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 1.125 by root, Sun Feb 21 15:37:53 2016 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 < 0) 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 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 } 441 }
176 } 442 }
177 443
178 --len; 444 --len;
179 } 445 }
180} 446}
181 447
182#define INDENT SB \ 448INLINE void
449encode_indent (enc_t *enc)
450{
183 if (enc->flags & F_INDENT) \ 451 if (enc->json.flags & F_INDENT)
184 { \ 452 {
185 int i_; \ 453 int spaces = enc->indent * INDENT_STEP;
186 need (enc, enc->indent); \ 454
187 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);
188 encode_ch (enc, ' '); \ 474 encode_ch (enc, '\n');
189 } \ 475 }
190 SE 476}
191 477
192#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 478INLINE void
193#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 479encode_comma (enc_t *enc)
194#define COMMA SB \ 480{
195 encode_ch (enc, ','); \ 481 encode_ch (enc, ',');
482
196 if (enc->flags & F_INDENT) \ 483 if (enc->json.flags & F_INDENT)
197 NL; \ 484 encode_nl (enc);
198 else if (enc->flags & F_SPACE_AFTER) \ 485 else if (enc->json.flags & F_SPACE_AFTER)
199 SPACE; \ 486 encode_space (enc);
200 SE 487}
201 488
202static void encode_sv (enc_t *enc, SV *sv); 489static void encode_sv (enc_t *enc, SV *sv);
203 490
204static void 491static void
205encode_av (enc_t *enc, AV *av) 492encode_av (enc_t *enc, AV *av)
206{ 493{
207 int i, len = av_len (av); 494 int i, len = av_len (av);
208 495
496 if (enc->indent >= enc->json.max_depth)
497 croak (ERR_NESTING_EXCEEDED);
498
209 encode_ch (enc, '['); NL; 499 encode_ch (enc, '[');
210 ++enc->indent;
211 500
501 if (len >= 0)
502 {
503 encode_nl (enc); ++enc->indent;
504
212 for (i = 0; i <= len; ++i) 505 for (i = 0; i <= len; ++i)
213 { 506 {
214 INDENT; 507 SV **svp = av_fetch (av, i, 0);
215 encode_sv (enc, *av_fetch (av, i, 0));
216 508
509 encode_indent (enc);
510
511 if (svp)
512 encode_sv (enc, *svp);
513 else
514 encode_str (enc, "null", 4, 0);
515
217 if (i < len) 516 if (i < len)
218 COMMA; 517 encode_comma (enc);
219 } 518 }
220 519
221 NL; 520 encode_nl (enc); --enc->indent; encode_indent (enc);
521 }
222 522
223 --enc->indent;
224 INDENT; encode_ch (enc, ']'); 523 encode_ch (enc, ']');
225} 524}
226 525
227static void 526static void
228encode_he (enc_t *enc, HE *he) 527encode_hk (enc_t *enc, HE *he)
229{ 528{
230 encode_ch (enc, '"'); 529 encode_ch (enc, '"');
231 530
232 if (HeKLEN (he) == HEf_SVKEY) 531 if (HeKLEN (he) == HEf_SVKEY)
233 { 532 {
234 SV *sv = HeSVKEY (he); 533 SV *sv = HeSVKEY (he);
235 STRLEN len; 534 STRLEN len;
236 char *str; 535 char *str;
237 536
238 SvGETMAGIC (sv); 537 SvGETMAGIC (sv);
239 str = SvPV (sv, len); 538 str = SvPV (sv, len);
240 539
241 encode_str (enc, str, len, SvUTF8 (sv)); 540 encode_str (enc, str, len, SvUTF8 (sv));
242 } 541 }
243 else 542 else
244 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 543 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245 544
246 encode_ch (enc, '"'); 545 encode_ch (enc, '"');
247 546
248 if (enc->flags & F_SPACE_BEFORE) SPACE; 547 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249 encode_ch (enc, ':'); 548 encode_ch (enc, ':');
250 if (enc->flags & F_SPACE_AFTER ) SPACE; 549 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251 encode_sv (enc, HeVAL (he));
252} 550}
253 551
254// compare hash entries, used when all keys are bytestrings 552// compare hash entries, used when all keys are bytestrings
255static int 553static int
256he_cmp_fast (const void *a_, const void *b_) 554he_cmp_fast (const void *a_, const void *b_)
… …
261 HE *b = *(HE **)b_; 559 HE *b = *(HE **)b_;
262 560
263 STRLEN la = HeKLEN (a); 561 STRLEN la = HeKLEN (a);
264 STRLEN lb = HeKLEN (b); 562 STRLEN lb = HeKLEN (b);
265 563
266 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 564 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267 cmp = la < lb ? -1 : la == lb ? 0 : 1; 565 cmp = lb - la;
268 566
269 return cmp; 567 return cmp;
270} 568}
271 569
272// 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
273static int 571static int
274he_cmp_slow (const void *a, const void *b) 572he_cmp_slow (const void *a, const void *b)
275{ 573{
276 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 574 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
277} 575}
278 576
279static void 577static void
280encode_hv (enc_t *enc, HV *hv) 578encode_hv (enc_t *enc, HV *hv)
281{ 579{
282 int count, i; 580 HE *he;
283 581
284 encode_ch (enc, '{'); NL; ++enc->indent; 582 if (enc->indent >= enc->json.max_depth)
583 croak (ERR_NESTING_EXCEEDED);
285 584
286 if ((count = hv_iterinit (hv))) 585 encode_ch (enc, '{');
287 { 586
288 // for canonical output we have to sort by keys first 587 // for canonical output we have to sort by keys first
289 // actually, this is mostly due to the stupid so-called 588 // actually, this is mostly due to the stupid so-called
290 // security workaround added somewhere in 5.8.x. 589 // security workaround added somewhere in 5.8.x
291 // that randomises hash orderings 590 // that randomises hash orderings
292 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))
293 { 596 {
294 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 {
295 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 }
296 621
297 i = 0; 622 i = 0;
298 while ((he = hv_iternext (hv))) 623 while ((he = hv_iternext (hv)))
299 { 624 {
300 hes [i++] = he; 625 hes [i++] = he;
… …
306 631
307 if (fast) 632 if (fast)
308 qsort (hes, count, sizeof (HE *), he_cmp_fast); 633 qsort (hes, count, sizeof (HE *), he_cmp_fast);
309 else 634 else
310 { 635 {
311 // hack to disable "use bytes" 636 // hack to forcefully disable "use bytes"
312 COP *oldcop = PL_curcop, cop; 637 COP cop = *PL_curcop;
313 cop.op_private = 0; 638 cop.op_private = 0;
639
640 ENTER;
641 SAVETMPS;
642
643 SAVEVPTR (PL_curcop);
314 PL_curcop = &cop; 644 PL_curcop = &cop;
315 645
316 SAVETMPS;
317 qsort (hes, count, sizeof (HE *), he_cmp_slow); 646 qsort (hes, count, sizeof (HE *), he_cmp_slow);
647
318 FREETMPS; 648 FREETMPS;
319 649 LEAVE;
320 PL_curcop = oldcop;
321 } 650 }
322 651
323 for (i = 0; i < count; ++i) 652 encode_nl (enc); ++enc->indent;
653
654 while (count--)
324 { 655 {
325 INDENT; 656 encode_indent (enc);
657 he = hes [count];
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 (count)
329 COMMA; 662 encode_comma (enc);
330 } 663 }
331 664
332 NL; 665 encode_nl (enc); --enc->indent; encode_indent (enc);
333 } 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;
724 SAVESTACK_POS ();
725 PUSHMARK (SP);
726 EXTEND (SP, 2);
727 // we re-bless the reference to get overload and other niceties right
728 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
729 PUSHs (sv_json);
730
731 PUTBACK;
732 count = call_sv ((SV *)GvCV (method), G_ARRAY);
733 SPAGAIN;
734
735 // catch this surprisingly common error
736 if (SvROK (TOPs) && SvRV (TOPs) == sv)
737 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
738
739 encode_ch (enc, '(');
740 encode_ch (enc, '"');
