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 1.2 by root, Thu Mar 22 17:28:50 2007 UTC vs.
Revision 1.83 by root, Wed Mar 26 22:54:38 2008 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
4 4
5#include "assert.h" 5#include <assert.h>
6#include "string.h" 6#include <string.h>
7#include "stdlib.h" 7#include <stdlib.h>
8#include <stdio.h>
9#include <limits.h>
10#include <float.h>
8 11
12#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h
14#endif
15
16// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13
20#endif
21
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
23
9#define F_ASCII 0x00000001 24#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL
10#define F_UTF8 0x00000002 26#define F_UTF8 0x00000004UL
11#define F_INDENT 0x00000004 27#define F_INDENT 0x00000008UL
12#define F_CANONICAL 0x00000008 28#define F_CANONICAL 0x00000010UL
13#define F_SPACE_BEFORE 0x00000010 29#define F_SPACE_BEFORE 0x00000020UL
14#define F_SPACE_AFTER 0x00000020 30#define F_SPACE_AFTER 0x00000040UL
15#define F_JSON_RPC 0x00000040 31#define F_ALLOW_NONREF 0x00000100UL
32#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL
36
37#define F_MAXDEPTH 0xf8000000UL
38#define S_MAXDEPTH 27
39#define F_MAXSIZE 0x01f00000UL
40#define S_MAXSIZE 20
41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
42
43#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
44#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
16 45
17#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 46#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
18#define F_DEFAULT 0 47#define F_DEFAULT (9UL << S_MAXDEPTH)
19 48
20#define INIT_SIZE 32 // initial scalar size to be allocated 49#define INIT_SIZE 32 // initial scalar size to be allocated
50#define INDENT_STEP 3 // spaces per indentation level
51
52#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
21 53
22#define SB do { 54#define SB do {
23#define SE } while (0) 55#define SE } while (0)
24 56
25static HV *json_stash; 57#if __GNUC__ >= 3
58# define expect(expr,value) __builtin_expect ((expr), (value))
59# define INLINE static inline
60#else
61# define expect(expr,value) (expr)
62# define INLINE static
63#endif
64
65#define expect_false(expr) expect ((expr) != 0, 0)
66#define expect_true(expr) expect ((expr) != 0, 1)
67
68#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
71
72#ifdef USE_ITHREADS
73# define JSON_SLOW 1
74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
75#else
76# define JSON_SLOW 0
77# define JSON_STASH json_stash
78#endif
79
80static HV *json_stash, *json_boolean_stash; // JSON::XS::
81static SV *json_true, *json_false;
82
83enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash
87 INCR_M_JSON // outside anything, count nesting
88};
89
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
91
92typedef struct {
93 U32 flags;
94 SV *cb_object;
95 HV *cb_sk_object;
96
97 // for the incremental parser
98 SV *incr_text; // the source text so far
99 STRLEN incr_pos; // the current offset into the text
100 int incr_nest; // {[]}-nesting level
101 int incr_mode;
102} JSON;
103
104/////////////////////////////////////////////////////////////////////////////
105// utility functions
106
107INLINE SV *
108get_bool (const char *name)
109{
110 SV *sv = get_sv (name, 1);
111
112 SvREADONLY_on (sv);
113 SvREADONLY_on (SvRV (sv));
114
115 return sv;
116}
117
118INLINE void
119shrink (SV *sv)
120{
121 sv_utf8_downgrade (sv, 1);
122 if (SvLEN (sv) > SvCUR (sv) + 1)
123 {
124#ifdef SvPV_shrink_to_cur
125 SvPV_shrink_to_cur (sv);
126#elif defined (SvPV_renew)
127 SvPV_renew (sv, SvCUR (sv) + 1);
128#endif
129 }
130}
131
132// decode an utf-8 character and return it, or (UV)-1 in
133// case of an error.
134// we special-case "safe" characters from U+80 .. U+7FF,
135// but use the very good perl function to parse anything else.
136// note that we never call this function for a ascii codepoints
137INLINE UV
138decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
139{
140 if (expect_true (len >= 2
141 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
142 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
143 {
144 *clen = 2;
145 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
146 }
147 else
148 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
149}
150
151// likewise for encoding, also never called for ascii codepoints
152// this function takes advantage of this fact, although current gccs
153// seem to optimise the check for >= 0x80 away anyways
154INLINE unsigned char *
155encode_utf8 (unsigned char *s, UV ch)
156{
157 if (expect_false (ch < 0x000080))
158 *s++ = ch;
159 else if (expect_true (ch < 0x000800))
160 *s++ = 0xc0 | ( ch >> 6),
161 *s++ = 0x80 | ( ch & 0x3f);
162 else if ( ch < 0x010000)
163 *s++ = 0xe0 | ( ch >> 12),
164 *s++ = 0x80 | ((ch >> 6) & 0x3f),
165 *s++ = 0x80 | ( ch & 0x3f);
166 else if ( ch < 0x110000)
167 *s++ = 0xf0 | ( ch >> 18),
168 *s++ = 0x80 | ((ch >> 12) & 0x3f),
169 *s++ = 0x80 | ((ch >> 6) & 0x3f),
170 *s++ = 0x80 | ( ch & 0x3f);
171
172 return s;
173}
174
175/////////////////////////////////////////////////////////////////////////////
176// encoder
26 177
27// structure used for encoding JSON 178// structure used for encoding JSON
28typedef struct 179typedef struct
29{ 180{
30 char *cur; 181 char *cur; // SvPVX (sv) + current output position
31 STRLEN len; // SvLEN (sv)
32 char *end; // SvEND (sv) 182 char *end; // SvEND (sv)
33 SV *sv; 183 SV *sv; // result scalar
34 UV flags; 184 JSON json;
35 int max_recurse; 185 U32 indent; // indentation level
36 int indent; 186 U32 maxdepth; // max. indentation/recursion level
187 UV limit; // escape character values >= this value when encoding
37} enc_t; 188} enc_t;
38 189
39// structure used for decoding JSON 190INLINE void
40typedef struct
41{
42 char *cur;
43 char *end;
44 char *err;
45 UV flags;
46} dec_t;
47
48static UV *
49SvJSON (SV *sv)
50{
51 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
52 croak ("object is not of type JSON::XS");
53
54 return &SvUVX (SvRV (sv));
55}
56
57/////////////////////////////////////////////////////////////////////////////
58
59static void
60need (enc_t *enc, STRLEN len) 191need (enc_t *enc, STRLEN len)
61{ 192{
62 if (enc->cur + len >= enc->end) 193 if (expect_false (enc->cur + len >= enc->end))
63 { 194 {
64 STRLEN cur = enc->cur - SvPVX (enc->sv); 195 STRLEN cur = enc->cur - SvPVX (enc->sv);
65 SvGROW (enc->sv, cur + len + 1); 196 SvGROW (enc->sv, cur + len + 1);
66 enc->cur = SvPVX (enc->sv) + cur; 197 enc->cur = SvPVX (enc->sv) + cur;
67 enc->end = SvEND (enc->sv); 198 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
68 } 199 }
69} 200}
70 201
71static void 202INLINE void
72encode_ch (enc_t *enc, char ch) 203encode_ch (enc_t *enc, char ch)
73{ 204{
74 need (enc, 1); 205 need (enc, 1);
75 *enc->cur++ = ch; 206 *enc->cur++ = ch;
76} 207}
… …
78static void 209static void
79encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) 210encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
80{ 211{
81 char *end = str + len; 212 char *end = str + len;
82 213
214 need (enc, len);
215
83 while (str < end) 216 while (str < end)
84 { 217 {
85 unsigned char ch = *(unsigned char *)str; 218 unsigned char ch = *(unsigned char *)str;
219
86 if (ch >= 0x20 && ch < 0x80) // most common case 220 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
87 { 221 {
222 if (expect_false (ch == '"')) // but with slow exceptions
223 {
224 need (enc, len += 1);
225 *enc->cur++ = '\\';
226 *enc->cur++ = '"';
227 }
228 else if (expect_false (ch == '\\'))
229 {
230 need (enc, len += 1);
231 *enc->cur++ = '\\';
232 *enc->cur++ = '\\';
233 }
234 else
88 *enc->cur++ = ch; 235 *enc->cur++ = ch;
236
