 651
326 for (i = 0; i < count; ++i) 652 encode_nl (enc); ++enc->indent;
653
654 while (count--)
327 { 655 {
328 INDENT; 656 encode_indent (enc);
657 he = hes [count];
329 encode_he (enc, hes [i]); 658 encode_hk (enc, he);
659 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
330 660
331 if (i < count - 1) 661 if (count)
332 COMMA; 662 encode_comma (enc);
333 } 663 }
334 664
335 NL; 665 encode_nl (enc); --enc->indent; encode_indent (enc);
336 } 666 }
667 }
668 else
669 {
670 if (hv_iterinit (hv) || SvMAGICAL (hv))
671 if ((he = hv_iternext (hv)))
672 {
673 encode_nl (enc); ++enc->indent;
674
675 for (;;)
676 {
677 encode_indent (enc);
678 encode_hk (enc, he);
679 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
680
681 if (!(he = hv_iternext (hv)))
682 break;
683
684 encode_comma (enc);
685 }
686
687 encode_nl (enc); --enc->indent; encode_indent (enc);
688 }
689 }
690
691 encode_ch (enc, '}');
692}
693
694// encode objects, arrays and special \0=false and \1=true values.
695static void
696encode_rv (enc_t *enc, SV *sv)
697{
698 svtype svt;
699 GV *method;
700
701 SvGETMAGIC (sv);
702 svt = SvTYPE (sv);
703
704 if (expect_false (SvOBJECT (sv)))
705 {
706 HV *boolean_stash = !JSON_SLOW || types_boolean_stash
707 ? types_boolean_stash
708 : gv_stashpv ("Types::Serialiser::Boolean", 1);
709 HV *stash = SvSTASH (sv);
710
711 if (stash == boolean_stash)
712 {
713 if (SvIV (sv))
714 encode_str (enc, "true", 4, 0);
715 else
716 encode_str (enc, "false", 5, 0);
717 }
718 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
719 {
720 int count;
721 dSP;
722
723 ENTER; SAVETMPS; PUSHMARK (SP);
724 EXTEND (SP, 2);
725 // we re-bless the reference to get overload and other niceties right
726 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
727 PUSHs (sv_json);
728
729 PUTBACK;
730 count = call_sv ((SV *)GvCV (method), G_ARRAY);
731 SPAGAIN;
732
733 // catch this surprisingly common error
734 if (SvROK (TOPs) && SvRV (TOPs) == sv)
735 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
736
737 encode_ch (enc, '(');
738 encode_ch (enc, '"');
739 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
740 encode_ch (enc, '"');
741 encode_ch (enc, ')');
742 encode_ch (enc, '[');
743
744 while (count)
745 {
746 encode_sv (enc, SP[1 - count--]);
747
748 if (count)
749 encode_ch (enc, ',');
750 }
751
752 encode_ch (enc, ']');
753
754 PUTBACK;
755
756 FREETMPS; LEAVE;
757 }
758 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
759 {
760 dSP;
761
762 ENTER; SAVETMPS; PUSHMARK (SP);
763 // we re-bless the reference to get overload and other niceties right
764 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
765
766 // calling with G_SCALAR ensures that we always get a 1 return value
767 PUTBACK;
768 call_sv ((SV *)GvCV (method), G_SCALAR);
769 SPAGAIN;
770
771 // catch this surprisingly common error
772 if (SvROK (TOPs) && SvRV (TOPs) == sv)
773 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
774
775 sv = POPs;
776 PUTBACK;
777
778 encode_sv (enc, sv);
779
780 FREETMPS; LEAVE;
781 }
782 else if (enc->json.flags & F_ALLOW_BLESSED)
783 encode_str (enc, "null", 4, 0);
337 else 784 else
338 { 785 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
339 SV *sv; 786 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
340 HE *he = hv_iternext (hv); 787 }
341 788 else if (svt == SVt_PVHV)
342 for (;;) 789 encode_hv (enc, (HV *)sv);
343 { 790 else if (svt == SVt_PVAV)
344 INDENT; 791 encode_av (enc, (AV *)sv);
345 encode_he (enc, he); 792 else if (svt < SVt_PVAV)
346
347 if (!(he = hv_iternext (hv)))
348 break;
349
350 COMMA;
351 }
352
353 NL;
354 }
355 } 793 {
794 STRLEN len = 0;
795 char *pv = svt ? SvPV (sv, len) : 0;
356 796
357 --enc->indent; INDENT; encode_ch (enc, '}'); 797 if (len == 1 && *pv == '1')
798 encode_str (enc, "true", 4, 0);
799 else if (len == 1 && *pv == '0')
800 encode_str (enc, "false", 5, 0);
801 else if (enc->json.flags & F_ALLOW_UNKNOWN)
802 encode_str (enc, "null", 4, 0);
803 else
804 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
805 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
806 }
807 else if (enc->json.flags & F_ALLOW_UNKNOWN)
808 encode_str (enc, "null", 4, 0);
809 else
810 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
811 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
358} 812}
359 813
360static void 814static void
361encode_sv (enc_t *enc, SV *sv) 815encode_sv (enc_t *enc, SV *sv)
362{ 816{
370 encode_str (enc, str, len, SvUTF8 (sv)); 824 encode_str (enc, str, len, SvUTF8 (sv));
371 encode_ch (enc, '"'); 825 encode_ch (enc, '"');
372 } 826 }
373 else if (SvNOKp (sv)) 827 else if (SvNOKp (sv))
374 { 828 {
829 // trust that perl will do the right thing w.r.t. JSON syntax.
375 need (enc, NV_DIG + 32); 830 need (enc, NV_DIG + 32);
376 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 831 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
377 enc->cur += strlen (enc->cur); 832 enc->cur += strlen (enc->cur);
378 } 833 }
379 else if (SvIOKp (sv)) 834 else if (SvIOKp (sv))
380 { 835 {
836 // we assume we can always read an IV as a UV and vice versa
837 // we assume two's complement
838 // we assume no aliasing issues in the union
839 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
840 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
841 {
842 // optimise the "small number case"
843 // code will likely be branchless and use only a single multiplication
844 // works for numbers up to 59074
845 I32 i = SvIVX (sv);
846 U32 u;
847 char digit, nz = 0;
848
381 need (enc, 64); 849 need (enc, 6);
850
851 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
852 u = i < 0 ? -i : i;
853
854 // convert to 4.28 fixed-point representation
855 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
856
857 // now output digit by digit, each time masking out the integer part
858 // and multiplying by 5 while moving the decimal point one to the right,
859 // resulting in a net multiplication by 10.
