… | |
… | |
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) |
… | |
… | |
123 | // this function takes advantage of this fact, although current gccs |
160 | // this function takes advantage of this fact, although current gccs |
124 | // seem to optimise the check for >= 0x80 away anyways |
161 | // seem to optimise the check for >= 0x80 away anyways |
125 | INLINE unsigned char * |
162 | INLINE unsigned char * |
126 | encode_utf8 (unsigned char *s, UV ch) |
163 | encode_utf8 (unsigned char *s, UV ch) |
127 | { |
164 | { |
128 | if (ch <= 0x7FF) |
165 | if (expect_false (ch < 0x000080)) |
129 | { |
166 | *s++ = ch; |
130 | *s++ = (ch >> 6) | 0xc0; |
167 | else if (expect_true (ch < 0x000800)) |
131 | *s++ = (ch & 0x3f) | 0x80; |
168 | *s++ = 0xc0 | ( ch >> 6), |
132 | } |
169 | *s++ = 0x80 | ( ch & 0x3f); |
133 | else |
170 | else if ( ch < 0x010000) |
134 | s = uvuni_to_utf8_flags (s, ch, 0); |
171 | *s++ = 0xe0 | ( ch >> 12), |
|
|
172 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
173 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
174 | else if ( ch < 0x110000) |
|
|
175 | *s++ = 0xf0 | ( ch >> 18), |
|
|
176 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
177 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
178 | *s++ = 0x80 | ( ch & 0x3f); |
135 | |
179 | |
136 | return s; |
180 | return s; |
|
|
181 | } |
|
|
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); |
137 | } |
190 | } |
138 | |
191 | |
139 | ///////////////////////////////////////////////////////////////////////////// |
192 | ///////////////////////////////////////////////////////////////////////////// |
140 | // encoder |
193 | // encoder |
141 | |
194 | |
… | |
… | |
145 | char *cur; // SvPVX (sv) + current output position |
198 | char *cur; // SvPVX (sv) + current output position |
146 | char *end; // SvEND (sv) |
199 | char *end; // SvEND (sv) |
147 | SV *sv; // result scalar |
200 | SV *sv; // result scalar |
148 | JSON json; |
201 | JSON json; |
149 | U32 indent; // indentation level |
202 | U32 indent; // indentation level |
150 | U32 maxdepth; // max. indentation/recursion level |
|
|
151 | UV limit; // escape character values >= this value when encoding |
203 | UV limit; // escape character values >= this value when encoding |
152 | } enc_t; |
204 | } enc_t; |
153 | |
205 | |
154 | INLINE void |
206 | INLINE void |
155 | need (enc_t *enc, STRLEN len) |
207 | need (enc_t *enc, STRLEN len) |
156 | { |
208 | { |
157 | if (expect_false (enc->cur + len >= enc->end)) |
209 | if (expect_false (enc->cur + len >= enc->end)) |
158 | { |
210 | { |
159 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
211 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
160 | SvGROW (enc->sv, cur + len + 1); |
212 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
161 | enc->cur = SvPVX (enc->sv) + cur; |
213 | enc->cur = SvPVX (enc->sv) + cur; |
162 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
214 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
163 | } |
215 | } |
164 | } |
216 | } |
165 | |
217 | |
… | |
… | |
227 | clen = 1; |
279 | clen = 1; |
228 | } |
280 | } |
229 | |
281 | |
230 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
282 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
231 | { |
283 | { |
232 | if (uch > 0xFFFFUL) |
284 | if (uch >= 0x10000UL) |
233 | { |
285 | { |
234 | if (uch > 0x10FFFFUL) |
286 | if (uch >= 0x110000UL) |
235 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
287 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
236 | |
288 | |
237 | need (enc, len += 11); |
289 | need (enc, len += 11); |
238 | sprintf (enc->cur, "\\u%04x\\u%04x", |
290 | sprintf (enc->cur, "\\u%04x\\u%04x", |
239 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
291 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
240 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
292 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
241 | enc->cur += 12; |
293 | enc->cur += 12; |
242 | } |
294 | } |
243 | else |
295 | else |
244 | { |
296 | { |
245 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
246 | need (enc, len += 5); |
297 | need (enc, len += 5); |
247 | *enc->cur++ = '\\'; |
298 | *enc->cur++ = '\\'; |
248 | *enc->cur++ = 'u'; |
299 | *enc->cur++ = 'u'; |
249 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
300 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
250 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
301 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
251 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
302 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
252 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
303 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
253 | } |
304 | } |
254 | |
305 | |
255 | str += clen; |
306 | str += clen; |
256 | } |
307 | } |
257 | else if (enc->json.flags & F_LATIN1) |
