… | |
… | |
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 (13UL << 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, *cb_sk_object; |
|
|
68 | } JSON; |
38 | |
69 | |
39 | ///////////////////////////////////////////////////////////////////////////// |
70 | ///////////////////////////////////////////////////////////////////////////// |
40 | // utility functions |
71 | // utility functions |
41 | |
72 | |
42 | static UV * |
73 | 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) |
74 | shrink (SV *sv) |
53 | { |
75 | { |
54 | sv_utf8_downgrade (sv, 1); |
76 | sv_utf8_downgrade (sv, 1); |
55 | if (SvLEN (sv) > SvCUR (sv) + 1) |
77 | if (SvLEN (sv) > SvCUR (sv) + 1) |
56 | { |
78 | { |
… | |
… | |
65 | // decode an utf-8 character and return it, or (UV)-1 in |
87 | // decode an utf-8 character and return it, or (UV)-1 in |
66 | // case of an error. |
88 | // case of an error. |
67 | // we special-case "safe" characters from U+80 .. U+7FF, |
89 | // we special-case "safe" characters from U+80 .. U+7FF, |
68 | // but use the very good perl function to parse anything else. |
90 | // but use the very good perl function to parse anything else. |
69 | // note that we never call this function for a ascii codepoints |
91 | // note that we never call this function for a ascii codepoints |
70 | static UV |
92 | inline UV |
71 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
93 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
72 | { |
94 | { |
73 | if (s[0] > 0xdf || s[0] < 0xc2) |
95 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
74 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
96 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
75 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
97 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
76 | { |
98 | { |
77 | *clen = 2; |
99 | *clen = 2; |
78 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
100 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
79 | } |
101 | } |
80 | else |
102 | else |
|
|
103 | { |
|
|
104 | *clen = (STRLEN)-1; |
81 | return (UV)-1; |
105 | return (UV)-1; |
|
|
106 | } |
82 | } |
107 | } |
83 | |
108 | |
84 | ///////////////////////////////////////////////////////////////////////////// |
109 | ///////////////////////////////////////////////////////////////////////////// |
85 | // encoder |
110 | // encoder |
86 | |
111 | |
… | |
… | |
88 | typedef struct |
113 | typedef struct |
89 | { |
114 | { |
90 | char *cur; // SvPVX (sv) + current output position |
115 | char *cur; // SvPVX (sv) + current output position |
91 | char *end; // SvEND (sv) |
116 | char *end; // SvEND (sv) |
92 | SV *sv; // result scalar |
117 | SV *sv; // result scalar |
93 | U32 flags; // F_* |
118 | JSON json; |
94 | U32 indent; // indentation level |
119 | U32 indent; // indentation level |
95 | U32 maxdepth; // max. indentation/recursion level |
120 | U32 maxdepth; // max. indentation/recursion level |
96 | } enc_t; |
121 | } enc_t; |
97 | |
122 | |
98 | static void |
123 | inline void |
99 | need (enc_t *enc, STRLEN len) |
124 | need (enc_t *enc, STRLEN len) |
100 | { |
125 | { |
101 | if (enc->cur + len >= enc->end) |
126 | if (expect_false (enc->cur + len >= enc->end)) |
102 | { |
127 | { |
103 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
128 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
104 | SvGROW (enc->sv, cur + len + 1); |
129 | SvGROW (enc->sv, cur + len + 1); |
105 | enc->cur = SvPVX (enc->sv) + cur; |
130 | enc->cur = SvPVX (enc->sv) + cur; |
106 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); |
131 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
107 | } |
132 | } |
108 | } |
133 | } |
109 | |
134 | |
110 | static void |
135 | inline void |
111 | encode_ch (enc_t *enc, char ch) |
136 | encode_ch (enc_t *enc, char ch) |
112 | { |
137 | { |
113 | need (enc, 1); |
138 | need (enc, 1); |
114 | *enc->cur++ = ch; |
139 | *enc->cur++ = ch; |
115 | } |
140 | } |
… | |
… | |
123 | |
148 | |
124 | while (str < end) |
149 | while (str < end) |
125 | { |
150 | { |
126 | unsigned char ch = *(unsigned char *)str; |
151 | unsigned char ch = *(unsigned char *)str; |
127 | |
152 | |
128 | if (ch >= 0x20 && ch < 0x80) // most common case |
153 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
129 | { |
154 | { |
130 | if (ch == '"') // but with slow exceptions |
155 | if (expect_false (ch == '"')) // but with slow exceptions |
131 | { |
156 | { |
132 | need (enc, len += 1); |
157 | need (enc, len += 1); |
133 | *enc->cur++ = '\\'; |
158 | *enc->cur++ = '\\'; |
134 | *enc->cur++ = '"'; |
159 | *enc->cur++ = '"'; |
135 | } |
160 | } |
136 | else if (ch == '\\') |
161 | else if (expect_false (ch == '\\')) |
137 | { |
162 | { |
138 | need (enc, len += 1); |
163 | need (enc, len += 1); |
139 | *enc->cur++ = '\\'; |
164 | *enc->cur++ = '\\'; |
140 | *enc->cur++ = '\\'; |
165 | *enc->cur++ = '\\'; |
141 | } |
166 | } |
… | |
… | |
159 | STRLEN clen; |
184 | STRLEN clen; |
160 | UV uch; |
185 | UV uch; |
161 | |
186 | |
162 | if (is_utf8) |
187 | if (is_utf8) |
163 | { |
188 | { |
164 | //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
|
|
165 | uch = decode_utf8 (str, end - str, &clen); |
189 | uch = decode_utf8 (str, end - str, &clen); |
166 | if (clen == (STRLEN)-1) |
190 | if (clen == (STRLEN)-1) |
167 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
191 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
168 | } |
192 | } |
169 | else |
193 | else |
… | |
… | |
173 | } |
197 | } |
174 | |
198 | |
175 | if (uch > 0x10FFFFUL) |
199 | if (uch > 0x10FFFFUL) |
176 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
200 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
177 | |
201 | |
178 | if (uch < 0x80 || enc->flags & F_ASCII) |
202 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
179 | { |
203 | { |
180 | if (uch > 0xFFFFUL) |
204 | if (uch > 0xFFFFUL) |
181 | { |