860 // we always write the digit to memory but conditionally increment
861 // the pointer, to enable the use of conditional move instructions.
862 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
863 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
864 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
865 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
866 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
867 }
868 else
869 {
870 // large integer, use the (rather slow) snprintf way.
871 need (enc, IVUV_MAXCHARS);
382 enc->cur += 872 enc->cur +=
383 SvIsUV(sv) 873 SvIsUV(sv)
384 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 874 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
385 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 875 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
876 }
386 } 877 }
387 else if (SvROK (sv)) 878 else if (SvROK (sv))
388 { 879 encode_rv (enc, SvRV (sv));
389 if (!--enc->max_recurse) 880 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
390 croak ("data structure too deep (hit recursion limit)");
391
392 sv = SvRV (sv);
393
394 switch (SvTYPE (sv))
395 {
396 case SVt_PVAV: encode_av (enc, (AV *)sv); break;
397 case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
398
399 default:
400 croak ("JSON can only represent references to arrays or hashes");
401 }
402 }
403 else if (!SvOK (sv))
404 encode_str (enc, "null", 4, 0); 881 encode_str (enc, "null", 4, 0);
405 else 882 else
406 croak ("encountered perl type that JSON cannot handle"); 883 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
884 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
407} 885}
408 886
409static SV * 887static SV *
410encode_json (SV *scalar, UV flags) 888encode_json (SV *scalar, JSON *json)
411{ 889{
412 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
413 croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
414
415 enc_t enc; 890 enc_t enc;
416 enc.flags = flags; 891
892 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
893 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
894
895 enc.json = *json;
417 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 896 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
418 enc.cur = SvPVX (enc.sv); 897 enc.cur = SvPVX (enc.sv);
419 enc.end = SvEND (enc.sv); 898 enc.end = SvEND (enc.sv);
420 enc.max_recurse = 0;
421 enc.indent = 0; 899 enc.indent = 0;
900 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
901 : enc.json.flags & F_LATIN1 ? 0x000100UL
902 : 0x110000UL;
422 903
423 SvPOK_only (enc.sv); 904 SvPOK_only (enc.sv);
424 encode_sv (&enc, scalar); 905 encode_sv (&enc, scalar);
906 encode_nl (&enc);
425 907
908 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
909 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
910
426 if (!(flags & (F_ASCII | F_UTF8))) 911 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
427 SvUTF8_on (enc.sv); 912 SvUTF8_on (enc.sv);
428 913
429 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 914 if (enc.json.flags & F_SHRINK)
915 shrink (enc.sv);
916
430 return enc.sv; 917 return enc.sv;
431} 918}
432 919
433///////////////////////////////////////////////////////////////////////////// 920/////////////////////////////////////////////////////////////////////////////
921// decoder
434 922
435#define WS \ 923// structure used for decoding JSON
924typedef struct
925{
926 char *cur; // current parser pointer
927 char *end; // end of input string
928 const char *err; // parse error, if != 0
929 JSON json;
930 U32 depth; // recursion depth
931 U32 maxdepth; // recursion depth limit
932} dec_t;
933
934INLINE void
935decode_comment (dec_t *dec)
936{
937 // only '#'-style comments allowed a.t.m.
938
939 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
940 ++dec->cur;
941}
942
943INLINE void
944decode_ws (dec_t *dec)
945{
436 for (;;) \ 946 for (;;)
437 { \ 947 {
438 char ch = *dec->cur; \ 948 char ch = *dec->cur;
949
439 if (ch > 0x20 \ 950 if (ch > 0x20)
951 {
952 if (expect_false (ch == '#'))
953 {
954 if (dec->json.flags & F_RELAXED)
955 decode_comment (dec);
956 else
957 break;
958 }
959 else
960 break;
961 }
440 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 962 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
441 break; \ 963 break; // parse error, but let higher level handle it, gives better error messages
964
442 ++dec->cur; \ 965 ++dec->cur;
443 } 966 }
967}
444 968
445#define ERR(reason) SB dec->err = reason; goto fail; SE 969#define ERR(reason) SB dec->err = reason; goto fail; SE
970
446#define EXPECT_CH(ch) SB \ 971#define EXPECT_CH(ch) SB \
447 if (*dec->cur != ch) \ 972 if (*dec->cur != ch) \
448 ERR (# ch " expected"); \ 973 ERR (# ch " expected"); \
449 ++dec->cur; \ 974 ++dec->cur; \
450 SE 975 SE
451 976
977#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
978#define DEC_DEC_DEPTH --dec->depth
979
452static SV *decode_sv (dec_t *dec); 980static SV *decode_sv (dec_t *dec);
453 981
454static signed char decode_hexdigit[256]; 982static signed char decode_hexdigit[256];
455 983
456static UV 984static UV
457decode_4hex (dec_t *dec) 985decode_4hex (dec_t *dec)
458{ 986{
459 signed char d1, d2, d3, d4; 987 signed char d1, d2, d3, d4;
988 unsigned char *cur = (unsigned char *)dec->cur;
460 989
461 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 990 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
462 if (d1 < 0) ERR ("four hexadecimal digits expected"); 991 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
463 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 992 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
464 if (d2 < 0) ERR ("four hexadecimal digits expected"); 993 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
465 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
466 if (d3 < 0) ERR ("four hexadecimal digits expected");
467 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
468 if (d4 < 0) ERR ("four hexadecimal digits expected");
469 994
470 dec->cur += 4; 995 dec->cur += 4;
471 996