89 str++; 237 ++str;
90 } 238 }
91 else 239 else
92 { 240 {
93 STRLEN clen; 241 switch (ch)
94 UV uch;
95
96 if (is_utf8)
97 { 242 {
98 uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); 243 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
99 if (clen < 0) 244 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
100 croak ("malformed UTF-8 character in string, cannot convert to JSON"); 245 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
101 } 246 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
102 else 247 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
103 {
104 uch = ch;
105 clen = 1;
106 }
107 248
108 need (enc, len += 6); 249 default:
109
110 if (uch < 0xa0 || enc->flags & F_ASCII)
111 {
112 if (uch > 0xFFFFUL)
113 { 250 {
251 STRLEN clen;
252 UV uch;
253
254 if (is_utf8)
255 {
256 uch = decode_utf8 (str, end - str, &clen);
257 if (clen == (STRLEN)-1)
258 croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
259 }
260 else
261 {
262 uch = ch;
114 len += 6; 263 clen = 1;
264 }
265
266 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
267 {
268 if (uch >= 0x10000UL)
269 {
270 if (uch >= 0x110000UL)
271 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
272
115 need (enc, len += 6); 273 need (enc, len += 11);
116 sprintf (enc->cur, "\\u%04x\\u%04x", 274 sprintf (enc->cur, "\\u%04x\\u%04x",
117 (uch - 0x10000) / 0x400 + 0xD800, 275 (int)((uch - 0x10000) / 0x400 + 0xD800),
118 (uch - 0x10000) % 0x400 + 0xDC00); 276 (int)((uch - 0x10000) % 0x400 + 0xDC00));
119 enc->cur += 12; 277 enc->cur += 12;
278 }
279 else
280 {
281 static char hexdigit [16] = "0123456789abcdef";
282 need (enc, len += 5);
283 *enc->cur++ = '\\';
284 *enc->cur++ = 'u';
285 *enc->cur++ = hexdigit [ uch >> 12 ];
286 *enc->cur++ = hexdigit [(uch >> 8) & 15];
287 *enc->cur++ = hexdigit [(uch >> 4) & 15];
288 *enc->cur++ = hexdigit [(uch >> 0) & 15];
289 }
290
291 str += clen;
120 } 292 }
293 else if (enc->json.flags & F_LATIN1)
294 {
295 *enc->cur++ = uch;
296 str += clen;
297 }
298 else if (is_utf8)
299 {
300 need (enc, len += clen);
301 do
302 {
303 *enc->cur++ = *str++;
304 }
305 while (--clen);
306 }
121 else 307 else
122 { 308 {
123 sprintf (enc->cur, "\\u%04x", uch); 309 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
124 enc->cur += 6; 310 enc->cur = encode_utf8 (enc->cur, uch);
311 ++str;
312 }
125 } 313 }
126 } 314 }
127 else if (is_utf8)
128 {
129 memcpy (enc->cur, str, clen);
130 enc->cur += clen;
131 }
132 else
133 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
134
135 str += clen;
136 } 315 }
137 316
138 --len; 317 --len;
139 } 318 }
140} 319}
141 320
142#define INDENT SB \ 321INLINE void
322encode_indent (enc_t *enc)
323{
143 if (enc->flags & F_INDENT) \ 324 if (enc->json.flags & F_INDENT)
144 { \ 325 {
145 int i_; \ 326 int spaces = enc->indent * INDENT_STEP;
146 need (enc, enc->indent); \ 327
147 for (i_ = enc->indent * 3; i_--; )\ 328 need (enc, spaces);
329 memset (enc->cur, ' ', spaces);
330 enc->cur += spaces;
331 }
332}
333
334INLINE void
335encode_space (enc_t *enc)
336{
337 need (enc, 1);
338 encode_ch (enc, ' ');
339}
340
341INLINE void
342encode_nl (enc_t *enc)
343{
344 if (enc->json.flags & F_INDENT)
345 {
346 need (enc, 1);
148 encode_ch (enc, ' '); \ 347 encode_ch (enc, '\n');
149 } \ 348 }
150 SE 349}
151 350
152#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE 351INLINE void
153#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE 352encode_comma (enc_t *enc)
154#define COMMA SB \ 353{
155 encode_ch (enc, ','); \ 354 encode_ch (enc, ',');
355
156 if (enc->flags & F_INDENT) \ 356 if (enc->json.flags & F_INDENT)
157 NL; \ 357 encode_nl (enc);
158 else if (enc->flags & F_SPACE_AFTER) \ 358 else if (enc->json.flags & F_SPACE_AFTER)
159 SPACE; \ 359 encode_space (enc);
160 SE 360}
161 361
162static void encode_sv (enc_t *enc, SV *sv); 362static void encode_sv (enc_t *enc, SV *sv);
163 363
164static void 364static void
165encode_av (enc_t *enc, AV *av) 365encode_av (enc_t *enc, AV *av)
166{ 366{
167 int i, len = av_len (av); 367 int i, len = av_len (av);
168 368
369 if (enc->indent >= enc->maxdepth)
370 croak ("data structure too deep (hit recursion limit)");
371
169 encode_ch (enc, '['); NL; 372 encode_ch (enc, '[');
170 ++enc->indent; 373
374 if (len >= 0)
375 {
376 encode_nl (enc); ++enc->indent;
171 377
172 for (i = 0; i <= len; ++i) 378 for (i = 0; i <= len; ++i)
173 { 379 {
174 INDENT; 380 SV **svp = av_fetch (av, i, 0);
175 encode_sv (enc, *av_fetch (av, i, 0));
176 381
382 encode_indent (enc);
383
384 if (svp)
385 encode_sv (enc, *svp);
386 else
387 encode_str (enc, "null", 4, 0);
388
177 if (i < len) 389 if (i < len)
178 COMMA; 390 encode_comma (enc);
179 } 391 }
180 392
181 NL; 393 encode_nl (enc); --enc->indent; encode_indent (enc);
182 394 }
183 --enc->indent; 395
184 INDENT; encode_ch (enc, ']'); 396 encode_ch (enc, ']');
185} 397}
186 398
187static void 399static void
188encode_he (enc_t *enc, HE *he) 400encode_hk (enc_t *enc, HE *he)
189{ 401{
190 encode_ch (enc, '"'); 402 encode_ch (enc, '"');
191 403
192 if (HeKLEN (he) == HEf_SVKEY) 404 if (HeKLEN (he) == HEf_SVKEY)
193 { 405 {
194 SV *sv = HeSVKEY (he); 406 SV *sv = HeSVKEY (he);
195 STRLEN len; 407 STRLEN len;
408 char *str;
409
410 SvGETMAGIC (sv);
196 char *str = SvPV (sv, len); 411 str = SvPV (sv, len);
197 412
198 encode_str (enc, str, len, SvUTF8 (sv)); 413 encode_str (enc, str, len, SvUTF8 (sv));
199 } 414 }
200 else 415 else
201 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 416 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
202 417
203 encode_ch (enc, '"'); 418 encode_ch (enc, '"');
204 419
205 if (enc->flags & F_SPACE_BEFORE) SPACE; 420 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
206 encode_ch (enc, ':'); 421 encode_ch (enc, ':');
207 if (enc->flags & F_SPACE_AFTER ) SPACE; 422 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
208 encode_sv (enc, HeVAL (he));
209} 423}
210 424
211// compare hash entries, used when all keys are bytestrings 425// compare hash entries, used when all keys are bytestrings
212static int 426static int
213he_cmp_fast (const void *a_, const void *b_) 427he_cmp_fast (const void *a_, const void *b_)
… …
218 HE *b = *(HE **)b_; 432 HE *b = *(HE **)b_;
219 433
220 STRLEN la = HeKLEN (a); 434 STRLEN la = HeKLEN (a);
221 STRLEN lb = HeKLEN (b); 435 STRLEN lb = HeKLEN (b);
222 436
223 if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 437 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
224 cmp = la < lb ? -1 : la == lb ? 0 : 1; 438 cmp = lb - la;
225 439
226 return cmp; 440 return cmp;
227} 441}
228 442
229// compare hash entries, used when some keys are sv's or utf-x 443// compare hash entries, used when some keys are sv's or utf-x
230static int 444static int
231he_cmp_slow (const void *a, const void *b) 445he_cmp_slow (const void *a, const void *b)
232{ 446{
233 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 447 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
234} 448}
235 449
236static void 450static void
237encode_hv (enc_t *enc, HV *hv) 451encode_hv (enc_t *enc, HV *hv)
238{ 452{
239 int count, i; 453 HE *he;
240 454
241 encode_ch (enc, '{'); NL; ++enc->indent; 455 if (enc->indent >= enc->maxdepth)
456 croak ("data structure too deep (hit recursion limit)");
242 457
243 if ((count = hv_iterinit (hv))) 458 encode_ch (enc, '{');
244 { 459
245 // for canonical output we have to sort by keys first 460 // for canonical output we have to sort by keys first
246 // actually, this is mostly due to the stupid so-called 461 // actually, this is mostly due to the stupid so-called
247 // security workaround added somewhere in 5.8.x. 462 // security workaround added somewhere in 5.8.x.