205 | { |
182 | need (enc, len += 11); |
206 | need (enc, len += 11); |
183 | sprintf (enc->cur, "\\u%04x\\u%04x", |
207 | sprintf (enc->cur, "\\u%04x\\u%04x", |
… | |
… | |
197 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
221 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
198 | } |
222 | } |
199 | |
223 | |
200 | str += clen; |
224 | str += clen; |
201 | } |
225 | } |
|
|
226 | else if (enc->json.flags & F_LATIN1) |
|
|
227 | { |
|
|
228 | *enc->cur++ = uch; |
|
|
229 | str += clen; |
|
|
230 | } |
202 | else if (is_utf8) |
231 | else if (is_utf8) |
203 | { |
232 | { |
204 | need (enc, len += clen); |
233 | need (enc, len += clen); |
205 | do |
234 | do |
206 | { |
235 | { |
… | |
… | |
208 | } |
237 | } |
209 | while (--clen); |
238 | while (--clen); |
210 | } |
239 | } |
211 | else |
240 | else |
212 | { |
241 | { |
213 | need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed |
242 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
214 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
243 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
215 | ++str; |
244 | ++str; |
216 | } |
245 | } |
217 | } |
246 | } |
218 | } |
247 | } |
… | |
… | |
220 | |
249 | |
221 | --len; |
250 | --len; |
222 | } |
251 | } |
223 | } |
252 | } |
224 | |
253 | |
225 | static void |
254 | inline void |
226 | encode_indent (enc_t *enc) |
255 | encode_indent (enc_t *enc) |
227 | { |
256 | { |
228 | if (enc->flags & F_INDENT) |
257 | if (enc->json.flags & F_INDENT) |
229 | { |
258 | { |
230 | int spaces = enc->indent * INDENT_STEP; |
259 | int spaces = enc->indent * INDENT_STEP; |
231 | |
260 | |
232 | need (enc, spaces); |
261 | need (enc, spaces); |
233 | memset (enc->cur, ' ', spaces); |
262 | memset (enc->cur, ' ', spaces); |
234 | enc->cur += spaces; |
263 | enc->cur += spaces; |
235 | } |
264 | } |
236 | } |
265 | } |
237 | |
266 | |
238 | static void |
267 | inline void |
239 | encode_space (enc_t *enc) |
268 | encode_space (enc_t *enc) |
240 | { |
269 | { |
241 | need (enc, 1); |
270 | need (enc, 1); |
242 | encode_ch (enc, ' '); |
271 | encode_ch (enc, ' '); |
243 | } |
272 | } |
244 | |
273 | |
245 | static void |
274 | inline void |
246 | encode_nl (enc_t *enc) |
275 | encode_nl (enc_t *enc) |
247 | { |
276 | { |
248 | if (enc->flags & F_INDENT) |
277 | if (enc->json.flags & F_INDENT) |
249 | { |
278 | { |
250 | need (enc, 1); |
279 | need (enc, 1); |
251 | encode_ch (enc, '\n'); |
280 | encode_ch (enc, '\n'); |
252 | } |
281 | } |
253 | } |
282 | } |
254 | |
283 | |
255 | static void |
284 | inline void |
256 | encode_comma (enc_t *enc) |
285 | encode_comma (enc_t *enc) |
257 | { |
286 | { |
258 | encode_ch (enc, ','); |
287 | encode_ch (enc, ','); |
259 | |
288 | |
260 | if (enc->flags & F_INDENT) |
289 | if (enc->json.flags & F_INDENT) |
261 | encode_nl (enc); |
290 | encode_nl (enc); |
262 | else if (enc->flags & F_SPACE_AFTER) |
291 | else if (enc->json.flags & F_SPACE_AFTER) |
263 | encode_space (enc); |
292 | encode_space (enc); |
264 | } |
293 | } |
265 | |
294 | |
266 | static void encode_sv (enc_t *enc, SV *sv); |
295 | static void encode_sv (enc_t *enc, SV *sv); |
267 | |
296 | |
268 | static void |
297 | static void |
269 | encode_av (enc_t *enc, AV *av) |
298 | encode_av (enc_t *enc, AV *av) |
270 | { |
299 | { |
271 | int i, len = av_len (av); |
300 | int i, len = av_len (av); |
|
|
301 | |
|
|
302 | if (enc->indent >= enc->maxdepth) |
|
|
303 | croak ("data structure too deep (hit recursion limit)"); |
272 | |
304 | |
273 | encode_ch (enc, '['); encode_nl (enc); |
305 | encode_ch (enc, '['); encode_nl (enc); |
274 | ++enc->indent; |
306 | ++enc->indent; |
275 | |
307 | |
276 | for (i = 0; i <= len; ++i) |
308 | for (i = 0; i <= len; ++i) |
… | |
… | |
307 | else |
339 | else |
308 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
340 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
309 | |
341 | |
310 | encode_ch (enc, '"'); |
342 | encode_ch (enc, '"'); |
311 | |
343 | |
312 | if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
344 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
313 | encode_ch (enc, ':'); |
345 | encode_ch (enc, ':'); |
314 | if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
346 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
315 | encode_sv (enc, HeVAL (he)); |
347 | encode_sv (enc, HeVAL (he)); |
316 | } |
348 | } |
317 | |
349 | |
318 | // compare hash entries, used when all keys are bytestrings |
350 | // compare hash entries, used when all keys are bytestrings |
319 | static int |
351 | static int |
… | |
… | |
342 | |
374 | |
343 | static void |
375 | static void |
344 | encode_hv (enc_t *enc, HV *hv) |
376 | encode_hv (enc_t *enc, HV *hv) |
345 | { |
377 | { |
346 | int count, i; |
378 | int count, i; |
|
|
379 | |
|
|
380 | if (enc->indent >= enc->maxdepth) |
|
|
381 | croak ("data structure too deep (hit recursion limit)"); |
347 | |
382 | |
348 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
383 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
349 | |
384 | |
350 | if ((count = hv_iterinit (hv))) |
385 | if ((count = hv_iterinit (hv))) |
351 | { |
386 | { |
352 | // for canonical output we have to sort by keys first |
387 | // for canonical output we have to sort by keys first |
353 | // actually, this is mostly due to the stupid so-called |
388 | // actually, this is mostly due to the stupid so-called |
354 | // security workaround added somewhere in 5.8.x. |
389 | // security workaround added somewhere in 5.8.x. |
355 | // that randomises hash orderings |
390 | // that randomises hash orderings |
356 | if (enc->flags & F_CANONICAL) |
391 | if (enc->json.flags & F_CANONICAL) |
357 | { |
392 | { |
358 | HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
359 | int fast = 1; |
393 | int fast = 1; |
|
|
394 | HE *he; |
|
|
395 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
