… | |
… | |
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 | |
9 | |
|
|
10 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
11 | # define snprintf _snprintf // C compilers have this in stdio.h |
|
|
12 | #endif |
|
|
13 | |
|
|
14 | // some old perls do not have this, try to make it work, no |
|
|
15 | // guarentees, though. if it breaks, you get to keep the pieces. |
|
|
16 | #ifndef UTF8_MAXBYTES |
|
|
17 | # define UTF8_MAXBYTES 13 |
|
|
18 | #endif |
|
|
19 | |
9 | #define F_ASCII 0x00000001UL |
20 | #define F_ASCII 0x00000001UL |
|
|
21 | #define F_LATIN1 0x00000002UL |
10 | #define F_UTF8 0x00000002UL |
22 | #define F_UTF8 0x00000004UL |
11 | #define F_INDENT 0x00000004UL |
23 | #define F_INDENT 0x00000008UL |
12 | #define F_CANONICAL 0x00000008UL |
24 | #define F_CANONICAL 0x00000010UL |
13 | #define F_SPACE_BEFORE 0x00000010UL |
25 | #define F_SPACE_BEFORE 0x00000020UL |
14 | #define F_SPACE_AFTER 0x00000020UL |
26 | #define F_SPACE_AFTER 0x00000040UL |
15 | #define F_ALLOW_NONREF 0x00000080UL |
27 | #define F_ALLOW_NONREF 0x00000100UL |
16 | #define F_SHRINK 0x00000100UL |
28 | #define F_SHRINK 0x00000200UL |
|
|
29 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
30 | #define F_CONV_BLESSED 0x00000800UL |
17 | #define F_MAXDEPTH 0xf8000000UL |
31 | #define F_MAXDEPTH 0xf8000000UL |
18 | #define S_MAXDEPTH 27 |
32 | #define S_MAXDEPTH 27 |
|
|
33 | #define F_MAXSIZE 0x01f00000UL |
|
|
34 | #define S_MAXSIZE 20 |
|
|
35 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
19 | |
36 | |
20 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
37 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
21 | |
38 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
22 | // F_SELFCONVERT? <=> to_json/toJson |
|
|
23 | // F_BLESSED? <=> { $__class__$ => } |
|
|
24 | |
39 | |
25 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
26 | #define F_DEFAULT (12UL << S_MAXDEPTH) |
41 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
27 | |
42 | |
28 | #define INIT_SIZE 32 // initial scalar size to be allocated |
43 | #define INIT_SIZE 32 // initial scalar size to be allocated |
29 | #define INDENT_STEP 3 // spaces per indentation level |
44 | #define INDENT_STEP 3 // spaces per indentation level |
30 | |
45 | |
31 | #define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character |
|
|
32 | #define SHORT_STRING_LEN 512 // special-case strings of up to this size |
46 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
33 | |
47 | |
34 | #define SB do { |
48 | #define SB do { |
35 | #define SE } while (0) |
49 | #define SE } while (0) |
36 | |
50 | |
|
|
51 | #if __GNUC__ >= 3 |
|
|
52 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
53 | # define inline inline |
|
|
54 | #else |
|
|
55 | # define expect(expr,value) (expr) |
|
|
56 | # define inline static |
|
|
57 | #endif |
|
|
58 | |
|
|
59 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
60 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
61 | |
37 | static HV *json_stash; // JSON::XS:: |
62 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
|
|
63 | static SV *json_true, *json_false; |
|
|
64 | |
|
|
65 | typedef struct { |
|
|
66 | U32 flags; |
|
|
67 | SV *cb_object; |
|
|
68 | HV *cb_sk_object; |
|
|
69 | } JSON; |
38 | |
70 | |
39 | ///////////////////////////////////////////////////////////////////////////// |
71 | ///////////////////////////////////////////////////////////////////////////// |
40 | // utility functions |
72 | // utility functions |
41 | |
73 | |
42 | static UV * |
74 | inline void |
43 | SvJSON (SV *sv) |
|
|
44 | { |
|
|
45 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
|
|
46 | croak ("object is not of type JSON::XS"); |
|
|
47 | |
|
|
48 | return &SvUVX (SvRV (sv)); |
|
|
49 | } |
|
|
50 | |
|
|
51 | static void |
|
|
52 | shrink (SV *sv) |
75 | shrink (SV *sv) |
53 | { |
76 | { |
54 | sv_utf8_downgrade (sv, 1); |
77 | sv_utf8_downgrade (sv, 1); |
55 | if (SvLEN (sv) > SvCUR (sv) + 1) |
78 | if (SvLEN (sv) > SvCUR (sv) + 1) |
56 | { |
79 | { |
… | |
… | |
65 | // decode an utf-8 character and return it, or (UV)-1 in |
88 | // decode an utf-8 character and return it, or (UV)-1 in |
66 | // case of an error. |
89 | // case of an error. |
67 | // we special-case "safe" characters from U+80 .. U+7FF, |
90 | // we special-case "safe" characters from U+80 .. U+7FF, |
68 | // but use the very good perl function to parse anything else. |
91 | // but use the very good perl function to parse anything else. |
69 | // note that we never call this function for a ascii codepoints |
92 | // note that we never call this function for a ascii codepoints |
70 | static UV |
93 | inline UV |
71 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
94 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
72 | { |
95 | { |
73 | if (s[0] > 0xdf || s[0] < 0xc2) |
96 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
74 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
97 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
75 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
98 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
76 | { |
99 | { |
77 | *clen = 2; |
100 | *clen = 2; |
78 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
101 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
… | |
… | |
91 | typedef struct |
114 | typedef struct |
92 | { |
115 | { |
93 | char *cur; // SvPVX (sv) + current output position |
116 | char *cur; // SvPVX (sv) + current output position |
94 | char *end; // SvEND (sv) |
117 | char *end; // SvEND (sv) |
95 | SV *sv; // result scalar |
118 | SV *sv; // result scalar |
96 | U32 flags; // F_* |
119 | JSON json; |
97 | U32 indent; // indentation level |
120 | U32 indent; // indentation level |
98 | U32 maxdepth; // max. indentation/recursion level |
121 | U32 maxdepth; // max. indentation/recursion level |
99 | } enc_t; |
122 | } enc_t; |
100 | |
123 | |
101 | static void |
124 | inline void |
102 | need (enc_t *enc, STRLEN len) |
125 | need (enc_t *enc, STRLEN len) |
103 | { |
126 | { |
104 | if (enc->cur + len >= enc->end) |
127 | if (expect_false (enc->cur + len >= enc->end)) |
105 | { |
128 | { |
106 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
129 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
107 | SvGROW (enc->sv, cur + len + 1); |
