1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
|
|
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
|
|
10 | #include <float.h> |
|
|
11 | #include <inttypes.h> |
8 | |
12 | |
|
|
13 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
14 | # define snprintf _snprintf // C compilers have this in stdio.h |
|
|
15 | #endif |
|
|
16 | |
|
|
17 | // some old perls do not have this, try to make it work, no |
|
|
18 | // guarantees, though. if it breaks, you get to keep the pieces. |
|
|
19 | #ifndef UTF8_MAXBYTES |
|
|
20 | # define UTF8_MAXBYTES 13 |
|
|
21 | #endif |
|
|
22 | |
|
|
23 | // compatibility with perl <5.18 |
|
|
24 | #ifndef HvNAMELEN_get |
|
|
25 | # define HvNAMELEN_get(hv) strlen (HvNAME (hv)) |
|
|
26 | #endif |
|
|
27 | #ifndef HvNAMELEN |
|
|
28 | # define HvNAMELEN(hv) HvNAMELEN_get (hv) |
|
|
29 | #endif |
|
|
30 | #ifndef HvNAMEUTF8 |
|
|
31 | # define HvNAMEUTF8(hv) 0 |
|
|
32 | #endif |
|
|
33 | |
|
|
34 | // three extra for rounding, sign, and end of string |
|
|
35 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
|
|
36 | |
9 | #define F_ASCII 0x00000001 |
37 | #define F_ASCII 0x00000001UL |
|
|
38 | #define F_LATIN1 0x00000002UL |
10 | #define F_UTF8 0x00000002 |
39 | #define F_UTF8 0x00000004UL |
11 | #define F_INDENT 0x00000004 |
40 | #define F_INDENT 0x00000008UL |
12 | #define F_CANONICAL 0x00000008 |
41 | #define F_CANONICAL 0x00000010UL |
13 | #define F_SPACE_BEFORE 0x00000010 |
42 | #define F_SPACE_BEFORE 0x00000020UL |
14 | #define F_SPACE_AFTER 0x00000020 |
43 | #define F_SPACE_AFTER 0x00000040UL |
15 | #define F_JSON_RPC 0x00000040 |
44 | #define F_ALLOW_NONREF 0x00000100UL |
|
|
45 | #define F_SHRINK 0x00000200UL |
|
|
46 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
47 | #define F_CONV_BLESSED 0x00000800UL |
|
|
48 | #define F_RELAXED 0x00001000UL |
|
|
49 | #define F_ALLOW_UNKNOWN 0x00002000UL |
|
|
50 | #define F_ALLOW_TAGS 0x00004000UL |
|
|
51 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
16 | |
52 | |
17 | #define F_DEFAULT 0 |
53 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
18 | |
54 | |
19 | #define INIT_SIZE 32 // initial scalar size to be allocated |
55 | #define INIT_SIZE 64 // initial scalar size to be allocated |
|
|
56 | #define INDENT_STEP 3 // spaces per indentation level |
|
|
57 | |
|
|
58 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
|
|
59 | |
|
|
60 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
20 | |
61 | |
21 | #define SB do { |
62 | #define SB do { |
22 | #define SE } while (0) |
63 | #define SE } while (0) |
23 | |
64 | |
24 | static HV *json_stash; |
65 | #if __GNUC__ >= 3 |
|
|
66 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
|
|
67 | # define INLINE static inline |
|
|
68 | #else |
|
|
69 | # define expect(expr,value) (expr) |
|
|
70 | # define INLINE static |
|
|
71 | #endif |
|
|
72 | |
|
|
73 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
74 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
75 | |
|
|
76 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
77 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
78 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
79 | |
|
|
80 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
81 | |
|
|
82 | #ifdef USE_ITHREADS |
|
|
83 | # define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
84 | # define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1)) |
|
|
85 | # define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value)) |
|
|
86 | #else |
|
|
87 | # define JSON_STASH json_stash |
|
|
88 | # define BOOL_STASH bool_stash |
|
|
89 | # define GET_BOOL(value) bool_ ## value |
|
|
90 | #endif |
|
|
91 | |
|
|
92 | // the amount of HEs to allocate on the stack, when sorting keys |
|
|
93 | #define STACK_HES 64 |
|
|
94 | |
|
|
95 | static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean:: |
|
|
96 | static SV *bool_false, *bool_true; |
|
|
97 | static SV *sv_json; |
|
|
98 | |
|
|
99 | enum { |
|
|
100 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
101 | INCR_M_TFN, // inside true/false/null |
|
|
102 | INCR_M_NUM, // inside number |
|
|
103 | INCR_M_STR, // inside string |
|
|
104 | INCR_M_BS, // inside backslash |
|
|
105 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
106 | INCR_M_C1, // inside comment in other places |
|
|
107 | INCR_M_JSON // outside anything, count nesting |
|
|
108 | }; |
|
|
109 | |
|
|
110 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
111 | |
|
|
112 | typedef struct { |
|
|
113 | U32 flags; |
|
|
114 | U32 max_depth; |
|
|
115 | STRLEN max_size; |
|
|
116 | |
|
|
117 | SV *cb_object; |
|
|
118 | HV *cb_sk_object; |
|
|
119 | |
|
|
120 | // for the incremental parser |
|
|
121 | SV *incr_text; // the source text so far |
|
|
122 | STRLEN incr_pos; // the current offset into the text |
|
|
123 | int incr_nest; // {[]}-nesting level |
|
|
124 | unsigned char incr_mode; |
|
|
125 | |
|
|
126 | SV *v_false, *v_true; |
|
|
127 | } JSON; |
|
|
128 | |
|
|
129 | INLINE void |
|
|
130 | json_init (JSON *json) |
|
|
131 | { |
|
|
132 | static const JSON init = { F_ALLOW_NONREF, 512 }; |
|
|
133 | |
|
|
134 | *json = init; |
|
|
135 | } |
|
|
136 | |
|
|
137 | ///////////////////////////////////////////////////////////////////////////// |
|
|
138 | // utility functions |
|
|
139 | |
|
|
140 | INLINE SV * |
|
|
141 | get_bool (const char *name) |
|
|
142 | { |
|
|
143 | SV *sv = get_sv (name, 1); |
|
|
144 | |
|
|
145 | SvREADONLY_on (sv); |
|
|
146 | SvREADONLY_on (SvRV (sv)); |
|
|
147 | |
|
|
148 | return sv; |
|
|
149 | } |
|
|
150 | |
|
|
151 | INLINE void |
|
|
152 | shrink (SV *sv) |
|
|
153 | { |
|
|
154 | sv_utf8_downgrade (sv, 1); |
|
|
155 | |
|
|
156 | if (SvLEN (sv) > SvCUR (sv) + 1) |
|
|
157 | { |
|
|
158 | #ifdef SvPV_shrink_to_cur |
|
|
159 | SvPV_shrink_to_cur (sv); |
|
|
160 | #elif defined (SvPV_renew) |
|
|
161 | SvPV_renew (sv, SvCUR (sv) + 1); |
|
|
162 | #endif |
|
|
163 | } |
|
|
164 | } |
|
|
165 | |
|
|
166 | /* adds two STRLENs together, slow, and with paranoia */ |
|
|
167 | STRLEN |
|
|
168 | strlen_sum (STRLEN l1, STRLEN l2) |
|
|
169 | { |
|
|
170 | size_t sum = l1 + l2; |
|
|
171 | |
|
|
172 | if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum) |
|
|
173 | croak ("JSON::XS: string size overflow"); |
|
|
174 | |
|
|
175 | return sum; |
|
|
176 | } |
|
|
177 | |
|
|
178 | /* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */ |
|
|
179 | static char * |
|
|
180 | json_sv_grow (SV *sv, size_t len1, size_t len2) |
|
|
181 | { |
|
|
182 | len1 = strlen_sum (len1, len2); |
|
|
183 | len1 = strlen_sum (len1, len1 >> 1); |
|
|
184 | |
|
|
185 | if (len1 > 4096 - 24) |
|
|
186 | len1 = (len1 | 4095) - 24; |
|
|
187 | |
|
|
188 | return SvGROW (sv, len1); |
|
|
189 | } |
|
|
190 | |
|
|
191 | // decode a utf-8 character and return it, or (UV)-1 in |
|
|
192 | // case of an error. |
|
|
193 | // we special-case "safe" characters from U+80 .. U+7FF, |
|
|
194 | // but use the very good perl function to parse anything else. |
|
|
195 | // note that we never call this function for a ascii codepoints |
|
|
196 | INLINE UV |
|
|
197 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
|
|
198 | { |
|
|
199 | if (expect_true (len >= 2 |
|
|
200 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
|
|
201 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
|
|
202 | { |
|
|
203 | *clen = 2; |
|
|
204 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
|
|
205 | } |
|
|
206 | else |
|
|
207 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
|
|
208 | } |
|
|
209 | |
|
|
210 | // likewise for encoding, also never called for ascii codepoints |
|
|
211 | // this function takes advantage of this fact, although current gccs |
|
|
212 | // seem to optimise the check for >= 0x80 away anyways |
|
|
213 | INLINE unsigned char * |
|
|
214 | encode_utf8 (unsigned char *s, UV ch) |
|
|
215 | { |
|
|
216 | if (expect_false (ch < 0x000080)) |
|
|
217 | *s++ = ch; |
|
|
218 | else if (expect_true (ch < 0x000800)) |
|
|
219 | *s++ = 0xc0 | ( ch >> 6), |
|
|
220 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
221 | else if ( ch < 0x010000) |
|
|
222 | *s++ = 0xe0 | ( ch >> 12), |
|
|
223 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
224 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
225 | else if ( ch < 0x110000) |
|
|
226 | *s++ = 0xf0 | ( ch >> 18), |
|
|
227 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
228 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
229 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
230 | |
|
|
231 | return s; |
|
|
232 | } |
|
|
233 | |
|
|
234 | // convert offset pointer to character index, sv must be string |
|
|
235 | static STRLEN |
|
|
236 | ptr_to_index (SV *sv, char *offset) |
|
|
237 | { |
|
|
238 | return SvUTF8 (sv) |
|
|
239 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
240 | : offset - SvPVX (sv); |
|
|
241 | } |
|
|
242 | |
|
|
243 | ///////////////////////////////////////////////////////////////////////////// |
|
|
244 | // fp hell |
|
|
245 | |
|
|
246 | // scan a group of digits, and a trailing exponent |
|
|
247 | static void |
|
|
248 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
249 | { |
|
|
250 | UV uaccum = 0; |
|
|
251 | int eaccum = 0; |
|
|
252 | |
|
|
253 | // if we recurse too deep, skip all remaining digits |
|
|
254 | // to avoid a stack overflow attack |
|
|
255 | if (expect_false (--maxdepth <= 0)) |
|
|
256 | while (((U8)*s - '0') < 10) |
|
|
257 | ++s; |
|
|
258 | |
|
|
259 | for (;;) |
|
|
260 | { |
|
|
261 | U8 dig = (U8)*s - '0'; |
|
|
262 | |
|
|
263 | if (expect_false (dig >= 10)) |
|
|
264 | { |
|
|
265 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
266 | { |
|
|
267 | ++s; |
|
|
268 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
269 | } |
|
|
270 | else if ((dig | ' ') == 'e' - '0') |
|
|
271 | { |
|
|
272 | int exp2 = 0; |
|
|
273 | int neg = 0; |
|
|
274 | |
|
|
275 | ++s; |
|
|
276 | |
|
|
277 | if (*s == '-') |
|
|
278 | { |
|
|
279 | ++s; |
|
|
280 | neg = 1; |
|
|
281 | } |
|
|
282 | else if (*s == '+') |
|
|
283 | ++s; |
|
|
284 | |
|
|
285 | while ((dig = (U8)*s - '0') < 10) |
|
|
286 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
287 | |
|
|
288 | *expo += neg ? -exp2 : exp2; |
|
|
289 | } |
|
|
290 | |
|
|
291 | break; |
|
|
292 | } |
|
|
293 | |
|
|
294 | ++s; |
|
|
295 | |
|
|
296 | uaccum = uaccum * 10 + dig; |
|
|
297 | ++eaccum; |
|
|
298 | |
|
|
299 | // if we have too many digits, then recurse for more |
|
|
300 | // we actually do this for rather few digits |
|
|
301 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
302 | { |
|
|
303 | if (postdp) *expo -= eaccum; |
|
|
304 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
305 | if (postdp) *expo += eaccum; |
|
|
306 | |
|
|
307 | break; |
|
|
308 | } |
|
|
309 | } |
|
|
310 | |
|
|
311 | // this relies greatly on the quality of the pow () |
|
|
312 | // implementation of the platform, but a good |
|
|
313 | // implementation is hard to beat. |
|
|
314 | // (IEEE 754 conformant ones are required to be exact) |
|
|
315 | if (postdp) *expo -= eaccum; |
|
|
316 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
317 | *expo += eaccum; |
|
|
318 | } |
|
|
319 | |
