… | |
… | |
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 |
… | |
… | |
64 | |
60 | |
65 | #define IN_RANGE_INC(type,val,beg,end) \ |
61 | #define IN_RANGE_INC(type,val,beg,end) \ |
66 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
62 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
67 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
63 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
68 | |
64 | |
|
|
65 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
66 | |
69 | #ifdef USE_ITHREADS |
67 | #ifdef USE_ITHREADS |
70 | # define JSON_SLOW 1 |
68 | # define JSON_SLOW 1 |
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
69 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
72 | #else |
70 | #else |
73 | # define JSON_SLOW 0 |
71 | # define JSON_SLOW 0 |
… | |
… | |
75 | #endif |
73 | #endif |
76 | |
74 | |
77 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
75 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
78 | static SV *json_true, *json_false; |
76 | static SV *json_true, *json_false; |
79 | |
77 | |
|
|
78 | enum { |
|
|
79 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
80 | INCR_M_STR, // inside string |
|
|
81 | INCR_M_BS, // inside backslash |
|
|
82 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
83 | INCR_M_C1, // inside comment in other places |
|
|
84 | INCR_M_JSON // outside anything, count nesting |
|
|
85 | }; |
|
|
86 | |
|
|
87 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
88 | |
80 | typedef struct { |
89 | typedef struct { |
81 | U32 flags; |
90 | U32 flags; |
|
|
91 | U32 max_depth; |
|
|
92 | STRLEN max_size; |
|
|
93 | |
82 | SV *cb_object; |
94 | SV *cb_object; |
83 | HV *cb_sk_object; |
95 | HV *cb_sk_object; |
|
|
96 | |
|
|
97 | // for the incremental parser |
|
|
98 | SV *incr_text; // the source text so far |
|
|
99 | STRLEN incr_pos; // the current offset into the text |
|
|
100 | int incr_nest; // {[]}-nesting level |
|
|
101 | unsigned char incr_mode; |
84 | } JSON; |
102 | } JSON; |
|
|
103 | |
|
|
104 | INLINE void |
|
|
105 | json_init (JSON *json) |
|
|
106 | { |
|
|
107 | Zero (json, 1, JSON); |
|
|
108 | json->max_depth = 512; |
|
|
109 | } |
85 | |
110 | |
86 | ///////////////////////////////////////////////////////////////////////////// |
111 | ///////////////////////////////////////////////////////////////////////////// |
87 | // utility functions |
112 | // utility functions |
88 | |
113 | |
|
|
114 | INLINE SV * |
|
|
115 | get_bool (const char *name) |
|
|
116 | { |
|
|
117 | SV *sv = get_sv (name, 1); |
|
|
118 | |
|
|
119 | SvREADONLY_on (sv); |
|
|
120 | SvREADONLY_on (SvRV (sv)); |
|
|
121 | |
|
|
122 | return sv; |
|
|
123 | } |
|
|
124 | |
89 | INLINE void |
125 | INLINE void |
90 | shrink (SV *sv) |
126 | shrink (SV *sv) |
91 | { |
127 | { |
92 | sv_utf8_downgrade (sv, 1); |
128 | sv_utf8_downgrade (sv, 1); |
|
|
129 | |
93 | if (SvLEN (sv) > SvCUR (sv) + 1) |
130 | if (SvLEN (sv) > SvCUR (sv) + 1) |
94 | { |
131 | { |
95 | #ifdef SvPV_shrink_to_cur |
132 | #ifdef SvPV_shrink_to_cur |
96 | SvPV_shrink_to_cur (sv); |
133 | SvPV_shrink_to_cur (sv); |
97 | #elif defined (SvPV_renew) |
134 | #elif defined (SvPV_renew) |
… | |
… | |
141 | *s++ = 0x80 | ( ch & 0x3f); |
178 | *s++ = 0x80 | ( ch & 0x3f); |
142 | |
179 | |
143 | return s; |
180 | return s; |
144 | } |
181 | } |
145 | |
182 | |
|
|
183 | // convert offset pointer to character index, sv must be string |
|
|
184 | static STRLEN |
|
|
185 | ptr_to_index (SV *sv, char *offset) |
|
|
186 | { |
|
|
187 | return SvUTF8 (sv) |
|
|
188 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
189 | : offset - SvPVX (sv); |
|
|
190 | } |
|
|
191 | |
|
|
192 | ///////////////////////////////////////////////////////////////////////////// |
|
|
193 | // fp hell |
|
|
194 | |
|
|
195 | // scan a group of digits, and a trailing exponent |
|
|
196 | static void |
|
|
197 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp) |
|
|
198 | { |
|
|
199 | UV uaccum = 0; |
|
|
200 | int eaccum = 0; |
|
|
201 | |
|
|
202 | for (;;) |
|
|
203 | { |
|
|
204 | U8 dig = (U8)*s - '0'; |
|
|
205 | |
|
|
206 | if (expect_false (dig >= 10)) |
|
|
207 | { |
|
|
208 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
209 | { |
|
|
210 | ++s; |
|
|
211 | json_atof_scan1 (s, accum, expo, 1); |
|
|
212 | } |
|
|
213 | else if ((dig | ' ') == 'e' - '0') |
|
|
214 | { |
|
|
215 | int exp2 = 0; |
|
|
216 | int neg = 0; |
|
|
217 | |
|
|
218 | ++s; |
|
|
219 | |
|
|
220 | if (*s == '-') |
|
|
221 | { |
|
|
222 | ++s; |
|
|
223 | neg = 1; |
|
|
224 | } |
|
|
225 | else if (*s == '+') |
|
|
226 | ++s; |
|
|
227 | |
|
|
228 | while ((dig = (U8)*s - '0') < 10) |
|
|
229 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
230 | |
|
|
231 | *expo += neg ? -exp2 : exp2; |
|
|
232 | } |
|
|
233 | |
|
|
234 | break; |
|
|
235 | } |
|
|
236 | |
|
|
237 | ++s; |
|
|
238 | |
|
|
239 | uaccum = uaccum * 10 + dig; |
|
|
240 | ++eaccum; |
|
|
241 | |
|
|
242 | // if we have too many digits, then recurse for more |
|
|
243 | // we actually do this for rather few digits |