308 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
328 | static void |
379 | static void |
329 | encode_av (enc_t *enc, AV *av) |
380 | encode_av (enc_t *enc, AV *av) |
330 | { |
381 | { |
331 | int i, len = av_len (av); |
382 | int i, len = av_len (av); |
332 | |
383 | |
333 | if (enc->indent >= enc->maxdepth) |
384 | if (enc->indent >= enc->json.max_depth) |
334 | croak ("data structure too deep (hit recursion limit)"); |
385 | croak (ERR_NESTING_EXCEEDED); |
335 | |
386 | |
336 | encode_ch (enc, '['); |
387 | encode_ch (enc, '['); |
337 | |
388 | |
338 | if (len >= 0) |
389 | if (len >= 0) |
339 | { |
390 | { |
… | |
… | |
413 | |
464 | |
414 | static void |
465 | static void |
415 | encode_hv (enc_t *enc, HV *hv) |
466 | encode_hv (enc_t *enc, HV *hv) |
416 | { |
467 | { |
417 | HE *he; |
468 | HE *he; |
418 | int count; |
|
|
419 | |
469 | |
420 | if (enc->indent >= enc->maxdepth) |
470 | if (enc->indent >= enc->json.max_depth) |
421 | croak ("data structure too deep (hit recursion limit)"); |
471 | croak (ERR_NESTING_EXCEEDED); |
422 | |
472 | |
423 | encode_ch (enc, '{'); |
473 | encode_ch (enc, '{'); |
424 | |
474 | |
425 | // for canonical output we have to sort by keys first |
475 | // for canonical output we have to sort by keys first |
426 | // actually, this is mostly due to the stupid so-called |
476 | // actually, this is mostly due to the stupid so-called |
427 | // security workaround added somewhere in 5.8.x. |
477 | // security workaround added somewhere in 5.8.x |
428 | // that randomises hash orderings |
478 | // that randomises hash orderings |
429 | if (enc->json.flags & F_CANONICAL) |
479 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
430 | { |
480 | { |
431 | int count = hv_iterinit (hv); |
481 | int count = hv_iterinit (hv); |
432 | |
482 | |
433 | if (SvMAGICAL (hv)) |
483 | if (SvMAGICAL (hv)) |
434 | { |
484 | { |
… | |
… | |
607 | |
657 | |
608 | if (len == 1 && *pv == '1') |
658 | if (len == 1 && *pv == '1') |
609 | encode_str (enc, "true", 4, 0); |
659 | encode_str (enc, "true", 4, 0); |
610 | else if (len == 1 && *pv == '0') |
660 | else if (len == 1 && *pv == '0') |
611 | encode_str (enc, "false", 5, 0); |
661 | encode_str (enc, "false", 5, 0); |
|
|
662 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
663 | encode_str (enc, "null", 4, 0); |
612 | else |
664 | else |
613 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
665 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
614 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
666 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
615 | } |
667 | } |
|
|
668 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
669 | encode_str (enc, "null", 4, 0); |
616 | else |
670 | else |
617 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
671 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
618 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
672 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
619 | } |
673 | } |
620 | |
674 | |
… | |
… | |
638 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
692 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
639 | enc->cur += strlen (enc->cur); |
693 | enc->cur += strlen (enc->cur); |
640 | } |
694 | } |
641 | else if (SvIOKp (sv)) |
695 | else if (SvIOKp (sv)) |
642 | { |
696 | { |
643 | // we assume we can always read an IV as a UV |
697 | // we assume we can always read an IV as a UV and vice versa |
644 | if (SvUV (sv) & ~(UV)0x7fff) |
698 | // we assume two's complement |
645 | { |
699 | // we assume no aliasing issues in the union |
646 | // large integer, use the (rather slow) snprintf way. |
700 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
647 | need (enc, sizeof (UV) * 3); |
701 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
648 | enc->cur += |
|
|
649 | SvIsUV(sv) |
|
|
650 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
|
|
651 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
652 | } |
|
|
653 | else |
|
|
654 | { |
702 | { |
655 | // optimise the "small number case" |
703 | // optimise the "small number case" |
656 | // code will likely be branchless and use only a single multiplication |
704 | // code will likely be branchless and use only a single multiplication |
|
|
705 | // works for numbers up to 59074 |
657 | I32 i = SvIV (sv); |
706 | I32 i = SvIVX (sv); |
658 | U32 u; |
707 | U32 u; |
659 | char digit, nz = 0; |
708 | char digit, nz = 0; |
660 | |
709 | |
661 | need (enc, 6); |
710 | need (enc, 6); |
662 | |
711 | |
… | |
… | |
668 | |
717 | |
669 | // now output digit by digit, each time masking out the integer part |
718 | // now output digit by digit, each time masking out the integer part |
670 | // and multiplying by 5 while moving the decimal point one to the right, |
719 | // and multiplying by 5 while moving the decimal point one to the right, |
671 | // resulting in a net multiplication by 10. |
