… | |
… | |
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 | // three extra for rounding, sign, and end of string |
|
|
23 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
20 | |
24 | |
21 | #define F_ASCII 0x00000001UL |
25 | #define F_ASCII 0x00000001UL |
22 | #define F_LATIN1 0x00000002UL |
26 | #define F_LATIN1 0x00000002UL |
23 | #define F_UTF8 0x00000004UL |
27 | #define F_UTF8 0x00000004UL |
24 | #define F_INDENT 0x00000008UL |
28 | #define F_INDENT 0x00000008UL |
… | |
… | |
28 | #define F_ALLOW_NONREF 0x00000100UL |
32 | #define F_ALLOW_NONREF 0x00000100UL |
29 | #define F_SHRINK 0x00000200UL |
33 | #define F_SHRINK 0x00000200UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
34 | #define F_ALLOW_BLESSED 0x00000400UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
35 | #define F_CONV_BLESSED 0x00000800UL |
32 | #define F_RELAXED 0x00001000UL |
36 | #define F_RELAXED 0x00001000UL |
33 | |
37 | #define F_ALLOW_UNKNOWN 0x00002000UL |
34 | #define F_MAXDEPTH 0xf8000000UL |
|
|
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 |
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
39 | |
39 | |
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 |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
44 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
45 | |
41 | |
46 | #define INIT_SIZE 32 // initial scalar size to be allocated |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
47 | #define INDENT_STEP 3 // spaces per indentation level |
43 | #define INDENT_STEP 3 // spaces per indentation level |
48 | |
44 | |
49 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
45 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
|
|
46 | |
|
|
47 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
50 | |
48 | |
51 | #define SB do { |
49 | #define SB do { |
52 | #define SE } while (0) |
50 | #define SE } while (0) |
53 | |
51 | |
54 | #if __GNUC__ >= 3 |
52 | #if __GNUC__ >= 3 |
… | |
… | |
64 | |
62 | |
65 | #define IN_RANGE_INC(type,val,beg,end) \ |
63 | #define IN_RANGE_INC(type,val,beg,end) \ |
66 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
64 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
67 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
65 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
68 | |
66 | |
|
|
67 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
68 | |
69 | #ifdef USE_ITHREADS |
69 | #ifdef USE_ITHREADS |
70 | # define JSON_SLOW 1 |
70 | # define JSON_SLOW 1 |
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
72 | #else |
72 | #else |
73 | # define JSON_SLOW 0 |
73 | # define JSON_SLOW 0 |
… | |
… | |
75 | #endif |
75 | #endif |
76 | |
76 | |
77 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
77 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
78 | static SV *json_true, *json_false; |
78 | static SV *json_true, *json_false; |
79 | |
79 | |
|
|
80 | enum { |
|
|
81 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
82 | INCR_M_STR, // inside string |
|
|
83 | INCR_M_BS, // inside backslash |
|
|
84 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
85 | INCR_M_C1, // inside comment in other places |
|
|
86 | INCR_M_JSON // outside anything, count nesting |
|
|
87 | }; |
|
|
88 | |
|
|
89 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
90 | |
80 | typedef struct { |
91 | typedef struct { |
81 | U32 flags; |
92 | U32 flags; |
|
|
93 | U32 max_depth; |
|
|
94 | STRLEN max_size; |
|
|
95 | |
82 | SV *cb_object; |
96 | SV *cb_object; |
83 | HV *cb_sk_object; |
97 | HV *cb_sk_object; |
|
|
98 | |
|
|
99 | // for the incremental parser |
|
|
100 | SV *incr_text; // the source text so far |
|
|
101 | STRLEN incr_pos; // the current offset into the text |
|
|
102 | int incr_nest; // {[]}-nesting level |
|
|
103 | unsigned char incr_mode; |
84 | } JSON; |
104 | } JSON; |
|
|
105 | |
|
|
106 | INLINE void |
|
|
107 | json_init (JSON *json) |
|
|
108 | { |
|
|
109 | Zero (json, 1, JSON); |
|
|
110 | json->max_depth = 512; |
|
|
111 | } |
85 | |
112 | |
86 | ///////////////////////////////////////////////////////////////////////////// |
113 | ///////////////////////////////////////////////////////////////////////////// |
87 | // utility functions |
114 | // utility functions |
88 | |
115 | |
|
|
116 | INLINE SV * |
|
|
117 | get_bool (const char *name) |
|
|
118 | { |
|
|
119 | SV *sv = get_sv (name, 1); |
|
|
120 | |
|
|
121 | SvREADONLY_on (sv); |
|
|
122 | SvREADONLY_on (SvRV (sv)); |
|
|
123 | |
|
|
124 | return sv; |
|
|
125 | } |
|
|
126 | |
89 | INLINE void |
127 | INLINE void |
90 | shrink (SV *sv) |
128 | shrink (SV *sv) |
91 | { |
129 | { |
92 | sv_utf8_downgrade (sv, 1); |
130 | sv_utf8_downgrade (sv, 1); |
|
|
131 | |
93 | if (SvLEN (sv) > SvCUR (sv) + 1) |
132 | if (SvLEN (sv) > SvCUR (sv) + 1) |
94 | { |
133 | { |
95 | #ifdef SvPV_shrink_to_cur |
134 | #ifdef SvPV_shrink_to_cur |
96 | SvPV_shrink_to_cur (sv); |
135 | SvPV_shrink_to_cur (sv); |
97 | #elif defined (SvPV_renew) |
136 | #elif defined (SvPV_renew) |
… | |
… | |
141 | *s++ = 0x80 | ( ch & 0x3f); |
180 | *s++ = 0x80 | ( ch & 0x3f); |
142 | |
181 | |
143 | return s; |
182 | return s; |
144 | } |
183 | } |
145 | |
184 | |
|
|
185 | // convert offset pointer to character index, sv must be string |
|
|
186 | static STRLEN |
|
|
187 | ptr_to_index (SV *sv, char *offset) |
|
|
188 | { |
|
|
189 | return SvUTF8 (sv) |
|
|
190 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
191 | : offset - SvPVX (sv); |
|
|
192 | } |
|
|
193 | |
|
|
194 | ///////////////////////////////////////////////////////////////////////////// |