|
|
320 | static NV |
|
|
321 | json_atof (const char *s) |
|
|
322 | { |
|
|
323 | NV accum = 0.; |
|
|
324 | int expo = 0; |
|
|
325 | int neg = 0; |
|
|
326 | |
|
|
327 | if (*s == '-') |
|
|
328 | { |
|
|
329 | ++s; |
|
|
330 | neg = 1; |
|
|
331 | } |
|
|
332 | |
|
|
333 | // a recursion depth of ten gives us >>500 bits |
|
|
334 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
335 | |
|
|
336 | return neg ? -accum : accum; |
|
|
337 | } |
|
|
338 | |
|
|
339 | // target of scalar reference is bool? -1 == nope, 0 == false, 1 == true |
|
|
340 | static int |
|
|
341 | ref_bool_type (SV *sv) |
|
|
342 | { |
|
|
343 | svtype svt = SvTYPE (sv); |
|
|
344 | |
|
|
345 | if (svt < SVt_PVAV) |
|
|
346 | { |
|
|
347 | STRLEN len = 0; |
|
|
348 | char *pv = svt ? SvPV (sv, len) : 0; |
|
|
349 | |
|
|
350 | if (len == 1) |
|
|
351 | if (*pv == '1') |
|
|
352 | return 1; |
|
|
353 | else if (*pv == '0') |
|
|
354 | return 0; |
|
|
355 | } |
|
|
356 | |
|
|
357 | return -1; |
|
|
358 | } |
|
|
359 | |
|
|
360 | // returns whether scalar is not a reference in the sense of allow_nonref |
|
|
361 | static int |
|
|
362 | json_nonref (SV *scalar) |
|
|
363 | { |
|
|
364 | if (!SvROK (scalar)) |
|
|
365 | return 1; |
|
|
366 | |
|
|
367 | scalar = SvRV (scalar); |
|
|
368 | |
|
|
369 | if (SvTYPE (scalar) >= SVt_PVMG) |
|
|
370 | { |
|
|
371 | if (SvSTASH (scalar) == bool_stash) |
|
|
372 | return 1; |
|
|
373 | |
|
|
374 | if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0) |
|
|
375 | return 1; |
|
|
376 | } |
|
|
377 | |
|
|
378 | return 0; |
|
|
379 | } |
|
|
380 | |
|
|
381 | ///////////////////////////////////////////////////////////////////////////// |
|
|
382 | // encoder |
25 | |
383 | |
26 | // structure used for encoding JSON |
384 | // structure used for encoding JSON |
27 | typedef struct |
385 | typedef struct |
28 | { |
386 | { |
29 | char *cur; |
387 | char *cur; // SvPVX (sv) + current output position |
30 | STRLEN len; // SvLEN (sv) |
|
|
31 | char *end; // SvEND (sv) |
388 | char *end; // SvEND (sv) |
32 | SV *sv; |
389 | SV *sv; // result scalar |
33 | UV flags; |
390 | JSON json; |
34 | int max_recurse; |
391 | U32 indent; // indentation level |
35 | int indent; |
392 | UV limit; // escape character values >= this value when encoding |
36 | } enc_t; |
393 | } enc_t; |
37 | |
394 | |
38 | // structure used for decoding JSON |
395 | INLINE void |
39 | typedef struct |
|
|
40 | { |
|
|
41 | char *cur; |
|
|
42 | char *end; |
|
|
43 | char *err; |
|
|
44 | UV flags; |
|
|
45 | } dec_t; |
|
|
46 | |
|
|
47 | static UV * |
|
|
48 | SvJSON (SV *sv) |
|
|
49 | { |
|
|
50 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
|
|
51 | croak ("object is not of type JSON::XS"); |
|
|
52 | |
|
|
53 | return &SvUVX (SvRV (sv)); |
|
|
54 | } |
|
|
55 | |
|
|
56 | ///////////////////////////////////////////////////////////////////////////// |
|
|
57 | |
|
|
58 | static void |
|
|
59 | need (enc_t *enc, STRLEN len) |
396 | need (enc_t *enc, STRLEN len) |
60 | { |
397 | { |
61 | if (enc->cur + len >= enc->end) |
398 | if (expect_false ((uintptr_t)(enc->end - enc->cur) < len)) |
62 | { |
399 | { |
63 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
400 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
64 | SvGROW (enc->sv, cur + len + 1); |
401 | char *buf = json_sv_grow (enc->sv, cur, len); |
65 | enc->cur = SvPVX (enc->sv) + cur; |
402 | enc->cur = buf + cur; |
66 | enc->end = SvEND (enc->sv); |
403 | enc->end = buf + SvLEN (enc->sv) - 1; |
67 | } |
404 | } |
68 | } |
405 | } |
69 | |
406 | |
70 | static void |
407 | INLINE void |
71 | encode_ch (enc_t *enc, char ch) |
408 | encode_ch (enc_t *enc, char ch) |
72 | { |
409 | { |
73 | need (enc, 1); |
410 | need (enc, 1); |
74 | *enc->cur++ = ch; |
411 | *enc->cur++ = ch; |
75 | } |
412 | } |
… | |
… | |
77 | static void |
414 | static void |
78 | encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
415 | encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
79 | { |
416 | { |
80 | char *end = str + len; |
417 | char *end = str + len; |
81 | |
418 | |
|
|
419 | need (enc, len); |
|
|
420 | |
82 | while (str < end) |
421 | while (str < end) |
83 | { |
422 | { |
84 | unsigned char ch = *(unsigned char *)str; |
423 | unsigned char ch = *(unsigned char *)str; |
|
|
424 | |
85 | if (ch >= 0x20 && ch < 0x80) // most common case |
425 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
86 | { |
426 | { |
|
|
427 | if (expect_false (ch == '"')) // but with slow exceptions |
|
|
428 | { |
|
|
429 | need (enc, len + 1); |
|
|
430 | *enc->cur++ = '\\'; |
|
|
431 | *enc->cur++ = '"'; |
|
|
432 | } |
|
|
433 | else if (expect_false (ch == '\\')) |
|
|
434 | { |
|
|
435 | need (enc, len + 1); |
|
|
436 | *enc->cur++ = '\\'; |
|
|
437 | *enc->cur++ = '\\'; |
|
|
438 | } |
|
|
439 | else |
87 | *enc->cur++ = ch; |
440 | *enc->cur++ = ch; |
|
|
441 | |
88 | str++; |
442 | ++str; |
89 | } |
443 | } |
90 | else |
444 | else |
91 | { |
445 | { |
92 | STRLEN clen; |
446 | switch (ch) |
93 | UV uch; |
|
|
94 | |
|
|
95 | if (is_utf8) |
|
|
96 | { |
447 | { |
97 | uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
448 | case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
98 | if (clen < 0) |
449 | case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
99 | croak ("malformed UTF-8 character in string, cannot convert to JSON"); |
450 | case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
100 | } |
451 | case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
101 | else |
452 | case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
102 | { |
|
|
103 | uch = ch; |
|
|
104 | clen = 1; |
|
|
105 | } |
|
|
106 | |
453 | |
107 | need (enc, len += 6); |
454 | default: |
108 | |
|
|
109 | if (uch < 0xa0 || enc->flags & F_ASCII) |
|
|
110 | { |
|
|
111 | if (uch > 0xFFFFUL) |
|
|
112 | { |
455 | { |
|
|
456 | STRLEN clen; |
|
|
457 | UV uch; |
|
|
458 | |
|
|
459 | if (is_utf8) |
|
|
460 | { |
|
|
461 | uch = decode_utf8 (str, end - str, &clen); |
|
|
462 | if (clen == (STRLEN)-1) |
|
|
463 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
|
|
464 | } |
|
|
465 | else |
|
|
466 | { |
|
|
467 | uch = ch; |
113 | len += 6; |
468 | clen = 1; |
|
|
469 | } |
|
|
470 | |
|
|
471 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
|
|
472 | { |
|
|
473 | if (uch >= 0x10000UL) |
|
|
474 | { |
|
|
475 | if (uch >= 0x110000UL) |
|
|
476 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
477 | |
114 | need (enc, len += 6); |
478 | need (enc, len + 11); |
115 | sprintf (enc->cur, "\\u%04x\\u%04x", |
479 | sprintf (enc->cur, "\\u%04x\\u%04x", |
116 | (uch - 0x10000) / 0x400 + 0xD800, |
480 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
117 | (uch - 0x10000) % 0x400 + 0xDC00); |
481 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
118 | enc->cur += 12; |
482 | enc->cur += 12; |
|
|
483 | } |
|
|
484 | else |
|
|
485 | { |
|
|
486 | need (enc, len + 5); |
|
|
487 | *enc->cur++ = '\\'; |
|
|
488 | *enc->cur++ = 'u'; |
|
|
489 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
|
|
490 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
|
|
491 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
|
|
492 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
|
|
493 | } |
|
|
494 | |
|
|
495 | str += clen; |
119 | } |
496 | } |
|
|
497 | else if (enc->json.flags & F_LATIN1) |
|
|
498 | { |
|
|
499 | *enc->cur++ = uch; |
|
|
500 | str += clen; |
|
|
501 | } |
|
|
502 | else if (is_utf8) |
|
|
503 | { |
|
|
504 | need (enc, len + clen); |
|
|
505 | do |
|
|
506 | { |
|
|
507 | *enc->cur++ = *str++; |
|
|
508 | } |
|
|
509 | while (--clen); |
|
|
510 | } |
120 | else |
511 | else |
121 | { |
512 | { |
122 | sprintf (enc->cur, "\\u%04x", uch); |
513 | need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
123 | enc->cur += 6; |
514 | enc->cur = encode_utf8 (enc->cur, uch); |
|
|
515 | ++str; |
|
|
516 | } |
124 | } |
517 | } |
125 | } |
518 | } |
126 | else if (is_utf8) |
|
|
127 | { |
|
|
128 | memcpy (enc->cur, str, clen); |
|
|
129 | enc->cur += clen; |
|
|
130 | } |
|
|
131 | else |
|
|
132 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
|
|
133 | |
|
|
134 | str += clen; |
|
|
135 | } |
519 | } |
136 | |
520 | |
137 | --len; |
521 | --len; |
138 | } |
522 | } |
139 | } |
523 | } |
140 | |
524 | |
141 | #define INDENT SB \ |
525 | INLINE void |
|
|
526 | encode_indent (enc_t *enc) |
|
|
527 | { |
142 | if (enc->flags & F_INDENT) \ |
528 | if (enc->json.flags & F_INDENT) |
143 | { \ |
529 | { |
144 | int i_; \ |
530 | int spaces = enc->indent * INDENT_STEP; |
145 | need (enc, enc->indent); \ |
531 | |
146 | for (i_ = enc->indent * 3; i_--; )\ |
532 | need (enc, spaces); |
|
|
533 | memset (enc->cur, ' ', spaces); |
|
|
534 | enc->cur += spaces; |
|
|
535 | } |
|
|
536 | } |
|
|
537 | |
|
|
538 | INLINE void |
|
|
539 | encode_space (enc_t *enc) |
|
|
540 | { |
|
|
541 | need (enc, 1); |
|
|
542 | encode_ch (enc, ' '); |
|
|
543 | } |
|
|
544 | |
|
|
545 | INLINE void |
|
|
546 | encode_nl (enc_t *enc) |
|
|
547 | { |
|
|
548 | if (enc->json.flags & F_INDENT) |
|
|
549 | { |
|
|
550 | need (enc, 1); |
147 | encode_ch (enc, ' '); \ |
551 | encode_ch (enc, '\n'); |
148 | } \ |
552 | } |
149 | SE |
553 | } |
150 | |
554 | |
151 | #define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE |
555 | INLINE void |
152 | #define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE |
556 | encode_comma (enc_t *enc) |
153 | #define COMMA SB \ |
557 | { |
154 | encode_ch (enc, ','); \ |
558 | encode_ch (enc, ','); |
|
|
559 | |
155 | if (enc->flags & F_INDENT) \ |
560 | if (enc->json.flags & F_INDENT) |
156 | NL; \ |
561 | encode_nl (enc); |
157 | else if (enc->flags & F_SPACE_AFTER) \ |
562 | else if (enc->json.flags & F_SPACE_AFTER) |
158 | SPACE; \ |
563 | encode_space (enc); |
159 | SE |
564 | } |
160 | |
565 | |
161 | static void encode_sv (enc_t *enc, SV *sv); |
566 | static void encode_sv (enc_t *enc, SV *sv); |
162 | |
567 | |
163 | static void |
568 | static void |
164 | encode_av (enc_t *enc, AV *av) |
569 | encode_av (enc_t *enc, AV *av) |
165 | { |
570 | { |
166 | int i, len = av_len (av); |
571 | int i, len = av_len (av); |
167 | |
572 | |
|
|
573 | if (enc->indent >= enc->json.max_depth) |
|
|
574 | croak (ERR_NESTING_EXCEEDED); |
|
|
575 | |
168 | encode_ch (enc, '['); NL; |
576 | encode_ch (enc, '['); |
169 | ++enc->indent; |
|
|
170 | |
577 | |
|
|
578 | if (len >= 0) |
|
|
579 | { |
|
|
580 | encode_nl (enc); ++enc->indent; |
|
|
581 | |
171 | for (i = 0; i <= len; ++i) |
582 | for (i = 0; i <= len; ++i) |
172 | { |
583 | { |
173 | INDENT; |
584 | SV **svp = av_fetch (av, i, 0); |
174 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
175 | |
585 | |
|
|
586 | encode_indent (enc); |
|
|
587 | |
|
|
588 | if (svp) |
|
|
589 | encode_sv (enc, *svp); |
|
|
590 | else |
|
|
591 | encode_str (enc, "null", 4, 0); |
|
|
592 | |
176 | if (i < len) |
593 | if (i < len) |
177 | COMMA; |
594 | encode_comma (enc); |
178 | } |
595 | } |
179 | |
596 | |
180 | NL; |
597 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
598 | } |
181 | |
599 | |
182 | --enc->indent; |
|
|
183 | INDENT; encode_ch (enc, ']'); |
600 | encode_ch (enc, ']'); |
184 | } |
601 | } |
185 | |