472 return ((UV)d1) << 12 997 return ((UV)d1) << 12
473 | ((UV)d2) << 8 998 | ((UV)d2) << 8
476 1001
477fail: 1002fail:
478 return (UV)-1; 1003 return (UV)-1;
479} 1004}
480 1005
481#define APPEND_GROW(n) SB \
482 if (cur + (n) >= end) \
483 { \
484 STRLEN ofs = cur - SvPVX (sv); \
485 SvGROW (sv, ofs + (n) + 1); \
486 cur = SvPVX (sv) + ofs; \
487 end = SvEND (sv); \
488 } \
489 SE
490
491#define APPEND_CH(ch) SB \
492 APPEND_GROW (1); \
493 *cur++ = (ch); \
494 SE
495
496static SV * 1006static SV *
497decode_str (dec_t *dec) 1007decode_str (dec_t *dec)
498{ 1008{
499 SV *sv = NEWSV (0,2); 1009 SV *sv = 0;
500 int utf8 = 0; 1010 int utf8 = 0;
501 char *cur = SvPVX (sv); 1011 char *dec_cur = dec->cur;
502 char *end = SvEND (sv);
503 1012
504 for (;;) 1013 do
505 { 1014 {
506 unsigned char ch = *(unsigned char *)dec->cur; 1015 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
1016 char *cur = buf;
507 1017
508 if (ch == '"') 1018 do
509 break;
510 else if (ch == '\\')
511 { 1019 {
512 switch (*++dec->cur) 1020 unsigned char ch = *(unsigned char *)dec_cur++;
1021
1022 if (expect_false (ch == '"'))
513 { 1023 {
514 case '\\': 1024 --dec_cur;
515 case '/': 1025 break;
516 case '"': APPEND_CH (*dec->cur++); break; 1026 }
517 1027 else if (expect_false (ch == '\\'))
518 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 1028 {
519 case 't': APPEND_CH ('\011'); ++dec->cur; break; 1029 switch (*dec_cur)
520 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
521 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
522 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
523
524 case 'u':
525 { 1030 {
526 UV lo, hi; 1031 case '\\':
527 ++dec->cur; 1032 case '/':
1033 case '"': *cur++ = *dec_cur++; break;
528 1034
529 hi = decode_4hex (dec); 1035 case 'b': ++dec_cur; *cur++ = '\010'; break;
530 if (hi == (UV)-1) 1036 case 't': ++dec_cur; *cur++ = '\011'; break;
531 goto fail; 1037 case 'n': ++dec_cur; *cur++ = '\012'; break;
1038 case 'f': ++dec_cur; *cur++ = '\014'; break;
1039 case 'r': ++dec_cur; *cur++ = '\015'; break;
532 1040
533 // possibly a surrogate pair 1041 case 'u':
534 if (hi >= 0xd800 && hi < 0xdc00)
535 { 1042 {
536 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 1043 UV lo, hi;
537 ERR ("missing low surrogate character in surrogate pair"); 1044 ++dec_cur;
538 1045
539 dec->cur += 2; 1046 dec->cur = dec_cur;
540
541 lo = decode_4hex (dec); 1047 hi = decode_4hex (dec);
1048 dec_cur = dec->cur;
542 if (lo == (UV)-1) 1049 if (hi == (UV)-1)
543 goto fail; 1050 goto fail;
544 1051
1052 // possibly a surrogate pair
1053 if (hi >= 0xd800)
1054 if (hi < 0xdc00)
1055 {
1056 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
1057 ERR ("missing low surrogate character in surrogate pair");
1058
1059 dec_cur += 2;
1060
1061 dec->cur = dec_cur;
1062 lo = decode_4hex (dec);
1063 dec_cur = dec->cur;
1064 if (lo == (UV)-1)
1065 goto fail;
1066
545 if (lo < 0xdc00 || lo >= 0xe000) 1067 if (lo < 0xdc00 || lo >= 0xe000)
546 ERR ("surrogate pair expected"); 1068 ERR ("surrogate pair expected");
547 1069
548 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 1070 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
1071 }
1072 else if (hi < 0xe000)
1073 ERR ("missing high surrogate character in surrogate pair");
1074
1075 if (hi >= 0x80)
1076 {
1077 utf8 = 1;
1078
1079 cur = encode_utf8 (cur, hi);
1080 }
1081 else
1082 *cur++ = hi;
549 } 1083 }
550 else if (hi >= 0xdc00 && hi < 0xe000)
551 ERR ("missing high surrogate character in surrogate pair");
552
553 if (hi >= 0x80)
554 { 1084 break;
555 utf8 = 1;
556 1085
557 APPEND_GROW (4); // at most 4 bytes for 21 bits
558 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
559 }
560 else 1086 default:
561 APPEND_CH (hi); 1087 --dec_cur;
1088 ERR ("illegal backslash escape sequence in string");
562 } 1089 }
563 break; 1090 }
1091 else if (expect_true (ch >= 0x20 && ch < 0x80))
1092 *cur++ = ch;
1093 else if (ch >= 0x80)
1094 {
1095 STRLEN clen;
564 1096
565 default:
566 --dec->cur; 1097 --dec_cur;
567 ERR ("illegal backslash escape sequence in string"); 1098
1099 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
1100 if (clen == (STRLEN)-1)
1101 ERR ("malformed UTF-8 character in JSON string");
1102
1103 do
1104 *cur++ = *dec_cur++;
1105 while (--clen);
1106
1107 utf8 = 1;
1108 }
1109 else
1110 {
1111 --dec_cur;
1112
1113 if (!ch)
1114 ERR ("unexpected end of string while parsing JSON string");
1115 else
1116 ERR ("invalid character encountered while parsing JSON string");
568 } 1117 }
569 } 1118 }
570 else if (ch >= 0x20 && ch <= 0x7f) 1119 while (cur < buf + SHORT_STRING_LEN);
571 APPEND_CH (*dec->cur++); 1120
572 else if (ch >= 0x80)
573 { 1121 {
574 STRLEN clen; 1122 STRLEN len = cur - buf;
575 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
576 if (clen == (STRLEN)-1)
577 ERR ("malformed UTF-8 character in string, cannot convert to JSON");
578 1123
579 APPEND_GROW (clen); 1124 if (sv)
580 do
581 { 1125 {
582 *cur++ = *dec->cur++; 1126 STRLEN cur = SvCUR (sv);
1127
1128 if (SvLEN (sv) <= cur + len)
1129 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
1130
1131 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
1132 SvCUR_set (sv, SvCUR (sv) + len);
583 } 1133 }
584 while (--clen);
585
586 utf8 = 1;
587 }
588 else if (dec->cur == dec->end)
589 ERR ("unexpected end of string while parsing json string");
590 else 1134 else
591 ERR ("invalid character encountered"); 1135 sv = newSVpvn (buf, len);
592 } 1136 }
1137 }
1138 while (*dec_cur != '"');
593 1139
594 ++dec->cur; 1140 ++dec_cur;
595 1141
596 SvCUR_set (sv, cur - SvPVX (sv)); 1142 if (sv)
597 1143 {
598 SvPOK_only (sv); 1144 SvPOK_only (sv);
599 *SvEND (sv) = 0; 1145 *SvEND (sv) = 0;
600 1146
601 if (utf8) 1147 if (utf8)
602 SvUTF8_on (sv); 1148 SvUTF8_on (sv);
1149 }
1150 else
1151 sv = newSVpvn ("", 0);
603 1152
1153 dec->cur = dec_cur;
604 return sv; 1154 return sv;
605 1155
606fail: 1156fail:
607 SvREFCNT_dec (sv); 1157 dec->cur = dec_cur;
608 return 0; 1158 return 0;
609} 1159}
610 1160
611static SV * 1161static SV *