396 | HE **hes = _alloca (count * sizeof (HE)); |
|
|
397 | #else |
|
|
398 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
399 | #endif |
360 | |
400 | |
361 | i = 0; |
401 | i = 0; |
362 | while ((he = hv_iternext (hv))) |
402 | while ((he = hv_iternext (hv))) |
363 | { |
403 | { |
364 | hes [i++] = he; |
404 | hes [i++] = he; |
… | |
… | |
399 | |
439 | |
400 | encode_nl (enc); |
440 | encode_nl (enc); |
401 | } |
441 | } |
402 | else |
442 | else |
403 | { |
443 | { |
404 | SV *sv; |
|
|
405 | HE *he = hv_iternext (hv); |
444 | HE *he = hv_iternext (hv); |
406 | |
445 | |
407 | for (;;) |
446 | for (;;) |
408 | { |
447 | { |
409 | encode_indent (enc); |
448 | encode_indent (enc); |
… | |
… | |
418 | encode_nl (enc); |
457 | encode_nl (enc); |
419 | } |
458 | } |
420 | } |
459 | } |
421 | |
460 | |
422 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
461 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
|
|
462 | } |
|
|
463 | |
|
|
464 | // encode objects, arrays and special \0=false and \1=true values. |
|
|
465 | static void |
|
|
466 | encode_rv (enc_t *enc, SV *sv) |
|
|
467 | { |
|
|
468 | svtype svt; |
|
|
469 | |
|
|
470 | SvGETMAGIC (sv); |
|
|
471 | svt = SvTYPE (sv); |
|
|
472 | |
|
|
473 | if (expect_false (SvOBJECT (sv))) |
|
|
474 | { |
|
|
475 | if (SvSTASH (sv) == json_boolean_stash) |
|
|
476 | { |
|
|
477 | if (SvIV (sv) == 0) |
|
|
478 | encode_str (enc, "false", 5, 0); |
|
|
479 | else |
|
|
480 | encode_str (enc, "true", 4, 0); |
|
|
481 | } |
|
|
482 | else |
|
|
483 | { |
|
|
484 | #if 0 |
|
|
485 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
|
|
486 | { |
|
|
487 | // not yet |
|
|
488 | } |
|
|
489 | #endif |
|
|
490 | if (enc->json.flags & F_CONV_BLESSED) |
|
|
491 | { |
|
|
492 | // we re-bless the reference to get overload and other niceties right |
|
|
493 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1); |
|
|
494 | |
|
|
495 | if (to_json) |
|
|
496 | { |
|
|
497 | dSP; ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
498 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
499 | |
|
|
500 | // calling with G_SCALAR ensures that we always get a 1 reutrn value |
|
|
501 | // check anyways. |
|
|
502 | PUTBACK; |
|
|
503 | assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR)); |
|
|
504 | SPAGAIN; |
|
|
505 | |
|
|
506 | encode_sv (enc, POPs); |
|
|
507 | |
|
|
508 | FREETMPS; LEAVE; |
|
|
509 | } |
|
|
510 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
511 | encode_str (enc, "null", 4, 0); |
|
|
512 | else |
|
|
513 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
514 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
515 | } |
|
|
516 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
517 | encode_str (enc, "null", 4, 0); |
|
|
518 | else |
|
|
519 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
|
|
520 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
521 | } |
|
|
522 | } |
|
|
523 | else if (svt == SVt_PVHV) |
|
|
524 | encode_hv (enc, (HV *)sv); |
|
|
525 | else if (svt == SVt_PVAV) |
|
|
526 | encode_av (enc, (AV *)sv); |
|
|
527 | else if (svt < SVt_PVAV) |
|
|
528 | { |
|
|
529 | if (SvNIOK (sv) && SvIV (sv) == 0) |
|
|
530 | encode_str (enc, "false", 5, 0); |
|
|
531 | else if (SvNIOK (sv) && SvIV (sv) == 1) |
|
|
532 | encode_str (enc, "true", 4, 0); |
|
|
533 | else |
|
|
534 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
|
|
535 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
536 | } |
|
|
537 | else |
|
|
538 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
539 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
423 | } |
540 | } |
424 | |
541 | |
425 | static void |
542 | static void |
426 | encode_sv (enc_t *enc, SV *sv) |
543 | encode_sv (enc_t *enc, SV *sv) |
427 | { |
544 | { |
… | |
… | |
435 | encode_str (enc, str, len, SvUTF8 (sv)); |
552 | encode_str (enc, str, len, SvUTF8 (sv)); |
436 | encode_ch (enc, '"'); |
553 | encode_ch (enc, '"'); |
437 | } |
554 | } |
438 | else if (SvNOKp (sv)) |
555 | else if (SvNOKp (sv)) |
439 | { |
556 | { |
|
|
557 | // trust that perl will do the right thing w.r.t. JSON syntax. |
440 | need (enc, NV_DIG + 32); |
558 | need (enc, NV_DIG + 32); |
441 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
559 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
442 | enc->cur += strlen (enc->cur); |
560 | enc->cur += strlen (enc->cur); |
443 | } |
561 | } |
444 | else if (SvIOKp (sv)) |
562 | else if (SvIOKp (sv)) |
445 | { |
563 | { |
446 | need (enc, 64); |
564 | // we assume we can always read an IV as a UV |
|
|
565 | if (SvUV (sv) & ~(UV)0x7fff) |
|
|
566 | { |
|
|
567 | // large integer, use the (rather slow) snprintf way. |
|
|
568 | need (enc, sizeof (UV) * 3); |
447 | enc->cur += |
569 | enc->cur += |
448 | SvIsUV(sv) |
570 | SvIsUV(sv) |
449 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
571 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
450 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
572 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
573 | } |
|
|
574 | else |
|
|
575 | { |
|
|
576 | // optimise the "small number case" |
|
|
577 | // code will likely be branchless and use only a single multiplication |
|
|
578 | I32 i = SvIV (sv); |
|
|
579 | U32 u; |
|
|
580 | char digit, nz = 0; |
|
|
581 | |
|
|
582 | need (enc, 6); |
|
|
583 | |
|
|
584 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
585 | u = i < 0 ? -i : i; |
|
|
586 | |
|
|
587 | // convert to 4.28 fixed-point representation |
|
|
588 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
589 | |
|
|
590 | // now output digit by digit, each time masking out the integer part |
|
|
591 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
592 | // resulting in a net multiplication by 10. |
|
|
593 | // we always write the digit to memory but conditionally increment |
|
|
594 | // the pointer, to ease the usage of conditional move instructions. |
