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 |
|
|
16 | #define F_ALLOW_NONREF 0x00000080 |
32 | #define F_ALLOW_NONREF 0x00000100UL |
17 | #define F_SHRINK 0x00000100 |
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 |
18 | |
39 | |
19 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
20 | #define F_DEFAULT 0 |
|
|
21 | |
41 | |
22 | #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 |
|
|
46 | |
|
|
47 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
23 | |
48 | |
24 | #define SB do { |
49 | #define SB do { |
25 | #define SE } while (0) |
50 | #define SE } while (0) |
26 | |
51 | |
27 | static HV *json_stash; |
52 | #if __GNUC__ >= 3 |
|
|
53 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
|
|
54 | # define INLINE static inline |
|
|
55 | #else |
|
|
56 | # define expect(expr,value) (expr) |
|
|
57 | # define INLINE static |
|
|
58 | #endif |
|
|
59 | |
|
|
60 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
61 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
62 | |
|
|
63 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
64 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
65 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
66 | |
|
|
67 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
68 | |
|
|
69 | #ifdef USE_ITHREADS |
|
|
70 | # define JSON_SLOW 1 |
|
|
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
72 | #else |
|
|
73 | # define JSON_SLOW 0 |
|
|
74 | # define JSON_STASH json_stash |
|
|
75 | #endif |
|
|
76 | |
|
|
77 | // the amount of HEs to allocate on the stack, when sorting keys |
|
|
78 | #define STACK_HES 64 |
|
|
79 | |
|
|
80 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
|
|
81 | static SV *json_true, *json_false; |
|
|
82 | |
|
|
83 | enum { |
|
|
84 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
85 | INCR_M_STR, // inside string |
|
|
86 | INCR_M_BS, // inside backslash |
|
|
87 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
88 | INCR_M_C1, // inside comment in other places |
|
|
89 | INCR_M_JSON // outside anything, count nesting |
|
|
90 | }; |
|
|
91 | |
|
|
92 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
93 | |
|
|
94 | typedef struct { |
|
|
95 | U32 flags; |
|
|
96 | U32 max_depth; |
|
|
97 | STRLEN max_size; |
|
|
98 | |
|
|
99 | SV *cb_object; |
|
|
100 | HV *cb_sk_object; |
|
|
101 | |
|
|
102 | // for the incremental parser |
|
|
103 | SV *incr_text; // the source text so far |
|
|
104 | STRLEN incr_pos; // the current offset into the text |
|
|
105 | int incr_nest; // {[]}-nesting level |
|
|
106 | unsigned char incr_mode; |
|
|
107 | } JSON; |
|
|
108 | |
|
|
109 | INLINE void |
|
|
110 | json_init (JSON *json) |
|
|
111 | { |
|
|
112 | Zero (json, 1, JSON); |
|
|
113 | json->max_depth = 512; |
|
|
114 | } |
|
|
115 | |
|
|
116 | ///////////////////////////////////////////////////////////////////////////// |
|
|
117 | // utility functions |
|
|
118 | |
|
|
119 | INLINE SV * |
|
|
120 | get_bool (const char *name) |
|
|
121 | { |
|
|
122 | SV *sv = get_sv (name, 1); |
|
|
123 | |
|
|
124 | SvREADONLY_on (sv); |
|
|
125 | SvREADONLY_on (SvRV (sv)); |
|
|
126 | |
|
|
127 | return sv; |
|
|
128 | } |
|
|
129 | |
|
|
130 | INLINE void |
|
|
131 | shrink (SV *sv) |
|
|
132 | { |
|
|
133 | sv_utf8_downgrade (sv, 1); |
|
|
134 | |
|
|
135 | if (SvLEN (sv) > SvCUR (sv) + 1) |
|
|
136 | { |
|
|
137 | #ifdef SvPV_shrink_to_cur |
|
|
138 | SvPV_shrink_to_cur (sv); |
