… | |
… | |
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> |
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
9 | #include <float.h> |
10 | #include <float.h> |
10 | |
11 | |
11 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | #endif |
14 | #endif |
14 | |
15 | |
15 | // some old perls do not have this, try to make it work, no |
16 | // some old perls do not have this, try to make it work, no |
16 | // guarentees, though. if it breaks, you get to keep the pieces. |
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
17 | #ifndef UTF8_MAXBYTES |
18 | #ifndef UTF8_MAXBYTES |
18 | # define UTF8_MAXBYTES 13 |
19 | # define UTF8_MAXBYTES 13 |
19 | #endif |
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) |
20 | |
35 | |
21 | #define F_ASCII 0x00000001UL |
36 | #define F_ASCII 0x00000001UL |
22 | #define F_LATIN1 0x00000002UL |
37 | #define F_LATIN1 0x00000002UL |
23 | #define F_UTF8 0x00000004UL |
38 | #define F_UTF8 0x00000004UL |
24 | #define F_INDENT 0x00000008UL |
39 | #define F_INDENT 0x00000008UL |
… | |
… | |
28 | #define F_ALLOW_NONREF 0x00000100UL |
43 | #define F_ALLOW_NONREF 0x00000100UL |
29 | #define F_SHRINK 0x00000200UL |
44 | #define F_SHRINK 0x00000200UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
45 | #define F_ALLOW_BLESSED 0x00000400UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
46 | #define F_CONV_BLESSED 0x00000800UL |
32 | #define F_RELAXED 0x00001000UL |
47 | #define F_RELAXED 0x00001000UL |
33 | |
48 | #define F_ALLOW_UNKNOWN 0x00002000UL |
34 | #define F_MAXDEPTH 0xf8000000UL |
49 | #define F_ALLOW_TAGS 0x00004000UL |
35 | #define S_MAXDEPTH 27 |
|
|
36 | #define F_MAXSIZE 0x01f00000UL |
|
|
37 | #define S_MAXSIZE 20 |
|
|
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
50 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
39 | |
51 | |
40 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
|
|
41 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
|
|
42 | |
|
|
43 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
52 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
44 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
45 | |
53 | |
46 | #define INIT_SIZE 32 // initial scalar size to be allocated |
54 | #define INIT_SIZE 32 // initial scalar size to be allocated |
47 | #define INDENT_STEP 3 // spaces per indentation level |
55 | #define INDENT_STEP 3 // spaces per indentation level |
48 | |
56 | |
49 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
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) |
50 | |
60 | |
51 | #define SB do { |
61 | #define SB do { |
52 | #define SE } while (0) |
62 | #define SE } while (0) |
53 | |
63 | |
54 | #if __GNUC__ >= 3 |
64 | #if __GNUC__ >= 3 |
… | |
… | |
64 | |
74 | |
65 | #define IN_RANGE_INC(type,val,beg,end) \ |
75 | #define IN_RANGE_INC(type,val,beg,end) \ |
66 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
76 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
67 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
77 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
68 | |
78 | |
|
|
79 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
80 | |
69 | #ifdef USE_ITHREADS |
81 | #ifdef USE_ITHREADS |
70 | # define JSON_SLOW 1 |
82 | # define JSON_SLOW 1 |
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
83 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
72 | #else |
84 | #else |
73 | # define JSON_SLOW 0 |
85 | # define JSON_SLOW 0 |
74 | # define JSON_STASH json_stash |
86 | # define JSON_STASH json_stash |
75 | #endif |
87 | #endif |
76 | |
88 | |
|
|
89 | // the amount of HEs to allocate on the stack, when sorting keys |
|
|
90 | #define STACK_HES 64 |
|
|
91 | |
77 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
92 | static HV *json_stash, *types_boolean_stash; // JSON::XS:: |
78 | static SV *json_true, *json_false; |
93 | static SV *types_true, *types_false, *sv_json; |
|
|
94 | |
|
|
95 | enum { |
|
|
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) |
79 | |
105 | |
80 | typedef struct { |
106 | typedef struct { |
81 | U32 flags; |
107 | U32 flags; |
|
|
108 | U32 max_depth; |
|
|
109 | STRLEN max_size; |
|
|
110 | |
82 | SV *cb_object; |
111 | SV *cb_object; |
83 | HV *cb_sk_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; |
84 | } JSON; |
119 | } JSON; |
|
|
120 | |
|
|
121 | INLINE void |
|
|
122 | json_init (JSON *json) |
|
|
123 | { |
|
|
124 | Zero (json, 1, JSON); |
|
|
125 | json->max_depth = 512; |
|
|
126 | } |
85 | |
127 | |
86 | ///////////////////////////////////////////////////////////////////////////// |
128 | ///////////////////////////////////////////////////////////////////////////// |
87 | // utility functions |
129 | // utility functions |
88 | |
130 | |
|
|
131 | INLINE SV * |
|
|
132 | get_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 | |
89 | INLINE void |
142 | INLINE void |
90 | shrink (SV *sv) |
143 | shrink (SV *sv) |
91 | { |
144 | { |
92 | sv_utf8_downgrade (sv, 1); |
145 | sv_utf8_downgrade (sv, 1); |
|
|
146 | |
93 | if (SvLEN (sv) > SvCUR (sv) + 1) |
147 | if (SvLEN (sv) > SvCUR (sv) + 1) |
94 | { |
148 | { |
95 | #ifdef SvPV_shrink_to_cur |
149 | #ifdef SvPV_shrink_to_cur |
96 | SvPV_shrink_to_cur (sv); |
150 | SvPV_shrink_to_cur (sv); |
97 | #elif defined (SvPV_renew) |
151 | #elif defined (SvPV_renew) |
… | |
… | |
141 | *s++ = 0x80 | ( ch & 0x3f); |
195 | *s++ = 0x80 | ( ch & 0x3f); |
142 | |
196 | |
143 | return s; |
197 | return s; |
144 | } |
198 | } |
145 | |
199 | |
|
|
200 | // convert offset pointer to character index, sv must be string |
|
|
201 | static STRLEN |
|
|
202 | ptr_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 |
|
|
213 | static void |
|
|
214 | json_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 | |
|
|
286 | static NV |
|
|
287 | json_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 | } |
146 | ///////////////////////////////////////////////////////////////////////////// |
304 | ///////////////////////////////////////////////////////////////////////////// |