|
|
595 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
|
|
596 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
|
|
597 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
|
|
598 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
|
|
599 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
600 | } |
451 | } |
601 | } |
452 | else if (SvROK (sv)) |
602 | else if (SvROK (sv)) |
453 | { |
603 | encode_rv (enc, SvRV (sv)); |
454 | SV *rv = SvRV (sv); |
|
|
455 | |
|
|
456 | if (enc->indent >= enc->maxdepth) |
|
|
457 | croak ("data structure too deep (hit recursion limit)"); |
|
|
458 | |
|
|
459 | switch (SvTYPE (rv)) |
|
|
460 | { |
|
|
461 | case SVt_PVAV: encode_av (enc, (AV *)rv); break; |
|
|
462 | case SVt_PVHV: encode_hv (enc, (HV *)rv); break; |
|
|
463 | |
|
|
464 | default: |
|
|
465 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
466 | SvPV_nolen (sv)); |
|
|
467 | } |
|
|
468 | } |
|
|
469 | else if (!SvOK (sv)) |
604 | else if (!SvOK (sv)) |
470 | encode_str (enc, "null", 4, 0); |
605 | encode_str (enc, "null", 4, 0); |
471 | else |
606 | else |
472 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
607 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
473 | SvPV_nolen (sv), SvFLAGS (sv)); |
608 | SvPV_nolen (sv), SvFLAGS (sv)); |
474 | } |
609 | } |
475 | |
610 | |
476 | static SV * |
611 | static SV * |
477 | encode_json (SV *scalar, U32 flags) |
612 | encode_json (SV *scalar, JSON *json) |
478 | { |
613 | { |
|
|
614 | enc_t enc; |
|
|
615 | |
479 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
616 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
480 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
617 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
481 | |
618 | |
482 | enc_t enc; |
619 | enc.json = *json; |
483 | enc.flags = flags; |
|
|
484 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
620 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
485 | enc.cur = SvPVX (enc.sv); |
621 | enc.cur = SvPVX (enc.sv); |
486 | enc.end = SvEND (enc.sv); |
622 | enc.end = SvEND (enc.sv); |
487 | enc.indent = 0; |
623 | enc.indent = 0; |
488 | enc.maxdepth = DEC_DEPTH (flags); |
624 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
489 | |
625 | |
490 | SvPOK_only (enc.sv); |
626 | SvPOK_only (enc.sv); |
491 | encode_sv (&enc, scalar); |
627 | encode_sv (&enc, scalar); |
492 | |
628 | |
|
|
629 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
630 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
631 | |
493 | if (!(flags & (F_ASCII | F_UTF8))) |
632 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
494 | SvUTF8_on (enc.sv); |
633 | SvUTF8_on (enc.sv); |
495 | |
634 | |
496 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
497 | |
|
|
498 | if (enc.flags & F_SHRINK) |
635 | if (enc.json.flags & F_SHRINK) |
499 | shrink (enc.sv); |
636 | shrink (enc.sv); |
500 | |
637 | |
501 | return enc.sv; |
638 | return enc.sv; |
502 | } |
639 | } |
503 | |
640 | |
… | |
… | |
508 | typedef struct |
645 | typedef struct |
509 | { |
646 | { |
510 | char *cur; // current parser pointer |
647 | char *cur; // current parser pointer |
511 | char *end; // end of input string |
648 | char *end; // end of input string |
512 | const char *err; // parse error, if != 0 |
649 | const char *err; // parse error, if != 0 |
513 | U32 flags; // F_* |
650 | JSON json; |
514 | U32 depth; // recursion depth |
651 | U32 depth; // recursion depth |
515 | U32 maxdepth; // recursion depth limit |
652 | U32 maxdepth; // recursion depth limit |
516 | } dec_t; |
653 | } dec_t; |
517 | |
654 | |
518 | static void |
655 | inline void |
519 | decode_ws (dec_t *dec) |
656 | decode_ws (dec_t *dec) |
520 | { |
657 | { |
521 | for (;;) |
658 | for (;;) |
522 | { |
659 | { |
523 | char ch = *dec->cur; |
660 | char ch = *dec->cur; |
… | |
… | |
549 | decode_4hex (dec_t *dec) |
686 | decode_4hex (dec_t *dec) |
550 | { |
687 | { |
551 | signed char d1, d2, d3, d4; |
688 | signed char d1, d2, d3, d4; |
552 | unsigned char *cur = (unsigned char *)dec->cur; |
689 | unsigned char *cur = (unsigned char *)dec->cur; |
553 | |
690 | |
554 | d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); |
691 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
555 | d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); |
692 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
556 | d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); |
693 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
557 | d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); |
694 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
558 | |
695 | |
559 | dec->cur += 4; |
696 | dec->cur += 4; |
560 | |
697 | |
561 | return ((UV)d1) << 12 |
698 | return ((UV)d1) << 12 |
562 | | ((UV)d2) << 8 |
699 | | ((UV)d2) << 8 |
… | |
… | |
570 | static SV * |
707 | static SV * |
571 | decode_str (dec_t *dec) |
708 | decode_str (dec_t *dec) |
572 | { |
709 | { |
573 | SV *sv = 0; |
710 | SV *sv = 0; |
574 | int utf8 = 0; |
711 | int utf8 = 0; |
|
|
712 | char *dec_cur = dec->cur; |
575 | |
713 | |
576 | do |
714 | do |
577 | { |
715 | { |
578 | char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; |
716 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
579 | char *cur = buf; |
717 | char *cur = buf; |
580 | |
718 | |
581 | do |
719 | do |
582 | { |
720 | { |
583 | unsigned char ch = *(unsigned char *)dec->cur++; |
721 | unsigned char ch = *(unsigned char *)dec_cur++; |
584 | |
722 | |
585 | if (ch == '"') |
723 | if (expect_false (ch == '"')) |
586 | { |
724 | { |
587 | --dec->cur; |
725 | --dec_cur; |
588 | break; |
726 | break; |
589 | } |
727 | } |
590 | else if (ch == '\\') |
728 | else if (expect_false (ch == '\\')) |
591 | { |
729 | { |
592 | switch (*dec->cur) |
730 | switch (*dec_cur) |
593 | { |
731 | { |
594 | case '\\': |
732 | case '\\': |
595 | case '/': |
733 | case '/': |