|
|
244 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
245 | { |
|
|
246 | if (postdp) *expo -= eaccum; |
|
|
247 | json_atof_scan1 (s, accum, expo, postdp); |
|
|
248 | if (postdp) *expo += eaccum; |
|
|
249 | |
|
|
250 | break; |
|
|
251 | } |
|
|
252 | } |
|
|
253 | |
|
|
254 | if (postdp) *expo -= eaccum; |
|
|
255 | *accum += uaccum * pow (10., *expo); |
|
|
256 | *expo += eaccum; |
|
|
257 | } |
|
|
258 | |
|
|
259 | static NV |
|
|
260 | json_atof (const char *s) |
|
|
261 | { |
|
|
262 | NV accum = 0.; |
|
|
263 | int expo = 0; |
|
|
264 | int neg = 0; |
|
|
265 | |
|
|
266 | if (*s == '-') |
|
|
267 | { |
|
|
268 | ++s; |
|
|
269 | neg = 1; |
|
|
270 | } |
|
|
271 | |
|
|
272 | json_atof_scan1 (s, &accum, &expo, 0); |
|
|
273 | |
|
|
274 | return neg ? -accum : accum; |
|
|
275 | } |
146 | ///////////////////////////////////////////////////////////////////////////// |
276 | ///////////////////////////////////////////////////////////////////////////// |
147 | // encoder |
277 | // encoder |
148 | |
278 | |
149 | // structure used for encoding JSON |
279 | // structure used for encoding JSON |
150 | typedef struct |
280 | typedef struct |
… | |
… | |
152 | char *cur; // SvPVX (sv) + current output position |
282 | char *cur; // SvPVX (sv) + current output position |
153 | char *end; // SvEND (sv) |
283 | char *end; // SvEND (sv) |
154 | SV *sv; // result scalar |
284 | SV *sv; // result scalar |
155 | JSON json; |
285 | JSON json; |
156 | U32 indent; // indentation level |
286 | U32 indent; // indentation level |
157 | U32 maxdepth; // max. indentation/recursion level |
|
|
158 | UV limit; // escape character values >= this value when encoding |
287 | UV limit; // escape character values >= this value when encoding |
159 | } enc_t; |
288 | } enc_t; |
160 | |
289 | |
161 | INLINE void |
290 | INLINE void |
162 | need (enc_t *enc, STRLEN len) |
291 | need (enc_t *enc, STRLEN len) |
163 | { |
292 | { |
164 | if (expect_false (enc->cur + len >= enc->end)) |
293 | if (expect_false (enc->cur + len >= enc->end)) |
165 | { |
294 | { |
166 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
295 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
167 | SvGROW (enc->sv, cur + len + 1); |
296 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
168 | enc->cur = SvPVX (enc->sv) + cur; |
297 | enc->cur = SvPVX (enc->sv) + cur; |
169 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
298 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
170 | } |
299 | } |
171 | } |
300 | } |
172 | |
301 | |
… | |
… | |
247 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
376 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
248 | enc->cur += 12; |
377 | enc->cur += 12; |
249 | } |
378 | } |
250 | else |
379 | else |
251 | { |
380 | { |
252 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
253 | need (enc, len += 5); |
381 | need (enc, len += 5); |
254 | *enc->cur++ = '\\'; |
382 | *enc->cur++ = '\\'; |
255 | *enc->cur++ = 'u'; |
383 | *enc->cur++ = 'u'; |
256 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
384 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
257 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
385 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
258 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
386 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
259 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
387 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
260 | } |
388 | } |
261 | |
389 | |
262 | str += clen; |
390 | str += clen; |
263 | } |
391 | } |
264 | else if (enc->json.flags & F_LATIN1) |
392 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
335 | static void |
463 | static void |
336 | encode_av (enc_t *enc, AV *av) |
464 | encode_av (enc_t *enc, AV *av) |
337 | { |
465 | { |
338 | int i, len = av_len (av); |
466 | int i, len = av_len (av); |
339 | |
467 | |
340 | if (enc->indent >= enc->maxdepth) |
468 | if (enc->indent >= enc->json.max_depth) |
341 | croak ("data structure too deep (hit recursion limit)"); |
469 | croak (ERR_NESTING_EXCEEDED); |
342 | |
470 | |
343 | encode_ch (enc, '['); |
471 | encode_ch (enc, '['); |
344 | |
472 | |
345 | if (len >= 0) |
473 | if (len >= 0) |
346 | { |
474 | { |
… | |
… | |
420 | |
548 | |
421 | static void |
549 | static void |
422 | encode_hv (enc_t *enc, HV *hv) |
550 | encode_hv (enc_t *enc, HV *hv) |
423 | { |
551 | { |
424 | HE *he; |
552 | HE *he; |
425 | int count; |
|
|
426 | |
553 | |
427 | if (enc->indent >= enc->maxdepth) |
554 | if (enc->indent >= enc->json.max_depth) |
428 | croak ("data structure too deep (hit recursion limit)"); |
555 | croak (ERR_NESTING_EXCEEDED); |
429 | |
556 | |
430 | encode_ch (enc, '{'); |
557 | encode_ch (enc, '{'); |
431 | |
558 | |
432 | // for canonical output we have to sort by keys first |
559 | // for canonical output we have to sort by keys first |
433 | // actually, this is mostly due to the stupid so-called |
560 | // actually, this is mostly due to the stupid so-called |
434 | // security workaround added somewhere in 5.8.x. |