147 | // encoder |
305 | // encoder |
148 | |
306 | |
149 | // structure used for encoding JSON |
307 | // structure used for encoding JSON |
150 | typedef struct |
308 | typedef struct |
… | |
… | |
152 | char *cur; // SvPVX (sv) + current output position |
310 | char *cur; // SvPVX (sv) + current output position |
153 | char *end; // SvEND (sv) |
311 | char *end; // SvEND (sv) |
154 | SV *sv; // result scalar |
312 | SV *sv; // result scalar |
155 | JSON json; |
313 | JSON json; |
156 | U32 indent; // indentation level |
314 | U32 indent; // indentation level |
157 | U32 maxdepth; // max. indentation/recursion level |
|
|
158 | UV limit; // escape character values >= this value when encoding |
315 | UV limit; // escape character values >= this value when encoding |
159 | } enc_t; |
316 | } enc_t; |
160 | |
317 | |
161 | INLINE void |
318 | INLINE void |
162 | need (enc_t *enc, STRLEN len) |
319 | need (enc_t *enc, STRLEN len) |
163 | { |
320 | { |
164 | if (expect_false (enc->cur + len >= enc->end)) |
321 | if (expect_false (enc->cur + len >= enc->end)) |
165 | { |
322 | { |
166 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
323 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
167 | SvGROW (enc->sv, cur + len + 1); |
324 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
168 | enc->cur = SvPVX (enc->sv) + cur; |
325 | enc->cur = SvPVX (enc->sv) + cur; |
169 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
326 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
170 | } |
327 | } |
171 | } |
328 | } |
172 | |
329 | |
… | |
… | |
247 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
404 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
248 | enc->cur += 12; |
405 | enc->cur += 12; |
249 | } |
406 | } |
250 | else |
407 | else |
251 | { |
408 | { |
252 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
253 | need (enc, len += 5); |
409 | need (enc, len += 5); |
254 | *enc->cur++ = '\\'; |
410 | *enc->cur++ = '\\'; |
255 | *enc->cur++ = 'u'; |
411 | *enc->cur++ = 'u'; |
256 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
412 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
257 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
413 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
258 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
414 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
259 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
415 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
260 | } |
416 | } |
261 | |
417 | |
262 | str += clen; |
418 | str += clen; |
263 | } |
419 | } |
264 | else if (enc->json.flags & F_LATIN1) |
420 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
335 | static void |
491 | static void |
336 | encode_av (enc_t *enc, AV *av) |
492 | encode_av (enc_t *enc, AV *av) |
337 | { |
493 | { |
338 | int i, len = av_len (av); |
494 | int i, len = av_len (av); |
339 | |
495 | |
340 | if (enc->indent >= enc->maxdepth) |
496 | if (enc->indent >= enc->json.max_depth) |
341 | croak ("data structure too deep (hit recursion limit)"); |
497 | croak (ERR_NESTING_EXCEEDED); |
342 | |
498 | |
343 | encode_ch (enc, '['); |
499 | encode_ch (enc, '['); |
344 | |
500 | |
345 | if (len >= 0) |
501 | if (len >= 0) |
346 | { |
502 | { |
347 | encode_nl (enc); ++enc->indent; |
503 | encode_nl (enc); ++enc->indent; |
348 | |
504 | |
349 | for (i = 0; i <= len; ++i) |
505 | for (i = 0; i <= len; ++i) |
… | |
… | |
361 | encode_comma (enc); |
517 | encode_comma (enc); |
362 | } |
518 | } |
363 | |
519 | |
364 | encode_nl (enc); --enc->indent; encode_indent (enc); |
520 | encode_nl (enc); --enc->indent; encode_indent (enc); |
365 | } |
521 | } |
366 | |
522 | |
367 | encode_ch (enc, ']'); |
523 | encode_ch (enc, ']'); |
368 | } |
524 | } |
369 | |
525 | |
370 | static void |
526 | static void |
371 | encode_hk (enc_t *enc, HE *he) |
527 | encode_hk (enc_t *enc, HE *he) |
… | |
… | |
375 | if (HeKLEN (he) == HEf_SVKEY) |
531 | if (HeKLEN (he) == HEf_SVKEY) |
376 | { |
532 | { |
377 | SV *sv = HeSVKEY (he); |
533 | SV *sv = HeSVKEY (he); |
378 | STRLEN len; |
534 | STRLEN len; |
379 | char *str; |
535 | char *str; |
380 | |
536 | |
381 | SvGETMAGIC (sv); |
537 | SvGETMAGIC (sv); |
382 | str = SvPV (sv, len); |
538 | str = SvPV (sv, len); |
383 | |
539 | |
384 | encode_str (enc, str, len, SvUTF8 (sv)); |
540 | encode_str (enc, str, len, SvUTF8 (sv)); |
385 | } |
541 | } |
… | |
… | |
420 | |
576 | |
421 | static void |
577 | static void |
422 | encode_hv (enc_t *enc, HV *hv) |
578 | encode_hv (enc_t *enc, HV *hv) |
423 | { |
579 | { |
424 | HE *he; |
580 | HE *he; |
425 | int count; |
|
|
426 | |
581 | |
427 | if (enc->indent >= enc->maxdepth) |
582 | if (enc->indent >= enc->json.max_depth) |
428 | croak ("data structure too deep (hit recursion limit)"); |
583 | croak (ERR_NESTING_EXCEEDED); |
429 | |
584 | |
430 | encode_ch (enc, '{'); |
585 | encode_ch (enc, '{'); |
431 | |
586 | |
432 | // for canonical output we have to sort by keys first |
587 | // for canonical output we have to sort by keys first |
433 | // actually, this is mostly due to the stupid so-called |
588 | // actually, this is mostly due to the stupid so-called |
434 | // security workaround added somewhere in 5.8.x. |
589 | // security workaround added somewhere in 5.8.x |
435 | // that randomises hash orderings |
590 | // that randomises hash orderings |
436 | if (enc->json.flags & F_CANONICAL) |
591 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
437 | { |
592 | { |
438 | int count = hv_iterinit (hv); |
593 | int count = hv_iterinit (hv); |
439 | |
594 | |
440 | if (SvMAGICAL (hv)) |
595 | if (SvMAGICAL (hv)) |
441 | { |
596 | { |
… | |
… | |
452 | } |
607 | } |
453 | |
608 | |
454 | if (count) |
609 | if (count) |
455 | { |
610 | { |
456 | int i, fast = 1; |
611 | int i, fast = 1; |
457 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
612 | HE *hes_stack [STACK_HES]; |
458 | HE **hes = _alloca (count * sizeof (HE)); |