|
|
139 | #elif defined (SvPV_renew) |
|
|
140 | SvPV_renew (sv, SvCUR (sv) + 1); |
|
|
141 | #endif |
|
|
142 | } |
|
|
143 | } |
|
|
144 | |
|
|
145 | // decode an utf-8 character and return it, or (UV)-1 in |
|
|
146 | // case of an error. |
|
|
147 | // we special-case "safe" characters from U+80 .. U+7FF, |
|
|
148 | // but use the very good perl function to parse anything else. |
|
|
149 | // note that we never call this function for a ascii codepoints |
|
|
150 | INLINE UV |
|
|
151 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
|
|
152 | { |
|
|
153 | if (expect_true (len >= 2 |
|
|
154 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
|
|
155 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
|
|
156 | { |
|
|
157 | *clen = 2; |
|
|
158 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
|
|
159 | } |
|
|
160 | else |
|
|
161 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
|
|
162 | } |
|
|
163 | |
|
|
164 | // likewise for encoding, also never called for ascii codepoints |
|
|
165 | // this function takes advantage of this fact, although current gccs |
|
|
166 | // seem to optimise the check for >= 0x80 away anyways |
|
|
167 | INLINE unsigned char * |
|
|
168 | encode_utf8 (unsigned char *s, UV ch) |
|
|
169 | { |
|
|
170 | if (expect_false (ch < 0x000080)) |
|
|
171 | *s++ = ch; |
|
|
172 | else if (expect_true (ch < 0x000800)) |
|
|
173 | *s++ = 0xc0 | ( ch >> 6), |
|
|
174 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
175 | else if ( ch < 0x010000) |
|
|
176 | *s++ = 0xe0 | ( ch >> 12), |
|
|
177 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
178 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
179 | else if ( ch < 0x110000) |
|
|
180 | *s++ = 0xf0 | ( ch >> 18), |
|
|
181 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
182 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
183 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
184 | |
|
|
185 | return s; |
|
|
186 | } |
|
|
187 | |
|
|
188 | // convert offset pointer to character index, sv must be string |
|
|
189 | static STRLEN |
|
|
190 | ptr_to_index (SV *sv, char *offset) |
|
|
191 | { |
|
|
192 | return SvUTF8 (sv) |
|
|
193 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
194 | : offset - SvPVX (sv); |
|
|
195 | } |
|
|
196 | |
|
|
197 | ///////////////////////////////////////////////////////////////////////////// |
|
|
198 | // fp hell |
|
|
199 | |
|
|
200 | // scan a group of digits, and a trailing exponent |
|
|
201 | static void |
|
|
202 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
203 | { |
|
|
204 | UV uaccum = 0; |
|
|
205 | int eaccum = 0; |
|
|
206 | |
|
|
207 | // if we recurse too deep, skip all remaining digits |
|
|
208 | // to avoid a stack overflow attack |
|
|
209 | if (expect_false (--maxdepth <= 0)) |
|
|
210 | while (((U8)*s - '0') < 10) |
|
|
211 | ++s; |
|
|
212 | |
|
|
213 | for (;;) |
|
|
214 | { |
|
|
215 | U8 dig = (U8)*s - '0'; |
|
|
216 | |
|
|
217 | if (expect_false (dig >= 10)) |
|
|
218 | { |
|
|
219 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
220 | { |
|
|
221 | ++s; |
|
|
222 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
223 | } |
|
|
224 | else if ((dig | ' ') == 'e' - '0') |
|
|
225 | { |
|
|
226 | int exp2 = 0; |
|
|
227 | int neg = 0; |
|
|
228 | |
|
|
229 | ++s; |
|
|
230 | |
|
|
231 | if (*s == '-') |
|
|
232 | { |
|
|
233 | ++s; |
|
|
234 | neg = 1; |
|
|
235 | } |