248 // that randomises hash orderings 463 // that randomises hash orderings
249 if (enc->flags & F_CANONICAL) 464 if (enc->json.flags & F_CANONICAL)
465 {
466 int count = hv_iterinit (hv);
467
468 if (SvMAGICAL (hv))
250 { 469 {
251 HE *he, *hes [count]; 470 // need to count by iterating. could improve by dynamically building the vector below
471 // but I don't care for the speed of this special case.
472 // note also that we will run into undefined behaviour when the two iterations
473 // do not result in the same count, something I might care for in some later release.
474
475 count = 0;
476 while (hv_iternext (hv))
477 ++count;
478
479 hv_iterinit (hv);
480 }
481
482 if (count)
483 {
252 int fast = 1; 484 int i, fast = 1;
485#if defined(__BORLANDC__) || defined(_MSC_VER)
486 HE **hes = _alloca (count * sizeof (HE));
487#else
488 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
489#endif
253 490
254 i = 0; 491 i = 0;
255 while ((he = hv_iternext (hv))) 492 while ((he = hv_iternext (hv)))
256 { 493 {
257 hes [i++] = he; 494 hes [i++] = he;
… …
263 500
264 if (fast) 501 if (fast)
265 qsort (hes, count, sizeof (HE *), he_cmp_fast); 502 qsort (hes, count, sizeof (HE *), he_cmp_fast);
266 else 503 else
267 { 504 {
268 // hack to disable "use bytes" 505 // hack to forcefully disable "use bytes"
269 COP *oldcop = PL_curcop, cop; 506 COP cop = *PL_curcop;
270 cop.op_private = 0; 507 cop.op_private = 0;
508
509 ENTER;
510 SAVETMPS;
511
512 SAVEVPTR (PL_curcop);
271 PL_curcop = &cop; 513 PL_curcop = &cop;
272 514
273 SAVETMPS;
274 qsort (hes, count, sizeof (HE *), he_cmp_slow); 515 qsort (hes, count, sizeof (HE *), he_cmp_slow);
516
275 FREETMPS; 517 FREETMPS;
276 518 LEAVE;
277 PL_curcop = oldcop;
278 } 519 }
279 520
280 for (i = 0; i < count; ++i) 521 encode_nl (enc); ++enc->indent;
522
523 while (count--)
281 { 524 {
282 INDENT; 525 encode_indent (enc);
526 he = hes [count];
283 encode_he (enc, hes [i]); 527 encode_hk (enc, he);
528 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
284 529
285 if (i < count - 1) 530 if (count)
286 COMMA; 531 encode_comma (enc);
287 } 532 }
288 533
534 encode_nl (enc); --enc->indent; encode_indent (enc);
535 }
536 }
537 else
538 {
539 if (hv_iterinit (hv) || SvMAGICAL (hv))
540 if ((he = hv_iternext (hv)))
289 NL; 541 {
542 encode_nl (enc); ++enc->indent;
543
544 for (;;)
545 {
546 encode_indent (enc);
547 encode_hk (enc, he);
548 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
549
550 if (!(he = hv_iternext (hv)))
551 break;
552
553 encode_comma (enc);
554 }
555
556 encode_nl (enc); --enc->indent; encode_indent (enc);
557 }
558 }
559
560 encode_ch (enc, '}');
561}
562
563// encode objects, arrays and special \0=false and \1=true values.
564static void
565encode_rv (enc_t *enc, SV *sv)
566{
567 svtype svt;
568
569 SvGETMAGIC (sv);
570 svt = SvTYPE (sv);
571
572 if (expect_false (SvOBJECT (sv)))
573 {
574 HV *stash = !JSON_SLOW || json_boolean_stash
575 ? json_boolean_stash
576 : gv_stashpv ("JSON::XS::Boolean", 1);
577
578 if (SvSTASH (sv) == stash)
579 {
580 if (SvIV (sv))
581 encode_str (enc, "true", 4, 0);
582 else
583 encode_str (enc, "false", 5, 0);
290 } 584 }
291 else 585 else
292 { 586 {
293 SV *sv; 587#if 0
294 HE *he = hv_iternext (hv); 588 if (0 && sv_derived_from (rv, "JSON::Literal"))
295
296 for (;;)
297 { 589 {
298 INDENT; 590 // not yet
299 encode_he (enc, he);
300
301 if (!(he = hv_iternext (hv)))
302 break;
303
304 COMMA;
305 } 591 }
306 592#endif
593 if (enc->json.flags & F_CONV_BLESSED)
307 NL; 594 {
595 // we re-bless the reference to get overload and other niceties right
596 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
597
598 if (to_json)
599 {
600 dSP;
601
602 ENTER; SAVETMPS; PUSHMARK (SP);
603 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
604
605 // calling with G_SCALAR ensures that we always get a 1 return value
606 PUTBACK;
607 call_sv ((SV *)GvCV (to_json), G_SCALAR);
608 SPAGAIN;
609
610 // catch this surprisingly common error
611 if (SvROK (TOPs) && SvRV (TOPs) == sv)
612 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
613
614 sv = POPs;
615 PUTBACK;
616
617 encode_sv (enc, sv);
618
619 FREETMPS; LEAVE;
620 }
621 else if (enc->json.flags & F_ALLOW_BLESSED)
622 encode_str (enc, "null", 4, 0);
623 else
624 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
625 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
626 }
627 else if (enc->json.flags & F_ALLOW_BLESSED)
628 encode_str (enc, "null", 4, 0);
629 else
630 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
631 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
308 } 632 }
309 } 633 }
634 else if (svt == SVt_PVHV)
635 encode_hv (enc, (HV *)sv);
636 else if (svt == SVt_PVAV)
637 encode_av (enc, (AV *)sv);
638 else if (svt < SVt_PVAV)
639 {
640 STRLEN len = 0;
641 char *pv = svt ? SvPV (sv, len) : 0;
310 642
311 --enc->indent; INDENT; encode_ch (enc, '}'); 643 if (len == 1 && *pv == '1')
644 encode_str (enc, "true", 4, 0);
645 else if (len == 1 && *pv == '0')
646 encode_str (enc, "false", 5, 0);
647 else
648 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
649 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
650 }
651 else
652 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
653 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
312} 654}
313 655
314static void 656static void
315encode_sv (enc_t *enc, SV *sv) 657encode_sv (enc_t *enc, SV *sv)
316{ 658{
659 SvGETMAGIC (sv);
660
317 if (SvPOKp (sv)) 661 if (SvPOKp (sv))
318 { 662 {
319 STRLEN len; 663 STRLEN len;
320 char *str = SvPV (sv, len); 664 char *str = SvPV (sv, len);
321 encode_ch (enc, '"'); 665 encode_ch (enc, '"');
322 encode_str (enc, str, len, SvUTF8 (sv)); 666 encode_str (enc, str, len, SvUTF8 (sv));
323 encode_ch (enc, '"'); 667 encode_ch (enc, '"');
324 } 668 }
325 else if (SvNOKp (sv)) 669 else if (SvNOKp (sv))
326 { 670 {
671 // trust that perl will do the right thing w.r.t. JSON syntax.
327 need (enc, NV_DIG + 32); 672 need (enc, NV_DIG + 32);
328 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 673 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
329 enc->cur += strlen (enc->cur); 674 enc->cur += strlen (enc->cur);
330 } 675 }
331 else if (SvIOKp (sv)) 676 else if (SvIOKp (sv))
332 { 677 {
678 // we assume we can always read an IV as a UV and vice versa
679 // we assume two's complement
680 // we assume no aliasing issues in the union
681 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
682 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
683 {
684 // optimise the "small number case"
685 // code will likely be branchless and use only a single multiplication
686 // works for numbers up to 59074
687 I32 i = SvIVX (sv);
688 U32 u;
689 char digit, nz = 0;
690
333 need (enc, 64); 691 need (enc, 6);
692
693 *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
694 u = i < 0 ? -i : i;
695
696 // convert to 4.28 fixed-point representation
697 u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
698
699 // now output digit by digit, each time masking out the integer part
700 // and multiplying by 5 while moving the decimal point one to the right,
701 // resulting in a net multiplication by 10.