612decode_num (dec_t *dec) 1162decode_num (dec_t *dec)
670 is_nv = 1; 1220 is_nv = 1;
671 } 1221 }
672 1222
673 if (!is_nv) 1223 if (!is_nv)
674 { 1224 {
675 UV uv; 1225 int len = dec->cur - start;
676 int numtype = grok_number (start, dec->cur - start, &uv); 1226
677 if (numtype & IS_NUMBER_IN_UV) 1227 // special case the rather common 1..5-digit-int case
678 if (numtype & IS_NUMBER_NEG) 1228 if (*start == '-')
1229 switch (len)
679 { 1230 {
680 if (uv < (UV)IV_MIN) 1231 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
681 return newSViv (-(IV)uv); 1232 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1233 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1234 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1235 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
682 } 1236 }
1237 else
1238 switch (len)
1239 {
1240 case 1: return newSViv ( start [0] - '0' * 1);
1241 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1242 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1243 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1244 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1245 }
1246
1247 {
1248 UV uv;
1249 int numtype = grok_number (start, len, &uv);
1250 if (numtype & IS_NUMBER_IN_UV)
1251 if (numtype & IS_NUMBER_NEG)
1252 {
1253 if (uv < (UV)IV_MIN)
1254 return newSViv (-(IV)uv);
1255 }
683 else 1256 else
684 return newSVuv (uv); 1257 return newSVuv (uv);
685 } 1258 }
686 1259
1260 len -= *start == '-' ? 1 : 0;
1261
1262 // does not fit into IV or UV, try NV
1263 if (len <= NV_DIG)
1264 // fits into NV without loss of precision
1265 return newSVnv (json_atof (start));
1266
1267 // everything else fails, convert it to a string
1268 return newSVpvn (start, dec->cur - start);
1269 }
1270
1271 // loss of precision here
687 return newSVnv (Atof (start)); 1272 return newSVnv (json_atof (start));
688 1273
689fail: 1274fail:
690 return 0; 1275 return 0;
691} 1276}
692 1277
693static SV * 1278static SV *
694decode_av (dec_t *dec) 1279decode_av (dec_t *dec)
695{ 1280{
696 AV *av = newAV (); 1281 AV *av = newAV ();
697 1282
698 WS; 1283 DEC_INC_DEPTH;
1284 decode_ws (dec);
1285
699 if (*dec->cur == ']') 1286 if (*dec->cur == ']')
700 ++dec->cur; 1287 ++dec->cur;
701 else 1288 else
702 for (;;) 1289 for (;;)
703 { 1290 {
707 if (!value) 1294 if (!value)
708 goto fail; 1295 goto fail;
709 1296
710 av_push (av, value); 1297 av_push (av, value);
711 1298
712 WS; 1299 decode_ws (dec);
713 1300
714 if (*dec->cur == ']') 1301 if (*dec->cur == ']')
715 { 1302 {
716 ++dec->cur; 1303 ++dec->cur;
717 break; 1304 break;
718 } 1305 }
719 1306
720 if (*dec->cur != ',') 1307 if (*dec->cur != ',')
721 ERR (", or ] expected while parsing array"); 1308 ERR (", or ] expected while parsing array");
722 1309
723 ++dec->cur; 1310 ++dec->cur;
1311
1312 decode_ws (dec);
1313
1314 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1315 {
1316 ++dec->cur;
1317 break;
1318 }
724 } 1319 }
725 1320
1321 DEC_DEC_DEPTH;
726 return newRV_noinc ((SV *)av); 1322 return newRV_noinc ((SV *)av);
727 1323
728fail: 1324fail:
729 SvREFCNT_dec (av); 1325 SvREFCNT_dec (av);
1326 DEC_DEC_DEPTH;
730 return 0; 1327 return 0;
731} 1328}
732 1329
733static SV * 1330static SV *
734decode_hv (dec_t *dec) 1331decode_hv (dec_t *dec)
735{ 1332{
1333 SV *sv;
736 HV *hv = newHV (); 1334 HV *hv = newHV ();
737 1335
738 WS; 1336 DEC_INC_DEPTH;
1337 decode_ws (dec);
1338
739 if (*dec->cur == '}') 1339 if (*dec->cur == '}')
740 ++dec->cur; 1340 ++dec->cur;
741 else 1341 else
742 for (;;) 1342 for (;;)
743 { 1343 {
744 SV *key, *value;
745
746 WS; EXPECT_CH ('"'); 1344 EXPECT_CH ('"');
747 1345
748 key = decode_str (dec); 1346 // heuristic: assume that
749 if (!key) 1347 // a) decode_str + hv_store_ent are abysmally slow.
750 goto fail; 1348 // b) most hash keys are short, simple ascii text.
1349 // => try to "fast-match" such strings to avoid
1350 // the overhead of decode_str + hv_store_ent.
1351 {
1352 SV *value;
1353 char *p = dec->cur;
1354 char *e = p + 24; // only try up to 24 bytes
751 1355
752 WS; EXPECT_CH (':'); 1356 for (;;)
753
754 value = decode_sv (dec);
755 if (!value)
756 { 1357 {
1358 // the >= 0x80 is false on most architectures
1359 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1360 {
1361 // slow path, back up and use decode_str
1362 SV *key = decode_str (dec);
1363 if (!key)
1364 goto fail;
1365
1366 decode_ws (dec); EXPECT_CH (':');
1367
1368 decode_ws (dec);
1369 value = decode_sv (dec);
1370 if (!value)
1371 {
1372 SvREFCNT_dec (key);
1373 goto fail;
1374 }
1375
1376 hv_store_ent (hv, key, value, 0);
757 SvREFCNT_dec (key); 1377 SvREFCNT_dec (key);
1378
1379 break;
1380 }
1381 else if (*p == '"')
1382 {
1383 // fast path, got a simple key
1384 char *key = dec->cur;
1385 int len = p - key;
1386 dec->cur = p + 1;
1387
1388 decode_ws (dec); EXPECT_CH (':');
1389
1390 decode_ws (dec);
1391 value = decode_sv (dec);
1392 if (!value)
758 goto fail; 1393 goto fail;
1394
1395 hv_store (hv, key, len, value, 0);
1396
1397 break;
1398 }
1399
1400 ++p;
759 } 1401 }
760
761 //TODO: optimise
762 hv_store_ent (hv, key, value, 0);
763
764 WS; 1402 }
1403
1404 decode_ws (dec);
765 1405
766 if (*dec->cur == '}') 1406 if (*dec->cur == '}')
767 { 1407 {
768 ++dec->cur; 1408 ++dec->cur;
769 break; 1409 break;
771 1411
772 if (*dec->cur != ',') 1412 if (*dec->cur != ',')
773 ERR (", or } expected while parsing object/hash"); 1413 ERR (", or } expected while parsing object/hash");
774 1414
775 ++dec->cur; 1415 ++dec->cur;
1416
1417 decode_ws (dec);
1418
1419 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1420 {
1421 ++dec->cur;
1422 break;
1423 }
776 } 1424 }
777 1425
1426 DEC_DEC_DEPTH;
778 return newRV_noinc ((SV *)hv); 1427 sv = newRV_noinc ((SV *)hv);
1428
1429 // check filter callbacks
1430 if (dec->json.flags & F_HOOK)
1431 {
1432 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1433 {
1434 HE *cb, *he;
1435
1436 hv_iterinit (hv);
1437 he = hv_iternext (hv);
1438 hv_iterinit (hv);
1439
1440 // the next line creates a mortal sv each time its called.