613 | HE **hes = hes_stack; |
459 | #else |
614 | |
460 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
615 | // allocate larger arrays on the heap |
461 | #endif |
616 | if (count > STACK_HES) |
|
|
617 | { |
|
|
618 | SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes))); |
|
|
619 | hes = (HE **)SvPVX (sv); |
|
|
620 | } |
462 | |
621 | |
463 | i = 0; |
622 | i = 0; |
464 | while ((he = hv_iternext (hv))) |
623 | while ((he = hv_iternext (hv))) |
465 | { |
624 | { |
466 | hes [i++] = he; |
625 | hes [i++] = he; |
… | |
… | |
535 | // encode objects, arrays and special \0=false and \1=true values. |
694 | // encode objects, arrays and special \0=false and \1=true values. |
536 | static void |
695 | static void |
537 | encode_rv (enc_t *enc, SV *sv) |
696 | encode_rv (enc_t *enc, SV *sv) |
538 | { |
697 | { |
539 | svtype svt; |
698 | svtype svt; |
|
|
699 | GV *method; |
540 | |
700 | |
541 | SvGETMAGIC (sv); |
701 | SvGETMAGIC (sv); |
542 | svt = SvTYPE (sv); |
702 | svt = SvTYPE (sv); |
543 | |
703 | |
544 | if (expect_false (SvOBJECT (sv))) |
704 | if (expect_false (SvOBJECT (sv))) |
545 | { |
705 | { |
546 | HV *stash = !JSON_SLOW || json_boolean_stash |
706 | HV *boolean_stash = !JSON_SLOW || types_boolean_stash |
547 | ? json_boolean_stash |
707 | ? types_boolean_stash |
548 | : gv_stashpv ("JSON::XS::Boolean", 1); |
708 | : gv_stashpv ("Types::Serialiser::Boolean", 1); |
|
|
709 | HV *stash = SvSTASH (sv); |
549 | |
710 | |
550 | if (SvSTASH (sv) == stash) |
711 | if (stash == boolean_stash) |
551 | { |
712 | { |
552 | if (SvIV (sv)) |
713 | if (SvIV (sv)) |
553 | encode_str (enc, "true", 4, 0); |
714 | encode_str (enc, "true", 4, 0); |
554 | else |
715 | else |
555 | encode_str (enc, "false", 5, 0); |
716 | encode_str (enc, "false", 5, 0); |
556 | } |
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); |
557 | else |
785 | else |
558 | { |
|
|
559 | #if 0 |
|
|
560 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
|
|
561 | { |
|
|
562 | // not yet |
|
|
563 | } |
|
|
564 | #endif |
|
|
565 | if (enc->json.flags & F_CONV_BLESSED) |
|
|
566 | { |
|
|
567 | // we re-bless the reference to get overload and other niceties right |
|
|
568 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
|
|
569 | |
|
|
570 | if (to_json) |
|
|
571 | { |
|
|
572 | dSP; |
|
|
573 | |
|
|
574 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
575 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
576 | |
|
|
577 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
578 | PUTBACK; |
|
|
579 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
|
|
580 | SPAGAIN; |
|
|
581 | |
|
|
582 | // catch this surprisingly common error |
|
|
583 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
584 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
585 | |
|
|
586 | sv = POPs; |
|
|
587 | PUTBACK; |
|
|
588 | |
|
|
589 | encode_sv (enc, sv); |
|
|
590 | |
|
|
591 | FREETMPS; LEAVE; |
|
|
592 | } |
|
|
593 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
594 | encode_str (enc, "null", 4, 0); |
|
|
595 | else |
|
|
596 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
597 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
598 | } |
|
|
599 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
600 | encode_str (enc, "null", 4, 0); |
|
|
601 | else |
|
|
602 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
786 | croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)", |
603 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
787 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
604 | } |
|
|
605 | } |
788 | } |
606 | else if (svt == SVt_PVHV) |
789 | else if (svt == SVt_PVHV) |
607 | encode_hv (enc, (HV *)sv); |
790 | encode_hv (enc, (HV *)sv); |
608 | else if (svt == SVt_PVAV) |
791 | else if (svt == SVt_PVAV) |
609 | encode_av (enc, (AV *)sv); |
792 | encode_av (enc, (AV *)sv); |
… | |
… | |
614 | |
797 | |
615 | if (len == 1 && *pv == '1') |
798 | if (len == 1 && *pv == '1') |
616 | encode_str (enc, "true", 4, 0); |
799 | encode_str (enc, "true", 4, 0); |
617 | else if (len == 1 && *pv == '0') |
800 | else if (len == 1 && *pv == '0') |
618 | encode_str (enc, "false", 5, 0); |
801 | encode_str (enc, "false", 5, 0); |
|
|
802 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
803 | encode_str (enc, "null", 4, 0); |
619 | else |
804 | else |
620 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
805 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
621 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
806 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
622 | } |
807 | } |
|
|
808 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
809 | encode_str (enc, "null", 4, 0); |
623 | else |
810 | else |
624 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
811 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
625 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
812 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
626 | } |
813 | } |
627 | |
814 | |
… | |
… | |
648 | else if (SvIOKp (sv)) |
835 | else if (SvIOKp (sv)) |
649 | { |
836 | { |
650 | // we assume we can always read an IV as a UV and vice versa |
837 | // we assume we can always read an IV as a UV and vice versa |
651 | // we assume two's complement |
838 | // we assume two's complement |
652 | // we assume no aliasing issues in the union |
839 | // we assume no aliasing issues in the union |
653 | if (SvIsUV (sv) ? SvUVX (sv) > 59000 |
840 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
654 | : SvIVX (sv) > 59000 || SvIVX (sv) < -59000) |
841 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
655 | { |
|
|
656 | // large integer, use the (rather slow) snprintf way. |
|
|
657 | need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8 |
|
|
658 | enc->cur += |
|
|
659 | SvIsUV(sv) |
|
|
660 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
|
|
661 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
662 | } |
|
|
663 | else |
|
|
664 | { |
842 | { |
665 | // optimise the "small number case" |