602 | |
186 | static void |
603 | static void |
187 | encode_he (enc_t *enc, HE *he) |
604 | encode_hk (enc_t *enc, HE *he) |
188 | { |
605 | { |
189 | encode_ch (enc, '"'); |
606 | encode_ch (enc, '"'); |
190 | |
607 | |
191 | if (HeKLEN (he) == HEf_SVKEY) |
608 | if (HeKLEN (he) == HEf_SVKEY) |
192 | { |
609 | { |
193 | SV *sv = HeSVKEY (he); |
610 | SV *sv = HeSVKEY (he); |
194 | STRLEN len; |
611 | STRLEN len; |
|
|
612 | char *str; |
|
|
613 | |
|
|
614 | SvGETMAGIC (sv); |
195 | char *str = SvPV (sv, len); |
615 | str = SvPV (sv, len); |
196 | |
616 | |
197 | encode_str (enc, str, len, SvUTF8 (sv)); |
617 | encode_str (enc, str, len, SvUTF8 (sv)); |
198 | } |
618 | } |
199 | else |
619 | else |
200 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
620 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
201 | |
621 | |
202 | encode_ch (enc, '"'); |
622 | encode_ch (enc, '"'); |
203 | |
623 | |
204 | if (enc->flags & F_SPACE_BEFORE) SPACE; |
624 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
205 | encode_ch (enc, ':'); |
625 | encode_ch (enc, ':'); |
206 | if (enc->flags & F_SPACE_AFTER ) SPACE; |
626 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
207 | encode_sv (enc, HeVAL (he)); |
|
|
208 | } |
627 | } |
209 | |
628 | |
210 | // compare hash entries, used when all keys are bytestrings |
629 | // compare hash entries, used when all keys are bytestrings |
211 | static int |
630 | static int |
212 | he_cmp_fast (const void *a_, const void *b_) |
631 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
217 | HE *b = *(HE **)b_; |
636 | HE *b = *(HE **)b_; |
218 | |
637 | |
219 | STRLEN la = HeKLEN (a); |
638 | STRLEN la = HeKLEN (a); |
220 | STRLEN lb = HeKLEN (b); |
639 | STRLEN lb = HeKLEN (b); |
221 | |
640 | |
222 | if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
641 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
223 | cmp = la < lb ? -1 : la == lb ? 0 : 1; |
642 | cmp = lb - la; |
224 | |
643 | |
225 | return cmp; |
644 | return cmp; |
226 | } |
645 | } |
227 | |
646 | |
228 | // compare hash entries, used when some keys are sv's or utf-x |
647 | // compare hash entries, used when some keys are sv's or utf-x |
229 | static int |
648 | static int |
230 | he_cmp_slow (const void *a, const void *b) |
649 | he_cmp_slow (const void *a, const void *b) |
231 | { |
650 | { |
232 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
651 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
233 | } |
652 | } |
234 | |
653 | |
235 | static void |
654 | static void |
236 | encode_hv (enc_t *enc, HV *hv) |
655 | encode_hv (enc_t *enc, HV *hv) |
237 | { |
656 | { |
238 | int count, i; |
657 | HE *he; |
239 | |
658 | |
240 | encode_ch (enc, '{'); NL; ++enc->indent; |
659 | if (enc->indent >= enc->json.max_depth) |
|
|
660 | croak (ERR_NESTING_EXCEEDED); |
241 | |
661 | |
242 | if ((count = hv_iterinit (hv))) |
662 | encode_ch (enc, '{'); |
243 | { |
663 | |
244 | // for canonical output we have to sort by keys first |
664 | // for canonical output we have to sort by keys first |
245 | // actually, this is mostly due to the stupid so-called |
665 | // actually, this is mostly due to the stupid so-called |
246 | // security workaround added somewhere in 5.8.x. |
666 | // security workaround added somewhere in 5.8.x |
247 | // that randomises hash orderings |
667 | // that randomises hash orderings |
248 | if (enc->flags & F_CANONICAL) |
668 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
669 | { |
|
|
670 | int count = hv_iterinit (hv); |
|
|
671 | |
|
|
672 | if (SvMAGICAL (hv)) |
249 | { |
673 | { |
250 | HE *he, *hes [count]; |
674 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
675 | // but I don't care for the speed of this special case. |
|
|
676 | // note also that we will run into undefined behaviour when the two iterations |
|
|
677 | // do not result in the same count, something I might care for in some later release. |
|
|
678 | |
|
|
679 | count = 0; |
|
|
680 | while (hv_iternext (hv)) |
|
|
681 | ++count; |
|
|
682 | |
|
|
683 | hv_iterinit (hv); |
|
|
684 | } |
|
|
685 | |
|
|
686 | if (count) |
|
|
687 | { |
251 | int fast = 1; |
688 | int i, fast = 1; |
|
|
689 | HE *hes_stack [STACK_HES]; |
|
|
690 | HE **hes = hes_stack; |
|
|
691 | |
|
|
692 | // allocate larger arrays on the heap |
|
|
693 | if (count > STACK_HES) |
|
|
694 | { |
|
|
695 | SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes))); |
|
|
696 | hes = (HE **)SvPVX (sv); |
|
|
697 | } |
252 | |
698 | |
253 | i = 0; |
699 | i = 0; |
254 | while ((he = hv_iternext (hv))) |
700 | while ((he = hv_iternext (hv))) |
255 | { |
701 | { |
256 | hes [i++] = he; |
702 | hes [i++] = he; |
… | |
… | |
262 | |
708 | |
263 | if (fast) |
709 | if (fast) |
264 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
710 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
265 | else |
711 | else |
266 | { |
712 | { |
267 | // hack to disable "use bytes" |
713 | // hack to forcefully disable "use bytes" |
268 | COP *oldcop = PL_curcop, cop; |
714 | COP cop = *PL_curcop; |
269 | cop.op_private = 0; |
715 | cop.op_private = 0; |
|
|
716 | |
|
|
717 | ENTER; |
|
|
718 | SAVETMPS; |
|
|
719 | |
|
|
720 | SAVEVPTR (PL_curcop); |
270 | PL_curcop = &cop; |
721 | PL_curcop = &cop; |
271 | |
722 | |
272 | SAVETMPS; |
|
|
273 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
723 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
|
|
724 | |
274 | FREETMPS; |
725 | FREETMPS; |
275 | |
726 | LEAVE; |
276 | PL_curcop = oldcop; |
|
|
277 | } |
727 | } |
278 | |
728 | |
279 | for (i = 0; i < count; ++i) |
729 | encode_nl (enc); ++enc->indent; |
|
|
730 | |
|
|
731 | while (count--) |
280 | { |
732 | { |
281 | INDENT; |
733 | encode_indent (enc); |
|
|
734 | he = hes [count]; |
282 | encode_he (enc, hes [i]); |
735 | encode_hk (enc, he); |
|
|
736 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
283 | |
737 | |
284 | if (i < count - 1) |
738 | if (count) |
285 | COMMA; |
739 | encode_comma (enc); |
286 | } |
740 | } |
287 | |
741 | |
288 | NL; |
742 | encode_nl (enc); --enc->indent; encode_indent (enc); |
289 | } |
743 | } |
|
|
744 | } |
|
|
745 | else |
|
|
746 | { |
|
|
747 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
748 | if ((he = hv_iternext (hv))) |
|
|
749 | { |
|
|
750 | encode_nl (enc); ++enc->indent; |
|
|
751 | |
|
|
752 | for (;;) |
|
|
753 | { |
|
|
754 | encode_indent (enc); |
|
|
755 | encode_hk (enc, he); |
|
|
756 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
|
|
757 | |
|
|
758 | if (!(he = hv_iternext (hv))) |
|
|
759 | break; |
|
|
760 | |
|
|
761 | encode_comma (enc); |
|
|
762 | } |
|
|
763 | |
|
|
764 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
765 | } |
|
|
766 | } |
|
|
767 | |
|
|
768 | encode_ch (enc, '}'); |
|
|
769 | } |
|
|
770 | |
|
|
771 | // encode objects, arrays and special \0=false and \1=true values. |
|
|
772 | static void |
|
|
773 | encode_rv (enc_t *enc, SV *sv) |
|
|
774 | { |
|
|
775 | svtype svt; |
|
|
776 | GV *method; |
|
|
777 | |
|
|
778 | SvGETMAGIC (sv); |
|
|
779 | svt = SvTYPE (sv); |
|
|
780 | |
|
|
781 | if (expect_false (SvOBJECT (sv))) |
|
|
782 | { |
|
|
783 | HV *stash = SvSTASH (sv); |
|
|
784 | |
|
|
785 | if (stash == bool_stash) |
|
|
786 | { |
|
|
787 | if (SvIV (sv)) encode_str (enc, "true" , 4, 0); |
|
|
788 | else encode_str (enc, "false", 5, 0); |
|
|
789 | } |
|
|
790 | else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0))) |
|
|
791 | { |
|
|
792 | int count; |
|
|
793 | dSP; |
|
|
794 | |
|
|
795 | ENTER; SAVETMPS; |
|
|
796 | PUSHMARK (SP); |
|
|
797 | EXTEND (SP, 2); |
|
|
798 | // we re-bless the reference to get overload and other niceties right |
|
|
799 | PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
|
|
800 | PUSHs (sv_json); |
|
|
801 | |
|
|
802 | PUTBACK; |
|
|
803 | count = call_sv ((SV *)GvCV (method), G_ARRAY); |
|
|
804 | SPAGAIN; |
|
|
805 | |
|
|
806 | // catch this surprisingly common error |
|
|
807 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
808 | croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
809 | |
|
|
810 | encode_ch (enc, '('); |
|
|
811 | encode_ch (enc, '"'); |
|
|
812 | encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash)); |
|
|
813 | encode_ch (enc, '"'); |
|
|
814 | encode_ch (enc, ')'); |
|
|
815 | encode_ch (enc, '['); |
|
|
816 | |
|
|
817 | if (count) |
|
|
818 | { |
|
|
819 | int i; |
|
|
820 | |
|
|
821 | for (i = 0; i < count - 1; ++i) |
|
|
822 | { |
|
|
823 | encode_sv (enc, SP[i + 1 - count]); |
|
|
824 | encode_ch (enc, ','); |
|
|
825 | } |
|
|
826 | |
|
|
827 | encode_sv (enc, TOPs); |
|
|
828 | SP -= count; |
|
|
829 | } |
|
|
830 | |
|
|
831 | encode_ch (enc, ']'); |
|
|
832 | |
|
|
833 | FREETMPS; LEAVE; |
|
|
834 | } |
|
|
835 | else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0))) |
|
|
836 | { |
|
|
837 | dSP; |
|
|
838 | |
|
|
839 | ENTER; SAVETMPS; |
|
|
840 | PUSHMARK (SP); |
|
|
841 | // we re-bless the reference to get overload and other niceties right |
|
|
842 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
|
|
843 | |
|
|
844 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
845 | PUTBACK; |
|
|
846 | call_sv ((SV *)GvCV (method), G_SCALAR); |
|
|
847 | SPAGAIN; |
|
|
848 | |
|
|
849 | // catch this surprisingly common error |
|
|
850 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
851 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
852 | |
|
|
853 | sv = POPs; |
|
|
854 | PUTBACK; |
|
|
855 | |
|
|
856 | encode_sv (enc, sv); |
|
|
857 | |
|
|
858 | FREETMPS; LEAVE; |
|
|
859 | } |
|
|
860 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
861 | encode_str (enc, "null", 4, 0); |
290 | else |
862 | else |
291 | { |
863 | croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)", |
292 | SV *sv; |
864 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
293 | HE *he = hv_iternext (hv); |
865 | } |
294 | |
866 | else if (svt == SVt_PVHV) |
295 | for (;;) |
867 | encode_hv (enc, (HV *)sv); |
296 | { |
868 | else if (svt == SVt_PVAV) |
297 | INDENT; |
869 | encode_av (enc, (AV *)sv); |
298 | encode_he (enc, he); |
870 | else if (svt < SVt_PVAV) |
299 | |
|
|
300 | if (!(he = hv_iternext (hv))) |
|
|
301 | break; |
|
|
302 | |
|
|
303 | COMMA; |
|
|
304 | } |
|
|
305 | |
|
|
306 | NL; |
|
|
307 | } |
|
|
308 | } |
871 | { |
|
|
872 | int bool_type = ref_bool_type (sv); |
309 | |
873 | |
310 | --enc->indent; INDENT; encode_ch (enc, '}'); |
874 | if (bool_type == 1) |
|
|
875 | encode_str (enc, "true", 4, 0); |
|
|
876 | else if (bool_type == 0) |
|
|
877 | encode_str (enc, "false", 5, 0); |
|
|
878 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
879 | encode_str (enc, "null", 4, 0); |
|
|
880 | else |
|
|
881 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
|
|
882 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
883 | } |
|
|
884 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
885 | encode_str (enc, "null", 4, 0); |
|
|
886 | else |
|
|
887 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
888 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
311 | } |
889 | } |
312 | |
890 | |
313 | static void |
891 | static void |
314 | encode_sv (enc_t *enc, SV *sv) |
892 | encode_sv (enc_t *enc, SV *sv) |