1441 // might want to optimise this for common cases.
1442 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1443
1444 if (cb)
1445 {
1446 dSP;
1447 int count;
1448
1449 ENTER; SAVETMPS; PUSHMARK (SP);
1450 XPUSHs (HeVAL (he));
1451 sv_2mortal (sv);
1452
1453 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1454
1455 if (count == 1)
1456 {
1457 sv = newSVsv (POPs);
1458 FREETMPS; LEAVE;
1459 return sv;
1460 }
1461
1462 SvREFCNT_inc (sv);
1463 FREETMPS; LEAVE;
1464 }
1465 }
1466
1467 if (dec->json.cb_object)
1468 {
1469 dSP;
1470 int count;
1471
1472 ENTER; SAVETMPS; PUSHMARK (SP);
1473 XPUSHs (sv_2mortal (sv));
1474
1475 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1476
1477 if (count == 1)
1478 {
1479 sv = newSVsv (POPs);
1480 FREETMPS; LEAVE;
1481 return sv;
1482 }
1483
1484 SvREFCNT_inc (sv);
1485 FREETMPS; LEAVE;
1486 }
1487 }
1488
1489 return sv;
779 1490
780fail: 1491fail:
781 SvREFCNT_dec (hv); 1492 SvREFCNT_dec (hv);
1493 DEC_DEC_DEPTH;
1494 return 0;
1495}
1496
1497static SV *
1498decode_tag (dec_t *dec)
1499{
1500 SV *tag = 0;
1501 SV *val = 0;
1502
1503 if (!(dec->json.flags & F_ALLOW_TAGS))
1504 ERR ("malformed JSON string, neither array, object, number, string or atom");
1505
1506 ++dec->cur;
1507
1508 decode_ws (dec);
1509
1510 tag = decode_sv (dec);
1511 if (!tag)
1512 goto fail;
1513
1514 if (!SvPOK (tag))
1515 ERR ("malformed JSON string, (tag) must be a string");
1516
1517 decode_ws (dec);
1518
1519 if (*dec->cur != ')')
1520 ERR (") expected after tag");
1521
1522 ++dec->cur;
1523
1524 decode_ws (dec);
1525
1526 val = decode_sv (dec);
1527 if (!val)
1528 goto fail;
1529
1530 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1531 ERR ("malformed JSON string, tag value must be an array");
1532
1533 {
1534 AV *av = (AV *)SvRV (val);
1535 int i, len = av_len (av) + 1;
1536 HV *stash = gv_stashsv (tag, 0);
1537 SV *sv;
1538
1539 if (!stash)
1540 ERR ("cannot decode perl-object (package does not exist)");
1541
1542 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1543
1544 if (!method)
1545 ERR ("cannot decode perl-object (package does not have a THAW method)");
1546
1547 dSP;
1548
1549 ENTER; SAVETMPS; PUSHMARK (SP);
1550 EXTEND (SP, len + 2);
1551 // we re-bless the reference to get overload and other niceties right
1552 PUSHs (tag);
1553 PUSHs (sv_json);
1554
1555 for (i = 0; i < len; ++i)
1556 PUSHs (*av_fetch (av, i, 1));
1557
1558 PUTBACK;
1559 call_sv ((SV *)GvCV (method), G_SCALAR);
1560 SPAGAIN;
1561
1562 SvREFCNT_dec (tag);
1563 SvREFCNT_dec (val);
1564 sv = SvREFCNT_inc (POPs);
1565
1566 PUTBACK;
1567
1568 FREETMPS; LEAVE;
1569
1570 return sv;
1571 }
1572
1573fail:
1574 SvREFCNT_dec (tag);
1575 SvREFCNT_dec (val);
782 return 0; 1576 return 0;
783} 1577}
784 1578
785static SV * 1579static SV *
786decode_sv (dec_t *dec) 1580decode_sv (dec_t *dec)
787{ 1581{
788 WS; 1582 // the beauty of JSON: you need exactly one character lookahead
1583 // to parse everything.
789 switch (*dec->cur) 1584 switch (*dec->cur)
790 { 1585 {
791 case '"': ++dec->cur; return decode_str (dec); 1586 case '"': ++dec->cur; return decode_str (dec);
792 case '[': ++dec->cur; return decode_av (dec); 1587 case '[': ++dec->cur; return decode_av (dec);
793 case '{': ++dec->cur; return decode_hv (dec); 1588 case '{': ++dec->cur; return decode_hv (dec);
1589 case '(': return decode_tag (dec);
794 1590
795 case '-': 1591 case '-':
796 case '0': case '1': case '2': case '3': case '4': 1592 case '0': case '1': case '2': case '3': case '4':
797 case '5': case '6': case '7': case '8': case '9': 1593 case '5': case '6': case '7': case '8': case '9':
798 return decode_num (dec); 1594 return decode_num (dec);
799 1595
800 case 't': 1596 case 't':
801 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1597 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
802 { 1598 {
803 dec->cur += 4; 1599 dec->cur += 4;
1600#if JSON_SLOW
1601 types_true = get_bool ("Types::Serialiser::true");
1602#endif
804 return newSViv (1); 1603 return newSVsv (types_true);
805 } 1604 }
806 else 1605 else
807 ERR ("'true' expected"); 1606 ERR ("'true' expected");
808 1607
809 break; 1608 break;
810 1609
811 case 'f': 1610 case 'f':
812 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1611 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
813 { 1612 {
814 dec->cur += 5; 1613 dec->cur += 5;
1614#if JSON_SLOW
1615 types_false = get_bool ("Types::Serialiser::false");
1616#endif
815 return newSViv (0); 1617 return newSVsv (types_false);
816 } 1618 }
817 else 1619 else
818 ERR ("'false' expected"); 1620 ERR ("'false' expected");
819 1621
820 break; 1622 break;
829 ERR ("'null' expected"); 1631 ERR ("'null' expected");
830 1632
831 break; 1633 break;
832 1634
833 default: 1635 default:
834 ERR ("malformed json string"); 1636 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
835 break; 1637 break;
836 } 1638 }
837 1639
838fail: 1640fail:
839 return 0; 1641 return 0;
840} 1642}
841 1643
842static SV * 1644static SV *
843decode_json (SV *string, UV flags) 1645decode_json (SV *string, JSON *json, char **offset_return)
844{ 1646{
1647 dec_t dec;
845 SV *sv; 1648 SV *sv;
846 1649
847 if (flags & F_UTF8) 1650 /* work around bugs in 5.10 where manipulating magic values
1651 * makes perl ignore the magic in subsequent accesses.