|
|
236 | else if (*s == '+') |
|
|
237 | ++s; |
|
|
238 | |
|
|
239 | while ((dig = (U8)*s - '0') < 10) |
|
|
240 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
241 | |
|
|
242 | *expo += neg ? -exp2 : exp2; |
|
|
243 | } |
|
|
244 | |
|
|
245 | break; |
|
|
246 | } |
|
|
247 | |
|
|
248 | ++s; |
|
|
249 | |
|
|
250 | uaccum = uaccum * 10 + dig; |
|
|
251 | ++eaccum; |
|
|
252 | |
|
|
253 | // if we have too many digits, then recurse for more |
|
|
254 | // we actually do this for rather few digits |
|
|
255 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
256 | { |
|
|
257 | if (postdp) *expo -= eaccum; |
|
|
258 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
259 | if (postdp) *expo += eaccum; |
|
|
260 | |
|
|
261 | break; |
|
|
262 | } |
|
|
263 | } |
|
|
264 | |
|
|
265 | // this relies greatly on the quality of the pow () |
|
|
266 | // implementation of the platform, but a good |
|
|
267 | // implementation is hard to beat. |
|
|
268 | // (IEEE 754 conformant ones are required to be exact) |
|
|
269 | if (postdp) *expo -= eaccum; |
|
|
270 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
271 | *expo += eaccum; |
|
|
272 | } |
|
|
273 | |
|
|
274 | static NV |
|
|
275 | json_atof (const char *s) |
|
|
276 | { |
|
|
277 | NV accum = 0.; |
|
|
278 | int expo = 0; |
|
|
279 | int neg = 0; |
|
|
280 | |
|
|
281 | if (*s == '-') |
|
|
282 | { |
|
|
283 | ++s; |
|
|
284 | neg = 1; |
|
|
285 | } |
|
|
286 | |
|
|
287 | // a recursion depth of ten gives us >>500 bits |
|
|
288 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
289 | |
|
|
290 | return neg ? -accum : accum; |
|
|
291 | } |
|
|
292 | ///////////////////////////////////////////////////////////////////////////// |
|
|
293 | // encoder |
28 | |
294 | |
29 | // structure used for encoding JSON |
295 | // structure used for encoding JSON |
30 | typedef struct |
296 | typedef struct |
31 | { |
297 | { |
32 | char *cur; |
298 | char *cur; // SvPVX (sv) + current output position |
33 | STRLEN len; // SvLEN (sv) |
|
|
34 | char *end; // SvEND (sv) |
299 | char *end; // SvEND (sv) |
35 | SV *sv; |
300 | SV *sv; // result scalar |
36 | UV flags; |
301 | JSON json; |
37 | int max_recurse; |
302 | U32 indent; // indentation level |
38 | int indent; |
303 | UV limit; // escape character values >= this value when encoding |
39 | } enc_t; |
304 | } enc_t; |
40 | |
305 | |
41 | // structure used for decoding JSON |
306 | INLINE void |
42 | typedef struct |
|
|
43 | { |
|
|
44 | char *cur; |
|
|
45 | char *end; |
|
|
46 | const char *err; |
|
|
47 | UV flags; |
|
|
48 | } dec_t; |
|
|
49 | |
|
|
50 | static UV * |
|
|
51 | SvJSON (SV *sv) |
|
|
52 | { |
|
|
53 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
|
|
54 | croak ("object is not of type JSON::XS"); |
|
|
55 | |
|
|
56 | return &SvUVX (SvRV (sv)); |
|
|
57 | } |
|
|
58 | |
|
|
59 | ///////////////////////////////////////////////////////////////////////////// |
|
|
60 | |
|
|
61 | static void |
|
|
62 | need (enc_t *enc, STRLEN len) |
307 | need (enc_t *enc, STRLEN len) |
63 | { |
308 | { |
64 | if (enc->cur + len >= enc->end) |
309 | if (expect_false (enc->cur + len >= enc->end)) |
65 | { |
310 | { |
66 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
311 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
67 | SvGROW (enc->sv, cur + len + 1); |
312 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
68 | enc->cur = SvPVX (enc->sv) + cur; |