130 | SvGROW (enc->sv, cur + len + 1); |
108 | enc->cur = SvPVX (enc->sv) + cur; |
131 | enc->cur = SvPVX (enc->sv) + cur; |
109 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); |
132 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
110 | } |
133 | } |
111 | } |
134 | } |
112 | |
135 | |
113 | static void |
136 | inline void |
114 | encode_ch (enc_t *enc, char ch) |
137 | encode_ch (enc_t *enc, char ch) |
115 | { |
138 | { |
116 | need (enc, 1); |
139 | need (enc, 1); |
117 | *enc->cur++ = ch; |
140 | *enc->cur++ = ch; |
118 | } |
141 | } |
… | |
… | |
126 | |
149 | |
127 | while (str < end) |
150 | while (str < end) |
128 | { |
151 | { |
129 | unsigned char ch = *(unsigned char *)str; |
152 | unsigned char ch = *(unsigned char *)str; |
130 | |
153 | |
131 | if (ch >= 0x20 && ch < 0x80) // most common case |
154 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
132 | { |
155 | { |
133 | if (ch == '"') // but with slow exceptions |
156 | if (expect_false (ch == '"')) // but with slow exceptions |
134 | { |
157 | { |
135 | need (enc, len += 1); |
158 | need (enc, len += 1); |
136 | *enc->cur++ = '\\'; |
159 | *enc->cur++ = '\\'; |
137 | *enc->cur++ = '"'; |
160 | *enc->cur++ = '"'; |
138 | } |
161 | } |
139 | else if (ch == '\\') |
162 | else if (expect_false (ch == '\\')) |
140 | { |
163 | { |
141 | need (enc, len += 1); |
164 | need (enc, len += 1); |
142 | *enc->cur++ = '\\'; |
165 | *enc->cur++ = '\\'; |
143 | *enc->cur++ = '\\'; |
166 | *enc->cur++ = '\\'; |
144 | } |
167 | } |
… | |
… | |
162 | STRLEN clen; |
185 | STRLEN clen; |
163 | UV uch; |
186 | UV uch; |
164 | |
187 | |
165 | if (is_utf8) |
188 | if (is_utf8) |
166 | { |
189 | { |
167 | //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
|
|
168 | uch = decode_utf8 (str, end - str, &clen); |
190 | uch = decode_utf8 (str, end - str, &clen); |
169 | if (clen == (STRLEN)-1) |
191 | if (clen == (STRLEN)-1) |
170 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
192 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
171 | } |
193 | } |
172 | else |
194 | else |
… | |
… | |
176 | } |
198 | } |
177 | |
199 | |
178 | if (uch > 0x10FFFFUL) |
200 | if (uch > 0x10FFFFUL) |
179 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
201 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
180 | |
202 | |
181 | if (uch < 0x80 || enc->flags & F_ASCII) |
203 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
182 | { |
204 | { |
183 | if (uch > 0xFFFFUL) |
205 | if (uch > 0xFFFFUL) |
184 | { |
206 | { |
185 | need (enc, len += 11); |
207 | need (enc, len += 11); |
186 | sprintf (enc->cur, "\\u%04x\\u%04x", |
208 | sprintf (enc->cur, "\\u%04x\\u%04x", |
… | |
… | |
200 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
222 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
201 | } |
223 | } |
202 | |
224 | |
203 | str += clen; |
225 | str += clen; |
204 | } |
226 | } |
|
|
227 | else if (enc->json.flags & F_LATIN1) |
|
|
228 | { |
|
|
229 | *enc->cur++ = uch; |
|
|
230 | str += clen; |
|
|
231 | } |
205 | else if (is_utf8) |
232 | else if (is_utf8) |
206 | { |
233 | { |
207 | need (enc, len += clen); |
234 | need (enc, len += clen); |
208 | do |
235 | do |
209 | { |
236 | { |
… | |
… | |
211 | } |
238 | } |
212 | while (--clen); |
239 | while (--clen); |
213 | } |
240 | } |
214 | else |
241 | else |
215 | { |
242 | { |
216 | need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed |
243 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
217 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
244 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
218 | ++str; |
245 | ++str; |
219 | } |
246 | } |
220 | } |
247 | } |
221 | } |
248 | } |
… | |
… | |
223 | |
250 | |
224 | --len; |
251 | --len; |
225 | } |
252 | } |
226 | } |
253 | } |
227 | |
254 | |
228 | static void |
255 | inline void |
229 | encode_indent (enc_t *enc) |
256 | encode_indent (enc_t *enc) |
230 | { |
257 | { |
231 | if (enc->flags & F_INDENT) |
258 | if (enc->json.flags & F_INDENT) |
232 | { |
259 | { |
233 | int spaces = enc->indent * INDENT_STEP; |
260 | int spaces = enc->indent * INDENT_STEP; |
234 | |
261 | |
235 | need (enc, spaces); |
262 | need (enc, spaces); |
236 | memset (enc->cur, ' ', spaces); |
263 | memset (enc->cur, ' ', spaces); |
237 | enc->cur += spaces; |
264 | enc->cur += spaces; |
238 | } |
265 | } |
239 | } |
266 | } |
240 | |
267 | |
241 | static void |
268 | inline void |
242 | encode_space (enc_t *enc) |
269 | encode_space (enc_t *enc) |
243 | { |
270 | { |
244 | need (enc, 1); |
271 | need (enc, 1); |
245 | encode_ch (enc, ' '); |
272 | encode_ch (enc, ' '); |
246 | } |
273 | } |
247 | |
274 | |
248 | static void |
275 | inline void |
249 | encode_nl (enc_t *enc) |
276 | encode_nl (enc_t *enc) |
250 | { |
277 | { |
251 | if (enc->flags & F_INDENT) |
278 | if (enc->json.flags & F_INDENT) |
252 | { |
279 | { |
253 | need (enc, 1); |
280 | need (enc, 1); |
254 | encode_ch (enc, '\n'); |
281 | encode_ch (enc, '\n'); |
255 | } |
282 | } |
256 | } |
283 | } |
257 | |
284 | |
258 | static void |
285 | inline void |
259 | encode_comma (enc_t *enc) |
286 | encode_comma (enc_t *enc) |
260 | { |
287 | { |
261 | encode_ch (enc, ','); |
288 | encode_ch (enc, ','); |
262 | |
289 | |
263 | if (enc->flags & F_INDENT) |
290 | if (enc->json.flags & F_INDENT) |
264 | encode_nl (enc); |
291 | encode_nl (enc); |
265 | else if (enc->flags & F_SPACE_AFTER) |
292 | else if (enc->json.flags & F_SPACE_AFTER) |
266 | encode_space (enc); |
293 | encode_space (enc); |
267 | } |
294 | } |
268 | |
295 | |
269 | static void encode_sv (enc_t *enc, SV *sv); |
296 | static void encode_sv (enc_t *enc, SV *sv); |
270 | |
297 | |
… | |
… | |
279 | encode_ch (enc, '['); encode_nl (enc); |
306 | encode_ch (enc, '['); encode_nl (enc); |
280 | ++enc->indent; |
307 | ++enc->indent; |
281 | |
308 | |
282 | for (i = 0; i <= len; ++i) |
309 | for (i = 0; i <= len; ++i) |
283 | { |
310 | { |
|
|
311 | SV **svp = av_fetch (av, i, 0); |
|
|
312 | |
284 | encode_indent (enc); |
313 | encode_indent (enc); |
285 | encode_sv (enc, *av_fetch (av, i, 0)); |
314 | |
|
|
315 | if (svp) |
|
|
316 | encode_sv (enc, *svp); |
|
|
317 | else |
|
|
318 | encode_str (enc, "null", 4, 0); |
286 | |
319 | |
287 | if (i < len) |
320 | if (i < len) |
288 | encode_comma (enc); |