843 | // optimise the "small number case" |
666 | // code will likely be branchless and use only a single multiplication |
844 | // code will likely be branchless and use only a single multiplication |
667 | // works for numbers up to 59074 |
845 | // works for numbers up to 59074 |
668 | I32 i = SvIVX (sv); |
846 | I32 i = SvIVX (sv); |
… | |
… | |
679 | |
857 | |
680 | // now output digit by digit, each time masking out the integer part |
858 | // now output digit by digit, each time masking out the integer part |
681 | // and multiplying by 5 while moving the decimal point one to the right, |
859 | // and multiplying by 5 while moving the decimal point one to the right, |
682 | // resulting in a net multiplication by 10. |
860 | // resulting in a net multiplication by 10. |
683 | // we always write the digit to memory but conditionally increment |
861 | // we always write the digit to memory but conditionally increment |
684 | // the pointer, to ease the usage of conditional move instructions. |
862 | // the pointer, to enable the use of conditional move instructions. |
685 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
863 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
686 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
864 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
687 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
865 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
688 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
866 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
689 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
867 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
690 | } |
868 | } |
|
|
869 | else |
|
|
870 | { |
|
|
871 | // large integer, use the (rather slow) snprintf way. |
|
|
872 | need (enc, IVUV_MAXCHARS); |
|
|
873 | enc->cur += |
|
|
874 | SvIsUV(sv) |
|
|
875 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
876 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
877 | } |
691 | } |
878 | } |
692 | else if (SvROK (sv)) |
879 | else if (SvROK (sv)) |
693 | encode_rv (enc, SvRV (sv)); |
880 | encode_rv (enc, SvRV (sv)); |
694 | else if (!SvOK (sv)) |
881 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
695 | encode_str (enc, "null", 4, 0); |
882 | encode_str (enc, "null", 4, 0); |
696 | else |
883 | else |
697 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
884 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data", |
698 | SvPV_nolen (sv), SvFLAGS (sv)); |
885 | SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
|
|
886 | } |
|
|
887 | |
|
|
888 | static int |
|
|
889 | json_scalar (SV *scalar) |
|
|
890 | { |
|
|
891 | return 0;//D |
|
|
892 | if (!SvROK (scalar)) |
|
|
893 | return 1; |
699 | } |
894 | } |
700 | |
895 | |
701 | static SV * |
896 | static SV * |
702 | encode_json (SV *scalar, JSON *json) |
897 | encode_json (SV *scalar, JSON *json) |
703 | { |
898 | { |
704 | enc_t enc; |
899 | enc_t enc; |
705 | |
900 | |
706 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
901 | if (!(json->flags & F_ALLOW_NONREF) && json_scalar (scalar)) |
707 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
902 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
708 | |
903 | |
709 | enc.json = *json; |
904 | enc.json = *json; |
710 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
905 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
711 | enc.cur = SvPVX (enc.sv); |
906 | enc.cur = SvPVX (enc.sv); |
712 | enc.end = SvEND (enc.sv); |
907 | enc.end = SvEND (enc.sv); |
713 | enc.indent = 0; |
908 | enc.indent = 0; |
714 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
|
|
715 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
909 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
716 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
910 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
717 | : 0x110000UL; |
911 | : 0x110000UL; |
718 | |
912 | |
719 | SvPOK_only (enc.sv); |
913 | SvPOK_only (enc.sv); |
720 | encode_sv (&enc, scalar); |
914 | encode_sv (&enc, scalar); |
|
|
915 | encode_nl (&enc); |
721 | |
916 | |
722 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
917 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
723 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
918 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
724 | |
919 | |
725 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
920 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
786 | if (*dec->cur != ch) \ |
981 | if (*dec->cur != ch) \ |
787 | ERR (# ch " expected"); \ |
982 | ERR (# ch " expected"); \ |
788 | ++dec->cur; \ |
983 | ++dec->cur; \ |
789 | SE |
984 | SE |
790 | |
985 | |
791 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
986 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
792 | #define DEC_DEC_DEPTH --dec->depth |
987 | #define DEC_DEC_DEPTH --dec->depth |
793 | |
988 | |
794 | static SV *decode_sv (dec_t *dec); |
989 | static SV *decode_sv (dec_t *dec); |
795 | |
990 | |
796 | static signed char decode_hexdigit[256]; |
991 | static signed char decode_hexdigit[256]; |
… | |
… | |
905 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
1100 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
906 | *cur++ = ch; |
1101 | *cur++ = ch; |
907 | else if (ch >= 0x80) |
1102 | else if (ch >= 0x80) |
908 | { |
1103 | { |
909 | STRLEN clen; |
1104 | STRLEN clen; |
910 | UV uch; |
|
|
911 | |
1105 | |
912 | --dec_cur; |
1106 | --dec_cur; |
913 | |
1107 | |
914 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1108 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
915 | if (clen == (STRLEN)-1) |
1109 | if (clen == (STRLEN)-1) |
916 | ERR ("malformed UTF-8 character in JSON string"); |
1110 | ERR ("malformed UTF-8 character in JSON string"); |
917 | |
1111 | |
918 | do |
1112 | do |
919 | *cur++ = *dec_cur++; |
1113 | *cur++ = *dec_cur++; |
920 | while (--clen); |
1114 | while (--clen); |
921 | |
1115 | |
922 | utf8 = 1; |