596 | case '"': *cur++ = *dec->cur++; break; |
734 | case '"': *cur++ = *dec_cur++; break; |
597 | |
735 | |
598 | case 'b': ++dec->cur; *cur++ = '\010'; break; |
736 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
599 | case 't': ++dec->cur; *cur++ = '\011'; break; |
737 | case 't': ++dec_cur; *cur++ = '\011'; break; |
600 | case 'n': ++dec->cur; *cur++ = '\012'; break; |
738 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
601 | case 'f': ++dec->cur; *cur++ = '\014'; break; |
739 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
602 | case 'r': ++dec->cur; *cur++ = '\015'; break; |
740 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
603 | |
741 | |
604 | case 'u': |
742 | case 'u': |
605 | { |
743 | { |
606 | UV lo, hi; |
744 | UV lo, hi; |
607 | ++dec->cur; |
745 | ++dec_cur; |
608 | |
746 | |
|
|
747 | dec->cur = dec_cur; |
609 | hi = decode_4hex (dec); |
748 | hi = decode_4hex (dec); |
|
|
749 | dec_cur = dec->cur; |
610 | if (hi == (UV)-1) |
750 | if (hi == (UV)-1) |
611 | goto fail; |
751 | goto fail; |
612 | |
752 | |
613 | // possibly a surrogate pair |
753 | // possibly a surrogate pair |
614 | if (hi >= 0xd800) |
754 | if (hi >= 0xd800) |
615 | if (hi < 0xdc00) |
755 | if (hi < 0xdc00) |
616 | { |
756 | { |
617 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
757 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
618 | ERR ("missing low surrogate character in surrogate pair"); |
758 | ERR ("missing low surrogate character in surrogate pair"); |
619 | |
759 | |
620 | dec->cur += 2; |
760 | dec_cur += 2; |
621 | |
761 | |
|
|
762 | dec->cur = dec_cur; |
622 | lo = decode_4hex (dec); |
763 | lo = decode_4hex (dec); |
|
|
764 | dec_cur = dec->cur; |
623 | if (lo == (UV)-1) |
765 | if (lo == (UV)-1) |
624 | goto fail; |
766 | goto fail; |
625 | |
767 | |
626 | if (lo < 0xdc00 || lo >= 0xe000) |
768 | if (lo < 0xdc00 || lo >= 0xe000) |
627 | ERR ("surrogate pair expected"); |
769 | ERR ("surrogate pair expected"); |
… | |
… | |
641 | *cur++ = hi; |
783 | *cur++ = hi; |
642 | } |
784 | } |
643 | break; |
785 | break; |
644 | |
786 | |
645 | default: |
787 | default: |
646 | --dec->cur; |
788 | --dec_cur; |
647 | ERR ("illegal backslash escape sequence in string"); |
789 | ERR ("illegal backslash escape sequence in string"); |
648 | } |
790 | } |
649 | } |
791 | } |
650 | else if (ch >= 0x20 && ch <= 0x7f) |
792 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
651 | *cur++ = ch; |
793 | *cur++ = ch; |
652 | else if (ch >= 0x80) |
794 | else if (ch >= 0x80) |
653 | { |
795 | { |
654 | --dec->cur; |
|
|
655 | |
|
|
656 | STRLEN clen; |
796 | STRLEN clen; |
|
|
797 | UV uch; |
|
|
798 | |
|
|
799 | --dec_cur; |
|
|
800 | |
657 | UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); |
801 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
658 | if (clen == (STRLEN)-1) |
802 | if (clen == (STRLEN)-1) |
659 | ERR ("malformed UTF-8 character in JSON string"); |
803 | ERR ("malformed UTF-8 character in JSON string"); |
660 | |
804 | |
661 | do |
805 | do |
662 | { |
|
|
663 | *cur++ = *dec->cur++; |
806 | *cur++ = *dec_cur++; |
664 | } |
|
|
665 | while (--clen); |
807 | while (--clen); |
666 | |
808 | |
667 | utf8 = 1; |
809 | utf8 = 1; |
668 | } |
810 | } |
669 | else if (!ch) |
|
|
670 | ERR ("unexpected end of string while parsing json string"); |
|
|
671 | else |
811 | else |
|
|
812 | { |
|
|
813 | --dec_cur; |
|
|
814 | |
|
|
815 | if (!ch) |
|
|
816 | ERR ("unexpected end of string while parsing JSON string"); |
|
|
817 | else |
672 | ERR ("invalid character encountered"); |
818 | ERR ("invalid character encountered while parsing JSON string"); |
673 | |
819 | } |
674 | } |
820 | } |
675 | while (cur < buf + SHORT_STRING_LEN); |
821 | while (cur < buf + SHORT_STRING_LEN); |
676 | |
822 | |
|
|
823 | { |
677 | STRLEN len = cur - buf; |
824 | STRLEN len = cur - buf; |
678 | |
825 | |
679 | if (sv) |
826 | if (sv) |
680 | { |
827 | { |
681 | SvGROW (sv, SvCUR (sv) + len + 1); |
828 | SvGROW (sv, SvCUR (sv) + len + 1); |
682 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
829 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
683 | SvCUR_set (sv, SvCUR (sv) + len); |
830 | SvCUR_set (sv, SvCUR (sv) + len); |
684 | } |
831 | } |
685 | else |
832 | else |
686 | sv = newSVpvn (buf, len); |
833 | sv = newSVpvn (buf, len); |
687 | } |
834 | } |
|
|
835 | } |
688 | while (*dec->cur != '"'); |
836 | while (*dec_cur != '"'); |
689 | |
837 | |
690 | ++dec->cur; |
838 | ++dec_cur; |
691 | |
839 | |
692 | if (sv) |
840 | if (sv) |
693 | { |
841 | { |
694 | SvPOK_only (sv); |
842 | SvPOK_only (sv); |
695 | *SvEND (sv) = 0; |
843 | *SvEND (sv) = 0; |
… | |
… | |
698 | SvUTF8_on (sv); |
846 | SvUTF8_on (sv); |
699 | } |
847 | } |
700 | else |
848 | else |
701 | sv = newSVpvn ("", 0); |
849 | sv = newSVpvn ("", 0); |
702 | |
850 | |
|
|
851 | dec->cur = dec_cur; |
703 | return sv; |
852 | return sv; |
704 | |
853 | |
705 | fail: |
854 | fail: |
|
|
855 | dec->cur = dec_cur; |
706 | return 0; |
856 | return 0; |
707 | } |
857 | } |
708 | |
858 | |
709 | static SV * |
859 | static SV * |
710 | decode_num (dec_t *dec) |
860 | decode_num (dec_t *dec) |
… | |
… | |
768 | is_nv = 1; |
918 | is_nv = 1; |
769 | } |
919 | } |
770 | |
920 | |
771 | if (!is_nv) |
921 | if (!is_nv) |
772 | { |
922 | { |
773 | UV uv; |
923 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
774 | int numtype = grok_number (start, dec->cur - start, &uv); |
924 | if (*start == '-') |
775 | if (numtype & IS_NUMBER_IN_UV) |
925 | switch (dec->cur - start) |
776 | if (numtype & IS_NUMBER_NEG) |
|
|
777 | { |
926 | { |
778 | if (uv < (UV)IV_MIN) |
927 | case 2: return newSViv (-( start [1] - '0' * 1)); |
779 | return newSViv (-(IV)uv); |
928 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
|
|
929 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
930 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