313 | enc->cur = SvPVX (enc->sv) + cur; |
69 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); |
314 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
70 | } |
315 | } |
71 | } |
316 | } |
72 | |
317 | |
73 | static void |
318 | INLINE void |
74 | encode_ch (enc_t *enc, char ch) |
319 | encode_ch (enc_t *enc, char ch) |
75 | { |
320 | { |
76 | need (enc, 1); |
321 | need (enc, 1); |
77 | *enc->cur++ = ch; |
322 | *enc->cur++ = ch; |
78 | } |
323 | } |
… | |
… | |
86 | |
331 | |
87 | while (str < end) |
332 | while (str < end) |
88 | { |
333 | { |
89 | unsigned char ch = *(unsigned char *)str; |
334 | unsigned char ch = *(unsigned char *)str; |
90 | |
335 | |
91 | if (ch >= 0x20 && ch < 0x80) // most common case |
336 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
92 | { |
337 | { |
93 | if (ch == '"') // but with slow exceptions |
338 | if (expect_false (ch == '"')) // but with slow exceptions |
94 | { |
339 | { |
95 | need (enc, len += 1); |
340 | need (enc, len += 1); |
96 | *enc->cur++ = '\\'; |
341 | *enc->cur++ = '\\'; |
97 | *enc->cur++ = '"'; |
342 | *enc->cur++ = '"'; |
98 | } |
343 | } |
99 | else if (ch == '\\') |
344 | else if (expect_false (ch == '\\')) |
100 | { |
345 | { |
101 | need (enc, len += 1); |
346 | need (enc, len += 1); |
102 | *enc->cur++ = '\\'; |
347 | *enc->cur++ = '\\'; |
103 | *enc->cur++ = '\\'; |
348 | *enc->cur++ = '\\'; |
104 | } |
349 | } |
… | |
… | |
122 | STRLEN clen; |
367 | STRLEN clen; |
123 | UV uch; |
368 | UV uch; |
124 | |
369 | |
125 | if (is_utf8) |
370 | if (is_utf8) |
126 | { |
371 | { |
127 | uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
372 | uch = decode_utf8 (str, end - str, &clen); |
128 | if (clen == (STRLEN)-1) |
373 | if (clen == (STRLEN)-1) |
129 | croak ("malformed UTF-8 character in string, cannot convert to JSON"); |
374 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
130 | } |
375 | } |
131 | else |
376 | else |
132 | { |
377 | { |
133 | uch = ch; |
378 | uch = ch; |
134 | clen = 1; |
379 | clen = 1; |
135 | } |
380 | } |
136 | |
381 | |
137 | if (uch < 0x80 || enc->flags & F_ASCII) |
382 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
138 | { |
383 | { |
139 | if (uch > 0xFFFFUL) |
384 | if (uch >= 0x10000UL) |
140 | { |
385 | { |
|
|
386 | if (uch >= 0x110000UL) |
|
|
387 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
388 | |
141 | need (enc, len += 11); |
389 | need (enc, len += 11); |
142 | sprintf (enc->cur, "\\u%04x\\u%04x", |
390 | sprintf (enc->cur, "\\u%04x\\u%04x", |
143 | (uch - 0x10000) / 0x400 + 0xD800, |
391 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
144 | (uch - 0x10000) % 0x400 + 0xDC00); |
392 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
145 | enc->cur += 12; |
393 | enc->cur += 12; |
146 | } |
394 | } |
147 | else |
395 | else |
148 | { |
396 | { |
149 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
150 | need (enc, len += 5); |
397 | need (enc, len += 5); |
151 | *enc->cur++ = '\\'; |
398 | *enc->cur++ = '\\'; |
152 | *enc->cur++ = 'u'; |
399 | *enc->cur++ = 'u'; |
153 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
400 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
154 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
401 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
155 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
402 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