1652 * also make a copy of non-PV values, to get them into a clean
1653 * state (SvPV should do that, but it's buggy, see below).
1654 */
1655 /*SvGETMAGIC (string);*/
1656 if (SvMAGICAL (string) || !SvPOK (string))
1657 string = sv_2mortal (newSVsv (string));
1658
1659 SvUPGRADE (string, SVt_PV);
1660
1661 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1662 * assertion with -DDEBUGGING, although SvCUR is documented to
1663 * return the xpv_cur field which certainly exists after upgrading.
1664 * according to nicholas clark, calling SvPOK fixes this.
1665 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1666 * and hope for the best.
1667 * Damnit, SvPV_nolen still trips over yet another assertion. This
1668 * assertion business is seriously broken, try yet another workaround
1669 * for the broken -DDEBUGGING.
1670 */
1671 {
1672#ifdef DEBUGGING
1673 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1674#else
1675 STRLEN offset = SvCUR (string);
1676#endif
1677
1678 if (offset > json->max_size && json->max_size)
1679 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1680 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1681 }
1682
1683 if (DECODE_WANTS_OCTETS (json))
848 sv_utf8_downgrade (string, 0); 1684 sv_utf8_downgrade (string, 0);
849 else 1685 else
850 sv_utf8_upgrade (string); 1686 sv_utf8_upgrade (string);
851 1687
852 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1688 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
853 1689
854 dec_t dec; 1690 dec.json = *json;
855 dec.flags = flags;
856 dec.cur = SvPVX (string); 1691 dec.cur = SvPVX (string);
857 dec.end = SvEND (string); 1692 dec.end = SvEND (string);
858 dec.err = 0; 1693 dec.err = 0;
1694 dec.depth = 0;
859 1695
1696 if (dec.json.cb_object || dec.json.cb_sk_object)
1697 dec.json.flags |= F_HOOK;
1698
1699 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1700
1701 decode_ws (&dec);
860 sv = decode_sv (&dec); 1702 sv = decode_sv (&dec);
861 1703
1704 if (offset_return)
1705 *offset_return = dec.cur;
1706
1707 if (!(offset_return || !sv))
1708 {
1709 // check for trailing garbage
1710 decode_ws (&dec);
1711
1712 if (*dec.cur)
1713 {
1714 dec.err = "garbage after JSON object";
1715 SvREFCNT_dec (sv);
1716 sv = 0;
1717 }
1718 }
1719
862 if (!sv) 1720 if (!sv)
863 { 1721 {
864 IV offset = utf8_distance (dec.cur, SvPVX (string));
865 SV *uni = sv_newmortal (); 1722 SV *uni = sv_newmortal ();
1723
866 // horrible hack to silence warning inside pv_uni_display 1724 // horrible hack to silence warning inside pv_uni_display
867 COP cop; 1725 COP cop = *PL_curcop;
868 memset (&cop, 0, sizeof (cop));
869 cop.cop_warnings = pWARN_NONE; 1726 cop.cop_warnings = pWARN_NONE;
1727 ENTER;
870 SAVEVPTR (PL_curcop); 1728 SAVEVPTR (PL_curcop);
871 PL_curcop = &cop; 1729 PL_curcop = &cop;
872
873 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1730 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1731 LEAVE;
1732
874 croak ("%s, at character offset %d (%s)", 1733 croak ("%s, at character offset %d (before \"%s\")",
875 dec.err, 1734 dec.err,
876 (int)offset, 1735 (int)ptr_to_index (string, dec.cur),
877 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1736 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
878 } 1737 }
879 1738
880 sv = sv_2mortal (sv); 1739 sv = sv_2mortal (sv);
881 1740
882 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1741 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
883 croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); 1742 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
884 1743
885 return sv; 1744 return sv;
886} 1745}
887 1746
1747/////////////////////////////////////////////////////////////////////////////
1748// incremental parser
1749
1750static void
1751incr_parse (JSON *self)
1752{
1753 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1754
1755 // the state machine here is a bit convoluted and could be simplified a lot
1756 // but this would make it slower, so...
