1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
8 | #include "stdio.h" |
8 | #include <stdio.h> |
|
|
9 | #include <float.h> |
9 | |
10 | |
10 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
11 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
11 | # define snprintf _snprintf // C compilers have this in stdio.h |
12 | # define snprintf _snprintf // C compilers have this in stdio.h |
12 | #endif |
13 | #endif |
13 | |
14 | |
|
|
15 | // 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 | #ifndef UTF8_MAXBYTES |
|
|
18 | # define UTF8_MAXBYTES 13 |
|
|
19 | #endif |
|
|
20 | |
14 | #define F_ASCII 0x00000001UL |
21 | #define F_ASCII 0x00000001UL |
15 | #define F_LATIN1 0x00000002UL |
22 | #define F_LATIN1 0x00000002UL |
16 | #define F_UTF8 0x00000004UL |
23 | #define F_UTF8 0x00000004UL |
17 | #define F_INDENT 0x00000008UL |
24 | #define F_INDENT 0x00000008UL |
18 | #define F_CANONICAL 0x00000010UL |
25 | #define F_CANONICAL 0x00000010UL |
19 | #define F_SPACE_BEFORE 0x00000020UL |
26 | #define F_SPACE_BEFORE 0x00000020UL |
20 | #define F_SPACE_AFTER 0x00000040UL |
27 | #define F_SPACE_AFTER 0x00000040UL |
21 | #define F_ALLOW_NONREF 0x00000100UL |
28 | #define F_ALLOW_NONREF 0x00000100UL |
22 | #define F_SHRINK 0x00000200UL |
29 | #define F_SHRINK 0x00000200UL |
|
|
30 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
31 | #define F_CONV_BLESSED 0x00000800UL |
23 | #define F_MAXDEPTH 0xf8000000UL |
32 | #define F_MAXDEPTH 0xf8000000UL |
24 | #define S_MAXDEPTH 27 |
33 | #define S_MAXDEPTH 27 |
|
|
34 | #define F_MAXSIZE 0x01f00000UL |
|
|
35 | #define S_MAXSIZE 20 |
|
|
36 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
25 | |
37 | |
26 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
38 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
27 | |
39 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
28 | // F_SELFCONVERT? <=> to_json/toJson |
|
|
29 | // F_BLESSED? <=> { $__class__$ => } |
|
|
30 | |
40 | |
31 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
41 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
32 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
42 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
33 | |
43 | |
34 | #define INIT_SIZE 32 // initial scalar size to be allocated |
44 | #define INIT_SIZE 32 // initial scalar size to be allocated |
35 | #define INDENT_STEP 3 // spaces per indentation level |
45 | #define INDENT_STEP 3 // spaces per indentation level |
36 | |
46 | |
37 | #define SHORT_STRING_LEN 512 // special-case strings of up to this size |
47 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
38 | |
48 | |
39 | #define SB do { |
49 | #define SB do { |
40 | #define SE } while (0) |
50 | #define SE } while (0) |
41 | |
51 | |
|
|
52 | #if __GNUC__ >= 3 |
|
|
53 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
54 | # define inline inline |
|
|
55 | #else |
|
|
56 | # define expect(expr,value) (expr) |
|
|
57 | # define inline static |
|
|
58 | #endif |
|
|
59 | |
|
|
60 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
61 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
62 | |
|
|
63 | #ifdef USE_ITHREADS |
|
|
64 | # define JSON_SLOW 1 |
|
|
65 | #else |
|
|
66 | # define JSON_SLOW 1 |
|
|
67 | #endif |
|
|
68 | |
42 | static HV *json_stash; // JSON::XS:: |
69 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
|
|
70 | static SV *json_true, *json_false; |
|
|
71 | |
|
|
72 | typedef struct { |
|
|
73 | U32 flags; |
|
|
74 | SV *cb_object; |
|
|
75 | HV *cb_sk_object; |
|
|
76 | } JSON; |
43 | |
77 | |
44 | ///////////////////////////////////////////////////////////////////////////// |
78 | ///////////////////////////////////////////////////////////////////////////// |
45 | // utility functions |
79 | // utility functions |
46 | |
80 | |
47 | static UV * |
81 | inline void |
48 | SvJSON (SV *sv) |
|
|
49 | { |
|
|
50 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
|
|
51 | croak ("object is not of type JSON::XS"); |
|
|
52 | |
|
|
53 | return &SvUVX (SvRV (sv)); |
|
|
54 | } |
|
|
55 | |
|
|
56 | static void |
|
|
57 | shrink (SV *sv) |
82 | shrink (SV *sv) |
58 | { |
83 | { |
59 | sv_utf8_downgrade (sv, 1); |
84 | sv_utf8_downgrade (sv, 1); |
60 | if (SvLEN (sv) > SvCUR (sv) + 1) |
85 | if (SvLEN (sv) > SvCUR (sv) + 1) |
61 | { |
86 | { |
… | |
… | |
70 | // decode an utf-8 character and return it, or (UV)-1 in |
95 | // decode an utf-8 character and return it, or (UV)-1 in |
71 | // case of an error. |
96 | // case of an error. |
72 | // we special-case "safe" characters from U+80 .. U+7FF, |
97 | // we special-case "safe" characters from U+80 .. U+7FF, |
73 | // but use the very good perl function to parse anything else. |
98 | // but use the very good perl function to parse anything else. |
74 | // note that we never call this function for a ascii codepoints |
99 | // note that we never call this function for a ascii codepoints |
75 | static UV |
100 | inline UV |
76 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
101 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
77 | { |
102 | { |
78 | if (s[0] > 0xdf || s[0] < 0xc2) |
103 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
79 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
104 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
80 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
105 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
81 | { |
106 | { |
82 | *clen = 2; |
107 | *clen = 2; |
83 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
108 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
… | |
… | |
96 | typedef struct |
121 | typedef struct |
97 | { |
122 | { |
98 | char *cur; // SvPVX (sv) + current output position |
123 | char *cur; // SvPVX (sv) + current output position |
99 | char *end; // SvEND (sv) |
124 | char *end; // SvEND (sv) |
100 | SV *sv; // result scalar |
125 | SV *sv; // result scalar |
101 | U32 flags; // F_* |
126 | JSON json; |
102 | U32 indent; // indentation level |
127 | U32 indent; // indentation level |
103 | U32 maxdepth; // max. indentation/recursion level |
128 | U32 maxdepth; // max. indentation/recursion level |