741 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
742 encode_ch (enc, '"');
743 encode_ch (enc, ')');
744 encode_ch (enc, '[');
745
746 while (count)
747 {
748 encode_sv (enc, SP[1 - count--]);
749
750 if (count)
751 encode_ch (enc, ',');
752 }
753
754 encode_ch (enc, ']');
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;
763 PUSHMARK (SP);
764 // we re-bless the reference to get overload and other niceties right
765 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
766
767 // calling with G_SCALAR ensures that we always get a 1 return value
768 PUTBACK;
769 call_sv ((SV *)GvCV (method), G_SCALAR);
770 SPAGAIN;
771
772 // catch this surprisingly common error
773 if (SvROK (TOPs) && SvRV (TOPs) == sv)
774 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
775
776 sv = POPs;
777 PUTBACK;
778
779 encode_sv (enc, sv);
780
781 FREETMPS; LEAVE;
782 }
783 else if (enc->json.flags & F_ALLOW_BLESSED)
784 encode_str (enc, "null", 4, 0);
334 else 785 else
335 { 786 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
336 SV *sv; 787 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
337 HE *he = hv_iternext (hv); 788 }
338 789 else if (svt == SVt_PVHV)
339 for (;;) 790 encode_hv (enc, (HV *)sv);
340 { 791 else if (svt == SVt_PVAV)
341 INDENT; 792 encode_av (enc, (AV *)sv);
342 encode_he (enc, he); 793 else if (svt < SVt_PVAV)
343
344 if (!(he = hv_iternext (hv)))
345 break;
346
347 COMMA;
348 }
349
350 NL;
351 }
352 } 794 {
795 STRLEN len = 0;
796 char *pv = svt ? SvPV (sv, len) : 0;
353 797
354 --enc->indent; INDENT; encode_ch (enc, '}'); 798 if (len == 1 && *pv == '1')
799 encode_str (enc, "true", 4, 0);
800 else if (len == 1 && *pv == '0')
801 encode_str (enc, "false", 5, 0);
802 else if (enc->json.flags & F_ALLOW_UNKNOWN)
803 encode_str (enc, "null", 4, 0);
804 else
805 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
806 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
807 }
808 else if (enc->json.flags & F_ALLOW_UNKNOWN)
809 encode_str (enc, "null", 4, 0);
810 else
811 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
812 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355} 813}
356 814
357static void 815static void
358encode_sv (enc_t *enc, SV *sv) 816encode_sv (enc_t *enc, SV *sv)
359{ 817{
… …
367 encode_str (enc, str, len, SvUTF8 (sv)); 825 encode_str (enc, str, len, SvUTF8 (sv));
368 encode_ch (enc, '"'); 826 encode_ch (enc, '"');
369 } 827 }
370 else if (SvNOKp (sv)) 828 else if (SvNOKp (sv))
371 { 829 {
830 // trust that perl will do the right thing w.r.t. JSON syntax.
372 need (enc, NV_DIG + 32); 831 need (enc, NV_DIG + 32);
373 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 832 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374 enc->cur += strlen (enc->cur); 833 enc->cur += strlen (enc->cur);
375 } 834 }
376 else if (SvIOKp (sv)) 835 else if (SvIOKp (sv))
377 { 836 {
837 // we assume we can always read an IV as a UV and vice versa
838 // we assume two's complement
839 // we assume no aliasing issues in the union
840 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
841 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
842 {
843 // optimise the "small number case"
844 // code will likely be branchless and use only a single multiplication
845 // works for numbers up to 59074
846 I32 i = SvIVX (sv);
847 U32 u;
848 char digit, nz = 0;
849
378 need (enc, 64); 850 need (enc, 6);
851
852 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
853 u = i < 0 ? -i : i;
854
855 // convert to 4.28 fixed-point representation
856 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
857
858 // now output digit by digit, each time masking out the integer part
859 // and multiplying by 5 while moving the decimal point one to the right,
860 // resulting in a net multiplication by 10.
861 // we always write the digit to memory but conditionally increment
862 // the pointer, to enable the use of conditional move instructions.
863 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
864 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
865 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
866 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
867 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
868 }
869 else
870 {
871 // large integer, use the (rather slow) snprintf way.
872 need (enc, IVUV_MAXCHARS);
379 enc->cur += 873 enc->cur +=
380 SvIsUV(sv) 874 SvIsUV(sv)
381 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 875 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 876 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
877 }
383 } 878 }
384 else if (SvROK (sv)) 879 else if (SvROK (sv))
385 { 880 encode_rv (enc, SvRV (sv));
386 if (!--enc->max_recurse) 881 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); 882 encode_str (enc, "null", 4, 0);
402 else 883 else
403 croak ("encountered perl type that JSON cannot handle"); 884 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
885 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
886}
887
888static int
889json_scalar (SV *scalar)
890{
891 return 0;//D
892 if (!SvROK (scalar))
893 return 1;
404} 894}
405 895
406static SV * 896static SV *
407encode_json (SV *scalar, UV flags) 897encode_json (SV *scalar, JSON *json)
408{ 898{
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; 899 enc_t enc;
413 enc.flags = flags; 900
901 if (!(json->flags & F_ALLOW_NONREF) && json_scalar (scalar))
902 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
903
904 enc.json = *json;
414 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 905 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415 enc.cur = SvPVX (enc.sv); 906 enc.cur = SvPVX (enc.sv);
416 enc.end = SvEND (enc.sv); 907 enc.end = SvEND (enc.sv);
417 enc.max_recurse = 0;
418 enc.indent = 0; 908 enc.indent = 0;
909 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
910 : enc.json.flags & F_LATIN1 ? 0x000100UL
911 : 0x110000UL;
419 912
420 SvPOK_only (enc.sv); 913 SvPOK_only (enc.sv);
421 encode_sv (&enc, scalar); 914 encode_sv (&enc, scalar);
915 encode_nl (&enc);
422 916
917 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
918 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
919
423 if (!(flags & (F_ASCII | F_UTF8))) 920 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424 SvUTF8_on (enc.sv); 921 SvUTF8_on (enc.sv);
425 922
426 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 923 if (enc.json.flags & F_SHRINK)
924 shrink (enc.sv);
925
427 return enc.sv; 926 return enc.sv;
428} 927}
429 928
430///////////////////////////////////////////////////////////////////////////// 929/////////////////////////////////////////////////////////////////////////////
930// decoder
431 931
432#define WS \ 932// structure used for decoding JSON
933typedef struct
934{
935 char *cur; // current parser pointer
936 char *end; // end of input string
937 const char *err; // parse error, if != 0
938 JSON json;
939 U32 depth; // recursion depth
940 U32 maxdepth; // recursion depth limit
941} dec_t;
942
943INLINE void
944decode_comment (dec_t *dec)
945{
946 // only '#'-style comments allowed a.t.m.