321 | encode_comma (enc); |
289 | } |
322 | } |
290 | |
323 | |
… | |
… | |
313 | else |
346 | else |
314 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
347 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
315 | |
348 | |
316 | encode_ch (enc, '"'); |
349 | encode_ch (enc, '"'); |
317 | |
350 | |
318 | if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
351 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
319 | encode_ch (enc, ':'); |
352 | encode_ch (enc, ':'); |
320 | if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
353 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
321 | encode_sv (enc, HeVAL (he)); |
354 | encode_sv (enc, HeVAL (he)); |
322 | } |
355 | } |
323 | |
356 | |
324 | // compare hash entries, used when all keys are bytestrings |
357 | // compare hash entries, used when all keys are bytestrings |
325 | static int |
358 | static int |
… | |
… | |
360 | { |
393 | { |
361 | // for canonical output we have to sort by keys first |
394 | // for canonical output we have to sort by keys first |
362 | // actually, this is mostly due to the stupid so-called |
395 | // actually, this is mostly due to the stupid so-called |
363 | // security workaround added somewhere in 5.8.x. |
396 | // security workaround added somewhere in 5.8.x. |
364 | // that randomises hash orderings |
397 | // that randomises hash orderings |
365 | if (enc->flags & F_CANONICAL) |
398 | if (enc->json.flags & F_CANONICAL) |
366 | { |
399 | { |
367 | HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
368 | int fast = 1; |
400 | int fast = 1; |
|
|
401 | HE *he; |
|
|
402 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
403 | HE **hes = _alloca (count * sizeof (HE)); |
|
|
404 | #else |
|
|
405 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
406 | #endif |
369 | |
407 | |
370 | i = 0; |
408 | i = 0; |
371 | while ((he = hv_iternext (hv))) |
409 | while ((he = hv_iternext (hv))) |
372 | { |
410 | { |
373 | hes [i++] = he; |
411 | hes [i++] = he; |
… | |
… | |
408 | |
446 | |
409 | encode_nl (enc); |
447 | encode_nl (enc); |
410 | } |
448 | } |
411 | else |
449 | else |
412 | { |
450 | { |
413 | SV *sv; |
|
|
414 | HE *he = hv_iternext (hv); |
451 | HE *he = hv_iternext (hv); |
415 | |
452 | |
416 | for (;;) |
453 | for (;;) |
417 | { |
454 | { |
418 | encode_indent (enc); |
455 | encode_indent (enc); |
… | |
… | |
433 | |
470 | |
434 | // encode objects, arrays and special \0=false and \1=true values. |
471 | // encode objects, arrays and special \0=false and \1=true values. |
435 | static void |
472 | static void |
436 | encode_rv (enc_t *enc, SV *sv) |
473 | encode_rv (enc_t *enc, SV *sv) |
437 | { |
474 | { |
|
|
475 | svtype svt; |
|
|
476 | |
438 | SvGETMAGIC (sv); |
477 | SvGETMAGIC (sv); |
439 | |
|
|
440 | svtype svt = SvTYPE (sv); |
478 | svt = SvTYPE (sv); |
441 | |
479 | |
|
|
480 | if (expect_false (SvOBJECT (sv))) |
|
|
481 | { |
|
|
482 | if (SvSTASH (sv) == json_boolean_stash) |
|
|
483 | { |
|
|
484 | if (SvIV (sv)) |
|
|
485 | encode_str (enc, "true", 4, 0); |
|
|
486 | else |
|
|
487 | encode_str (enc, "false", 5, 0); |
|
|
488 | } |
|
|
489 | else |
|
|
490 | { |
|
|
491 | #if 0 |
|
|
492 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
|
|
493 | { |
|
|
494 | // not yet |
|
|
495 | } |
|
|
496 | #endif |
|
|
497 | if (enc->json.flags & F_CONV_BLESSED) |
|
|
498 | { |
|
|
499 | // we re-bless the reference to get overload and other niceties right |
|
|
500 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
|
|
501 | |
|
|
502 | if (to_json) |
|
|
503 | { |
|
|
504 | int count; |
|
|
505 | dSP; |
|
|
506 | |
|
|
507 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
508 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
509 | |
|
|
510 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
511 | PUTBACK; |
|
|
512 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
|
|
513 | SPAGAIN; |
|
|
514 | |
|
|
515 | // catch this surprisingly common error |
|
|
516 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
517 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
518 | |
|
|
519 | sv = POPs; |
|
|
520 | PUTBACK; |
|
|
521 | |
|
|
522 | encode_sv (enc, sv); |
|
|
523 | |
|
|
524 | FREETMPS; LEAVE; |
|
|
525 | } |
|
|
526 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
527 | encode_str (enc, "null", 4, 0); |
|
|
528 | else |
|
|
529 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
530 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
531 | } |
|
|
532 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
533 | encode_str (enc, "null", 4, 0); |
|
|
534 | else |
|
|
535 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
|
|
536 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
537 | } |
|
|
538 | } |
442 | if (svt == SVt_PVHV) |
539 | else if (svt == SVt_PVHV) |
443 | encode_hv (enc, (HV *)sv); |
540 | encode_hv (enc, (HV *)sv); |
444 | else if (svt == SVt_PVAV) |
541 | else if (svt == SVt_PVAV) |
445 | encode_av (enc, (AV *)sv); |
542 | encode_av (enc, (AV *)sv); |
446 | else if (svt < SVt_PVAV) |
543 | else if (svt < SVt_PVAV) |
447 | { |
544 | { |
448 | if (SvNIOK (sv) && SvIV (sv) == 0) |
545 | STRLEN len = 0; |
|
|
546 | char *pv = svt ? SvPV (sv, len) : 0; |
|
|
547 | |
|
|
548 | if (len == 1 && *pv == '1') |
|
|
549 | encode_str (enc, "true", 4, 0); |
|
|
550 | else if (len == 1 && *pv == '0') |
449 | encode_str (enc, "false", 5, 0); |
551 | encode_str (enc, "false", 5, 0); |
450 | else if (SvNIOK (sv) && SvIV (sv) == 1) |
|
|
451 | encode_str (enc, "true", 4, 0); |
|
|
452 | else |
552 | else |
453 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
553 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
454 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
554 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
455 | } |
555 | } |
456 | else |
556 | else |
… | |
… | |
471 | encode_str (enc, str, len, SvUTF8 (sv)); |
571 | encode_str (enc, str, len, SvUTF8 (sv)); |
472 | encode_ch (enc, '"'); |
572 | encode_ch (enc, '"'); |
473 | } |
573 | } |
474 | else if (SvNOKp (sv)) |
574 | else if (SvNOKp (sv)) |
475 | { |
575 | { |
|
|
576 | // trust that perl will do the right thing w.r.t. JSON syntax. |
476 | need (enc, NV_DIG + 32); |
577 | need (enc, NV_DIG + 32); |
477 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