|
|
195 | // fp hell |
|
|
196 | |
|
|
197 | // scan a group of digits, and a trailing exponent |
|
|
198 | static void |
|
|
199 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
200 | { |
|
|
201 | UV uaccum = 0; |
|
|
202 | int eaccum = 0; |
|
|
203 | |
|
|
204 | // if we recurse too deep, skip all remaining digits |
|
|
205 | // to avoid a stack overflow attack |
|
|
206 | if (expect_false (--maxdepth <= 0)) |
|
|
207 | while (((U8)*s - '0') < 10) |
|
|
208 | ++s; |
|
|
209 | |
|
|
210 | for (;;) |
|
|
211 | { |
|
|
212 | U8 dig = (U8)*s - '0'; |
|
|
213 | |
|
|
214 | if (expect_false (dig >= 10)) |
|
|
215 | { |
|
|
216 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
217 | { |
|
|
218 | ++s; |
|
|
219 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
220 | } |
|
|
221 | else if ((dig | ' ') == 'e' - '0') |
|
|
222 | { |
|
|
223 | int exp2 = 0; |
|
|
224 | int neg = 0; |
|
|
225 | |
|
|
226 | ++s; |
|
|
227 | |
|
|
228 | if (*s == '-') |
|
|
229 | { |
|
|
230 | ++s; |
|
|
231 | neg = 1; |
|
|
232 | } |
|
|
233 | else if (*s == '+') |
|
|
234 | ++s; |
|
|
235 | |
|
|
236 | while ((dig = (U8)*s - '0') < 10) |
|
|
237 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
238 | |
|
|
239 | *expo += neg ? -exp2 : exp2; |
|
|
240 | } |
|
|
241 | |
|
|
242 | break; |
|
|
243 | } |
|
|
244 | |
|
|
245 | ++s; |
|
|
246 | |
|
|
247 | uaccum = uaccum * 10 + dig; |
|
|
248 | ++eaccum; |
|
|
249 | |
|
|
250 | // if we have too many digits, then recurse for more |
|
|
251 | // we actually do this for rather few digits |
|
|
252 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
253 | { |
|
|
254 | if (postdp) *expo -= eaccum; |
|
|
255 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
256 | if (postdp) *expo += eaccum; |
|
|
257 | |
|
|
258 | break; |
|
|
259 | } |
|
|
260 | } |
|
|
261 | |
|
|
262 | // this relies greatly on the quality of the pow () |
|
|
263 | // implementation of the platform, but a good |
|
|
264 | // implementation is hard to beat. |
|
|
265 | // (IEEE 754 conformant ones are required to be exact) |
|
|
266 | if (postdp) *expo -= eaccum; |
|
|
267 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
268 | *expo += eaccum; |
|
|
269 | } |
|
|
270 | |
|
|
271 | static NV |
|
|
272 | json_atof (const char *s) |
|
|
273 | { |
|
|
274 | NV accum = 0.; |
|
|
275 | int expo = 0; |
|
|
276 | int neg = 0; |
|
|
277 | |
|
|
278 | if (*s == '-') |
|
|
279 | { |
|
|
280 | ++s; |
|
|
281 | neg = 1; |
|
|
282 | } |
|
|
283 | |
|
|
284 | // a recursion depth of ten gives us >>500 bits |
|
|
285 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
286 | |
|
|
287 | return neg ? -accum : accum; |
|
|
288 | } |
146 | ///////////////////////////////////////////////////////////////////////////// |
289 | ///////////////////////////////////////////////////////////////////////////// |
147 | // encoder |
290 | // encoder |
148 | |
291 | |
149 | // structure used for encoding JSON |
292 | // structure used for encoding JSON |
150 | typedef struct |
293 | typedef struct |
… | |
… | |
152 | char *cur; // SvPVX (sv) + current output position |
295 | char *cur; // SvPVX (sv) + current output position |
153 | char *end; // SvEND (sv) |
296 | char *end; // SvEND (sv) |
154 | SV *sv; // result scalar |
297 | SV *sv; // result scalar |
155 | JSON json; |
298 | JSON json; |
156 | U32 indent; // indentation level |
299 | U32 indent; // indentation level |
157 | U32 maxdepth; // max. indentation/recursion level |
|
|
158 | UV limit; // escape character values >= this value when encoding |
300 | UV limit; // escape character values >= this value when encoding |
159 | } enc_t; |
301 | } enc_t; |
160 | |
302 | |
161 | INLINE void |
303 | INLINE void |
162 | need (enc_t *enc, STRLEN len) |
304 | need (enc_t *enc, STRLEN len) |
163 | { |
305 | { |
164 | if (expect_false (enc->cur + len >= enc->end)) |
306 | if (expect_false (enc->cur + len >= enc->end)) |
165 | { |
307 | { |
166 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
308 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
167 | SvGROW (enc->sv, cur + len + 1); |
309 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
168 | enc->cur = SvPVX (enc->sv) + cur; |
310 | enc->cur = SvPVX (enc->sv) + cur; |
169 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
311 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
170 | } |
312 | } |
171 | } |
313 | } |
172 | |
314 | |
… | |
… | |
247 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
389 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
248 | enc->cur += 12; |
390 | enc->cur += 12; |
249 | } |
391 | } |
250 | else |
392 | else |
251 | { |
393 | { |
252 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
253 | need (enc, len += 5); |
394 | need (enc, len += 5); |
254 | *enc->cur++ = '\\'; |
395 | *enc->cur++ = '\\'; |
255 | *enc->cur++ = 'u'; |
396 | *enc->cur++ = 'u'; |
256 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
397 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
257 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
398 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
258 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
399 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
259 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
400 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
260 | } |
401 | } |
261 | |
402 | |
262 | str += clen; |
403 | str += clen; |
263 | } |
404 | } |
264 | else if (enc->json.flags & F_LATIN1) |
405 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
335 | static void |