720 | // resulting in a net multiplication by 10. |
672 | // we always write the digit to memory but conditionally increment |
721 | // we always write the digit to memory but conditionally increment |
673 | // the pointer, to ease the usage of conditional move instructions. |
722 | // the pointer, to enable the use of conditional move instructions. |
674 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
723 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
675 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
724 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
676 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
725 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
677 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
726 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
678 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
727 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
679 | } |
728 | } |
|
|
729 | else |
|
|
730 | { |
|
|
731 | // large integer, use the (rather slow) snprintf way. |
|
|
732 | need (enc, IVUV_MAXCHARS); |
|
|
733 | enc->cur += |
|
|
734 | SvIsUV(sv) |
|
|
735 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
736 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
737 | } |
680 | } |
738 | } |
681 | else if (SvROK (sv)) |
739 | else if (SvROK (sv)) |
682 | encode_rv (enc, SvRV (sv)); |
740 | encode_rv (enc, SvRV (sv)); |
683 | else if (!SvOK (sv)) |
741 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
684 | encode_str (enc, "null", 4, 0); |
742 | encode_str (enc, "null", 4, 0); |
685 | else |
743 | else |
686 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
744 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
687 | SvPV_nolen (sv), SvFLAGS (sv)); |
745 | SvPV_nolen (sv), SvFLAGS (sv)); |
688 | } |
746 | } |
… | |
… | |
698 | enc.json = *json; |
756 | enc.json = *json; |
699 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
757 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
700 | enc.cur = SvPVX (enc.sv); |
758 | enc.cur = SvPVX (enc.sv); |
701 | enc.end = SvEND (enc.sv); |
759 | enc.end = SvEND (enc.sv); |
702 | enc.indent = 0; |
760 | enc.indent = 0; |
703 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
|
|
704 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
761 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
705 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
762 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
706 | : 0x10FFFFUL; |
763 | : 0x110000UL; |
707 | |
764 | |
708 | SvPOK_only (enc.sv); |
765 | SvPOK_only (enc.sv); |
709 | encode_sv (&enc, scalar); |
766 | encode_sv (&enc, scalar); |
|
|
767 | encode_nl (&enc); |
710 | |
768 | |
711 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
769 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
712 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
770 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
713 | |
771 | |
714 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
772 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
775 | if (*dec->cur != ch) \ |
833 | if (*dec->cur != ch) \ |
776 | ERR (# ch " expected"); \ |
834 | ERR (# ch " expected"); \ |
777 | ++dec->cur; \ |
835 | ++dec->cur; \ |
778 | SE |
836 | SE |
779 | |
837 | |
780 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
838 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
781 | #define DEC_DEC_DEPTH --dec->depth |
839 | #define DEC_DEC_DEPTH --dec->depth |
782 | |
840 | |
783 | static SV *decode_sv (dec_t *dec); |
841 | static SV *decode_sv (dec_t *dec); |
784 | |
842 | |
785 | static signed char decode_hexdigit[256]; |
843 | static signed char decode_hexdigit[256]; |
… | |
… | |
889 | default: |
947 | default: |
890 | --dec_cur; |
948 | --dec_cur; |
891 | ERR ("illegal backslash escape sequence in string"); |
949 | ERR ("illegal backslash escape sequence in string"); |
892 | } |
950 | } |
893 | } |
951 | } |
894 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
952 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
895 | *cur++ = ch; |
953 | *cur++ = ch; |
896 | else if (ch >= 0x80) |
954 | else if (ch >= 0x80) |
897 | { |
955 | { |
898 | STRLEN clen; |
956 | STRLEN clen; |
899 | UV uch; |
|
|
900 | |
957 | |
901 | --dec_cur; |
958 | --dec_cur; |
902 | |
959 | |
903 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
960 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
904 | if (clen == (STRLEN)-1) |
961 | if (clen == (STRLEN)-1) |
905 | ERR ("malformed UTF-8 character in JSON string"); |
962 | ERR ("malformed UTF-8 character in JSON string"); |
906 | |
963 | |
907 | do |
964 | do |
908 | *cur++ = *dec_cur++; |
965 | *cur++ = *dec_cur++; |
… | |
… | |
925 | { |
982 | { |
926 | STRLEN len = cur - buf; |
983 | STRLEN len = cur - buf; |
927 | |
984 | |
928 | if (sv) |
985 | if (sv) |
929 | { |
986 | { |
930 | SvGROW (sv, SvCUR (sv) + len + 1); |