780 | } |
931 | } |
|
|
932 | else |
|
|
933 | switch (dec->cur - start) |
|
|
934 | { |
|
|
935 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
936 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
937 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
938 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
939 | } |
|
|
940 | |
|
|
941 | { |
|
|
942 | UV uv; |
|
|
943 | int numtype = grok_number (start, dec->cur - start, &uv); |
|
|
944 | if (numtype & IS_NUMBER_IN_UV) |
|
|
945 | if (numtype & IS_NUMBER_NEG) |
|
|
946 | { |
|
|
947 | if (uv < (UV)IV_MIN) |
|
|
948 | return newSViv (-(IV)uv); |
|
|
949 | } |
781 | else |
950 | else |
782 | return newSVuv (uv); |
951 | return newSVuv (uv); |
|
|
952 | |
|
|
953 | // here would likely be the place for bigint support |
783 | } |
954 | } |
|
|
955 | } |
784 | |
956 | |
|
|
957 | // if we ever support bigint or bigfloat, this is the place for bigfloat |
785 | return newSVnv (Atof (start)); |
958 | return newSVnv (Atof (start)); |
786 | |
959 | |
787 | fail: |
960 | fail: |
788 | return 0; |
961 | return 0; |
789 | } |
962 | } |
… | |
… | |
833 | } |
1006 | } |
834 | |
1007 | |
835 | static SV * |
1008 | static SV * |
836 | decode_hv (dec_t *dec) |
1009 | decode_hv (dec_t *dec) |
837 | { |
1010 | { |
|
|
1011 | SV *sv; |
838 | HV *hv = newHV (); |
1012 | HV *hv = newHV (); |
839 | |
1013 | |
840 | DEC_INC_DEPTH; |
1014 | DEC_INC_DEPTH; |
841 | decode_ws (dec); |
1015 | decode_ws (dec); |
842 | |
1016 | |
843 | if (*dec->cur == '}') |
1017 | if (*dec->cur == '}') |
844 | ++dec->cur; |
1018 | ++dec->cur; |
845 | else |
1019 | else |
846 | for (;;) |
1020 | for (;;) |
847 | { |
1021 | { |
848 | SV *key, *value; |
|
|
849 | |
|
|
850 | decode_ws (dec); EXPECT_CH ('"'); |
1022 | decode_ws (dec); EXPECT_CH ('"'); |
851 | |
1023 | |
852 | key = decode_str (dec); |
1024 | // heuristic: assume that |
853 | if (!key) |
1025 | // a) decode_str + hv_store_ent are abysmally slow. |
854 | goto fail; |
1026 | // b) most hash keys are short, simple ascii text. |
|
|
1027 | // => try to "fast-match" such strings to avoid |
|
|
1028 | // the overhead of decode_str + hv_store_ent. |
|
|
1029 | { |
|
|
1030 | SV *value; |
|
|
1031 | char *p = dec->cur; |
|
|
1032 | char *e = p + 24; // only try up to 24 bytes |
855 | |
1033 | |
856 | decode_ws (dec); EXPECT_CH (':'); |
1034 | for (;;) |
857 | |
|
|
858 | value = decode_sv (dec); |
|
|
859 | if (!value) |
|
|
860 | { |
1035 | { |
|
|
1036 | // the >= 0x80 is true on most architectures |
|
|
1037 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1038 | { |
|
|
1039 | // slow path, back up and use decode_str |
|
|
1040 | SV *key = decode_str (dec); |
|
|
1041 | if (!key) |
|
|
1042 | goto fail; |
|
|
1043 | |
|
|
1044 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1045 | |
|
|
1046 | value = decode_sv (dec); |
|
|
1047 | if (!value) |
|
|
1048 | { |
|
|
1049 | SvREFCNT_dec (key); |
|
|
1050 | goto fail; |
|
|
1051 | } |
|
|
1052 | |
|
|
1053 | hv_store_ent (hv, key, value, 0); |
861 | SvREFCNT_dec (key); |
1054 | SvREFCNT_dec (key); |
|
|
1055 | |
|
|
1056 | break; |
|
|
1057 | } |
|
|
1058 | else if (*p == '"') |
|
|
1059 | { |
|
|
1060 | // fast path, got a simple key |
|
|
1061 | char *key = dec->cur; |
|
|
1062 | int len = p - key; |
|
|
1063 | dec->cur = p + 1; |
|
|
1064 | |
|
|
1065 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1066 | |
|
|
1067 | value = decode_sv (dec); |
|
|
1068 | if (!value) |
862 | goto fail; |
1069 | goto fail; |
|
|
1070 | |
|
|
1071 | hv_store (hv, key, len, value, 0); |
|
|
1072 | |
|
|
1073 | break; |
|
|
1074 | } |
|
|
1075 | |
|
|
1076 | ++p; |
863 | } |
1077 | } |
864 | |
1078 | } |
865 | hv_store_ent (hv, key, value, 0); |
|
|
866 | SvREFCNT_dec (key); |
|
|
867 | |
1079 | |
868 | decode_ws (dec); |
1080 | decode_ws (dec); |
869 | |
1081 | |
870 | if (*dec->cur == '}') |
1082 | if (*dec->cur == '}') |
871 | { |
1083 | { |
… | |
… | |
878 | |
1090 | |
879 | ++dec->cur; |
1091 | ++dec->cur; |
880 | } |
1092 | } |
881 | |
1093 | |
882 | DEC_DEC_DEPTH; |
1094 | DEC_DEC_DEPTH; |
|
|
1095 | sv = newRV_noinc ((SV *)hv); |
|
|
1096 | |
|
|
1097 | // check filter callbacks |
|
|
1098 | if (dec->json.flags & F_HOOK) |
|
|
1099 | { |
|
|
1100 | ENTER; SAVETMPS; |
|
|
1101 | |
|
|
1102 | if (HvKEYS (hv) == 1 && dec->json.cb_sk_object) |
|
|
1103 | { |
|
|
1104 | int count; |
|
|
1105 | |
|
|
1106 | dSP; PUSHMARK (SP); |
|
|
1107 | XPUSHs (sv_2mortal (sv)); |
|
|
1108 | |
|
|
1109 | PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN; |
|
|
1110 | |
|
|
1111 | if (count == 1) |
|
|
1112 | sv = newSVsv (POPs); |
|
|
1113 | else |
|
|
1114 | SvREFCNT_inc (sv); |
|
|
1115 | } |
|
|
1116 | |
|
|
1117 | if (dec->json.cb_object) |
|
|
1118 | { |
|
|
1119 | int count; |
|
|
1120 | |
|
|
1121 | dSP; ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1122 | XPUSHs (sv_2mortal (sv)); |
|
|
1123 | |
|
|
1124 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1125 | |
|
|
1126 | if (count == 1) |
|
|
1127 | sv = newSVsv (POPs); |
|
|
1128 | else |
|
|
1129 | SvREFCNT_inc (sv); |
|
|
1130 | } |
|
|
1131 | |
|
|
1132 | FREETMPS; LEAVE; |
|
|
1133 | } |
|
|
1134 | |
883 | return newRV_noinc ((SV *)hv); |
1135 | return newRV_noinc ((SV *)hv); |
884 | |
1136 | |
885 | fail: |
1137 | fail: |
886 | SvREFCNT_dec (hv); |
1138 | SvREFCNT_dec (hv); |
887 | DEC_DEC_DEPTH; |
1139 | DEC_DEC_DEPTH; |
… | |
… | |
890 | |
1142 | |
891 | static SV * |
1143 | static SV * |
892 | decode_sv (dec_t *dec) |
1144 | decode_sv (dec_t *dec) |
893 | { |
1145 | { |
894 | decode_ws (dec); |
1146 | decode_ws (dec); |
|
|
1147 | |
|
|
1148 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1149 | // to parse anything. |
895 | switch (*dec->cur) |
1150 | switch (*dec->cur) |
896 | { |
1151 | { |
897 | case '"': ++dec->cur; return decode_str (dec); |
1152 | case '"': ++dec->cur; return decode_str (dec); |