156 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
403 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
157 | } |
404 | } |
158 | |
405 | |
|
|
406 | str += clen; |
|
|
407 | } |
|
|
408 | else if (enc->json.flags & F_LATIN1) |
|
|
409 | { |
|
|
410 | *enc->cur++ = uch; |
159 | str += clen; |
411 | str += clen; |
160 | } |
412 | } |
161 | else if (is_utf8) |
413 | else if (is_utf8) |
162 | { |
414 | { |
163 | need (enc, len += clen); |
415 | need (enc, len += clen); |
… | |
… | |
167 | } |
419 | } |
168 | while (--clen); |
420 | while (--clen); |
169 | } |
421 | } |
170 | else |
422 | else |
171 | { |
423 | { |
172 | need (enc, 10); // never more than 11 bytes needed |
424 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
173 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
425 | enc->cur = encode_utf8 (enc->cur, uch); |
174 | ++str; |
426 | ++str; |
175 | } |
427 | } |
176 | } |
428 | } |
177 | } |
429 | } |
178 | } |
430 | } |
179 | |
431 | |
180 | --len; |
432 | --len; |
181 | } |
433 | } |
182 | } |
434 | } |
183 | |
435 | |
184 | #define INDENT SB \ |
436 | INLINE void |
|
|
437 | encode_indent (enc_t *enc) |
|
|
438 | { |
185 | if (enc->flags & F_INDENT) \ |
439 | if (enc->json.flags & F_INDENT) |
186 | { \ |
440 | { |
187 | int i_; \ |
441 | int spaces = enc->indent * INDENT_STEP; |
188 | need (enc, enc->indent); \ |
442 | |
189 | for (i_ = enc->indent * 3; i_--; )\ |
443 | need (enc, spaces); |
|
|
444 | memset (enc->cur, ' ', spaces); |
|
|
445 | enc->cur += spaces; |
|
|
446 | } |
|
|
447 | } |
|
|
448 | |
|
|
449 | INLINE void |
|
|
450 | encode_space (enc_t *enc) |
|
|
451 | { |
|
|
452 | need (enc, 1); |
|
|
453 | encode_ch (enc, ' '); |
|
|
454 | } |
|
|
455 | |
|
|
456 | INLINE void |
|
|
457 | encode_nl (enc_t *enc) |
|
|
458 | { |
|
|
459 | if (enc->json.flags & F_INDENT) |
|
|
460 | { |
|
|
461 | need (enc, 1); |
190 | encode_ch (enc, ' '); \ |
462 | encode_ch (enc, '\n'); |
191 | } \ |
463 | } |
192 | SE |
464 | } |
193 | |
465 | |
194 | #define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE |
466 | INLINE void |
195 | #define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE |
467 | encode_comma (enc_t *enc) |
196 | #define COMMA SB \ |
468 | { |
197 | encode_ch (enc, ','); \ |
469 | encode_ch (enc, ','); |
|
|
470 | |
198 | if (enc->flags & F_INDENT) \ |
471 | if (enc->json.flags & F_INDENT) |
199 | NL; \ |
472 | encode_nl (enc); |
200 | else if (enc->flags & F_SPACE_AFTER) \ |
473 | else if (enc->json.flags & F_SPACE_AFTER) |
201 | SPACE; \ |
474 | encode_space (enc); |
202 | SE |
475 | } |
203 | |
476 | |
204 | static void encode_sv (enc_t *enc, SV *sv); |
477 | static void encode_sv (enc_t *enc, SV *sv); |
205 | |
478 | |
206 | static void |
479 | static void |
207 | encode_av (enc_t *enc, AV *av) |
480 | encode_av (enc_t *enc, AV *av) |
208 | { |
481 | { |
209 | int i, len = av_len (av); |
482 | int i, len = av_len (av); |
210 | |
483 | |
|
|
484 | if (enc->indent >= enc->json.max_depth) |
|
|
485 | croak (ERR_NESTING_EXCEEDED); |
|
|
486 | |
211 | encode_ch (enc, '['); NL; |
487 | encode_ch (enc, '['); |
212 | ++enc->indent; |
|
|
213 | |
488 | |
|
|
489 | if (len >= 0) |
|
|
490 | { |
|
|
491 | encode_nl (enc); ++enc->indent; |
|
|
492 | |
214 | for (i = 0; i <= len; ++i) |
493 | for (i = 0; i <= len; ++i) |
215 | { |
494 | { |
216 | INDENT; |
495 | SV **svp = av_fetch (av, i, 0); |
217 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