947
948 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
949 ++dec->cur;
950}
951
952INLINE void
953decode_ws (dec_t *dec)
954{
433 for (;;) \ 955 for (;;)
434 { \ 956 {
435 char ch = *dec->cur; \ 957 char ch = *dec->cur;
958
436 if (ch > 0x20 \ 959 if (ch > 0x20)
960 {
961 if (expect_false (ch == '#'))
962 {
963 if (dec->json.flags & F_RELAXED)
964 decode_comment (dec);
965 else
966 break;
967 }
968 else
969 break;
970 }
437 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 971 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438 break; \ 972 break; // parse error, but let higher level handle it, gives better error messages
973
439 ++dec->cur; \ 974 ++dec->cur;
440 } 975 }
976}
441 977
442#define ERR(reason) SB dec->err = reason; goto fail; SE 978#define ERR(reason) SB dec->err = reason; goto fail; SE
979
443#define EXPECT_CH(ch) SB \ 980#define EXPECT_CH(ch) SB \
444 if (*dec->cur != ch) \ 981 if (*dec->cur != ch) \
445 ERR (# ch " expected"); \ 982 ERR (# ch " expected"); \
446 ++dec->cur; \ 983 ++dec->cur; \
447 SE 984 SE
448 985
986#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
987#define DEC_DEC_DEPTH --dec->depth
988
449static SV *decode_sv (dec_t *dec); 989static SV *decode_sv (dec_t *dec);
450 990
451static signed char decode_hexdigit[256]; 991static signed char decode_hexdigit[256];
452 992
453static UV 993static UV
454decode_4hex (dec_t *dec) 994decode_4hex (dec_t *dec)
455{ 995{
456 signed char d1, d2, d3, d4; 996 signed char d1, d2, d3, d4;
997 unsigned char *cur = (unsigned char *)dec->cur;
457 998
458 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 999 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"); 1000 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]]; 1001 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"); 1002 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 1003
467 dec->cur += 4; 1004 dec->cur += 4;
468 1005
469 return ((UV)d1) << 12 1006 return ((UV)d1) << 12
470 | ((UV)d2) << 8 1007 | ((UV)d2) << 8
… …
473 1010
474fail: 1011fail:
475 return (UV)-1; 1012 return (UV)-1;
476} 1013}
477 1014
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 * 1015static SV *
494decode_str (dec_t *dec) 1016decode_str (dec_t *dec)
495{ 1017{
496 SV *sv = NEWSV (0,2); 1018 SV *sv = 0;
497 int utf8 = 0; 1019 int utf8 = 0;
498 char *cur = SvPVX (sv); 1020 char *dec_cur = dec->cur;
499 char *end = SvEND (sv);
500 1021
501 for (;;) 1022 do
502 { 1023 {
503 unsigned char ch = *(unsigned char *)dec->cur; 1024 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
1025 char *cur = buf;
504 1026
505 if (ch == '"') 1027 do
506 break;
507 else if (ch == '\\')
508 { 1028 {
509 switch (*++dec->cur) 1029 unsigned char ch = *(unsigned char *)dec_cur++;
1030
1031 if (expect_false (ch == '"'))
510 { 1032 {
511 case '\\': 1033 --dec_cur;
512 case '/': 1034 break;
513 case '"': APPEND_CH (*dec->cur++); break; 1035 }
514 1036 else if (expect_false (ch == '\\'))
515 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 1037 {
516 case 't': APPEND_CH ('\011'); ++dec->cur; break; 1038 switch (*dec_cur)
517 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521 case 'u':
522 { 1039 {
523 UV lo, hi; 1040 case '\\':
524 ++dec->cur; 1041 case '/':
1042 case '"': *cur++ = *dec_cur++; break;
525 1043
526 hi = decode_4hex (dec); 1044 case 'b': ++dec_cur; *cur++ = '\010'; break;
527 if (hi == (UV)-1) 1045 case 't': ++dec_cur; *cur++ = '\011'; break;
528 goto fail; 1046 case 'n': ++dec_cur; *cur++ = '\012'; break;
1047 case 'f': ++dec_cur; *cur++ = '\014'; break;
1048 case 'r': ++dec_cur; *cur++ = '\015'; break;
529 1049
530 // possibly a surrogate pair 1050 case 'u':
531 if (hi >= 0xd800 && hi < 0xdc00)
532 { 1051 {
533 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 1052 UV lo, hi;
534 ERR ("illegal surrogate character"); 1053 ++dec_cur;
535 1054
536 dec->cur += 2; 1055 dec->cur = dec_cur;
537
538 lo = decode_4hex (dec); 1056 hi = decode_4hex (dec);
1057 dec_cur = dec->cur;
539 if (lo == (UV)-1) 1058 if (hi == (UV)-1)
540 goto fail; 1059 goto fail;
541 1060
1061 // possibly a surrogate pair
1062 if (hi >= 0xd800)
1063 if (hi < 0xdc00)
1064 {
1065 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
1066 ERR ("missing low surrogate character in surrogate pair");
1067
1068 dec_cur += 2;
1069
1070 dec->cur = dec_cur;
1071 lo = decode_4hex (dec);
1072 dec_cur = dec->cur;
1073 if (lo == (UV)-1)
1074 goto fail;
1075
542 if (lo < 0xdc00 || lo >= 0xe000) 1076 if (lo < 0xdc00 || lo >= 0xe000)
543 ERR ("surrogate pair expected"); 1077 ERR ("surrogate pair expected");
544 1078
545 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 1079 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
1080 }
1081 else if (hi < 0xe000)
1082 ERR ("missing high surrogate character in surrogate pair");
1083
1084 if (hi >= 0x80)
1085 {
1086 utf8 = 1;
1087
1088 cur = encode_utf8 (cur, hi);
1089 }
1090 else
1091 *cur++ = hi;
546 } 1092 }
547 else if (lo >= 0xdc00 && lo < 0xe000)
548 ERR ("illegal surrogate character");
549
550 if (hi >= 0x80)
551 { 1093 break;
552 utf8 = 1;
553 1094
554 APPEND_GROW (4); // at most 4 bytes for 21 bits
555 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556 }
557 else 1095 default:
558 APPEND_CH (hi); 1096 --dec_cur;
1097 ERR ("illegal backslash escape sequence in string");
559 } 1098 }
1099 }
1100 else if (expect_true (ch >= 0x20 && ch < 0x80))
1101 *cur++ = ch;
1102 else if (ch >= 0x80)
1103 {
1104 STRLEN clen;
1105
1106 --dec_cur;
1107
1108 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
1109 if (clen == (STRLEN)-1)
1110 ERR ("malformed UTF-8 character in JSON string");
1111
1112 do
1113 *cur++ = *dec_cur++;
1114 while (--clen);
1115
1116 utf8 = 1;
1117 }
1118 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1119 *cur++ = ch;
1120 else
1121 {
1122 --dec_cur;
1123
1124 if (!ch)
1125 ERR ("unexpected end of string while parsing JSON string");
560 break; 1126 else
1127 ERR ("invalid character encountered while parsing JSON string");
561 } 1128 }
562 } 1129 }
563 else if (ch >= 0x20 && ch <= 0x7f) 1130 while (cur < buf + SHORT_STRING_LEN);
564 APPEND_CH (*dec->cur++); 1131
565 else if (ch >= 0x80) 1132 {
1133 STRLEN len = cur - buf;
1134
1135 if (sv)
566 { 1136 {
567 STRLEN clen; 1137 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 1138
572 APPEND_GROW (clen); 1139 if (SvLEN (sv) <= cur + len)