315 | { |
893 | { |
|
|
894 | SvGETMAGIC (sv); |
|
|
895 | |
316 | if (SvPOKp (sv)) |
896 | if (SvPOKp (sv)) |
317 | { |
897 | { |
318 | STRLEN len; |
898 | STRLEN len; |
319 | char *str = SvPV (sv, len); |
899 | char *str = SvPV (sv, len); |
320 | encode_ch (enc, '"'); |
900 | encode_ch (enc, '"'); |
321 | encode_str (enc, str, len, SvUTF8 (sv)); |
901 | encode_str (enc, str, len, SvUTF8 (sv)); |
322 | encode_ch (enc, '"'); |
902 | encode_ch (enc, '"'); |
323 | } |
903 | } |
324 | else if (SvNOKp (sv)) |
904 | else if (SvNOKp (sv)) |
325 | { |
905 | { |
|
|
906 | // trust that perl will do the right thing w.r.t. JSON syntax. |
326 | need (enc, NV_DIG + 32); |
907 | need (enc, NV_DIG + 32); |
327 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
908 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
328 | enc->cur += strlen (enc->cur); |
909 | enc->cur += strlen (enc->cur); |
329 | } |
910 | } |
330 | else if (SvIOKp (sv)) |
911 | else if (SvIOKp (sv)) |
331 | { |
912 | { |
|
|
913 | // we assume we can always read an IV as a UV and vice versa |
|
|
914 | // we assume two's complement |
|
|
915 | // we assume no aliasing issues in the union |
|
|
916 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
|
|
917 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
|
|
918 | { |
|
|
919 | // optimise the "small number case" |
|
|
920 | // code will likely be branchless and use only a single multiplication |
|
|
921 | // works for numbers up to 59074 |
|
|
922 | I32 i = SvIVX (sv); |
|
|
923 | U32 u; |
|
|
924 | char digit, nz = 0; |
|
|
925 | |
332 | need (enc, 64); |
926 | need (enc, 6); |
|
|
927 | |
|
|
928 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
929 | u = i < 0 ? -i : i; |
|
|
930 | |
|
|
931 | // convert to 4.28 fixed-point representation |
|
|
932 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
933 | |
|
|
934 | // now output digit by digit, each time masking out the integer part |
|
|
935 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
936 | // resulting in a net multiplication by 10. |
|
|
937 | // we always write the digit to memory but conditionally increment |
|
|
938 | // the pointer, to enable the use of conditional move instructions. |
|
|
939 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
|
|
940 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
|
|
941 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
|
|
942 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
|
|
943 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
944 | } |
|
|
945 | else |
|
|
946 | { |
|
|
947 | // large integer, use the (rather slow) snprintf way. |
|
|
948 | need (enc, IVUV_MAXCHARS); |
333 | enc->cur += |
949 | enc->cur += |
334 | SvIsUV(sv) |
950 | SvIsUV(sv) |
335 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
951 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
336 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
952 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
953 | } |
337 | } |
954 | } |
338 | else if (SvROK (sv)) |
955 | else if (SvROK (sv)) |
339 | { |
956 | encode_rv (enc, SvRV (sv)); |
340 | if (!--enc->max_recurse) |
957 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
341 | croak ("data structure too deep (hit recursion limit)"); |
|
|
342 | |
|
|
343 | sv = SvRV (sv); |
|
|
344 | |
|
|
345 | switch (SvTYPE (sv)) |
|
|
346 | { |
|
|
347 | case SVt_PVAV: encode_av (enc, (AV *)sv); break; |
|
|
348 | case SVt_PVHV: encode_hv (enc, (HV *)sv); break; |
|
|
349 | |
|
|
350 | default: |
|
|
351 | croak ("JSON can only represent references to arrays or hashes"); |
|
|
352 | } |
|
|
353 | } |
|
|
354 | else if (!SvOK (sv)) |
|
|
355 | encode_str (enc, "null", 4, 0); |
958 | encode_str (enc, "null", 4, 0); |
356 | else |
959 | else |
357 | croak ("encountered perl type that JSON cannot handle"); |
960 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data", |
|
|
961 | SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
358 | } |
962 | } |
359 | |
963 | |
360 | static SV * |
964 | static SV * |
361 | encode_json (SV *scalar, UV flags) |
965 | encode_json (SV *scalar, JSON *json) |
362 | { |
966 | { |
363 | enc_t enc; |
967 | enc_t enc; |
364 | enc.flags = flags; |
968 | |
|
|
969 | if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar)) |
|
|
970 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
|
|
971 | |
|
|
972 | enc.json = *json; |
365 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
973 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
366 | enc.cur = SvPVX (enc.sv); |
974 | enc.cur = SvPVX (enc.sv); |
367 | enc.end = SvEND (enc.sv); |
975 | enc.end = SvEND (enc.sv); |
368 | enc.max_recurse = 0; |
|
|
369 | enc.indent = 0; |
976 | enc.indent = 0; |
|
|
977 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
978 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
979 | : 0x110000UL; |
370 | |
980 | |
371 | SvPOK_only (enc.sv); |
981 | SvPOK_only (enc.sv); |
372 | encode_sv (&enc, scalar); |
982 | encode_sv (&enc, scalar); |
|
|
983 | encode_nl (&enc); |
373 | |
984 | |
|
|
985 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
986 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
987 | |
374 | if (!(flags & (F_ASCII | F_UTF8))) |
988 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
375 | SvUTF8_on (enc.sv); |
989 | SvUTF8_on (enc.sv); |
376 | |
990 | |
377 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
991 | if (enc.json.flags & F_SHRINK) |
|
|
992 | shrink (enc.sv); |
|
|
993 | |
378 | return enc.sv; |
994 | return enc.sv; |
379 | } |
995 | } |
380 | |
996 | |
381 | ///////////////////////////////////////////////////////////////////////////// |
997 | ///////////////////////////////////////////////////////////////////////////// |
|
|
998 | // decoder |
382 | |
999 | |
383 | #define WS \ |
1000 | // structure used for decoding JSON |
|
|
1001 | typedef struct |
|
|
1002 | { |
|
|
1003 | char *cur; // current parser pointer |
|
|
1004 | char *end; // end of input string |
|
|
1005 | const char *err; // parse error, if != 0 |
|
|
1006 | JSON json; |
|
|
1007 | U32 depth; // recursion depth |
|
|
1008 | U32 maxdepth; // recursion depth limit |
|
|
1009 | } dec_t; |
|
|
1010 | |
|
|
1011 | INLINE void |
|
|
1012 | decode_comment (dec_t *dec) |
|
|
1013 | { |
|
|
1014 | // only '#'-style comments allowed a.t.m. |
|
|
1015 | |
|
|
1016 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
1017 | ++dec->cur; |
|
|
1018 | } |
|
|
1019 | |
|
|
1020 | INLINE void |
|
|
1021 | decode_ws (dec_t *dec) |
|
|
1022 | { |
384 | for (;;) \ |
1023 | for (;;) |
385 | { \ |
1024 | { |
386 | char ch = *dec->cur; \ |
1025 | char ch = *dec->cur; |
|
|
1026 | |
387 | if (ch > 0x20 \ |
1027 | if (ch > 0x20) |
|
|
1028 | { |
|
|
1029 | if (expect_false (ch == '#')) |
|
|
1030 | { |
|
|
1031 | if (dec->json.flags & F_RELAXED) |
|
|
1032 | decode_comment (dec); |
|
|
1033 | else |
|
|
1034 | break; |
|
|
1035 | } |
|
|
1036 | else |
|
|
1037 | break; |
|
|
1038 | } |
388 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ |
1039 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
389 | break; \ |
1040 | break; // parse error, but let higher level handle it, gives better error messages |
|
|
1041 | |
390 | ++dec->cur; \ |
1042 | ++dec->cur; |
391 | } |
1043 | } |
|
|
1044 | } |
392 | |
1045 | |
393 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
1046 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
|
|
1047 | |
394 | #define EXPECT_CH(ch) SB \ |
1048 | #define EXPECT_CH(ch) SB \ |
395 | if (*dec->cur != ch) \ |
1049 | if (*dec->cur != ch) \ |
396 | ERR (# ch " expected"); \ |
1050 | ERR (# ch " expected"); \ |
397 | ++dec->cur; \ |
1051 | ++dec->cur; \ |
398 | SE |
1052 | SE |
399 | |
1053 | |
|
|
1054 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
|
|
1055 | #define DEC_DEC_DEPTH --dec->depth |
|
|
1056 | |
400 | static SV *decode_sv (dec_t *dec); |
1057 | static SV *decode_sv (dec_t *dec); |
401 | |
|
|
402 | #define APPEND_CH(ch) SB \ |
|
|
403 | SvGROW (sv, cur + 1 + 1); \ |
|
|
404 | SvPVX (sv)[cur++] = (ch); \ |
|
|
405 | SE |
|
|
406 | |
1058 | |
407 | static signed char decode_hexdigit[256]; |
1059 | static signed char decode_hexdigit[256]; |
408 | |
1060 | |
409 | static UV |
1061 | static UV |
410 | decode_4hex (dec_t *dec) |
1062 | decode_4hex (dec_t *dec) |
411 | { |
1063 | { |
412 | signed char d1, d2, d3, d4; |
1064 | signed char d1, d2, d3, d4; |
|
|
1065 | unsigned char *cur = (unsigned char *)dec->cur; |
413 | |
1066 | |
414 | d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; |
1067 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
415 | if (d1 < 0) ERR ("four hexadecimal digits expected"); |
1068 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
416 | d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; |
1069 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
417 | if (d2 < 0) ERR ("four hexadecimal digits expected"); |
1070 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
418 | d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]]; |
|
|
419 | if (d3 < 0) ERR ("four hexadecimal digits expected"); |
|
|
420 | d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]]; |
|
|
421 | if (d4 < 0) ERR ("four hexadecimal digits expected"); |
|
|
422 | |
1071 | |
423 | dec->cur += 4; |
1072 | dec->cur += 4; |
424 | |
1073 | |
425 | return ((UV)d1) << 12 |
1074 | return ((UV)d1) << 12 |
426 | | ((UV)d2) << 8 |
1075 | | ((UV)d2) << 8 |
… | |
… | |
432 | } |
1081 | } |
433 | |
1082 | |
434 | static SV * |
1083 | static SV * |
435 | decode_str (dec_t *dec) |
1084 | decode_str (dec_t *dec) |
436 | { |
1085 | { |
437 | SV *sv = NEWSV (0,2); |
1086 | SV *sv = 0; |
438 | STRLEN cur = 0; |
|
|
439 | int utf8 = 0; |
1087 | int utf8 = 0; |
|
|
1088 | char *dec_cur = dec->cur; |
440 | |
1089 | |
441 | for (;;) |
1090 | do |
442 | { |
1091 | { |
443 | unsigned char ch = *(unsigned char *)dec->cur; |
1092 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
|
|
1093 | char *cur = buf; |
444 | |
1094 | |
445 | if (ch == '"') |
1095 | do |
446 | break; |
|
|
447 | else if (ch == '\\') |
|
|
448 | { |
1096 | { |
449 | switch (*++dec->cur) |
1097 | unsigned char ch = *(unsigned char *)dec_cur++; |
|
|
1098 | |
|
|
1099 | if (expect_false (ch == '"')) |
450 | { |
1100 | { |
451 | case '\\': |
1101 | --dec_cur; |
452 | case '/': |
1102 | break; |
453 | case '"': APPEND_CH (*dec->cur++); break; |
1103 | } |
454 | |
1104 | else if (expect_false (ch == '\\')) |
455 | case 'b': APPEND_CH ('\010'); ++dec->cur; break; |
1105 | { |
456 | case 't': APPEND_CH ('\011'); ++dec->cur; break; |
1106 | switch (*dec_cur) |
457 | case 'n': APPEND_CH ('\012'); ++dec->cur; break; |
|
|
458 | case 'f': APPEND_CH ('\014'); ++dec->cur; break; |
|
|
459 | case 'r': APPEND_CH ('\015'); ++dec->cur; break; |
|
|
460 | |
|
|
461 | case 'u': |
|
|
462 | { |
1107 | { |
463 | UV lo, hi; |
1108 | case '\\': |
464 | ++dec->cur; |
1109 | case '/': |
|
|
1110 | case '"': *cur++ = *dec_cur++; break; |
465 | |
1111 | |
466 | hi = decode_4hex (dec); |
1112 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
467 | if (hi == (UV)-1) |
1113 | case 't': ++dec_cur; *cur++ = '\011'; break; |
468 | goto fail; |