702 // we always write the digit to memory but conditionally increment
703 // the pointer, to enable the use of conditional move instructions.
704 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
705 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
706 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
707 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
708 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
709 }
710 else
711 {
712 // large integer, use the (rather slow) snprintf way.
713 need (enc, IVUV_MAXCHARS);
334 enc->cur += 714 enc->cur +=
335 SvIsUV(sv) 715 SvIsUV(sv)
336 ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) 716 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
337 : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); 717 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
718 }
338 } 719 }
339 else if (SvROK (sv)) 720 else if (SvROK (sv))
340 { 721 encode_rv (enc, SvRV (sv));
341 if (!--enc->max_recurse)
342 croak ("data structure too deep (hit recursion limit)");
343
344 sv = SvRV (sv);
345
346 switch (SvTYPE (sv))
347 {
348 case SVt_PVAV: encode_av (enc, (AV *)sv); break;
349 case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
350
351 default:
352 croak ("JSON can only represent references to arrays or hashes");
353 }
354 }
355 else if (!SvOK (sv)) 722 else if (!SvOK (sv))
356 encode_str (enc, "null", 4, 0); 723 encode_str (enc, "null", 4, 0);
357 else 724 else
358 croak ("encountered perl type that JSON cannot handle"); 725 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
726 SvPV_nolen (sv), SvFLAGS (sv));
359} 727}
360 728
361static SV * 729static SV *
362encode_json (SV *scalar, UV flags) 730encode_json (SV *scalar, JSON *json)
363{ 731{
364 enc_t enc; 732 enc_t enc;
365 enc.flags = flags; 733
734 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
735 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
736
737 enc.json = *json;
366 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 738 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
367 enc.cur = SvPVX (enc.sv); 739 enc.cur = SvPVX (enc.sv);
368 enc.end = SvEND (enc.sv); 740 enc.end = SvEND (enc.sv);
369 enc.max_recurse = 0;
370 enc.indent = 0; 741 enc.indent = 0;
742 enc.maxdepth = DEC_DEPTH (enc.json.flags);
743 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
744 : enc.json.flags & F_LATIN1 ? 0x000100UL
745 : 0x110000UL;
371 746
372 SvPOK_only (enc.sv); 747 SvPOK_only (enc.sv);
373 encode_sv (&enc, scalar); 748 encode_sv (&enc, scalar);
374 749
750 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
751 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
752
375 if (!(flags & (F_ASCII | F_UTF8))) 753 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
376 SvUTF8_on (enc.sv); 754 SvUTF8_on (enc.sv);
377 755
378 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 756 if (enc.json.flags & F_SHRINK)
757 shrink (enc.sv);
758
379 return enc.sv; 759 return enc.sv;
380} 760}
381 761
382///////////////////////////////////////////////////////////////////////////// 762/////////////////////////////////////////////////////////////////////////////
763// decoder
383 764
384#define WS \ 765// structure used for decoding JSON
766typedef struct
767{
768 char *cur; // current parser pointer
769 char *end; // end of input string
770 const char *err; // parse error, if != 0
771 JSON json;
772 U32 depth; // recursion depth
773 U32 maxdepth; // recursion depth limit
774} dec_t;
775
776INLINE void
777decode_comment (dec_t *dec)
778{
779 // only '#'-style comments allowed a.t.m.
780
781 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
782 ++dec->cur;
783}
784
785INLINE void
786decode_ws (dec_t *dec)
787{
385 for (;;) \ 788 for (;;)
386 { \ 789 {
387 char ch = *dec->cur; \ 790 char ch = *dec->cur;
791
388 if (ch > 0x20 \ 792 if (ch > 0x20)
793 {
794 if (expect_false (ch == '#'))
795 {
796 if (dec->json.flags & F_RELAXED)
797 decode_comment (dec);
798 else
799 break;
800 }
801 else
802 break;
803 }
389 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ 804 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
390 break; \ 805 break; // parse error, but let higher level handle it, gives better error messages
806
391 ++dec->cur; \ 807 ++dec->cur;
392 } 808 }
809}
393 810
394#define ERR(reason) SB dec->err = reason; goto fail; SE 811#define ERR(reason) SB dec->err = reason; goto fail; SE
812
395#define EXPECT_CH(ch) SB \ 813#define EXPECT_CH(ch) SB \
396 if (*dec->cur != ch) \ 814 if (*dec->cur != ch) \
397 ERR (# ch " expected"); \ 815 ERR (# ch " expected"); \
398 ++dec->cur; \ 816 ++dec->cur; \
399 SE 817 SE
400 818
819#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
820#define DEC_DEC_DEPTH --dec->depth
821
401static SV *decode_sv (dec_t *dec); 822static SV *decode_sv (dec_t *dec);
402
403#define APPEND_CH(ch) SB \
404 SvGROW (sv, cur + 1 + 1); \
405 SvPVX (sv)[cur++] = (ch); \
406 SE
407 823
408static signed char decode_hexdigit[256]; 824static signed char decode_hexdigit[256];
409 825
410static UV 826static UV
411decode_4hex (dec_t *dec) 827decode_4hex (dec_t *dec)
412{ 828{
413 signed char d1, d2, d3, d4; 829 signed char d1, d2, d3, d4;
830 unsigned char *cur = (unsigned char *)dec->cur;
414 831
415 d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; 832 d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
416 if (d1 < 0) ERR ("four hexadecimal digits expected"); 833 d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
417 d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; 834 d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
418 if (d2 < 0) ERR ("four hexadecimal digits expected"); 835 d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
419 d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
420 if (d3 < 0) ERR ("four hexadecimal digits expected");
421 d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
422 if (d4 < 0) ERR ("four hexadecimal digits expected");
423 836
424 dec->cur += 4; 837 dec->cur += 4;
425 838
426 return ((UV)d1) << 12 839 return ((UV)d1) << 12
427 | ((UV)d2) << 8 840 | ((UV)d2) << 8
… …
433} 846}
434 847
435static SV * 848static SV *
436decode_str (dec_t *dec) 849decode_str (dec_t *dec)
437{ 850{
438 SV *sv = NEWSV (0,2); 851 SV *sv = 0;
439 STRLEN cur = 0;
440 int utf8 = 0; 852 int utf8 = 0;
853 char *dec_cur = dec->cur;
441 854
442 for (;;) 855 do
443 { 856 {
444 unsigned char ch = *(unsigned char *)dec->cur; 857 char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
858 char *cur = buf;
445 859
446 if (ch == '"') 860 do
447 break;
448 else if (ch == '\\')
449 { 861 {
450 switch (*++dec->cur) 862 unsigned char ch = *(unsigned char *)dec_cur++;
863
864 if (expect_false (ch == '"'))
451 { 865 {
452 case '\\': 866 --dec_cur;