987 | STRLEN cur = SvCUR (sv); |
|
|
988 | |
|
|
989 | if (SvLEN (sv) <= cur + len) |
|
|
990 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
991 | |
931 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
992 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
932 | SvCUR_set (sv, SvCUR (sv) + len); |
993 | SvCUR_set (sv, SvCUR (sv) + len); |
933 | } |
994 | } |
934 | else |
995 | else |
935 | sv = newSVpvn (buf, len); |
996 | sv = newSVpvn (buf, len); |
… | |
… | |
1026 | |
1087 | |
1027 | // special case the rather common 1..5-digit-int case |
1088 | // special case the rather common 1..5-digit-int case |
1028 | if (*start == '-') |
1089 | if (*start == '-') |
1029 | switch (len) |
1090 | switch (len) |
1030 | { |
1091 | { |
1031 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1092 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
1032 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1093 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
1033 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1094 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1034 | case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1095 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1035 | case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1096 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1036 | } |
1097 | } |
1037 | else |
1098 | else |
1038 | switch (len) |
1099 | switch (len) |
1039 | { |
1100 | { |
1040 | case 1: return newSViv ( start [0] - '0' * 1); |
1101 | case 1: return newSViv ( start [0] - '0' * 1); |
1041 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1102 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1042 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1103 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1043 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1104 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1044 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1105 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1045 | } |
1106 | } |
1046 | |
1107 | |
1047 | { |
1108 | { |
1048 | UV uv; |
1109 | UV uv; |
1049 | int numtype = grok_number (start, len, &uv); |
1110 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1157 | char *p = dec->cur; |
1218 | char *p = dec->cur; |
1158 | char *e = p + 24; // only try up to 24 bytes |
1219 | char *e = p + 24; // only try up to 24 bytes |
1159 | |
1220 | |
1160 | for (;;) |
1221 | for (;;) |
1161 | { |
1222 | { |
1162 | // the >= 0x80 is true on most architectures |
1223 | // the >= 0x80 is false on most architectures |
1163 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1224 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1164 | { |
1225 | { |
1165 | // slow path, back up and use decode_str |
1226 | // slow path, back up and use decode_str |
1166 | SV *key = decode_str (dec); |
1227 | SV *key = decode_str (dec); |
1167 | if (!key) |
1228 | if (!key) |
… | |
… | |
1298 | |
1359 | |
1299 | static SV * |
1360 | static SV * |
1300 | decode_sv (dec_t *dec) |
1361 | decode_sv (dec_t *dec) |
1301 | { |
1362 | { |
1302 | // the beauty of JSON: you need exactly one character lookahead |
1363 | // the beauty of JSON: you need exactly one character lookahead |
1303 | // to parse anything. |
1364 | // to parse everything. |
1304 | switch (*dec->cur) |
1365 | switch (*dec->cur) |
1305 | { |
1366 | { |
1306 | case '"': ++dec->cur; return decode_str (dec); |
1367 | case '"': ++dec->cur; return decode_str (dec); |
1307 | case '[': ++dec->cur; return decode_av (dec); |
1368 | case '[': ++dec->cur; return decode_av (dec); |
1308 | case '{': ++dec->cur; return decode_hv (dec); |
1369 | case '{': ++dec->cur; return decode_hv (dec); |
1309 | |
1370 | |
1310 | case '-': |
1371 | case '-': |
1311 | case '0': case '1': case '2': case '3': case '4': |
1372 | case '0': case '1': case '2': case '3': case '4': |
1312 | case '5': case '6': case '7': case '8': case '9': |
1373 | case '5': case '6': case '7': case '8': case '9': |
1313 | return decode_num (dec); |
1374 | return decode_num (dec); |
… | |
… | |
1315 | case 't': |
1376 | case 't': |
1316 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1377 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1317 | { |
1378 | { |
1318 | dec->cur += 4; |
1379 | dec->cur += 4; |
1319 | #if JSON_SLOW |
1380 | #if JSON_SLOW |
1320 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
1381 | json_true = get_bool ("JSON::XS::true"); |
1321 | #endif |
1382 | #endif |
1322 | return SvREFCNT_inc (json_true); |
1383 | return newSVsv (json_true); |
1323 | } |
1384 | } |
1324 | else |
1385 | else |
1325 | ERR ("'true' expected"); |
1386 | ERR ("'true' expected"); |
1326 | |
1387 | |
1327 | break; |
1388 | break; |
… | |
… | |
1329 | case 'f': |
1390 | case 'f': |
1330 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1391 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1331 | { |