1114 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
|
|
1115 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
|
|
1116 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
469 | |
1117 | |
470 | // possibly a surrogate pair |
1118 | case 'u': |
471 | if (hi >= 0xd800 && hi < 0xdc00) |
|
|
472 | { |
1119 | { |
473 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
1120 | UV lo, hi; |
474 | ERR ("illegal surrogate character"); |
1121 | ++dec_cur; |
475 | |
1122 | |
476 | dec->cur += 2; |
1123 | dec->cur = dec_cur; |
477 | |
|
|
478 | lo = decode_4hex (dec); |
1124 | hi = decode_4hex (dec); |
|
|
1125 | dec_cur = dec->cur; |
479 | if (lo == (UV)-1) |
1126 | if (hi == (UV)-1) |
480 | goto fail; |
1127 | goto fail; |
481 | |
1128 | |
|
|
1129 | // possibly a surrogate pair |
|
|
1130 | if (hi >= 0xd800) |
|
|
1131 | if (hi < 0xdc00) |
|
|
1132 | { |
|
|
1133 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
|
|
1134 | ERR ("missing low surrogate character in surrogate pair"); |
|
|
1135 | |
|
|
1136 | dec_cur += 2; |
|
|
1137 | |
|
|
1138 | dec->cur = dec_cur; |
|
|
1139 | lo = decode_4hex (dec); |
|
|
1140 | dec_cur = dec->cur; |
|
|
1141 | if (lo == (UV)-1) |
|
|
1142 | goto fail; |
|
|
1143 | |
482 | if (lo < 0xdc00 || lo >= 0xe000) |
1144 | if (lo < 0xdc00 || lo >= 0xe000) |
483 | ERR ("surrogate pair expected"); |
1145 | ERR ("surrogate pair expected"); |
484 | |
1146 | |
485 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
1147 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
|
|
1148 | } |
|
|
1149 | else if (hi < 0xe000) |
|
|
1150 | ERR ("missing high surrogate character in surrogate pair"); |
|
|
1151 | |
|
|
1152 | if (hi >= 0x80) |
|
|
1153 | { |
|
|
1154 | utf8 = 1; |
|
|
1155 | |
|
|
1156 | cur = encode_utf8 (cur, hi); |
|
|
1157 | } |
|
|
1158 | else |
|
|
1159 | *cur++ = hi; |
486 | } |
1160 | } |
487 | else if (lo >= 0xdc00 && lo < 0xe000) |
|
|
488 | ERR ("illegal surrogate character"); |
|
|
489 | |
|
|
490 | if (hi >= 0x80) |
|
|
491 | { |
1161 | break; |
492 | utf8 = 1; |
|
|
493 | |
1162 | |
494 | SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits |
|
|
495 | cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv); |
|
|
496 | } |
|
|
497 | else |
1163 | default: |
498 | APPEND_CH (hi); |
1164 | --dec_cur; |
|
|
1165 | ERR ("illegal backslash escape sequence in string"); |
499 | } |
1166 | } |
|
|
1167 | } |
|
|
1168 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
|
|
1169 | *cur++ = ch; |
|
|
1170 | else if (ch >= 0x80) |
|
|
1171 | { |
|
|
1172 | STRLEN clen; |
|
|
1173 | |
|
|
1174 | --dec_cur; |
|
|
1175 | |
|
|
1176 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
|
|
1177 | if (clen == (STRLEN)-1) |
|
|
1178 | ERR ("malformed UTF-8 character in JSON string"); |
|
|
1179 | |
|
|
1180 | do |
|
|
1181 | *cur++ = *dec_cur++; |
|
|
1182 | while (--clen); |
|
|
1183 | |
|
|
1184 | utf8 = 1; |
|
|
1185 | } |
|
|
1186 | else if (ch == '\t' && dec->json.flags & F_RELAXED) |
|
|
1187 | *cur++ = ch; |
|
|
1188 | else |
|
|
1189 | { |
|
|
1190 | --dec_cur; |
|
|
1191 | |
|
|
1192 | if (!ch) |
|
|
1193 | ERR ("unexpected end of string while parsing JSON string"); |
500 | break; |
1194 | else |
|
|
1195 | ERR ("invalid character encountered while parsing JSON string"); |
501 | } |
1196 | } |
502 | } |
1197 | } |
503 | else if (ch >= 0x20 && ch <= 0x7f) |
1198 | while (cur < buf + SHORT_STRING_LEN); |
504 | APPEND_CH (*dec->cur++); |
1199 | |
505 | else if (ch >= 0x80) |
1200 | { |
|
|
1201 | STRLEN len = cur - buf; |
|
|
1202 | |
|
|
1203 | if (sv) |
506 | { |
1204 | { |
507 | STRLEN clen; |
1205 | STRLEN cur = SvCUR (sv); |
508 | UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); |
|
|
509 | if (clen < 0) |
|
|
510 | ERR ("malformed UTF-8 character in string, cannot convert to JSON"); |
|
|
511 | |
1206 | |
512 | SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits |
1207 | if (SvLEN (sv) - cur <= len) |
513 | memcpy (SvPVX (sv) + cur, dec->cur, clen); |
1208 | json_sv_grow (sv, cur, len); |
514 | dec->cur += clen; |
1209 | |
|
|
1210 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
|
|
1211 | SvCUR_set (sv, SvCUR (sv) + len); |
515 | } |
1212 | } |
516 | else |
1213 | else |
517 | ERR ("invalid character encountered"); |
1214 | sv = newSVpvn (buf, len); |
518 | } |
1215 | } |
|
|
1216 | } |
|
|
1217 | while (*dec_cur != '"'); |
519 | |
1218 | |
520 | ++dec->cur; |
1219 | ++dec_cur; |
521 | |
1220 | |
|
|
1221 | if (sv) |
|
|
1222 | { |
522 | SvPOK_only (sv); |
1223 | SvPOK_only (sv); |
523 | |
|
|
524 | SvCUR_set (sv, cur); |
|
|
525 | *SvEND (sv) = 0; |
1224 | *SvEND (sv) = 0; |
526 | |
1225 | |
527 | if (utf8) |
1226 | if (utf8) |
528 | SvUTF8_on (sv); |
1227 | SvUTF8_on (sv); |
|
|
1228 | } |
|
|
1229 | else |
|
|
1230 | sv = newSVpvn ("", 0); |
529 | |
1231 | |
|
|
1232 | dec->cur = dec_cur; |
530 | return sv; |
1233 | return sv; |
531 | |
1234 | |
532 | fail: |
1235 | fail: |
533 | SvREFCNT_dec (sv); |
1236 | dec->cur = dec_cur; |
534 | return 0; |
1237 | return 0; |
535 | } |
1238 | } |
536 | |
1239 | |
537 | static SV * |
1240 | static SV * |
538 | decode_num (dec_t *dec) |
1241 | decode_num (dec_t *dec) |
… | |
… | |
548 | { |
1251 | { |
549 | ++dec->cur; |
1252 | ++dec->cur; |
550 | if (*dec->cur >= '0' && *dec->cur <= '9') |
1253 | if (*dec->cur >= '0' && *dec->cur <= '9') |
551 | ERR ("malformed number (leading zero must not be followed by another digit)"); |
1254 | ERR ("malformed number (leading zero must not be followed by another digit)"); |
552 | } |
1255 | } |
553 | |
1256 | else if (*dec->cur < '0' || *dec->cur > '9') |
554 | // int |
1257 | ERR ("malformed number (no digits after initial minus)"); |
|
|
1258 | else |
|
|
1259 | do |
|
|
1260 | { |
|
|
1261 | ++dec->cur; |
|
|
1262 | } |
555 | while (*dec->cur >= '0' && *dec->cur <= '9') |
1263 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
556 | ++dec->cur; |
|
|
557 | |
1264 | |
558 | // [frac] |
1265 | // [frac] |
559 | if (*dec->cur == '.') |
1266 | if (*dec->cur == '.') |
560 | { |
1267 | { |
561 | is_nv = 1; |
1268 | ++dec->cur; |
|
|
1269 | |
|
|
1270 | if (*dec->cur < '0' || *dec->cur > '9') |
|
|
1271 | ERR ("malformed number (no digits after decimal point)"); |
562 | |
1272 | |
563 | do |
1273 | do |
564 | { |
1274 | { |
565 | ++dec->cur; |
1275 | ++dec->cur; |
566 | } |
1276 | } |
567 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
1277 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
|
|
1278 | |
|
|
1279 | is_nv = 1; |
568 | } |
1280 | } |
569 | |
1281 | |
570 | // [exp] |
1282 | // [exp] |
571 | if (*dec->cur == 'e' || *dec->cur == 'E') |
1283 | if (*dec->cur == 'e' || *dec->cur == 'E') |
572 | { |
1284 | { |
573 | is_nv = 1; |
|
|
574 | |
|
|
575 | ++dec->cur; |
1285 | ++dec->cur; |
|
|
1286 | |
576 | if (*dec->cur == '-' || *dec->cur == '+') |
1287 | if (*dec->cur == '-' || *dec->cur == '+') |
577 | ++dec->cur; |
1288 | ++dec->cur; |
578 | |
1289 | |
|
|
1290 | if (*dec->cur < '0' || *dec->cur > '9') |
|
|
1291 | ERR ("malformed number (no digits after exp sign)"); |
|
|
1292 | |
|
|
1293 | do |
|
|
1294 | { |
|
|
1295 | ++dec->cur; |
|
|
1296 | } |
579 | while (*dec->cur >= '0' && *dec->cur <= '9') |
1297 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
580 | ++dec->cur; |
1298 | |
|
|
1299 | is_nv = 1; |
581 | } |
1300 | } |
582 | |
1301 | |
583 | if (!is_nv) |
1302 | if (!is_nv) |
584 | { |
1303 | { |
585 | UV uv; |
1304 | int len = dec->cur - start; |
586 | int numtype = grok_number (start, dec->cur - start, &uv); |
1305 | |
587 | if (numtype & IS_NUMBER_IN_UV) |
1306 | // special case the rather common 1..5-digit-int case |
588 | if (numtype & IS_NUMBER_NEG) |
1307 | if (*start == '-') |
|
|
1308 | switch (len) |
589 | { |
1309 | { |
590 | if (uv < (UV)IV_MIN) |
1310 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
591 | return newSViv (-(IV)uv); |
1311 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
|
|
1312 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
1313 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1314 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
592 | } |
1315 | } |
|
|
1316 | else |
|
|
1317 | switch (len) |
|
|
1318 | { |
|
|
1319 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
1320 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
1321 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
1322 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1323 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
|
|
1324 | } |
|
|
1325 | |
|
|
1326 | { |
|
|
1327 | UV uv; |
|
|
1328 | int numtype = grok_number (start, len, &uv); |
|
|
1329 | if (numtype & IS_NUMBER_IN_UV) |
|
|
1330 | if (numtype & IS_NUMBER_NEG) |
|
|
1331 | { |
|
|
1332 | if (uv < (UV)IV_MIN) |
|
|
1333 | return newSViv (-(IV)uv); |
|
|
1334 | } |
593 | else |
1335 | else |
594 | return newSVuv (uv); |
1336 | return newSVuv (uv); |
595 | } |
1337 | } |
596 | |
1338 | |
|
|
1339 | len -= *start == '-' ? 1 : 0; |
|
|
1340 | |
|
|
1341 | // does not fit into IV or UV, try NV |
|
|
1342 | if (len <= NV_DIG) |
|
|
1343 | // fits into NV without loss of precision |
|
|
1344 | return newSVnv (json_atof (start)); |
|
|
1345 | |
|
|
1346 | // everything else fails, convert it to a string |
|
|
1347 | return newSVpvn (start, dec->cur - start); |
|
|
1348 | } |
|
|
1349 | |
|
|
1350 | // loss of precision here |
597 | return newSVnv (Atof (start)); |
1351 | return newSVnv (json_atof (start)); |
598 | |
1352 | |
599 | fail: |
1353 | fail: |
600 | return 0; |
1354 | return 0; |
601 | } |
1355 | } |
602 | |
1356 | |
603 | static SV * |
1357 | static SV * |
604 | decode_av (dec_t *dec) |
1358 | decode_av (dec_t *dec) |
605 | { |
1359 | { |
606 | AV *av = newAV (); |
1360 | AV *av = newAV (); |
607 | |
1361 | |
|
|
1362 | DEC_INC_DEPTH; |
|
|
1363 | decode_ws (dec); |
|
|
1364 | |
|
|
1365 | if (*dec->cur == ']') |
|
|
1366 | ++dec->cur; |
|
|
1367 | else |
608 | for (;;) |
1368 | for (;;) |
609 | { |
1369 | { |
610 | SV *value; |
1370 | SV *value; |
611 | |
1371 | |
612 | value = decode_sv (dec); |
1372 | value = decode_sv (dec); |
613 | if (!value) |
1373 | if (!value) |
614 | goto fail; |
1374 | goto fail; |
615 | |
1375 | |
616 | av_push (av, value); |
1376 | av_push (av, value); |
617 | |
1377 | |
618 | WS; |
1378 | decode_ws (dec); |
619 | |
1379 | |
620 | if (*dec->cur == ']') |
1380 | if (*dec->cur == ']') |
621 | { |
1381 | { |
622 | ++dec->cur; |
1382 | ++dec->cur; |
623 | break; |
1383 | break; |
624 | } |
1384 | } |
625 | |
1385 | |
626 | if (*dec->cur != ',') |
1386 | if (*dec->cur != ',') |
627 | ERR (", or ] expected while parsing array"); |
1387 | ERR (", or ] expected while parsing array"); |
628 | |
1388 | |
629 | ++dec->cur; |
1389 | ++dec->cur; |
|
|
1390 | |
|
|
1391 | decode_ws (dec); |
|
|
1392 | |
|
|
1393 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1394 | { |
|
|
1395 | ++dec->cur; |
|
|
1396 | break; |
|
|
1397 | } |
630 | } |
1398 | } |
631 | |
1399 | |
|
|
1400 | DEC_DEC_DEPTH; |
632 | return newRV_noinc ((SV *)av); |
1401 | return newRV_noinc ((SV *)av); |
633 | |
1402 | |
634 | fail: |
1403 | fail: |
635 | SvREFCNT_dec (av); |
1404 | SvREFCNT_dec (av); |
|
|
1405 | DEC_DEC_DEPTH; |
636 | return 0; |
1406 | return 0; |
637 | } |
1407 | } |