898 | case '[': ++dec->cur; return decode_av (dec); |
1153 | case '[': ++dec->cur; return decode_av (dec); |
899 | case '{': ++dec->cur; return decode_hv (dec); |
1154 | case '{': ++dec->cur; return decode_hv (dec); |
… | |
… | |
905 | |
1160 | |
906 | case 't': |
1161 | case 't': |
907 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1162 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
908 | { |
1163 | { |
909 | dec->cur += 4; |
1164 | dec->cur += 4; |
910 | return newSViv (1); |
1165 | return SvREFCNT_inc (json_true); |
911 | } |
1166 | } |
912 | else |
1167 | else |
913 | ERR ("'true' expected"); |
1168 | ERR ("'true' expected"); |
914 | |
1169 | |
915 | break; |
1170 | break; |
916 | |
1171 | |
917 | case 'f': |
1172 | case 'f': |
918 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1173 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
919 | { |
1174 | { |
920 | dec->cur += 5; |
1175 | dec->cur += 5; |
921 | return newSViv (0); |
1176 | return SvREFCNT_inc (json_false); |
922 | } |
1177 | } |
923 | else |
1178 | else |
924 | ERR ("'false' expected"); |
1179 | ERR ("'false' expected"); |
925 | |
1180 | |
926 | break; |
1181 | break; |
… | |
… | |
935 | ERR ("'null' expected"); |
1190 | ERR ("'null' expected"); |
936 | |
1191 | |
937 | break; |
1192 | break; |
938 | |
1193 | |
939 | default: |
1194 | default: |
940 | ERR ("malformed json string, neither array, object, number, string or atom"); |
1195 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
941 | break; |
1196 | break; |
942 | } |
1197 | } |
943 | |
1198 | |
944 | fail: |
1199 | fail: |
945 | return 0; |
1200 | return 0; |
946 | } |
1201 | } |
947 | |
1202 | |
948 | static SV * |
1203 | static SV * |
949 | decode_json (SV *string, U32 flags) |
1204 | decode_json (SV *string, JSON *json, UV *offset_return) |
950 | { |
1205 | { |
|
|
1206 | dec_t dec; |
|
|
1207 | UV offset; |
951 | SV *sv; |
1208 | SV *sv; |
952 | |
1209 | |
|
|
1210 | SvGETMAGIC (string); |
|
|
1211 | SvUPGRADE (string, SVt_PV); |
|
|
1212 | |
|
|
1213 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
|
|
1214 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1215 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
|
|
1216 | |
953 | if (flags & F_UTF8) |
1217 | if (json->flags & F_UTF8) |
954 | sv_utf8_downgrade (string, 0); |
1218 | sv_utf8_downgrade (string, 0); |
955 | else |
1219 | else |
956 | sv_utf8_upgrade (string); |
1220 | sv_utf8_upgrade (string); |
957 | |
1221 | |
958 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1222 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
959 | |
1223 | |
960 | dec_t dec; |
1224 | dec.json = *json; |
961 | dec.flags = flags; |
|
|
962 | dec.cur = SvPVX (string); |
1225 | dec.cur = SvPVX (string); |
963 | dec.end = SvEND (string); |
1226 | dec.end = SvEND (string); |
964 | dec.err = 0; |
1227 | dec.err = 0; |
965 | dec.depth = 0; |
1228 | dec.depth = 0; |
966 | dec.maxdepth = DEC_DEPTH (dec.flags); |
1229 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
967 | |
1230 | |
968 | *SvEND (sv) = 0; // this shou[ld basically be a nop, too |
1231 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1232 | dec.json.flags |= F_HOOK; |
|
|
1233 | |
|
|
1234 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
969 | sv = decode_sv (&dec); |
1235 | sv = decode_sv (&dec); |
970 | |
1236 | |
|
|
1237 | if (!(offset_return || !sv)) |
|
|
1238 | { |
|
|
1239 | // check for trailing garbage |
|
|
1240 | decode_ws (&dec); |
|
|
1241 | |
|
|
1242 | if (*dec.cur) |
|
|
1243 | { |
|
|
1244 | dec.err = "garbage after JSON object"; |
|
|
1245 | SvREFCNT_dec (sv); |
|
|
1246 | sv = 0; |
|
|
1247 | } |
|
|
1248 | } |
|
|
1249 | |
|
|
1250 | if (offset_return || !sv) |
|
|
1251 | { |
|
|
1252 | offset = dec.json.flags & F_UTF8 |
|
|
1253 | ? dec.cur - SvPVX (string) |
|
|
1254 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1255 | |
|
|
1256 | if (offset_return) |
|
|
1257 | *offset_return = offset; |
|
|
1258 | } |
|
|
1259 | |
971 | if (!sv) |
1260 | if (!sv) |
972 | { |
1261 | { |
973 | IV offset = dec.flags & F_UTF8 |
|
|
974 | ? dec.cur - SvPVX (string) |
|
|
975 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
976 | SV *uni = sv_newmortal (); |
1262 | SV *uni = sv_newmortal (); |
977 | |
1263 | |
978 | // horrible hack to silence warning inside pv_uni_display |
1264 | // horrible hack to silence warning inside pv_uni_display |
979 | COP cop = *PL_curcop; |
1265 | COP cop = *PL_curcop; |
980 | cop.cop_warnings = pWARN_NONE; |
1266 | cop.cop_warnings = pWARN_NONE; |
… | |
… | |
982 | SAVEVPTR (PL_curcop); |
1268 | SAVEVPTR (PL_curcop); |
983 | PL_curcop = &cop; |
1269 | PL_curcop = &cop; |
984 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1270 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
985 | LEAVE; |
1271 | LEAVE; |
986 | |
1272 | |
987 | croak ("%s, at character offset %d (%s)", |
1273 | croak ("%s, at character offset %d [\"%s\"]", |
988 | dec.err, |
1274 | dec.err, |
989 | (int)offset, |
1275 | (int)offset, |
990 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1276 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
991 | } |
1277 | } |
992 | |
1278 | |
993 | sv = sv_2mortal (sv); |
1279 | sv = sv_2mortal (sv); |
994 | |
1280 | |
995 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1281 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
996 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1282 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
997 | |
1283 | |
998 | return sv; |
1284 | return sv; |
999 | } |
1285 | } |
1000 | |
1286 | |
… | |
… | |
1004 | MODULE = JSON::XS PACKAGE = JSON::XS |
1290 | MODULE = JSON::XS PACKAGE = JSON::XS |
1005 | |
1291 | |
1006 | BOOT: |
1292 | BOOT: |
1007 | { |
1293 | { |
1008 | int i; |
1294 | int i; |
1009 | |
|
|
1010 | memset (decode_hexdigit, 0xff, 256); |
|
|
1011 | |
1295 | |
1012 | for (i = 0; i < 256; ++i) |