573 memcpy (cur, dec->cur, clen); 1140 SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
574 cur += clen; 1141
575 dec->cur += clen; 1142 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
1143 SvCUR_set (sv, SvCUR (sv) + len);
576 } 1144 }
577 else 1145 else
578 ERR ("invalid character encountered"); 1146 sv = newSVpvn (buf, len);
579 } 1147 }
1148 }
1149 while (*dec_cur != '"');
580 1150
581 ++dec->cur; 1151 ++dec_cur;
582 1152
583 SvCUR_set (sv, cur - SvPVX (sv)); 1153 if (sv)
584 1154 {
585 SvPOK_only (sv); 1155 SvPOK_only (sv);
586 *SvEND (sv) = 0; 1156 *SvEND (sv) = 0;
587 1157
588 if (utf8) 1158 if (utf8)
589 SvUTF8_on (sv); 1159 SvUTF8_on (sv);
1160 }
1161 else
1162 sv = newSVpvn ("", 0);
590 1163
1164 dec->cur = dec_cur;
591 return sv; 1165 return sv;
592 1166
593fail: 1167fail:
594 SvREFCNT_dec (sv); 1168 dec->cur = dec_cur;
595 return 0; 1169 return 0;
596} 1170}
597 1171
598static SV * 1172static SV *
599decode_num (dec_t *dec) 1173decode_num (dec_t *dec)
… …
609 { 1183 {
610 ++dec->cur; 1184 ++dec->cur;
611 if (*dec->cur >= '0' && *dec->cur <= '9') 1185 if (*dec->cur >= '0' && *dec->cur <= '9')
612 ERR ("malformed number (leading zero must not be followed by another digit)"); 1186 ERR ("malformed number (leading zero must not be followed by another digit)");
613 } 1187 }
614 1188 else if (*dec->cur < '0' || *dec->cur > '9')
615 // int 1189 ERR ("malformed number (no digits after initial minus)");
1190 else
1191 do
1192 {
1193 ++dec->cur;
1194 }
616 while (*dec->cur >= '0' && *dec->cur <= '9') 1195 while (*dec->cur >= '0' && *dec->cur <= '9');
617 ++dec->cur;
618 1196
619 // [frac] 1197 // [frac]
620 if (*dec->cur == '.') 1198 if (*dec->cur == '.')
621 { 1199 {
622 is_nv = 1; 1200 ++dec->cur;
1201
1202 if (*dec->cur < '0' || *dec->cur > '9')
1203 ERR ("malformed number (no digits after decimal point)");
623 1204
624 do 1205 do
625 { 1206 {
626 ++dec->cur; 1207 ++dec->cur;
627 } 1208 }
628 while (*dec->cur >= '0' && *dec->cur <= '9'); 1209 while (*dec->cur >= '0' && *dec->cur <= '9');
1210
1211 is_nv = 1;
629 } 1212 }
630 1213
631 // [exp] 1214 // [exp]
632 if (*dec->cur == 'e' || *dec->cur == 'E') 1215 if (*dec->cur == 'e' || *dec->cur == 'E')
633 { 1216 {
634 is_nv = 1;
635
636 ++dec->cur; 1217 ++dec->cur;
1218
637 if (*dec->cur == '-' || *dec->cur == '+') 1219 if (*dec->cur == '-' || *dec->cur == '+')
638 ++dec->cur; 1220 ++dec->cur;
639 1221
1222 if (*dec->cur < '0' || *dec->cur > '9')
1223 ERR ("malformed number (no digits after exp sign)");
1224
1225 do
1226 {
1227 ++dec->cur;
1228 }
640 while (*dec->cur >= '0' && *dec->cur <= '9') 1229 while (*dec->cur >= '0' && *dec->cur <= '9');
641 ++dec->cur; 1230
1231 is_nv = 1;
642 } 1232 }
643 1233
644 if (!is_nv) 1234 if (!is_nv)
645 { 1235 {
646 UV uv; 1236 int len = dec->cur - start;
647 int numtype = grok_number (start, dec->cur - start, &uv); 1237
648 if (numtype & IS_NUMBER_IN_UV) 1238 // special case the rather common 1..5-digit-int case
649 if (numtype & IS_NUMBER_NEG) 1239 if (*start == '-')
1240 switch (len)
650 { 1241 {
651 if (uv < (UV)IV_MIN) 1242 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
652 return newSViv (-(IV)uv); 1243 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1244 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1245 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1246 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
653 } 1247 }
1248 else
1249 switch (len)
1250 {
1251 case 1: return newSViv ( start [0] - '0' * 1);
1252 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1253 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1254 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1255 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1256 }
1257
1258 {
1259 UV uv;
1260 int numtype = grok_number (start, len, &uv);
1261 if (numtype & IS_NUMBER_IN_UV)
1262 if (numtype & IS_NUMBER_NEG)
1263 {
1264 if (uv < (UV)IV_MIN)
1265 return newSViv (-(IV)uv);
1266 }
654 else 1267 else
655 return newSVuv (uv); 1268 return newSVuv (uv);
656 } 1269 }
657 1270
1271 len -= *start == '-' ? 1 : 0;
1272
1273 // does not fit into IV or UV, try NV
1274 if (len <= NV_DIG)
1275 // fits into NV without loss of precision
1276 return newSVnv (json_atof (start));
1277
1278 // everything else fails, convert it to a string
1279 return newSVpvn (start, dec->cur - start);
1280 }
1281
1282 // loss of precision here
658 return newSVnv (Atof (start)); 1283 return newSVnv (json_atof (start));
659 1284
660fail: 1285fail:
661 return 0; 1286 return 0;
662} 1287}
663 1288
664static SV * 1289static SV *
665decode_av (dec_t *dec) 1290decode_av (dec_t *dec)
666{ 1291{
667 AV *av = newAV (); 1292 AV *av = newAV ();
668 1293
1294 DEC_INC_DEPTH;
1295 decode_ws (dec);
1296
1297 if (*dec->cur == ']')
1298 ++dec->cur;
1299 else
669 for (;;) 1300 for (;;)
670 { 1301 {
671 SV *value; 1302 SV *value;
672 1303
673 value = decode_sv (dec); 1304 value = decode_sv (dec);
674 if (!value) 1305 if (!value)
675 goto fail; 1306 goto fail;
676 1307
677 av_push (av, value); 1308 av_push (av, value);
678 1309
679 WS; 1310 decode_ws (dec);
680 1311
681 if (*dec->cur == ']') 1312 if (*dec->cur == ']')
682 { 1313 {
683 ++dec->cur; 1314 ++dec->cur;
684 break; 1315 break;
685 } 1316 }
686 1317
687 if (*dec->cur != ',') 1318 if (*dec->cur != ',')
688 ERR (", or ] expected while parsing array"); 1319 ERR (", or ] expected while parsing array");
689 1320
690 ++dec->cur; 1321 ++dec->cur;
1322
1323 decode_ws (dec);
1324
1325 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1326 {
1327 ++dec->cur;
1328 break;
1329 }
691 } 1330 }
692 1331
1332 DEC_DEC_DEPTH;
693 return newRV_noinc ((SV *)av); 1333 return newRV_noinc ((SV *)av);
694 1334
695fail: 1335fail:
696 SvREFCNT_dec (av); 1336 SvREFCNT_dec (av);
1337 DEC_DEC_DEPTH;
697 return 0; 1338 return 0;
698} 1339}
699 1340
700static SV * 1341static SV *
701decode_hv (dec_t *dec) 1342decode_hv (dec_t *dec)
702{ 1343{
1344 SV *sv;
703 HV *hv = newHV (); 1345 HV *hv = newHV ();
704 1346
1347 DEC_INC_DEPTH;
1348 decode_ws (dec);
1349
1350 if (*dec->cur == '}')
1351 ++dec->cur;
1352 else
705 for (;;) 1353 for (;;)
706 { 1354 {
707 SV *key, *value;
708
709 WS; EXPECT_CH ('"'); 1355 EXPECT_CH ('"');
710 1356
711 key = decode_str (dec); 1357 // heuristic: assume that
712 if (!key) 1358 // a) decode_str + hv_store_ent are abysmally slow.