561 | // security workaround added somewhere in 5.8.x |
435 | // that randomises hash orderings |
562 | // that randomises hash orderings |
436 | if (enc->json.flags & F_CANONICAL) |
563 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
437 | { |
564 | { |
438 | int count = hv_iterinit (hv); |
565 | int count = hv_iterinit (hv); |
439 | |
566 | |
440 | if (SvMAGICAL (hv)) |
567 | if (SvMAGICAL (hv)) |
441 | { |
568 | { |
… | |
… | |
614 | |
741 | |
615 | if (len == 1 && *pv == '1') |
742 | if (len == 1 && *pv == '1') |
616 | encode_str (enc, "true", 4, 0); |
743 | encode_str (enc, "true", 4, 0); |
617 | else if (len == 1 && *pv == '0') |
744 | else if (len == 1 && *pv == '0') |
618 | encode_str (enc, "false", 5, 0); |
745 | encode_str (enc, "false", 5, 0); |
|
|
746 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
747 | encode_str (enc, "null", 4, 0); |
619 | else |
748 | else |
620 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
749 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
621 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
750 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
622 | } |
751 | } |
|
|
752 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
753 | encode_str (enc, "null", 4, 0); |
623 | else |
754 | else |
624 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
755 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
625 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
756 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
626 | } |
757 | } |
627 | |
758 | |
… | |
… | |
648 | else if (SvIOKp (sv)) |
779 | else if (SvIOKp (sv)) |
649 | { |
780 | { |
650 | // we assume we can always read an IV as a UV and vice versa |
781 | // we assume we can always read an IV as a UV and vice versa |
651 | // we assume two's complement |
782 | // we assume two's complement |
652 | // we assume no aliasing issues in the union |
783 | // we assume no aliasing issues in the union |
653 | if (SvIsUV (sv) ? SvUVX (sv) > 59000 |
784 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
654 | : SvIVX (sv) > 59000 || SvIVX (sv) < -59000) |
785 | : 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 | { |
786 | { |
665 | // optimise the "small number case" |
787 | // optimise the "small number case" |
666 | // code will likely be branchless and use only a single multiplication |
788 | // code will likely be branchless and use only a single multiplication |
667 | // works for numbers up to 59074 |
789 | // works for numbers up to 59074 |
668 | I32 i = SvIVX (sv); |
790 | I32 i = SvIVX (sv); |
… | |
… | |
679 | |
801 | |
680 | // now output digit by digit, each time masking out the integer part |
802 | // 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, |
803 | // and multiplying by 5 while moving the decimal point one to the right, |
682 | // resulting in a net multiplication by 10. |
804 | // resulting in a net multiplication by 10. |
683 | // we always write the digit to memory but conditionally increment |
805 | // we always write the digit to memory but conditionally increment |
684 | // the pointer, to ease the usage of conditional move instructions. |
806 | // 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; |
807 | 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; |
808 | 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; |
809 | 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; |
810 | 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' |
811 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
690 | } |
812 | } |
|
|
813 | else |
|
|
814 | { |
|
|
815 | // large integer, use the (rather slow) snprintf way. |
|
|
816 | need (enc, IVUV_MAXCHARS); |
|
|
817 | enc->cur += |
|
|
818 | SvIsUV(sv) |
|
|
819 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
820 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
821 | } |
691 | } |
822 | } |
692 | else if (SvROK (sv)) |
823 | else if (SvROK (sv)) |
693 | encode_rv (enc, SvRV (sv)); |
824 | encode_rv (enc, SvRV (sv)); |
694 | else if (!SvOK (sv)) |
825 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
695 | encode_str (enc, "null", 4, 0); |
826 | encode_str (enc, "null", 4, 0); |
696 | else |
827 | else |
697 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
828 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
698 | SvPV_nolen (sv), SvFLAGS (sv)); |
829 | SvPV_nolen (sv), SvFLAGS (sv)); |
699 | } |
830 | } |
… | |
… | |
709 | enc.json = *json; |
840 | enc.json = *json; |
710 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
841 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
711 | enc.cur = SvPVX (enc.sv); |
842 | enc.cur = SvPVX (enc.sv); |
712 | enc.end = SvEND (enc.sv); |
843 | enc.end = SvEND (enc.sv); |
713 | enc.indent = 0; |
844 | enc.indent = 0; |
714 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
|
|
715 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
845 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
716 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
846 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