453 case '/': 867 break;
454 case '"': APPEND_CH (*dec->cur++); break; 868 }
455 869 else if (expect_false (ch == '\\'))
456 case 'b': APPEND_CH ('\010'); ++dec->cur; break; 870 {
457 case 't': APPEND_CH ('\011'); ++dec->cur; break; 871 switch (*dec_cur)
458 case 'n': APPEND_CH ('\012'); ++dec->cur; break;
459 case 'f': APPEND_CH ('\014'); ++dec->cur; break;
460 case 'r': APPEND_CH ('\015'); ++dec->cur; break;
461
462 case 'u':
463 { 872 {
464 UV lo, hi; 873 case '\\':
465 ++dec->cur; 874 case '/':
875 case '"': *cur++ = *dec_cur++; break;
466 876
467 hi = decode_4hex (dec); 877 case 'b': ++dec_cur; *cur++ = '\010'; break;
468 if (hi == (UV)-1) 878 case 't': ++dec_cur; *cur++ = '\011'; break;
469 goto fail; 879 case 'n': ++dec_cur; *cur++ = '\012'; break;
880 case 'f': ++dec_cur; *cur++ = '\014'; break;
881 case 'r': ++dec_cur; *cur++ = '\015'; break;
470 882
471 // possibly a surrogate pair 883 case 'u':
472 if (hi >= 0xd800 && hi < 0xdc00)
473 { 884 {
474 if (dec->cur [0] != '\\' || dec->cur [1] != 'u') 885 UV lo, hi;
475 ERR ("illegal surrogate character"); 886 ++dec_cur;
476 887
477 dec->cur += 2; 888 dec->cur = dec_cur;
478
479 lo = decode_4hex (dec); 889 hi = decode_4hex (dec);
890 dec_cur = dec->cur;
480 if (lo == (UV)-1) 891 if (hi == (UV)-1)
481 goto fail; 892 goto fail;
482 893
894 // possibly a surrogate pair
895 if (hi >= 0xd800)
896 if (hi < 0xdc00)
897 {
898 if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
899 ERR ("missing low surrogate character in surrogate pair");
900
901 dec_cur += 2;
902
903 dec->cur = dec_cur;
904 lo = decode_4hex (dec);
905 dec_cur = dec->cur;
906 if (lo == (UV)-1)
907 goto fail;
908
483 if (lo < 0xdc00 || lo >= 0xe000) 909 if (lo < 0xdc00 || lo >= 0xe000)
484 ERR ("surrogate pair expected"); 910 ERR ("surrogate pair expected");
485 911
486 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; 912 hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
913 }
914 else if (hi < 0xe000)
915 ERR ("missing high surrogate character in surrogate pair");
916
917 if (hi >= 0x80)
918 {
919 utf8 = 1;
920
921 cur = encode_utf8 (cur, hi);
922 }
923 else
924 *cur++ = hi;
487 } 925 }
488 else if (lo >= 0xdc00 && lo < 0xe000)
489 ERR ("illegal surrogate character");
490
491 if (hi >= 0x80)
492 { 926 break;
493 utf8 = 1;
494 927
495 SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
496 cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
497 }
498 else 928 default:
499 APPEND_CH (hi); 929 --dec_cur;
930 ERR ("illegal backslash escape sequence in string");
500 } 931 }
932 }
933 else if (expect_true (ch >= 0x20 && ch < 0x80))
934 *cur++ = ch;
935 else if (ch >= 0x80)
936 {
937 STRLEN clen;
938 UV uch;
939
940 --dec_cur;
941
942 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
943 if (clen == (STRLEN)-1)
944 ERR ("malformed UTF-8 character in JSON string");
945
946 do
947 *cur++ = *dec_cur++;
948 while (--clen);
949
950 utf8 = 1;
951 }
952 else
953 {
954 --dec_cur;
955
956 if (!ch)
957 ERR ("unexpected end of string while parsing JSON string");
501 break; 958 else
959 ERR ("invalid character encountered while parsing JSON string");
502 } 960 }
503 } 961 }
504 else if (ch >= 0x20 && ch <= 0x7f) 962 while (cur < buf + SHORT_STRING_LEN);
505 APPEND_CH (*dec->cur++); 963
506 else if (ch >= 0x80) 964 {
965 STRLEN len = cur - buf;
966
967 if (sv)
507 { 968 {
508 STRLEN clen; 969 SvGROW (sv, SvCUR (sv) + len + 1);
509 UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); 970 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
510 if (clen < 0) 971 SvCUR_set (sv, SvCUR (sv) + len);
511 ERR ("malformed UTF-8 character in string, cannot convert to JSON");
512
513 SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
514 memcpy (SvPVX (sv) + cur, dec->cur, clen);
515 dec->cur += clen;
516 } 972 }
517 else 973 else
518 ERR ("invalid character encountered"); 974 sv = newSVpvn (buf, len);
519 } 975 }
976 }
977 while (*dec_cur != '"');
520 978
521 ++dec->cur; 979 ++dec_cur;
522 980
981 if (sv)
982 {
523 SvPOK_only (sv); 983 SvPOK_only (sv);
524
525 SvCUR_set (sv, cur);
526 *SvEND (sv) = 0; 984 *SvEND (sv) = 0;
527 985
528 if (utf8) 986 if (utf8)
529 SvUTF8_on (sv); 987 SvUTF8_on (sv);
988 }
989 else
990 sv = newSVpvn ("", 0);
530 991
992 dec->cur = dec_cur;
531 return sv; 993 return sv;
532 994
533fail: 995fail:
534 SvREFCNT_dec (sv); 996 dec->cur = dec_cur;
535 return 0; 997 return 0;
536} 998}
537 999
538static SV * 1000static SV *
539decode_num (dec_t *dec) 1001decode_num (dec_t *dec)
… …
549 { 1011 {
550 ++dec->cur; 1012 ++dec->cur;
551 if (*dec->cur >= '0' && *dec->cur <= '9') 1013 if (*dec->cur >= '0' && *dec->cur <= '9')
552 ERR ("malformed number (leading zero must not be followed by another digit)"); 1014 ERR ("malformed number (leading zero must not be followed by another digit)");
553 } 1015 }
554 1016 else if (*dec->cur < '0' || *dec->cur > '9')
555 // int 1017 ERR ("malformed number (no digits after initial minus)");
1018 else
1019 do
1020 {
1021 ++dec->cur;
1022 }
556 while (*dec->cur >= '0' && *dec->cur <= '9') 1023 while (*dec->cur >= '0' && *dec->cur <= '9');
557 ++dec->cur;
558 1024
559 // [frac] 1025 // [frac]
560 if (*dec->cur == '.') 1026 if (*dec->cur == '.')
561 { 1027 {
562 is_nv = 1; 1028 ++dec->cur;
1029
1030 if (*dec->cur < '0' || *dec->cur > '9')
1031 ERR ("malformed number (no digits after decimal point)");
563 1032
564 do 1033 do
565 { 1034 {
566 ++dec->cur; 1035 ++dec->cur;
567 } 1036 }
568 while (*dec->cur >= '0' && *dec->cur <= '9'); 1037 while (*dec->cur >= '0' && *dec->cur <= '9');
1038
1039 is_nv = 1;
569 } 1040 }
570 1041
571 // [exp] 1042 // [exp]
572 if (*dec->cur == 'e' || *dec->cur == 'E') 1043 if (*dec->cur == 'e' || *dec->cur == 'E')
573 { 1044 {
574 is_nv = 1;
575
576 ++dec->cur; 1045 ++dec->cur;
1046
577 if (*dec->cur == '-' || *dec->cur == '+') 1047 if (*dec->cur == '-' || *dec->cur == '+')
578 ++dec->cur; 1048 ++dec->cur;
579 1049
1050 if (*dec->cur < '0' || *dec->cur > '9')
1051 ERR ("malformed number (no digits after exp sign)");
1052
1053 do
1054 {
1055 ++dec->cur;
1056 }
580 while (*dec->cur >= '0' && *dec->cur <= '9') 1057 while (*dec->cur >= '0' && *dec->cur <= '9');
581 ++dec->cur; 1058
1059 is_nv = 1;
582 } 1060 }
583 1061
584 if (!is_nv) 1062 if (!is_nv)
585 { 1063 {
586 UV uv; 1064 int len = dec->cur - start;
587 int numtype = grok_number (start, dec->cur - start, &uv); 1065
588 if (numtype & IS_NUMBER_IN_UV) 1066 // special case the rather common 1..5-digit-int case
589 if (numtype & IS_NUMBER_NEG) 1067 if (*start == '-')
1068 switch (len)
590 { 1069 {
591 if (uv < (UV)IV_MIN) 1070 case 2: return newSViv (-( start [1] - '0' * 1));
592 return newSViv (-(IV)uv); 1071 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
1072 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1073 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1074 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
593 } 1075 }
1076 else
1077 switch (len)
1078 {
1079 case 1: return newSViv ( start [0] - '0' * 1);
1080 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1081 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1082 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1083 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1084 }
1085
1086 {
1087 UV uv;
1088 int numtype = grok_number (start, len, &uv);
1089 if (numtype & IS_NUMBER_IN_UV)
1090 if (numtype & IS_NUMBER_NEG)
1091 {
1092 if (uv < (UV)IV_MIN)
1093 return newSViv (-(IV)uv);
1094 }
594 else 1095 else
595 return newSVuv (uv); 1096 return newSVuv (uv);
596 } 1097 }
597 1098
1099 len -= *start == '-' ? 1 : 0;
1100
1101 // does not fit into IV or UV, try NV
1102 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
1103 #if defined (LDBL_DIG)
1104 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1105 #endif
1106 )
1107 // fits into NV without loss of precision
1108 return newSVnv (Atof (start));
1109
1110 // everything else fails, convert it to a string
1111 return newSVpvn (start, dec->cur - start);
1112 }
1113
1114 // loss of precision here
598 return newSVnv (Atof (start)); 1115 return newSVnv (Atof (start));
599 1116
600fail: 1117fail:
601 return 0; 1118 return 0;
602} 1119}
… …
604static SV * 1121static SV *
605decode_av (dec_t *dec) 1122decode_av (dec_t *dec)
606{ 1123{
607 AV *av = newAV (); 1124 AV *av = newAV ();
608 1125
1126 DEC_INC_DEPTH;
1127 decode_ws (dec);
1128
1129 if (*dec->cur == ']')
1130 ++dec->cur;
1131 else
609 for (;;) 1132 for (;;)
610 { 1133 {
611 SV *value; 1134 SV *value;
612 1135
613 value = decode_sv (dec); 1136 value = decode_sv (dec);
614 if (!value) 1137 if (!value)
615 goto fail; 1138 goto fail;
616 1139
617 av_push (av, value); 1140 av_push (av, value);
618 1141
619 WS; 1142 decode_ws (dec);
620 1143
621 if (*dec->cur == ']') 1144 if (*dec->cur == ']')
622 { 1145 {
623 ++dec->cur; 1146 ++dec->cur;
624 break; 1147 break;
1148 }
625 } 1149
626
627 if (*dec->cur != ',') 1150 if (*dec->cur != ',')
628 ERR (", or ] expected while parsing array"); 1151 ERR (", or ] expected while parsing array");
629 1152
630 ++dec->cur; 1153 ++dec->cur;
1154
1155 decode_ws (dec);
1156
1157 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1158 {
1159 ++dec->cur;
1160 break;
1161 }
631 } 1162 }
632 1163
1164 DEC_DEC_DEPTH;
633 return newRV_noinc ((SV *)av); 1165 return newRV_noinc ((SV *)av);
634 1166
635fail: 1167fail:
636 SvREFCNT_dec (av); 1168 SvREFCNT_dec (av);
1169 DEC_DEC_DEPTH;
637 return 0; 1170 return 0;
638} 1171}
639 1172
640static SV * 1173static SV *
641decode_hv (dec_t *dec) 1174decode_hv (dec_t *dec)
642{ 1175{
1176 SV *sv;
643 HV *hv = newHV (); 1177 HV *hv = newHV ();
644 1178
1179 DEC_INC_DEPTH;
1180 decode_ws (dec);
1181
1182 if (*dec->cur == '}')
1183 ++dec->cur;
1184 else
645 for (;;) 1185 for (;;)
646 { 1186 {
647 SV *key, *value;
648
649 WS; EXPECT_CH ('"'); 1187 EXPECT_CH ('"');
650 1188
651 key = decode_str (dec); 1189 // heuristic: assume that
652 if (!key) 1190 // a) decode_str + hv_store_ent are abysmally slow.
653 goto fail; 1191 // b) most hash keys are short, simple ascii text.
654 1192 // => try to "fast-match" such strings to avoid
655 WS; EXPECT_CH (':'); 1193 // the overhead of decode_str + hv_store_ent.
656
657 value = decode_sv (dec);
658 if (!value)
659 { 1194 {
1195 SV *value;
1196 char *p = dec->cur;
1197 char *e = p + 24; // only try up to 24 bytes
1198
1199 for (;;)
1200 {
1201 // the >= 0x80 is false on most architectures
1202 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1203 {
1204 // slow path, back up and use decode_str
1205 SV *key = decode_str (dec);
1206 if (!key)
1207 goto fail;
1208
1209 decode_ws (dec); EXPECT_CH (':');
1210
1211 decode_ws (dec);
1212 value = decode_sv (dec);
1213 if (!value)
1214 {
1215 SvREFCNT_dec (key);
1216 goto fail;
1217 }
1218
1219 hv_store_ent (hv, key, value, 0);
660 SvREFCNT_dec (key); 1220 SvREFCNT_dec (key);
1221
1222 break;
1223 }
1224 else if (*p == '"')
1225 {
1226 // fast path, got a simple key
1227 char *key = dec->cur;
1228 int len = p - key;
1229 dec->cur = p + 1;
1230
1231 decode_ws (dec); EXPECT_CH (':');
1232
1233 decode_ws (dec);
1234 value = decode_sv (dec);
1235 if (!value)
661 goto fail; 1236 goto fail;
1237
1238 hv_store (hv, key, len, value, 0);
1239
1240 break;
1241 }
1242
1243 ++p;
1244 }
662 } 1245 }
663 1246
664 //TODO: optimise 1247 decode_ws (dec);
665 hv_store_ent (hv, key, value, 0);
666 1248
667 WS;
668
669 if (*dec->cur == '}') 1249 if (*dec->cur == '}')
1250 {
1251 ++dec->cur;
1252 break;
1253 }
1254
1255 if (*dec->cur != ',')
1256 ERR (", or } expected while parsing object/hash");
1257
1258 ++dec->cur;
1259
1260 decode_ws (dec);
1261
1262 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1263 {
1264 ++dec->cur;
1265 break;
1266 }
1267 }
1268
1269 DEC_DEC_DEPTH;
1270 sv = newRV_noinc ((SV *)hv);
1271
1272 // check filter callbacks
1273 if (dec->json.flags & F_HOOK)
1274 {
1275 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
670 { 1276 {
671 ++dec->cur; 1277 HE *cb, *he;
672 break; 1278
1279 hv_iterinit (hv);
1280 he = hv_iternext (hv);
1281 hv_iterinit (hv);
1282
1283 // the next line creates a mortal sv each time its called.
1284 // might want to optimise this for common cases.
1285 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1286
1287 if (cb)
1288 {
1289 dSP;
1290 int count;
1291
1292 ENTER; SAVETMPS; PUSHMARK (SP);
1293 XPUSHs (HeVAL (he));
1294
1295 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1296
1297 if (count == 1)
1298 {
1299 sv = newSVsv (POPs);
1300 FREETMPS; LEAVE;
1301 return sv;
1302 }
1303
1304 FREETMPS; LEAVE;
1305 }
673 } 1306 }
674 1307
675 if (*dec->cur != ',') 1308 if (dec->json.cb_object)
676 ERR (", or } expected while parsing object/hash"); 1309 {
1310 dSP;
1311 int count;
677 1312
678 ++dec->cur; 1313 ENTER; SAVETMPS; PUSHMARK (SP);
679 } 1314 XPUSHs (sv_2mortal (sv));
680 1315
681 return newRV_noinc ((SV *)hv); 1316 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1317
1318 if (count == 1)
1319 {
1320 sv = newSVsv (POPs);
1321 FREETMPS; LEAVE;
1322 return sv;
1323 }
1324
1325 SvREFCNT_inc (sv);
1326 FREETMPS; LEAVE;
1327 }
1328 }
1329
1330 return sv;
682 1331
683fail: 1332fail:
684 SvREFCNT_dec (hv); 1333 SvREFCNT_dec (hv);
1334 DEC_DEC_DEPTH;
685 return 0; 1335 return 0;
686} 1336}
687 1337
688static SV * 1338static SV *
689decode_sv (dec_t *dec) 1339decode_sv (dec_t *dec)
690{ 1340{
691 WS; 1341 // the beauty of JSON: you need exactly one character lookahead
1342 // to parse everything.