713 goto fail; 1359 // b) most hash keys are short, simple ascii text.
714 1360 // => try to "fast-match" such strings to avoid
715 WS; EXPECT_CH (':'); 1361 // the overhead of decode_str + hv_store_ent.
716
717 value = decode_sv (dec);
718 if (!value)
719 { 1362 {
1363 SV *value;
1364 char *p = dec->cur;
1365 char *e = p + 24; // only try up to 24 bytes
1366
1367 for (;;)
1368 {
1369 // the >= 0x80 is false on most architectures
1370 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1371 {
1372 // slow path, back up and use decode_str
1373 SV *key = decode_str (dec);
1374 if (!key)
1375 goto fail;
1376
1377 decode_ws (dec); EXPECT_CH (':');
1378
1379 decode_ws (dec);
1380 value = decode_sv (dec);
1381 if (!value)
1382 {
1383 SvREFCNT_dec (key);
1384 goto fail;
1385 }
1386
1387 hv_store_ent (hv, key, value, 0);
720 SvREFCNT_dec (key); 1388 SvREFCNT_dec (key);
1389
1390 break;
1391 }
1392 else if (*p == '"')
1393 {
1394 // fast path, got a simple key
1395 char *key = dec->cur;
1396 int len = p - key;
1397 dec->cur = p + 1;
1398
1399 decode_ws (dec); EXPECT_CH (':');
1400
1401 decode_ws (dec);
1402 value = decode_sv (dec);
1403 if (!value)
721 goto fail; 1404 goto fail;
1405
1406 hv_store (hv, key, len, value, 0);
1407
1408 break;
1409 }
1410
1411 ++p;
1412 }
722 } 1413 }
723 1414
724 //TODO: optimise 1415 decode_ws (dec);
725 hv_store_ent (hv, key, value, 0);
726 1416
727 WS;
728
729 if (*dec->cur == '}') 1417 if (*dec->cur == '}')
1418 {
1419 ++dec->cur;
1420 break;
1421 }
1422
1423 if (*dec->cur != ',')
1424 ERR (", or } expected while parsing object/hash");
1425
1426 ++dec->cur;
1427
1428 decode_ws (dec);
1429
1430 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1431 {
1432 ++dec->cur;
1433 break;
1434 }
1435 }
1436
1437 DEC_DEC_DEPTH;
1438 sv = newRV_noinc ((SV *)hv);
1439
1440 // check filter callbacks
1441 if (dec->json.flags & F_HOOK)
1442 {
1443 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
730 { 1444 {
731 ++dec->cur; 1445 HE *cb, *he;
732 break; 1446
1447 hv_iterinit (hv);
1448 he = hv_iternext (hv);
1449 hv_iterinit (hv);
1450
1451 // the next line creates a mortal sv each time it's called.
1452 // might want to optimise this for common cases.
1453 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1454
1455 if (cb)
1456 {
1457 dSP;
1458 int count;
1459
1460 ENTER; SAVETMPS;
1461 SAVESTACK_POS ();
1462 PUSHMARK (SP);
1463 XPUSHs (HeVAL (he));
1464 sv_2mortal (sv);
1465
1466 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1467
1468 if (count == 1)
1469 {
1470 sv = newSVsv (POPs);
1471 FREETMPS; LEAVE;
1472 return sv;
1473 }
1474
1475 SvREFCNT_inc (sv);
1476 FREETMPS; LEAVE;
1477 }
733 } 1478 }
734 1479
735 if (*dec->cur != ',') 1480 if (dec->json.cb_object)
736 ERR (", or } expected while parsing object/hash"); 1481 {
1482 dSP;
1483 int count;
737 1484
738 ++dec->cur; 1485 ENTER; SAVETMPS;
739 } 1486 SAVESTACK_POS ();
1487 PUSHMARK (SP);
1488 XPUSHs (sv_2mortal (sv));
740 1489
741 return newRV_noinc ((SV *)hv); 1490 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1491
1492 if (count == 1)
1493 {
1494 sv = newSVsv (POPs);
1495 FREETMPS; LEAVE;
1496 return sv;
1497 }
1498
1499 SvREFCNT_inc (sv);
1500 FREETMPS; LEAVE;
1501 }
1502 }
1503
1504 return sv;
742 1505
743fail: 1506fail:
744 SvREFCNT_dec (hv); 1507 SvREFCNT_dec (hv);
1508 DEC_DEC_DEPTH;
1509 return 0;
1510}
1511
1512static SV *
1513decode_tag (dec_t *dec)
1514{
1515 SV *tag = 0;
1516 SV *val = 0;
1517
1518 if (!(dec->json.flags & F_ALLOW_TAGS))
1519 ERR ("malformed JSON string, neither array, object, number, string or atom");
1520
1521 ++dec->cur;
1522
1523 decode_ws (dec);
1524
1525 tag = decode_sv (dec);
1526 if (!tag)
1527 goto fail;
1528
1529 if (!SvPOK (tag))
1530 ERR ("malformed JSON string, (tag) must be a string");
1531
1532 decode_ws (dec);
1533
1534 if (*dec->cur != ')')
1535 ERR (") expected after tag");
1536
1537 ++dec->cur;
1538
1539 decode_ws (dec);
1540
1541 val = decode_sv (dec);
1542 if (!val)
1543 goto fail;
1544
1545 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1546 ERR ("malformed JSON string, tag value must be an array");
1547
1548 {
1549 AV *av = (AV *)SvRV (val);
1550 int i, len = av_len (av) + 1;
1551 HV *stash = gv_stashsv (tag, 0);
1552 SV *sv;
1553
1554 if (!stash)
1555 ERR ("cannot decode perl-object (package does not exist)");
1556
1557 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1558
1559 if (!method)
1560 ERR ("cannot decode perl-object (package does not have a THAW method)");
1561
1562 dSP;
1563
1564 ENTER; SAVETMPS;
1565 PUSHMARK (SP);
1566 EXTEND (SP, len + 2);
1567 // we re-bless the reference to get overload and other niceties right
1568 PUSHs (tag);
1569 PUSHs (sv_json);
1570
1571 for (i = 0; i < len; ++i)
1572 PUSHs (*av_fetch (av, i, 1));
1573
1574 PUTBACK;
1575 call_sv ((SV *)GvCV (method), G_SCALAR);
1576 SPAGAIN;
1577
1578 SvREFCNT_dec (tag);
1579 SvREFCNT_dec (val);
1580 sv = SvREFCNT_inc (POPs);
1581
1582 PUTBACK;
1583
1584 FREETMPS; LEAVE;
1585
1586 return sv;
1587 }
1588
1589fail:
1590 SvREFCNT_dec (tag);
1591 SvREFCNT_dec (val);
745 return 0; 1592 return 0;
746} 1593}
747 1594
748static SV * 1595static SV *
749decode_sv (dec_t *dec) 1596decode_sv (dec_t *dec)
750{ 1597{
751 WS; 1598 // the beauty of JSON: you need exactly one character lookahead
1599 // to parse everything.