476 | static void |
336 | encode_av (enc_t *enc, AV *av) |
477 | encode_av (enc_t *enc, AV *av) |
337 | { |
478 | { |
338 | int i, len = av_len (av); |
479 | int i, len = av_len (av); |
339 | |
480 | |
340 | if (enc->indent >= enc->maxdepth) |
481 | if (enc->indent >= enc->json.max_depth) |
341 | croak ("data structure too deep (hit recursion limit)"); |
482 | croak (ERR_NESTING_EXCEEDED); |
342 | |
483 | |
343 | encode_ch (enc, '['); |
484 | encode_ch (enc, '['); |
344 | |
485 | |
345 | if (len >= 0) |
486 | if (len >= 0) |
346 | { |
487 | { |
… | |
… | |
420 | |
561 | |
421 | static void |
562 | static void |
422 | encode_hv (enc_t *enc, HV *hv) |
563 | encode_hv (enc_t *enc, HV *hv) |
423 | { |
564 | { |
424 | HE *he; |
565 | HE *he; |
425 | int count; |
|
|
426 | |
566 | |
427 | if (enc->indent >= enc->maxdepth) |
567 | if (enc->indent >= enc->json.max_depth) |
428 | croak ("data structure too deep (hit recursion limit)"); |
568 | croak (ERR_NESTING_EXCEEDED); |
429 | |
569 | |
430 | encode_ch (enc, '{'); |
570 | encode_ch (enc, '{'); |
431 | |
571 | |
432 | // for canonical output we have to sort by keys first |
572 | // for canonical output we have to sort by keys first |
433 | // actually, this is mostly due to the stupid so-called |
573 | // actually, this is mostly due to the stupid so-called |
434 | // security workaround added somewhere in 5.8.x. |
574 | // security workaround added somewhere in 5.8.x |
435 | // that randomises hash orderings |
575 | // that randomises hash orderings |
436 | if (enc->json.flags & F_CANONICAL) |
576 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
437 | { |
577 | { |
438 | int count = hv_iterinit (hv); |
578 | int count = hv_iterinit (hv); |
439 | |
579 | |
440 | if (SvMAGICAL (hv)) |
580 | if (SvMAGICAL (hv)) |
441 | { |
581 | { |
… | |
… | |
614 | |
754 | |
615 | if (len == 1 && *pv == '1') |
755 | if (len == 1 && *pv == '1') |
616 | encode_str (enc, "true", 4, 0); |
756 | encode_str (enc, "true", 4, 0); |
617 | else if (len == 1 && *pv == '0') |
757 | else if (len == 1 && *pv == '0') |
618 | encode_str (enc, "false", 5, 0); |
758 | encode_str (enc, "false", 5, 0); |
|
|
759 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
760 | encode_str (enc, "null", 4, 0); |
619 | else |
761 | else |
620 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
762 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
621 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
763 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
622 | } |
764 | } |
|
|
765 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
766 | encode_str (enc, "null", 4, 0); |
623 | else |
767 | else |
624 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
768 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
625 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
769 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
626 | } |
770 | } |
627 | |
771 | |
… | |
… | |
648 | else if (SvIOKp (sv)) |
792 | else if (SvIOKp (sv)) |
649 | { |
793 | { |
650 | // we assume we can always read an IV as a UV and vice versa |
794 | // we assume we can always read an IV as a UV and vice versa |
651 | // we assume two's complement |
795 | // we assume two's complement |
652 | // we assume no aliasing issues in the union |
796 | // we assume no aliasing issues in the union |
653 | if (SvIsUV (sv) ? SvUVX (sv) > 59000 |
797 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
654 | : SvIVX (sv) > 59000 || SvIVX (sv) < -59000) |
798 | : 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 | { |
799 | { |
665 | // optimise the "small number case" |
800 | // optimise the "small number case" |
666 | // code will likely be branchless and use only a single multiplication |
801 | // code will likely be branchless and use only a single multiplication |
667 | // works for numbers up to 59074 |
802 | // works for numbers up to 59074 |
668 | I32 i = SvIVX (sv); |
803 | I32 i = SvIVX (sv); |
… | |
… | |
679 | |
814 | |
680 | // now output digit by digit, each time masking out the integer part |
815 | // 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, |
816 | // and multiplying by 5 while moving the decimal point one to the right, |
682 | // resulting in a net multiplication by 10. |
817 | // resulting in a net multiplication by 10. |
683 | // we always write the digit to memory but conditionally increment |
818 | // we always write the digit to memory but conditionally increment |
684 | // the pointer, to ease the usage of conditional move instructions. |
819 | // 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; |
820 | 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; |
821 | 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; |
822 | 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; |
823 | 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' |
824 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
690 | } |
825 | } |
|
|
826 | else |
|
|
827 | { |
|
|
828 | // large integer, use the (rather slow) snprintf way. |
|
|
829 | need (enc, IVUV_MAXCHARS); |
|
|
830 | enc->cur += |
|
|
831 | SvIsUV(sv) |
|
|
832 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
833 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
834 | } |
691 | } |
835 | } |
692 | else if (SvROK (sv)) |
836 | else if (SvROK (sv)) |
693 | encode_rv (enc, SvRV (sv)); |
837 | encode_rv (enc, SvRV (sv)); |
694 | else if (!SvOK (sv)) |
838 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
695 | encode_str (enc, "null", 4, 0); |
839 | encode_str (enc, "null", 4, 0); |