638 | |
1408 | |
639 | static SV * |
1409 | static SV * |
640 | decode_hv (dec_t *dec) |
1410 | decode_hv (dec_t *dec) |
641 | { |
1411 | { |
|
|
1412 | SV *sv; |
642 | HV *hv = newHV (); |
1413 | HV *hv = newHV (); |
643 | |
1414 | |
|
|
1415 | DEC_INC_DEPTH; |
|
|
1416 | decode_ws (dec); |
|
|
1417 | |
|
|
1418 | if (*dec->cur == '}') |
|
|
1419 | ++dec->cur; |
|
|
1420 | else |
644 | for (;;) |
1421 | for (;;) |
645 | { |
1422 | { |
646 | SV *key, *value; |
|
|
647 | |
|
|
648 | WS; EXPECT_CH ('"'); |
1423 | EXPECT_CH ('"'); |
649 | |
1424 | |
650 | key = decode_str (dec); |
1425 | // heuristic: assume that |
651 | if (!key) |
1426 | // a) decode_str + hv_store_ent are abysmally slow. |
652 | goto fail; |
1427 | // b) most hash keys are short, simple ascii text. |
653 | |
1428 | // => try to "fast-match" such strings to avoid |
654 | WS; EXPECT_CH (':'); |
1429 | // the overhead of decode_str + hv_store_ent. |
655 | |
|
|
656 | value = decode_sv (dec); |
|
|
657 | if (!value) |
|
|
658 | { |
1430 | { |
|
|
1431 | SV *value; |
|
|
1432 | char *p = dec->cur; |
|
|
1433 | char *e = p + 24; // only try up to 24 bytes |
|
|
1434 | |
|
|
1435 | for (;;) |
|
|
1436 | { |
|
|
1437 | // the >= 0x80 is false on most architectures |
|
|
1438 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1439 | { |
|
|
1440 | // slow path, back up and use decode_str |
|
|
1441 | SV *key = decode_str (dec); |
|
|
1442 | if (!key) |
|
|
1443 | goto fail; |
|
|
1444 | |
|
|
1445 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1446 | |
|
|
1447 | decode_ws (dec); |
|
|
1448 | value = decode_sv (dec); |
|
|
1449 | if (!value) |
|
|
1450 | { |
|
|
1451 | SvREFCNT_dec (key); |
|
|
1452 | goto fail; |
|
|
1453 | } |
|
|
1454 | |
|
|
1455 | hv_store_ent (hv, key, value, 0); |
659 | SvREFCNT_dec (key); |
1456 | SvREFCNT_dec (key); |
|
|
1457 | |
|
|
1458 | break; |
|
|
1459 | } |
|
|
1460 | else if (*p == '"') |
|
|
1461 | { |
|
|
1462 | // fast path, got a simple key |
|
|
1463 | char *key = dec->cur; |
|
|
1464 | int len = p - key; |
|
|
1465 | dec->cur = p + 1; |
|
|
1466 | |
|
|
1467 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1468 | |
|
|
1469 | decode_ws (dec); |
|
|
1470 | value = decode_sv (dec); |
|
|
1471 | if (!value) |
660 | goto fail; |
1472 | goto fail; |
|
|
1473 | |
|
|
1474 | hv_store (hv, key, len, value, 0); |
|
|
1475 | |
|
|
1476 | break; |
|
|
1477 | } |
|
|
1478 | |
|
|
1479 | ++p; |
|
|
1480 | } |
661 | } |
1481 | } |
662 | |
1482 | |
663 | //TODO: optimise |
1483 | decode_ws (dec); |
664 | hv_store_ent (hv, key, value, 0); |
|
|
665 | |
1484 | |
666 | WS; |
|
|
667 | |
|
|
668 | if (*dec->cur == '}') |
1485 | if (*dec->cur == '}') |
|
|
1486 | { |
|
|
1487 | ++dec->cur; |
|
|
1488 | break; |
|
|
1489 | } |
|
|
1490 | |
|
|
1491 | if (*dec->cur != ',') |
|
|
1492 | ERR (", or } expected while parsing object/hash"); |
|
|
1493 | |
|
|
1494 | ++dec->cur; |
|
|
1495 | |
|
|
1496 | decode_ws (dec); |
|
|
1497 | |
|
|
1498 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1499 | { |
|
|
1500 | ++dec->cur; |
|
|
1501 | break; |
|
|
1502 | } |
|
|
1503 | } |
|
|
1504 | |
|
|
1505 | DEC_DEC_DEPTH; |
|
|
1506 | sv = newRV_noinc ((SV *)hv); |
|
|
1507 | |
|
|
1508 | // check filter callbacks |
|
|
1509 | if (dec->json.flags & F_HOOK) |
|
|
1510 | { |
|
|
1511 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
669 | { |
1512 | { |
670 | ++dec->cur; |
1513 | HE *cb, *he; |
671 | break; |
1514 | |
|
|
1515 | hv_iterinit (hv); |
|
|
1516 | he = hv_iternext (hv); |
|
|
1517 | hv_iterinit (hv); |
|
|
1518 | |
|
|
1519 | // the next line creates a mortal sv each time it's called. |
|
|
1520 | // might want to optimise this for common cases. |
|
|
1521 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1522 | |
|
|
1523 | if (cb) |
|
|
1524 | { |
|
|
1525 | dSP; |
|
|
1526 | int count; |
|
|
1527 | |
|
|
1528 | ENTER; SAVETMPS; |
|
|
1529 | PUSHMARK (SP); |
|
|
1530 | XPUSHs (HeVAL (he)); |
|
|
1531 | sv_2mortal (sv); |
|
|
1532 | |
|
|
1533 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1534 | |
|
|
1535 | if (count == 1) |
|
|
1536 | { |
|
|
1537 | sv = newSVsv (POPs); |
|
|
1538 | FREETMPS; LEAVE; |
|
|
1539 | return sv; |
|
|
1540 | } |
|
|
1541 | else if (count) |
|
|
1542 | croak ("filter_json_single_key_object callbacks must not return more than one scalar"); |
|
|
1543 | |
|
|
1544 | SvREFCNT_inc (sv); |
|
|
1545 | FREETMPS; LEAVE; |
|
|
1546 | } |
672 | } |
1547 | } |
673 | |
1548 | |
674 | if (*dec->cur != ',') |
1549 | if (dec->json.cb_object) |
675 | ERR (", or } expected while parsing object/hash"); |
1550 | { |
|
|
1551 | dSP; |
|
|
1552 | int count; |
676 | |
1553 | |
677 | ++dec->cur; |
1554 | ENTER; SAVETMPS; |
678 | } |
1555 | PUSHMARK (SP); |
|
|
1556 | XPUSHs (sv_2mortal (sv)); |
679 | |
1557 | |
680 | return newRV_noinc ((SV *)hv); |
1558 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1559 | |
|
|
1560 | if (count == 1) |
|
|
1561 | { |
|
|
1562 | sv = newSVsv (POPs); |
|
|
1563 | FREETMPS; LEAVE; |
|
|
1564 | return sv; |
|
|
1565 | } |
|
|
1566 | else if (count) |
|
|
1567 | croak ("filter_json_object callbacks must not return more than one scalar"); |
|
|
1568 | |
|
|
1569 | SvREFCNT_inc (sv); |
|
|
1570 | FREETMPS; LEAVE; |
|
|
1571 | } |
|
|
1572 | } |
|
|
1573 | |
|
|
1574 | return sv; |
681 | |
1575 | |
682 | fail: |
1576 | fail: |
683 | SvREFCNT_dec (hv); |
1577 | SvREFCNT_dec (hv); |
|
|
1578 | DEC_DEC_DEPTH; |
|
|
1579 | return 0; |
|
|
1580 | } |
|
|
1581 | |
|
|
1582 | static SV * |
|
|
1583 | decode_tag (dec_t *dec) |
|
|
1584 | { |
|
|
1585 | SV *tag = 0; |
|
|
1586 | SV *val = 0; |
|
|
1587 | |
|
|
1588 | if (!(dec->json.flags & F_ALLOW_TAGS)) |
|
|
1589 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
|
|
1590 | |
|
|
1591 | ++dec->cur; |
|
|
1592 | |
|
|
1593 | decode_ws (dec); |
|
|
1594 | |
|
|
1595 | tag = decode_sv (dec); |
|
|
1596 | if (!tag) |
|
|
1597 | goto fail; |
|
|
1598 | |
|
|
1599 | if (!SvPOK (tag)) |
|
|
1600 | ERR ("malformed JSON string, (tag) must be a string"); |
|
|
1601 | |
|
|
1602 | decode_ws (dec); |
|
|
1603 | |
|
|
1604 | if (*dec->cur != ')') |
|
|
1605 | ERR (") expected after tag"); |
|
|
1606 | |
|
|
1607 | ++dec->cur; |
|
|
1608 | |
|
|
1609 | decode_ws (dec); |
|
|
1610 | |
|
|
1611 | val = decode_sv (dec); |
|
|
1612 | if (!val) |
|
|
1613 | goto fail; |
|
|
1614 | |
|
|
1615 | if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV) |
|
|
1616 | ERR ("malformed JSON string, tag value must be an array"); |
|
|
1617 | |
|
|
1618 | { |
|
|
1619 | AV *av = (AV *)SvRV (val); |
|
|
1620 | int i, len = av_len (av) + 1; |
|
|
1621 | HV *stash = gv_stashsv (tag, 0); |
|
|
1622 | SV *sv; |
|
|
1623 | |
|
|
1624 | if (!stash) |
|
|
1625 | ERR ("cannot decode perl-object (package does not exist)"); |
|
|
1626 | |
|
|
1627 | GV *method = gv_fetchmethod_autoload (stash, "THAW", 0); |
|
|
1628 | |
|
|
1629 | if (!method) |
|
|
1630 | ERR ("cannot decode perl-object (package does not have a THAW method)"); |
|
|
1631 | |
|
|
1632 | dSP; |
|
|
1633 | |
|
|
1634 | ENTER; SAVETMPS; |
|
|
1635 | PUSHMARK (SP); |
|
|
1636 | EXTEND (SP, len + 2); |
|
|
1637 | // we re-bless the reference to get overload and other niceties right |
|
|
1638 | PUSHs (tag); |
|
|
1639 | PUSHs (sv_json); |
|
|
1640 | |
|
|
1641 | for (i = 0; i < len; ++i) |
|
|
1642 | PUSHs (*av_fetch (av, i, 1)); |
|
|
1643 | |
|
|
1644 | PUTBACK; |
|
|
1645 | call_sv ((SV *)GvCV (method), G_SCALAR); |
|
|
1646 | SPAGAIN; |
|
|
1647 | |
|
|
1648 | SvREFCNT_dec (tag); |
|
|
1649 | SvREFCNT_dec (val); |
|
|
1650 | sv = SvREFCNT_inc (POPs); |
|
|
1651 | |
|
|
1652 | PUTBACK; |
|
|
1653 | |
|
|
1654 | FREETMPS; LEAVE; |
|
|
1655 | |
|
|
1656 | return sv; |
|
|
1657 | } |
|
|
1658 | |
|
|
1659 | fail: |
|
|
1660 | SvREFCNT_dec (tag); |
|
|
1661 | SvREFCNT_dec (val); |
684 | return 0; |
1662 | return 0; |
685 | } |
1663 | } |
686 | |
1664 | |
687 | static SV * |
1665 | static SV * |
688 | decode_sv (dec_t *dec) |
1666 | decode_sv (dec_t *dec) |
689 | { |
1667 | { |
690 | WS; |
1668 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1669 | // to parse everything. |
691 | switch (*dec->cur) |
1670 | switch (*dec->cur) |
692 | { |
1671 | { |
693 | case '"': ++dec->cur; return decode_str (dec); |
1672 | case '"': ++dec->cur; return decode_str (dec); |
694 | case '[': ++dec->cur; return decode_av (dec); |
1673 | case '[': ++dec->cur; return decode_av (dec); |
695 | case '{': ++dec->cur; return decode_hv (dec); |
1674 | case '{': ++dec->cur; return decode_hv (dec); |
|
|
1675 | case '(': return decode_tag (dec); |
696 | |
1676 | |
697 | case '-': |
1677 | case '-': |
698 | case '0': case '1': case '2': case '3': case '4': |
1678 | case '0': case '1': case '2': case '3': case '4': |
699 | case '5': case '6': case '7': case '8': case '9': |
1679 | case '5': case '6': case '7': case '8': case '9': |
700 | return decode_num (dec); |
1680 | return decode_num (dec); |
701 | |
1681 | |
|
|
1682 | case 'f': |
|
|
1683 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
|
|
1684 | { |
|
|
1685 | dec->cur += 5; |
|
|
1686 | |
|
|
1687 | if (expect_false (!dec->json.v_false)) |
|
|
1688 | dec->json.v_false = GET_BOOL (false); |
|
|
1689 | |
|
|
1690 | return newSVsv (dec->json.v_false); |
|
|
1691 | } |
|
|
1692 | else |
|
|
1693 | ERR ("'false' expected"); |
|
|
1694 | |
|
|
1695 | break; |
|
|
1696 | |
702 | case 't': |
1697 | case 't': |
703 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1698 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
704 | { |
1699 | { |
705 | dec->cur += 4; |
1700 | dec->cur += 4; |
|
|
1701 | |
|
|
1702 | if (expect_false (!dec->json.v_true)) |
|
|
1703 | dec->json.v_true = GET_BOOL (true); |
|
|
1704 | |
706 | return newSViv (1); |
1705 | return newSVsv (dec->json.v_true); |
707 | } |
1706 | } |
708 | else |
1707 | else |
709 | ERR ("'true' expected"); |
1708 | ERR ("'true' expected"); |
710 | |
|
|
711 | break; |
|
|
712 | |
|
|
713 | case 'f': |
|
|
714 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
|
|
715 | { |
|
|
716 | dec->cur += 5; |
|
|
717 | return newSViv (0); |
|
|
718 | } |
|
|
719 | else |
|
|
720 | ERR ("'false' expected"); |
|
|
721 | |
1709 | |
722 | break; |
1710 | break; |
723 | |
1711 | |
724 | case 'n': |
1712 | case 'n': |
725 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
1713 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
726 | { |
1714 | { |
727 | dec->cur += 4; |
1715 | dec->cur += 4; |
728 | return newSViv (1); |
1716 | return newSVsv (&PL_sv_undef); |
729 | } |
1717 | } |
730 | else |
1718 | else |
731 | ERR ("'null' expected"); |
1719 | ERR ("'null' expected"); |
732 | |
1720 | |
733 | break; |
1721 | break; |
734 | |
1722 | |
735 | default: |
1723 | default: |
736 | ERR ("malformed json string"); |
1724 | ERR ("malformed JSON string, neither tag, array, object, number, string or atom"); |
737 | break; |
1725 | break; |
738 | } |
1726 | } |
739 | |
1727 | |
740 | fail: |
1728 | fail: |
741 | return 0; |
1729 | return 0; |
742 | } |
1730 | } |
743 | |
1731 | |
744 | static SV * |
1732 | static SV * |
745 | decode_json (SV *string, UV flags) |
1733 | decode_json (SV *string, JSON *json, STRLEN *offset_return) |
746 | { |
1734 | { |
|
|