1757
1758 for (;;)
1759 {
1760 //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
1761 switch (self->incr_mode)
1762 {
1763 // only used for initial whitespace skipping
1764 case INCR_M_WS:
1765 for (;;)
1766 {
1767 if (*p > 0x20)
1768 {
1769 if (*p == '#')
1770 {
1771 self->incr_mode = INCR_M_C0;
1772 goto incr_m_c;
1773 }
1774 else
1775 {
1776 self->incr_mode = INCR_M_JSON;
1777 goto incr_m_json;
1778 }
1779 }
1780 else if (!*p)
1781 goto interrupt;
1782
1783 ++p;
1784 }
1785
1786 // skip a single char inside a string (for \\-processing)
1787 case INCR_M_BS:
1788 if (!*p)
1789 goto interrupt;
1790
1791 ++p;
1792 self->incr_mode = INCR_M_STR;
1793 goto incr_m_str;
1794
1795 // inside #-style comments
1796 case INCR_M_C0:
1797 case INCR_M_C1:
1798 incr_m_c:
1799 for (;;)
1800 {
1801 if (*p == '\n')
1802 {
1803 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1804 break;
1805 }
1806 else if (!*p)
1807 goto interrupt;
1808
1809 ++p;
1810 }
1811
1812 break;
1813
1814 // inside a string
1815 case INCR_M_STR:
1816 incr_m_str:
1817 for (;;)
1818 {
1819 if (*p == '"')
1820 {
1821 ++p;
1822 self->incr_mode = INCR_M_JSON;
1823
1824 if (!self->incr_nest)
1825 goto interrupt;
1826
1827 goto incr_m_json;
1828 }
1829 else if (*p == '\\')
1830 {
1831 ++p; // "virtually" consumes character after \
1832
1833 if (!*p) // if at end of string we have to switch modes
1834 {
1835 self->incr_mode = INCR_M_BS;
1836 goto interrupt;
1837 }
1838 }
1839 else if (!*p)
1840 goto interrupt;
1841
1842 ++p;
1843 }
1844
1845 // after initial ws, outside string
1846 case INCR_M_JSON:
1847 incr_m_json:
1848 for (;;)
1849 {
1850 switch (*p++)
1851 {
1852 case 0:
1853 --p;
1854 goto interrupt;
1855
1856 case 0x09:
1857 case 0x0a:
1858 case 0x0d:
1859 case 0x20:
1860 if (!self->incr_nest)
1861 {
1862 --p; // do not eat the whitespace, let the next round do it
1863 goto interrupt;
1864 }
1865 break;
1866
1867 case '"':
1868 self->incr_mode = INCR_M_STR;
1869 goto incr_m_str;
1870
1871 case '[':
1872 case '{':
1873 case '(':
1874 if (++self->incr_nest > self->max_depth)
1875 croak (ERR_NESTING_EXCEEDED);
1876 break;
1877
1878 case ']':
1879 case '}':
1880 if (--self->incr_nest <= 0)
1881 goto interrupt;
1882 break;
1883
1884 case ')':
1885 --self->incr_nest;
1886 break;
1887
1888 case '#':
1889 self->incr_mode = INCR_M_C1;
1890 goto incr_m_c;
1891 }
1892 }
1893 }
1894
1895 modechange:
1896 ;
1897 }
1898
1899interrupt:
1900 self->incr_pos = p - SvPVX (self->incr_text);
1901 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1902 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1903}
1904
1905/////////////////////////////////////////////////////////////////////////////
1906// XS interface functions
1907
888MODULE = JSON::XS PACKAGE = JSON::XS 1908MODULE = JSON::XS PACKAGE = JSON::XS
889 1909
890BOOT: 1910BOOT:
891{ 1911{
892 int i; 1912 int i;
893 1913
894 memset (decode_hexdigit, 0xff, 256);
895 for (i = 10; i--; ) 1914 for (i = 0; i < 256; ++i)
896 decode_hexdigit ['0' + i] = i; 1915 decode_hexdigit [i] =
1916 i >= '0' && i <= '9' ? i - '0'
1917 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1918 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1919 : -1;
897 1920
898 for (i = 7; i--; )
899 {
900 decode_hexdigit ['a' + i] = 10 + i;
901 decode_hexdigit ['A' + i] = 10 + i;
902 }
903
904 json_stash = gv_stashpv ("JSON::XS", 1); 1921 json_stash = gv_stashpv ("JSON::XS" , 1);
1922 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1923
1924 types_true = get_bool ("Types::Serialiser::true");
1925 types_false = get_bool ("Types::Serialiser::false");
1926
1927 sv_json = newSVpv ("JSON", 0);
1928 SvREADONLY_on (sv_json);
1929
1930 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
905} 1931}
906 1932
907PROTOTYPES: DISABLE 1933PROTOTYPES: DISABLE
908 1934
909SV *new (char *dummy) 1935void CLONE (...)
910 CODE: 1936 CODE:
911 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1937 json_stash = 0;
1938 types_boolean_stash = 0;
1939
1940void new (char *klass)
1941 PPCODE:
1942{
1943 SV *pv = NEWSV (0, sizeof (JSON));
1944 SvPOK_only (pv);
1945 json_init ((JSON *)SvPVX (pv));
1946 XPUSHs (sv_2mortal (sv_bless (
1947 newRV_noinc (pv),
1948 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1949 )));
1950}
1951
1952void ascii (JSON *self, int enable = 1)
1953 ALIAS:
1954 ascii = F_ASCII
1955 latin1 = F_LATIN1
1956 utf8 = F_UTF8
1957 indent = F_INDENT
1958 canonical = F_CANONICAL
1959 space_before = F_SPACE_BEFORE
1960 space_after = F_SPACE_AFTER
1961 pretty = F_PRETTY
1962 allow_nonref = F_ALLOW_NONREF
1963 shrink = F_SHRINK
1964 allow_blessed = F_ALLOW_BLESSED
1965 convert_blessed = F_CONV_BLESSED
1966 relaxed = F_RELAXED
1967 allow_unknown = F_ALLOW_UNKNOWN
1968 allow_tags = F_ALLOW_TAGS
1969 PPCODE:
1970{
1971 if (enable)
1972 self->flags |= ix;
1973 else
1974 self->flags &= ~ix;
1975
1976 XPUSHs (ST (0));
1977}
1978
1979void get_ascii (JSON *self)
1980 ALIAS:
1981 get_ascii = F_ASCII
1982 get_latin1 = F_LATIN1
1983 get_utf8 = F_UTF8
1984 get_indent = F_INDENT
1985 get_canonical = F_CANONICAL
1986 get_space_before = F_SPACE_BEFORE
1987 get_space_after = F_SPACE_AFTER
1988 get_allow_nonref = F_ALLOW_NONREF
1989 get_shrink = F_SHRINK
1990 get_allow_blessed = F_ALLOW_BLESSED
1991 get_convert_blessed = F_CONV_BLESSED
1992 get_relaxed = F_RELAXED
1993 get_allow_unknown = F_ALLOW_UNKNOWN
1994 get_allow_tags = F_ALLOW_TAGS
1995 PPCODE:
1996 XPUSHs (boolSV (self->flags & ix));
1997
1998void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1999 PPCODE:
2000 self->max_depth = max_depth;
2001 XPUSHs (ST (0));
2002
2003U32 get_max_depth (JSON *self)
2004 CODE:
2005 RETVAL = self->max_depth;
912 OUTPUT: 2006 OUTPUT:
913 RETVAL 2007 RETVAL
914 2008
915SV *ascii (SV *self, int enable) 2009void max_size (JSON *self, U32 max_size = 0)
916 ALIAS: 2010 PPCODE:
917 ascii = F_ASCII 2011 self->max_size = max_size;