1392 | { |
1332 | dec->cur += 5; |
1393 | dec->cur += 5; |
1333 | #if JSON_SLOW |
1394 | #if JSON_SLOW |
1334 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1395 | json_false = get_bool ("JSON::XS::false"); |
1335 | #endif |
1396 | #endif |
1336 | return SvREFCNT_inc (json_false); |
1397 | return newSVsv (json_false); |
1337 | } |
1398 | } |
1338 | else |
1399 | else |
1339 | ERR ("'false' expected"); |
1400 | ERR ("'false' expected"); |
1340 | |
1401 | |
1341 | break; |
1402 | break; |
… | |
… | |
1359 | fail: |
1420 | fail: |
1360 | return 0; |
1421 | return 0; |
1361 | } |
1422 | } |
1362 | |
1423 | |
1363 | static SV * |
1424 | static SV * |
1364 | decode_json (SV *string, JSON *json, UV *offset_return) |
1425 | decode_json (SV *string, JSON *json, char **offset_return) |
1365 | { |
1426 | { |
1366 | dec_t dec; |
1427 | dec_t dec; |
1367 | UV offset; |
|
|
1368 | SV *sv; |
1428 | SV *sv; |
1369 | |
1429 | |
|
|
1430 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1431 | * will perl ignore the magic in subsequent accesses |
|
|
1432 | */ |
1370 | SvGETMAGIC (string); |
1433 | /*SvGETMAGIC (string);*/ |
|
|
1434 | if (SvMAGICAL (string)) |
|
|
1435 | string = sv_2mortal (newSVsv (string)); |
|
|
1436 | |
1371 | SvUPGRADE (string, SVt_PV); |
1437 | SvUPGRADE (string, SVt_PV); |
1372 | |
1438 | |
1373 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1439 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1440 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1441 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1442 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1443 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1444 | * and hope for the best. |
|
|
1445 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1446 | * assertion business is seriously broken, try yet another workaround |
|
|
1447 | * for the broken -DDEBUGGING. |
|
|
1448 | */ |
|
|
1449 | { |
|
|
1450 | #ifdef DEBUGGING |
|
|
1451 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1452 | #else |
|
|
1453 | STRLEN offset = SvCUR (string); |
|
|
1454 | #endif |
|
|
1455 | |
|
|
1456 | if (offset > json->max_size && json->max_size) |
1374 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1457 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1375 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1458 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1459 | } |
1376 | |
1460 | |
1377 | if (json->flags & F_UTF8) |
1461 | if (json->flags & F_UTF8) |
1378 | sv_utf8_downgrade (string, 0); |
1462 | sv_utf8_downgrade (string, 0); |
1379 | else |
1463 | else |
1380 | sv_utf8_upgrade (string); |
1464 | sv_utf8_upgrade (string); |
1381 | |
1465 | |
1382 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1466 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1383 | |
1467 | |
1384 | dec.json = *json; |
1468 | dec.json = *json; |
1385 | dec.cur = SvPVX (string); |
1469 | dec.cur = SvPVX (string); |
1386 | dec.end = SvEND (string); |
1470 | dec.end = SvEND (string); |
1387 | dec.err = 0; |
1471 | dec.err = 0; |
1388 | dec.depth = 0; |
1472 | dec.depth = 0; |
1389 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1390 | |
1473 | |
1391 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1474 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1392 | dec.json.flags |= F_HOOK; |
1475 | dec.json.flags |= F_HOOK; |
1393 | |
1476 | |
1394 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1477 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1395 | |
1478 | |
1396 | decode_ws (&dec); |
1479 | decode_ws (&dec); |
1397 | sv = decode_sv (&dec); |
1480 | sv = decode_sv (&dec); |
|
|
1481 | |
|
|
1482 | if (offset_return) |
|
|
1483 | *offset_return = dec.cur; |
1398 | |
1484 | |
1399 | if (!(offset_return || !sv)) |
1485 | if (!(offset_return || !sv)) |
1400 | { |
1486 | { |
1401 | // check for trailing garbage |
1487 | // check for trailing garbage |
1402 | decode_ws (&dec); |
1488 | decode_ws (&dec); |
… | |
… | |
1405 | { |
1491 | { |
1406 | dec.err = "garbage after JSON object"; |
1492 | dec.err = "garbage after JSON object"; |
1407 | SvREFCNT_dec (sv); |
1493 | SvREFCNT_dec (sv); |
1408 | sv = 0; |
1494 | sv = 0; |
1409 | } |
1495 | } |
1410 | } |
|
|
1411 | |
|
|
1412 | if (offset_return || !sv) |
|
|
1413 | { |
|
|
1414 | offset = dec.json.flags & F_UTF8 |
|
|
1415 | ? dec.cur - SvPVX (string) |
|
|
1416 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1417 | |
|
|
1418 | if (offset_return) |
|
|
1419 | *offset_return = offset; |
|
|
1420 | } |
1496 | } |
1421 | |
1497 | |
1422 | if (!sv) |
1498 | if (!sv) |
1423 | { |
1499 | { |
1424 | SV *uni = sv_newmortal (); |
1500 | SV *uni = sv_newmortal (); |
… | |
… | |
1430 | SAVEVPTR (PL_curcop); |
1506 | SAVEVPTR (PL_curcop); |
1431 | PL_curcop = &cop; |
1507 | PL_curcop = &cop; |
1432 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1508 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1433 | LEAVE; |