752 switch (*dec->cur) 1600 switch (*dec->cur)
753 { 1601 {
754 case '"': ++dec->cur; return decode_str (dec); 1602 case '"': ++dec->cur; return decode_str (dec);
755 case '[': ++dec->cur; return decode_av (dec); 1603 case '[': ++dec->cur; return decode_av (dec);
756 case '{': ++dec->cur; return decode_hv (dec); 1604 case '{': ++dec->cur; return decode_hv (dec);
1605 case '(': return decode_tag (dec);
757 1606
758 case '-': 1607 case '-':
759 case '0': case '1': case '2': case '3': case '4': 1608 case '0': case '1': case '2': case '3': case '4':
760 case '5': case '6': case '7': case '8': case '9': 1609 case '5': case '6': case '7': case '8': case '9':
761 return decode_num (dec); 1610 return decode_num (dec);
762 1611
763 case 't': 1612 case 't':
764 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1613 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765 { 1614 {
766 dec->cur += 4; 1615 dec->cur += 4;
1616#if JSON_SLOW
1617 types_true = get_bool ("Types::Serialiser::true");
1618#endif
767 return newSViv (1); 1619 return newSVsv (types_true);
768 } 1620 }
769 else 1621 else
770 ERR ("'true' expected"); 1622 ERR ("'true' expected");
771 1623
772 break; 1624 break;
773 1625
774 case 'f': 1626 case 'f':
775 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1627 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776 { 1628 {
777 dec->cur += 5; 1629 dec->cur += 5;
1630#if JSON_SLOW
1631 types_false = get_bool ("Types::Serialiser::false");
1632#endif
778 return newSViv (0); 1633 return newSVsv (types_false);
779 } 1634 }
780 else 1635 else
781 ERR ("'false' expected"); 1636 ERR ("'false' expected");
782 1637
783 break; 1638 break;
784 1639
785 case 'n': 1640 case 'n':
786 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1641 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787 { 1642 {
788 dec->cur += 4; 1643 dec->cur += 4;
789 return newSViv (1); 1644 return newSVsv (&PL_sv_undef);
790 } 1645 }
791 else 1646 else
792 ERR ("'null' expected"); 1647 ERR ("'null' expected");
793 1648
794 break; 1649 break;
795 1650
796 default: 1651 default:
797 ERR ("malformed json string"); 1652 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
798 break; 1653 break;
799 } 1654 }
800 1655
801fail: 1656fail:
802 return 0; 1657 return 0;
803} 1658}
804 1659
805static SV * 1660static SV *
806decode_json (SV *string, UV flags) 1661decode_json (SV *string, JSON *json, char **offset_return)
807{ 1662{
1663 dec_t dec;
808 SV *sv; 1664 SV *sv;
809 1665
810 if (!(flags & F_UTF8)) 1666 /* work around bugs in 5.10 where manipulating magic values
1667 * makes perl ignore the magic in subsequent accesses.
1668 * also make a copy of non-PV values, to get them into a clean
1669 * state (SvPV should do that, but it's buggy, see below).
1670 */
1671 /*SvGETMAGIC (string);*/
1672 if (SvMAGICAL (string) || !SvPOK (string))
1673 string = sv_2mortal (newSVsv (string));
1674
1675 SvUPGRADE (string, SVt_PV);
1676
1677 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1678 * assertion with -DDEBUGGING, although SvCUR is documented to
1679 * return the xpv_cur field which certainly exists after upgrading.
1680 * according to nicholas clark, calling SvPOK fixes this.
1681 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1682 * and hope for the best.
1683 * Damnit, SvPV_nolen still trips over yet another assertion. This
1684 * assertion business is seriously broken, try yet another workaround
1685 * for the broken -DDEBUGGING.
1686 */
1687 {
1688#ifdef DEBUGGING
1689 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1690#else
1691 STRLEN offset = SvCUR (string);
1692#endif
1693
1694 if (offset > json->max_size && json->max_size)
1695 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1696 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1697 }
1698
1699 if (DECODE_WANTS_OCTETS (json))
1700 sv_utf8_downgrade (string, 0);
1701 else
811 sv_utf8_upgrade (string); 1702 sv_utf8_upgrade (string);
812 1703
813 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1704 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814 1705
815 dec_t dec; 1706 dec.json = *json;
816 dec.flags = flags;
817 dec.cur = SvPVX (string); 1707 dec.cur = SvPVX (string);
818 dec.end = SvEND (string); 1708 dec.end = SvEND (string);
819 dec.err = 0; 1709 dec.err = 0;
1710 dec.depth = 0;
820 1711
821 *dec.end = 1; // invalid anywhere 1712 if (dec.json.cb_object || dec.json.cb_sk_object)
1713 dec.json.flags |= F_HOOK;
1714
1715 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1716
1717 decode_ws (&dec);
822 sv = decode_sv (&dec); 1718 sv = decode_sv (&dec);
823 *dec.end = 0; 1719
1720 if (offset_return)
1721 *offset_return = dec.cur;
1722
1723 if (!(offset_return || !sv))
1724 {
1725 // check for trailing garbage
1726 decode_ws (&dec);
1727
1728 if (*dec.cur)
1729 {
1730 dec.err = "garbage after JSON object";
1731 SvREFCNT_dec (sv);
1732 sv = 0;
1733 }
1734 }
824 1735
825 if (!sv) 1736 if (!sv)
826 { 1737 {
827 IV offset = utf8_distance (dec.cur, SvPVX (string));
828 SV *uni = sv_newmortal (); 1738 SV *uni = sv_newmortal ();
829 1739
1740 // horrible hack to silence warning inside pv_uni_display
1741 COP cop = *PL_curcop;
1742 cop.cop_warnings = pWARN_NONE;
1743 ENTER;
1744 SAVEVPTR (PL_curcop);
1745 PL_curcop = &cop;
830 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1746 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1747 LEAVE;
1748
831 croak ("%s, at character %d (%s)", 1749 croak ("%s, at character offset %d (before \"%s\")",
832 dec.err, 1750 dec.err,
833 (int)offset, 1751 (int)ptr_to_index (string, dec.cur),
834 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1752 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835 } 1753 }
836 1754
837 sv = sv_2mortal (sv); 1755 sv = sv_2mortal (sv);
838 1756
839 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1757 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)"); 1758 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841 1759
842 return sv; 1760 return sv;
843} 1761}
844 1762
1763/////////////////////////////////////////////////////////////////////////////
1764// incremental parser
1765
1766static void
1767incr_parse (JSON *self)
1768{
1769 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1770
1771 // the state machine here is a bit convoluted and could be simplified a lot
1772 // but this would make it slower, so...