104 | } enc_t; |
129 | } enc_t; |
105 | |
130 | |
106 | static void |
131 | inline void |
107 | need (enc_t *enc, STRLEN len) |
132 | need (enc_t *enc, STRLEN len) |
108 | { |
133 | { |
109 | if (enc->cur + len >= enc->end) |
134 | if (expect_false (enc->cur + len >= enc->end)) |
110 | { |
135 | { |
111 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
136 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
112 | SvGROW (enc->sv, cur + len + 1); |
137 | SvGROW (enc->sv, cur + len + 1); |
113 | enc->cur = SvPVX (enc->sv) + cur; |
138 | enc->cur = SvPVX (enc->sv) + cur; |
114 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
139 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
115 | } |
140 | } |
116 | } |
141 | } |
117 | |
142 | |
118 | static void |
143 | inline void |
119 | encode_ch (enc_t *enc, char ch) |
144 | encode_ch (enc_t *enc, char ch) |
120 | { |
145 | { |
121 | need (enc, 1); |
146 | need (enc, 1); |
122 | *enc->cur++ = ch; |
147 | *enc->cur++ = ch; |
123 | } |
148 | } |
… | |
… | |
131 | |
156 | |
132 | while (str < end) |
157 | while (str < end) |
133 | { |
158 | { |
134 | unsigned char ch = *(unsigned char *)str; |
159 | unsigned char ch = *(unsigned char *)str; |
135 | |
160 | |
136 | if (ch >= 0x20 && ch < 0x80) // most common case |
161 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
137 | { |
162 | { |
138 | if (ch == '"') // but with slow exceptions |
163 | if (expect_false (ch == '"')) // but with slow exceptions |
139 | { |
164 | { |
140 | need (enc, len += 1); |
165 | need (enc, len += 1); |
141 | *enc->cur++ = '\\'; |
166 | *enc->cur++ = '\\'; |
142 | *enc->cur++ = '"'; |
167 | *enc->cur++ = '"'; |
143 | } |
168 | } |
144 | else if (ch == '\\') |
169 | else if (expect_false (ch == '\\')) |
145 | { |
170 | { |
146 | need (enc, len += 1); |
171 | need (enc, len += 1); |
147 | *enc->cur++ = '\\'; |
172 | *enc->cur++ = '\\'; |
148 | *enc->cur++ = '\\'; |
173 | *enc->cur++ = '\\'; |
149 | } |
174 | } |
… | |
… | |
167 | STRLEN clen; |
192 | STRLEN clen; |
168 | UV uch; |
193 | UV uch; |
169 | |
194 | |
170 | if (is_utf8) |
195 | if (is_utf8) |
171 | { |
196 | { |
172 | //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
|
|
173 | uch = decode_utf8 (str, end - str, &clen); |
197 | uch = decode_utf8 (str, end - str, &clen); |
174 | if (clen == (STRLEN)-1) |
198 | if (clen == (STRLEN)-1) |
175 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
199 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
176 | } |
200 | } |
177 | else |
201 | else |
… | |
… | |
181 | } |
205 | } |
182 | |
206 | |
183 | if (uch > 0x10FFFFUL) |
207 | if (uch > 0x10FFFFUL) |
184 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
208 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
185 | |
209 | |
186 | if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF)) |
210 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
187 | { |
211 | { |
188 | if (uch > 0xFFFFUL) |
212 | if (uch > 0xFFFFUL) |
189 | { |
213 | { |
190 | need (enc, len += 11); |
214 | need (enc, len += 11); |
191 | sprintf (enc->cur, "\\u%04x\\u%04x", |
215 | sprintf (enc->cur, "\\u%04x\\u%04x", |
… | |
… | |
205 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
229 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
206 | } |
230 | } |
207 | |
231 | |
208 | str += clen; |
232 | str += clen; |
209 | } |
233 | } |
210 | else if (enc->flags & F_LATIN1) |
234 | else if (enc->json.flags & F_LATIN1) |
211 | { |
235 | { |
212 | *enc->cur++ = uch; |
236 | *enc->cur++ = uch; |
213 | str += clen; |
237 | str += clen; |
214 | } |
238 | } |
215 | else if (is_utf8) |
239 | else if (is_utf8) |
… | |
… | |
233 | |
257 | |
234 | --len; |
258 | --len; |
235 | } |
259 | } |
236 | } |
260 | } |
237 | |
261 | |
238 | static void |
262 | inline void |
239 | encode_indent (enc_t *enc) |
263 | encode_indent (enc_t *enc) |
240 | { |
264 | { |
241 | if (enc->flags & F_INDENT) |
265 | if (enc->json.flags & F_INDENT) |
242 | { |
266 | { |
243 | int spaces = enc->indent * INDENT_STEP; |
267 | int spaces = enc->indent * INDENT_STEP; |
244 | |
268 | |
245 | need (enc, spaces); |
269 | need (enc, spaces); |
246 | memset (enc->cur, ' ', spaces); |
270 | memset (enc->cur, ' ', spaces); |
247 | enc->cur += spaces; |
271 | enc->cur += spaces; |
248 | } |
272 | } |
249 | } |
273 | } |
250 | |
274 | |
251 | static void |
275 | inline void |
252 | encode_space (enc_t *enc) |
276 | encode_space (enc_t *enc) |
253 | { |
277 | { |
254 | need (enc, 1); |
278 | need (enc, 1); |
255 | encode_ch (enc, ' '); |
279 | encode_ch (enc, ' '); |
256 | } |
280 | } |
257 | |
281 | |
258 | static void |
282 | inline void |
259 | encode_nl (enc_t *enc) |
283 | encode_nl (enc_t *enc) |
260 | { |
284 | { |
261 | if (enc->flags & F_INDENT) |
285 | if (enc->json.flags & F_INDENT) |
262 | { |
286 | { |
263 | need (enc, 1); |
287 | need (enc, 1); |
264 | encode_ch (enc, '\n'); |
288 | encode_ch (enc, '\n'); |
265 | } |
289 | } |
266 | } |
290 | } |
267 | |
291 | |
268 | static void |
292 | inline void |
269 | encode_comma (enc_t *enc) |
293 | encode_comma (enc_t *enc) |
270 | { |
294 | { |
271 | encode_ch (enc, ','); |
295 | encode_ch (enc, ','); |
272 | |
296 | |
273 | if (enc->flags & F_INDENT) |
297 | if (enc->json.flags & F_INDENT) |
274 | encode_nl (enc); |
298 | encode_nl (enc); |
275 | else if (enc->flags & F_SPACE_AFTER) |
299 | else if (enc->json.flags & F_SPACE_AFTER) |
276 | encode_space (enc); |
300 | encode_space (enc); |
277 | } |
301 | } |
278 | |
302 | |
279 | static void encode_sv (enc_t *enc, SV *sv); |
303 | static void encode_sv (enc_t *enc, SV *sv); |
280 | |
304 | |
… | |
… | |
289 | encode_ch (enc, '['); encode_nl (enc); |
313 | encode_ch (enc, '['); encode_nl (enc); |
290 | ++enc->indent; |
314 | ++enc->indent; |
291 | |
315 | |
292 | for (i = 0; i <= len; ++i) |
316 | for (i = 0; i <= len; ++i) |
293 | { |
317 | { |
|
|
318 | SV **svp = av_fetch (av, i, 0); |
|
|
319 | |
294 | encode_indent (enc); |
320 | encode_indent (enc); |
295 | encode_sv (enc, *av_fetch (av, i, 0)); |
321 | |
|
|
322 | if (svp) |
|
|
323 | encode_sv (enc, *svp); |
|
|
324 | else |
|
|
325 | encode_str (enc, "null", 4, 0); |