218 | |
496 | |
|
|
497 | encode_indent (enc); |
|
|
498 | |
|
|
499 | if (svp) |
|
|
500 | encode_sv (enc, *svp); |
|
|
501 | else |
|
|
502 | encode_str (enc, "null", 4, 0); |
|
|
503 | |
219 | if (i < len) |
504 | if (i < len) |
220 | COMMA; |
505 | encode_comma (enc); |
221 | } |
506 | } |
222 | |
507 | |
223 | NL; |
508 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
509 | } |
224 | |
510 | |
225 | --enc->indent; |
|
|
226 | INDENT; encode_ch (enc, ']'); |
511 | encode_ch (enc, ']'); |
227 | } |
512 | } |
228 | |
513 | |
229 | static void |
514 | static void |
230 | encode_he (enc_t *enc, HE *he) |
515 | encode_hk (enc_t *enc, HE *he) |
231 | { |
516 | { |
232 | encode_ch (enc, '"'); |
517 | encode_ch (enc, '"'); |
233 | |
518 | |
234 | if (HeKLEN (he) == HEf_SVKEY) |
519 | if (HeKLEN (he) == HEf_SVKEY) |
235 | { |
520 | { |
236 | SV *sv = HeSVKEY (he); |
521 | SV *sv = HeSVKEY (he); |
237 | STRLEN len; |
522 | STRLEN len; |
238 | char *str; |
523 | char *str; |
239 | |
524 | |
240 | SvGETMAGIC (sv); |
525 | SvGETMAGIC (sv); |
241 | str = SvPV (sv, len); |
526 | str = SvPV (sv, len); |
242 | |
527 | |
243 | encode_str (enc, str, len, SvUTF8 (sv)); |
528 | encode_str (enc, str, len, SvUTF8 (sv)); |
244 | } |
529 | } |
245 | else |
530 | else |
246 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
531 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
247 | |
532 | |
248 | encode_ch (enc, '"'); |
533 | encode_ch (enc, '"'); |
249 | |
534 | |
250 | if (enc->flags & F_SPACE_BEFORE) SPACE; |
535 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
251 | encode_ch (enc, ':'); |
536 | encode_ch (enc, ':'); |
252 | if (enc->flags & F_SPACE_AFTER ) SPACE; |
537 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
253 | encode_sv (enc, HeVAL (he)); |
|
|
254 | } |
538 | } |
255 | |
539 | |
256 | // compare hash entries, used when all keys are bytestrings |
540 | // compare hash entries, used when all keys are bytestrings |
257 | static int |
541 | static int |
258 | he_cmp_fast (const void *a_, const void *b_) |
542 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
263 | HE *b = *(HE **)b_; |
547 | HE *b = *(HE **)b_; |
264 | |
548 | |
265 | STRLEN la = HeKLEN (a); |
549 | STRLEN la = HeKLEN (a); |
266 | STRLEN lb = HeKLEN (b); |
550 | STRLEN lb = HeKLEN (b); |
267 | |
551 | |
268 | if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
552 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
269 | cmp = la < lb ? -1 : la == lb ? 0 : 1; |
553 | cmp = lb - la; |
270 | |
554 | |
271 | return cmp; |
555 | return cmp; |
272 | } |
556 | } |
273 | |
557 | |
274 | // compare hash entries, used when some keys are sv's or utf-x |
558 | // compare hash entries, used when some keys are sv's or utf-x |
275 | static int |
559 | static int |
276 | he_cmp_slow (const void *a, const void *b) |
560 | he_cmp_slow (const void *a, const void *b) |
277 | { |
561 | { |
278 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
562 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
279 | } |
563 | } |
280 | |
564 | |
281 | static void |
565 | static void |
282 | encode_hv (enc_t *enc, HV *hv) |
566 | encode_hv (enc_t *enc, HV *hv) |
283 | { |
567 | { |
284 | int count, i; |
568 | HE *he; |
285 | |
569 | |
286 | encode_ch (enc, '{'); NL; ++enc->indent; |
570 | if (enc->indent >= enc->json.max_depth) |
|
|
571 | croak (ERR_NESTING_EXCEEDED); |
287 | |