696 | else |
840 | else |
697 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
841 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
698 | SvPV_nolen (sv), SvFLAGS (sv)); |
842 | SvPV_nolen (sv), SvFLAGS (sv)); |
699 | } |
843 | } |
… | |
… | |
709 | enc.json = *json; |
853 | enc.json = *json; |
710 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
854 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
711 | enc.cur = SvPVX (enc.sv); |
855 | enc.cur = SvPVX (enc.sv); |
712 | enc.end = SvEND (enc.sv); |
856 | enc.end = SvEND (enc.sv); |
713 | enc.indent = 0; |
857 | enc.indent = 0; |
714 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
|
|
715 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
858 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
716 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
859 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
717 | : 0x110000UL; |
860 | : 0x110000UL; |
718 | |
861 | |
719 | SvPOK_only (enc.sv); |
862 | SvPOK_only (enc.sv); |
720 | encode_sv (&enc, scalar); |
863 | encode_sv (&enc, scalar); |
|
|
864 | encode_nl (&enc); |
721 | |
865 | |
722 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
866 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
723 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
867 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
724 | |
868 | |
725 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
869 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
786 | if (*dec->cur != ch) \ |
930 | if (*dec->cur != ch) \ |
787 | ERR (# ch " expected"); \ |
931 | ERR (# ch " expected"); \ |
788 | ++dec->cur; \ |
932 | ++dec->cur; \ |
789 | SE |
933 | SE |
790 | |
934 | |
791 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
935 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
792 | #define DEC_DEC_DEPTH --dec->depth |
936 | #define DEC_DEC_DEPTH --dec->depth |
793 | |
937 | |
794 | static SV *decode_sv (dec_t *dec); |
938 | static SV *decode_sv (dec_t *dec); |
795 | |
939 | |
796 | static signed char decode_hexdigit[256]; |
940 | static signed char decode_hexdigit[256]; |
… | |
… | |
905 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
1049 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
906 | *cur++ = ch; |
1050 | *cur++ = ch; |
907 | else if (ch >= 0x80) |
1051 | else if (ch >= 0x80) |
908 | { |
1052 | { |
909 | STRLEN clen; |
1053 | STRLEN clen; |
910 | UV uch; |
|
|
911 | |
1054 | |
912 | --dec_cur; |
1055 | --dec_cur; |
913 | |
1056 | |
914 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1057 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
915 | if (clen == (STRLEN)-1) |
1058 | if (clen == (STRLEN)-1) |
916 | ERR ("malformed UTF-8 character in JSON string"); |
1059 | ERR ("malformed UTF-8 character in JSON string"); |
917 | |
1060 | |
918 | do |
1061 | do |
919 | *cur++ = *dec_cur++; |
1062 | *cur++ = *dec_cur++; |
… | |
… | |
936 | { |
1079 | { |
937 | STRLEN len = cur - buf; |
1080 | STRLEN len = cur - buf; |
938 | |
1081 | |
939 | if (sv) |
1082 | if (sv) |
940 | { |
1083 | { |
941 | SvGROW (sv, SvCUR (sv) + len + 1); |
1084 | STRLEN cur = SvCUR (sv); |
|
|
1085 | |
|
|
1086 | if (SvLEN (sv) <= cur + len) |
|
|
1087 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1088 | |
942 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1089 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
943 | SvCUR_set (sv, SvCUR (sv) + len); |
1090 | SvCUR_set (sv, SvCUR (sv) + len); |
944 | } |
1091 | } |
945 | else |
1092 | else |
946 | sv = newSVpvn (buf, len); |
1093 | sv = newSVpvn (buf, len); |
… | |
… | |
1037 | |
1184 | |
1038 | // special case the rather common 1..5-digit-int case |
1185 | // special case the rather common 1..5-digit-int case |
1039 | if (*start == '-') |
1186 | if (*start == '-') |
1040 | switch (len) |
1187 | switch (len) |
1041 | { |
1188 | { |
1042 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1189 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
1043 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1190 | 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)); |
1191 | 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)); |
1192 | 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)); |
1193 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1047 | } |
1194 | } |
1048 | else |
1195 | else |
1049 | switch (len) |
1196 | switch (len) |
1050 | { |
1197 | { |
1051 | case 1: return newSViv ( start [0] - '0' * 1); |
1198 | case 1: return newSViv ( start [0] - '0' * 1); |
1052 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1199 | 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); |
1200 | 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); |
1201 | 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); |
1202 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1056 | } |
1203 | } |
1057 | |
1204 | |
1058 | { |
1205 | { |
1059 | UV uv; |
1206 | UV uv; |
1060 | int numtype = grok_number (start, len, &uv); |
1207 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1069 | } |
1216 | } |
1070 | |
1217 | |
1071 | len -= *start == '-' ? 1 : 0; |
1218 | len -= *start == '-' ? 1 : 0; |
1072 | |
1219 | |
1073 | // does not fit into IV or UV, try NV |
1220 | // does not fit into IV or UV, try NV |
1074 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
1221 | 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 |
1222 | // fits into NV without loss of precision |
1080 | return newSVnv (Atof (start)); |
1223 | return newSVnv (json_atof (start)); |
1081 | |
1224 | |