578 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
478 | enc->cur += strlen (enc->cur); |
579 | enc->cur += strlen (enc->cur); |
479 | } |
580 | } |
480 | else if (SvIOKp (sv)) |
581 | else if (SvIOKp (sv)) |
481 | { |
582 | { |
482 | need (enc, 64); |
583 | // we assume we can always read an IV as a UV |
|
|
584 | if (SvUV (sv) & ~(UV)0x7fff) |
|
|
585 | { |
|
|
586 | // large integer, use the (rather slow) snprintf way. |
|
|
587 | need (enc, sizeof (UV) * 3); |
483 | enc->cur += |
588 | enc->cur += |
484 | SvIsUV(sv) |
589 | SvIsUV(sv) |
485 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
590 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
486 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
591 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
592 | } |
|
|
593 | else |
|
|
594 | { |
|
|
595 | // optimise the "small number case" |
|
|
596 | // code will likely be branchless and use only a single multiplication |
|
|
597 | I32 i = SvIV (sv); |
|
|
598 | U32 u; |
|
|
599 | char digit, nz = 0; |
|
|
600 | |
|
|
601 | need (enc, 6); |
|
|
602 | |
|
|
603 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
604 | u = i < 0 ? -i : i; |
|
|
605 | |
|
|
606 | // convert to 4.28 fixed-point representation |
|
|
607 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
608 | |
|
|
609 | // now output digit by digit, each time masking out the integer part |
|
|
610 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
611 | // resulting in a net multiplication by 10. |
|
|
612 | // we always write the digit to memory but conditionally increment |
|
|
613 | // the pointer, to ease the usage of conditional move instructions. |
|
|
614 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
|
|
615 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
|
|
616 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
|
|
617 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
|
|
618 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
619 | } |
487 | } |
620 | } |
488 | else if (SvROK (sv)) |
621 | else if (SvROK (sv)) |
489 | encode_rv (enc, SvRV (sv)); |
622 | encode_rv (enc, SvRV (sv)); |
490 | else if (!SvOK (sv)) |
623 | else if (!SvOK (sv)) |
491 | encode_str (enc, "null", 4, 0); |
624 | encode_str (enc, "null", 4, 0); |
… | |
… | |
493 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
626 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
494 | SvPV_nolen (sv), SvFLAGS (sv)); |
627 | SvPV_nolen (sv), SvFLAGS (sv)); |
495 | } |
628 | } |
496 | |
629 | |
497 | static SV * |
630 | static SV * |
498 | encode_json (SV *scalar, U32 flags) |
631 | encode_json (SV *scalar, JSON *json) |
499 | { |
632 | { |
|
|
633 | enc_t enc; |
|
|
634 | |
500 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
635 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
501 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
636 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
502 | |
637 | |
503 | enc_t enc; |
638 | enc.json = *json; |
504 | enc.flags = flags; |
|
|
505 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
639 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
506 | enc.cur = SvPVX (enc.sv); |
640 | enc.cur = SvPVX (enc.sv); |
507 | enc.end = SvEND (enc.sv); |
641 | enc.end = SvEND (enc.sv); |
508 | enc.indent = 0; |
642 | enc.indent = 0; |
509 | enc.maxdepth = DEC_DEPTH (flags); |
643 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
510 | |
644 | |
511 | SvPOK_only (enc.sv); |
645 | SvPOK_only (enc.sv); |
512 | encode_sv (&enc, scalar); |
646 | encode_sv (&enc, scalar); |
513 | |
647 | |
|
|
648 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
649 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
650 | |
514 | if (!(flags & (F_ASCII | F_UTF8))) |
651 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
515 | SvUTF8_on (enc.sv); |
652 | SvUTF8_on (enc.sv); |
516 | |
653 | |
517 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
518 | |
|
|
519 | if (enc.flags & F_SHRINK) |
654 | if (enc.json.flags & F_SHRINK) |
520 | shrink (enc.sv); |
655 | shrink (enc.sv); |
521 | |
656 | |
522 | return enc.sv; |
657 | return enc.sv; |
523 | } |
658 | } |
524 | |
659 | |
… | |
… | |
529 | typedef struct |
664 | typedef struct |
530 | { |
665 | { |
531 | char *cur; // current parser pointer |
666 | char *cur; // current parser pointer |
532 | char *end; // end of input string |
667 | char *end; // end of input string |
533 | const char *err; // parse error, if != 0 |
668 | const char *err; // parse error, if != 0 |
534 | U32 flags; // F_* |
669 | JSON json; |
535 | U32 depth; // recursion depth |
670 | U32 depth; // recursion depth |
536 | U32 maxdepth; // recursion depth limit |
671 | U32 maxdepth; // recursion depth limit |
537 | } dec_t; |
672 | } dec_t; |
538 | |
673 | |
539 | static void |
674 | inline void |
540 | decode_ws (dec_t *dec) |
675 | decode_ws (dec_t *dec) |
541 | { |
676 | { |
542 | for (;;) |
677 | for (;;) |
543 | { |
678 | { |
544 | char ch = *dec->cur; |
679 | char ch = *dec->cur; |
… | |
… | |
570 | decode_4hex (dec_t *dec) |
705 | decode_4hex (dec_t *dec) |
571 | { |
706 | { |
572 | signed char d1, d2, d3, d4; |
707 | signed char d1, d2, d3, d4; |
573 | unsigned char *cur = (unsigned char *)dec->cur; |
708 | unsigned char *cur = (unsigned char *)dec->cur; |
574 | |
709 | |
575 | d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); |
710 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
576 | d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); |
711 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
577 | d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); |
712 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
578 | d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); |
713 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
579 | |
714 | |
580 | dec->cur += 4; |
715 | dec->cur += 4; |
581 | |
716 | |
582 | return ((UV)d1) << 12 |
717 | return ((UV)d1) << 12 |
583 | | ((UV)d2) << 8 |
718 | | ((UV)d2) << 8 |
… | |
… | |
591 | static SV * |
726 | static SV * |
592 | decode_str (dec_t *dec) |
727 | decode_str (dec_t *dec) |
593 | { |
728 | { |
594 | SV *sv = 0; |
729 | SV *sv = 0; |
595 | int utf8 = 0; |
730 | int utf8 = 0; |
|
|
731 | char *dec_cur = dec->cur; |
596 | |
732 | |
597 | do |