1509 | LEAVE; |
1434 | |
1510 | |
1435 | croak ("%s, at character offset %d [\"%s\"]", |
1511 | croak ("%s, at character offset %d (before \"%s\")", |
1436 | dec.err, |
1512 | dec.err, |
1437 | (int)offset, |
1513 | ptr_to_index (string, dec.cur), |
1438 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1514 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1439 | } |
1515 | } |
1440 | |
1516 | |
1441 | sv = sv_2mortal (sv); |
1517 | sv = sv_2mortal (sv); |
1442 | |
1518 | |
1443 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1519 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1444 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1520 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1445 | |
1521 | |
1446 | return sv; |
1522 | return sv; |
|
|
1523 | } |
|
|
1524 | |
|
|
1525 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1526 | // incremental parser |
|
|
1527 | |
|
|
1528 | static void |
|
|
1529 | incr_parse (JSON *self) |
|
|
1530 | { |
|
|
1531 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1532 | |
|
|
1533 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1534 | // but this would make it slower, so... |
|
|
1535 | |
|
|
1536 | for (;;) |
|
|
1537 | { |
|
|
1538 | //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 |
|
|
1539 | switch (self->incr_mode) |
|
|
1540 | { |
|
|
1541 | // only used for initial whitespace skipping |
|
|
1542 | case INCR_M_WS: |
|
|
1543 | for (;;) |
|
|
1544 | { |
|
|
1545 | if (*p > 0x20) |
|
|
1546 | { |
|
|
1547 | if (*p == '#') |
|
|
1548 | { |
|
|
1549 | self->incr_mode = INCR_M_C0; |
|
|
1550 | goto incr_m_c; |
|
|
1551 | } |
|
|
1552 | else |
|
|
1553 | { |
|
|
1554 | self->incr_mode = INCR_M_JSON; |
|
|
1555 | goto incr_m_json; |
|
|
1556 | } |
|
|
1557 | } |
|
|
1558 | else if (!*p) |
|
|
1559 | goto interrupt; |
|
|
1560 | |
|
|
1561 | ++p; |
|
|
1562 | } |
|
|
1563 | |
|
|
1564 | // skip a single char inside a string (for \\-processing) |
|
|
1565 | case INCR_M_BS: |
|
|
1566 | if (!*p) |
|
|
1567 | goto interrupt; |
|
|
1568 | |
|
|
1569 | ++p; |
|
|
1570 | self->incr_mode = INCR_M_STR; |
|
|
1571 | goto incr_m_str; |
|
|
1572 | |
|
|
1573 | // inside #-style comments |
|
|
1574 | case INCR_M_C0: |
|
|
1575 | case INCR_M_C1: |
|
|
1576 | incr_m_c: |
|
|
1577 | for (;;) |
|
|
1578 | { |
|
|
1579 | if (*p == '\n') |
|
|
1580 | { |
|
|
1581 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1582 | break; |
|
|
1583 | } |
|
|
1584 | else if (!*p) |
|
|
1585 | goto interrupt; |
|
|
1586 | |
|
|
1587 | ++p; |
|
|
1588 | } |
|
|
1589 | |
|
|
1590 | break; |
|
|
1591 | |
|
|
1592 | // inside a string |
|
|
1593 | case INCR_M_STR: |
|
|
1594 | incr_m_str: |
|
|
1595 | for (;;) |
|
|
1596 | { |
|
|
1597 | if (*p == '"') |
|
|
1598 | { |
|
|
1599 | ++p; |
|
|
1600 | self->incr_mode = INCR_M_JSON; |
|
|
1601 | |
|
|
1602 | if (!self->incr_nest) |
|
|
1603 | goto interrupt; |
|
|
1604 | |
|
|
1605 | goto incr_m_json; |
|
|
1606 | } |
|
|
1607 | else if (*p == '\\') |
|
|
1608 | { |
|
|
1609 | ++p; // "virtually" consumes character after \ |
|
|
1610 | |
|
|
1611 | if (!*p) // if at end of string we have to switch modes |
|
|
1612 | { |
|
|
1613 | self->incr_mode = INCR_M_BS; |
|
|
1614 | goto interrupt; |
|
|
1615 | } |
|
|
1616 | } |
|
|
1617 | else if (!*p) |
|
|
1618 | goto interrupt; |
|
|
1619 | |
|
|
1620 | ++p; |
|
|
1621 | } |
|
|
1622 | |
|
|
1623 | // after initial ws, outside string |
|
|
1624 | case INCR_M_JSON: |
|
|
1625 | incr_m_json: |
|
|
1626 | for (;;) |
|
|
1627 | { |
|
|
1628 | switch (*p++) |
|
|
1629 | { |
|
|
1630 | case 0: |
|
|
1631 | --p; |
|
|
1632 | goto interrupt; |
|
|
1633 | |
|
|
1634 | case 0x09: |
|
|
1635 | case 0x0a: |
|
|
1636 | case 0x0d: |
|
|
1637 | case 0x20: |
|
|
1638 | if (!self->incr_nest) |
|
|
1639 | { |
|
|
1640 | --p; // do not eat the whitespace, let the next round do it |
|
|
1641 | goto interrupt; |
|
|
1642 | } |
|
|
1643 | break; |
|
|
1644 | |
|
|
1645 | case '"': |
|
|
1646 | self->incr_mode = INCR_M_STR; |
|
|
1647 | goto incr_m_str; |
|
|
1648 | |
|
|
1649 | case '[': |
|
|
1650 | case '{': |
|
|
1651 | if (++self->incr_nest > self->max_depth) |
|
|
1652 | croak (ERR_NESTING_EXCEEDED); |
|
|
1653 | break; |
|
|
1654 | |
|
|
1655 | case ']': |
|
|
1656 | case '}': |
|
|
1657 | if (--self->incr_nest <= 0) |
|
|
1658 | goto interrupt; |
|
|
1659 | break; |
|
|
1660 | |
|
|
1661 | case '#': |
|
|
1662 | self->incr_mode = INCR_M_C1; |
|
|
1663 | goto incr_m_c; |
|
|
1664 | } |
|
|
1665 | } |
|
|
1666 | } |
|
|
1667 | |
|
|
1668 | modechange: |
|
|
1669 | ; |
|
|
1670 | } |
|
|
1671 | |
|
|
1672 | interrupt: |
|
|
1673 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1674 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1675 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1447 | } |
1676 | } |
1448 | |
1677 | |