1082 | // everything else fails, convert it to a string |
1225 | // everything else fails, convert it to a string |
1083 | return newSVpvn (start, dec->cur - start); |
1226 | return newSVpvn (start, dec->cur - start); |
1084 | } |
1227 | } |
1085 | |
1228 | |
1086 | // loss of precision here |
1229 | // loss of precision here |
1087 | return newSVnv (Atof (start)); |
1230 | return newSVnv (json_atof (start)); |
1088 | |
1231 | |
1089 | fail: |
1232 | fail: |
1090 | return 0; |
1233 | return 0; |
1091 | } |
1234 | } |
1092 | |
1235 | |
… | |
… | |
1168 | char *p = dec->cur; |
1311 | char *p = dec->cur; |
1169 | char *e = p + 24; // only try up to 24 bytes |
1312 | char *e = p + 24; // only try up to 24 bytes |
1170 | |
1313 | |
1171 | for (;;) |
1314 | for (;;) |
1172 | { |
1315 | { |
1173 | // the >= 0x80 is true on most architectures |
1316 | // the >= 0x80 is false on most architectures |
1174 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1317 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1175 | { |
1318 | { |
1176 | // slow path, back up and use decode_str |
1319 | // slow path, back up and use decode_str |
1177 | SV *key = decode_str (dec); |
1320 | SV *key = decode_str (dec); |
1178 | if (!key) |
1321 | if (!key) |
… | |
… | |
1261 | dSP; |
1404 | dSP; |
1262 | int count; |
1405 | int count; |
1263 | |
1406 | |
1264 | ENTER; SAVETMPS; PUSHMARK (SP); |
1407 | ENTER; SAVETMPS; PUSHMARK (SP); |
1265 | XPUSHs (HeVAL (he)); |
1408 | XPUSHs (HeVAL (he)); |
|
|
1409 | sv_2mortal (sv); |
1266 | |
1410 | |
1267 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1411 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1268 | |
1412 | |
1269 | if (count == 1) |
1413 | if (count == 1) |
1270 | { |
1414 | { |
1271 | sv = newSVsv (POPs); |
1415 | sv = newSVsv (POPs); |
1272 | FREETMPS; LEAVE; |
1416 | FREETMPS; LEAVE; |
1273 | return sv; |
1417 | return sv; |
1274 | } |
1418 | } |
1275 | |
1419 | |
|
|
1420 | SvREFCNT_inc (sv); |
1276 | FREETMPS; LEAVE; |
1421 | FREETMPS; LEAVE; |
1277 | } |
1422 | } |
1278 | } |
1423 | } |
1279 | |
1424 | |
1280 | if (dec->json.cb_object) |
1425 | if (dec->json.cb_object) |
… | |
… | |
1326 | case 't': |
1471 | case 't': |
1327 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1472 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1328 | { |
1473 | { |
1329 | dec->cur += 4; |
1474 | dec->cur += 4; |
1330 | #if JSON_SLOW |
1475 | #if JSON_SLOW |
1331 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
1476 | json_true = get_bool ("JSON::XS::true"); |
1332 | #endif |
1477 | #endif |
1333 | return SvREFCNT_inc (json_true); |
1478 | return newSVsv (json_true); |
1334 | } |
1479 | } |
1335 | else |
1480 | else |
1336 | ERR ("'true' expected"); |
1481 | ERR ("'true' expected"); |
1337 | |
1482 | |
1338 | break; |
1483 | break; |
… | |
… | |
1340 | case 'f': |
1485 | case 'f': |
1341 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1486 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1342 | { |
1487 | { |
1343 | dec->cur += 5; |
1488 | dec->cur += 5; |
1344 | #if JSON_SLOW |
1489 | #if JSON_SLOW |
1345 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1490 | json_false = get_bool ("JSON::XS::false"); |
1346 | #endif |
1491 | #endif |
1347 | return SvREFCNT_inc (json_false); |
1492 | return newSVsv (json_false); |
1348 | } |
1493 | } |
1349 | else |
1494 | else |
1350 | ERR ("'false' expected"); |
1495 | ERR ("'false' expected"); |
1351 | |
1496 | |
1352 | break; |
1497 | break; |
… | |
… | |
1370 | fail: |
1515 | fail: |
1371 | return 0; |
1516 | return 0; |
1372 | } |
1517 | } |
1373 | |
1518 | |
1374 | static SV * |
1519 | static SV * |
1375 | decode_json (SV *string, JSON *json, UV *offset_return) |
1520 | decode_json (SV *string, JSON *json, char **offset_return) |
1376 | { |
1521 | { |
1377 | dec_t dec; |
1522 | dec_t dec; |
1378 | UV offset; |
|
|
1379 | SV *sv; |
1523 | SV *sv; |
1380 | |
1524 | |
|
|
1525 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1526 | * will perl ignore the magic in subsequent accesses. |
|
|
1527 | * also make a copy of non-PV values, to get them into a clean |
|
|
1528 | * state (SvPV should do that, but it's buggy, see below). |
|
|
1529 | */ |
1381 | SvGETMAGIC (string); |
1530 | /*SvGETMAGIC (string);*/ |
|
|
1531 | if (SvMAGICAL (string) || !SvPOK (string)) |
|
|
1532 | string = sv_2mortal (newSVsv (string)); |
|
|
1533 | |
1382 | SvUPGRADE (string, SVt_PV); |
1534 | SvUPGRADE (string, SVt_PV); |
1383 | |
1535 | |
1384 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1536 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1537 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1538 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1539 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1540 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1541 | * and hope for the best. |
|
|
1542 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1543 | * assertion business is seriously broken, try yet another workaround |
|
|
1544 | * for the broken -DDEBUGGING. |
|
|
1545 | */ |
|
|
1546 | { |
|
|
1547 | #ifdef DEBUGGING |
|
|
1548 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1549 | #else |
|
|
1550 | STRLEN offset = SvCUR (string); |
|
|
1551 | #endif |
|
|
1552 | |
|
|
1553 | 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", |
1554 | 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)); |
1555 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1556 | } |
1387 | |
1557 | |
1388 | if (json->flags & F_UTF8) |
1558 | if (DECODE_WANTS_OCTETS (json)) |
1389 | sv_utf8_downgrade (string, 0); |
1559 | sv_utf8_downgrade (string, 0); |
1390 | else |
1560 | else |
1391 | sv_utf8_upgrade (string); |
1561 | sv_utf8_upgrade (string); |
1392 | |
1562 | |
1393 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1563 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1394 | |
1564 | |
1395 | dec.json = *json; |
1565 | dec.json = *json; |
1396 | dec.cur = SvPVX (string); |
1566 | dec.cur = SvPVX (string); |
1397 | dec.end = SvEND (string); |
1567 | dec.end = SvEND (string); |
1398 | dec.err = 0; |
1568 | dec.err = 0; |
1399 | dec.depth = 0; |
1569 | dec.depth = 0; |
1400 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1401 | |
1570 | |
1402 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1571 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1403 | dec.json.flags |= F_HOOK; |
1572 | dec.json.flags |= F_HOOK; |
1404 | |
1573 | |
1405 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1574 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1406 | |
1575 | |
1407 | decode_ws (&dec); |
1576 | decode_ws (&dec); |
1408 | sv = decode_sv (&dec); |
1577 | sv = decode_sv (&dec); |
|
|
1578 | |
|
|
1579 | if (offset_return) |
|
|
1580 | *offset_return = dec.cur; |
1409 | |
1581 | |
1410 | if (!(offset_return || !sv)) |
1582 | if (!(offset_return || !sv)) |
1411 | { |
1583 | { |
1412 | // check for trailing garbage |
1584 | // check for trailing garbage |
1413 | decode_ws (&dec); |
1585 | decode_ws (&dec); |
… | |
… | |
1416 | { |
1588 | { |
1417 | dec.err = "garbage after JSON object"; |
1589 | dec.err = "garbage after JSON object"; |
1418 | SvREFCNT_dec (sv); |
1590 | SvREFCNT_dec (sv); |
1419 | sv = 0; |
1591 | sv = 0; |
1420 | } |
1592 | } |
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 | } |
1593 | } |
1432 | |
1594 | |
1433 | if (!sv) |
1595 | if (!sv) |
1434 | { |
1596 | { |
1435 | SV *uni = sv_newmortal (); |
1597 | SV *uni = sv_newmortal (); |
… | |
… | |
1441 | SAVEVPTR (PL_curcop); |
1603 | SAVEVPTR (PL_curcop); |
1442 | PL_curcop = &cop; |
1604 | PL_curcop = &cop; |
1443 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1605 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1444 | LEAVE; |
1606 | LEAVE; |
1445 | |
1607 | |
1446 | croak ("%s, at character offset %d [\"%s\"]", |
1608 | croak ("%s, at character offset %d (before \"%s\")", |
1447 | dec.err, |
1609 | dec.err, |
1448 | (int)offset, |
1610 | ptr_to_index (string, dec.cur), |
1449 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1611 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1450 | } |
1612 | } |
1451 | |
1613 | |
1452 | sv = sv_2mortal (sv); |
1614 | sv = sv_2mortal (sv); |
1453 | |
1615 | |
1454 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1616 | 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)"); |
1617 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1456 | |
1618 | |
1457 | return sv; |
1619 | return sv; |
|
|
1620 | } |
|
|
1621 | |
|
|
1622 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1623 | // incremental parser |
|
|
1624 | |
|
|
1625 | static void |
|
|
1626 | incr_parse (JSON *self) |
|
|
1627 | { |
|
|
1628 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1629 | |
|
|
1630 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1631 | // but this would make it slower, so... |
|
|
1632 | |
|
|
1633 | for (;;) |
|
|
1634 | { |
|
|
1635 | //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 |
|
|
1636 | switch (self->incr_mode) |
|
|
1637 | { |
|
|
1638 | // only used for initial whitespace skipping |
|
|
1639 | case INCR_M_WS: |
|
|
1640 | for (;;) |
|
|
1641 | { |
|
|
1642 | if (*p > 0x20) |
|
|
1643 | { |
|
|
1644 | if (*p == '#') |
|
|
1645 | { |
|
|
1646 | self->incr_mode = INCR_M_C0; |
|
|
1647 | goto incr_m_c; |
|
|
1648 | } |
|
|
1649 | else |
|
|
1650 | { |
|
|
1651 | self->incr_mode = INCR_M_JSON; |
|
|
1652 | goto incr_m_json; |
|
|
1653 | } |
|
|
1654 | } |
|
|
1655 | else if (!*p) |
|
|
1656 | goto interrupt; |
|
|
1657 | |
|
|
1658 | ++p; |
|
|
1659 | } |
|
|
1660 | |
|
|
1661 | // skip a single char inside a string (for \\-processing) |
|
|
1662 | case INCR_M_BS: |
|
|
1663 | if (!*p) |
|
|
1664 | goto interrupt; |
|
|
1665 | |
|
|
1666 | ++p; |
|
|
1667 | self->incr_mode = INCR_M_STR; |
|
|
1668 | goto incr_m_str; |
|
|
1669 | |
|
|
1670 | // inside #-style comments |
|
|
1671 | case INCR_M_C0: |
|
|
1672 | case INCR_M_C1: |
|
|
1673 | incr_m_c: |
|
|
1674 | for (;;) |
|
|
1675 | { |
|
|
1676 | if (*p == '\n') |
|
|
1677 | { |
|
|
1678 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1679 | break; |
|
|
1680 | } |
|
|
1681 | else if (!*p) |
|
|
1682 | goto interrupt; |
|
|
1683 | |
|
|
1684 | ++p; |
|
|
1685 | } |
|
|
1686 | |
|
|
1687 | break; |
|
|
1688 | |
|
|
1689 | // inside a string |
|
|
1690 | case INCR_M_STR: |
|
|
1691 | incr_m_str: |
|
|
1692 | for (;;) |
|
|
1693 | { |
|
|
1694 | if (*p == '"') |
|
|
1695 | { |
|
|
1696 | ++p; |
|
|
1697 | self->incr_mode = INCR_M_JSON; |
|
|
1698 | |
|
|
1699 | if (!self->incr_nest) |
|
|
1700 | goto interrupt; |
|
|
1701 | |
|
|
1702 | goto incr_m_json; |
|
|
1703 | } |
|
|
1704 | else if (*p == '\\') |
|
|
1705 | { |
|
|
1706 | ++p; // "virtually" consumes character after \ |
|
|
1707 | |
|
|
1708 | if (!*p) // if at end of string we have to switch modes |
|
|
1709 | { |
|
|
1710 | self->incr_mode = INCR_M_BS; |
|
|
1711 | goto interrupt; |
|
|
1712 | } |
|
|
1713 | } |
|
|
1714 | else if (!*p) |
|
|
1715 | goto interrupt; |
|
|
1716 | |
|
|
1717 | ++p; |
|
|
1718 | } |
|
|
1719 | |
|
|
1720 | // after initial ws, outside string |
|