733 | do |
598 | { |
734 | { |
599 | char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; |
735 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
600 | char *cur = buf; |
736 | char *cur = buf; |
601 | |
737 | |
602 | do |
738 | do |
603 | { |
739 | { |
604 | unsigned char ch = *(unsigned char *)dec->cur++; |
740 | unsigned char ch = *(unsigned char *)dec_cur++; |
605 | |
741 | |
606 | if (ch == '"') |
742 | if (expect_false (ch == '"')) |
607 | { |
743 | { |
608 | --dec->cur; |
744 | --dec_cur; |
609 | break; |
745 | break; |
610 | } |
746 | } |
611 | else if (ch == '\\') |
747 | else if (expect_false (ch == '\\')) |
612 | { |
748 | { |
613 | switch (*dec->cur) |
749 | switch (*dec_cur) |
614 | { |
750 | { |
615 | case '\\': |
751 | case '\\': |
616 | case '/': |
752 | case '/': |
617 | case '"': *cur++ = *dec->cur++; break; |
753 | case '"': *cur++ = *dec_cur++; break; |
618 | |
754 | |
619 | case 'b': ++dec->cur; *cur++ = '\010'; break; |
755 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
620 | case 't': ++dec->cur; *cur++ = '\011'; break; |
756 | case 't': ++dec_cur; *cur++ = '\011'; break; |
621 | case 'n': ++dec->cur; *cur++ = '\012'; break; |
757 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
622 | case 'f': ++dec->cur; *cur++ = '\014'; break; |
758 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
623 | case 'r': ++dec->cur; *cur++ = '\015'; break; |
759 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
624 | |
760 | |
625 | case 'u': |
761 | case 'u': |
626 | { |
762 | { |
627 | UV lo, hi; |
763 | UV lo, hi; |
628 | ++dec->cur; |
764 | ++dec_cur; |
629 | |
765 | |
|
|
766 | dec->cur = dec_cur; |
630 | hi = decode_4hex (dec); |
767 | hi = decode_4hex (dec); |
|
|
768 | dec_cur = dec->cur; |
631 | if (hi == (UV)-1) |
769 | if (hi == (UV)-1) |
632 | goto fail; |
770 | goto fail; |
633 | |
771 | |
634 | // possibly a surrogate pair |
772 | // possibly a surrogate pair |
635 | if (hi >= 0xd800) |
773 | if (hi >= 0xd800) |
636 | if (hi < 0xdc00) |
774 | if (hi < 0xdc00) |
637 | { |
775 | { |
638 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
776 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
639 | ERR ("missing low surrogate character in surrogate pair"); |
777 | ERR ("missing low surrogate character in surrogate pair"); |
640 | |
778 | |
641 | dec->cur += 2; |
779 | dec_cur += 2; |
642 | |
780 | |
|
|
781 | dec->cur = dec_cur; |
643 | lo = decode_4hex (dec); |
782 | lo = decode_4hex (dec); |
|
|
783 | dec_cur = dec->cur; |
644 | if (lo == (UV)-1) |
784 | if (lo == (UV)-1) |
645 | goto fail; |
785 | goto fail; |
646 | |
786 | |
647 | if (lo < 0xdc00 || lo >= 0xe000) |
787 | if (lo < 0xdc00 || lo >= 0xe000) |
648 | ERR ("surrogate pair expected"); |
788 | ERR ("surrogate pair expected"); |
… | |
… | |
662 | *cur++ = hi; |
802 | *cur++ = hi; |
663 | } |
803 | } |
664 | break; |
804 | break; |
665 | |
805 | |
666 | default: |
806 | default: |
667 | --dec->cur; |
807 | --dec_cur; |
668 | ERR ("illegal backslash escape sequence in string"); |
808 | ERR ("illegal backslash escape sequence in string"); |
669 | } |
809 | } |
670 | } |
810 | } |
671 | else if (ch >= 0x20 && ch <= 0x7f) |
811 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
672 | *cur++ = ch; |
812 | *cur++ = ch; |
673 | else if (ch >= 0x80) |
813 | else if (ch >= 0x80) |
674 | { |
814 | { |
675 | --dec->cur; |
|
|
676 | |
|
|
677 | STRLEN clen; |
815 | STRLEN clen; |
|
|
816 | UV uch; |
|
|
817 | |
|
|
818 | --dec_cur; |
|
|
819 | |
678 | UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); |
820 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
679 | if (clen == (STRLEN)-1) |
821 | if (clen == (STRLEN)-1) |
680 | ERR ("malformed UTF-8 character in JSON string"); |
822 | ERR ("malformed UTF-8 character in JSON string"); |
681 | |
823 | |
682 | do |
824 | do |
683 | *cur++ = *dec->cur++; |
825 | *cur++ = *dec_cur++; |
684 | while (--clen); |
826 | while (--clen); |
685 | |
827 | |
686 | utf8 = 1; |
828 | utf8 = 1; |
687 | } |
829 | } |
688 | else |
830 | else |
689 | { |
831 | { |
690 | --dec->cur; |
832 | --dec_cur; |
691 | |
833 | |
692 | if (!ch) |
834 | if (!ch) |
693 | ERR ("unexpected end of string while parsing json string"); |
835 | ERR ("unexpected end of string while parsing JSON string"); |
694 | else |
836 | else |
695 | ERR ("invalid character encountered while parsing json string"); |
837 | ERR ("invalid character encountered while parsing JSON string"); |
696 | } |
838 | } |
697 | } |
839 | } |
698 | while (cur < buf + SHORT_STRING_LEN); |
840 | while (cur < buf + SHORT_STRING_LEN); |
699 | |
841 | |
|
|
842 | { |
700 | STRLEN len = cur - buf; |
843 | STRLEN len = cur - buf; |
701 | |
844 | |
702 | if (sv) |
845 | if (sv) |
703 | { |
846 | { |
704 | SvGROW (sv, SvCUR (sv) + len + 1); |
847 | SvGROW (sv, SvCUR (sv) + len + 1); |
705 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
848 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
706 | SvCUR_set (sv, SvCUR (sv) + len); |
849 | SvCUR_set (sv, SvCUR (sv) + len); |
707 | } |
850 | } |
708 | else |
851 | else |
709 | sv = newSVpvn (buf, len); |
852 | sv = newSVpvn (buf, len); |
710 | } |
853 | } |
|
|
854 | } |
711 | while (*dec->cur != '"'); |
855 | while (*dec_cur != '"'); |
712 | |
856 | |
713 | ++dec->cur; |
857 | ++dec_cur; |
714 | |
858 | |
715 | if (sv) |
859 | if (sv) |
716 | { |
860 | { |
717 | SvPOK_only (sv); |
861 | SvPOK_only (sv); |
718 | *SvEND (sv) = 0; |
862 | *SvEND (sv) = 0; |
… | |
… | |
721 | SvUTF8_on (sv); |
865 | SvUTF8_on (sv); |
722 | } |
866 | } |
723 | else |
867 | else |
724 | sv = newSVpvn ("", 0); |
868 | sv = newSVpvn ("", 0); |
725 | |
869 | |
|
|
870 | dec->cur = dec_cur; |
726 | return sv; |
871 | return sv; |
727 | |
872 | |
728 | fail: |
873 | fail: |
|
|
874 | dec->cur = dec_cur; |
729 | return 0; |
875 | return 0; |
730 | } |
876 | } |
731 | |
877 | |
732 | static SV * |
878 | static SV * |
733 | decode_num (dec_t *dec) |
879 | decode_num (dec_t *dec) |
… | |
… | |
791 | is_nv = 1; |
937 | is_nv = 1; |
792 | } |
938 | } |
793 | |
939 | |
794 | if (!is_nv) |
940 | if (!is_nv) |
795 | { |
941 | { |
796 | UV uv; |
942 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
797 | int numtype = grok_number (start, dec->cur - start, &uv); |
943 | if (*start == '-') |