572 | |
288 | if ((count = hv_iterinit (hv))) |
573 | encode_ch (enc, '{'); |
289 | { |
574 | |
290 | // for canonical output we have to sort by keys first |
575 | // for canonical output we have to sort by keys first |
291 | // actually, this is mostly due to the stupid so-called |
576 | // actually, this is mostly due to the stupid so-called |
292 | // security workaround added somewhere in 5.8.x. |
577 | // security workaround added somewhere in 5.8.x |
293 | // that randomises hash orderings |
578 | // that randomises hash orderings |
294 | if (enc->flags & F_CANONICAL) |
579 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
580 | { |
|
|
581 | int count = hv_iterinit (hv); |
|
|
582 | |
|
|
583 | if (SvMAGICAL (hv)) |
295 | { |
584 | { |
296 | HE *he, *hes [count]; |
585 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
586 | // but I don't care for the speed of this special case. |
|
|
587 | // note also that we will run into undefined behaviour when the two iterations |
|
|
588 | // do not result in the same count, something I might care for in some later release. |
|
|
589 | |
|
|
590 | count = 0; |
|
|
591 | while (hv_iternext (hv)) |
|
|
592 | ++count; |
|
|
593 | |
|
|
594 | hv_iterinit (hv); |
|
|
595 | } |
|
|
596 | |
|
|
597 | if (count) |
|
|
598 | { |
297 | int fast = 1; |
599 | int i, fast = 1; |
|
|
600 | HE *hes_stack [STACK_HES]; |
|
|
601 | HE **hes = hes_stack; |
|
|
602 | |
|
|
603 | // allocate larger arrays on the heap |
|
|
604 | if (count > STACK_HES) |
|
|
605 | { |
|
|
606 | SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes))); |
|
|
607 | hes = (HE **)SvPVX (sv); |
|
|
608 | } |
298 | |
609 | |
299 | i = 0; |
610 | i = 0; |
300 | while ((he = hv_iternext (hv))) |
611 | while ((he = hv_iternext (hv))) |
301 | { |
612 | { |
302 | hes [i++] = he; |
613 | hes [i++] = he; |
… | |
… | |
308 | |
619 | |
309 | if (fast) |
620 | if (fast) |
310 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
621 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
311 | else |
622 | else |
312 | { |
623 | { |
313 | // hack to disable "use bytes" |
624 | // hack to forcefully disable "use bytes" |
314 | COP *oldcop = PL_curcop, cop; |
625 | COP cop = *PL_curcop; |
315 | cop.op_private = 0; |
626 | cop.op_private = 0; |
|
|
627 | |
|
|
628 | ENTER; |
|
|
629 | SAVETMPS; |
|
|
630 | |
|
|
631 | SAVEVPTR (PL_curcop); |
316 | PL_curcop = &cop; |
632 | PL_curcop = &cop; |
317 | |
633 | |
318 | SAVETMPS; |
|
|
319 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
634 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
|
|
635 | |
320 | FREETMPS; |
636 | FREETMPS; |
|
|
637 | LEAVE; |
|
|
638 | } |
321 | |
639 | |
322 | PL_curcop = oldcop; |
640 | encode_nl (enc); ++enc->indent; |
|
|
641 | |
|
|
642 | while (count--) |
323 | } |
643 | { |
|
|
644 | encode_indent (enc); |
|
|
645 | he = hes [count]; |
|
|
646 | encode_hk (enc, he); |
|
|
647 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
324 | |
648 | |
325 | for (i = 0; i < count; ++i) |
649 | if (count) |
|
|
650 | encode_comma (enc); |
|
|
651 | } |
|
|
652 | |
|
|
653 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
654 | } |
|
|
655 | } |
|
|
656 | else |
|
|
657 | { |
|
|
658 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
659 | if ((he = hv_iternext (hv))) |
|
|
660 | { |
|
|
661 | encode_nl (enc); ++enc->indent; |
|
|
662 | |
|
|
663 | for (;;) |
326 | { |
664 | { |
327 | INDENT; |