|
1721 | case INCR_M_JSON: |
|
|
1722 | incr_m_json: |
|
|
1723 | for (;;) |
|
|
1724 | { |
|
|
1725 | switch (*p++) |
|
|
1726 | { |
|
|
1727 | case 0: |
|
|
1728 | --p; |
|
|
1729 | goto interrupt; |
|
|
1730 | |
|
|
1731 | case 0x09: |
|
|
1732 | case 0x0a: |
|
|
1733 | case 0x0d: |
|
|
1734 | case 0x20: |
|
|
1735 | if (!self->incr_nest) |
|
|
1736 | { |
|
|
1737 | --p; // do not eat the whitespace, let the next round do it |
|
|
1738 | goto interrupt; |
|
|
1739 | } |
|
|
1740 | break; |
|
|
1741 | |
|
|
1742 | case '"': |
|
|
1743 | self->incr_mode = INCR_M_STR; |
|
|
1744 | goto incr_m_str; |
|
|
1745 | |
|
|
1746 | case '[': |
|
|
1747 | case '{': |
|
|
1748 | if (++self->incr_nest > self->max_depth) |
|
|
1749 | croak (ERR_NESTING_EXCEEDED); |
|
|
1750 | break; |
|
|
1751 | |
|
|
1752 | case ']': |
|
|
1753 | case '}': |
|
|
1754 | if (--self->incr_nest <= 0) |
|
|
1755 | goto interrupt; |
|
|
1756 | break; |
|
|
1757 | |
|
|
1758 | case '#': |
|
|
1759 | self->incr_mode = INCR_M_C1; |
|
|
1760 | goto incr_m_c; |
|
|
1761 | } |
|
|
1762 | } |
|
|
1763 | } |
|
|
1764 | |
|
|
1765 | modechange: |
|
|
1766 | ; |
|
|
1767 | } |
|
|
1768 | |
|
|
1769 | interrupt: |
|
|
1770 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1771 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1772 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1458 | } |
1773 | } |
1459 | |
1774 | |
1460 | ///////////////////////////////////////////////////////////////////////////// |
1775 | ///////////////////////////////////////////////////////////////////////////// |
1461 | // XS interface functions |
1776 | // XS interface functions |
1462 | |
1777 | |
… | |
… | |
1474 | : -1; |
1789 | : -1; |
1475 | |
1790 | |
1476 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1791 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1477 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1792 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1478 | |
1793 | |
1479 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1794 | json_true = get_bool ("JSON::XS::true"); |
1480 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1795 | json_false = get_bool ("JSON::XS::false"); |
|
|
1796 | |
|
|
1797 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1481 | } |
1798 | } |
1482 | |
1799 | |
1483 | PROTOTYPES: DISABLE |
1800 | PROTOTYPES: DISABLE |
1484 | |
1801 | |
1485 | void CLONE (...) |
1802 | void CLONE (...) |
… | |
… | |
1490 | void new (char *klass) |
1807 | void new (char *klass) |
1491 | PPCODE: |
1808 | PPCODE: |
1492 | { |
1809 | { |
1493 | SV *pv = NEWSV (0, sizeof (JSON)); |
1810 | SV *pv = NEWSV (0, sizeof (JSON)); |
1494 | SvPOK_only (pv); |
1811 | SvPOK_only (pv); |
1495 | Zero (SvPVX (pv), 1, JSON); |
1812 | json_init ((JSON *)SvPVX (pv)); |
1496 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1497 | XPUSHs (sv_2mortal (sv_bless ( |
1813 | XPUSHs (sv_2mortal (sv_bless ( |
1498 | newRV_noinc (pv), |
1814 | newRV_noinc (pv), |
1499 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1815 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1500 | ))); |
1816 | ))); |
1501 | } |
1817 | } |
… | |
… | |
1513 | allow_nonref = F_ALLOW_NONREF |
1829 | allow_nonref = F_ALLOW_NONREF |
1514 | shrink = F_SHRINK |
1830 | shrink = F_SHRINK |
1515 | allow_blessed = F_ALLOW_BLESSED |
1831 | allow_blessed = F_ALLOW_BLESSED |
1516 | convert_blessed = F_CONV_BLESSED |
1832 | convert_blessed = F_CONV_BLESSED |
1517 | relaxed = F_RELAXED |
1833 | relaxed = F_RELAXED |
|
|
1834 | allow_unknown = F_ALLOW_UNKNOWN |
1518 | PPCODE: |
1835 | PPCODE: |
1519 | { |
1836 | { |
1520 | if (enable) |
1837 | if (enable) |
1521 | self->flags |= ix; |
1838 | self->flags |= ix; |
1522 | else |
1839 | else |
… | |
… | |
1537 | get_allow_nonref = F_ALLOW_NONREF |
1854 | get_allow_nonref = F_ALLOW_NONREF |
1538 | get_shrink = F_SHRINK |
1855 | get_shrink = F_SHRINK |
1539 | get_allow_blessed = F_ALLOW_BLESSED |
1856 | get_allow_blessed = F_ALLOW_BLESSED |
1540 | get_convert_blessed = F_CONV_BLESSED |
1857 | get_convert_blessed = F_CONV_BLESSED |
1541 | get_relaxed = F_RELAXED |
1858 | get_relaxed = F_RELAXED |
|
|
1859 | get_allow_unknown = F_ALLOW_UNKNOWN |
1542 | PPCODE: |
1860 | PPCODE: |
1543 | XPUSHs (boolSV (self->flags & ix)); |
1861 | XPUSHs (boolSV (self->flags & ix)); |
1544 | |
1862 | |
1545 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1863 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1546 | PPCODE: |
1864 | PPCODE: |
1547 | { |
1865 | 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)); |
1866 | XPUSHs (ST (0)); |
1558 | } |
|
|
1559 | |
1867 | |
1560 | U32 get_max_depth (JSON *self) |
1868 | U32 get_max_depth (JSON *self) |
1561 | CODE: |
1869 | CODE: |
1562 | RETVAL = DEC_DEPTH (self->flags); |
1870 | RETVAL = self->max_depth; |
1563 | OUTPUT: |
1871 | OUTPUT: |
1564 | RETVAL |
1872 | RETVAL |
1565 | |
1873 | |
1566 | void max_size (JSON *self, UV max_size = 0) |
1874 | void max_size (JSON *self, U32 max_size = 0) |
1567 | PPCODE: |
1875 | PPCODE: |
1568 | { |
1876 | 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)); |
1877 | XPUSHs (ST (0)); |
1580 | } |
|
|
1581 | |
1878 | |
1582 | int get_max_size (JSON *self) |
1879 | int get_max_size (JSON *self) |
1583 | CODE: |
1880 | CODE: |
1584 | RETVAL = DEC_SIZE (self->flags); |
1881 | RETVAL = self->max_size; |
1585 | OUTPUT: |
1882 | OUTPUT: |
1586 | RETVAL |
1883 | RETVAL |
1587 | |
1884 | |
1588 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1885 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1589 | PPCODE: |