918 utf8 = F_UTF8 2012 XPUSHs (ST (0));
919 indent = F_INDENT 2013
920 canonical = F_CANONICAL 2014int get_max_size (JSON *self)
921 space_before = F_SPACE_BEFORE
922 space_after = F_SPACE_AFTER
923 json_rpc = F_JSON_RPC
924 pretty = F_PRETTY
925 allow_nonref = F_ALLOW_NONREF
926 CODE: 2015 CODE:
927{ 2016 RETVAL = self->max_size;
928 UV *uv = SvJSON (self);
929 if (enable)
930 *uv |= ix;
931 else
932 *uv &= ~ix;
933
934 RETVAL = newSVsv (self);
935}
936 OUTPUT: 2017 OUTPUT:
937 RETVAL 2018 RETVAL
938 2019
939void encode (SV *self, SV *scalar) 2020void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
940 PPCODE: 2021 PPCODE:
941 XPUSHs (encode_json (scalar, *SvJSON (self))); 2022{
2023 SvREFCNT_dec (self->cb_object);
2024 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
942 2025
943void decode (SV *self, SV *jsonstr) 2026 XPUSHs (ST (0));
2027}
2028
2029void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
944 PPCODE: 2030 PPCODE:
945 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 2031{
2032 if (!self->cb_sk_object)
2033 self->cb_sk_object = newHV ();
2034
2035 if (SvOK (cb))
2036 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
2037 else
2038 {
2039 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
2040
2041 if (!HvKEYS (self->cb_sk_object))
2042 {
2043 SvREFCNT_dec (self->cb_sk_object);
2044 self->cb_sk_object = 0;
2045 }
2046 }
2047
2048 XPUSHs (ST (0));
2049}
2050
2051void encode (JSON *self, SV *scalar)
2052 PPCODE:
2053 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2054 XPUSHs (scalar);
2055
2056void decode (JSON *self, SV *jsonstr)
2057 PPCODE:
2058 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2059 XPUSHs (jsonstr);
2060
2061void decode_prefix (JSON *self, SV *jsonstr)
2062 PPCODE:
2063{
2064 SV *sv;
2065 char *offset;
2066 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
2067 EXTEND (SP, 2);
2068 PUSHs (sv);
2069 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
2070}
2071
2072void incr_parse (JSON *self, SV *jsonstr = 0)
2073 PPCODE:
2074{
2075 if (!self->incr_text)
2076 self->incr_text = newSVpvn ("", 0);
2077
2078 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2079 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2080 if (DECODE_WANTS_OCTETS (self))
2081 {
2082 if (self->incr_pos)
2083 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2084 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2085
2086 sv_utf8_downgrade (self->incr_text, 0);
2087 }
2088 else
2089 {
2090 sv_utf8_upgrade (self->incr_text);
2091
2092 if (self->incr_pos)
2093 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2094 - (U8 *)SvPVX (self->incr_text);
2095 }
2096
2097 // append data, if any
2098 if (jsonstr)
2099 {
2100 /* make sure both strings have same encoding */
2101 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2102 if (SvUTF8 (jsonstr))
2103 sv_utf8_downgrade (jsonstr, 0);
2104 else
2105 sv_utf8_upgrade (jsonstr);
2106
2107 /* and then just blindly append */
2108 {
2109 STRLEN len;
2110 const char *str = SvPV (jsonstr, len);
2111 STRLEN cur = SvCUR (self->incr_text);
2112
2113 if (SvLEN (self->incr_text) <= cur + len)
2114 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2115
2116 Move (str, SvEND (self->incr_text), len, char);
2117 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2118 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2119 }
2120 }
2121
2122 if (GIMME_V != G_VOID)
2123 do
2124 {
2125 SV *sv;
2126 char *offset;
2127
2128 if (!INCR_DONE (self))
2129 {
2130 incr_parse (self);
2131
2132 if (self->incr_pos > self->max_size && self->max_size)
2133 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2134 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2135
2136 if (!INCR_DONE (self))
2137 {
2138 // as an optimisation, do not accumulate white space in the incr buffer
2139 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2140 {
2141 self->incr_pos = 0;
2142 SvCUR_set (self->incr_text, 0);
2143 }
2144
2145 break;
2146 }
2147 }
2148
2149 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2150 XPUSHs (sv);
2151
2152 self->incr_pos -= offset - SvPVX (self->incr_text);
2153 self->incr_nest = 0;
2154 self->incr_mode = 0;
2155
2156 sv_chop (self->incr_text, offset);
2157 }
2158 while (GIMME_V == G_ARRAY);
2159}
2160
2161SV *incr_text (JSON *self)
2162 ATTRS: lvalue
2163 CODE:
2164{
2165 if (self->incr_pos)
2166 croak ("incr_text can not be called when the incremental parser already started parsing");
2167
2168 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2169}
2170 OUTPUT:
2171 RETVAL
2172
2173void incr_skip (JSON *self)
2174 CODE:
2175{
2176 if (self->incr_pos)
2177 {
2178 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2179 self->incr_pos = 0;
2180 self->incr_nest = 0;
2181 self->incr_mode = 0;
2182 }
2183}
2184
2185void incr_reset (JSON *self)
2186 CODE:
2187{
2188 SvREFCNT_dec (self->incr_text);
2189 self->incr_text = 0;
2190 self->incr_pos = 0;
2191 self->incr_nest = 0;
2192 self->incr_mode = 0;
2193}
2194
2195void DESTROY (JSON *self)
2196 CODE:
2197 SvREFCNT_dec (self->cb_sk_object);
2198 SvREFCNT_dec (self->cb_object);
2199 SvREFCNT_dec (self->incr_text);
946 2200
947PROTOTYPES: ENABLE 2201PROTOTYPES: ENABLE
948 2202
949void to_json (SV *scalar) 2203void encode_json (SV *scalar)
950 PPCODE: 2204 PPCODE:
951 XPUSHs (encode_json (scalar, F_UTF8)); 2205{
2206 JSON json;
2207 json_init (&json);
2208 json.flags |= F_UTF8;
2209 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2210 XPUSHs (scalar);
2211}
952 2212
953void from_json (SV *jsonstr) 2213void decode_json (SV *jsonstr)
954 PPCODE: 2214 PPCODE:
955 XPUSHs (decode_json (jsonstr, F_UTF8)); 2215{
2216 JSON json;
2217 json_init (&json);
2218 json.flags |= F_UTF8;
2219 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2220 XPUSHs (jsonstr);
2221}
956 2222

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines