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> |
8 | |
11 | |
|
|
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
|
|
14 | #endif |
|
|
15 | |
|
|
16 | // some old perls do not have this, try to make it work, no |
|
|
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
|
|
18 | #ifndef UTF8_MAXBYTES |
|
|
19 | # define UTF8_MAXBYTES 13 |
|
|
20 | #endif |
|
|
21 | |
|
|
22 | // three extra for rounding, sign, and end of string |
|
|
23 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
|
|
24 | |
9 | #define F_ASCII 0x00000001 |
25 | #define F_ASCII 0x00000001UL |
|
|
26 | #define F_LATIN1 0x00000002UL |
10 | #define F_UTF8 0x00000002 |
27 | #define F_UTF8 0x00000004UL |
11 | #define F_INDENT 0x00000004 |
28 | #define F_INDENT 0x00000008UL |
12 | #define F_CANONICAL 0x00000008 |
29 | #define F_CANONICAL 0x00000010UL |
13 | #define F_SPACE_BEFORE 0x00000010 |
30 | #define F_SPACE_BEFORE 0x00000020UL |
14 | #define F_SPACE_AFTER 0x00000020 |
31 | #define F_SPACE_AFTER 0x00000040UL |
15 | #define F_JSON_RPC 0x00000040 |
32 | #define F_ALLOW_NONREF 0x00000100UL |
|
|
33 | #define F_SHRINK 0x00000200UL |
|
|
34 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
35 | #define F_CONV_BLESSED 0x00000800UL |
|
|
36 | #define F_RELAXED 0x00001000UL |
|
|
37 | #define F_ALLOW_UNKNOWN 0x00002000UL |
|
|
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
16 | |
39 | |
17 | #define F_DEFAULT 0 |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
18 | |
41 | |
19 | #define INIT_SIZE 32 // initial scalar size to be allocated |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
|
|
43 | #define INDENT_STEP 3 // spaces per indentation level |
|
|
44 | |
|
|
45 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
20 | |
46 | |
21 | #define SB do { |
47 | #define SB do { |
22 | #define SE } while (0) |
48 | #define SE } while (0) |
23 | |
49 | |
24 | static HV *json_stash; |
50 | #if __GNUC__ >= 3 |
|
|
51 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
|
|
52 | # define INLINE static inline |
|
|
53 | #else |
|
|
54 | # define expect(expr,value) (expr) |
|
|
55 | # define INLINE static |
|
|
56 | #endif |
|
|
57 | |
|
|
58 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
59 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
60 | |
|
|
61 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
62 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
63 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
64 | |
|
|
65 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
66 | |
|
|
67 | #ifdef USE_ITHREADS |
|
|
68 | # define JSON_SLOW 1 |
|
|
69 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
70 | #else |
|
|
71 | # define JSON_SLOW 0 |
|
|
72 | # define JSON_STASH json_stash |
|
|
73 | #endif |
|
|
74 | |
|
|
75 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
|
|
76 | static SV *json_true, *json_false; |
|
|
77 | |
|
|
78 | enum { |
|
|
79 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
80 | INCR_M_STR, // inside string |
|
|
81 | INCR_M_BS, // inside backslash |
|
|
82 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
83 | INCR_M_C1, // inside comment in other places |
|
|
84 | INCR_M_JSON // outside anything, count nesting |
|
|
85 | }; |
|
|
86 | |
|
|
87 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
88 | |
|
|
89 | typedef struct { |
|
|
90 | U32 flags; |
|
|
91 | U32 max_depth; |
|
|
92 | STRLEN max_size; |
|
|
93 | |
|
|
94 | SV *cb_object; |
|
|
95 | HV *cb_sk_object; |
|
|
96 | |
|
|
97 | // for the incremental parser |
|
|
98 | SV *incr_text; // the source text so far |
|
|
99 | STRLEN incr_pos; // the current offset into the text |
|
|
100 | int incr_nest; // {[]}-nesting level |
|
|
101 | unsigned char incr_mode; |
|
|
102 | } JSON; |
|
|
103 | |
|
|
104 | INLINE void |
|
|
105 | json_init (JSON *json) |
|
|
106 | { |
|
|
107 | Zero (json, 1, JSON); |
|
|
108 | json->max_depth = 512; |
|
|
109 | } |
|
|
110 | |
|
|
111 | ///////////////////////////////////////////////////////////////////////////// |
|
|
112 | // utility functions |
|
|
113 | |
|
|
114 | INLINE SV * |
|
|
115 | get_bool (const char *name) |
|
|
116 | { |
|
|
117 | SV *sv = get_sv (name, 1); |
|
|
118 | |
|
|
119 | SvREADONLY_on (sv); |
|
|
120 | SvREADONLY_on (SvRV (sv)); |
|
|
121 | |
|
|
122 | return sv; |
|
|
123 | } |
|
|
124 | |
|
|
125 | INLINE void |
|
|
126 | shrink (SV *sv) |
|
|
127 | { |
|
|
128 | sv_utf8_downgrade (sv, 1); |
|
|
129 | |
|
|
130 | if (SvLEN (sv) > SvCUR (sv) + 1) |
|
|
131 | { |
|
|
132 | #ifdef SvPV_shrink_to_cur |
|
|
133 | SvPV_shrink_to_cur (sv); |
|
|
134 | #elif defined (SvPV_renew) |
|
|
135 | SvPV_renew (sv, SvCUR (sv) + 1); |
|
|
136 | #endif |
|
|
137 | } |
|
|
138 | } |
|
|
139 | |
|
|
140 | // decode an utf-8 character and return it, or (UV)-1 in |
|
|
141 | // case of an error. |
|
|
142 | // we special-case "safe" characters from U+80 .. U+7FF, |
|
|
143 | // but use the very good perl function to parse anything else. |
|
|
144 | // note that we never call this function for a ascii codepoints |
|
|
145 | INLINE UV |
|
|
146 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
|
|
147 | { |
|
|
148 | if (expect_true (len >= 2 |
|
|
149 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
|
|
150 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
|
|
151 | { |
|
|
152 | *clen = 2; |
|
|
153 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
|
|
154 | } |
|
|
155 | else |
|
|
156 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
|
|
157 | } |
|
|
158 | |
|
|
159 | // likewise for encoding, also never called for ascii codepoints |
|
|
160 | // this function takes advantage of this fact, although current gccs |
|
|
161 | // seem to optimise the check for >= 0x80 away anyways |
|
|
162 | INLINE unsigned char * |
|
|
163 | encode_utf8 (unsigned char *s, UV ch) |
|
|
164 | { |
|
|
165 | if (expect_false (ch < 0x000080)) |
|
|
166 | *s++ = ch; |
|
|
167 | else if (expect_true (ch < 0x000800)) |
|
|
168 | *s++ = 0xc0 | ( ch >> 6), |
|
|
169 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
170 | else if ( ch < 0x010000) |
|
|
171 | *s++ = 0xe0 | ( ch >> 12), |
|
|
172 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
173 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
174 | else if ( ch < 0x110000) |
|
|
175 | *s++ = 0xf0 | ( ch >> 18), |
|
|
176 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
177 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
178 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
179 | |
|
|
180 | return s; |
|
|
181 | } |
|
|
182 | |
|
|
183 | // convert offset pointer to character index, sv must be string |
|
|
184 | static STRLEN |
|
|
185 | ptr_to_index (SV *sv, char *offset) |
|
|
186 | { |
|
|
187 | return SvUTF8 (sv) |
|
|
188 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
189 | : offset - SvPVX (sv); |
|
|
190 | } |
|
|
191 | |
|
|
192 | ///////////////////////////////////////////////////////////////////////////// |
|
|
193 | // fp hell |
|
|
194 | |
|
|
195 | // scan a group of digits, and a trailing exponent |
|
|
196 | static void |
|
|
197 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
198 | { |
|
|
199 | UV uaccum = 0; |
|
|
200 | int eaccum = 0; |
|
|
201 | |
|
|
202 | // if we recurse too deep, skip all remaining digits |
|
|
203 | // to avoid a stack overflow attack |
|
|
204 | if (expect_false (--maxdepth <= 0)) |
|
|
205 | while (((U8)*s - '0') < 10) |
|
|
206 | ++s; |
|
|
207 | |
|
|
208 | for (;;) |
|
|
209 | { |
|
|
210 | U8 dig = (U8)*s - '0'; |
|
|
211 | |
|
|
212 | if (expect_false (dig >= 10)) |
|
|
213 | { |
|
|
214 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
215 | { |
|
|
216 | ++s; |
|
|
217 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
218 | } |
|
|
219 | else if ((dig | ' ') == 'e' - '0') |
|
|
220 | { |
|
|
221 | int exp2 = 0; |
|
|
222 | int neg = 0; |
|
|
223 | |
|
|
224 | ++s; |
|
|
225 | |
|
|
226 | if (*s == '-') |
|
|
227 | { |
|
|
228 | ++s; |
|
|
229 | neg = 1; |
|
|
230 | } |
|
|
231 | else if (*s == '+') |
|
|
232 | ++s; |
|
|
233 | |
|
|
234 | while ((dig = (U8)*s - '0') < 10) |
|
|
235 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
236 | |
|
|
237 | *expo += neg ? -exp2 : exp2; |
|
|
238 | } |
|
|
239 | |
|
|
240 | break; |
|
|
241 | } |
|
|
242 | |
|
|
243 | ++s; |
|
|
244 | |
|
|
245 | uaccum = uaccum * 10 + dig; |
|
|
246 | ++eaccum; |
|
|
247 | |
|
|
248 | // if we have too many digits, then recurse for more |
|
|
249 | // we actually do this for rather few digits |
|
|
250 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
251 | { |
|
|
252 | if (postdp) *expo -= eaccum; |
|
|
253 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
254 | if (postdp) *expo += eaccum; |
|
|
255 | |
|
|
256 | break; |
|
|
257 | } |
|
|
258 | } |
|
|
259 | |
|
|
260 | // this relies greatly on the quality of the pow () |
|
|
261 | // implementation of the platform, but a good |
|
|
262 | // implementation is hard to beat. |
|
|
263 | if (postdp) *expo -= eaccum; |
|
|
264 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
265 | *expo += eaccum; |
|
|
266 | } |
|
|
267 | |
|
|
268 | static NV |
|
|
269 | json_atof (const char *s) |
|
|
270 | { |
|
|
271 | NV accum = 0.; |
|
|
272 | int expo = 0; |
|
|
273 | int neg = 0; |
|
|
274 | |
|
|
275 | if (*s == '-') |
|
|
276 | { |
|
|
277 | ++s; |
|
|
278 | neg = 1; |
|
|
279 | } |
|
|
280 | |
|
|
281 | // a recursion depth of ten gives us >>500 bits |
|
|
282 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
283 | |
|
|
284 | return neg ? -accum : accum; |
|
|
285 | } |
|
|
286 | ///////////////////////////////////////////////////////////////////////////// |
|
|
287 | // encoder |
25 | |
288 | |
26 | // structure used for encoding JSON |
289 | // structure used for encoding JSON |
27 | typedef struct |
290 | typedef struct |
28 | { |
291 | { |
29 | char *cur; |
292 | char *cur; // SvPVX (sv) + current output position |
30 | STRLEN len; // SvLEN (sv) |
|
|
31 | char *end; // SvEND (sv) |
293 | char *end; // SvEND (sv) |
32 | SV *sv; |
294 | SV *sv; // result scalar |
33 | UV flags; |
295 | JSON json; |
34 | int max_recurse; |
296 | U32 indent; // indentation level |
35 | int indent; |
297 | UV limit; // escape character values >= this value when encoding |
36 | } enc_t; |
298 | } enc_t; |
37 | |
299 | |
38 | // structure used for decoding JSON |
300 | 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) |
301 | need (enc_t *enc, STRLEN len) |
60 | { |
302 | { |
61 | if (enc->cur + len >= enc->end) |
303 | if (expect_false (enc->cur + len >= enc->end)) |
62 | { |
304 | { |
63 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
305 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
64 | SvGROW (enc->sv, cur + len + 1); |
306 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
65 | enc->cur = SvPVX (enc->sv) + cur; |
307 | enc->cur = SvPVX (enc->sv) + cur; |
66 | enc->end = SvEND (enc->sv); |
308 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
67 | } |
309 | } |
68 | } |
310 | } |
69 | |
311 | |
70 | static void |
312 | INLINE void |
71 | encode_ch (enc_t *enc, char ch) |
313 | encode_ch (enc_t *enc, char ch) |
72 | { |
314 | { |
73 | need (enc, 1); |
315 | need (enc, 1); |
74 | *enc->cur++ = ch; |
316 | *enc->cur++ = ch; |
75 | } |
317 | } |
… | |
… | |
77 | static void |
319 | static void |
78 | encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
320 | encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
79 | { |
321 | { |
80 | char *end = str + len; |
322 | char *end = str + len; |
81 | |
323 | |
|
|
324 | need (enc, len); |
|
|
325 | |
82 | while (str < end) |
326 | while (str < end) |
83 | { |
327 | { |
84 | unsigned char ch = *(unsigned char *)str; |
328 | unsigned char ch = *(unsigned char *)str; |
|
|
329 | |
85 | if (ch >= 0x20 && ch < 0x80) // most common case |
330 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
86 | { |
331 | { |
|
|
332 | if (expect_false (ch == '"')) // but with slow exceptions |
|
|
333 | { |
|
|
334 | need (enc, len += 1); |
|
|
335 | *enc->cur++ = '\\'; |
|
|
336 | *enc->cur++ = '"'; |
|
|
337 | } |
|
|
338 | else if (expect_false (ch == '\\')) |
|
|
339 | { |
|
|
340 | need (enc, len += 1); |
|
|
341 | *enc->cur++ = '\\'; |
|
|
342 | *enc->cur++ = '\\'; |
|
|
343 | } |
|
|
344 | else |
87 | *enc->cur++ = ch; |
345 | *enc->cur++ = ch; |
|
|
346 | |
88 | str++; |
347 | ++str; |
89 | } |
348 | } |
90 | else |
349 | else |
91 | { |
350 | { |
92 | STRLEN clen; |
351 | switch (ch) |
93 | UV uch; |
|
|
94 | |
|
|
95 | if (is_utf8) |
|
|
96 | { |
352 | { |
97 | uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
353 | case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
98 | if (clen < 0) |
354 | 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"); |
355 | case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
100 | } |
356 | case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
101 | else |
357 | case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
102 | { |
|
|
103 | uch = ch; |
|
|
104 | clen = 1; |
|
|
105 | } |
|
|
106 | |
358 | |
107 | need (enc, len += 6); |
359 | default: |
108 | |
|
|
109 | if (uch < 0xa0 || enc->flags & F_ASCII) |
|
|
110 | { |
|
|
111 | if (uch > 0xFFFFUL) |
|
|
112 | { |
360 | { |
|
|
361 | STRLEN clen; |
|
|
362 | UV uch; |
|
|
363 | |
|
|
364 | if (is_utf8) |
|
|
365 | { |
|
|
366 | uch = decode_utf8 (str, end - str, &clen); |
|
|
367 | if (clen == (STRLEN)-1) |
|
|
368 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
|
|
369 | } |
|
|
370 | else |
|
|
371 | { |
|
|
372 | uch = ch; |
113 | len += 6; |
373 | clen = 1; |
|
|
374 | } |
|
|
375 | |
|
|
376 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
|
|
377 | { |
|
|
378 | if (uch >= 0x10000UL) |
|
|
379 | { |
|
|
380 | if (uch >= 0x110000UL) |
|
|
381 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
382 | |
114 | need (enc, len += 6); |
383 | need (enc, len += 11); |
115 | sprintf (enc->cur, "\\u%04x\\u%04x", |
384 | sprintf (enc->cur, "\\u%04x\\u%04x", |
116 | (uch - 0x10000) / 0x400 + 0xD800, |
385 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
117 | (uch - 0x10000) % 0x400 + 0xDC00); |
386 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
118 | enc->cur += 12; |
387 | enc->cur += 12; |
|
|
388 | } |
|
|
389 | else |
|
|
390 | { |
|
|
391 | need (enc, len += 5); |
|
|
392 | *enc->cur++ = '\\'; |
|
|
393 | *enc->cur++ = 'u'; |
|
|
394 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
|
|
395 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
|
|
396 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
|
|
397 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
|
|
398 | } |
|
|
399 | |
|
|
400 | str += clen; |
119 | } |
401 | } |
|
|
402 | else if (enc->json.flags & F_LATIN1) |
|
|
403 | { |
|
|
404 | *enc->cur++ = uch; |
|
|
405 | str += clen; |
|
|
406 | } |
|
|
407 | else if (is_utf8) |
|
|
408 | { |
|
|
409 | need (enc, len += clen); |
|
|
410 | do |
|
|
411 | { |
|
|
412 | *enc->cur++ = *str++; |
|
|
413 | } |
|
|
414 | while (--clen); |
|
|
415 | } |
120 | else |
416 | else |
121 | { |
417 | { |
122 | sprintf (enc->cur, "\\u%04x", uch); |
418 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
123 | enc->cur += 6; |
419 | enc->cur = encode_utf8 (enc->cur, uch); |
|
|
420 | ++str; |
|
|
421 | } |
124 | } |
422 | } |
125 | } |
423 | } |
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 | } |
424 | } |
136 | |
425 | |
137 | --len; |
426 | --len; |
138 | } |
427 | } |
139 | } |
428 | } |
140 | |
429 | |
141 | #define INDENT SB \ |
430 | INLINE void |
|
|
431 | encode_indent (enc_t *enc) |
|
|
432 | { |
142 | if (enc->flags & F_INDENT) \ |
433 | if (enc->json.flags & F_INDENT) |
143 | { \ |
434 | { |
144 | int i_; \ |
435 | int spaces = enc->indent * INDENT_STEP; |
145 | need (enc, enc->indent); \ |
436 | |
146 | for (i_ = enc->indent * 3; i_--; )\ |
437 | need (enc, spaces); |
|
|
438 | memset (enc->cur, ' ', spaces); |
|
|
439 | enc->cur += spaces; |
|
|
440 | } |
|
|
441 | } |
|
|
442 | |
|
|
443 | INLINE void |
|
|
444 | encode_space (enc_t *enc) |
|
|
445 | { |
|
|
446 | need (enc, 1); |
|
|
447 | encode_ch (enc, ' '); |
|
|
448 | } |
|
|
449 | |
|
|
450 | INLINE void |
|
|
451 | encode_nl (enc_t *enc) |
|
|
452 | { |
|
|
453 | if (enc->json.flags & F_INDENT) |
|
|
454 | { |
|
|
455 | need (enc, 1); |
147 | encode_ch (enc, ' '); \ |
456 | encode_ch (enc, '\n'); |
148 | } \ |
457 | } |
149 | SE |
458 | } |
150 | |
459 | |
151 | #define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE |
460 | INLINE void |
152 | #define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE |
461 | encode_comma (enc_t *enc) |
153 | #define COMMA SB \ |
462 | { |
154 | encode_ch (enc, ','); \ |
463 | encode_ch (enc, ','); |
|
|
464 | |
155 | if (enc->flags & F_INDENT) \ |
465 | if (enc->json.flags & F_INDENT) |
156 | NL; \ |
466 | encode_nl (enc); |
157 | else if (enc->flags & F_SPACE_AFTER) \ |
467 | else if (enc->json.flags & F_SPACE_AFTER) |
158 | SPACE; \ |
468 | encode_space (enc); |
159 | SE |
469 | } |
160 | |
470 | |
161 | static void encode_sv (enc_t *enc, SV *sv); |
471 | static void encode_sv (enc_t *enc, SV *sv); |
162 | |
472 | |
163 | static void |
473 | static void |
164 | encode_av (enc_t *enc, AV *av) |
474 | encode_av (enc_t *enc, AV *av) |
165 | { |
475 | { |
166 | int i, len = av_len (av); |
476 | int i, len = av_len (av); |
167 | |
477 | |
|
|
478 | if (enc->indent >= enc->json.max_depth) |
|
|
479 | croak (ERR_NESTING_EXCEEDED); |
|
|
480 | |
168 | encode_ch (enc, '['); NL; |
481 | encode_ch (enc, '['); |
169 | ++enc->indent; |
482 | |
|
|
483 | if (len >= 0) |
|
|
484 | { |
|
|
485 | encode_nl (enc); ++enc->indent; |
170 | |
486 | |
171 | for (i = 0; i <= len; ++i) |
487 | for (i = 0; i <= len; ++i) |
172 | { |
488 | { |
173 | INDENT; |
489 | SV **svp = av_fetch (av, i, 0); |
174 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
175 | |
490 | |
|
|
491 | encode_indent (enc); |
|
|
492 | |
|
|
493 | if (svp) |
|
|
494 | encode_sv (enc, *svp); |
|
|
495 | else |
|
|
496 | encode_str (enc, "null", 4, 0); |
|
|
497 | |
176 | if (i < len) |
498 | if (i < len) |
177 | COMMA; |
499 | encode_comma (enc); |
178 | } |
500 | } |
179 | |
501 | |
180 | NL; |
502 | encode_nl (enc); --enc->indent; encode_indent (enc); |
181 | |
503 | } |
182 | --enc->indent; |
504 | |
183 | INDENT; encode_ch (enc, ']'); |
505 | encode_ch (enc, ']'); |
184 | } |
506 | } |
185 | |
507 | |
186 | static void |
508 | static void |
187 | encode_he (enc_t *enc, HE *he) |
509 | encode_hk (enc_t *enc, HE *he) |
188 | { |
510 | { |
189 | encode_ch (enc, '"'); |
511 | encode_ch (enc, '"'); |
190 | |
512 | |
191 | if (HeKLEN (he) == HEf_SVKEY) |
513 | if (HeKLEN (he) == HEf_SVKEY) |
192 | { |
514 | { |
193 | SV *sv = HeSVKEY (he); |
515 | SV *sv = HeSVKEY (he); |
194 | STRLEN len; |
516 | STRLEN len; |
|
|
517 | char *str; |
|
|
518 | |
|
|
519 | SvGETMAGIC (sv); |
195 | char *str = SvPV (sv, len); |
520 | str = SvPV (sv, len); |
196 | |
521 | |
197 | encode_str (enc, str, len, SvUTF8 (sv)); |
522 | encode_str (enc, str, len, SvUTF8 (sv)); |
198 | } |
523 | } |
199 | else |
524 | else |
200 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
525 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
201 | |
526 | |
202 | encode_ch (enc, '"'); |
527 | encode_ch (enc, '"'); |
203 | |
528 | |
204 | if (enc->flags & F_SPACE_BEFORE) SPACE; |
529 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
205 | encode_ch (enc, ':'); |
530 | encode_ch (enc, ':'); |
206 | if (enc->flags & F_SPACE_AFTER ) SPACE; |
531 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
207 | encode_sv (enc, HeVAL (he)); |
|
|
208 | } |
532 | } |
209 | |
533 | |
210 | // compare hash entries, used when all keys are bytestrings |
534 | // compare hash entries, used when all keys are bytestrings |
211 | static int |
535 | static int |
212 | he_cmp_fast (const void *a_, const void *b_) |
536 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
217 | HE *b = *(HE **)b_; |
541 | HE *b = *(HE **)b_; |
218 | |
542 | |
219 | STRLEN la = HeKLEN (a); |
543 | STRLEN la = HeKLEN (a); |
220 | STRLEN lb = HeKLEN (b); |
544 | STRLEN lb = HeKLEN (b); |
221 | |
545 | |
222 | if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
546 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
223 | cmp = la < lb ? -1 : la == lb ? 0 : 1; |
547 | cmp = lb - la; |
224 | |
548 | |
225 | return cmp; |
549 | return cmp; |
226 | } |
550 | } |
227 | |
551 | |
228 | // compare hash entries, used when some keys are sv's or utf-x |
552 | // compare hash entries, used when some keys are sv's or utf-x |
229 | static int |
553 | static int |
230 | he_cmp_slow (const void *a, const void *b) |
554 | he_cmp_slow (const void *a, const void *b) |
231 | { |
555 | { |
232 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
556 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
233 | } |
557 | } |
234 | |
558 | |
235 | static void |
559 | static void |
236 | encode_hv (enc_t *enc, HV *hv) |
560 | encode_hv (enc_t *enc, HV *hv) |
237 | { |
561 | { |
238 | int count, i; |
562 | HE *he; |
239 | |
563 | |
240 | encode_ch (enc, '{'); NL; ++enc->indent; |
564 | if (enc->indent >= enc->json.max_depth) |
|
|
565 | croak (ERR_NESTING_EXCEEDED); |
241 | |
566 | |
242 | if ((count = hv_iterinit (hv))) |
567 | encode_ch (enc, '{'); |
243 | { |
568 | |
244 | // for canonical output we have to sort by keys first |
569 | // for canonical output we have to sort by keys first |
245 | // actually, this is mostly due to the stupid so-called |
570 | // actually, this is mostly due to the stupid so-called |
246 | // security workaround added somewhere in 5.8.x. |
571 | // security workaround added somewhere in 5.8.x |
247 | // that randomises hash orderings |
572 | // that randomises hash orderings |
248 | if (enc->flags & F_CANONICAL) |
573 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
574 | { |
|
|
575 | int count = hv_iterinit (hv); |
|
|
576 | |
|
|
577 | if (SvMAGICAL (hv)) |
249 | { |
578 | { |
250 | HE *he, *hes [count]; |
579 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
580 | // but I don't care for the speed of this special case. |
|
|
581 | // note also that we will run into undefined behaviour when the two iterations |
|
|
582 | // do not result in the same count, something I might care for in some later release. |
|
|
583 | |
|
|
584 | count = 0; |
|
|
585 | while (hv_iternext (hv)) |
|
|
586 | ++count; |
|
|
587 | |
|
|
588 | hv_iterinit (hv); |
|
|
589 | } |
|
|
590 | |
|
|
591 | if (count) |
|
|
592 | { |
251 | int fast = 1; |
593 | int i, fast = 1; |
|
|
594 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
595 | HE **hes = _alloca (count * sizeof (HE)); |
|
|
596 | #else |
|
|
597 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
598 | #endif |
252 | |
599 | |
253 | i = 0; |
600 | i = 0; |
254 | while ((he = hv_iternext (hv))) |
601 | while ((he = hv_iternext (hv))) |
255 | { |
602 | { |
256 | hes [i++] = he; |
603 | hes [i++] = he; |
… | |
… | |
262 | |
609 | |
263 | if (fast) |
610 | if (fast) |
264 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
611 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
265 | else |
612 | else |
266 | { |
613 | { |
267 | // hack to disable "use bytes" |
614 | // hack to forcefully disable "use bytes" |
268 | COP *oldcop = PL_curcop, cop; |
615 | COP cop = *PL_curcop; |
269 | cop.op_private = 0; |
616 | cop.op_private = 0; |
|
|
617 | |
|
|
618 | ENTER; |
|
|
619 | SAVETMPS; |
|
|
620 | |
|
|
621 | SAVEVPTR (PL_curcop); |
270 | PL_curcop = &cop; |
622 | PL_curcop = &cop; |
271 | |
623 | |
272 | SAVETMPS; |
|
|
273 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
624 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
|
|
625 | |
274 | FREETMPS; |
626 | FREETMPS; |
275 | |
627 | LEAVE; |
276 | PL_curcop = oldcop; |
|
|
277 | } |
628 | } |
278 | |
629 | |
279 | for (i = 0; i < count; ++i) |
630 | encode_nl (enc); ++enc->indent; |
|
|
631 | |
|
|
632 | while (count--) |
280 | { |
633 | { |
281 | INDENT; |
634 | encode_indent (enc); |
|
|
635 | he = hes [count]; |
282 | encode_he (enc, hes [i]); |
636 | encode_hk (enc, he); |
|
|
637 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
283 | |
638 | |
284 | if (i < count - 1) |
639 | if (count) |
285 | COMMA; |
640 | encode_comma (enc); |
286 | } |
641 | } |
287 | |
642 | |
|
|
643 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
644 | } |
|
|
645 | } |
|
|
646 | else |
|
|
647 | { |
|
|
648 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
649 | if ((he = hv_iternext (hv))) |
288 | NL; |
650 | { |
|
|
651 | encode_nl (enc); ++enc->indent; |
|
|
652 | |
|
|
653 | for (;;) |
|
|
654 | { |
|
|
655 | encode_indent (enc); |
|
|
656 | encode_hk (enc, he); |
|
|
657 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
|
|
658 | |
|
|
659 | if (!(he = hv_iternext (hv))) |
|
|
660 | break; |
|
|
661 | |
|
|
662 | encode_comma (enc); |
|
|
663 | } |
|
|
664 | |
|
|
665 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
666 | } |
|
|
667 | } |
|
|
668 | |
|
|
669 | encode_ch (enc, '}'); |
|
|
670 | } |
|
|
671 | |
|
|
672 | // encode objects, arrays and special \0=false and \1=true values. |
|
|
673 | static void |
|
|
674 | encode_rv (enc_t *enc, SV *sv) |
|
|
675 | { |
|
|
676 | svtype svt; |
|
|
677 | |
|
|
678 | SvGETMAGIC (sv); |
|
|
679 | svt = SvTYPE (sv); |
|
|
680 | |
|
|
681 | if (expect_false (SvOBJECT (sv))) |
|
|
682 | { |
|
|
683 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
684 | ? json_boolean_stash |
|
|
685 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
686 | |
|
|
687 | if (SvSTASH (sv) == stash) |
|
|
688 | { |
|
|
689 | if (SvIV (sv)) |
|
|
690 | encode_str (enc, "true", 4, 0); |
|
|
691 | else |
|
|
692 | encode_str (enc, "false", 5, 0); |
289 | } |
693 | } |
290 | else |
694 | else |
291 | { |
695 | { |
292 | SV *sv; |
696 | #if 0 |
293 | HE *he = hv_iternext (hv); |
697 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
294 | |
|
|
295 | for (;;) |
|
|
296 | { |
698 | { |
297 | INDENT; |
699 | // not yet |
298 | encode_he (enc, he); |
|
|
299 | |
|
|
300 | if (!(he = hv_iternext (hv))) |
|
|
301 | break; |
|
|
302 | |
|
|
303 | COMMA; |
|
|
304 | } |
700 | } |
305 | |
701 | #endif |
|
|
702 | if (enc->json.flags & F_CONV_BLESSED) |
306 | NL; |
703 | { |
|
|
704 | // we re-bless the reference to get overload and other niceties right |
|
|
705 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
|
|
706 | |
|
|
707 | if (to_json) |
|
|
708 | { |
|
|
709 | dSP; |
|
|
710 | |
|
|
711 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
712 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
713 | |
|
|
714 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
715 | PUTBACK; |
|
|
716 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
|
|
717 | SPAGAIN; |
|
|
718 | |
|
|
719 | // catch this surprisingly common error |
|
|
720 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
721 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
722 | |
|
|
723 | sv = POPs; |
|
|
724 | PUTBACK; |
|
|
725 | |
|
|
726 | encode_sv (enc, sv); |
|
|
727 | |
|
|
728 | FREETMPS; LEAVE; |
|
|
729 | } |
|
|
730 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
731 | encode_str (enc, "null", 4, 0); |
|
|
732 | else |
|
|
733 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
734 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
735 | } |
|
|
736 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
737 | encode_str (enc, "null", 4, 0); |
|
|
738 | else |
|
|
739 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
|
|
740 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
307 | } |
741 | } |
308 | } |
742 | } |
|
|
743 | else if (svt == SVt_PVHV) |
|
|
744 | encode_hv (enc, (HV *)sv); |
|
|
745 | else if (svt == SVt_PVAV) |
|
|
746 | encode_av (enc, (AV *)sv); |
|
|
747 | else if (svt < SVt_PVAV) |
|
|
748 | { |
|
|
749 | STRLEN len = 0; |
|
|
750 | char *pv = svt ? SvPV (sv, len) : 0; |
309 | |
751 | |
310 | --enc->indent; INDENT; encode_ch (enc, '}'); |
752 | if (len == 1 && *pv == '1') |
|
|
753 | encode_str (enc, "true", 4, 0); |
|
|
754 | else if (len == 1 && *pv == '0') |
|
|
755 | encode_str (enc, "false", 5, 0); |
|
|
756 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
757 | encode_str (enc, "null", 4, 0); |
|
|
758 | else |
|
|
759 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
|
|
760 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
761 | } |
|
|
762 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
763 | encode_str (enc, "null", 4, 0); |
|
|
764 | else |
|
|
765 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
766 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
311 | } |
767 | } |
312 | |
768 | |
313 | static void |
769 | static void |
314 | encode_sv (enc_t *enc, SV *sv) |
770 | encode_sv (enc_t *enc, SV *sv) |
315 | { |
771 | { |
|
|
772 | SvGETMAGIC (sv); |
|
|
773 | |
316 | if (SvPOKp (sv)) |
774 | if (SvPOKp (sv)) |
317 | { |
775 | { |
318 | STRLEN len; |
776 | STRLEN len; |
319 | char *str = SvPV (sv, len); |
777 | char *str = SvPV (sv, len); |
320 | encode_ch (enc, '"'); |
778 | encode_ch (enc, '"'); |
321 | encode_str (enc, str, len, SvUTF8 (sv)); |
779 | encode_str (enc, str, len, SvUTF8 (sv)); |
322 | encode_ch (enc, '"'); |
780 | encode_ch (enc, '"'); |
323 | } |
781 | } |
324 | else if (SvNOKp (sv)) |
782 | else if (SvNOKp (sv)) |
325 | { |
783 | { |
|
|
784 | // trust that perl will do the right thing w.r.t. JSON syntax. |
326 | need (enc, NV_DIG + 32); |
785 | need (enc, NV_DIG + 32); |
327 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
786 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
328 | enc->cur += strlen (enc->cur); |
787 | enc->cur += strlen (enc->cur); |
329 | } |
788 | } |
330 | else if (SvIOKp (sv)) |
789 | else if (SvIOKp (sv)) |
331 | { |
790 | { |
|
|
791 | // we assume we can always read an IV as a UV and vice versa |
|
|
792 | // we assume two's complement |
|
|
793 | // we assume no aliasing issues in the union |
|
|
794 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
|
|
795 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
|
|
796 | { |
|
|
797 | // optimise the "small number case" |
|
|
798 | // code will likely be branchless and use only a single multiplication |
|
|
799 | // works for numbers up to 59074 |
|
|
800 | I32 i = SvIVX (sv); |
|
|
801 | U32 u; |
|
|
802 | char digit, nz = 0; |
|
|
803 | |
332 | need (enc, 64); |
804 | need (enc, 6); |
|
|
805 | |
|
|
806 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
807 | u = i < 0 ? -i : i; |
|
|
808 | |
|
|
809 | // convert to 4.28 fixed-point representation |
|
|
810 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
811 | |
|
|
812 | // now output digit by digit, each time masking out the integer part |
|
|
813 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
814 | // resulting in a net multiplication by 10. |
|
|
815 | // we always write the digit to memory but conditionally increment |
|
|
816 | // the pointer, to enable the use of conditional move instructions. |
|
|
817 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
|
|
818 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
|
|
819 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
|
|
820 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
|
|
821 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
822 | } |
|
|
823 | else |
|
|
824 | { |
|
|
825 | // large integer, use the (rather slow) snprintf way. |
|
|
826 | need (enc, IVUV_MAXCHARS); |
333 | enc->cur += |
827 | enc->cur += |
334 | SvIsUV(sv) |
828 | SvIsUV(sv) |
335 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
829 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
336 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
830 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
831 | } |
337 | } |
832 | } |
338 | else if (SvROK (sv)) |
833 | else if (SvROK (sv)) |
339 | { |
834 | encode_rv (enc, SvRV (sv)); |
340 | if (!--enc->max_recurse) |
835 | 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); |
836 | encode_str (enc, "null", 4, 0); |
356 | else |
837 | else |
357 | croak ("encountered perl type that JSON cannot handle"); |
838 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
|
|
839 | SvPV_nolen (sv), SvFLAGS (sv)); |
358 | } |
840 | } |
359 | |
841 | |
360 | static SV * |
842 | static SV * |
361 | encode_json (SV *scalar, UV flags) |
843 | encode_json (SV *scalar, JSON *json) |
362 | { |
844 | { |
363 | enc_t enc; |
845 | enc_t enc; |
364 | enc.flags = flags; |
846 | |
|
|
847 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
|
|
848 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
|
|
849 | |
|
|
850 | enc.json = *json; |
365 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
851 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
366 | enc.cur = SvPVX (enc.sv); |
852 | enc.cur = SvPVX (enc.sv); |
367 | enc.end = SvEND (enc.sv); |
853 | enc.end = SvEND (enc.sv); |
368 | enc.max_recurse = 0; |
|
|
369 | enc.indent = 0; |
854 | enc.indent = 0; |
|
|
855 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
856 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
857 | : 0x110000UL; |
370 | |
858 | |
371 | SvPOK_only (enc.sv); |
859 | SvPOK_only (enc.sv); |
372 | encode_sv (&enc, scalar); |
860 | encode_sv (&enc, scalar); |
|
|
861 | encode_nl (&enc); |
373 | |
862 | |
|
|
863 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
864 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
865 | |
374 | if (!(flags & (F_ASCII | F_UTF8))) |
866 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
375 | SvUTF8_on (enc.sv); |
867 | SvUTF8_on (enc.sv); |
376 | |
868 | |
377 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
869 | if (enc.json.flags & F_SHRINK) |
|
|
870 | shrink (enc.sv); |
|
|
871 | |
378 | return enc.sv; |
872 | return enc.sv; |
379 | } |
873 | } |
380 | |
874 | |
381 | ///////////////////////////////////////////////////////////////////////////// |
875 | ///////////////////////////////////////////////////////////////////////////// |
|
|
876 | // decoder |
382 | |
877 | |
383 | #define WS \ |
878 | // structure used for decoding JSON |
|
|
879 | typedef struct |
|
|
880 | { |
|
|
881 | char *cur; // current parser pointer |
|
|
882 | char *end; // end of input string |
|
|
883 | const char *err; // parse error, if != 0 |
|
|
884 | JSON json; |
|
|
885 | U32 depth; // recursion depth |
|
|
886 | U32 maxdepth; // recursion depth limit |
|
|
887 | } dec_t; |
|
|
888 | |
|
|
889 | INLINE void |
|
|
890 | decode_comment (dec_t *dec) |
|
|
891 | { |
|
|
892 | // only '#'-style comments allowed a.t.m. |
|
|
893 | |
|
|
894 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
895 | ++dec->cur; |
|
|
896 | } |
|
|
897 | |
|
|
898 | INLINE void |
|
|
899 | decode_ws (dec_t *dec) |
|
|
900 | { |
384 | for (;;) \ |
901 | for (;;) |
385 | { \ |
902 | { |
386 | char ch = *dec->cur; \ |
903 | char ch = *dec->cur; |
|
|
904 | |
387 | if (ch > 0x20 \ |
905 | if (ch > 0x20) |
|
|
906 | { |
|
|
907 | if (expect_false (ch == '#')) |
|
|
908 | { |
|
|
909 | if (dec->json.flags & F_RELAXED) |
|
|
910 | decode_comment (dec); |
|
|
911 | else |
|
|
912 | break; |
|
|
913 | } |
|
|
914 | else |
|
|
915 | break; |
|
|
916 | } |
388 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ |
917 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
389 | break; \ |
918 | break; // parse error, but let higher level handle it, gives better error messages |
|
|
919 | |
390 | ++dec->cur; \ |
920 | ++dec->cur; |
391 | } |
921 | } |
|
|
922 | } |
392 | |
923 | |
393 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
924 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
|
|
925 | |
394 | #define EXPECT_CH(ch) SB \ |
926 | #define EXPECT_CH(ch) SB \ |
395 | if (*dec->cur != ch) \ |
927 | if (*dec->cur != ch) \ |
396 | ERR (# ch " expected"); \ |
928 | ERR (# ch " expected"); \ |
397 | ++dec->cur; \ |
929 | ++dec->cur; \ |
398 | SE |
930 | SE |
399 | |
931 | |
|
|
932 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
|
|
933 | #define DEC_DEC_DEPTH --dec->depth |
|
|
934 | |
400 | static SV *decode_sv (dec_t *dec); |
935 | 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 | |
936 | |
407 | static signed char decode_hexdigit[256]; |
937 | static signed char decode_hexdigit[256]; |
408 | |
938 | |
409 | static UV |
939 | static UV |
410 | decode_4hex (dec_t *dec) |
940 | decode_4hex (dec_t *dec) |
411 | { |
941 | { |
412 | signed char d1, d2, d3, d4; |
942 | signed char d1, d2, d3, d4; |
|
|
943 | unsigned char *cur = (unsigned char *)dec->cur; |
413 | |
944 | |
414 | d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; |
945 | 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"); |
946 | 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]]; |
947 | 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"); |
948 | 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 | |
949 | |
423 | dec->cur += 4; |
950 | dec->cur += 4; |
424 | |
951 | |
425 | return ((UV)d1) << 12 |
952 | return ((UV)d1) << 12 |
426 | | ((UV)d2) << 8 |
953 | | ((UV)d2) << 8 |
… | |
… | |
432 | } |
959 | } |
433 | |
960 | |
434 | static SV * |
961 | static SV * |
435 | decode_str (dec_t *dec) |
962 | decode_str (dec_t *dec) |
436 | { |
963 | { |
437 | SV *sv = NEWSV (0,2); |
964 | SV *sv = 0; |
438 | STRLEN cur = 0; |
|
|
439 | int utf8 = 0; |
965 | int utf8 = 0; |
|
|
966 | char *dec_cur = dec->cur; |
440 | |
967 | |
441 | for (;;) |
968 | do |
442 | { |
969 | { |
443 | unsigned char ch = *(unsigned char *)dec->cur; |
970 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
|
|
971 | char *cur = buf; |
444 | |
972 | |
445 | if (ch == '"') |
973 | do |
446 | break; |
|
|
447 | else if (ch == '\\') |
|
|
448 | { |
974 | { |
449 | switch (*++dec->cur) |
975 | unsigned char ch = *(unsigned char *)dec_cur++; |
|
|
976 | |
|
|
977 | if (expect_false (ch == '"')) |
450 | { |
978 | { |
451 | case '\\': |
979 | --dec_cur; |
452 | case '/': |
980 | break; |
453 | case '"': APPEND_CH (*dec->cur++); break; |
981 | } |
454 | |
982 | else if (expect_false (ch == '\\')) |
455 | case 'b': APPEND_CH ('\010'); ++dec->cur; break; |
983 | { |
456 | case 't': APPEND_CH ('\011'); ++dec->cur; break; |
984 | 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 | { |
985 | { |
463 | UV lo, hi; |
986 | case '\\': |
464 | ++dec->cur; |
987 | case '/': |
|
|
988 | case '"': *cur++ = *dec_cur++; break; |
465 | |
989 | |
466 | hi = decode_4hex (dec); |
990 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
467 | if (hi == (UV)-1) |
991 | case 't': ++dec_cur; *cur++ = '\011'; break; |
468 | goto fail; |
992 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
|
|
993 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
|
|
994 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
469 | |
995 | |
470 | // possibly a surrogate pair |
996 | case 'u': |
471 | if (hi >= 0xd800 && hi < 0xdc00) |
|
|
472 | { |
997 | { |
473 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
998 | UV lo, hi; |
474 | ERR ("illegal surrogate character"); |
999 | ++dec_cur; |
475 | |
1000 | |
476 | dec->cur += 2; |
1001 | dec->cur = dec_cur; |
477 | |
|
|
478 | lo = decode_4hex (dec); |
1002 | hi = decode_4hex (dec); |
|
|
1003 | dec_cur = dec->cur; |
479 | if (lo == (UV)-1) |
1004 | if (hi == (UV)-1) |
480 | goto fail; |
1005 | goto fail; |
481 | |
1006 | |
|
|
1007 | // possibly a surrogate pair |
|
|
1008 | if (hi >= 0xd800) |
|
|
1009 | if (hi < 0xdc00) |
|
|
1010 | { |
|
|
1011 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
|
|
1012 | ERR ("missing low surrogate character in surrogate pair"); |
|
|
1013 | |
|
|
1014 | dec_cur += 2; |
|
|
1015 | |
|
|
1016 | dec->cur = dec_cur; |
|
|
1017 | lo = decode_4hex (dec); |
|
|
1018 | dec_cur = dec->cur; |
|
|
1019 | if (lo == (UV)-1) |
|
|
1020 | goto fail; |
|
|
1021 | |
482 | if (lo < 0xdc00 || lo >= 0xe000) |
1022 | if (lo < 0xdc00 || lo >= 0xe000) |
483 | ERR ("surrogate pair expected"); |
1023 | ERR ("surrogate pair expected"); |
484 | |
1024 | |
485 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
1025 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
|
|
1026 | } |
|
|
1027 | else if (hi < 0xe000) |
|
|
1028 | ERR ("missing high surrogate character in surrogate pair"); |
|
|
1029 | |
|
|
1030 | if (hi >= 0x80) |
|
|
1031 | { |
|
|
1032 | utf8 = 1; |
|
|
1033 | |
|
|
1034 | cur = encode_utf8 (cur, hi); |
|
|
1035 | } |
|
|
1036 | else |
|
|
1037 | *cur++ = hi; |
486 | } |
1038 | } |
487 | else if (lo >= 0xdc00 && lo < 0xe000) |
|
|
488 | ERR ("illegal surrogate character"); |
|
|
489 | |
|
|
490 | if (hi >= 0x80) |
|
|
491 | { |
1039 | break; |
492 | utf8 = 1; |
|
|
493 | |
1040 | |
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 |
1041 | default: |
498 | APPEND_CH (hi); |
1042 | --dec_cur; |
|
|
1043 | ERR ("illegal backslash escape sequence in string"); |
499 | } |
1044 | } |
|
|
1045 | } |
|
|
1046 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
|
|
1047 | *cur++ = ch; |
|
|
1048 | else if (ch >= 0x80) |
|
|
1049 | { |
|
|
1050 | STRLEN clen; |
|
|
1051 | |
|
|
1052 | --dec_cur; |
|
|
1053 | |
|
|
1054 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
|
|
1055 | if (clen == (STRLEN)-1) |
|
|
1056 | ERR ("malformed UTF-8 character in JSON string"); |
|
|
1057 | |
|
|
1058 | do |
|
|
1059 | *cur++ = *dec_cur++; |
|
|
1060 | while (--clen); |
|
|
1061 | |
|
|
1062 | utf8 = 1; |
|
|
1063 | } |
|
|
1064 | else |
|
|
1065 | { |
|
|
1066 | --dec_cur; |
|
|
1067 | |
|
|
1068 | if (!ch) |
|
|
1069 | ERR ("unexpected end of string while parsing JSON string"); |
500 | break; |
1070 | else |
|
|
1071 | ERR ("invalid character encountered while parsing JSON string"); |
501 | } |
1072 | } |
502 | } |
1073 | } |
503 | else if (ch >= 0x20 && ch <= 0x7f) |
1074 | while (cur < buf + SHORT_STRING_LEN); |
504 | APPEND_CH (*dec->cur++); |
1075 | |
505 | else if (ch >= 0x80) |
1076 | { |
|
|
1077 | STRLEN len = cur - buf; |
|
|
1078 | |
|
|
1079 | if (sv) |
506 | { |
1080 | { |
507 | STRLEN clen; |
1081 | 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 | |
1082 | |
512 | SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits |
1083 | if (SvLEN (sv) <= cur + len) |
513 | memcpy (SvPVX (sv) + cur, dec->cur, clen); |
1084 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
514 | dec->cur += clen; |
1085 | |
|
|
1086 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
|
|
1087 | SvCUR_set (sv, SvCUR (sv) + len); |
515 | } |
1088 | } |
516 | else |
1089 | else |
517 | ERR ("invalid character encountered"); |
1090 | sv = newSVpvn (buf, len); |
518 | } |
1091 | } |
|
|
1092 | } |
|
|
1093 | while (*dec_cur != '"'); |
519 | |
1094 | |
520 | ++dec->cur; |
1095 | ++dec_cur; |
521 | |
1096 | |
|
|
1097 | if (sv) |
|
|
1098 | { |
522 | SvPOK_only (sv); |
1099 | SvPOK_only (sv); |
523 | |
|
|
524 | SvCUR_set (sv, cur); |
|
|
525 | *SvEND (sv) = 0; |
1100 | *SvEND (sv) = 0; |
526 | |
1101 | |
527 | if (utf8) |
1102 | if (utf8) |
528 | SvUTF8_on (sv); |
1103 | SvUTF8_on (sv); |
|
|
1104 | } |
|
|
1105 | else |
|
|
1106 | sv = newSVpvn ("", 0); |
529 | |
1107 | |
|
|
1108 | dec->cur = dec_cur; |
530 | return sv; |
1109 | return sv; |
531 | |
1110 | |
532 | fail: |
1111 | fail: |
533 | SvREFCNT_dec (sv); |
1112 | dec->cur = dec_cur; |
534 | return 0; |
1113 | return 0; |
535 | } |
1114 | } |
536 | |
1115 | |
537 | static SV * |
1116 | static SV * |
538 | decode_num (dec_t *dec) |
1117 | decode_num (dec_t *dec) |
… | |
… | |
548 | { |
1127 | { |
549 | ++dec->cur; |
1128 | ++dec->cur; |
550 | if (*dec->cur >= '0' && *dec->cur <= '9') |
1129 | if (*dec->cur >= '0' && *dec->cur <= '9') |
551 | ERR ("malformed number (leading zero must not be followed by another digit)"); |
1130 | ERR ("malformed number (leading zero must not be followed by another digit)"); |
552 | } |
1131 | } |
553 | |
1132 | else if (*dec->cur < '0' || *dec->cur > '9') |
554 | // int |
1133 | ERR ("malformed number (no digits after initial minus)"); |
|
|
1134 | else |
|
|
1135 | do |
|
|
1136 | { |
|
|
1137 | ++dec->cur; |
|
|
1138 | } |
555 | while (*dec->cur >= '0' && *dec->cur <= '9') |
1139 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
556 | ++dec->cur; |
|
|
557 | |
1140 | |
558 | // [frac] |
1141 | // [frac] |
559 | if (*dec->cur == '.') |
1142 | if (*dec->cur == '.') |
560 | { |
1143 | { |
561 | is_nv = 1; |
1144 | ++dec->cur; |
|
|
1145 | |
|
|
1146 | if (*dec->cur < '0' || *dec->cur > '9') |
|
|
1147 | ERR ("malformed number (no digits after decimal point)"); |
562 | |
1148 | |
563 | do |
1149 | do |
564 | { |
1150 | { |
565 | ++dec->cur; |
1151 | ++dec->cur; |
566 | } |
1152 | } |
567 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
1153 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
|
|
1154 | |
|
|
1155 | is_nv = 1; |
568 | } |
1156 | } |
569 | |
1157 | |
570 | // [exp] |
1158 | // [exp] |
571 | if (*dec->cur == 'e' || *dec->cur == 'E') |
1159 | if (*dec->cur == 'e' || *dec->cur == 'E') |
572 | { |
1160 | { |
573 | is_nv = 1; |
|
|
574 | |
|
|
575 | ++dec->cur; |
1161 | ++dec->cur; |
|
|
1162 | |
576 | if (*dec->cur == '-' || *dec->cur == '+') |
1163 | if (*dec->cur == '-' || *dec->cur == '+') |
577 | ++dec->cur; |
1164 | ++dec->cur; |
578 | |
1165 | |
|
|
1166 | if (*dec->cur < '0' || *dec->cur > '9') |
|
|
1167 | ERR ("malformed number (no digits after exp sign)"); |
|
|
1168 | |
|
|
1169 | do |
|
|
1170 | { |
|
|
1171 | ++dec->cur; |
|
|
1172 | } |
579 | while (*dec->cur >= '0' && *dec->cur <= '9') |
1173 | while (*dec->cur >= '0' && *dec->cur <= '9'); |
580 | ++dec->cur; |
1174 | |
|
|
1175 | is_nv = 1; |
581 | } |
1176 | } |
582 | |
1177 | |
583 | if (!is_nv) |
1178 | if (!is_nv) |
584 | { |
1179 | { |
585 | UV uv; |
1180 | int len = dec->cur - start; |
586 | int numtype = grok_number (start, dec->cur - start, &uv); |
1181 | |
587 | if (numtype & IS_NUMBER_IN_UV) |
1182 | // special case the rather common 1..5-digit-int case |
588 | if (numtype & IS_NUMBER_NEG) |
1183 | if (*start == '-') |
|
|
1184 | switch (len) |
589 | { |
1185 | { |
590 | if (uv < (UV)IV_MIN) |
1186 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
591 | return newSViv (-(IV)uv); |
1187 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
|
|
1188 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
1189 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1190 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
592 | } |
1191 | } |
|
|
1192 | else |
|
|
1193 | switch (len) |
|
|
1194 | { |
|
|
1195 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
1196 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
1197 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
1198 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1199 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
|
|
1200 | } |
|
|
1201 | |
|
|
1202 | { |
|
|
1203 | UV uv; |
|
|
1204 | int numtype = grok_number (start, len, &uv); |
|
|
1205 | if (numtype & IS_NUMBER_IN_UV) |
|
|
1206 | if (numtype & IS_NUMBER_NEG) |
|
|
1207 | { |
|
|
1208 | if (uv < (UV)IV_MIN) |
|
|
1209 | return newSViv (-(IV)uv); |
|
|
1210 | } |
593 | else |
1211 | else |
594 | return newSVuv (uv); |
1212 | return newSVuv (uv); |
595 | } |
1213 | } |
596 | |
1214 | |
|
|
1215 | len -= *start == '-' ? 1 : 0; |
|
|
1216 | |
|
|
1217 | // does not fit into IV or UV, try NV |
|
|
1218 | if (len <= NV_DIG) |
|
|
1219 | // fits into NV without loss of precision |
|
|
1220 | return newSVnv (json_atof (start)); |
|
|
1221 | |
|
|
1222 | // everything else fails, convert it to a string |
|
|
1223 | return newSVpvn (start, dec->cur - start); |
|
|
1224 | } |
|
|
1225 | |
|
|
1226 | // loss of precision here |
597 | return newSVnv (Atof (start)); |
1227 | return newSVnv (json_atof (start)); |
598 | |
1228 | |
599 | fail: |
1229 | fail: |
600 | return 0; |
1230 | return 0; |
601 | } |
1231 | } |
602 | |
1232 | |
603 | static SV * |
1233 | static SV * |
604 | decode_av (dec_t *dec) |
1234 | decode_av (dec_t *dec) |
605 | { |
1235 | { |
606 | AV *av = newAV (); |
1236 | AV *av = newAV (); |
607 | |
1237 | |
|
|
1238 | DEC_INC_DEPTH; |
|
|
1239 | decode_ws (dec); |
|
|
1240 | |
|
|
1241 | if (*dec->cur == ']') |
|
|
1242 | ++dec->cur; |
|
|
1243 | else |
608 | for (;;) |
1244 | for (;;) |
609 | { |
1245 | { |
610 | SV *value; |
1246 | SV *value; |
611 | |
1247 | |
612 | value = decode_sv (dec); |
1248 | value = decode_sv (dec); |
613 | if (!value) |
1249 | if (!value) |
614 | goto fail; |
1250 | goto fail; |
615 | |
1251 | |
616 | av_push (av, value); |
1252 | av_push (av, value); |
617 | |
1253 | |
618 | WS; |
1254 | decode_ws (dec); |
619 | |
1255 | |
620 | if (*dec->cur == ']') |
1256 | if (*dec->cur == ']') |
621 | { |
1257 | { |
622 | ++dec->cur; |
1258 | ++dec->cur; |
623 | break; |
1259 | break; |
|
|
1260 | } |
624 | } |
1261 | |
625 | |
|
|
626 | if (*dec->cur != ',') |
1262 | if (*dec->cur != ',') |
627 | ERR (", or ] expected while parsing array"); |
1263 | ERR (", or ] expected while parsing array"); |
628 | |
1264 | |
629 | ++dec->cur; |
1265 | ++dec->cur; |
|
|
1266 | |
|
|
1267 | decode_ws (dec); |
|
|
1268 | |
|
|
1269 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1270 | { |
|
|
1271 | ++dec->cur; |
|
|
1272 | break; |
|
|
1273 | } |
630 | } |
1274 | } |
631 | |
1275 | |
|
|
1276 | DEC_DEC_DEPTH; |
632 | return newRV_noinc ((SV *)av); |
1277 | return newRV_noinc ((SV *)av); |
633 | |
1278 | |
634 | fail: |
1279 | fail: |
635 | SvREFCNT_dec (av); |
1280 | SvREFCNT_dec (av); |
|
|
1281 | DEC_DEC_DEPTH; |
636 | return 0; |
1282 | return 0; |
637 | } |
1283 | } |
638 | |
1284 | |
639 | static SV * |
1285 | static SV * |
640 | decode_hv (dec_t *dec) |
1286 | decode_hv (dec_t *dec) |
641 | { |
1287 | { |
|
|
1288 | SV *sv; |
642 | HV *hv = newHV (); |
1289 | HV *hv = newHV (); |
643 | |
1290 | |
|
|
1291 | DEC_INC_DEPTH; |
|
|
1292 | decode_ws (dec); |
|
|
1293 | |
|
|
1294 | if (*dec->cur == '}') |
|
|
1295 | ++dec->cur; |
|
|
1296 | else |
644 | for (;;) |
1297 | for (;;) |
645 | { |
1298 | { |
646 | SV *key, *value; |
|
|
647 | |
|
|
648 | WS; EXPECT_CH ('"'); |
1299 | EXPECT_CH ('"'); |
649 | |
1300 | |
650 | key = decode_str (dec); |
1301 | // heuristic: assume that |
651 | if (!key) |
1302 | // a) decode_str + hv_store_ent are abysmally slow. |
652 | goto fail; |
1303 | // b) most hash keys are short, simple ascii text. |
653 | |
1304 | // => try to "fast-match" such strings to avoid |
654 | WS; EXPECT_CH (':'); |
1305 | // the overhead of decode_str + hv_store_ent. |
655 | |
|
|
656 | value = decode_sv (dec); |
|
|
657 | if (!value) |
|
|
658 | { |
1306 | { |
|
|
1307 | SV *value; |
|
|
1308 | char *p = dec->cur; |
|
|
1309 | char *e = p + 24; // only try up to 24 bytes |
|
|
1310 | |
|
|
1311 | for (;;) |
|
|
1312 | { |
|
|
1313 | // the >= 0x80 is false on most architectures |
|
|
1314 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1315 | { |
|
|
1316 | // slow path, back up and use decode_str |
|
|
1317 | SV *key = decode_str (dec); |
|
|
1318 | if (!key) |
|
|
1319 | goto fail; |
|
|
1320 | |
|
|
1321 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1322 | |
|
|
1323 | decode_ws (dec); |
|
|
1324 | value = decode_sv (dec); |
|
|
1325 | if (!value) |
|
|
1326 | { |
|
|
1327 | SvREFCNT_dec (key); |
|
|
1328 | goto fail; |
|
|
1329 | } |
|
|
1330 | |
|
|
1331 | hv_store_ent (hv, key, value, 0); |
659 | SvREFCNT_dec (key); |
1332 | SvREFCNT_dec (key); |
|
|
1333 | |
|
|
1334 | break; |
|
|
1335 | } |
|
|
1336 | else if (*p == '"') |
|
|
1337 | { |
|
|
1338 | // fast path, got a simple key |
|
|
1339 | char *key = dec->cur; |
|
|
1340 | int len = p - key; |
|
|
1341 | dec->cur = p + 1; |
|
|
1342 | |
|
|
1343 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1344 | |
|
|
1345 | decode_ws (dec); |
|
|
1346 | value = decode_sv (dec); |
|
|
1347 | if (!value) |
660 | goto fail; |
1348 | goto fail; |
|
|
1349 | |
|
|
1350 | hv_store (hv, key, len, value, 0); |
|
|
1351 | |
|
|
1352 | break; |
|
|
1353 | } |
|
|
1354 | |
|
|
1355 | ++p; |
|
|
1356 | } |
661 | } |
1357 | } |
662 | |
1358 | |
663 | //TODO: optimise |
1359 | decode_ws (dec); |
664 | hv_store_ent (hv, key, value, 0); |
|
|
665 | |
1360 | |
666 | WS; |
|
|
667 | |
|
|
668 | if (*dec->cur == '}') |
1361 | if (*dec->cur == '}') |
|
|
1362 | { |
|
|
1363 | ++dec->cur; |
|
|
1364 | break; |
|
|
1365 | } |
|
|
1366 | |
|
|
1367 | if (*dec->cur != ',') |
|
|
1368 | ERR (", or } expected while parsing object/hash"); |
|
|
1369 | |
|
|
1370 | ++dec->cur; |
|
|
1371 | |
|
|
1372 | decode_ws (dec); |
|
|
1373 | |
|
|
1374 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1375 | { |
|
|
1376 | ++dec->cur; |
|
|
1377 | break; |
|
|
1378 | } |
|
|
1379 | } |
|
|
1380 | |
|
|
1381 | DEC_DEC_DEPTH; |
|
|
1382 | sv = newRV_noinc ((SV *)hv); |
|
|
1383 | |
|
|
1384 | // check filter callbacks |
|
|
1385 | if (dec->json.flags & F_HOOK) |
|
|
1386 | { |
|
|
1387 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
669 | { |
1388 | { |
670 | ++dec->cur; |
1389 | HE *cb, *he; |
671 | break; |
1390 | |
|
|
1391 | hv_iterinit (hv); |
|
|
1392 | he = hv_iternext (hv); |
|
|
1393 | hv_iterinit (hv); |
|
|
1394 | |
|
|
1395 | // the next line creates a mortal sv each time its called. |
|
|
1396 | // might want to optimise this for common cases. |
|
|
1397 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1398 | |
|
|
1399 | if (cb) |
|
|
1400 | { |
|
|
1401 | dSP; |
|
|
1402 | int count; |
|
|
1403 | |
|
|
1404 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1405 | XPUSHs (HeVAL (he)); |
|
|
1406 | sv_2mortal (sv); |
|
|
1407 | |
|
|
1408 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1409 | |
|
|
1410 | if (count == 1) |
|
|
1411 | { |
|
|
1412 | sv = newSVsv (POPs); |
|
|
1413 | FREETMPS; LEAVE; |
|
|
1414 | return sv; |
|
|
1415 | } |
|
|
1416 | |
|
|
1417 | SvREFCNT_inc (sv); |
|
|
1418 | FREETMPS; LEAVE; |
|
|
1419 | } |
672 | } |
1420 | } |
673 | |
1421 | |
674 | if (*dec->cur != ',') |
1422 | if (dec->json.cb_object) |
675 | ERR (", or } expected while parsing object/hash"); |
1423 | { |
|
|
1424 | dSP; |
|
|
1425 | int count; |
676 | |
1426 | |
677 | ++dec->cur; |
1427 | ENTER; SAVETMPS; PUSHMARK (SP); |
678 | } |
1428 | XPUSHs (sv_2mortal (sv)); |
679 | |
1429 | |
680 | return newRV_noinc ((SV *)hv); |
1430 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1431 | |
|
|
1432 | if (count == 1) |
|
|
1433 | { |
|
|
1434 | sv = newSVsv (POPs); |
|
|
1435 | FREETMPS; LEAVE; |
|
|
1436 | return sv; |
|
|
1437 | } |
|
|
1438 | |
|
|
1439 | SvREFCNT_inc (sv); |
|
|
1440 | FREETMPS; LEAVE; |
|
|
1441 | } |
|
|
1442 | } |
|
|
1443 | |
|
|
1444 | return sv; |
681 | |
1445 | |
682 | fail: |
1446 | fail: |
683 | SvREFCNT_dec (hv); |
1447 | SvREFCNT_dec (hv); |
|
|
1448 | DEC_DEC_DEPTH; |
684 | return 0; |
1449 | return 0; |
685 | } |
1450 | } |
686 | |
1451 | |
687 | static SV * |
1452 | static SV * |
688 | decode_sv (dec_t *dec) |
1453 | decode_sv (dec_t *dec) |
689 | { |
1454 | { |
690 | WS; |
1455 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1456 | // to parse everything. |
691 | switch (*dec->cur) |
1457 | switch (*dec->cur) |
692 | { |
1458 | { |
693 | case '"': ++dec->cur; return decode_str (dec); |
1459 | case '"': ++dec->cur; return decode_str (dec); |
694 | case '[': ++dec->cur; return decode_av (dec); |
1460 | case '[': ++dec->cur; return decode_av (dec); |
695 | case '{': ++dec->cur; return decode_hv (dec); |
1461 | case '{': ++dec->cur; return decode_hv (dec); |
696 | |
1462 | |
697 | case '-': |
1463 | case '-': |
698 | case '0': case '1': case '2': case '3': case '4': |
1464 | case '0': case '1': case '2': case '3': case '4': |
699 | case '5': case '6': case '7': case '8': case '9': |
1465 | case '5': case '6': case '7': case '8': case '9': |
700 | return decode_num (dec); |
1466 | return decode_num (dec); |
701 | |
1467 | |
702 | case 't': |
1468 | case 't': |
703 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1469 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
704 | { |
1470 | { |
705 | dec->cur += 4; |
1471 | dec->cur += 4; |
|
|
1472 | #if JSON_SLOW |
|
|
1473 | json_true = get_bool ("JSON::XS::true"); |
|
|
1474 | #endif |
706 | return newSViv (1); |
1475 | return newSVsv (json_true); |
707 | } |
1476 | } |
708 | else |
1477 | else |
709 | ERR ("'true' expected"); |
1478 | ERR ("'true' expected"); |
710 | |
1479 | |
711 | break; |
1480 | break; |
712 | |
1481 | |
713 | case 'f': |
1482 | case 'f': |
714 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1483 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
715 | { |
1484 | { |
716 | dec->cur += 5; |
1485 | dec->cur += 5; |
|
|
1486 | #if JSON_SLOW |
|
|
1487 | json_false = get_bool ("JSON::XS::false"); |
|
|
1488 | #endif |
717 | return newSViv (0); |
1489 | return newSVsv (json_false); |
718 | } |
1490 | } |
719 | else |
1491 | else |
720 | ERR ("'false' expected"); |
1492 | ERR ("'false' expected"); |
721 | |
1493 | |
722 | break; |
1494 | break; |
723 | |
1495 | |
724 | case 'n': |
1496 | case 'n': |
725 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
1497 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
726 | { |
1498 | { |
727 | dec->cur += 4; |
1499 | dec->cur += 4; |
728 | return newSViv (1); |
1500 | return newSVsv (&PL_sv_undef); |
729 | } |
1501 | } |
730 | else |
1502 | else |
731 | ERR ("'null' expected"); |
1503 | ERR ("'null' expected"); |
732 | |
1504 | |
733 | break; |
1505 | break; |
734 | |
1506 | |
735 | default: |
1507 | default: |
736 | ERR ("malformed json string"); |
1508 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
737 | break; |
1509 | break; |
738 | } |
1510 | } |
739 | |
1511 | |
740 | fail: |
1512 | fail: |
741 | return 0; |
1513 | return 0; |
742 | } |
1514 | } |
743 | |
1515 | |
744 | static SV * |
1516 | static SV * |
745 | decode_json (SV *string, UV flags) |
1517 | decode_json (SV *string, JSON *json, char **offset_return) |
746 | { |
1518 | { |
|
|
1519 | dec_t dec; |
747 | SV *sv; |
1520 | SV *sv; |
748 | |
1521 | |
|
|
1522 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1523 | * will perl ignore the magic in subsequent accesses |
|
|
1524 | */ |
|
|
1525 | /*SvGETMAGIC (string);*/ |
|
|
1526 | if (SvMAGICAL (string)) |
|
|
1527 | string = sv_2mortal (newSVsv (string)); |
|
|
1528 | |
|
|
1529 | SvUPGRADE (string, SVt_PV); |
|
|
1530 | |
|
|
1531 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1532 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1533 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1534 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1535 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1536 | * and hope for the best. |
|
|
1537 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1538 | * assertion business is seriously broken, try yet another workaround |
|
|
1539 | * for the broken -DDEBUGGING. |
|
|
1540 | */ |
|
|
1541 | { |
|
|
1542 | #ifdef DEBUGGING |
|
|
1543 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1544 | #else |
|
|
1545 | STRLEN offset = SvCUR (string); |
|
|
1546 | #endif |
|
|
1547 | |
|
|
1548 | if (offset > json->max_size && json->max_size) |
|
|
1549 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1550 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1551 | } |
|
|
1552 | |
749 | if (!(flags & F_UTF8)) |
1553 | if (json->flags & F_UTF8) |
|
|
1554 | sv_utf8_downgrade (string, 0); |
|
|
1555 | else |
750 | sv_utf8_upgrade (string); |
1556 | sv_utf8_upgrade (string); |
751 | |
1557 | |
752 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1558 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
753 | |
1559 | |
754 | dec_t dec; |
1560 | dec.json = *json; |
755 | dec.flags = flags; |
|
|
756 | dec.cur = SvPVX (string); |
1561 | dec.cur = SvPVX (string); |
757 | dec.end = SvEND (string); |
1562 | dec.end = SvEND (string); |
758 | dec.err = 0; |
1563 | dec.err = 0; |
|
|
1564 | dec.depth = 0; |
759 | |
1565 | |
760 | *dec.end = 1; // invalid anywhere |
1566 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1567 | dec.json.flags |= F_HOOK; |
|
|
1568 | |
|
|
1569 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1570 | |
|
|
1571 | decode_ws (&dec); |
761 | sv = decode_sv (&dec); |
1572 | sv = decode_sv (&dec); |
762 | *dec.end = 0; |
1573 | |
|
|
1574 | if (offset_return) |
|
|
1575 | *offset_return = dec.cur; |
|
|
1576 | |
|
|
1577 | if (!(offset_return || !sv)) |
|
|
1578 | { |
|
|
1579 | // check for trailing garbage |
|
|
1580 | decode_ws (&dec); |
|
|
1581 | |
|
|
1582 | if (*dec.cur) |
|
|
1583 | { |
|
|
1584 | dec.err = "garbage after JSON object"; |
|
|
1585 | SvREFCNT_dec (sv); |
|
|
1586 | sv = 0; |
|
|
1587 | } |
|
|
1588 | } |
763 | |
1589 | |
764 | if (!sv) |
1590 | if (!sv) |
765 | { |
1591 | { |
766 | IV offset = utf8_distance (dec.cur, SvPVX (string)); |
|
|
767 | SV *uni = sv_newmortal (); |
1592 | SV *uni = sv_newmortal (); |
768 | |
1593 | |
|
|
1594 | // horrible hack to silence warning inside pv_uni_display |
|
|
1595 | COP cop = *PL_curcop; |
|
|
1596 | cop.cop_warnings = pWARN_NONE; |
|
|
1597 | ENTER; |
|
|
1598 | SAVEVPTR (PL_curcop); |
|
|
1599 | PL_curcop = &cop; |
769 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1600 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
|
|
1601 | LEAVE; |
|
|
1602 | |
770 | croak ("%s, at character %d (%s)", |
1603 | croak ("%s, at character offset %d (before \"%s\")", |
771 | dec.err, |
1604 | dec.err, |
772 | (int)offset, |
1605 | ptr_to_index (string, dec.cur), |
773 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1606 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
774 | } |
1607 | } |
775 | |
1608 | |
776 | sv_dump (sv);//D |
|
|
777 | return sv_2mortal (sv); |
1609 | sv = sv_2mortal (sv); |
|
|
1610 | |
|
|
1611 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
|
|
1612 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
|
|
1613 | |
|
|
1614 | return sv; |
778 | } |
1615 | } |
|
|
1616 | |
|
|
1617 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1618 | // incremental parser |
|
|
1619 | |
|
|
1620 | static void |
|
|
1621 | incr_parse (JSON *self) |
|
|
1622 | { |
|
|
1623 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1624 | |
|
|
1625 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1626 | // but this would make it slower, so... |
|
|
1627 | |
|
|
1628 | for (;;) |
|
|
1629 | { |
|
|
1630 | //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D |
|
|
1631 | switch (self->incr_mode) |
|
|
1632 | { |
|
|
1633 | // only used for initial whitespace skipping |
|
|
1634 | case INCR_M_WS: |
|
|
1635 | for (;;) |
|
|
1636 | { |
|
|
1637 | if (*p > 0x20) |
|
|
1638 | { |
|
|
1639 | if (*p == '#') |
|
|
1640 | { |
|
|
1641 | self->incr_mode = INCR_M_C0; |
|
|
1642 | goto incr_m_c; |
|
|
1643 | } |
|
|
1644 | else |
|
|
1645 | { |
|
|
1646 | self->incr_mode = INCR_M_JSON; |
|
|
1647 | goto incr_m_json; |
|
|
1648 | } |
|
|
1649 | } |
|
|
1650 | else if (!*p) |
|
|
1651 | goto interrupt; |
|
|
1652 | |
|
|
1653 | ++p; |
|
|
1654 | } |
|
|
1655 | |
|
|
1656 | // skip a single char inside a string (for \\-processing) |
|
|
1657 | case INCR_M_BS: |
|
|
1658 | if (!*p) |
|
|
1659 | goto interrupt; |
|
|
1660 | |
|
|
1661 | ++p; |
|
|
1662 | self->incr_mode = INCR_M_STR; |
|
|
1663 | goto incr_m_str; |
|
|
1664 | |
|
|
1665 | // inside #-style comments |
|
|
1666 | case INCR_M_C0: |
|
|
1667 | case INCR_M_C1: |
|
|
1668 | incr_m_c: |
|
|
1669 | for (;;) |
|
|
1670 | { |
|
|
1671 | if (*p == '\n') |
|
|
1672 | { |
|
|
1673 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1674 | break; |
|
|
1675 | } |
|
|
1676 | else if (!*p) |
|
|
1677 | goto interrupt; |
|
|
1678 | |
|
|
1679 | ++p; |
|
|
1680 | } |
|
|
1681 | |
|
|
1682 | break; |
|
|
1683 | |
|
|
1684 | // inside a string |
|
|
1685 | case INCR_M_STR: |
|
|
1686 | incr_m_str: |
|
|
1687 | for (;;) |
|
|
1688 | { |
|
|
1689 | if (*p == '"') |
|
|
1690 | { |
|
|
1691 | ++p; |
|
|
1692 | self->incr_mode = INCR_M_JSON; |
|
|
1693 | |
|
|
1694 | if (!self->incr_nest) |
|
|
1695 | goto interrupt; |
|
|
1696 | |
|
|
1697 | goto incr_m_json; |
|
|
1698 | } |
|
|
1699 | else if (*p == '\\') |
|
|
1700 | { |
|
|
1701 | ++p; // "virtually" consumes character after \ |
|
|
1702 | |
|
|
1703 | if (!*p) // if at end of string we have to switch modes |
|
|
1704 | { |
|
|
1705 | self->incr_mode = INCR_M_BS; |
|
|
1706 | goto interrupt; |
|
|
1707 | } |
|
|
1708 | } |
|
|
1709 | else if (!*p) |
|
|
1710 | goto interrupt; |
|
|
1711 | |
|
|
1712 | ++p; |
|
|
1713 | } |
|
|
1714 | |
|
|
1715 | // after initial ws, outside string |
|
|
1716 | case INCR_M_JSON: |
|
|
1717 | incr_m_json: |
|
|
1718 | for (;;) |
|
|
1719 | { |
|
|
1720 | switch (*p++) |
|
|
1721 | { |
|
|
1722 | case 0: |
|
|
1723 | --p; |
|
|
1724 | goto interrupt; |
|
|
1725 | |
|
|
1726 | case 0x09: |
|
|
1727 | case 0x0a: |
|
|
1728 | case 0x0d: |
|
|
1729 | case 0x20: |
|
|
1730 | if (!self->incr_nest) |
|
|
1731 | { |
|
|
1732 | --p; // do not eat the whitespace, let the next round do it |
|
|
1733 | goto interrupt; |
|
|
1734 | } |
|
|
1735 | break; |
|
|
1736 | |
|
|
1737 | case '"': |
|
|
1738 | self->incr_mode = INCR_M_STR; |
|
|
1739 | goto incr_m_str; |
|
|
1740 | |
|
|
1741 | case '[': |
|
|
1742 | case '{': |
|
|
1743 | if (++self->incr_nest > self->max_depth) |
|
|
1744 | croak (ERR_NESTING_EXCEEDED); |
|
|
1745 | break; |
|
|
1746 | |
|
|
1747 | case ']': |
|
|
1748 | case '}': |
|
|
1749 | if (--self->incr_nest <= 0) |
|
|
1750 | goto interrupt; |
|
|
1751 | break; |
|
|
1752 | |
|
|
1753 | case '#': |
|
|
1754 | self->incr_mode = INCR_M_C1; |
|
|
1755 | goto incr_m_c; |
|
|
1756 | } |
|
|
1757 | } |
|
|
1758 | } |
|
|
1759 | |
|
|
1760 | modechange: |
|
|
1761 | ; |
|
|
1762 | } |
|
|
1763 | |
|
|
1764 | interrupt: |
|
|
1765 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1766 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1767 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
|
|
1768 | } |
|
|
1769 | |
|
|
1770 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1771 | // XS interface functions |
779 | |
1772 | |
780 | MODULE = JSON::XS PACKAGE = JSON::XS |
1773 | MODULE = JSON::XS PACKAGE = JSON::XS |
781 | |
1774 | |
782 | BOOT: |
1775 | BOOT: |
783 | { |
1776 | { |
784 | int i; |
1777 | int i; |
785 | |
1778 | |
786 | memset (decode_hexdigit, 0xff, 256); |
|
|
787 | for (i = 10; i--; ) |
1779 | for (i = 0; i < 256; ++i) |
788 | decode_hexdigit ['0' + i] = i; |
1780 | decode_hexdigit [i] = |