1116 | utf8 = 1; |
923 | } |
1117 | } |
|
|
1118 | else if (ch == '\t' && dec->json.flags & F_RELAXED) |
|
|
1119 | *cur++ = ch; |
924 | else |
1120 | else |
925 | { |
1121 | { |
926 | --dec_cur; |
1122 | --dec_cur; |
927 | |
1123 | |
928 | if (!ch) |
1124 | if (!ch) |
… | |
… | |
936 | { |
1132 | { |
937 | STRLEN len = cur - buf; |
1133 | STRLEN len = cur - buf; |
938 | |
1134 | |
939 | if (sv) |
1135 | if (sv) |
940 | { |
1136 | { |
941 | SvGROW (sv, SvCUR (sv) + len + 1); |
1137 | STRLEN cur = SvCUR (sv); |
|
|
1138 | |
|
|
1139 | if (SvLEN (sv) <= cur + len) |
|
|
1140 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1141 | |
942 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1142 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
943 | SvCUR_set (sv, SvCUR (sv) + len); |
1143 | SvCUR_set (sv, SvCUR (sv) + len); |
944 | } |
1144 | } |
945 | else |
1145 | else |
946 | sv = newSVpvn (buf, len); |
1146 | sv = newSVpvn (buf, len); |
… | |
… | |
1037 | |
1237 | |
1038 | // special case the rather common 1..5-digit-int case |
1238 | // special case the rather common 1..5-digit-int case |
1039 | if (*start == '-') |
1239 | if (*start == '-') |
1040 | switch (len) |
1240 | switch (len) |
1041 | { |
1241 | { |
1042 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1242 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
1043 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1243 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
1044 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1244 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1045 | case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1245 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1046 | case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1246 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1047 | } |
1247 | } |
1048 | else |
1248 | else |
1049 | switch (len) |
1249 | switch (len) |
1050 | { |
1250 | { |
1051 | case 1: return newSViv ( start [0] - '0' * 1); |
1251 | case 1: return newSViv ( start [0] - '0' * 1); |
1052 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1252 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1053 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1253 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1054 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1254 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1055 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1255 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1056 | } |
1256 | } |
1057 | |
1257 | |
1058 | { |
1258 | { |
1059 | UV uv; |
1259 | UV uv; |
1060 | int numtype = grok_number (start, len, &uv); |
1260 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1069 | } |
1269 | } |
1070 | |
1270 | |
1071 | len -= *start == '-' ? 1 : 0; |
1271 | len -= *start == '-' ? 1 : 0; |
1072 | |
1272 | |
1073 | // does not fit into IV or UV, try NV |
1273 | // does not fit into IV or UV, try NV |
1074 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
1274 | if (len <= NV_DIG) |
1075 | #if defined (LDBL_DIG) |
|
|
1076 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
1077 | #endif |
|
|
1078 | ) |
|
|
1079 | // fits into NV without loss of precision |
1275 | // fits into NV without loss of precision |
1080 | return newSVnv (Atof (start)); |
1276 | return newSVnv (json_atof (start)); |
1081 | |
1277 | |
1082 | // everything else fails, convert it to a string |
1278 | // everything else fails, convert it to a string |
1083 | return newSVpvn (start, dec->cur - start); |
1279 | return newSVpvn (start, dec->cur - start); |
1084 | } |
1280 | } |
1085 | |
1281 | |
1086 | // loss of precision here |
1282 | // loss of precision here |
1087 | return newSVnv (Atof (start)); |
1283 | return newSVnv (json_atof (start)); |
1088 | |
1284 | |
1089 | fail: |
1285 | fail: |
1090 | return 0; |
1286 | return 0; |
1091 | } |
1287 | } |
1092 | |
1288 | |
… | |
… | |
1116 | if (*dec->cur == ']') |
1312 | if (*dec->cur == ']') |
1117 | { |
1313 | { |
1118 | ++dec->cur; |
1314 | ++dec->cur; |
1119 | break; |
1315 | break; |
1120 | } |
1316 | } |
1121 | |
1317 | |
1122 | if (*dec->cur != ',') |
1318 | if (*dec->cur != ',') |
1123 | ERR (", or ] expected while parsing array"); |
1319 | ERR (", or ] expected while parsing array"); |
1124 | |
1320 | |
1125 | ++dec->cur; |
1321 | ++dec->cur; |
1126 | |
1322 | |
… | |
… | |
1168 | char *p = dec->cur; |
1364 | char *p = dec->cur; |
1169 | char *e = p + 24; // only try up to 24 bytes |
1365 | char *e = p + 24; // only try up to 24 bytes |
1170 | |
1366 | |
1171 | for (;;) |
1367 | for (;;) |
1172 | { |
1368 | { |
1173 | // the >= 0x80 is true on most architectures |
1369 | // the >= 0x80 is false on most architectures |
1174 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1370 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1175 | { |
1371 | { |
1176 | // slow path, back up and use decode_str |
1372 | // slow path, back up and use decode_str |
1177 | SV *key = decode_str (dec); |
1373 | SV *key = decode_str (dec); |
1178 | if (!key) |
1374 | if (!key) |
… | |
… | |
1250 | |
1446 | |
1251 | hv_iterinit (hv); |
1447 | hv_iterinit (hv); |
1252 | he = hv_iternext (hv); |
1448 | he = hv_iternext (hv); |
1253 | hv_iterinit (hv); |
1449 | hv_iterinit (hv); |
1254 | |
1450 | |
1255 | // the next line creates a mortal sv each time its called. |
1451 | // the next line creates a mortal sv each time it's called. |
1256 | // might want to optimise this for common cases. |
1452 | // might want to optimise this for common cases. |
1257 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1453 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1258 | |
1454 | |
1259 | if (cb) |
1455 | if (cb) |
1260 | { |
1456 | { |
1261 | dSP; |
1457 | dSP; |
1262 | int count; |
1458 | int count; |
1263 | |
1459 | |
1264 | ENTER; SAVETMPS; PUSHMARK (SP); |
1460 | ENTER; SAVETMPS; |
|
|
1461 | SAVESTACK_POS (); |
|
|
1462 | PUSHMARK (SP); |
1265 | XPUSHs (HeVAL (he)); |
1463 | XPUSHs (HeVAL (he)); |
|
|
1464 | sv_2mortal (sv); |
1266 | |
1465 | |
1267 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1466 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1268 | |
1467 | |
1269 | if (count == 1) |
1468 | if (count == 1) |
1270 | { |
1469 | { |
1271 | sv = newSVsv (POPs); |
1470 | sv = newSVsv (POPs); |
1272 | FREETMPS; LEAVE; |
1471 | FREETMPS; LEAVE; |
1273 | return sv; |
1472 | return sv; |
1274 | } |
1473 | } |
1275 | |
1474 | |
|
|
1475 | SvREFCNT_inc (sv); |
1276 | FREETMPS; LEAVE; |
1476 | FREETMPS; LEAVE; |
1277 | } |
1477 | } |
1278 | } |
1478 | } |
1279 | |
1479 | |
1280 | if (dec->json.cb_object) |
1480 | if (dec->json.cb_object) |
1281 | { |
1481 | { |
1282 | dSP; |
1482 | dSP; |
1283 | int count; |
1483 | int count; |
1284 | |
1484 | |
1285 | ENTER; SAVETMPS; PUSHMARK (SP); |
1485 | ENTER; SAVETMPS; |
|
|
1486 | SAVESTACK_POS (); |
|
|
1487 | PUSHMARK (SP); |
1286 | XPUSHs (sv_2mortal (sv)); |
1488 | XPUSHs (sv_2mortal (sv)); |
1287 | |
1489 | |
1288 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1490 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1289 | |
1491 | |
1290 | if (count == 1) |
1492 | if (count == 1) |
… | |
… | |
1306 | DEC_DEC_DEPTH; |
1508 | DEC_DEC_DEPTH; |
1307 | return 0; |
1509 | return 0; |
1308 | } |
1510 | } |
1309 | |
1511 | |
1310 | static SV * |
1512 | static SV * |
|
|
1513 | decode_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 | |
|
|
1589 | fail: |
|
|
1590 | SvREFCNT_dec (tag); |
|
|
1591 | SvREFCNT_dec (val); |
|
|
1592 | return 0; |
|
|
1593 | } |
|
|
1594 | |
|
|
1595 | static SV * |
1311 | decode_sv (dec_t *dec) |
1596 | decode_sv (dec_t *dec) |
1312 | { |
1597 | { |
1313 | // the beauty of JSON: you need exactly one character lookahead |
1598 | // the beauty of JSON: you need exactly one character lookahead |
1314 | // to parse everything. |
1599 | // to parse everything. |
1315 | switch (*dec->cur) |
1600 | switch (*dec->cur) |
1316 | { |
1601 | { |
1317 | case '"': ++dec->cur; return decode_str (dec); |
1602 | case '"': ++dec->cur; return decode_str (dec); |
1318 | case '[': ++dec->cur; return decode_av (dec); |
1603 | case '[': ++dec->cur; return decode_av (dec); |
1319 | case '{': ++dec->cur; return decode_hv (dec); |
1604 | case '{': ++dec->cur; return decode_hv (dec); |
|
|
1605 | case '(': return decode_tag (dec); |
1320 | |
1606 | |
1321 | case '-': |
1607 | case '-': |
1322 | case '0': case '1': case '2': case '3': case '4': |
1608 | case '0': case '1': case '2': case '3': case '4': |
1323 | case '5': case '6': case '7': case '8': case '9': |
1609 | case '5': case '6': case '7': case '8': case '9': |
1324 | return decode_num (dec); |
1610 | return decode_num (dec); |
… | |
… | |
1326 | case 't': |
1612 | case 't': |
1327 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1613 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1328 | { |
1614 | { |
1329 | dec->cur += 4; |
1615 | dec->cur += 4; |
1330 | #if JSON_SLOW |
1616 | #if JSON_SLOW |
1331 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
1617 | types_true = get_bool ("Types::Serialiser::true"); |
1332 | #endif |
1618 | #endif |
1333 | return SvREFCNT_inc (json_true); |
1619 | return newSVsv (types_true); |
1334 | } |
1620 | } |
1335 | else |
1621 | else |
1336 | ERR ("'true' expected"); |
1622 | ERR ("'true' expected"); |
1337 | |
1623 | |
1338 | break; |
1624 | break; |
… | |
… | |
1340 | case 'f': |
1626 | case 'f': |
1341 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1627 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1342 | { |
1628 | { |
1343 | dec->cur += 5; |
1629 | dec->cur += 5; |
1344 | #if JSON_SLOW |
1630 | #if JSON_SLOW |
1345 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1631 | types_false = get_bool ("Types::Serialiser::false"); |
1346 | #endif |
1632 | #endif |
1347 | return SvREFCNT_inc (json_false); |
1633 | return newSVsv (types_false); |
1348 | } |
1634 | } |
1349 | else |
1635 | else |
1350 | ERR ("'false' expected"); |
1636 | ERR ("'false' expected"); |
1351 | |
1637 | |
1352 | break; |
1638 | break; |
… | |
… | |
1361 | ERR ("'null' expected"); |
1647 | ERR ("'null' expected"); |
1362 | |
1648 | |
1363 | break; |
1649 | break; |
1364 | |
1650 | |
1365 | default: |
1651 | default: |
1366 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
1652 | ERR ("malformed JSON string, neither tag, array, object, number, string or atom"); |
1367 | break; |
1653 | break; |
1368 | } |
1654 | } |
1369 | |
1655 | |
1370 | fail: |
1656 | fail: |
1371 | return 0; |
1657 | return 0; |
1372 | } |
1658 | } |
1373 | |
1659 | |
1374 | static SV * |
1660 | static SV * |
1375 | decode_json (SV *string, JSON *json, UV *offset_return) |
1661 | decode_json (SV *string, JSON *json, char **offset_return) |
1376 | { |
1662 | { |
1377 | dec_t dec; |
1663 | dec_t dec; |
1378 | UV offset; |
|
|
1379 | SV *sv; |
1664 | SV *sv; |
1380 | |
1665 | |
|
|
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 | */ |
1381 | SvGETMAGIC (string); |
1671 | /*SvGETMAGIC (string);*/ |
|
|
1672 | if (SvMAGICAL (string) || !SvPOK (string)) |
|
|
1673 | string = sv_2mortal (newSVsv (string)); |
|
|
1674 | |
1382 | SvUPGRADE (string, SVt_PV); |
1675 | SvUPGRADE (string, SVt_PV); |
1383 | |
1676 | |
1384 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
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) |
1385 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1695 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1386 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1696 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1697 | } |
1387 | |
1698 | |
1388 | if (json->flags & F_UTF8) |
1699 | if (DECODE_WANTS_OCTETS (json)) |
1389 | sv_utf8_downgrade (string, 0); |
1700 | sv_utf8_downgrade (string, 0); |
1390 | else |
1701 | else |
1391 | sv_utf8_upgrade (string); |
1702 | sv_utf8_upgrade (string); |
1392 | |
1703 | |
1393 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1704 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1394 | |
1705 | |
1395 | dec.json = *json; |
1706 | dec.json = *json; |
1396 | dec.cur = SvPVX (string); |
1707 | dec.cur = SvPVX (string); |
1397 | dec.end = SvEND (string); |
1708 | dec.end = SvEND (string); |
1398 | dec.err = 0; |
1709 | dec.err = 0; |
1399 | dec.depth = 0; |
1710 | dec.depth = 0; |
1400 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1401 | |
1711 | |
1402 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1712 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1403 | dec.json.flags |= F_HOOK; |
1713 | dec.json.flags |= F_HOOK; |
1404 | |
1714 | |
1405 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1715 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1406 | |
1716 | |
1407 | decode_ws (&dec); |
1717 | decode_ws (&dec); |
1408 | sv = decode_sv (&dec); |
1718 | sv = decode_sv (&dec); |
|
|
1719 | |
|
|
1720 | if (offset_return) |
|
|
1721 | *offset_return = dec.cur; |
1409 | |
1722 | |
1410 | if (!(offset_return || !sv)) |
1723 | if (!(offset_return || !sv)) |
1411 | { |
1724 | { |
1412 | // check for trailing garbage |
1725 | // check for trailing garbage |
1413 | decode_ws (&dec); |
1726 | decode_ws (&dec); |
… | |
… | |
1416 | { |
1729 | { |
1417 | dec.err = "garbage after JSON object"; |
1730 | dec.err = "garbage after JSON object"; |
1418 | SvREFCNT_dec (sv); |
1731 | SvREFCNT_dec (sv); |
1419 | sv = 0; |
1732 | sv = 0; |
1420 | } |
1733 | } |
1421 | } |
|
|
1422 | |
|
|
1423 | if (offset_return || !sv) |
|
|
1424 | { |
|
|
1425 | offset = dec.json.flags & F_UTF8 |
|
|
1426 | ? dec.cur - SvPVX (string) |
|
|
1427 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1428 | |
|
|
1429 | if (offset_return) |
|
|
1430 | *offset_return = offset; |
|
|
1431 | } |
1734 | } |
1432 | |
1735 | |
1433 | if (!sv) |
1736 | if (!sv) |
1434 | { |
1737 | { |
1435 | SV *uni = sv_newmortal (); |
1738 | SV *uni = sv_newmortal (); |
… | |
… | |
1441 | SAVEVPTR (PL_curcop); |
1744 | SAVEVPTR (PL_curcop); |
1442 | PL_curcop = &cop; |
1745 | PL_curcop = &cop; |
1443 | 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); |
1444 | LEAVE; |
1747 | LEAVE; |
1445 | |
1748 | |
1446 | croak ("%s, at character offset %d [\"%s\"]", |
1749 | croak ("%s, at character offset %d (before \"%s\")", |
1447 | dec.err, |
1750 | dec.err, |
1448 | (int)offset, |
1751 | (int)ptr_to_index (string, dec.cur), |
1449 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1752 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1450 | } |
1753 | } |
1451 | |
1754 | |
1452 | sv = sv_2mortal (sv); |
1755 | sv = sv_2mortal (sv); |
1453 | |
1756 | |
1454 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1757 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1455 | croak ("JSON text must be an object or array (but found number, string, true, false or null, 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)"); |
1456 | |
1759 | |
1457 | return sv; |
1760 | return sv; |
|
|
1761 | } |
|
|
1762 | |
|
|
1763 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1764 | // incremental parser |
|
|
1765 | |
|
|
1766 | static void |
|
|
1767 | incr_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 | |
|
|
1915 | interrupt: |
|
|
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 |
1458 | } |
1919 | } |
1459 | |
1920 | |
1460 | ///////////////////////////////////////////////////////////////////////////// |
1921 | ///////////////////////////////////////////////////////////////////////////// |
1461 | // XS interface functions |
1922 | // XS interface functions |
1462 | |
1923 | |
… | |
… | |
1471 | i >= '0' && i <= '9' ? i - '0' |
1932 | i >= '0' && i <= '9' ? i - '0' |
1472 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1933 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1473 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1934 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1474 | : -1; |
1935 | : -1; |
1475 | |
1936 | |
1476 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1937 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1477 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1938 | types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1); |
1478 | |
1939 | |
1479 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1940 | types_true = get_bool ("Types::Serialiser::true"); |
1480 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
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 */ |
1481 | } |
1947 | } |
1482 | |
1948 | |
1483 | PROTOTYPES: DISABLE |
1949 | PROTOTYPES: DISABLE |
1484 | |
1950 | |
1485 | void CLONE (...) |
1951 | void CLONE (...) |
1486 | CODE: |
1952 | CODE: |
1487 | json_stash = 0; |
1953 | json_stash = 0; |
1488 | json_boolean_stash = 0; |
1954 | types_boolean_stash = 0; |
1489 | |
1955 | |
1490 | void new (char *klass) |
1956 | void new (char *klass) |
1491 | PPCODE: |
1957 | PPCODE: |
1492 | { |
1958 | { |
1493 | SV *pv = NEWSV (0, sizeof (JSON)); |
1959 | SV *pv = NEWSV (0, sizeof (JSON)); |
1494 | SvPOK_only (pv); |
1960 | SvPOK_only (pv); |
1495 | Zero (SvPVX (pv), 1, JSON); |
1961 | json_init ((JSON *)SvPVX (pv)); |
1496 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1497 | XPUSHs (sv_2mortal (sv_bless ( |
1962 | XPUSHs (sv_2mortal (sv_bless ( |
1498 | newRV_noinc (pv), |
1963 | newRV_noinc (pv), |
1499 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1964 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1500 | ))); |
1965 | ))); |
1501 | } |
1966 | } |
… | |
… | |
1513 | allow_nonref = F_ALLOW_NONREF |
1978 | allow_nonref = F_ALLOW_NONREF |
1514 | shrink = F_SHRINK |
1979 | shrink = F_SHRINK |
1515 | allow_blessed = F_ALLOW_BLESSED |
1980 | allow_blessed = F_ALLOW_BLESSED |
1516 | convert_blessed = F_CONV_BLESSED |
1981 | convert_blessed = F_CONV_BLESSED |
1517 | relaxed = F_RELAXED |
1982 | relaxed = F_RELAXED |
|
|
1983 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
1984 | allow_tags = F_ALLOW_TAGS |
1518 | PPCODE: |
1985 | PPCODE: |
1519 | { |
1986 | { |
1520 | if (enable) |
1987 | if (enable) |
1521 | self->flags |= ix; |