798 | if (numtype & IS_NUMBER_IN_UV) |
944 | switch (dec->cur - start) |
799 | if (numtype & IS_NUMBER_NEG) |
|
|
800 | { |
945 | { |
801 | if (uv < (UV)IV_MIN) |
946 | case 2: return newSViv (-( start [1] - '0' * 1)); |
802 | return newSViv (-(IV)uv); |
947 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
|
|
948 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
949 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
803 | } |
950 | } |
|
|
951 | else |
|
|
952 | switch (dec->cur - start) |
|
|
953 | { |
|
|
954 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
955 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
956 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
957 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
958 | } |
|
|
959 | |
|
|
960 | { |
|
|
961 | UV uv; |
|
|
962 | int numtype = grok_number (start, dec->cur - start, &uv); |
|
|
963 | if (numtype & IS_NUMBER_IN_UV) |
|
|
964 | if (numtype & IS_NUMBER_NEG) |
|
|
965 | { |
|
|
966 | if (uv < (UV)IV_MIN) |
|
|
967 | return newSViv (-(IV)uv); |
|
|
968 | } |
804 | else |
969 | else |
805 | return newSVuv (uv); |
970 | return newSVuv (uv); |
|
|
971 | |
|
|
972 | // here would likely be the place for bigint support |
806 | } |
973 | } |
|
|
974 | } |
807 | |
975 | |
|
|
976 | // if we ever support bigint or bigfloat, this is the place for bigfloat |
808 | return newSVnv (Atof (start)); |
977 | return newSVnv (Atof (start)); |
809 | |
978 | |
810 | fail: |
979 | fail: |
811 | return 0; |
980 | return 0; |
812 | } |
981 | } |
… | |
… | |
856 | } |
1025 | } |
857 | |
1026 | |
858 | static SV * |
1027 | static SV * |
859 | decode_hv (dec_t *dec) |
1028 | decode_hv (dec_t *dec) |
860 | { |
1029 | { |
|
|
1030 | SV *sv; |
861 | HV *hv = newHV (); |
1031 | HV *hv = newHV (); |
862 | |
1032 | |
863 | DEC_INC_DEPTH; |
1033 | DEC_INC_DEPTH; |
864 | decode_ws (dec); |
1034 | decode_ws (dec); |
865 | |
1035 | |
866 | if (*dec->cur == '}') |
1036 | if (*dec->cur == '}') |
867 | ++dec->cur; |
1037 | ++dec->cur; |
868 | else |
1038 | else |
869 | for (;;) |
1039 | for (;;) |
870 | { |
1040 | { |
871 | SV *key, *value; |
|
|
872 | |
|
|
873 | decode_ws (dec); EXPECT_CH ('"'); |
1041 | decode_ws (dec); EXPECT_CH ('"'); |
874 | |
1042 | |
875 | key = decode_str (dec); |
1043 | // heuristic: assume that |
876 | if (!key) |
1044 | // a) decode_str + hv_store_ent are abysmally slow. |
877 | goto fail; |
1045 | // b) most hash keys are short, simple ascii text. |
|
|
1046 | // => try to "fast-match" such strings to avoid |
|
|
1047 | // the overhead of decode_str + hv_store_ent. |
|
|
1048 | { |
|
|
1049 | SV *value; |
|
|
1050 | char *p = dec->cur; |
|
|
1051 | char *e = p + 24; // only try up to 24 bytes |
878 | |
1052 | |
879 | decode_ws (dec); EXPECT_CH (':'); |
1053 | for (;;) |
880 | |
|
|
881 | value = decode_sv (dec); |
|
|
882 | if (!value) |
|
|
883 | { |
1054 | { |
|
|
1055 | // the >= 0x80 is true on most architectures |
|
|
1056 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1057 | { |
|
|
1058 | // slow path, back up and use decode_str |
|
|
1059 | SV *key = decode_str (dec); |
|
|
1060 | if (!key) |
|
|
1061 | goto fail; |
|
|
1062 | |
|
|
1063 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1064 | |
|
|
1065 | value = decode_sv (dec); |
|
|
1066 | if (!value) |
|
|
1067 | { |
|
|
1068 | SvREFCNT_dec (key); |
|
|
1069 | goto fail; |
|
|
1070 | } |
|
|
1071 | |
|
|
1072 | hv_store_ent (hv, key, value, 0); |
884 | SvREFCNT_dec (key); |
1073 | SvREFCNT_dec (key); |
|
|
1074 | |
|
|
1075 | break; |
|
|
1076 | } |
|
|
1077 | else if (*p == '"') |
|
|
1078 | { |
|
|
1079 | // fast path, got a simple key |
|
|
1080 | char *key = dec->cur; |
|
|
1081 | int len = p - key; |
|
|
1082 | dec->cur = p + 1; |
|
|
1083 | |
|
|
1084 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1085 | |
|
|
1086 | value = decode_sv (dec); |
|
|
1087 | if (!value) |
885 | goto fail; |
1088 | goto fail; |
|
|
1089 | |
|
|
1090 | hv_store (hv, key, len, value, 0); |
|
|
1091 | |
|
|
1092 | break; |
|
|
1093 | } |
|
|
1094 | |
|
|
1095 | ++p; |
886 | } |
1096 | } |
887 | |
1097 | } |
888 | hv_store_ent (hv, key, value, 0); |
|
|
889 | SvREFCNT_dec (key); |
|
|
890 | |
1098 | |
891 | decode_ws (dec); |
1099 | decode_ws (dec); |
892 | |
1100 | |
893 | if (*dec->cur == '}') |
1101 | if (*dec->cur == '}') |
894 | { |
1102 | { |
… | |
… | |
901 | |
1109 | |
902 | ++dec->cur; |
1110 | ++dec->cur; |
903 | } |
1111 | } |
904 | |
1112 | |
905 | DEC_DEC_DEPTH; |
1113 | DEC_DEC_DEPTH; |
906 | return newRV_noinc ((SV *)hv); |
1114 | sv = newRV_noinc ((SV *)hv); |
|
|
1115 | |
|
|
1116 | // check filter callbacks |
|
|
1117 | if (dec->json.flags & F_HOOK) |
|
|
1118 | { |
|
|
1119 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1120 | { |
|
|
1121 | HE *cb, *he; |
|
|
1122 | |
|
|
1123 | hv_iterinit (hv); |
|
|
1124 | he = hv_iternext (hv); |
|
|
1125 | hv_iterinit (hv); |
|
|
1126 | |
|
|
1127 | // the next line creates a mortal sv each time its called. |
|
|
1128 | // might want to optimise this for common cases. |
|
|
1129 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1130 | |
|
|
1131 | if (cb) |
|
|
1132 | { |
|
|
1133 | dSP; |
|
|
1134 | int count; |
|
|
1135 | |
|
|
1136 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1137 | XPUSHs (HeVAL (he)); |
|
|
1138 | |
|
|
1139 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1140 | |
|
|
1141 | if (count == 1) |
|
|
1142 | { |
|
|
1143 | sv = newSVsv (POPs); |
|
|
1144 | FREETMPS; LEAVE; |
|
|
1145 | return sv; |
|
|
1146 | } |
|
|
1147 | |
|
|
1148 | FREETMPS; LEAVE; |
|
|
1149 | } |
|
|
1150 | } |
|
|
1151 | |
|
|
1152 | if (dec->json.cb_object) |
|
|
1153 | { |
|
|
1154 | dSP; |
|
|
1155 | int count; |
|
|
1156 | |
|
|
1157 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1158 | XPUSHs (sv_2mortal (sv)); |
|
|
1159 | |
|
|
1160 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1161 | |
|
|
1162 | if (count == 1) |
|
|
1163 | { |
|
|
1164 | sv = newSVsv (POPs); |
|
|
1165 | FREETMPS; LEAVE; |
|
|
1166 | return sv; |
|
|
1167 | } |
|
|
1168 | |
|
|
1169 | SvREFCNT_inc (sv); |
|
|
1170 | FREETMPS; LEAVE; |
|
|
1171 | } |
|
|
1172 | } |
|
|
1173 | |
|
|
1174 | return sv; |
907 | |
1175 | |
908 | fail: |
1176 | fail: |
909 | SvREFCNT_dec (hv); |
1177 | SvREFCNT_dec (hv); |
910 | DEC_DEC_DEPTH; |
1178 | DEC_DEC_DEPTH; |
911 | return 0; |
1179 | return 0; |
… | |
… | |
913 | |
1181 | |
914 | static SV * |
1182 | static SV * |
915 | decode_sv (dec_t *dec) |
1183 | decode_sv (dec_t *dec) |
916 | { |
1184 | { |
917 | decode_ws (dec); |
1185 | decode_ws (dec); |
|
|
1186 | |
|
|
1187 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1188 | // to parse anything. |
918 | switch (*dec->cur) |
1189 | switch (*dec->cur) |
919 | { |
1190 | { |
920 | case '"': ++dec->cur; return decode_str (dec); |
1191 | case '"': ++dec->cur; return decode_str (dec); |
921 | case '[': ++dec->cur; return decode_av (dec); |
1192 | case '[': ++dec->cur; return decode_av (dec); |
922 | case '{': ++dec->cur; return decode_hv (dec); |
1193 | case '{': ++dec->cur; return decode_hv (dec); |
… | |
… | |
928 | |
1199 | |
929 | case 't': |
1200 | case 't': |
930 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1201 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
931 | { |
1202 | { |
932 | dec->cur += 4; |
1203 | dec->cur += 4; |
933 | return newSViv (1); |
1204 | return SvREFCNT_inc (json_true); |
934 | } |
1205 | } |
935 | else |
1206 | else |
936 | ERR ("'true' expected"); |
1207 | ERR ("'true' expected"); |
937 | |
1208 | |
938 | break; |
1209 | break; |
939 | |
1210 | |
940 | case 'f': |
1211 | case 'f': |
941 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1212 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
942 | { |
1213 | { |
943 | dec->cur += 5; |
1214 | dec->cur += 5; |
944 | return newSViv (0); |
1215 | return SvREFCNT_inc (json_false); |
945 | } |
1216 | } |
946 | else |
1217 | else |
947 | ERR ("'false' expected"); |
1218 | ERR ("'false' expected"); |
948 | |
1219 | |
949 | break; |
1220 | break; |
… | |
… | |
958 | ERR ("'null' expected"); |
1229 | ERR ("'null' expected"); |
959 | |
1230 | |
960 | break; |
1231 | break; |
961 | |
1232 | |
962 | default: |
1233 | default: |
963 | ERR ("malformed json string, neither array, object, number, string or atom"); |
1234 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
964 | break; |
1235 | break; |
965 | } |
1236 | } |
966 | |
1237 | |
967 | fail: |
1238 | fail: |
968 | return 0; |
1239 | return 0; |
969 | } |
1240 | } |
970 | |
1241 | |
971 | static SV * |
1242 | static SV * |
972 | decode_json (SV *string, U32 flags) |
1243 | decode_json (SV *string, JSON *json, UV *offset_return) |
973 | { |
1244 | { |
|
|
1245 | dec_t dec; |
|
|
1246 | UV offset; |
974 | SV *sv; |
1247 | SV *sv; |
975 | |
1248 | |
|
|
1249 | SvGETMAGIC (string); |
976 | SvUPGRADE (string, SVt_PV); |
1250 | SvUPGRADE (string, SVt_PV); |
977 | |
1251 | |
|
|
1252 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
|
|
1253 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1254 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
|
|
1255 | |
978 | if (flags & F_UTF8) |
1256 | if (json->flags & F_UTF8) |
979 | sv_utf8_downgrade (string, 0); |
1257 | sv_utf8_downgrade (string, 0); |
980 | else |
1258 | else |
981 | sv_utf8_upgrade (string); |
1259 | sv_utf8_upgrade (string); |
982 | |
1260 | |
983 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1261 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
984 | |
1262 | |
985 | dec_t dec; |
1263 | dec.json = *json; |
986 | dec.flags = flags; |
|
|
987 | dec.cur = SvPVX (string); |
1264 | dec.cur = SvPVX (string); |
988 | dec.end = SvEND (string); |
1265 | dec.end = SvEND (string); |
989 | dec.err = 0; |
1266 | dec.err = 0; |
990 | dec.depth = 0; |
1267 | dec.depth = 0; |
991 | dec.maxdepth = DEC_DEPTH (dec.flags); |
1268 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
992 | |
1269 | |
|
|
1270 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1271 | dec.json.flags |= F_HOOK; |
|
|
1272 | |
993 | *dec.end = 0; // this should basically be a nop, too, but make sure its there |
1273 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
994 | sv = decode_sv (&dec); |
1274 | sv = decode_sv (&dec); |
995 | |
1275 | |
|
|
1276 | if (!(offset_return || !sv)) |
|
|
1277 | { |
|
|
1278 | // check for trailing garbage |
|
|
1279 | decode_ws (&dec); |
|
|
1280 | |
|
|
1281 | if (*dec.cur) |
|
|
1282 | { |
|
|
1283 | dec.err = "garbage after JSON object"; |
|
|
1284 | SvREFCNT_dec (sv); |
|
|
1285 | sv = 0; |
|
|
1286 | } |
|
|
1287 | } |
|
|
1288 | |
|
|
1289 | if (offset_return || !sv) |
|
|
1290 | { |
|
|
1291 | offset = dec.json.flags & F_UTF8 |
|
|
1292 | ? dec.cur - SvPVX (string) |
|
|
1293 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1294 | |
|
|
1295 | if (offset_return) |
|
|
1296 | *offset_return = offset; |
|
|
1297 | } |
|
|
1298 | |
996 | if (!sv) |
1299 | if (!sv) |
997 | { |
1300 | { |
998 | IV offset = dec.flags & F_UTF8 |
|
|
999 | ? dec.cur - SvPVX (string) |
|
|
1000 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1001 | SV *uni = sv_newmortal (); |
1301 | SV *uni = sv_newmortal (); |
1002 | |
1302 | |
1003 | // horrible hack to silence warning inside pv_uni_display |
1303 | // horrible hack to silence warning inside pv_uni_display |
1004 | COP cop = *PL_curcop; |
1304 | COP cop = *PL_curcop; |
1005 | cop.cop_warnings = pWARN_NONE; |
1305 | cop.cop_warnings = pWARN_NONE; |
… | |
… | |
1015 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1315 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1016 | } |
1316 | } |
1017 | |
1317 | |
1018 | sv = sv_2mortal (sv); |
1318 | sv = sv_2mortal (sv); |
1019 | |
1319 | |
1020 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1320 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1021 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1321 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1022 | |
1322 | |
1023 | return sv; |
1323 | return sv; |
1024 | } |
1324 | } |
1025 | |
1325 | |
… | |
… | |
1029 | MODULE = JSON::XS PACKAGE = JSON::XS |
1329 | MODULE = JSON::XS PACKAGE = JSON::XS |
1030 | |
1330 | |
1031 | BOOT: |
1331 | BOOT: |
1032 | { |
1332 | { |
1033 | int i; |
1333 | int i; |
1034 | |
|
|
1035 | memset (decode_hexdigit, 0xff, 256); |
|
|
1036 | |
1334 | |
1037 | for (i = 0; i < 256; ++i) |
1335 | for (i = 0; i < 256; ++i) |
1038 | decode_hexdigit [i] = |
1336 | decode_hexdigit [i] = |
1039 | i >= '0' && i <= '9' ? i - '0' |
1337 | i >= '0' && i <= '9' ? i - '0' |