692 switch (*dec->cur) 1343 switch (*dec->cur)
693 { 1344 {
694 case '"': ++dec->cur; return decode_str (dec); 1345 case '"': ++dec->cur; return decode_str (dec);
695 case '[': ++dec->cur; return decode_av (dec); 1346 case '[': ++dec->cur; return decode_av (dec);
696 case '{': ++dec->cur; return decode_hv (dec); 1347 case '{': ++dec->cur; return decode_hv (dec);
697 1348
698 case '-': 1349 case '-':
699 case '0': case '1': case '2': case '3': case '4': 1350 case '0': case '1': case '2': case '3': case '4':
700 case '5': case '6': case '7': case '8': case '9': 1351 case '5': case '6': case '7': case '8': case '9':
701 return decode_num (dec); 1352 return decode_num (dec);
702 1353
703 case 't': 1354 case 't':
704 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1355 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
705 { 1356 {
706 dec->cur += 4; 1357 dec->cur += 4;
1358#if JSON_SLOW
1359 json_true = get_bool ("JSON::XS::true");
1360#endif
707 return newSViv (1); 1361 return newSVsv (json_true);
708 } 1362 }
709 else 1363 else
710 ERR ("'true' expected"); 1364 ERR ("'true' expected");
711 1365
712 break; 1366 break;
713 1367
714 case 'f': 1368 case 'f':
715 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1369 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
716 { 1370 {
717 dec->cur += 5; 1371 dec->cur += 5;
1372#if JSON_SLOW
1373 json_false = get_bool ("JSON::XS::false");
1374#endif
718 return newSViv (0); 1375 return newSVsv (json_false);
719 } 1376 }
720 else 1377 else
721 ERR ("'false' expected"); 1378 ERR ("'false' expected");
722 1379
723 break; 1380 break;
724 1381
725 case 'n': 1382 case 'n':
726 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1383 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
727 { 1384 {
728 dec->cur += 4; 1385 dec->cur += 4;
729 return newSViv (1); 1386 return newSVsv (&PL_sv_undef);
730 } 1387 }
731 else 1388 else
732 ERR ("'null' expected"); 1389 ERR ("'null' expected");
733 1390
734 break; 1391 break;
735 1392
736 default: 1393 default:
737 ERR ("malformed json string"); 1394 ERR ("malformed JSON string, neither array, object, number, string or atom");
738 break; 1395 break;
739 } 1396 }
740 1397
741fail: 1398fail:
742 return 0; 1399 return 0;
743} 1400}
744 1401
745static SV * 1402static SV *
746decode_json (SV *string, UV flags) 1403decode_json (SV *string, JSON *json, STRLEN *offset_return)
747{ 1404{
1405 dec_t dec;
1406 STRLEN offset;
748 SV *sv; 1407 SV *sv;
749 1408
1409 SvGETMAGIC (string);
1410 SvUPGRADE (string, SVt_PV);
1411
1412 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
1413 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1414 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
1415
750 if (!(flags & F_UTF8)) 1416 if (json->flags & F_UTF8)
1417 sv_utf8_downgrade (string, 0);
1418 else
751 sv_utf8_upgrade (string); 1419 sv_utf8_upgrade (string);
752 1420
753 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1421 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
754 1422
755 dec_t dec; 1423 dec.json = *json;
756 dec.flags = flags;
757 dec.cur = SvPVX (string); 1424 dec.cur = SvPVX (string);
758 dec.end = SvEND (string); 1425 dec.end = SvEND (string);
759 dec.err = 0; 1426 dec.err = 0;
1427 dec.depth = 0;
1428 dec.maxdepth = DEC_DEPTH (dec.json.flags);
760 1429
761 *dec.end = 1; // invalid anywhere 1430 if (dec.json.cb_object || dec.json.cb_sk_object)
1431 dec.json.flags |= F_HOOK;
1432
1433 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1434
1435 decode_ws (&dec);
762 sv = decode_sv (&dec); 1436 sv = decode_sv (&dec);
763 *dec.end = 0; 1437
1438 if (!(offset_return || !sv))
1439 {
1440 // check for trailing garbage
1441 decode_ws (&dec);
1442
1443 if (*dec.cur)
1444 {
1445 dec.err = "garbage after JSON object";
1446 SvREFCNT_dec (sv);
1447 sv = 0;
1448 }
1449 }
1450
1451 if (offset_return || !sv)
1452 {
1453 offset = dec.json.flags & F_UTF8
1454 ? dec.cur - SvPVX (string)
1455 : utf8_distance (dec.cur, SvPVX (string));
1456
1457 if (offset_return)
1458 *offset_return = offset;
1459 }
764 1460
765 if (!sv) 1461 if (!sv)
766 { 1462 {
767 IV offset = utf8_distance (dec.cur, SvPVX (string));
768 SV *uni = sv_newmortal (); 1463 SV *uni = sv_newmortal ();
769 1464
1465 // horrible hack to silence warning inside pv_uni_display
1466 COP cop = *PL_curcop;
1467 cop.cop_warnings = pWARN_NONE;
1468 ENTER;
1469 SAVEVPTR (PL_curcop);
1470 PL_curcop = &cop;
770 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1471 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1472 LEAVE;
1473
771 croak ("%s, at character %d (%s)", 1474 croak ("%s, at character offset %d [\"%s\"]",
772 dec.err, 1475 dec.err,
773 (int)offset, 1476 (int)offset,
774 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1477 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
775 } 1478 }
776 1479
777 sv_dump (sv);//D
778 return sv_2mortal (sv); 1480 sv = sv_2mortal (sv);
1481
1482 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1483 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1484
1485 return sv;
779} 1486}
1487
1488/////////////////////////////////////////////////////////////////////////////
1489// incremental parser
1490
1491static void
1492incr_parse (JSON *self)
1493{
1494 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1495
1496 for (;;)
1497 {
1498 //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
1499 switch (self->incr_mode)
1500 {
1501 // only used for intiial whitespace skipping
1502 case INCR_M_WS:
1503 for (;;)
1504 {
1505 if (*p > 0x20)
1506 {
1507 self->incr_mode = INCR_M_JSON;
1508 goto incr_m_json;
1509 }
1510 else if (!*p)
1511 goto interrupt;
1512
1513 ++p;
1514 }
1515
1516 // skip a single char inside a string (for \\-processing)
1517 case INCR_M_BS:
1518 if (!*p)
1519 goto interrupt;
1520
1521 ++p;
1522 self->incr_mode = INCR_M_STR;
1523 goto incr_m_str;
1524
1525 // inside a string
1526 case INCR_M_STR:
1527 incr_m_str:
1528 for (;;)
1529 {
1530 if (*p == '"')
1531 {
1532 ++p;
1533 self->incr_mode = INCR_M_JSON;
1534
1535 if (!self->incr_nest)
1536 goto interrupt;
1537
1538 goto incr_m_json;
1539 }
1540 else if (*p == '\\')
1541 {
1542 ++p; // "virtually" consumes character after \
1543
1544 if (!*p) // if at end of string we have to switch modes
1545 {
1546 self->incr_mode = INCR_M_BS;
1547 goto interrupt;
1548 }
1549 }
1550 else if (!*p)
1551 goto interrupt;
1552
1553 ++p;
1554 }
1555
1556 // after initial ws, outside string
1557 case INCR_M_JSON:
1558 incr_m_json:
1559 for (;;)
1560 {
1561 switch (*p++)
1562 {
1563 case 0:
1564 --p;
1565 goto interrupt;
1566
1567 case 0x09:
1568 case 0x0a:
1569 case 0x0d:
1570 case 0x20:
1571 if (!self->incr_nest)
1572 {
1573 --p; // do not eat the whitespace, let the next round do it
1574 goto interrupt;
1575 }
1576 break;
1577
1578 case '"':
1579 self->incr_mode = INCR_M_STR;
1580 goto incr_m_str;
1581
1582 case '[':
1583 case '{':
1584 ++self->incr_nest;
1585 break;
1586
1587 case ']':
1588 case '}':
1589 if (!--self->incr_nest)
1590 goto interrupt;
1591 }
1592 }
1593 }
1594
1595 modechange:
1596 ;
1597 }
1598
1599interrupt:
1600 self->incr_pos = p - SvPVX (self->incr_text);
1601 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1602}
1603
1604/////////////////////////////////////////////////////////////////////////////
1605// XS interface functions
780 1606
781MODULE = JSON::XS PACKAGE = JSON::XS 1607MODULE = JSON::XS PACKAGE = JSON::XS
782 1608
783BOOT: 1609BOOT:
784{ 1610{
785 int i; 1611 int i;
786 1612
787 memset (decode_hexdigit, 0xff, 256);
788 for (i = 10; i--; ) 1613 for (i = 0; i < 256; ++i)
789 decode_hexdigit ['0' + i] = i; 1614 decode_hexdigit [i] =
1615 i >= '0' && i <= '9' ? i - '0'
1616 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1617 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1618 : -1;
790 1619
791 for (i = 6; --i; )
792 {
793 decode_hexdigit ['a' + i] = 10 + i;
794 decode_hexdigit ['A' + i] = 10 + i;
795 }
796
797 json_stash = gv_stashpv ("JSON::XS", 1); 1620 json_stash = gv_stashpv ("JSON::XS" , 1);
798} 1621 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
799 1622
800SV *new (char *dummy) 1623 json_true = get_bool ("JSON::XS::true");
1624 json_false = get_bool ("JSON::XS::false");
1625}
1626
1627PROTOTYPES: DISABLE
1628
1629void CLONE (...)