|
|
1781 | i >= '0' && i <= '9' ? i - '0' |
|
|
1782 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
|
|
1783 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
|
|
1784 | : -1; |
789 | |
1785 | |
790 | for (i = 6; --i; ) |
|
|
791 | { |
|
|
792 | decode_hexdigit ['a' + i] = 10 + i; |
|
|
793 | decode_hexdigit ['A' + i] = 10 + i; |
|
|
794 | } |
|
|
795 | |
|
|
796 | json_stash = gv_stashpv ("JSON::XS", 1); |
1786 | json_stash = gv_stashpv ("JSON::XS" , 1); |
797 | } |
1787 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
798 | |
1788 | |
799 | SV *new (char *dummy) |
1789 | json_true = get_bool ("JSON::XS::true"); |
|
|
1790 | json_false = get_bool ("JSON::XS::false"); |
|
|
1791 | |
|
|
1792 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
|
|
1793 | } |
|
|
1794 | |
|
|
1795 | PROTOTYPES: DISABLE |
|
|
1796 | |
|
|
1797 | void CLONE (...) |
800 | CODE: |
1798 | CODE: |
801 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1799 | json_stash = 0; |
|
|
1800 | json_boolean_stash = 0; |
|
|
1801 | |
|
|
1802 | void new (char *klass) |
|
|
1803 | PPCODE: |
|
|
1804 | { |
|
|
1805 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1806 | SvPOK_only (pv); |
|
|
1807 | json_init ((JSON *)SvPVX (pv)); |
|
|
1808 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1809 | newRV_noinc (pv), |
|
|
1810 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1811 | ))); |
|
|
1812 | } |
|
|
1813 | |
|
|
1814 | void ascii (JSON *self, int enable = 1) |
|
|
1815 | ALIAS: |
|
|
1816 | ascii = F_ASCII |
|
|
1817 | latin1 = F_LATIN1 |
|
|
1818 | utf8 = F_UTF8 |
|
|
1819 | indent = F_INDENT |
|
|
1820 | canonical = F_CANONICAL |
|
|
1821 | space_before = F_SPACE_BEFORE |
|
|
1822 | space_after = F_SPACE_AFTER |
|
|
1823 | pretty = F_PRETTY |
|
|
1824 | allow_nonref = F_ALLOW_NONREF |
|
|
1825 | shrink = F_SHRINK |
|
|
1826 | allow_blessed = F_ALLOW_BLESSED |
|
|
1827 | convert_blessed = F_CONV_BLESSED |
|
|
1828 | relaxed = F_RELAXED |
|
|
1829 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
1830 | PPCODE: |
|
|
1831 | { |
|
|
1832 | if (enable) |
|
|
1833 | self->flags |= ix; |
|
|
1834 | else |
|
|
1835 | self->flags &= ~ix; |
|
|
1836 | |
|
|
1837 | XPUSHs (ST (0)); |
|
|
1838 | } |
|
|
1839 | |
|
|
1840 | void get_ascii (JSON *self) |
|
|
1841 | ALIAS: |
|
|
1842 | get_ascii = F_ASCII |
|
|
1843 | get_latin1 = F_LATIN1 |
|
|
1844 | get_utf8 = F_UTF8 |
|
|
1845 | get_indent = F_INDENT |
|
|
1846 | get_canonical = F_CANONICAL |
|
|
1847 | get_space_before = F_SPACE_BEFORE |
|
|
1848 | get_space_after = F_SPACE_AFTER |
|
|
1849 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1850 | get_shrink = F_SHRINK |
|
|
1851 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1852 | get_convert_blessed = F_CONV_BLESSED |
|
|
1853 | get_relaxed = F_RELAXED |
|
|
1854 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
1855 | PPCODE: |
|
|
1856 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1857 | |
|
|
1858 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
|
|
1859 | PPCODE: |
|
|
1860 | self->max_depth = max_depth; |
|
|
1861 | XPUSHs (ST (0)); |
|
|
1862 | |
|
|
1863 | U32 get_max_depth (JSON *self) |
|
|
1864 | CODE: |
|
|
1865 | RETVAL = self->max_depth; |
802 | OUTPUT: |
1866 | OUTPUT: |
803 | RETVAL |
1867 | RETVAL |
804 | |
1868 | |
805 | SV *ascii (SV *self, int enable) |
1869 | void max_size (JSON *self, U32 max_size = 0) |
806 | ALIAS: |
1870 | PPCODE: |
807 | ascii = F_ASCII |
1871 | self->max_size = max_size; |
808 | utf8 = F_UTF8 |
1872 | XPUSHs (ST (0)); |
809 | indent = F_INDENT |
1873 | |
810 | canonical = F_CANONICAL |
1874 | 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: |
1875 | CODE: |
815 | { |
1876 | 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: |
1877 | OUTPUT: |
825 | RETVAL |
1878 | RETVAL |
826 | |
1879 | |
827 | void encode (SV *self, SV *scalar) |
1880 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
828 | PPCODE: |
1881 | PPCODE: |
829 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
1882 | { |
|
|
1883 | SvREFCNT_dec (self->cb_object); |
|
|
1884 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
830 | |
1885 | |
831 | void decode (SV *self, SV *jsondata) |
1886 | XPUSHs (ST (0)); |
|
|
1887 | } |
|
|
1888 | |
|
|
1889 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
832 | PPCODE: |
1890 | PPCODE: |
|
|
1891 | { |
|
|
1892 | if (!self->cb_sk_object) |
|
|
1893 | self->cb_sk_object = newHV (); |
|
|
1894 | |
|
|
1895 | if (SvOK (cb)) |
|
|
1896 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
1897 | else |
|
|
1898 | { |
|
|
1899 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
1900 | |
|
|
1901 | if (!HvKEYS (self->cb_sk_object)) |
|
|
1902 | { |
|
|
1903 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1904 | self->cb_sk_object = 0; |
|
|
1905 | } |
|
|
1906 | } |
|
|
1907 | |
|
|
1908 | XPUSHs (ST (0)); |
|
|
1909 | } |
|
|
1910 | |
|
|
1911 | void encode (JSON *self, SV *scalar) |
|
|
1912 | PPCODE: |
|
|
1913 | XPUSHs (encode_json (scalar, self)); |
|
|
1914 | |
|
|
1915 | void decode (JSON *self, SV *jsonstr) |
|
|
1916 | PPCODE: |
833 | XPUSHs (decode_json (jsondata, *SvJSON (self))); |
1917 | XPUSHs (decode_json (jsonstr, self, 0)); |
834 | |
1918 | |
|
|
1919 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
1920 | PPCODE: |
|
|
1921 | { |
|
|
1922 | char *offset; |
|
|
1923 | EXTEND (SP, 2); |
|
|
1924 | PUSHs (decode_json (jsonstr, self, &offset)); |
|
|
1925 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1926 | } |
|
|
1927 | |
|
|
1928 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1929 | PPCODE: |
|
|
1930 | { |
|
|
1931 | if (!self->incr_text) |
|
|
1932 | self->incr_text = newSVpvn ("", 0); |
|
|
1933 | |
|
|
1934 | // append data, if any |
|
|
1935 | if (jsonstr) |
|
|
1936 | { |
|
|
1937 | if (SvUTF8 (jsonstr)) |
|
|
1938 | { |
|
|
1939 | if (!SvUTF8 (self->incr_text)) |
|
|
1940 | { |
|
|
1941 | /* utf-8-ness differs, need to upgrade */ |
|
|
1942 | sv_utf8_upgrade (self->incr_text); |
|
|
1943 | |
|
|
1944 | if (self->incr_pos) |
|
|
1945 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1946 | - (U8 *)SvPVX (self->incr_text); |
|
|
1947 | } |
|
|
1948 | } |
|
|
1949 | else if (SvUTF8 (self->incr_text)) |
|
|
1950 | sv_utf8_upgrade (jsonstr); |
|
|
1951 | |
|
|
1952 | { |
|
|
1953 | STRLEN len; |
|
|
1954 | const char *str = SvPV (jsonstr, len); |
|
|
1955 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1956 | |
|
|
1957 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1958 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1959 | |
|
|
1960 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1961 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1962 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1963 | } |
|
|
1964 | } |
|
|
1965 | |
|
|
1966 | if (GIMME_V != G_VOID) |
|
|
1967 | do |
|
|
1968 | { |
|
|
1969 | char *offset; |
|
|
1970 | |
|
|
1971 | if (!INCR_DONE (self)) |
|
|
1972 | { |
|
|
1973 | incr_parse (self); |
|
|
1974 | |
|
|
1975 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1976 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1977 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1978 | |
|
|
1979 | if (!INCR_DONE (self)) |
|
|
1980 | break; |
|
|
1981 | } |
|
|
1982 | |
|
|
1983 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1984 | |
|
|
1985 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
1986 | self->incr_nest = 0; |
|
|
1987 | self->incr_mode = 0; |
|
|
1988 | |
|
|
1989 | sv_chop (self->incr_text, offset); |
|
|
1990 | } |
|
|
1991 | while (GIMME_V == G_ARRAY); |
|
|
1992 | } |
|
|
1993 | |
|
|
1994 | SV *incr_text (JSON *self) |
|
|
1995 | ATTRS: lvalue |
|
|
1996 | CODE: |
|
|
1997 | { |
|
|
1998 | if (self->incr_pos) |
|
|
1999 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
2000 | |
|
|
2001 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2002 | } |
|
|
2003 | OUTPUT: |
|
|
2004 | RETVAL |
|
|
2005 | |
|
|
2006 | void incr_skip (JSON *self) |
|
|
2007 | CODE: |
|
|
2008 | { |
|
|
2009 | if (self->incr_pos) |
|
|
2010 | { |
|
|
2011 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2012 | self->incr_pos = 0; |
|
|
2013 | self->incr_nest = 0; |
|
|
2014 | self->incr_mode = 0; |
|
|
2015 | } |
|
|
2016 | } |
|
|
2017 | |
|
|
2018 | void incr_reset (JSON *self) |
|
|
2019 | CODE: |
|
|
2020 | { |
|
|
2021 | SvREFCNT_dec (self->incr_text); |
|
|
2022 | self->incr_text = 0; |
|
|
2023 | self->incr_pos = 0; |
|
|
2024 | self->incr_nest = 0; |
|
|
2025 | self->incr_mode = 0; |
|
|
2026 | } |
|
|
2027 | |
|
|
2028 | void DESTROY (JSON *self) |
|
|
2029 | CODE: |
|
|
2030 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2031 | SvREFCNT_dec (self->cb_object); |
|
|
2032 | SvREFCNT_dec (self->incr_text); |
|
|
2033 | |
|
|
2034 | PROTOTYPES: ENABLE |
|
|
2035 | |
|
|
2036 | void encode_json (SV *scalar) |
|
|
2037 | ALIAS: |
|
|
2038 | to_json_ = 0 |
|
|
2039 | encode_json = F_UTF8 |
|
|
2040 | PPCODE: |
|
|
2041 | { |
|
|
2042 | JSON json; |
|
|
2043 | json_init (&json); |
|
|
2044 | json.flags |= ix; |
|
|
2045 | XPUSHs (encode_json (scalar, &json)); |
|
|
2046 | } |
|
|
2047 | |
|
|
2048 | void decode_json (SV *jsonstr) |
|
|
2049 | ALIAS: |
|
|
2050 | from_json_ = 0 |
|
|
2051 | decode_json = F_UTF8 |
|
|
2052 | PPCODE: |
|
|
2053 | { |
|
|
2054 | JSON json; |
|
|
2055 | json_init (&json); |
|
|
2056 | json.flags |= ix; |
|
|
2057 | XPUSHs (decode_json (jsonstr, &json, 0)); |
|
|
2058 | } |
|
|
2059 | |