1296 | for (i = 0; i < 256; ++i) |
1013 | decode_hexdigit [i] = |
1297 | decode_hexdigit [i] = |
1014 | i >= '0' && i <= '9' ? i - '0' |
1298 | i >= '0' && i <= '9' ? i - '0' |
1015 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1299 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1016 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1300 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1017 | : -1; |
1301 | : -1; |
1018 | |
1302 | |
1019 | json_stash = gv_stashpv ("JSON::XS", 1); |
1303 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1304 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
1305 | |
|
|
1306 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
|
|
1307 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1020 | } |
1308 | } |
1021 | |
1309 | |
1022 | PROTOTYPES: DISABLE |
1310 | PROTOTYPES: DISABLE |
1023 | |
1311 | |
1024 | SV *new (char *dummy) |
1312 | void new (char *klass) |
1025 | CODE: |
1313 | PPCODE: |
1026 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1314 | { |
1027 | OUTPUT: |
1315 | SV *pv = NEWSV (0, sizeof (JSON)); |
1028 | RETVAL |
1316 | SvPOK_only (pv); |
|
|
1317 | Zero (SvPVX (pv), 1, JSON); |
|
|
1318 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1319 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash))); |
|
|
1320 | } |
1029 | |
1321 | |
1030 | SV *ascii (SV *self, int enable = 1) |
1322 | void ascii (JSON *self, int enable = 1) |
1031 | ALIAS: |
1323 | ALIAS: |
1032 | ascii = F_ASCII |
1324 | ascii = F_ASCII |
|
|
1325 | latin1 = F_LATIN1 |
1033 | utf8 = F_UTF8 |
1326 | utf8 = F_UTF8 |
1034 | indent = F_INDENT |
1327 | indent = F_INDENT |
1035 | canonical = F_CANONICAL |
1328 | canonical = F_CANONICAL |
1036 | space_before = F_SPACE_BEFORE |
1329 | space_before = F_SPACE_BEFORE |
1037 | space_after = F_SPACE_AFTER |
1330 | space_after = F_SPACE_AFTER |
1038 | pretty = F_PRETTY |
1331 | pretty = F_PRETTY |
1039 | allow_nonref = F_ALLOW_NONREF |
1332 | allow_nonref = F_ALLOW_NONREF |
1040 | shrink = F_SHRINK |
1333 | shrink = F_SHRINK |
|
|
1334 | allow_blessed = F_ALLOW_BLESSED |
|
|
1335 | convert_blessed = F_CONV_BLESSED |
1041 | CODE: |
1336 | PPCODE: |
1042 | { |
1337 | { |
1043 | UV *uv = SvJSON (self); |
|
|
1044 | if (enable) |
1338 | if (enable) |
1045 | *uv |= ix; |
1339 | self->flags |= ix; |
1046 | else |
1340 | else |
1047 | *uv &= ~ix; |
1341 | self->flags &= ~ix; |
1048 | |
1342 | |
1049 | RETVAL = newSVsv (self); |
1343 | XPUSHs (ST (0)); |
1050 | } |
1344 | } |
1051 | OUTPUT: |
|
|
1052 | RETVAL |
|
|
1053 | |
1345 | |
1054 | SV *max_depth (SV *self, int max_depth = 0x80000000UL) |
1346 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1055 | CODE: |
1347 | PPCODE: |
1056 | { |
1348 | { |
1057 | UV *uv = SvJSON (self); |
|
|
1058 | UV log2 = 0; |
1349 | UV log2 = 0; |
1059 | |
1350 | |
1060 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1351 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1061 | |
1352 | |
1062 | while ((1UL << log2) < max_depth) |
1353 | while ((1UL << log2) < max_depth) |
1063 | ++log2; |
1354 | ++log2; |
1064 | |
1355 | |
1065 | *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1356 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1066 | |
1357 | |
1067 | RETVAL = newSVsv (self); |
1358 | XPUSHs (ST (0)); |
1068 | } |
1359 | } |
1069 | OUTPUT: |
|
|
1070 | RETVAL |
|
|
1071 | |
1360 | |
1072 | void encode (SV *self, SV *scalar) |
1361 | void max_size (JSON *self, UV max_size = 0) |
1073 | PPCODE: |
1362 | PPCODE: |
1074 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
1363 | { |
|
|
1364 | UV log2 = 0; |
1075 | |
1365 | |
1076 | void decode (SV *self, SV *jsonstr) |
1366 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1367 | if (max_size == 1) max_size = 2; |
|
|
1368 | |
|
|
1369 | while ((1UL << log2) < max_size) |
|
|
1370 | ++log2; |
|
|
1371 | |
|
|
1372 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1373 | |
|
|
1374 | XPUSHs (ST (0)); |
|
|
1375 | } |
|
|
1376 | |
|
|
1377 | void filter_json_objects (JSON *self, SV *cb = &PL_sv_undef) |
|
|
1378 | ALIAS: |
|
|
1379 | filter_sk_json_objects = 1 |
1077 | PPCODE: |
1380 | PPCODE: |
|
|
1381 | { |
|
|
1382 | if (!SvOK (cb)) |
|
|
1383 | cb = 0; |
|
|
1384 | |
|
|
1385 | switch (ix) |
|
|
1386 | { |
|
|
1387 | case 0: self->cb_object = cb; break; |
|
|
1388 | case 1: self->cb_sk_object = cb; break; |
|
|
1389 | } |
|
|
1390 | |
|
|
1391 | XPUSHs (ST (0)); |
|
|
1392 | } |
|
|
1393 | |
|
|
1394 | void encode (JSON *self, SV *scalar) |
|
|
1395 | PPCODE: |
|
|
1396 | XPUSHs (encode_json (scalar, self)); |
|
|
1397 | |
|
|
1398 | void decode (JSON *self, SV *jsonstr) |
|
|
1399 | PPCODE: |
1078 | XPUSHs (decode_json (jsonstr, *SvJSON (self))); |
1400 | XPUSHs (decode_json (jsonstr, self, 0)); |
|
|
1401 | |
|
|
1402 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
1403 | PPCODE: |
|
|
1404 | { |
|
|
1405 | UV offset; |
|
|
1406 | EXTEND (SP, 2); |
|
|
1407 | PUSHs (decode_json (jsonstr, self, &offset)); |
|
|
1408 | PUSHs (sv_2mortal (newSVuv (offset))); |
|
|
1409 | } |
1079 | |
1410 | |
1080 | PROTOTYPES: ENABLE |
1411 | PROTOTYPES: ENABLE |
1081 | |
1412 | |
1082 | void to_json (SV *scalar) |
1413 | void to_json (SV *scalar) |
1083 | ALIAS: |
|
|
1084 | objToJson = 0 |
|
|
1085 | PPCODE: |
1414 | PPCODE: |
|
|
1415 | { |
|
|
1416 | JSON json = { F_DEFAULT | F_UTF8 }; |
1086 | XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); |
1417 | XPUSHs (encode_json (scalar, &json)); |
|
|
1418 | } |
1087 | |
1419 | |
1088 | void from_json (SV *jsonstr) |
1420 | void from_json (SV *jsonstr) |
1089 | ALIAS: |
|
|
1090 | jsonToObj = 0 |
|
|
1091 | PPCODE: |
1421 | PPCODE: |
|
|
1422 | { |
|
|
1423 | JSON json = { F_DEFAULT | F_UTF8 }; |
1092 | XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8)); |
1424 | XPUSHs (decode_json (jsonstr, &json, 0)); |
|
|
1425 | } |
1093 | |
1426 | |