717 | : 0x110000UL; |
847 | : 0x110000UL; |
718 | |
848 | |
719 | SvPOK_only (enc.sv); |
849 | SvPOK_only (enc.sv); |
720 | encode_sv (&enc, scalar); |
850 | encode_sv (&enc, scalar); |
|
|
851 | encode_nl (&enc); |
721 | |
852 | |
722 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
853 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
723 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
854 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
724 | |
855 | |
725 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
856 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
786 | if (*dec->cur != ch) \ |
917 | if (*dec->cur != ch) \ |
787 | ERR (# ch " expected"); \ |
918 | ERR (# ch " expected"); \ |
788 | ++dec->cur; \ |
919 | ++dec->cur; \ |
789 | SE |
920 | SE |
790 | |
921 | |
791 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
922 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
792 | #define DEC_DEC_DEPTH --dec->depth |
923 | #define DEC_DEC_DEPTH --dec->depth |
793 | |
924 | |
794 | static SV *decode_sv (dec_t *dec); |
925 | static SV *decode_sv (dec_t *dec); |
795 | |
926 | |
796 | static signed char decode_hexdigit[256]; |
927 | static signed char decode_hexdigit[256]; |
… | |
… | |
905 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
1036 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
906 | *cur++ = ch; |
1037 | *cur++ = ch; |
907 | else if (ch >= 0x80) |
1038 | else if (ch >= 0x80) |
908 | { |
1039 | { |
909 | STRLEN clen; |
1040 | STRLEN clen; |
910 | UV uch; |
|
|
911 | |
1041 | |
912 | --dec_cur; |
1042 | --dec_cur; |
913 | |
1043 | |
914 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1044 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
915 | if (clen == (STRLEN)-1) |
1045 | if (clen == (STRLEN)-1) |
916 | ERR ("malformed UTF-8 character in JSON string"); |
1046 | ERR ("malformed UTF-8 character in JSON string"); |
917 | |
1047 | |
918 | do |
1048 | do |
919 | *cur++ = *dec_cur++; |
1049 | *cur++ = *dec_cur++; |
… | |
… | |
936 | { |
1066 | { |
937 | STRLEN len = cur - buf; |
1067 | STRLEN len = cur - buf; |
938 | |
1068 | |
939 | if (sv) |
1069 | if (sv) |
940 | { |
1070 | { |
941 | SvGROW (sv, SvCUR (sv) + len + 1); |
1071 | STRLEN cur = SvCUR (sv); |
|
|
1072 | |
|
|
1073 | if (SvLEN (sv) <= cur + len) |
|
|
1074 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1075 | |
942 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1076 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
943 | SvCUR_set (sv, SvCUR (sv) + len); |
1077 | SvCUR_set (sv, SvCUR (sv) + len); |
944 | } |
1078 | } |
945 | else |
1079 | else |
946 | sv = newSVpvn (buf, len); |
1080 | sv = newSVpvn (buf, len); |
… | |
… | |
1037 | |
1171 | |
1038 | // special case the rather common 1..5-digit-int case |
1172 | // special case the rather common 1..5-digit-int case |
1039 | if (*start == '-') |
1173 | if (*start == '-') |
1040 | switch (len) |
1174 | switch (len) |
1041 | { |
1175 | { |
1042 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1176 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
1043 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1177 | 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)); |
1178 | 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)); |
1179 | 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)); |
1180 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1047 | } |
1181 | } |
1048 | else |
1182 | else |
1049 | switch (len) |
1183 | switch (len) |
1050 | { |
1184 | { |
1051 | case 1: return newSViv ( start [0] - '0' * 1); |
1185 | case 1: return newSViv ( start [0] - '0' * 1); |
1052 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1186 | 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); |
1187 | 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); |
1188 | 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); |
1189 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1056 | } |
1190 | } |
1057 | |
1191 | |
1058 | { |
1192 | { |
1059 | UV uv; |
1193 | UV uv; |
1060 | int numtype = grok_number (start, len, &uv); |
1194 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1069 | } |
1203 | } |
1070 | |
1204 | |
1071 | len -= *start == '-' ? 1 : 0; |
1205 | len -= *start == '-' ? 1 : 0; |
1072 | |
1206 | |
1073 | // does not fit into IV or UV, try NV |
1207 | // does not fit into IV or UV, try NV |
1074 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
1208 | 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 |
1209 | // fits into NV without loss of precision |
1080 | return newSVnv (Atof (start)); |
1210 | return newSVnv (json_atof (start)); |
1081 | |
1211 | |
1082 | // everything else fails, convert it to a string |
1212 | // everything else fails, convert it to a string |
1083 | return newSVpvn (start, dec->cur - start); |
1213 | return newSVpvn (start, dec->cur - start); |
1084 | } |
1214 | } |
1085 | |
1215 | |
1086 | // loss of precision here |
1216 | // loss of precision here |
1087 | return newSVnv (Atof (start)); |
1217 | return newSVnv (json_atof (start)); |