1773
1774 for (;;)
1775 {
1776 //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
1777 switch (self->incr_mode)
1778 {
1779 // only used for initial whitespace skipping
1780 case INCR_M_WS:
1781 for (;;)
1782 {
1783 if (*p > 0x20)
1784 {
1785 if (*p == '#')
1786 {
1787 self->incr_mode = INCR_M_C0;
1788 goto incr_m_c;
1789 }
1790 else
1791 {
1792 self->incr_mode = INCR_M_JSON;
1793 goto incr_m_json;
1794 }
1795 }
1796 else if (!*p)
1797 goto interrupt;
1798
1799 ++p;
1800 }
1801
1802 // skip a single char inside a string (for \\-processing)
1803 case INCR_M_BS:
1804 if (!*p)
1805 goto interrupt;
1806
1807 ++p;
1808 self->incr_mode = INCR_M_STR;
1809 goto incr_m_str;
1810
1811 // inside #-style comments
1812 case INCR_M_C0:
1813 case INCR_M_C1:
1814 incr_m_c:
1815 for (;;)
1816 {
1817 if (*p == '\n')
1818 {
1819 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1820 break;
1821 }
1822 else if (!*p)
1823 goto interrupt;
1824
1825 ++p;
1826 }
1827
1828 break;
1829
1830 // inside a string
1831 case INCR_M_STR:
1832 incr_m_str:
1833 for (;;)
1834 {
1835 if (*p == '"')
1836 {
1837 ++p;
1838 self->incr_mode = INCR_M_JSON;
1839
1840 if (!self->incr_nest)
1841 goto interrupt;
1842
1843 goto incr_m_json;
1844 }
1845 else if (*p == '\\')
1846 {
1847 ++p; // "virtually" consumes character after \
1848
1849 if (!*p) // if at end of string we have to switch modes
1850 {
1851 self->incr_mode = INCR_M_BS;
1852 goto interrupt;
1853 }
1854 }
1855 else if (!*p)
1856 goto interrupt;
1857
1858 ++p;
1859 }
1860
1861 // after initial ws, outside string
1862 case INCR_M_JSON:
1863 incr_m_json:
1864 for (;;)
1865 {
1866 switch (*p++)
1867 {
1868 case 0:
1869 --p;
1870 goto interrupt;
1871
1872 case 0x09:
1873 case 0x0a:
1874 case 0x0d:
1875 case 0x20:
1876 if (!self->incr_nest)
1877 {
1878 --p; // do not eat the whitespace, let the next round do it
1879 goto interrupt;
1880 }
1881 break;
1882
1883 case '"':
1884 self->incr_mode = INCR_M_STR;
1885 goto incr_m_str;
1886
1887 case '[':
1888 case '{':
1889 case '(':
1890 if (++self->incr_nest > self->max_depth)
1891 croak (ERR_NESTING_EXCEEDED);
1892 break;
1893
1894 case ']':
1895 case '}':
1896 if (--self->incr_nest <= 0)
1897 goto interrupt;
1898 break;
1899
1900 case ')':
1901 --self->incr_nest;
1902 break;
1903
1904 case '#':
1905 self->incr_mode = INCR_M_C1;
1906 goto incr_m_c;
1907 }
1908 }
1909 }
1910
1911 modechange:
1912 ;
1913 }
1914
1915interrupt:
1916 self->incr_pos = p - SvPVX (self->incr_text);
1917 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1918 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1919}
1920
1921/////////////////////////////////////////////////////////////////////////////
1922// XS interface functions
1923
845MODULE = JSON::XS PACKAGE = JSON::XS 1924MODULE = JSON::XS PACKAGE = JSON::XS
846 1925
847BOOT: 1926BOOT:
848{ 1927{
849 int i; 1928 int i;
850 1929
851 memset (decode_hexdigit, 0xff, 256);
852 for (i = 10; i--; ) 1930 for (i = 0; i < 256; ++i)
853 decode_hexdigit ['0' + i] = i; 1931 decode_hexdigit [i] =
1932 i >= '0' && i <= '9' ? i - '0'
1933 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1934 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1935 : -1;
854 1936
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); 1937 json_stash = gv_stashpv ("JSON::XS" , 1);
1938 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1939
1940 types_true = get_bool ("Types::Serialiser::true");
1941 types_false = get_bool ("Types::Serialiser::false");
1942
1943 sv_json = newSVpv ("JSON", 0);
1944 SvREADONLY_on (sv_json);
1945
1946 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
862} 1947}
863 1948
864PROTOTYPES: DISABLE 1949PROTOTYPES: DISABLE
865 1950
866SV *new (char *dummy) 1951void CLONE (...)