1449 | ///////////////////////////////////////////////////////////////////////////// |
1678 | ///////////////////////////////////////////////////////////////////////////// |
1450 | // XS interface functions |
1679 | // XS interface functions |
1451 | |
1680 | |
… | |
… | |
1463 | : -1; |
1692 | : -1; |
1464 | |
1693 | |
1465 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1694 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1466 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1695 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1467 | |
1696 | |
1468 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1697 | json_true = get_bool ("JSON::XS::true"); |
1469 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1698 | json_false = get_bool ("JSON::XS::false"); |
|
|
1699 | |
|
|
1700 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1470 | } |
1701 | } |
1471 | |
1702 | |
1472 | PROTOTYPES: DISABLE |
1703 | PROTOTYPES: DISABLE |
1473 | |
1704 | |
1474 | void CLONE (...) |
1705 | void CLONE (...) |
… | |
… | |
1479 | void new (char *klass) |
1710 | void new (char *klass) |
1480 | PPCODE: |
1711 | PPCODE: |
1481 | { |
1712 | { |
1482 | SV *pv = NEWSV (0, sizeof (JSON)); |
1713 | SV *pv = NEWSV (0, sizeof (JSON)); |
1483 | SvPOK_only (pv); |
1714 | SvPOK_only (pv); |
1484 | Zero (SvPVX (pv), 1, JSON); |
1715 | json_init ((JSON *)SvPVX (pv)); |
1485 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1486 | XPUSHs (sv_2mortal (sv_bless ( |
1716 | XPUSHs (sv_2mortal (sv_bless ( |
1487 | newRV_noinc (pv), |
1717 | newRV_noinc (pv), |
1488 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1718 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1489 | ))); |
1719 | ))); |
1490 | } |
1720 | } |
… | |
… | |
1502 | allow_nonref = F_ALLOW_NONREF |
1732 | allow_nonref = F_ALLOW_NONREF |
1503 | shrink = F_SHRINK |
1733 | shrink = F_SHRINK |
1504 | allow_blessed = F_ALLOW_BLESSED |
1734 | allow_blessed = F_ALLOW_BLESSED |
1505 | convert_blessed = F_CONV_BLESSED |
1735 | convert_blessed = F_CONV_BLESSED |
1506 | relaxed = F_RELAXED |
1736 | relaxed = F_RELAXED |
|
|
1737 | allow_unknown = F_ALLOW_UNKNOWN |
1507 | PPCODE: |
1738 | PPCODE: |
1508 | { |
1739 | { |
1509 | if (enable) |
1740 | if (enable) |
1510 | self->flags |= ix; |
1741 | self->flags |= ix; |
1511 | else |
1742 | else |
… | |
… | |
1526 | get_allow_nonref = F_ALLOW_NONREF |
1757 | get_allow_nonref = F_ALLOW_NONREF |
1527 | get_shrink = F_SHRINK |
1758 | get_shrink = F_SHRINK |
1528 | get_allow_blessed = F_ALLOW_BLESSED |
1759 | get_allow_blessed = F_ALLOW_BLESSED |
1529 | get_convert_blessed = F_CONV_BLESSED |
1760 | get_convert_blessed = F_CONV_BLESSED |
1530 | get_relaxed = F_RELAXED |
1761 | get_relaxed = F_RELAXED |
|
|
1762 | get_allow_unknown = F_ALLOW_UNKNOWN |
1531 | PPCODE: |
1763 | PPCODE: |
1532 | XPUSHs (boolSV (self->flags & ix)); |
1764 | XPUSHs (boolSV (self->flags & ix)); |
1533 | |
1765 | |
1534 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1766 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1535 | PPCODE: |
1767 | PPCODE: |
1536 | { |
1768 | self->max_depth = max_depth; |
1537 | UV log2 = 0; |
|
|
1538 | |
|
|
1539 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
|
|
1540 | |
|
|
1541 | while ((1UL << log2) < max_depth) |
|
|
1542 | ++log2; |
|
|
1543 | |
|
|
1544 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1545 | |
|
|
1546 | XPUSHs (ST (0)); |
1769 | XPUSHs (ST (0)); |
1547 | } |
|
|
1548 | |
1770 | |
1549 | U32 get_max_depth (JSON *self) |
1771 | U32 get_max_depth (JSON *self) |
1550 | CODE: |
1772 | CODE: |
1551 | RETVAL = DEC_DEPTH (self->flags); |
1773 | RETVAL = self->max_depth; |
1552 | OUTPUT: |
1774 | OUTPUT: |
1553 | RETVAL |
1775 | RETVAL |
1554 | |
1776 | |
1555 | void max_size (JSON *self, UV max_size = 0) |
1777 | void max_size (JSON *self, U32 max_size = 0) |
1556 | PPCODE: |
1778 | PPCODE: |
1557 | { |
1779 | self->max_size = max_size; |
1558 | UV log2 = 0; |
|
|
1559 | |
|
|
1560 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1561 | if (max_size == 1) max_size = 2; |
|
|
1562 | |
|
|
1563 | while ((1UL << log2) < max_size) |
|
|
1564 | ++log2; |
|
|
1565 | |
|
|
1566 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1567 | |
|
|
1568 | XPUSHs (ST (0)); |
1780 | XPUSHs (ST (0)); |
1569 | } |
|
|
1570 | |
1781 | |
1571 | int get_max_size (JSON *self) |
1782 | int get_max_size (JSON *self) |
1572 | CODE: |
1783 | CODE: |
1573 | RETVAL = DEC_SIZE (self->flags); |
1784 | RETVAL = self->max_size; |
1574 | OUTPUT: |
1785 | OUTPUT: |
1575 | RETVAL |
1786 | RETVAL |
1576 | |
1787 | |
1577 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1788 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1578 | PPCODE: |
1789 | PPCODE: |
… | |
… | |
1614 | XPUSHs (decode_json (jsonstr, self, 0)); |
1825 | XPUSHs (decode_json (jsonstr, self, 0)); |
1615 | |
1826 | |