1088 | |
1218 | |
1089 | fail: |
1219 | fail: |
1090 | return 0; |
1220 | return 0; |
1091 | } |
1221 | } |
1092 | |
1222 | |
… | |
… | |
1168 | char *p = dec->cur; |
1298 | char *p = dec->cur; |
1169 | char *e = p + 24; // only try up to 24 bytes |
1299 | char *e = p + 24; // only try up to 24 bytes |
1170 | |
1300 | |
1171 | for (;;) |
1301 | for (;;) |
1172 | { |
1302 | { |
1173 | // the >= 0x80 is true on most architectures |
1303 | // the >= 0x80 is false on most architectures |
1174 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1304 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1175 | { |
1305 | { |
1176 | // slow path, back up and use decode_str |
1306 | // slow path, back up and use decode_str |
1177 | SV *key = decode_str (dec); |
1307 | SV *key = decode_str (dec); |
1178 | if (!key) |
1308 | if (!key) |
… | |
… | |
1326 | case 't': |
1456 | case 't': |
1327 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1457 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1328 | { |
1458 | { |
1329 | dec->cur += 4; |
1459 | dec->cur += 4; |
1330 | #if JSON_SLOW |
1460 | #if JSON_SLOW |
1331 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
1461 | json_true = get_bool ("JSON::XS::true"); |
1332 | #endif |
1462 | #endif |
1333 | return SvREFCNT_inc (json_true); |
1463 | return newSVsv (json_true); |
1334 | } |
1464 | } |
1335 | else |
1465 | else |
1336 | ERR ("'true' expected"); |
1466 | ERR ("'true' expected"); |
1337 | |
1467 | |
1338 | break; |
1468 | break; |
… | |
… | |
1340 | case 'f': |
1470 | case 'f': |
1341 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1471 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1342 | { |
1472 | { |
1343 | dec->cur += 5; |
1473 | dec->cur += 5; |
1344 | #if JSON_SLOW |
1474 | #if JSON_SLOW |
1345 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1475 | json_false = get_bool ("JSON::XS::false"); |
1346 | #endif |
1476 | #endif |
1347 | return SvREFCNT_inc (json_false); |
1477 | return newSVsv (json_false); |
1348 | } |
1478 | } |
1349 | else |
1479 | else |
1350 | ERR ("'false' expected"); |
1480 | ERR ("'false' expected"); |
1351 | |
1481 | |
1352 | break; |
1482 | break; |
… | |
… | |
1370 | fail: |
1500 | fail: |
1371 | return 0; |
1501 | return 0; |
1372 | } |
1502 | } |
1373 | |
1503 | |
1374 | static SV * |
1504 | static SV * |
1375 | decode_json (SV *string, JSON *json, UV *offset_return) |
1505 | decode_json (SV *string, JSON *json, char **offset_return) |
1376 | { |
1506 | { |
1377 | dec_t dec; |
1507 | dec_t dec; |
1378 | UV offset; |
|
|
1379 | SV *sv; |
1508 | SV *sv; |
1380 | |
1509 | |
|
|
1510 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1511 | * will perl ignore the magic in subsequent accesses |
|
|
1512 | */ |
1381 | SvGETMAGIC (string); |
1513 | /*SvGETMAGIC (string);*/ |
|
|
1514 | if (SvMAGICAL (string)) |
|
|
1515 | string = sv_2mortal (newSVsv (string)); |
|
|
1516 | |
1382 | SvUPGRADE (string, SVt_PV); |
1517 | SvUPGRADE (string, SVt_PV); |
1383 | |
1518 | |
1384 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1519 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1520 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1521 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1522 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1523 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1524 | * and hope for the best. |
|
|
1525 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1526 | * assertion business is seriously broken, try yet another workaround |
|
|
1527 | * for the broken -DDEBUGGING. |
|
|
1528 | */ |
|
|
1529 | { |
|
|
1530 | #ifdef DEBUGGING |
|
|
1531 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1532 | #else |
|
|
1533 | STRLEN offset = SvCUR (string); |
|
|
1534 | #endif |
|
|
1535 | |
|
|
1536 | 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", |
1537 | 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)); |
1538 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1539 | } |
1387 | |
1540 | |
1388 | if (json->flags & F_UTF8) |
1541 | if (json->flags & F_UTF8) |
1389 | sv_utf8_downgrade (string, 0); |
1542 | sv_utf8_downgrade (string, 0); |
1390 | else |
1543 | else |
1391 | sv_utf8_upgrade (string); |
1544 | sv_utf8_upgrade (string); |
1392 | |
1545 | |
1393 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1546 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1394 | |
1547 | |
1395 | dec.json = *json; |
1548 | dec.json = *json; |
1396 | dec.cur = SvPVX (string); |
1549 | dec.cur = SvPVX (string); |
1397 | dec.end = SvEND (string); |
1550 | dec.end = SvEND (string); |
1398 | dec.err = 0; |
1551 | dec.err = 0; |
1399 | dec.depth = 0; |
1552 | dec.depth = 0; |
1400 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1401 | |
1553 | |
1402 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1554 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1403 | dec.json.flags |= F_HOOK; |
1555 | dec.json.flags |= F_HOOK; |
1404 | |
1556 | |
1405 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1557 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1406 | |