296 | |
326 | |
297 | if (i < len) |
327 | if (i < len) |
298 | encode_comma (enc); |
328 | encode_comma (enc); |
299 | } |
329 | } |
300 | |
330 | |
… | |
… | |
323 | else |
353 | else |
324 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
354 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
325 | |
355 | |
326 | encode_ch (enc, '"'); |
356 | encode_ch (enc, '"'); |
327 | |
357 | |
328 | if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
358 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
329 | encode_ch (enc, ':'); |
359 | encode_ch (enc, ':'); |
330 | if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
360 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
331 | encode_sv (enc, HeVAL (he)); |
361 | encode_sv (enc, HeVAL (he)); |
332 | } |
362 | } |
333 | |
363 | |
334 | // compare hash entries, used when all keys are bytestrings |
364 | // compare hash entries, used when all keys are bytestrings |
335 | static int |
365 | static int |
… | |
… | |
370 | { |
400 | { |
371 | // for canonical output we have to sort by keys first |
401 | // for canonical output we have to sort by keys first |
372 | // actually, this is mostly due to the stupid so-called |
402 | // actually, this is mostly due to the stupid so-called |
373 | // security workaround added somewhere in 5.8.x. |
403 | // security workaround added somewhere in 5.8.x. |
374 | // that randomises hash orderings |
404 | // that randomises hash orderings |
375 | if (enc->flags & F_CANONICAL) |
405 | if (enc->json.flags & F_CANONICAL) |
376 | { |
406 | { |
377 | HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
378 | int fast = 1; |
407 | int fast = 1; |
|
|
408 | HE *he; |
|
|
409 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
410 | HE **hes = _alloca (count * sizeof (HE)); |
|
|
411 | #else |
|
|
412 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
413 | #endif |
379 | |
414 | |
380 | i = 0; |
415 | i = 0; |
381 | while ((he = hv_iternext (hv))) |
416 | while ((he = hv_iternext (hv))) |
382 | { |
417 | { |
383 | hes [i++] = he; |
418 | hes [i++] = he; |
… | |
… | |
447 | svtype svt; |
482 | svtype svt; |
448 | |
483 | |
449 | SvGETMAGIC (sv); |
484 | SvGETMAGIC (sv); |
450 | svt = SvTYPE (sv); |
485 | svt = SvTYPE (sv); |
451 | |
486 | |
|
|
487 | if (expect_false (SvOBJECT (sv))) |
|
|
488 | { |
|
|
489 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
490 | ? json_boolean_stash |
|
|
491 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
492 | |
|
|
493 | if (SvSTASH (sv) == stash) |
|
|
494 | { |
|
|
495 | if (SvIV (sv)) |
|
|
496 | encode_str (enc, "true", 4, 0); |
|
|
497 | else |
|
|
498 | encode_str (enc, "false", 5, 0); |
|
|
499 | } |
|
|
500 | else |
|
|
501 | { |
|
|
502 | #if 0 |
|
|
503 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
|
|
504 | { |
|
|
505 | // not yet |
|
|
506 | } |
|
|
507 | #endif |
|
|
508 | if (enc->json.flags & F_CONV_BLESSED) |
|
|
509 | { |
|
|
510 | // we re-bless the reference to get overload and other niceties right |
|
|
511 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
|
|
512 | |
|
|
513 | if (to_json) |
|
|
514 | { |
|
|
515 | int count; |
|
|
516 | dSP; |
|
|
517 | |
|
|
518 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
519 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
520 | |
|
|
521 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
522 | PUTBACK; |
|
|
523 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
|
|
524 | SPAGAIN; |
|
|
525 | |
|
|
526 | // catch this surprisingly common error |
|
|
527 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
528 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
529 | |
|
|
530 | sv = POPs; |
|
|
531 | PUTBACK; |
|
|
532 | |
|
|
533 | encode_sv (enc, sv); |
|
|
534 | |
|
|
535 | FREETMPS; LEAVE; |
|
|
536 | } |
|
|
537 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
538 | encode_str (enc, "null", 4, 0); |
|
|
539 | else |
|
|
540 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
541 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
542 | } |
|
|
543 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
544 | encode_str (enc, "null", 4, 0); |
|
|
545 | else |
|
|
546 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
|
|
547 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
548 | } |
|
|
549 | } |
452 | if (svt == SVt_PVHV) |
550 | else if (svt == SVt_PVHV) |
453 | encode_hv (enc, (HV *)sv); |
551 | encode_hv (enc, (HV *)sv); |
454 | else if (svt == SVt_PVAV) |
552 | else if (svt == SVt_PVAV) |
455 | encode_av (enc, (AV *)sv); |
553 | encode_av (enc, (AV *)sv); |
456 | else if (svt < SVt_PVAV) |
554 | else if (svt < SVt_PVAV) |
457 | { |
555 | { |
458 | if (SvNIOK (sv) && SvIV (sv) == 0) |
556 | STRLEN len = 0; |
|
|
557 | char *pv = svt ? SvPV (sv, len) : 0; |
|
|
558 | |
|
|
559 | if (len == 1 && *pv == '1') |
|
|
560 | encode_str (enc, "true", 4, 0); |
|
|
561 | else if (len == 1 && *pv == '0') |
459 | encode_str (enc, "false", 5, 0); |
562 | encode_str (enc, "false", 5, 0); |
460 | else if (SvNIOK (sv) && SvIV (sv) == 1) |
|
|
461 | encode_str (enc, "true", 4, 0); |
|
|
462 | else |
563 | else |
463 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
564 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
464 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
565 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
465 | } |
566 | } |
466 | else |
567 | else |
… | |
… | |
481 | encode_str (enc, str, len, SvUTF8 (sv)); |
582 | encode_str (enc, str, len, SvUTF8 (sv)); |
482 | encode_ch (enc, '"'); |
583 | encode_ch (enc, '"'); |
483 | } |
584 | } |
484 | else if (SvNOKp (sv)) |
585 | else if (SvNOKp (sv)) |
485 | { |
586 | { |
|
|
587 | // trust that perl will do the right thing w.r.t. JSON syntax. |
486 | need (enc, NV_DIG + 32); |
588 | need (enc, NV_DIG + 32); |
487 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
589 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
488 | enc->cur += strlen (enc->cur); |
590 | enc->cur += strlen (enc->cur); |
489 | } |
591 | } |
490 | else if (SvIOKp (sv)) |
592 | else if (SvIOKp (sv)) |
491 | { |
593 | { |
492 | need (enc, 64); |
594 | // we assume we can always read an IV as a UV |
|
|
595 | if (SvUV (sv) & ~(UV)0x7fff) |
|
|
596 | { |
|
|
597 | // large integer, use the (rather slow) snprintf way. |
|
|
598 | need (enc, sizeof (UV) * 3); |
493 | enc->cur += |
599 | enc->cur += |
494 | SvIsUV(sv) |
600 | SvIsUV(sv) |
495 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
601 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
496 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
602 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
603 | } |
|
|
604 | else |
|
|
605 | { |
|
|
606 | // optimise the "small number case" |
|
|
607 | // code will likely be branchless and use only a single multiplication |
|
|
608 | I32 i = SvIV (sv); |
|
|
609 | U32 u; |
|
|
610 | char digit, nz = 0; |
|
|
611 | |
|
|
612 | need (enc, 6); |
|
|
613 | |
|
|
614 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
615 | u = i < 0 ? -i : i; |
|
|
616 | |
|
|
617 | // convert to 4.28 fixed-point representation |
|
|
618 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
619 | |
|
|
620 | // now output digit by digit, each time masking out the integer part |
|
|
621 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
622 | // resulting in a net multiplication by 10. |
|
|
623 | // we always write the digit to memory but conditionally increment |
|
|
624 | // the pointer, to ease the usage of conditional move instructions. |
|
|
625 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
|
|
626 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
|
|
627 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
|
|
628 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
|
|
629 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
630 | } |
497 | } |
631 | } |
498 | else if (SvROK (sv)) |
632 | else if (SvROK (sv)) |
499 | encode_rv (enc, SvRV (sv)); |
633 | encode_rv (enc, SvRV (sv)); |
500 | else if (!SvOK (sv)) |
634 | else if (!SvOK (sv)) |
501 | encode_str (enc, "null", 4, 0); |
635 | encode_str (enc, "null", 4, 0); |
… | |
… | |
503 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
637 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
504 | SvPV_nolen (sv), SvFLAGS (sv)); |
638 | SvPV_nolen (sv), SvFLAGS (sv)); |
505 | } |
639 | } |
506 | |
640 | |
507 | static SV * |
641 | static SV * |
508 | encode_json (SV *scalar, U32 flags) |
642 | encode_json (SV *scalar, JSON *json) |
509 | { |
643 | { |
510 | enc_t enc; |
644 | enc_t enc; |
511 | |
645 | |
512 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
646 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
513 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
647 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
514 | |
648 | |
515 | enc.flags = flags; |
649 | enc.json = *json; |
516 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
650 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
517 | enc.cur = SvPVX (enc.sv); |
651 | enc.cur = SvPVX (enc.sv); |
518 | enc.end = SvEND (enc.sv); |
652 | enc.end = SvEND (enc.sv); |
519 | enc.indent = 0; |
653 | enc.indent = 0; |
520 | enc.maxdepth = DEC_DEPTH (flags); |
654 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
521 | |
655 | |
522 | SvPOK_only (enc.sv); |
656 | SvPOK_only (enc.sv); |
523 | encode_sv (&enc, scalar); |
657 | encode_sv (&enc, scalar); |
524 | |
658 | |
525 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
659 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
526 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
660 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
527 | |
661 | |
528 | if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
662 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
529 | SvUTF8_on (enc.sv); |
663 | SvUTF8_on (enc.sv); |
530 | |
664 | |
531 | if (enc.flags & F_SHRINK) |
665 | if (enc.json.flags & F_SHRINK) |
532 | shrink (enc.sv); |
666 | shrink (enc.sv); |
533 | |
667 | |
534 | return enc.sv; |
668 | return enc.sv; |
535 | } |
669 | } |
536 | |
670 | |
… | |
… | |
541 | typedef struct |
675 | typedef struct |
542 | { |
676 | { |
543 | char *cur; // current parser pointer |
677 | char *cur; // current parser pointer |
544 | char *end; // end of input string |
678 | char *end; // end of input string |
545 | const char *err; // parse error, if != 0 |
679 | const char *err; // parse error, if != 0 |
546 | U32 flags; // F_* |
680 | JSON json; |
547 | U32 depth; // recursion depth |
681 | U32 depth; // recursion depth |
548 | U32 maxdepth; // recursion depth limit |
682 | U32 maxdepth; // recursion depth limit |
549 | } dec_t; |
683 | } dec_t; |
550 | |
684 | |
551 | static void |
685 | inline void |
552 | decode_ws (dec_t *dec) |
686 | decode_ws (dec_t *dec) |
553 | { |
687 | { |
554 | for (;;) |
688 | for (;;) |
555 | { |
689 | { |
556 | char ch = *dec->cur; |
690 | char ch = *dec->cur; |
… | |
… | |
582 | decode_4hex (dec_t *dec) |
716 | decode_4hex (dec_t *dec) |
583 | { |
717 | { |
584 | signed char d1, d2, d3, d4; |
718 | signed char d1, d2, d3, d4; |
585 | unsigned char *cur = (unsigned char *)dec->cur; |
719 | unsigned char *cur = (unsigned char *)dec->cur; |
586 | |
720 | |
587 | d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); |
721 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
588 | d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); |
722 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
589 | d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); |
723 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
590 | d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); |
724 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
591 | |
725 | |
592 | dec->cur += 4; |
726 | dec->cur += 4; |
593 | |
727 | |
594 | return ((UV)d1) << 12 |
728 | return ((UV)d1) << 12 |
595 | | ((UV)d2) << 8 |
729 | | ((UV)d2) << 8 |
… | |
… | |
603 | static SV * |
737 | static SV * |
604 | decode_str (dec_t *dec) |
738 | decode_str (dec_t *dec) |
605 | { |
739 | { |
606 | SV *sv = 0; |
740 | SV *sv = 0; |
607 | int utf8 = 0; |
741 | int utf8 = 0; |
|
|
742 | char *dec_cur = dec->cur; |
608 | |
743 | |
609 | do |
744 | do |
610 | { |
745 | { |
611 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
746 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
612 | char *cur = buf; |
747 | char *cur = buf; |
613 | |
748 | |
614 | do |
749 | do |
615 | { |
750 | { |
616 | unsigned char ch = *(unsigned char *)dec->cur++; |
751 | unsigned char ch = *(unsigned char *)dec_cur++; |
617 | |
752 | |
618 | if (ch == '"') |
753 | if (expect_false (ch == '"')) |
619 | { |
754 | { |
620 | --dec->cur; |
755 | --dec_cur; |
621 | break; |
756 | break; |
622 | } |
757 | } |
623 | else if (ch == '\\') |
758 | else if (expect_false (ch == '\\')) |
624 | { |
759 | { |
625 | switch (*dec->cur) |
760 | switch (*dec_cur) |
626 | { |
761 | { |
627 | case '\\': |
762 | case '\\': |
628 | case '/': |
763 | case '/': |
629 | case '"': *cur++ = *dec->cur++; break; |
764 | case '"': *cur++ = *dec_cur++; break; |
630 | |
765 | |
631 | case 'b': ++dec->cur; *cur++ = '\010'; break; |
766 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
632 | case 't': ++dec->cur; *cur++ = '\011'; break; |
767 | case 't': ++dec_cur; *cur++ = '\011'; break; |
633 | case 'n': ++dec->cur; *cur++ = '\012'; break; |
768 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
634 | case 'f': ++dec->cur; *cur++ = '\014'; break; |
769 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
635 | case 'r': ++dec->cur; *cur++ = '\015'; break; |
770 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
636 | |
771 | |
637 | case 'u': |
772 | case 'u': |
638 | { |
773 | { |
639 | UV lo, hi; |
774 | UV lo, hi; |
640 | ++dec->cur; |
775 | ++dec_cur; |
641 | |
776 | |
|
|
777 | dec->cur = dec_cur; |
642 | hi = decode_4hex (dec); |
778 | hi = decode_4hex (dec); |
|
|
779 | dec_cur = dec->cur; |
643 | if (hi == (UV)-1) |
780 | if (hi == (UV)-1) |
644 | goto fail; |
781 | goto fail; |
645 | |
782 | |
646 | // possibly a surrogate pair |
783 | // possibly a surrogate pair |
647 | if (hi >= 0xd800) |
784 | if (hi >= 0xd800) |
648 | if (hi < 0xdc00) |
785 | if (hi < 0xdc00) |
649 | { |
786 | { |
650 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
787 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
651 | ERR ("missing low surrogate character in surrogate pair"); |
788 | ERR ("missing low surrogate character in surrogate pair"); |
652 | |
789 | |
653 | dec->cur += 2; |
790 | dec_cur += 2; |
654 | |
791 | |
|
|
792 | dec->cur = dec_cur; |
655 | lo = decode_4hex (dec); |
793 | lo = decode_4hex (dec); |
|
|
794 | dec_cur = dec->cur; |
656 | if (lo == (UV)-1) |
795 | if (lo == (UV)-1) |
657 | goto fail; |
796 | goto fail; |
658 | |
797 | |
659 | if (lo < 0xdc00 || lo >= 0xe000) |
798 | if (lo < 0xdc00 || lo >= 0xe000) |
660 | ERR ("surrogate pair expected"); |
799 | ERR ("surrogate pair expected"); |
… | |
… | |
674 | *cur++ = hi; |
813 | *cur++ = hi; |
675 | } |
814 | } |
676 | break; |
815 | break; |
677 | |
816 | |
678 | default: |
817 | default: |
679 | --dec->cur; |
818 | --dec_cur; |
680 | ERR ("illegal backslash escape sequence in string"); |
819 | ERR ("illegal backslash escape sequence in string"); |
681 | } |
820 | } |
682 | } |
821 | } |
683 | else if (ch >= 0x20 && ch <= 0x7f) |
822 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
684 | *cur++ = ch; |
823 | *cur++ = ch; |
685 | else if (ch >= 0x80) |
824 | else if (ch >= 0x80) |
686 | { |
825 | { |
687 | STRLEN clen; |
826 | STRLEN clen; |
688 | UV uch; |
827 | UV uch; |
689 | |
828 | |
690 | --dec->cur; |
829 | --dec_cur; |
691 | |
830 | |
692 | uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); |
831 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
693 | if (clen == (STRLEN)-1) |
832 | if (clen == (STRLEN)-1) |
694 | ERR ("malformed UTF-8 character in JSON string"); |
833 | ERR ("malformed UTF-8 character in JSON string"); |
695 | |
834 | |
696 | do |
835 | do |
697 | *cur++ = *dec->cur++; |
836 | *cur++ = *dec_cur++; |
698 | while (--clen); |
837 | while (--clen); |
699 | |
838 | |
700 | utf8 = 1; |
839 | utf8 = 1; |
701 | } |
840 | } |
702 | else |
841 | else |
703 | { |
842 | { |
704 | --dec->cur; |
843 | --dec_cur; |
705 | |
844 | |
706 | if (!ch) |
845 | if (!ch) |
707 | ERR ("unexpected end of string while parsing JSON string"); |
846 | ERR ("unexpected end of string while parsing JSON string"); |
708 | else |
847 | else |
709 | ERR ("invalid character encountered while parsing JSON string"); |
848 | ERR ("invalid character encountered while parsing JSON string"); |
… | |
… | |
722 | } |
861 | } |
723 | else |
862 | else |
724 | sv = newSVpvn (buf, len); |
863 | sv = newSVpvn (buf, len); |
725 | } |
864 | } |
726 | } |
865 | } |
727 | while (*dec->cur != '"'); |
866 | while (*dec_cur != '"'); |
728 | |
867 | |
729 | ++dec->cur; |
868 | ++dec_cur; |
730 | |
869 | |
731 | if (sv) |
870 | if (sv) |
732 | { |
871 | { |
733 | SvPOK_only (sv); |
872 | SvPOK_only (sv); |
734 | *SvEND (sv) = 0; |
873 | *SvEND (sv) = 0; |
… | |
… | |
737 | SvUTF8_on (sv); |
876 | SvUTF8_on (sv); |
738 | } |
877 | } |
739 | else |
878 | else |
740 | sv = newSVpvn ("", 0); |
879 | sv = newSVpvn ("", 0); |
741 | |
880 | |
|
|
881 | dec->cur = dec_cur; |
742 | return sv; |
882 | return sv; |
743 | |
883 | |
744 | fail: |
884 | fail: |
|
|
885 | dec->cur = dec_cur; |
745 | return 0; |
886 | return 0; |
746 | } |
887 | } |
747 | |
888 | |
748 | static SV * |
889 | static SV * |
749 | decode_num (dec_t *dec) |
890 | decode_num (dec_t *dec) |
… | |
… | |
807 | is_nv = 1; |
948 | is_nv = 1; |
808 | } |
949 | } |
809 | |
950 | |
810 | if (!is_nv) |
951 | if (!is_nv) |
811 | { |
952 | { |
812 | UV uv; |
953 | int len = dec->cur - start; |
813 | int numtype = grok_number (start, dec->cur - start, &uv); |
954 | |
814 | if (numtype & IS_NUMBER_IN_UV) |
955 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
815 | if (numtype & IS_NUMBER_NEG) |
956 | if (*start == '-') |
|
|
957 | switch (len) |
816 | { |
958 | { |
817 | if (uv < (UV)IV_MIN) |
959 | case 2: return newSViv (-( start [1] - '0' * 1)); |
818 | return newSViv (-(IV)uv); |
960 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
|
|
961 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
962 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
819 | } |