1886 | PPCODE: |
… | |
… | |
1625 | XPUSHs (decode_json (jsonstr, self, 0)); |
1922 | XPUSHs (decode_json (jsonstr, self, 0)); |
1626 | |
1923 | |
1627 | void decode_prefix (JSON *self, SV *jsonstr) |
1924 | void decode_prefix (JSON *self, SV *jsonstr) |
1628 | PPCODE: |
1925 | PPCODE: |
1629 | { |
1926 | { |
1630 | UV offset; |
1927 | char *offset; |
1631 | EXTEND (SP, 2); |
1928 | EXTEND (SP, 2); |
1632 | PUSHs (decode_json (jsonstr, self, &offset)); |
1929 | PUSHs (decode_json (jsonstr, self, &offset)); |
1633 | PUSHs (sv_2mortal (newSVuv (offset))); |
1930 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1931 | } |
|
|
1932 | |
|
|
1933 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1934 | PPCODE: |
|
|
1935 | { |
|
|
1936 | if (!self->incr_text) |
|
|
1937 | self->incr_text = newSVpvn ("", 0); |
|
|
1938 | |
|
|
1939 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
1940 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
1941 | if (DECODE_WANTS_OCTETS (self)) |
|
|
1942 | { |
|
|
1943 | if (self->incr_pos) |
|
|
1944 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
1945 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
1946 | |
|
|
1947 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
1948 | } |
|
|
1949 | else |
|
|
1950 | { |
|
|
1951 | sv_utf8_upgrade (self->incr_text); |
|
|
1952 | |
|
|
1953 | if (self->incr_pos) |
|
|
1954 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1955 | - (U8 *)SvPVX (self->incr_text); |
|
|
1956 | } |
|
|
1957 | |
|
|
1958 | // append data, if any |
|
|
1959 | if (jsonstr) |
|
|
1960 | { |
|
|
1961 | /* make sure both strings have same encoding */ |
|
|
1962 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
|
|
1963 | if (SvUTF8 (jsonstr)) |
|
|
1964 | sv_utf8_downgrade (jsonstr, 0); |
|
|
1965 | else |
|
|
1966 | sv_utf8_upgrade (jsonstr); |
|
|
1967 | |
|
|
1968 | /* and then just blindly append */ |
|
|
1969 | { |
|
|
1970 | STRLEN len; |
|
|
1971 | const char *str = SvPV (jsonstr, len); |
|
|
1972 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1973 | |
|
|
1974 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1975 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1976 | |
|
|
1977 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1978 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1979 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1980 | } |
|
|
1981 | } |
|
|
1982 | |
|
|
1983 | if (GIMME_V != G_VOID) |
|
|
1984 | do |
|
|
1985 | { |
|
|
1986 | char *offset; |
|
|
1987 | |
|
|
1988 | if (!INCR_DONE (self)) |
|
|
1989 | { |
|
|
1990 | incr_parse (self); |
|
|
1991 | |
|
|
1992 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1993 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1994 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1995 | |
|
|
1996 | if (!INCR_DONE (self)) |
|
|
1997 | break; |
|
|
1998 | } |
|
|
1999 | |
|
|
2000 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
2001 | |
|
|
2002 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
2003 | self->incr_nest = 0; |
|
|
2004 | self->incr_mode = 0; |
|
|
2005 | |
|
|
2006 | sv_chop (self->incr_text, offset); |
|
|
2007 | } |
|
|
2008 | while (GIMME_V == G_ARRAY); |
|
|
2009 | } |
|
|
2010 | |
|
|
2011 | SV *incr_text (JSON *self) |
|
|
2012 | ATTRS: lvalue |
|
|
2013 | CODE: |
|
|
2014 | { |
|
|
2015 | if (self->incr_pos) |
|
|
2016 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
2017 | |
|
|
2018 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2019 | } |
|
|
2020 | OUTPUT: |
|
|
2021 | RETVAL |
|
|
2022 | |
|
|
2023 | void incr_skip (JSON *self) |
|
|
2024 | CODE: |
|
|
2025 | { |
|
|
2026 | if (self->incr_pos) |
|
|
2027 | { |
|
|
2028 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2029 | self->incr_pos = 0; |
|
|
2030 | self->incr_nest = 0; |
|
|
2031 | self->incr_mode = 0; |
|
|
2032 | } |
|
|
2033 | } |
|
|
2034 | |
|
|
2035 | void incr_reset (JSON *self) |
|
|
2036 | CODE: |
|
|
2037 | { |
|
|
2038 | SvREFCNT_dec (self->incr_text); |
|
|
2039 | self->incr_text = 0; |
|
|
2040 | self->incr_pos = 0; |
|
|
2041 | self->incr_nest = 0; |
|
|
2042 | self->incr_mode = 0; |
1634 | } |
2043 | } |
1635 | |
2044 | |
1636 | void DESTROY (JSON *self) |
2045 | void DESTROY (JSON *self) |
1637 | CODE: |
2046 | CODE: |
1638 | SvREFCNT_dec (self->cb_sk_object); |
2047 | SvREFCNT_dec (self->cb_sk_object); |
1639 | SvREFCNT_dec (self->cb_object); |
2048 | SvREFCNT_dec (self->cb_object); |
|
|
2049 | SvREFCNT_dec (self->incr_text); |
1640 | |
2050 | |
1641 | PROTOTYPES: ENABLE |
2051 | PROTOTYPES: ENABLE |
1642 | |
2052 | |
1643 | void encode_json (SV *scalar) |
2053 | void encode_json (SV *scalar) |
|
|
2054 | ALIAS: |
|
|
2055 | to_json_ = 0 |
|
|
2056 | encode_json = F_UTF8 |
1644 | PPCODE: |
2057 | PPCODE: |
1645 | { |
2058 | { |
1646 | JSON json = { F_DEFAULT | F_UTF8 }; |
2059 | JSON json; |
|
|
2060 | json_init (&json); |
|
|
2061 | json.flags |= ix; |
1647 | XPUSHs (encode_json (scalar, &json)); |
2062 | XPUSHs (encode_json (scalar, &json)); |
1648 | } |
2063 | } |
1649 | |
2064 | |
1650 | void decode_json (SV *jsonstr) |
2065 | void decode_json (SV *jsonstr) |
|
|
2066 | ALIAS: |
|
|
2067 | from_json_ = 0 |
|
|
2068 | decode_json = F_UTF8 |
1651 | PPCODE: |
2069 | PPCODE: |
1652 | { |
2070 | { |
1653 | JSON json = { F_DEFAULT | F_UTF8 }; |
2071 | JSON json; |
|
|
2072 | json_init (&json); |
|
|
2073 | json.flags |= ix; |
1654 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2074 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1655 | } |
2075 | } |
1656 | |
2076 | |