1558 | |
1407 | decode_ws (&dec); |
1559 | decode_ws (&dec); |
1408 | sv = decode_sv (&dec); |
1560 | sv = decode_sv (&dec); |
|
|
1561 | |
|
|
1562 | if (offset_return) |
|
|
1563 | *offset_return = dec.cur; |
1409 | |
1564 | |
1410 | if (!(offset_return || !sv)) |
1565 | if (!(offset_return || !sv)) |
1411 | { |
1566 | { |
1412 | // check for trailing garbage |
1567 | // check for trailing garbage |
1413 | decode_ws (&dec); |
1568 | decode_ws (&dec); |
… | |
… | |
1416 | { |
1571 | { |
1417 | dec.err = "garbage after JSON object"; |
1572 | dec.err = "garbage after JSON object"; |
1418 | SvREFCNT_dec (sv); |
1573 | SvREFCNT_dec (sv); |
1419 | sv = 0; |
1574 | sv = 0; |
1420 | } |
1575 | } |
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 | } |
1576 | } |
1432 | |
1577 | |
1433 | if (!sv) |
1578 | if (!sv) |
1434 | { |
1579 | { |
1435 | SV *uni = sv_newmortal (); |
1580 | SV *uni = sv_newmortal (); |
… | |
… | |
1441 | SAVEVPTR (PL_curcop); |
1586 | SAVEVPTR (PL_curcop); |
1442 | PL_curcop = &cop; |
1587 | PL_curcop = &cop; |
1443 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1588 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1444 | LEAVE; |
1589 | LEAVE; |
1445 | |
1590 | |
1446 | croak ("%s, at character offset %d [\"%s\"]", |
1591 | croak ("%s, at character offset %d (before \"%s\")", |
1447 | dec.err, |
1592 | dec.err, |
1448 | (int)offset, |
1593 | ptr_to_index (string, dec.cur), |
1449 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1594 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1450 | } |
1595 | } |
1451 | |
1596 | |
1452 | sv = sv_2mortal (sv); |
1597 | sv = sv_2mortal (sv); |
1453 | |
1598 | |
1454 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1599 | 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)"); |
1600 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1456 | |
1601 | |
1457 | return sv; |
1602 | return sv; |
|
|
1603 | } |
|
|
1604 | |
|
|
1605 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1606 | // incremental parser |
|
|
1607 | |
|
|
1608 | static void |
|
|
1609 | incr_parse (JSON *self) |
|
|
1610 | { |
|
|
1611 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1612 | |
|
|
1613 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1614 | // but this would make it slower, so... |
|
|
1615 | |
|
|
1616 | for (;;) |
|
|
1617 | { |
|
|
1618 | //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 |
|
|
1619 | switch (self->incr_mode) |
|
|
1620 | { |
|
|
1621 | // only used for initial whitespace skipping |
|
|
1622 | case INCR_M_WS: |
|
|
1623 | for (;;) |
|
|
1624 | { |
|
|
1625 | if (*p > 0x20) |
|
|
1626 | { |
|
|
1627 | if (*p == '#') |
|
|
1628 | { |
|
|
1629 | self->incr_mode = INCR_M_C0; |
|
|
1630 | goto incr_m_c; |
|
|
1631 | } |
|
|
1632 | else |
|
|
1633 | { |
|
|
1634 | self->incr_mode = INCR_M_JSON; |
|
|
1635 | goto incr_m_json; |
|
|
1636 | } |
|
|
1637 | } |
|
|
1638 | else if (!*p) |
|
|
1639 | goto interrupt; |
|
|
1640 | |
|
|
1641 | ++p; |
|
|
1642 | } |
|
|
1643 | |
|
|
1644 | // skip a single char inside a string (for \\-processing) |
|
|
1645 | case INCR_M_BS: |
|
|
1646 | if (!*p) |
|
|
1647 | goto interrupt; |
|
|
1648 | |
|
|
1649 | ++p; |
|
|
1650 | self->incr_mode = INCR_M_STR; |
|
|
1651 | goto incr_m_str; |
|
|
1652 | |
|
|
1653 | // inside #-style comments |
|
|
1654 | case INCR_M_C0: |
|
|
1655 | case INCR_M_C1: |
|
|
1656 | incr_m_c: |
|
|
1657 | for (;;) |
|
|
1658 | { |
|
|
1659 | if (*p == '\n') |
|
|
1660 | { |
|
|
1661 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1662 | break; |
|
|
1663 | } |
|
|
1664 | else if (!*p) |
|
|
1665 | goto interrupt; |
|
|
1666 | |
|
|
1667 | ++p; |
|
|
1668 | } |
|
|
1669 | |
|
|
1670 | break; |
|
|
1671 | |
|
|
1672 | // inside a string |
|
|
1673 | case INCR_M_STR: |
|
|
1674 | incr_m_str: |
|
|
1675 | for (;;) |
|
|
1676 | { |
|
|
1677 | if (*p == '"') |
|
|
1678 | { |
|
|
1679 | ++p; |
|
|
1680 | self->incr_mode = INCR_M_JSON; |
|
|
1681 | |
|
|
1682 | if (!self->incr_nest) |
|
|
1683 | goto interrupt; |
|
|
1684 | |
|
|
1685 | goto incr_m_json; |
|
|
1686 | } |
|
|
1687 | else if (*p == '\\') |
|
|
1688 | { |
|
|
1689 | ++p; // "virtually" consumes character after \ |
|
|
1690 | |
|
|
1691 | if (!*p) // if at end of string we have to switch modes |
|
|
1692 | { |
|
|
1693 | self->incr_mode = INCR_M_BS; |
|
|
1694 | goto interrupt; |
|
|
1695 | } |
|
|
1696 | } |
|
|
1697 | else if (!*p) |
|
|
1698 | goto interrupt; |
|
|
1699 | |
|
|
1700 | ++p; |
|
|
1701 | } |
|
|
1702 | |
|
|
1703 | // after initial ws, outside string |
|
|
1704 | case INCR_M_JSON: |
|
|
1705 | incr_m_json: |
|
|
1706 | for (;;) |
|
|
1707 | { |
|
|
1708 | switch (*p++) |
|
|
1709 | { |
|
|
1710 | case 0: |
|
|
1711 | --p; |
|
|
1712 | goto interrupt; |
|
|
1713 | |
|
|
1714 | case 0x09: |
|
|
1715 | case 0x0a: |
|
|
1716 | case 0x0d: |
|
|
1717 | case 0x20: |
|
|
1718 | if (!self->incr_nest) |
|
|
1719 | { |
|
|
1720 | --p; // do not eat the whitespace, let the next round do it |
|
|
1721 | goto interrupt; |
|
|
1722 | } |
|
|