1988 | self->flags |= ix; |
1522 | else |
1989 | else |
… | |
… | |
1537 | get_allow_nonref = F_ALLOW_NONREF |
2004 | get_allow_nonref = F_ALLOW_NONREF |
1538 | get_shrink = F_SHRINK |
2005 | get_shrink = F_SHRINK |
1539 | get_allow_blessed = F_ALLOW_BLESSED |
2006 | get_allow_blessed = F_ALLOW_BLESSED |
1540 | get_convert_blessed = F_CONV_BLESSED |
2007 | get_convert_blessed = F_CONV_BLESSED |
1541 | get_relaxed = F_RELAXED |
2008 | get_relaxed = F_RELAXED |
|
|
2009 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
2010 | get_allow_tags = F_ALLOW_TAGS |
1542 | PPCODE: |
2011 | PPCODE: |
1543 | XPUSHs (boolSV (self->flags & ix)); |
2012 | XPUSHs (boolSV (self->flags & ix)); |
1544 | |
2013 | |
1545 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
2014 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1546 | PPCODE: |
2015 | PPCODE: |
1547 | { |
2016 | self->max_depth = max_depth; |
1548 | UV log2 = 0; |
|
|
1549 | |
|
|
1550 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
|
|
1551 | |
|
|
1552 | while ((1UL << log2) < max_depth) |
|
|
1553 | ++log2; |
|
|
1554 | |
|
|
1555 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1556 | |
|
|
1557 | XPUSHs (ST (0)); |
2017 | XPUSHs (ST (0)); |
1558 | } |
|
|
1559 | |
2018 | |
1560 | U32 get_max_depth (JSON *self) |
2019 | U32 get_max_depth (JSON *self) |
1561 | CODE: |
2020 | CODE: |
1562 | RETVAL = DEC_DEPTH (self->flags); |
2021 | RETVAL = self->max_depth; |
1563 | OUTPUT: |
2022 | OUTPUT: |
1564 | RETVAL |
2023 | RETVAL |
1565 | |
2024 | |
1566 | void max_size (JSON *self, UV max_size = 0) |
2025 | void max_size (JSON *self, U32 max_size = 0) |
1567 | PPCODE: |
2026 | PPCODE: |
1568 | { |
2027 | self->max_size = max_size; |
1569 | UV log2 = 0; |
|
|
1570 | |
|
|
1571 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1572 | if (max_size == 1) max_size = 2; |
|
|
1573 | |
|
|
1574 | while ((1UL << log2) < max_size) |
|
|
1575 | ++log2; |
|
|
1576 | |
|
|
1577 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1578 | |
|
|
1579 | XPUSHs (ST (0)); |
2028 | XPUSHs (ST (0)); |
1580 | } |
|
|
1581 | |
2029 | |
1582 | int get_max_size (JSON *self) |
2030 | int get_max_size (JSON *self) |
1583 | CODE: |
2031 | CODE: |
1584 | RETVAL = DEC_SIZE (self->flags); |
2032 | RETVAL = self->max_size; |
1585 | OUTPUT: |
2033 | OUTPUT: |
1586 | RETVAL |
2034 | RETVAL |
1587 | |
2035 | |
1588 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
2036 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1589 | PPCODE: |
2037 | PPCODE: |
… | |
… | |
1595 | } |
2043 | } |
1596 | |
2044 | |
1597 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
2045 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
1598 | PPCODE: |
2046 | PPCODE: |
1599 | { |
2047 | { |
1600 | if (!self->cb_sk_object) |
2048 | if (!self->cb_sk_object) |
1601 | self->cb_sk_object = newHV (); |
2049 | self->cb_sk_object = newHV (); |
1602 | |
2050 | |
1603 | if (SvOK (cb)) |
2051 | if (SvOK (cb)) |
1604 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
2052 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
1605 | else |
2053 | else |
… | |
… | |
1616 | XPUSHs (ST (0)); |
2064 | XPUSHs (ST (0)); |
1617 | } |
2065 | } |
1618 | |
2066 | |
1619 | void encode (JSON *self, SV *scalar) |
2067 | void encode (JSON *self, SV *scalar) |
1620 | PPCODE: |
2068 | PPCODE: |
1621 | XPUSHs (encode_json (scalar, self)); |
2069 | PUTBACK; scalar = encode_json (scalar, self); SPAGAIN; |
|
|
2070 | XPUSHs (scalar); |
1622 | |
2071 | |
1623 | void decode (JSON *self, SV *jsonstr) |
2072 | void decode (JSON *self, SV *jsonstr) |
1624 | PPCODE: |
2073 | PPCODE: |
1625 | XPUSHs (decode_json (jsonstr, self, 0)); |
2074 | PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN; |
|
|
2075 | XPUSHs (jsonstr); |
1626 | |
2076 | |
1627 | void decode_prefix (JSON *self, SV *jsonstr) |
2077 | void decode_prefix (JSON *self, SV *jsonstr) |
1628 | PPCODE: |
2078 | PPCODE: |
1629 | { |
2079 | { |
1630 | UV offset; |
2080 | SV *sv; |
|
|
2081 | char *offset; |
|
|
2082 | PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN; |
1631 | EXTEND (SP, 2); |
2083 | EXTEND (SP, 2); |
1632 | PUSHs (decode_json (jsonstr, self, &offset)); |
2084 | PUSHs (sv); |
1633 | PUSHs (sv_2mortal (newSVuv (offset))); |
2085 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
2086 | } |
|
|
2087 | |
|
|
2088 | void 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 | |
|
|
2177 | SV *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 | |
|
|
2189 | void 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 | |
|
|
2201 | void 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; |
1634 | } |
2209 | } |
1635 | |
2210 | |
1636 | void DESTROY (JSON *self) |
2211 | void DESTROY (JSON *self) |
1637 | CODE: |
2212 | CODE: |
1638 | SvREFCNT_dec (self->cb_sk_object); |
2213 | SvREFCNT_dec (self->cb_sk_object); |
1639 | SvREFCNT_dec (self->cb_object); |
2214 | SvREFCNT_dec (self->cb_object); |
|
|
2215 | SvREFCNT_dec (self->incr_text); |
1640 | |
2216 | |
1641 | PROTOTYPES: ENABLE |
2217 | PROTOTYPES: ENABLE |
1642 | |
2218 | |
1643 | void encode_json (SV *scalar) |
2219 | void encode_json (SV *scalar) |
1644 | PPCODE: |
2220 | PPCODE: |
1645 | { |
2221 | { |
1646 | JSON json = { F_DEFAULT | F_UTF8 }; |
2222 | JSON json; |
1647 | XPUSHs (encode_json (scalar, &json)); |
2223 | json_init (&json); |
|
|
2224 | json.flags |= F_UTF8; |
|
|
2225 | PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN; |
|
|
2226 | XPUSHs (scalar); |
1648 | } |
2227 | } |
1649 | |
2228 | |
1650 | void decode_json (SV *jsonstr) |
2229 | void decode_json (SV *jsonstr) |
1651 | PPCODE: |
2230 | PPCODE: |
1652 | { |
2231 | { |
1653 | JSON json = { F_DEFAULT | F_UTF8 }; |
2232 | JSON json; |
1654 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2233 | json_init (&json); |
|
|
2234 | json.flags |= F_UTF8; |
|
|
2235 | PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN; |
|
|
2236 | XPUSHs (jsonstr); |
1655 | } |
2237 | } |
1656 | |
2238 | |