1616 | void decode_prefix (JSON *self, SV *jsonstr) |
1827 | void decode_prefix (JSON *self, SV *jsonstr) |
1617 | PPCODE: |
1828 | PPCODE: |
1618 | { |
1829 | { |
1619 | UV offset; |
1830 | char *offset; |
1620 | EXTEND (SP, 2); |
1831 | EXTEND (SP, 2); |
1621 | PUSHs (decode_json (jsonstr, self, &offset)); |
1832 | PUSHs (decode_json (jsonstr, self, &offset)); |
1622 | PUSHs (sv_2mortal (newSVuv (offset))); |
1833 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1834 | } |
|
|
1835 | |
|
|
1836 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1837 | PPCODE: |
|
|
1838 | { |
|
|
1839 | if (!self->incr_text) |
|
|
1840 | self->incr_text = newSVpvn ("", 0); |
|
|
1841 | |
|
|
1842 | // append data, if any |
|
|
1843 | if (jsonstr) |
|
|
1844 | { |
|
|
1845 | if (SvUTF8 (jsonstr)) |
|
|
1846 | { |
|
|
1847 | if (!SvUTF8 (self->incr_text)) |
|
|
1848 | { |
|
|
1849 | /* utf-8-ness differs, need to upgrade */ |
|
|
1850 | sv_utf8_upgrade (self->incr_text); |
|
|
1851 | |
|
|
1852 | if (self->incr_pos) |
|
|
1853 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1854 | - (U8 *)SvPVX (self->incr_text); |
|
|
1855 | } |
|
|
1856 | } |
|
|
1857 | else if (SvUTF8 (self->incr_text)) |
|
|
1858 | sv_utf8_upgrade (jsonstr); |
|
|
1859 | |
|
|
1860 | { |
|
|
1861 | STRLEN len; |
|
|
1862 | const char *str = SvPV (jsonstr, len); |
|
|
1863 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1864 | |
|
|
1865 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1866 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1867 | |
|
|
1868 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1869 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1870 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1871 | } |
|
|
1872 | } |
|
|
1873 | |
|
|
1874 | if (GIMME_V != G_VOID) |
|
|
1875 | do |
|
|
1876 | { |
|
|
1877 | char *offset; |
|
|
1878 | |
|
|
1879 | if (!INCR_DONE (self)) |
|
|
1880 | { |
|
|
1881 | incr_parse (self); |
|
|
1882 | |
|
|
1883 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1884 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1885 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1886 | |
|
|
1887 | if (!INCR_DONE (self)) |
|
|
1888 | break; |
|
|
1889 | } |
|
|
1890 | |
|
|
1891 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1892 | |
|
|
1893 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
1894 | self->incr_nest = 0; |
|
|
1895 | self->incr_mode = 0; |
|
|
1896 | |
|
|
1897 | sv_chop (self->incr_text, offset); |
|
|
1898 | } |
|
|
1899 | while (GIMME_V == G_ARRAY); |
|
|
1900 | } |
|
|
1901 | |
|
|
1902 | SV *incr_text (JSON *self) |
|
|
1903 | ATTRS: lvalue |
|
|
1904 | CODE: |
|
|
1905 | { |
|
|
1906 | if (self->incr_pos) |
|
|
1907 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
1908 | |
|
|
1909 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
1910 | } |
|
|
1911 | OUTPUT: |
|
|
1912 | RETVAL |
|
|
1913 | |
|
|
1914 | void incr_skip (JSON *self) |
|
|
1915 | CODE: |
|
|
1916 | { |
|
|
1917 | if (self->incr_pos) |
|
|
1918 | { |
|
|
1919 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
1920 | self->incr_pos = 0; |
|
|
1921 | self->incr_nest = 0; |
|
|
1922 | self->incr_mode = 0; |
|
|
1923 | } |
|
|
1924 | } |
|
|
1925 | |
|
|
1926 | void incr_reset (JSON *self) |
|
|
1927 | CODE: |
|
|
1928 | { |
|
|
1929 | SvREFCNT_dec (self->incr_text); |
|
|
1930 | self->incr_text = 0; |
|
|
1931 | self->incr_pos = 0; |
|
|
1932 | self->incr_nest = 0; |
|
|
1933 | self->incr_mode = 0; |
1623 | } |
1934 | } |
1624 | |
1935 | |
1625 | void DESTROY (JSON *self) |
1936 | void DESTROY (JSON *self) |
1626 | CODE: |
1937 | CODE: |
1627 | SvREFCNT_dec (self->cb_sk_object); |
1938 | SvREFCNT_dec (self->cb_sk_object); |
1628 | SvREFCNT_dec (self->cb_object); |
1939 | SvREFCNT_dec (self->cb_object); |
|
|
1940 | SvREFCNT_dec (self->incr_text); |
1629 | |
1941 | |
1630 | PROTOTYPES: ENABLE |
1942 | PROTOTYPES: ENABLE |
1631 | |
1943 | |
1632 | void encode_json (SV *scalar) |
1944 | void encode_json (SV *scalar) |
|
|
1945 | ALIAS: |
|
|
1946 | to_json_ = 0 |
|
|
1947 | encode_json = F_UTF8 |
1633 | PPCODE: |
1948 | PPCODE: |
1634 | { |
1949 | { |
1635 | JSON json = { F_DEFAULT | F_UTF8 }; |
1950 | JSON json; |
|
|
1951 | json_init (&json); |
|
|
1952 | json.flags |= ix; |
1636 | XPUSHs (encode_json (scalar, &json)); |
1953 | XPUSHs (encode_json (scalar, &json)); |
1637 | } |
1954 | } |
1638 | |
1955 | |
1639 | void decode_json (SV *jsonstr) |
1956 | void decode_json (SV *jsonstr) |
|
|
1957 | ALIAS: |
|
|
1958 | from_json_ = 0 |
|
|
1959 | decode_json = F_UTF8 |
1640 | PPCODE: |
1960 | PPCODE: |
1641 | { |
1961 | { |
1642 | JSON json = { F_DEFAULT | F_UTF8 }; |
1962 | JSON json; |
|
|
1963 | json_init (&json); |
|
|
1964 | json.flags |= ix; |
1643 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1965 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1644 | } |
1966 | } |
1645 | |
1967 | |
|
|
1968 | |