1723 | break; |
|
|
1724 | |
|
|
1725 | case '"': |
|
|
1726 | self->incr_mode = INCR_M_STR; |
|
|
1727 | goto incr_m_str; |
|
|
1728 | |
|
|
1729 | case '[': |
|
|
1730 | case '{': |
|
|
1731 | if (++self->incr_nest > self->max_depth) |
|
|
1732 | croak (ERR_NESTING_EXCEEDED); |
|
|
1733 | break; |
|
|
1734 | |
|
|
1735 | case ']': |
|
|
1736 | case '}': |
|
|
1737 | if (--self->incr_nest <= 0) |
|
|
1738 | goto interrupt; |
|
|
1739 | break; |
|
|
1740 | |
|
|
1741 | case '#': |
|
|
1742 | self->incr_mode = INCR_M_C1; |
|
|
1743 | goto incr_m_c; |
|
|
1744 | } |
|
|
1745 | } |
|
|
1746 | } |
|
|
1747 | |
|
|
1748 | modechange: |
|
|
1749 | ; |
|
|
1750 | } |
|
|
1751 | |
|
|
1752 | interrupt: |
|
|
1753 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1754 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1755 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1458 | } |
1756 | } |
1459 | |
1757 | |
1460 | ///////////////////////////////////////////////////////////////////////////// |
1758 | ///////////////////////////////////////////////////////////////////////////// |
1461 | // XS interface functions |
1759 | // XS interface functions |
1462 | |
1760 | |
… | |
… | |
1474 | : -1; |
1772 | : -1; |
1475 | |
1773 | |
1476 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1774 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1477 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1775 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1478 | |
1776 | |
1479 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1777 | json_true = get_bool ("JSON::XS::true"); |
1480 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1778 | json_false = get_bool ("JSON::XS::false"); |
|
|
1779 | |
|
|
1780 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1481 | } |
1781 | } |
1482 | |
1782 | |
1483 | PROTOTYPES: DISABLE |
1783 | PROTOTYPES: DISABLE |
1484 | |
1784 | |
1485 | void CLONE (...) |
1785 | void CLONE (...) |
… | |
… | |
1490 | void new (char *klass) |
1790 | void new (char *klass) |
1491 | PPCODE: |
1791 | PPCODE: |
1492 | { |
1792 | { |
1493 | SV *pv = NEWSV (0, sizeof (JSON)); |
1793 | SV *pv = NEWSV (0, sizeof (JSON)); |
1494 | SvPOK_only (pv); |
1794 | SvPOK_only (pv); |
1495 | Zero (SvPVX (pv), 1, JSON); |
1795 | json_init ((JSON *)SvPVX (pv)); |
1496 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1497 | XPUSHs (sv_2mortal (sv_bless ( |
1796 | XPUSHs (sv_2mortal (sv_bless ( |
1498 | newRV_noinc (pv), |
1797 | newRV_noinc (pv), |
1499 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1798 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1500 | ))); |
1799 | ))); |
1501 | } |
1800 | } |
… | |
… | |
1513 | allow_nonref = F_ALLOW_NONREF |
1812 | allow_nonref = F_ALLOW_NONREF |
1514 | shrink = F_SHRINK |
1813 | shrink = F_SHRINK |
1515 | allow_blessed = F_ALLOW_BLESSED |
1814 | allow_blessed = F_ALLOW_BLESSED |
1516 | convert_blessed = F_CONV_BLESSED |
1815 | convert_blessed = F_CONV_BLESSED |
1517 | relaxed = F_RELAXED |
1816 | relaxed = F_RELAXED |
|
|
1817 | allow_unknown = F_ALLOW_UNKNOWN |
1518 | PPCODE: |
1818 | PPCODE: |
1519 | { |
1819 | { |
1520 | if (enable) |
1820 | if (enable) |
1521 | self->flags |= ix; |
1821 | self->flags |= ix; |
1522 | else |
1822 | else |
… | |
… | |
1537 | get_allow_nonref = F_ALLOW_NONREF |
1837 | get_allow_nonref = F_ALLOW_NONREF |
1538 | get_shrink = F_SHRINK |
1838 | get_shrink = F_SHRINK |
1539 | get_allow_blessed = F_ALLOW_BLESSED |
1839 | get_allow_blessed = F_ALLOW_BLESSED |
1540 | get_convert_blessed = F_CONV_BLESSED |
1840 | get_convert_blessed = F_CONV_BLESSED |
1541 | get_relaxed = F_RELAXED |
1841 | get_relaxed = F_RELAXED |
|
|
1842 | get_allow_unknown = F_ALLOW_UNKNOWN |
1542 | PPCODE: |
1843 | PPCODE: |
1543 | XPUSHs (boolSV (self->flags & ix)); |
1844 | XPUSHs (boolSV (self->flags & ix)); |
1544 | |
1845 | |
1545 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1846 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1546 | PPCODE: |
1847 | PPCODE: |
1547 | { |
1848 | 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)); |
1849 | XPUSHs (ST (0)); |
1558 | } |
|
|
1559 | |
1850 | |
1560 | U32 get_max_depth (JSON *self) |
1851 | U32 get_max_depth (JSON *self) |
1561 | CODE: |
1852 | CODE: |
1562 | RETVAL = DEC_DEPTH (self->flags); |
1853 | RETVAL = self->max_depth; |
1563 | OUTPUT: |
1854 | OUTPUT: |
1564 | RETVAL |
1855 | RETVAL |
1565 | |
1856 | |
1566 | void max_size (JSON *self, UV max_size = 0) |
1857 | void max_size (JSON *self, U32 max_size = 0) |
1567 | PPCODE: |
1858 | PPCODE: |
1568 | { |
1859 | 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)); |
1860 | XPUSHs (ST (0)); |
1580 | } |
|
|
1581 | |
1861 | |
1582 | int get_max_size (JSON *self) |
1862 | int get_max_size (JSON *self) |
1583 | CODE: |
1863 | CODE: |
1584 | RETVAL = DEC_SIZE (self->flags); |
1864 | RETVAL = self->max_size; |
1585 | OUTPUT: |
1865 | OUTPUT: |
1586 | RETVAL |
1866 | RETVAL |
1587 | |
1867 | |
1588 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1868 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1589 | PPCODE: |