1040 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1338 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1041 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1339 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1042 | : -1; |
1340 | : -1; |
1043 | |
1341 | |
1044 | json_stash = gv_stashpv ("JSON::XS", 1); |
1342 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1343 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
1344 | |
|
|
1345 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
|
|
1346 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1045 | } |
1347 | } |
1046 | |
1348 | |
1047 | PROTOTYPES: DISABLE |
1349 | PROTOTYPES: DISABLE |
1048 | |
1350 | |
1049 | SV *new (char *dummy) |
1351 | void new (char *klass) |
1050 | CODE: |
1352 | PPCODE: |
1051 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1353 | { |
1052 | OUTPUT: |
1354 | SV *pv = NEWSV (0, sizeof (JSON)); |
1053 | RETVAL |
1355 | SvPOK_only (pv); |
|
|
1356 | Zero (SvPVX (pv), 1, JSON); |
|
|
1357 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1358 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash))); |
|
|
1359 | } |
1054 | |
1360 | |
1055 | SV *ascii (SV *self, int enable = 1) |
1361 | void ascii (JSON *self, int enable = 1) |
1056 | ALIAS: |
1362 | ALIAS: |
1057 | ascii = F_ASCII |
1363 | ascii = F_ASCII |
|
|
1364 | latin1 = F_LATIN1 |
1058 | utf8 = F_UTF8 |
1365 | utf8 = F_UTF8 |
1059 | indent = F_INDENT |
1366 | indent = F_INDENT |
1060 | canonical = F_CANONICAL |
1367 | canonical = F_CANONICAL |
1061 | space_before = F_SPACE_BEFORE |
1368 | space_before = F_SPACE_BEFORE |
1062 | space_after = F_SPACE_AFTER |
1369 | space_after = F_SPACE_AFTER |
1063 | pretty = F_PRETTY |
1370 | pretty = F_PRETTY |
1064 | allow_nonref = F_ALLOW_NONREF |
1371 | allow_nonref = F_ALLOW_NONREF |
1065 | shrink = F_SHRINK |
1372 | shrink = F_SHRINK |
|
|
1373 | allow_blessed = F_ALLOW_BLESSED |
|
|
1374 | convert_blessed = F_CONV_BLESSED |
1066 | CODE: |
1375 | PPCODE: |
1067 | { |
1376 | { |
1068 | UV *uv = SvJSON (self); |
|
|
1069 | if (enable) |
1377 | if (enable) |
1070 | *uv |= ix; |
1378 | self->flags |= ix; |
1071 | else |
1379 | else |
1072 | *uv &= ~ix; |
1380 | self->flags &= ~ix; |
1073 | |
1381 | |
1074 | RETVAL = newSVsv (self); |
1382 | XPUSHs (ST (0)); |
1075 | } |
1383 | } |
1076 | OUTPUT: |
|
|
1077 | RETVAL |
|
|
1078 | |
1384 | |
1079 | SV *max_depth (SV *self, int max_depth = 0x80000000UL) |
1385 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1080 | CODE: |
1386 | PPCODE: |
1081 | { |
1387 | { |
1082 | UV *uv = SvJSON (self); |
|
|
1083 | UV log2 = 0; |
1388 | UV log2 = 0; |
1084 | |
1389 | |
1085 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1390 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1086 | |
1391 | |
1087 | while ((1UL << log2) < max_depth) |
1392 | while ((1UL << log2) < max_depth) |
1088 | ++log2; |
1393 | ++log2; |
1089 | |
1394 | |
1090 | *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1395 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1091 | |
1396 | |
1092 | RETVAL = newSVsv (self); |
1397 | XPUSHs (ST (0)); |
1093 | } |
1398 | } |
1094 | OUTPUT: |
|
|
1095 | RETVAL |
|
|
1096 | |
1399 | |
1097 | void encode (SV *self, SV *scalar) |
1400 | void max_size (JSON *self, UV max_size = 0) |
1098 | PPCODE: |
1401 | PPCODE: |
1099 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
1402 | { |
|
|
1403 | UV log2 = 0; |
1100 | |
1404 | |
1101 | void decode (SV *self, SV *jsonstr) |
1405 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1406 | if (max_size == 1) max_size = 2; |
|
|
1407 | |
|
|
1408 | while ((1UL << log2) < max_size) |
|
|
1409 | ++log2; |
|
|
1410 | |
|
|
1411 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1412 | |
|
|
1413 | XPUSHs (ST (0)); |
|
|
1414 | } |
|
|
1415 | |
|
|
1416 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1102 | PPCODE: |
1417 | PPCODE: |
|
|
1418 | { |
|
|
1419 | SvREFCNT_dec (self->cb_object); |
|
|
1420 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
|
|
1421 | |
|
|
1422 | XPUSHs (ST (0)); |
|
|
1423 | } |
|
|
1424 | |
|
|
1425 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
|
|
1426 | PPCODE: |
|
|
1427 | { |
|
|
1428 | if (!self->cb_sk_object) |
|
|
1429 | self->cb_sk_object = newHV (); |
|
|
1430 | |
|
|
1431 | if (SvOK (cb)) |
|
|
1432 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
1433 | else |
|
|
1434 | { |
|
|
1435 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
1436 | |
|
|
1437 | if (!HvKEYS (self->cb_sk_object)) |
|
|
1438 | { |
|
|
1439 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1440 | self->cb_sk_object = 0; |
|
|
1441 | } |
|
|
1442 | } |
|
|
1443 | |
|
|
1444 | XPUSHs (ST (0)); |
|
|
1445 | } |
|
|
1446 | |
|
|
1447 | void encode (JSON *self, SV *scalar) |
|
|
1448 | PPCODE: |
|
|
1449 | XPUSHs (encode_json (scalar, self)); |
|
|
1450 | |
|
|
1451 | void decode (JSON *self, SV *jsonstr) |
|
|
1452 | PPCODE: |
1103 | XPUSHs (decode_json (jsonstr, *SvJSON (self))); |
1453 | XPUSHs (decode_json (jsonstr, self, 0)); |
|
|
1454 | |
|
|
1455 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
1456 | PPCODE: |
|
|
1457 | { |
|
|
1458 | UV offset; |
|
|
1459 | EXTEND (SP, 2); |
|
|
1460 | PUSHs (decode_json (jsonstr, self, &offset)); |
|
|
1461 | PUSHs (sv_2mortal (newSVuv (offset))); |
|
|
1462 | } |
|
|
1463 | |
|
|
1464 | void DESTROY (JSON *self) |
|
|
1465 | CODE: |
|
|
1466 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1467 | SvREFCNT_dec (self->cb_object); |
1104 | |
1468 | |
1105 | PROTOTYPES: ENABLE |
1469 | PROTOTYPES: ENABLE |
1106 | |
1470 | |
1107 | void to_json (SV *scalar) |
1471 | void to_json (SV *scalar) |
1108 | ALIAS: |
|
|
1109 | objToJson = 0 |
|
|
1110 | PPCODE: |
1472 | PPCODE: |
|
|
1473 | { |
|
|
1474 | JSON json = { F_DEFAULT | F_UTF8 }; |
1111 | XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); |
1475 | XPUSHs (encode_json (scalar, &json)); |
|
|
1476 | } |
1112 | |
1477 | |
1113 | void from_json (SV *jsonstr) |
1478 | void from_json (SV *jsonstr) |
1114 | ALIAS: |
|
|
1115 | jsonToObj = 0 |
|
|
1116 | PPCODE: |
1479 | PPCODE: |
|
|
1480 | { |
|
|
1481 | JSON json = { F_DEFAULT | F_UTF8 }; |
1117 | XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8)); |
1482 | XPUSHs (decode_json (jsonstr, &json, 0)); |
|
|
1483 | } |
1118 | |
1484 | |