963 | } |
|
|
964 | else |
|
|
965 | switch (len) |
|
|
966 | { |
|
|
967 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
968 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
969 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
970 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
971 | } |
|
|
972 | |
|
|
973 | { |
|
|
974 | UV uv; |
|
|
975 | int numtype = grok_number (start, len, &uv); |
|
|
976 | if (numtype & IS_NUMBER_IN_UV) |
|
|
977 | if (numtype & IS_NUMBER_NEG) |
|
|
978 | { |
|
|
979 | if (uv < (UV)IV_MIN) |
|
|
980 | return newSViv (-(IV)uv); |
|
|
981 | } |
820 | else |
982 | else |
821 | return newSVuv (uv); |
983 | return newSVuv (uv); |
822 | } |
984 | } |
823 | |
985 | |
|
|
986 | len -= *start == '-' ? 1 : 0; |
|
|
987 | |
|
|
988 | // does not fit into IV or UV, try NV |
|
|
989 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
|
|
990 | #if defined (LDBL_DIG) |
|
|
991 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
992 | #endif |
|
|
993 | ) |
|
|
994 | // fits into NV without loss of precision |
|
|
995 | return newSVnv (Atof (start)); |
|
|
996 | |
|
|
997 | // everything else fails, convert it to a string |
|
|
998 | return newSVpvn (start, dec->cur - start); |
|
|
999 | } |
|
|
1000 | |
|
|
1001 | // loss of precision here |
824 | return newSVnv (Atof (start)); |
1002 | return newSVnv (Atof (start)); |
825 | |
1003 | |
826 | fail: |
1004 | fail: |
827 | return 0; |
1005 | return 0; |
828 | } |
1006 | } |
… | |
… | |
872 | } |
1050 | } |
873 | |
1051 | |
874 | static SV * |
1052 | static SV * |
875 | decode_hv (dec_t *dec) |
1053 | decode_hv (dec_t *dec) |
876 | { |
1054 | { |
|
|
1055 | SV *sv; |
877 | HV *hv = newHV (); |
1056 | HV *hv = newHV (); |
878 | |
1057 | |
879 | DEC_INC_DEPTH; |
1058 | DEC_INC_DEPTH; |
880 | decode_ws (dec); |
1059 | decode_ws (dec); |
881 | |
1060 | |
882 | if (*dec->cur == '}') |
1061 | if (*dec->cur == '}') |
883 | ++dec->cur; |
1062 | ++dec->cur; |
884 | else |
1063 | else |
885 | for (;;) |
1064 | for (;;) |
886 | { |
1065 | { |
887 | SV *key, *value; |
|
|
888 | |
|
|
889 | decode_ws (dec); EXPECT_CH ('"'); |
1066 | decode_ws (dec); EXPECT_CH ('"'); |
890 | |
1067 | |
891 | key = decode_str (dec); |
1068 | // heuristic: assume that |
892 | if (!key) |
1069 | // a) decode_str + hv_store_ent are abysmally slow. |
893 | goto fail; |
1070 | // b) most hash keys are short, simple ascii text. |
|
|
1071 | // => try to "fast-match" such strings to avoid |
|
|
1072 | // the overhead of decode_str + hv_store_ent. |
|
|
1073 | { |
|
|
1074 | SV *value; |
|
|
1075 | char *p = dec->cur; |
|
|
1076 | char *e = p + 24; // only try up to 24 bytes |
894 | |
1077 | |
895 | decode_ws (dec); EXPECT_CH (':'); |
1078 | for (;;) |
896 | |
|
|
897 | value = decode_sv (dec); |
|
|
898 | if (!value) |
|
|
899 | { |
1079 | { |
|
|
1080 | // the >= 0x80 is true on most architectures |
|
|
1081 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1082 | { |
|
|
1083 | // slow path, back up and use decode_str |
|
|
1084 | SV *key = decode_str (dec); |
|
|
1085 | if (!key) |
|
|
1086 | goto fail; |
|
|
1087 | |
|
|
1088 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1089 | |
|
|
1090 | value = decode_sv (dec); |
|
|
1091 | if (!value) |
|
|
1092 | { |
|
|
1093 | SvREFCNT_dec (key); |
|
|
1094 | goto fail; |
|
|
1095 | } |
|
|
1096 | |
|
|
1097 | hv_store_ent (hv, key, value, 0); |
900 | SvREFCNT_dec (key); |
1098 | SvREFCNT_dec (key); |
|
|
1099 | |
|
|
1100 | break; |
|
|
1101 | } |
|
|
1102 | else if (*p == '"') |
|
|
1103 | { |
|
|
1104 | // fast path, got a simple key |
|
|
1105 | char *key = dec->cur; |
|
|
1106 | int len = p - key; |
|
|
1107 | dec->cur = p + 1; |
|
|
1108 | |
|
|
1109 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1110 | |
|
|
1111 | value = decode_sv (dec); |
|
|
1112 | if (!value) |
901 | goto fail; |
1113 | goto fail; |
|
|
1114 | |
|
|
1115 | hv_store (hv, key, len, value, 0); |
|
|
1116 | |
|
|
1117 | break; |
|
|
1118 | } |
|
|
1119 | |
|
|
1120 | ++p; |
902 | } |
1121 | } |
903 | |
1122 | } |
904 | hv_store_ent (hv, key, value, 0); |
|
|
905 | SvREFCNT_dec (key); |
|
|
906 | |
1123 | |
907 | decode_ws (dec); |
1124 | decode_ws (dec); |
908 | |
1125 | |
909 | if (*dec->cur == '}') |
1126 | if (*dec->cur == '}') |
910 | { |
1127 | { |
… | |
… | |
917 | |
1134 | |
918 | ++dec->cur; |
1135 | ++dec->cur; |
919 | } |
1136 | } |
920 | |
1137 | |
921 | DEC_DEC_DEPTH; |
1138 | DEC_DEC_DEPTH; |
922 | return newRV_noinc ((SV *)hv); |
1139 | sv = newRV_noinc ((SV *)hv); |
|
|
1140 | |
|
|
1141 | // check filter callbacks |
|
|
1142 | if (dec->json.flags & F_HOOK) |
|
|
1143 | { |
|
|
1144 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1145 | { |
|
|
1146 | HE *cb, *he; |
|
|
1147 | |
|
|
1148 | hv_iterinit (hv); |
|
|
1149 | he = hv_iternext (hv); |
|
|
1150 | hv_iterinit (hv); |
|
|
1151 | |
|
|
1152 | // the next line creates a mortal sv each time its called. |
|
|
1153 | // might want to optimise this for common cases. |
|
|
1154 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1155 | |
|
|
1156 | if (cb) |
|
|
1157 | { |
|
|
1158 | dSP; |
|
|
1159 | int count; |
|
|
1160 | |
|
|
1161 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1162 | XPUSHs (HeVAL (he)); |
|
|
1163 | |
|
|
1164 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1165 | |
|
|
1166 | if (count == 1) |
|
|
1167 | { |
|
|
1168 | sv = newSVsv (POPs); |
|
|
1169 | FREETMPS; LEAVE; |
|
|
1170 | return sv; |
|
|
1171 | } |
|
|
1172 | |
|
|
1173 | FREETMPS; LEAVE; |
|
|
1174 | } |
|
|
1175 | } |
|
|
1176 | |
|
|
1177 | if (dec->json.cb_object) |
|
|
1178 | { |
|
|
1179 | dSP; |
|
|
1180 | int count; |
|
|
1181 | |
|
|
1182 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1183 | XPUSHs (sv_2mortal (sv)); |
|
|
1184 | |
|
|
1185 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1186 | |
|
|
1187 | if (count == 1) |
|
|
1188 | { |
|
|
1189 | sv = newSVsv (POPs); |
|
|
1190 | FREETMPS; LEAVE; |
|
|
1191 | return sv; |
|
|
1192 | } |
|
|
1193 | |
|
|
1194 | SvREFCNT_inc (sv); |
|
|
1195 | FREETMPS; LEAVE; |
|
|
1196 | } |
|
|
1197 | } |
|
|
1198 | |
|
|
1199 | return sv; |