1735 | dec_t dec; |
747 | SV *sv; |
1736 | SV *sv; |
748 | |
1737 | |
749 | if (!(flags & F_UTF8)) |
1738 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1739 | * makes perl ignore the magic in subsequent accesses. |
|
|
1740 | * also make a copy of non-PV values, to get them into a clean |
|
|
1741 | * state (SvPV should do that, but it's buggy, see below). |
|
|
1742 | * |
|
|
1743 | * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16), |
|
|
1744 | * as reported by Reini Urban. |
|
|
1745 | */ |
|
|
1746 | /*SvGETMAGIC (string);*/ |
|
|
1747 | if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string)) |
|
|
1748 | string = sv_2mortal (newSVsv (string)); |
|
|
1749 | |
|
|
1750 | SvUPGRADE (string, SVt_PV); |
|
|
1751 | |
|
|
1752 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1753 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1754 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1755 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1756 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1757 | * and hope for the best. |
|
|
1758 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1759 | * assertion business is seriously broken, try yet another workaround |
|
|
1760 | * for the broken -DDEBUGGING. |
|
|
1761 | */ |
|
|
1762 | { |
|
|
1763 | #ifdef DEBUGGING |
|
|
1764 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1765 | #else |
|
|
1766 | STRLEN offset = SvCUR (string); |
|
|
1767 | #endif |
|
|
1768 | |
|
|
1769 | if (offset > json->max_size && json->max_size) |
|
|
1770 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1771 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1772 | } |
|
|
1773 | |
|
|
1774 | if (DECODE_WANTS_OCTETS (json)) |
|
|
1775 | sv_utf8_downgrade (string, 0); |
|
|
1776 | else |
750 | sv_utf8_upgrade (string); |
1777 | sv_utf8_upgrade (string); |
751 | |
1778 | |
752 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1779 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
753 | |
1780 | |
754 | dec_t dec; |
1781 | dec.json = *json; |
755 | dec.flags = flags; |
|
|
756 | dec.cur = SvPVX (string); |
1782 | dec.cur = SvPVX (string); |
757 | dec.end = SvEND (string); |
1783 | dec.end = SvEND (string); |
758 | dec.err = 0; |
1784 | dec.err = 0; |
|
|
1785 | dec.depth = 0; |
759 | |
1786 | |
760 | *dec.end = 1; // invalid anywhere |
1787 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1788 | dec.json.flags |= F_HOOK; |
|
|
1789 | |
|
|
1790 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1791 | |
|
|
1792 | decode_ws (&dec); |
761 | sv = decode_sv (&dec); |
1793 | sv = decode_sv (&dec); |
762 | *dec.end = 0; |
1794 | |
|
|
1795 | if (offset_return) |
|
|
1796 | *offset_return = dec.cur - SvPVX (string); |
|
|
1797 | else if (sv) |
|
|
1798 | { |
|
|
1799 | // check for trailing garbage |
|
|
1800 | decode_ws (&dec); |
|
|
1801 | |
|
|
1802 | if (dec.cur != dec.end) |
|
|
1803 | { |
|
|
1804 | dec.err = "garbage after JSON object"; |
|
|
1805 | SvREFCNT_dec (sv); |
|
|
1806 | sv = 0; |
|
|
1807 | } |
|
|
1808 | } |
763 | |
1809 | |
764 | if (!sv) |
1810 | if (!sv) |
765 | { |
1811 | { |
766 | IV offset = utf8_distance (dec.cur, SvPVX (string)); |
|
|
767 | SV *uni = sv_newmortal (); |
1812 | SV *uni = sv_newmortal (); |
768 | |
1813 | |
|
|
1814 | // horrible hack to silence warning inside pv_uni_display |
|
|
1815 | COP cop = *PL_curcop; |
|
|
1816 | cop.cop_warnings = pWARN_NONE; |
|
|
1817 | ENTER; |
|
|
1818 | SAVEVPTR (PL_curcop); |
|
|
1819 | PL_curcop = &cop; |
769 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1820 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
|
|
1821 | LEAVE; |
|
|
1822 | |
770 | croak ("%s, at character %d (%s)", |
1823 | croak ("%s, at character offset %d (before \"%s\")", |
771 | dec.err, |
1824 | dec.err, |
772 | (int)offset, |
1825 | (int)ptr_to_index (string, dec.cur), |
773 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1826 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
774 | } |
1827 | } |
775 | |
1828 | |
776 | sv_dump (sv);//D |
|
|
777 | return sv_2mortal (sv); |
1829 | sv = sv_2mortal (sv); |
|
|
1830 | |
|
|
1831 | if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv)) |
|
|
1832 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
|
|
1833 | |
|
|
1834 | return sv; |
778 | } |
1835 | } |
|
|
1836 | |
|
|
1837 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1838 | // incremental parser |
|
|
1839 | |
|
|
1840 | static void |
|
|
1841 | incr_parse (JSON *self) |
|
|
1842 | { |
|
|
1843 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1844 | |
|
|
1845 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1846 | // but this would make it slower, so... |
|
|
1847 | |
|
|
1848 | for (;;) |
|
|
1849 | { |
|
|
1850 | switch (self->incr_mode) |
|
|
1851 | { |
|
|
1852 | // reached end of a scalar, see if we are inside a nested structure or not |
|
|
1853 | end_of_scalar: |
|
|
1854 | self->incr_mode = INCR_M_JSON; |
|
|
1855 | |
|
|
1856 | if (self->incr_nest) // end of a scalar inside array, object or tag |
|
|
1857 | goto incr_m_json; |
|
|
1858 | else // end of scalar outside structure, json text ends here |
|
|
1859 | goto interrupt; |
|
|
1860 | |
|
|
1861 | // only used for initial whitespace skipping |
|
|
1862 | case INCR_M_WS: |
|
|
1863 | for (;;) |
|
|
1864 | { |
|
|
1865 | if (*p > 0x20) |
|
|
1866 | { |
|
|
1867 | if (*p == '#') |
|
|
1868 | { |
|
|
1869 | self->incr_mode = INCR_M_C0; |
|
|
1870 | goto incr_m_c; |
|
|
1871 | } |
|
|
1872 | else |
|
|
1873 | { |
|
|
1874 | self->incr_mode = INCR_M_JSON; |
|
|
1875 | goto incr_m_json; |
|
|
1876 | } |
|
|
1877 | } |
|
|
1878 | else if (!*p) |
|
|
1879 | goto interrupt; |
|
|
1880 | |
|
|
1881 | ++p; |
|
|
1882 | } |
|
|
1883 | |
|
|
1884 | // skip a single char inside a string (for \\-processing) |
|
|
1885 | case INCR_M_BS: |
|
|
1886 | if (!*p) |
|
|
1887 | goto interrupt; |
|
|
1888 | |
|
|
1889 | ++p; |
|
|
1890 | self->incr_mode = INCR_M_STR; |
|
|
1891 | goto incr_m_str; |
|
|
1892 | |
|
|
1893 | // inside #-style comments |
|
|
1894 | case INCR_M_C0: |
|
|
1895 | case INCR_M_C1: |
|
|
1896 | incr_m_c: |
|
|
1897 | for (;;) |
|
|
1898 | { |
|
|
1899 | if (*p == '\n') |
|
|
1900 | { |
|
|
1901 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1902 | break; |
|
|
1903 | } |
|
|
1904 | else if (!*p) |
|
|
1905 | goto interrupt; |
|
|
1906 | |
|
|
1907 | ++p; |
|
|
1908 | } |
|
|
1909 | |
|
|
1910 | break; |
|
|
1911 | |
|
|
1912 | // inside true/false/null |
|
|
1913 | case INCR_M_TFN: |
|
|
1914 | incr_m_tfn: |
|
|
1915 | for (;;) |
|
|
1916 | switch (*p++) |
|
|
1917 | { |
|
|
1918 | case 'r': case 'u': case 'e': // tRUE, falsE, nUll |
|
|
1919 | case 'a': case 'l': case 's': // fALSe, nuLL |
|
|
1920 | // allowed |
|
|
1921 | break; |
|
|
1922 | |
|
|
1923 | default: |
|
|
1924 | --p; |
|
|
1925 | goto end_of_scalar; |
|
|
1926 | } |
|
|
1927 | |
|
|
1928 | // inside a number |
|
|
1929 | case INCR_M_NUM: |
|
|
1930 | incr_m_num: |
|
|
1931 | for (;;) |
|
|
1932 | switch (*p++) |
|
|
1933 | { |
|
|
1934 | case 'e': case 'E': case '.': case '+': |
|
|
1935 | case '-': |
|
|
1936 | case '0': case '1': case '2': case '3': case '4': |
|
|
1937 | case '5': case '6': case '7': case '8': case '9': |
|
|
1938 | // allowed |
|
|
1939 | break; |
|
|
1940 | |
|
|
1941 | default: |
|
|
1942 | --p; |
|
|
1943 | goto end_of_scalar; |
|
|
1944 | } |
|
|
1945 | |
|
|
1946 | // inside a string |
|
|
1947 | case INCR_M_STR: |
|
|
1948 | incr_m_str: |
|
|
1949 | for (;;) |
|
|
1950 | { |
|
|
1951 | if (*p == '"') |
|
|
1952 | { |
|
|
1953 | ++p; |
|
|
1954 | goto end_of_scalar; |
|
|
1955 | } |
|
|
1956 | else if (*p == '\\') |
|
|
1957 | { |
|
|
1958 | ++p; // "virtually" consumes character after \ |
|
|
1959 | |
|
|
1960 | if (!*p) // if at end of string we have to switch modes |
|
|
1961 | { |
|
|
1962 | self->incr_mode = INCR_M_BS; |
|
|
1963 | goto interrupt; |
|
|
1964 | } |
|
|
1965 | } |
|
|
1966 | else if (!*p) |
|
|
1967 | goto interrupt; |
|
|
1968 | |
|
|
1969 | ++p; |
|
|
1970 | } |
|
|
1971 | |
|
|
1972 | // after initial ws, outside string |
|
|
1973 | case INCR_M_JSON: |
|
|
1974 | incr_m_json: |
|
|
1975 | for (;;) |
|
|
1976 | { |
|
|
1977 | switch (*p++) |
|
|
1978 | { |
|
|
1979 | case 0: |
|
|
1980 | --p; |
|
|
1981 | goto interrupt; |
|
|
1982 | |
|
|
1983 | case 0x09: |
|
|
1984 | case 0x0a: |
|
|
1985 | case 0x0d: |
|
|
1986 | case 0x20: |
|
|
1987 | if (!self->incr_nest) |
|
|
1988 | { |
|
|
1989 | --p; // do not eat the whitespace, let the next round do it |
|
|
1990 | goto interrupt; |
|
|
1991 | } |
|
|
1992 | break; |
|
|
1993 | |
|
|
1994 | // the following three blocks handle scalars. this makes the parser |
|
|
1995 | // more strict than required inside arrays or objects, and could |
|
|
1996 | // be moved to a special case on the toplevel (except strings) |
|
|
1997 | case 't': |
|
|
1998 | case 'f': |
|
|
1999 | case 'n': |
|
|
2000 | self->incr_mode = INCR_M_TFN; |
|
|
2001 | goto incr_m_tfn; |
|
|
2002 | |
|
|
2003 | case '-': |
|
|
2004 | case '0': case '1': case '2': case '3': case '4': |
|
|
2005 | case '5': case '6': case '7': case '8': case '9': |
|
|
2006 | self->incr_mode = INCR_M_NUM; |
|
|
2007 | goto incr_m_num; |
|
|
2008 | |
|
|
2009 | case '"': |
|
|
2010 | self->incr_mode = INCR_M_STR; |
|
|
2011 | goto incr_m_str; |
|
|
2012 | |
|
|
2013 | case '[': |
|
|
2014 | case '{': |
|
|
2015 | case '(': |
|
|
2016 | if (++self->incr_nest > self->max_depth) |
|
|
2017 | croak (ERR_NESTING_EXCEEDED); |
|
|
2018 | break; |
|
|
2019 | |
|
|
2020 | case ']': |
|
|
2021 | case '}': |
|
|
2022 | if (--self->incr_nest <= 0) |
|
|
2023 | goto interrupt; |
|
|
2024 | break; |
|
|
2025 | |
|
|
2026 | case ')': |
|
|
2027 | --self->incr_nest; |
|
|
2028 | break; |
|
|
2029 | |
|
|
2030 | case '#': |
|
|
2031 | self->incr_mode = INCR_M_C1; |
|
|
2032 | goto incr_m_c; |
|
|
2033 | } |
|
|
2034 | } |
|
|
2035 | } |
|
|
2036 | |
|
|
2037 | modechange: |
|
|
2038 | ; |
|
|
2039 | } |
|
|
2040 | |
|
|
2041 | interrupt: |
|
|
2042 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
2043 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
2044 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
|
|
2045 | } |
|
|
2046 | |
|
|
2047 | ///////////////////////////////////////////////////////////////////////////// |
|
|
2048 | // XS interface functions |
779 | |
2049 | |
780 | MODULE = JSON::XS PACKAGE = JSON::XS |
2050 | MODULE = JSON::XS PACKAGE = JSON::XS |
781 | |
2051 | |
782 | BOOT: |
2052 | BOOT: |
783 | { |
2053 | { |
784 | int i; |
2054 | int i; |
785 | |
2055 | |
786 | memset (decode_hexdigit, 0xff, 256); |
|
|
787 | for (i = 10; i--; ) |
2056 | for (i = 0; i < 256; ++i) |
788 | decode_hexdigit ['0' + i] = i; |
2057 | decode_hexdigit [i] = |
|
|
2058 | i >= '0' && i <= '9' ? i - '0' |
|
|
2059 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
|
|
2060 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
|
|
2061 | : -1; |
789 | |
2062 | |
790 | for (i = 6; --i; ) |
2063 | json_stash = gv_stashpv ("JSON::XS" , 1); |
791 | { |
2064 | bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1); |
792 | decode_hexdigit ['a' + i] = 10 + i; |
2065 | bool_false = get_bool ("Types::Serialiser::false"); |