801 CODE: 1630 CODE:
802 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 1631 json_stash = 0;
1632 json_boolean_stash = 0;
1633
1634void new (char *klass)
1635 PPCODE:
1636{
1637 SV *pv = NEWSV (0, sizeof (JSON));
1638 SvPOK_only (pv);
1639 Zero (SvPVX (pv), 1, JSON);
1640 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1641 XPUSHs (sv_2mortal (sv_bless (
1642 newRV_noinc (pv),
1643 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1644 )));
1645}
1646
1647void ascii (JSON *self, int enable = 1)
1648 ALIAS:
1649 ascii = F_ASCII
1650 latin1 = F_LATIN1
1651 utf8 = F_UTF8
1652 indent = F_INDENT
1653 canonical = F_CANONICAL
1654 space_before = F_SPACE_BEFORE
1655 space_after = F_SPACE_AFTER
1656 pretty = F_PRETTY
1657 allow_nonref = F_ALLOW_NONREF
1658 shrink = F_SHRINK
1659 allow_blessed = F_ALLOW_BLESSED
1660 convert_blessed = F_CONV_BLESSED
1661 relaxed = F_RELAXED
1662 PPCODE:
1663{
1664 if (enable)
1665 self->flags |= ix;
1666 else
1667 self->flags &= ~ix;
1668
1669 XPUSHs (ST (0));
1670}
1671
1672void get_ascii (JSON *self)
1673 ALIAS:
1674 get_ascii = F_ASCII
1675 get_latin1 = F_LATIN1
1676 get_utf8 = F_UTF8
1677 get_indent = F_INDENT
1678 get_canonical = F_CANONICAL
1679 get_space_before = F_SPACE_BEFORE
1680 get_space_after = F_SPACE_AFTER
1681 get_allow_nonref = F_ALLOW_NONREF
1682 get_shrink = F_SHRINK
1683 get_allow_blessed = F_ALLOW_BLESSED
1684 get_convert_blessed = F_CONV_BLESSED
1685 get_relaxed = F_RELAXED
1686 PPCODE:
1687 XPUSHs (boolSV (self->flags & ix));
1688
1689void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1690 PPCODE:
1691{
1692 UV log2 = 0;
1693
1694 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1695
1696 while ((1UL << log2) < max_depth)
1697 ++log2;
1698
1699 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1700
1701 XPUSHs (ST (0));
1702}
1703
1704U32 get_max_depth (JSON *self)
1705 CODE:
1706 RETVAL = DEC_DEPTH (self->flags);
803 OUTPUT: 1707 OUTPUT:
804 RETVAL 1708 RETVAL
805 1709
806SV *ascii (SV *self, int enable) 1710void max_size (JSON *self, UV max_size = 0)
807 ALIAS: 1711 PPCODE:
808 ascii = F_ASCII 1712{
809 utf8 = F_UTF8 1713 UV log2 = 0;
810 indent = F_INDENT 1714
811 canonical = F_CANONICAL 1715 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
812 space_before = F_SPACE_BEFORE 1716 if (max_size == 1) max_size = 2;
813 space_after = F_SPACE_AFTER 1717
814 json_rpc = F_JSON_RPC 1718 while ((1UL << log2) < max_size)
815 pretty = F_PRETTY 1719 ++log2;
1720
1721 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1722
1723 XPUSHs (ST (0));
1724}
1725
1726int get_max_size (JSON *self)
816 CODE: 1727 CODE:
817{ 1728 RETVAL = DEC_SIZE (self->flags);
818 UV *uv = SvJSON (self);
819 if (enable)
820 *uv |= ix;
821 else
822 *uv &= ~ix;
823
824 RETVAL = newSVsv (self);
825}
826 OUTPUT: 1729 OUTPUT:
827 RETVAL 1730 RETVAL
828 1731
829void encode (SV *self, SV *scalar) 1732void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
830 PPCODE: 1733 PPCODE:
831 XPUSHs (encode_json (scalar, *SvJSON (self))); 1734{
1735 SvREFCNT_dec (self->cb_object);
1736 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
832 1737
833void decode (SV *self, SV *jsonstr) 1738 XPUSHs (ST (0));
1739}
1740
1741void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
834 PPCODE: 1742 PPCODE:
835 XPUSHs (decode_json (jsonstr, *SvJSON (self))); 1743{
1744 if (!self->cb_sk_object)
1745 self->cb_sk_object = newHV ();
836 1746
837void to_json (SV *scalar) 1747 if (SvOK (cb))
1748 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1749 else
1750 {
1751 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1752
1753 if (!HvKEYS (self->cb_sk_object))
1754 {
1755 SvREFCNT_dec (self->cb_sk_object);
1756 self->cb_sk_object = 0;
1757 }
1758 }
1759
1760 XPUSHs (ST (0));
1761}
1762
1763void encode (JSON *self, SV *scalar)
838 PPCODE: 1764 PPCODE:
839 XPUSHs (encode_json (scalar, F_UTF8)); 1765 XPUSHs (encode_json (scalar, self));
840 1766
841void from_json (SV *jsonstr) 1767void decode (JSON *self, SV *jsonstr)
842 PPCODE: 1768 PPCODE:
843 XPUSHs (decode_json (jsonstr, F_UTF8)); 1769 XPUSHs (decode_json (jsonstr, self, 0));
844 1770
1771void decode_prefix (JSON *self, SV *jsonstr)
1772 PPCODE:
1773{
1774 STRLEN offset;
1775 EXTEND (SP, 2);
1776 PUSHs (decode_json (jsonstr, self, &offset));
1777 PUSHs (sv_2mortal (newSVuv (offset)));
1778}
1779
1780void incr_parse (JSON *self, SV *jsonstr = 0)
1781 PPCODE:
1782{
1783 if (!self->incr_text)
1784 self->incr_text = newSVpvn ("", 0);
1785
1786 // append data, if any
1787 if (jsonstr)
1788 {
1789 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
1790 {
1791 /* utf-8-ness differs, need to upgrade */
1792 sv_utf8_upgrade (self->incr_text);
1793
1794 if (self->incr_pos)
1795 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1796 - (U8 *)SvPVX (self->incr_text);
1797 }
1798
1799 {
1800 STRLEN len;
1801 const char *str = SvPV (jsonstr, len);
1802 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
1803 Move (str, SvEND (self->incr_text), len, char);
1804 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1805 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1806 }
1807 }
1808
1809 if (GIMME_V != G_VOID)
1810 do
1811 {
1812 STRLEN offset;
1813
1814 if (!INCR_DONE (self))
1815 {
1816 incr_parse (self);
1817 if (!INCR_DONE (self))
1818 break;
1819 }
1820
1821 XPUSHs (decode_json (self->incr_text, self, &offset));
1822
1823 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1824 self->incr_pos -= offset;
1825 self->incr_nest = 0;
1826 self->incr_mode = 0;
1827 }
1828 while (GIMME_V == G_ARRAY);
1829}
1830
1831SV *incr_text (JSON *self)
1832 ATTRS: lvalue
1833 CODE:
1834{
1835 if (self->incr_pos)
1836 croak ("incr_text can not be called when the incremental parser already started parsing");
1837
1838 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1839}
1840 OUTPUT:
1841 RETVAL
1842
1843void incr_skip (JSON *self)
1844 CODE:
1845{
1846 if (self->incr_pos)
1847 {
1848 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
1849 self->incr_pos = 0;
1850 self->incr_nest = 0;
1851 self->incr_mode = 0;
1852 }
1853}
1854
1855void DESTROY (JSON *self)
1856 CODE:
1857 SvREFCNT_dec (self->cb_sk_object);
1858 SvREFCNT_dec (self->cb_object);
1859 SvREFCNT_dec (self->incr_text);
1860
1861PROTOTYPES: ENABLE
1862
1863void encode_json (SV *scalar)
1864 ALIAS:
1865 to_json_ = 0
1866 encode_json = F_UTF8
1867 PPCODE:
1868{
1869 JSON json = { F_DEFAULT | ix };
1870 XPUSHs (encode_json (scalar, &json));
1871}
1872
1873void decode_json (SV *jsonstr)
1874 ALIAS:
1875 from_json_ = 0
1876 decode_json = F_UTF8
1877 PPCODE:
1878{
1879 JSON json = { F_DEFAULT | ix };
1880 XPUSHs (decode_json (jsonstr, &json, 0));
1881}
1882
1883

Diff Legend

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