923 | |
1200 | |
924 | fail: |
1201 | fail: |
925 | SvREFCNT_dec (hv); |
1202 | SvREFCNT_dec (hv); |
926 | DEC_DEC_DEPTH; |
1203 | DEC_DEC_DEPTH; |
927 | return 0; |
1204 | return 0; |
… | |
… | |
929 | |
1206 | |
930 | static SV * |
1207 | static SV * |
931 | decode_sv (dec_t *dec) |
1208 | decode_sv (dec_t *dec) |
932 | { |
1209 | { |
933 | decode_ws (dec); |
1210 | decode_ws (dec); |
|
|
1211 | |
|
|
1212 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1213 | // to parse anything. |
934 | switch (*dec->cur) |
1214 | switch (*dec->cur) |
935 | { |
1215 | { |
936 | case '"': ++dec->cur; return decode_str (dec); |
1216 | case '"': ++dec->cur; return decode_str (dec); |
937 | case '[': ++dec->cur; return decode_av (dec); |
1217 | case '[': ++dec->cur; return decode_av (dec); |
938 | case '{': ++dec->cur; return decode_hv (dec); |
1218 | case '{': ++dec->cur; return decode_hv (dec); |
… | |
… | |
944 | |
1224 | |
945 | case 't': |
1225 | case 't': |
946 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1226 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
947 | { |
1227 | { |
948 | dec->cur += 4; |
1228 | dec->cur += 4; |
949 | return newSViv (1); |
1229 | return SvREFCNT_inc (json_true); |
950 | } |
1230 | } |
951 | else |
1231 | else |
952 | ERR ("'true' expected"); |
1232 | ERR ("'true' expected"); |
953 | |
1233 | |
954 | break; |
1234 | break; |
955 | |
1235 | |
956 | case 'f': |
1236 | case 'f': |
957 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1237 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
958 | { |
1238 | { |
959 | dec->cur += 5; |
1239 | dec->cur += 5; |
960 | return newSViv (0); |
1240 | return SvREFCNT_inc (json_false); |
961 | } |
1241 | } |
962 | else |
1242 | else |
963 | ERR ("'false' expected"); |
1243 | ERR ("'false' expected"); |
964 | |
1244 | |
965 | break; |
1245 | break; |
… | |
… | |
983 | fail: |
1263 | fail: |
984 | return 0; |
1264 | return 0; |
985 | } |
1265 | } |
986 | |
1266 | |
987 | static SV * |
1267 | static SV * |
988 | decode_json (SV *string, U32 flags, UV *offset_return) |
1268 | decode_json (SV *string, JSON *json, UV *offset_return) |
989 | { |
1269 | { |
990 | dec_t dec; |
1270 | dec_t dec; |
991 | UV offset; |
1271 | UV offset; |
992 | SV *sv; |
1272 | SV *sv; |
993 | |
1273 | |
994 | SvGETMAGIC (string); |
1274 | SvGETMAGIC (string); |
995 | SvUPGRADE (string, SVt_PV); |
1275 | SvUPGRADE (string, SVt_PV); |
996 | |
1276 | |
|
|
1277 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
|
|
1278 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1279 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
|
|
1280 | |
997 | if (flags & F_UTF8) |
1281 | if (json->flags & F_UTF8) |
998 | sv_utf8_downgrade (string, 0); |
1282 | sv_utf8_downgrade (string, 0); |
999 | else |
1283 | else |
1000 | sv_utf8_upgrade (string); |
1284 | sv_utf8_upgrade (string); |
1001 | |
1285 | |
1002 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1286 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1003 | |
1287 | |
1004 | dec.flags = flags; |
1288 | dec.json = *json; |
1005 | dec.cur = SvPVX (string); |
1289 | dec.cur = SvPVX (string); |
1006 | dec.end = SvEND (string); |
1290 | dec.end = SvEND (string); |
1007 | dec.err = 0; |
1291 | dec.err = 0; |
1008 | dec.depth = 0; |
1292 | dec.depth = 0; |
1009 | dec.maxdepth = DEC_DEPTH (dec.flags); |
1293 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1294 | |
|
|
1295 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1296 | dec.json.flags |= F_HOOK; |
1010 | |
1297 | |
1011 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1298 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1012 | sv = decode_sv (&dec); |
1299 | sv = decode_sv (&dec); |
1013 | |
1300 | |
1014 | if (!(offset_return || !sv)) |
1301 | if (!(offset_return || !sv)) |
… | |
… | |
1024 | } |
1311 | } |
1025 | } |
1312 | } |
1026 | |
1313 | |
1027 | if (offset_return || !sv) |
1314 | if (offset_return || !sv) |
1028 | { |
1315 | { |
1029 | offset = dec.flags & F_UTF8 |
1316 | offset = dec.json.flags & F_UTF8 |
1030 | ? dec.cur - SvPVX (string) |
1317 | ? dec.cur - SvPVX (string) |
1031 | : utf8_distance (dec.cur, SvPVX (string)); |
1318 | : utf8_distance (dec.cur, SvPVX (string)); |
1032 | |
1319 | |
1033 | if (offset_return) |
1320 | if (offset_return) |
1034 | *offset_return = offset; |
1321 | *offset_return = offset; |
… | |
… | |
1053 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1340 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1054 | } |
1341 | } |
1055 | |
1342 | |
1056 | sv = sv_2mortal (sv); |
1343 | sv = sv_2mortal (sv); |
1057 | |
1344 | |
1058 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1345 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1059 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1346 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1060 | |
1347 | |
1061 | return sv; |
1348 | return sv; |
1062 | } |
1349 | } |
1063 | |
1350 | |
… | |
… | |
1067 | MODULE = JSON::XS PACKAGE = JSON::XS |
1354 | MODULE = JSON::XS PACKAGE = JSON::XS |
1068 | |
1355 | |
1069 | BOOT: |
1356 | BOOT: |
1070 | { |
1357 | { |
1071 | int i; |
1358 | int i; |
1072 | |
|
|
1073 | memset (decode_hexdigit, 0xff, 256); |
|
|
1074 | |
1359 | |
1075 | for (i = 0; i < 256; ++i) |
1360 | for (i = 0; i < 256; ++i) |
1076 | decode_hexdigit [i] = |
1361 | decode_hexdigit [i] = |
1077 | i >= '0' && i <= '9' ? i - '0' |
1362 | i >= '0' && i <= '9' ? i - '0' |
1078 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1363 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1079 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1364 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1080 | : -1; |
1365 | : -1; |
1081 | |
1366 | |
1082 | json_stash = gv_stashpv ("JSON::XS", 1); |
1367 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1368 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
1369 | |
|
|
1370 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
|
|
1371 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1083 | } |
1372 | } |
1084 | |
1373 | |
1085 | PROTOTYPES: DISABLE |
1374 | PROTOTYPES: DISABLE |
1086 | |
1375 | |