793 | decode_hexdigit ['A' + i] = 10 + i; |
2066 | bool_true = get_bool ("Types::Serialiser::true"); |
|
|
2067 | |
|
|
2068 | sv_json = newSVpv ("JSON", 0); |
|
|
2069 | SvREADONLY_on (sv_json); |
|
|
2070 | |
|
|
2071 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
|
|
2072 | } |
|
|
2073 | |
|
|
2074 | PROTOTYPES: DISABLE |
|
|
2075 | |
|
|
2076 | void CLONE (...) |
|
|
2077 | CODE: |
|
|
2078 | // as long as these writes are atomic, the race should not matter |
|
|
2079 | // as existing threads either already use 0, or use the old value, |
|
|
2080 | // which is sitll correct for the initial thread. |
|
|
2081 | json_stash = 0; |
|
|
2082 | bool_stash = 0; |
|
|
2083 | bool_false = 0; |
|
|
2084 | bool_true = 0; |
|
|
2085 | |
|
|
2086 | void new (char *klass) |
|
|
2087 | PPCODE: |
|
|
2088 | { |
|
|
2089 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
2090 | SvPOK_only (pv); |
|
|
2091 | json_init ((JSON *)SvPVX (pv)); |
|
|
2092 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
2093 | newRV_noinc (pv), |
|
|
2094 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
2095 | ))); |
|
|
2096 | } |
|
|
2097 | |
|
|
2098 | void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0) |
|
|
2099 | PPCODE: |
|
|
2100 | self->v_false = newSVsv (v_false); |
|
|
2101 | self->v_true = newSVsv (v_true); |
|
|
2102 | XPUSHs (ST (0)); |
|
|
2103 | |
|
|
2104 | void get_boolean_values (JSON *self) |
|
|
2105 | PPCODE: |
|
|
2106 | if (self->v_false && self->v_true) |
|
|
2107 | { |
|
|
2108 | EXTEND (SP, 2); |
|
|
2109 | PUSHs (self->v_false); |
|
|
2110 | PUSHs (self->v_true); |
794 | } |
2111 | } |
795 | |
2112 | |
796 | json_stash = gv_stashpv ("JSON::XS", 1); |
2113 | void ascii (JSON *self, int enable = 1) |
797 | } |
2114 | ALIAS: |
|
|
2115 | ascii = F_ASCII |
|
|
2116 | latin1 = F_LATIN1 |
|
|
2117 | utf8 = F_UTF8 |
|
|
2118 | indent = F_INDENT |
|
|
2119 | canonical = F_CANONICAL |
|
|
2120 | space_before = F_SPACE_BEFORE |
|
|
2121 | space_after = F_SPACE_AFTER |
|
|
2122 | pretty = F_PRETTY |
|
|
2123 | allow_nonref = F_ALLOW_NONREF |
|
|
2124 | shrink = F_SHRINK |
|
|
2125 | allow_blessed = F_ALLOW_BLESSED |
|
|
2126 | convert_blessed = F_CONV_BLESSED |
|
|
2127 | relaxed = F_RELAXED |
|
|
2128 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
2129 | allow_tags = F_ALLOW_TAGS |
|
|
2130 | PPCODE: |
|
|
2131 | { |
|
|
2132 | if (enable) |
|
|
2133 | self->flags |= ix; |
|
|
2134 | else |
|
|
2135 | self->flags &= ~ix; |
798 | |
2136 | |
799 | SV *new (char *dummy) |
2137 | XPUSHs (ST (0)); |
|
|
2138 | } |
|
|
2139 | |
|
|
2140 | void get_ascii (JSON *self) |
|
|
2141 | ALIAS: |
|
|
2142 | get_ascii = F_ASCII |
|
|
2143 | get_latin1 = F_LATIN1 |
|
|
2144 | get_utf8 = F_UTF8 |
|
|
2145 | get_indent = F_INDENT |
|
|
2146 | get_canonical = F_CANONICAL |
|
|
2147 | get_space_before = F_SPACE_BEFORE |
|
|
2148 | get_space_after = F_SPACE_AFTER |
|
|
2149 | get_allow_nonref = F_ALLOW_NONREF |
|
|
2150 | get_shrink = F_SHRINK |
|
|
2151 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
2152 | get_convert_blessed = F_CONV_BLESSED |
|
|
2153 | get_relaxed = F_RELAXED |
|
|
2154 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
2155 | get_allow_tags = F_ALLOW_TAGS |
|
|
2156 | PPCODE: |
|
|
2157 | XPUSHs (boolSV (self->flags & ix)); |
|
|
2158 | |
|
|
2159 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
|
|
2160 | PPCODE: |
|
|
2161 | self->max_depth = max_depth; |
|
|
2162 | XPUSHs (ST (0)); |
|
|
2163 | |
|
|
2164 | U32 get_max_depth (JSON *self) |
800 | CODE: |
2165 | CODE: |
801 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
2166 | RETVAL = self->max_depth; |
802 | OUTPUT: |
2167 | OUTPUT: |
803 | RETVAL |
2168 | RETVAL |
804 | |
2169 | |
805 | SV *ascii (SV *self, int enable) |
2170 | void max_size (JSON *self, U32 max_size = 0) |
806 | ALIAS: |
2171 | PPCODE: |
807 | ascii = F_ASCII |
2172 | self->max_size = max_size; |
808 | utf8 = F_UTF8 |
2173 | XPUSHs (ST (0)); |
809 | indent = F_INDENT |
2174 | |
810 | canonical = F_CANONICAL |
2175 | int get_max_size (JSON *self) |
811 | space_before = F_SPACE_BEFORE |
|
|
812 | space_after = F_SPACE_AFTER |
|
|
813 | json_rpc = F_JSON_RPC |
|
|
814 | CODE: |
2176 | CODE: |
815 | { |
2177 | RETVAL = self->max_size; |
816 | UV *uv = SvJSON (self); |
|
|
817 | if (enable) |
|
|
818 | *uv |= ix; |
|
|
819 | else |
|
|
820 | *uv &= ~ix; |
|
|
821 | |
|
|
822 | RETVAL = newSVsv (self); |
|
|
823 | } |
|
|
824 | OUTPUT: |
2178 | OUTPUT: |
825 | RETVAL |
2179 | RETVAL |
826 | |
2180 | |
827 | void encode (SV *self, SV *scalar) |
2181 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
828 | PPCODE: |
2182 | PPCODE: |
829 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
2183 | { |
|
|
2184 | SvREFCNT_dec (self->cb_object); |
|
|
2185 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
830 | |
2186 | |
831 | void decode (SV *self, SV *jsondata) |
2187 | XPUSHs (ST (0)); |
|
|
2188 | } |
|
|
2189 | |
|
|
2190 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
832 | PPCODE: |
2191 | PPCODE: |
833 | XPUSHs (decode_json (jsondata, *SvJSON (self))); |
2192 | { |
|
|
2193 | if (!self->cb_sk_object) |
|
|
2194 | self->cb_sk_object = newHV (); |
834 | |
2195 | |
|
|
2196 | if (SvOK (cb)) |
|
|
2197 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
2198 | else |
|
|
2199 | { |
|
|
2200 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
2201 | |
|
|
2202 | if (!HvKEYS (self->cb_sk_object)) |
|
|
2203 | { |
|
|
2204 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2205 | self->cb_sk_object = 0; |
|
|
2206 | } |
|
|
2207 | } |
|
|
2208 | |
|
|
2209 | XPUSHs (ST (0)); |
|
|
2210 | } |
|
|
2211 | |
|
|
2212 | void encode (JSON *self, SV *scalar) |
|
|
2213 | PPCODE: |
|
|
2214 | PUTBACK; scalar = encode_json (scalar, self); SPAGAIN; |
|
|
2215 | XPUSHs (scalar); |
|
|
2216 | |
|
|
2217 | void decode (JSON *self, SV *jsonstr) |
|
|
2218 | PPCODE: |
|
|
2219 | PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN; |
|
|
2220 | XPUSHs (jsonstr); |
|
|
2221 | |
|
|
2222 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
2223 | PPCODE: |
|
|
2224 | { |
|
|
2225 | SV *sv; |
|
|
2226 | STRLEN offset; |
|
|
2227 | PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN; |
|
|
2228 | EXTEND (SP, 2); |
|
|
2229 | PUSHs (sv); |
|
|
2230 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset)))); |
|
|
2231 | } |
|
|
2232 | |
|
|
2233 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
2234 | PPCODE: |
|
|
2235 | { |
|
|
2236 | if (!self->incr_text) |
|
|
2237 | self->incr_text = newSVpvn ("", 0); |
|
|
2238 | |
|
|
2239 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
2240 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
2241 | if (DECODE_WANTS_OCTETS (self)) |
|
|
2242 | { |
|
|
2243 | if (self->incr_pos) |
|
|
2244 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
2245 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
2246 | |
|
|
2247 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
2248 | } |
|
|
2249 | else |
|
|
2250 | { |
|
|
2251 | sv_utf8_upgrade (self->incr_text); |
|
|
2252 | |
|
|
2253 | if (self->incr_pos) |
|
|
2254 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
2255 | - (U8 *)SvPVX (self->incr_text); |
|
|
2256 | } |
|
|
2257 | |
|
|
2258 | // append data, if any |
|
|
2259 | if (jsonstr) |
|
|
2260 | { |
|
|
2261 | /* make sure both strings have same encoding */ |
|
|
2262 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
|
|
2263 | if (SvUTF8 (jsonstr)) |
|
|
2264 | sv_utf8_downgrade (jsonstr, 0); |
|
|
2265 | else |
|
|
2266 | sv_utf8_upgrade (jsonstr); |
|
|
2267 | |
|
|
2268 | /* and then just blindly append */ |
|
|
2269 | { |
|
|
2270 | STRLEN len; |
|
|
2271 | const char *str = SvPV (jsonstr, len); |
|
|
2272 | STRLEN cur = SvCUR (self->incr_text); |
|
|
2273 | |
|
|
2274 | if (SvLEN (self->incr_text) - cur <= len) |
|
|
2275 | json_sv_grow (self->incr_text, cur, len); |
|
|
2276 | |
|
|
2277 | Move (str, SvEND (self->incr_text), len, char); |
|
|
2278 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
2279 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
2280 | } |
|
|
2281 | } |
|
|
2282 | |
|
|
2283 | if (GIMME_V != G_VOID) |
|
|
2284 | do |
|
|
2285 | { |
|
|
2286 | SV *sv; |
|
|
2287 | STRLEN offset; |
|
|
2288 | |
|
|
2289 | if (!INCR_DONE (self)) |
|
|
2290 | { |
|
|
2291 | incr_parse (self); |
|
|
2292 | |
|
|
2293 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
2294 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
2295 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
2296 | |
|
|
2297 | if (!INCR_DONE (self)) |
|
|
2298 | { |
|
|
2299 | // as an optimisation, do not accumulate white space in the incr buffer |
|
|
2300 | if (self->incr_mode == INCR_M_WS && self->incr_pos) |
|
|
2301 | { |
|
|
2302 | self->incr_pos = 0; |
|
|
2303 | SvCUR_set (self->incr_text, 0); |
|
|
2304 | } |
|
|
2305 | |
|
|
2306 | break; |
|
|
2307 | } |
|
|
2308 | } |
|
|
2309 | |
|
|
2310 | PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN; |
|
|
2311 | XPUSHs (sv); |
|
|
2312 | |
|
|
2313 | self->incr_pos -= offset; |
|
|
2314 | self->incr_nest = 0; |
|
|
2315 | self->incr_mode = 0; |
|
|
2316 | |
|
|
2317 | sv_chop (self->incr_text, SvPVX (self->incr_text) + offset); |
|
|
2318 | } |
|
|
2319 | while (GIMME_V == G_ARRAY); |
|
|
2320 | } |
|
|
2321 | |
|
|
2322 | SV *incr_text (JSON *self) |
|
|
2323 | ATTRS: lvalue |
|
|
2324 | CODE: |
|
|
2325 | { |
|
|
2326 | if (self->incr_pos) |
|
|
2327 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
2328 | |
|
|
2329 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2330 | } |
|
|
2331 | OUTPUT: |
|
|
2332 | RETVAL |
|
|
2333 | |
|
|
2334 | void incr_skip (JSON *self) |
|
|
2335 | CODE: |
|
|
2336 | { |
|
|
2337 | if (self->incr_pos) |
|
|
2338 | { |
|
|
2339 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2340 | self->incr_pos = 0; |
|
|
2341 | self->incr_nest = 0; |
|
|
2342 | self->incr_mode = 0; |
|
|
2343 | } |
|
|
2344 | } |
|
|
2345 | |
|
|
2346 | void incr_reset (JSON *self) |
|
|
2347 | CODE: |
|
|
2348 | { |
|
|
2349 | SvREFCNT_dec (self->incr_text); |
|
|
2350 | self->incr_text = 0; |
|
|
2351 | self->incr_pos = 0; |
|
|
2352 | self->incr_nest = 0; |
|
|
2353 | self->incr_mode = 0; |
|
|
2354 | } |
|
|
2355 | |
|
|
2356 | void DESTROY (JSON *self) |
|
|
2357 | CODE: |
|
|
2358 | SvREFCNT_dec (self->v_false); |
|
|
2359 | SvREFCNT_dec (self->v_true); |
|
|
2360 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2361 | SvREFCNT_dec (self->cb_object); |
|
|
2362 | SvREFCNT_dec (self->incr_text); |
|
|
2363 | |
|
|
2364 | PROTOTYPES: ENABLE |
|
|
2365 | |
|
|
2366 | void encode_json (SV *scalar) |
|
|
2367 | PPCODE: |
|
|
2368 | { |
|
|
2369 | JSON json; |
|
|
2370 | json_init (&json); |
|
|
2371 | json.flags |= F_UTF8; |
|
|
2372 | PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN; |
|
|
2373 | XPUSHs (scalar); |
|
|
2374 | } |
|
|
2375 | |
|
|
2376 | void decode_json (SV *jsonstr) |
|
|
2377 | PPCODE: |
|
|
2378 | { |
|
|
2379 | JSON json; |
|
|
2380 | json_init (&json); |
|
|
2381 | json.flags |= F_UTF8; |
|
|
2382 | PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN; |
|
|
2383 | XPUSHs (jsonstr); |
|
|
2384 | } |
|
|
2385 | |