1869 | PPCODE: |
… | |
… | |
1625 | XPUSHs (decode_json (jsonstr, self, 0)); |
1905 | XPUSHs (decode_json (jsonstr, self, 0)); |
1626 | |
1906 | |
1627 | void decode_prefix (JSON *self, SV *jsonstr) |
1907 | void decode_prefix (JSON *self, SV *jsonstr) |
1628 | PPCODE: |
1908 | PPCODE: |
1629 | { |
1909 | { |
1630 | UV offset; |
1910 | char *offset; |
1631 | EXTEND (SP, 2); |
1911 | EXTEND (SP, 2); |
1632 | PUSHs (decode_json (jsonstr, self, &offset)); |
1912 | PUSHs (decode_json (jsonstr, self, &offset)); |
1633 | PUSHs (sv_2mortal (newSVuv (offset))); |
1913 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1914 | } |
|
|
1915 | |
|
|
1916 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1917 | PPCODE: |
|
|
1918 | { |
|
|
1919 | if (!self->incr_text) |
|
|
1920 | self->incr_text = newSVpvn ("", 0); |
|
|
1921 | |
|
|
1922 | // append data, if any |
|
|
1923 | if (jsonstr) |
|
|
1924 | { |
|
|
1925 | if (SvUTF8 (jsonstr)) |
|
|
1926 | { |
|
|
1927 | if (!SvUTF8 (self->incr_text)) |
|
|
1928 | { |
|
|
1929 | /* utf-8-ness differs, need to upgrade */ |
|
|
1930 | sv_utf8_upgrade (self->incr_text); |
|
|
1931 | |
|
|
1932 | if (self->incr_pos) |
|
|
1933 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1934 | - (U8 *)SvPVX (self->incr_text); |
|
|
1935 | } |
|
|
1936 | } |
|
|
1937 | else if (SvUTF8 (self->incr_text)) |
|
|
1938 | sv_utf8_upgrade (jsonstr); |
|
|
1939 | |
|
|
1940 | { |
|
|
1941 | STRLEN len; |
|
|
1942 | const char *str = SvPV (jsonstr, len); |
|
|
1943 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1944 | |
|
|
1945 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1946 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1947 | |
|
|
1948 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1949 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1950 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1951 | } |
|
|
1952 | } |
|
|
1953 | |
|
|
1954 | if (GIMME_V != G_VOID) |
|
|
1955 | do |
|
|
1956 | { |
|
|
1957 | char *offset; |
|
|
1958 | |
|
|
1959 | if (!INCR_DONE (self)) |
|
|
1960 | { |
|
|
1961 | incr_parse (self); |
|
|
1962 | |
|
|
1963 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1964 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1965 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1966 | |
|
|
1967 | if (!INCR_DONE (self)) |
|
|
1968 | break; |
|
|
1969 | } |
|
|
1970 | |
|
|
1971 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1972 | |
|
|
1973 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
1974 | self->incr_nest = 0; |
|
|
1975 | self->incr_mode = 0; |
|
|
1976 | |
|
|
1977 | sv_chop (self->incr_text, offset); |
|
|
1978 | } |
|
|
1979 | while (GIMME_V == G_ARRAY); |
|
|
1980 | } |
|
|
1981 | |
|
|
1982 | SV *incr_text (JSON *self) |
|
|
1983 | ATTRS: lvalue |
|
|
1984 | CODE: |
|
|
1985 | { |
|
|
1986 | if (self->incr_pos) |
|
|
1987 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
1988 | |
|
|
1989 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
1990 | } |
|
|
1991 | OUTPUT: |
|
|
1992 | RETVAL |
|
|
1993 | |
|
|
1994 | void incr_skip (JSON *self) |
|
|
1995 | CODE: |
|
|
1996 | { |
|
|
1997 | if (self->incr_pos) |
|
|
1998 | { |
|
|
1999 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2000 | self->incr_pos = 0; |
|
|
2001 | self->incr_nest = 0; |
|
|
2002 | self->incr_mode = 0; |
|
|
2003 | } |
|
|
2004 | } |
|
|
2005 | |
|
|
2006 | void incr_reset (JSON *self) |
|
|
2007 | CODE: |
|
|
2008 | { |
|
|
2009 | SvREFCNT_dec (self->incr_text); |
|
|
2010 | self->incr_text = 0; |
|
|
2011 | self->incr_pos = 0; |
|
|
2012 | self->incr_nest = 0; |
|
|
2013 | self->incr_mode = 0; |
1634 | } |
2014 | } |
1635 | |
2015 | |
1636 | void DESTROY (JSON *self) |
2016 | void DESTROY (JSON *self) |
1637 | CODE: |
2017 | CODE: |
1638 | SvREFCNT_dec (self->cb_sk_object); |
2018 | SvREFCNT_dec (self->cb_sk_object); |
1639 | SvREFCNT_dec (self->cb_object); |
2019 | SvREFCNT_dec (self->cb_object); |
|
|
2020 | SvREFCNT_dec (self->incr_text); |
1640 | |
2021 | |
1641 | PROTOTYPES: ENABLE |
2022 | PROTOTYPES: ENABLE |
1642 | |
2023 | |
1643 | void encode_json (SV *scalar) |
2024 | void encode_json (SV *scalar) |
|
|
2025 | ALIAS: |
|
|
2026 | to_json_ = 0 |
|
|
2027 | encode_json = F_UTF8 |
1644 | PPCODE: |
2028 | PPCODE: |
1645 | { |
2029 | { |
1646 | JSON json = { F_DEFAULT | F_UTF8 }; |
2030 | JSON json; |
|
|
2031 | json_init (&json); |
|
|
2032 | json.flags |= ix; |
1647 | XPUSHs (encode_json (scalar, &json)); |
2033 | XPUSHs (encode_json (scalar, &json)); |
1648 | } |
2034 | } |
1649 | |
2035 | |
1650 | void decode_json (SV *jsonstr) |
2036 | void decode_json (SV *jsonstr) |
|
|
2037 | ALIAS: |
|
|
2038 | from_json_ = 0 |
|
|
2039 | decode_json = F_UTF8 |
1651 | PPCODE: |
2040 | PPCODE: |
1652 | { |
2041 | { |
1653 | JSON json = { F_DEFAULT | F_UTF8 }; |
2042 | JSON json; |
|
|
2043 | json_init (&json); |
|
|
2044 | json.flags |= ix; |
1654 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2045 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1655 | } |
2046 | } |
1656 | |
2047 | |