1087 | SV *new (char *dummy) |
1376 | void CLONE (...) |
1088 | CODE: |
1377 | CODE: |
1089 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1378 | json_stash = 0; |
1090 | OUTPUT: |
1379 | json_boolean_stash = 0; |
1091 | RETVAL |
|
|
1092 | |
1380 | |
|
|
1381 | void new (char *klass) |
|
|
1382 | PPCODE: |
|
|
1383 | { |
|
|
1384 | HV *stash = !JSON_SLOW || json_stash |
|
|
1385 | ? json_stash |
|
|
1386 | : gv_stashpv ("JSON::XS", 1); |
|
|
1387 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1388 | SvPOK_only (pv); |
|
|
1389 | Zero (SvPVX (pv), 1, JSON); |
|
|
1390 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1391 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash))); |
|
|
1392 | } |
|
|
1393 | |
1093 | SV *ascii (SV *self, int enable = 1) |
1394 | void ascii (JSON *self, int enable = 1) |
1094 | ALIAS: |
1395 | ALIAS: |
1095 | ascii = F_ASCII |
1396 | ascii = F_ASCII |
1096 | latin1 = F_LATIN1 |
1397 | latin1 = F_LATIN1 |
1097 | utf8 = F_UTF8 |
1398 | utf8 = F_UTF8 |
1098 | indent = F_INDENT |
1399 | indent = F_INDENT |
1099 | canonical = F_CANONICAL |
1400 | canonical = F_CANONICAL |
1100 | space_before = F_SPACE_BEFORE |
1401 | space_before = F_SPACE_BEFORE |
1101 | space_after = F_SPACE_AFTER |
1402 | space_after = F_SPACE_AFTER |
1102 | pretty = F_PRETTY |
1403 | pretty = F_PRETTY |
1103 | allow_nonref = F_ALLOW_NONREF |
1404 | allow_nonref = F_ALLOW_NONREF |
1104 | shrink = F_SHRINK |
1405 | shrink = F_SHRINK |
|
|
1406 | allow_blessed = F_ALLOW_BLESSED |
|
|
1407 | convert_blessed = F_CONV_BLESSED |
1105 | CODE: |
1408 | PPCODE: |
1106 | { |
1409 | { |
1107 | UV *uv = SvJSON (self); |
|
|
1108 | if (enable) |
1410 | if (enable) |
1109 | *uv |= ix; |
1411 | self->flags |= ix; |
1110 | else |
1412 | else |
1111 | *uv &= ~ix; |
1413 | self->flags &= ~ix; |
1112 | |
1414 | |
1113 | RETVAL = newSVsv (self); |
1415 | XPUSHs (ST (0)); |
1114 | } |
1416 | } |
1115 | OUTPUT: |
|
|
1116 | RETVAL |
|
|
1117 | |
1417 | |
1118 | SV *max_depth (SV *self, UV max_depth = 0x80000000UL) |
1418 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1119 | CODE: |
1419 | PPCODE: |
1120 | { |
1420 | { |
1121 | UV *uv = SvJSON (self); |
|
|
1122 | UV log2 = 0; |
1421 | UV log2 = 0; |
1123 | |
1422 | |
1124 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1423 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1125 | |
1424 | |
1126 | while ((1UL << log2) < max_depth) |
1425 | while ((1UL << log2) < max_depth) |
1127 | ++log2; |
1426 | ++log2; |
1128 | |
1427 | |
1129 | *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1428 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1130 | |
1429 | |
1131 | RETVAL = newSVsv (self); |
1430 | XPUSHs (ST (0)); |
1132 | } |
1431 | } |
1133 | OUTPUT: |
|
|
1134 | RETVAL |
|
|
1135 | |
1432 | |
1136 | void encode (SV *self, SV *scalar) |
1433 | void max_size (JSON *self, UV max_size = 0) |
1137 | PPCODE: |
1434 | PPCODE: |
1138 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
1435 | { |
|
|
1436 | UV log2 = 0; |
1139 | |
1437 | |
1140 | void decode (SV *self, SV *jsonstr) |
1438 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1439 | if (max_size == 1) max_size = 2; |
|
|
1440 | |
|
|
1441 | while ((1UL << log2) < max_size) |
|
|
1442 | ++log2; |
|
|
1443 | |
|
|
1444 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1445 | |
|
|
1446 | XPUSHs (ST (0)); |
|
|
1447 | } |
|
|
1448 | |
|
|
1449 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1141 | PPCODE: |
1450 | PPCODE: |
|
|
1451 | { |
|
|
1452 | SvREFCNT_dec (self->cb_object); |
|
|
1453 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
|
|
1454 | |
|
|
1455 | XPUSHs (ST (0)); |
|
|
1456 | } |
|
|
1457 | |
|
|
1458 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
|
|
1459 | PPCODE: |
|
|
1460 | { |
|
|
1461 | if (!self->cb_sk_object) |
|
|
1462 | self->cb_sk_object = newHV (); |
|
|
1463 | |
|
|
1464 | if (SvOK (cb)) |
|
|
1465 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
1466 | else |
|
|
1467 | { |
|
|
1468 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
1469 | |
|
|
1470 | if (!HvKEYS (self->cb_sk_object)) |
|
|
1471 | { |
|
|
1472 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1473 | self->cb_sk_object = 0; |
|
|
1474 | } |
|
|
1475 | } |
|
|
1476 | |
|
|
1477 | XPUSHs (ST (0)); |
|
|
1478 | } |
|
|
1479 | |
|
|
1480 | void encode (JSON *self, SV *scalar) |
|
|
1481 | PPCODE: |
|
|
1482 | XPUSHs (encode_json (scalar, self)); |
|
|
1483 | |
|
|
1484 | void decode (JSON *self, SV *jsonstr) |
|
|
1485 | PPCODE: |
1142 | XPUSHs (decode_json (jsonstr, *SvJSON (self), 0)); |
1486 | XPUSHs (decode_json (jsonstr, self, 0)); |
1143 | |
1487 | |
1144 | void decode_prefix (SV *self, SV *jsonstr) |
1488 | void decode_prefix (JSON *self, SV *jsonstr) |
1145 | PPCODE: |
1489 | PPCODE: |
1146 | { |
1490 | { |
1147 | UV offset; |
1491 | UV offset; |
1148 | EXTEND (SP, 2); |
1492 | EXTEND (SP, 2); |
1149 | PUSHs (decode_json (jsonstr, *SvJSON (self), &offset)); |
1493 | PUSHs (decode_json (jsonstr, self, &offset)); |
1150 | PUSHs (sv_2mortal (newSVuv (offset))); |
1494 | PUSHs (sv_2mortal (newSVuv (offset))); |
1151 | } |
1495 | } |
1152 | |
1496 | |
|
|
1497 | void DESTROY (JSON *self) |
|
|
1498 | CODE: |
|
|
1499 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1500 | SvREFCNT_dec (self->cb_object); |
|
|
1501 | |
1153 | PROTOTYPES: ENABLE |
1502 | PROTOTYPES: ENABLE |
1154 | |
1503 | |
1155 | void to_json (SV *scalar) |
1504 | void to_json (SV *scalar) |
1156 | ALIAS: |
|
|
1157 | objToJson = 0 |
|
|
1158 | PPCODE: |
1505 | PPCODE: |
|
|
1506 | { |
|
|
1507 | JSON json = { F_DEFAULT | F_UTF8 }; |
1159 | XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); |
1508 | XPUSHs (encode_json (scalar, &json)); |
|
|
1509 | } |
1160 | |
1510 | |
1161 | void from_json (SV *jsonstr) |
1511 | void from_json (SV *jsonstr) |
1162 | ALIAS: |
|
|
1163 | jsonToObj = 0 |
|
|
1164 | PPCODE: |
1512 | PPCODE: |
|
|
1513 | { |
|
|
1514 | JSON json = { F_DEFAULT | F_UTF8 }; |
1165 | XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0)); |
1515 | XPUSHs (decode_json (jsonstr, &json, 0)); |
|
|
1516 | } |
1166 | |
1517 | |