867 CODE: 1952 CODE:
868 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1953 json_stash = 0;
1954 types_boolean_stash = 0;
1955
1956void new (char *klass)
1957 PPCODE:
1958{
1959 SV *pv = NEWSV (0, sizeof (JSON));
1960 SvPOK_only (pv);
1961 json_init ((JSON *)SvPVX (pv));
1962 XPUSHs (sv_2mortal (sv_bless (
1963 newRV_noinc (pv),
1964 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1965 )));
1966}
1967
1968void ascii (JSON *self, int enable = 1)
1969 ALIAS:
1970 ascii = F_ASCII
1971 latin1 = F_LATIN1
1972 utf8 = F_UTF8
1973 indent = F_INDENT
1974 canonical = F_CANONICAL
1975 space_before = F_SPACE_BEFORE
1976 space_after = F_SPACE_AFTER
1977 pretty = F_PRETTY
1978 allow_nonref = F_ALLOW_NONREF
1979 shrink = F_SHRINK
1980 allow_blessed = F_ALLOW_BLESSED
1981 convert_blessed = F_CONV_BLESSED
1982 relaxed = F_RELAXED
1983 allow_unknown = F_ALLOW_UNKNOWN
1984 allow_tags = F_ALLOW_TAGS
1985 PPCODE:
1986{
1987 if (enable)
1988 self->flags |= ix;
1989 else
1990 self->flags &= ~ix;
1991
1992 XPUSHs (ST (0));
1993}
1994
1995void get_ascii (JSON *self)
1996 ALIAS:
1997 get_ascii = F_ASCII
1998 get_latin1 = F_LATIN1
1999 get_utf8 = F_UTF8
2000 get_indent = F_INDENT
2001 get_canonical = F_CANONICAL
2002 get_space_before = F_SPACE_BEFORE
2003 get_space_after = F_SPACE_AFTER
2004 get_allow_nonref = F_ALLOW_NONREF
2005 get_shrink = F_SHRINK
2006 get_allow_blessed = F_ALLOW_BLESSED
2007 get_convert_blessed = F_CONV_BLESSED
2008 get_relaxed = F_RELAXED
2009 get_allow_unknown = F_ALLOW_UNKNOWN
2010 get_allow_tags = F_ALLOW_TAGS
2011 PPCODE:
2012 XPUSHs (boolSV (self->flags & ix));
2013
2014void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
2015 PPCODE:
2016 self->max_depth = max_depth;
2017 XPUSHs (ST (0));
2018
2019U32 get_max_depth (JSON *self)
2020 CODE:
2021 RETVAL = self->max_depth;
869 OUTPUT: 2022 OUTPUT:
870 RETVAL 2023 RETVAL
871 2024
872SV *ascii (SV *self, int enable) 2025void max_size (JSON *self, U32 max_size = 0)
873 ALIAS: 2026 PPCODE:
874 ascii = F_ASCII 2027 self->max_size = max_size;
875 utf8 = F_UTF8 2028 XPUSHs (ST (0));
876 indent = F_INDENT 2029
877 canonical = F_CANONICAL 2030int 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: 2031 CODE:
884{ 2032 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: 2033 OUTPUT:
894 RETVAL 2034 RETVAL
895 2035
896void encode (SV *self, SV *scalar) 2036void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
897 PPCODE: 2037 PPCODE:
898 XPUSHs (encode_json (scalar, *SvJSON (self))); 2038{
2039 SvREFCNT_dec (self->cb_object);
2040 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899 2041
900void decode (SV *self, SV *jsonstr) 2042 XPUSHs (ST (0));
2043}
2044
2045void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
901 PPCODE: 2046 PPCODE:
902 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 2047{
2048 if (!self->cb_sk_object)
2049 self->cb_sk_object = newHV ();
2050
2051 if (SvOK (cb))
2052 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
2053 else
2054 {
2055 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
2056
2057 if (!HvKEYS (self->cb_sk_object))
2058 {
2059 SvREFCNT_dec (self->cb_sk_object);
2060 self->cb_sk_object = 0;
2061 }
2062 }
2063
2064 XPUSHs (ST (0));
2065}
2066
2067void encode (JSON *self, SV *scalar)
2068 PPCODE:
2069 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2070 XPUSHs (scalar);
2071
2072void decode (JSON *self, SV *jsonstr)
2073 PPCODE:
2074 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2075 XPUSHs (jsonstr);
2076
2077void decode_prefix (JSON *self, SV *jsonstr)
2078 PPCODE:
2079{
2080 SV *sv;
2081 char *offset;
2082 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
2083 EXTEND (SP, 2);
2084 PUSHs (sv);
2085 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
2086}
2087
2088void incr_parse (JSON *self, SV *jsonstr = 0)
2089 PPCODE:
2090{
2091 if (!self->incr_text)
2092 self->incr_text = newSVpvn ("", 0);
2093
2094 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2095 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2096 if (DECODE_WANTS_OCTETS (self))
2097 {
2098 if (self->incr_pos)
2099 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2100 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2101
2102 sv_utf8_downgrade (self->incr_text, 0);
2103 }
2104 else
2105 {
2106 sv_utf8_upgrade (self->incr_text);
2107
2108 if (self->incr_pos)
2109 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2110 - (U8 *)SvPVX (self->incr_text);
2111 }
2112
2113 // append data, if any
2114 if (jsonstr)
2115 {
2116 /* make sure both strings have same encoding */
2117 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2118 if (SvUTF8 (jsonstr))
2119 sv_utf8_downgrade (jsonstr, 0);
2120 else
2121 sv_utf8_upgrade (jsonstr);
2122
2123 /* and then just blindly append */
2124 {
2125 STRLEN len;
2126 const char *str = SvPV (jsonstr, len);
2127 STRLEN cur = SvCUR (self->incr_text);
2128
2129 if (SvLEN (self->incr_text) <= cur + len)
2130 SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
2131
2132 Move (str, SvEND (self->incr_text), len, char);
2133 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2134 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2135 }
2136 }
2137
2138 if (GIMME_V != G_VOID)
2139 do
2140 {
2141 SV *sv;
2142 char *offset;
2143
2144 if (!INCR_DONE (self))
2145 {
2146 incr_parse (self);
2147
2148 if (self->incr_pos > self->max_size && self->max_size)
2149 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2150 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2151
2152 if (!INCR_DONE (self))
2153 {
2154 // as an optimisation, do not accumulate white space in the incr buffer
2155 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2156 {
2157 self->incr_pos = 0;
2158 SvCUR_set (self->incr_text, 0);
2159 }
2160
2161 break;
2162 }
2163 }
2164
2165 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2166 XPUSHs (sv);
2167
2168 self->incr_pos -= offset - SvPVX (self->incr_text);
2169 self->incr_nest = 0;
2170 self->incr_mode = 0;
2171
2172 sv_chop (self->incr_text, offset);
2173 }
2174 while (GIMME_V == G_ARRAY);
2175}
2176
2177SV *incr_text (JSON *self)
2178 ATTRS: lvalue
2179 CODE:
2180{
2181 if (self->incr_pos)
2182 croak ("incr_text can not be called when the incremental parser already started parsing");
2183
2184 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
2185}
2186 OUTPUT:
2187 RETVAL
2188
2189void incr_skip (JSON *self)
2190 CODE:
2191{
2192 if (self->incr_pos)
2193 {
2194 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2195 self->incr_pos = 0;
2196 self->incr_nest = 0;
2197 self->incr_mode = 0;
2198 }
2199}
2200
2201void incr_reset (JSON *self)
2202 CODE:
2203{
2204 SvREFCNT_dec (self->incr_text);
2205 self->incr_text = 0;
2206 self->incr_pos = 0;
2207 self->incr_nest = 0;
2208 self->incr_mode = 0;
2209}
2210
2211void DESTROY (JSON *self)
2212 CODE:
2213 SvREFCNT_dec (self->cb_sk_object);
2214 SvREFCNT_dec (self->cb_object);
2215 SvREFCNT_dec (self->incr_text);
903 2216
904PROTOTYPES: ENABLE 2217PROTOTYPES: ENABLE
905 2218
906void to_json (SV *scalar) 2219void encode_json (SV *scalar)
907 PPCODE: 2220 PPCODE:
908 XPUSHs (encode_json (scalar, F_UTF8)); 2221{
2222 JSON json;
2223 json_init (&json);
2224 json.flags |= F_UTF8;
2225 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2226 XPUSHs (scalar);
2227}
909 2228
910void from_json (SV *jsonstr) 2229void decode_json (SV *jsonstr)
911 PPCODE: 2230 PPCODE:
912 XPUSHs (decode_json (jsonstr, F_UTF8)); 2231{
2232 JSON json;
2233 json_init (&json);
2234 json.flags |= F_UTF8;
2235 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2236 XPUSHs (jsonstr);
2237}
913 2238

Diff Legend

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