665 | encode_indent (enc); |
328 | encode_he (enc, hes [i]); |
666 | encode_hk (enc, he); |
|
|
667 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
329 | |
668 | |
330 | if (i < count - 1) |
669 | if (!(he = hv_iternext (hv))) |
331 | COMMA; |
670 | break; |
|
|
671 | |
|
|
672 | encode_comma (enc); |
332 | } |
673 | } |
333 | |
674 | |
|
|
675 | encode_nl (enc); --enc->indent; encode_indent (enc); |
334 | NL; |
676 | } |
|
|
677 | } |
|
|
678 | |
|
|
679 | encode_ch (enc, '}'); |
|
|
680 | } |
|
|
681 | |
|
|
682 | // encode objects, arrays and special \0=false and \1=true values. |
|
|
683 | static void |
|
|
684 | encode_rv (enc_t *enc, SV *sv) |
|
|
685 | { |
|
|
686 | svtype svt; |
|
|
687 | |
|
|
688 | SvGETMAGIC (sv); |
|
|
689 | svt = SvTYPE (sv); |
|
|
690 | |
|
|
691 | if (expect_false (SvOBJECT (sv))) |
|
|
692 | { |
|
|
693 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
694 | ? json_boolean_stash |
|
|
695 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
696 | |
|
|
697 | if (SvSTASH (sv) == stash) |
|
|
698 | { |
|
|
699 | if (SvIV (sv)) |
|
|
700 | encode_str (enc, "true", 4, 0); |
|
|
701 | else |
|
|
702 | encode_str (enc, "false", 5, 0); |
335 | } |
703 | } |
336 | else |
704 | else |
337 | { |
705 | { |
338 | SV *sv; |
706 | #if 0 |
339 | HE *he = hv_iternext (hv); |
707 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
340 | |
|
|
341 | for (;;) |
|
|
342 | { |
|
|
343 | INDENT; |
|
|
344 | encode_he (enc, he); |
|
|
345 | |
|
|
346 | if (!(he = hv_iternext (hv))) |
|
|
347 | break; |
|
|
348 | |
|
|
349 | COMMA; |
|
|
350 | } |
708 | { |
351 | |
709 | // not yet |
352 | NL; |
710 | } |
|
|
711 | #endif |
|
|
712 | if (enc->json.flags & F_CONV_BLESSED) |
|
|
713 | { |
|
|
714 | // we re-bless the reference to get overload and other niceties right |
|
|
715 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
|
|
716 | |
|
|
717 | if (to_json) |
|
|
718 | { |
|
|
719 | dSP; |
|
|
720 | |
|
|
721 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
722 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
723 | |
|
|
724 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
725 | PUTBACK; |
|
|
726 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
|
|
727 | SPAGAIN; |
|
|
728 | |
|
|
729 | // catch this surprisingly common error |
|
|
730 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
731 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
732 | |
|
|
733 | sv = POPs; |
|
|
734 | PUTBACK; |
|
|
735 | |
|
|
736 | encode_sv (enc, sv); |
|
|
737 | |
|
|
738 | FREETMPS; LEAVE; |
|
|
739 | } |
|
|
740 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
741 | encode_str (enc, "null", 4, 0); |
|
|
742 | else |
|
|
743 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
744 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
745 | } |
|
|
746 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
747 | encode_str (enc, "null", 4, 0); |
|
|
748 | else |
|
|
749 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
|
|
750 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
353 | } |
751 | } |
354 | } |
752 | } |
|
|
753 | else if (svt == SVt_PVHV) |
|
|
754 | encode_hv (enc, (HV *)sv); |
|
|
755 | else if (svt == SVt_PVAV) |
|
|
756 | encode_av (enc, (AV *)sv); |
|
|
757 | else if (svt < SVt_PVAV) |
|
|
758 | { |
|
|
759 | STRLEN len = 0; |
|
|
760 | char *pv = svt ? SvPV (sv, len) : 0; |
355 | |
761 | |
356 | --enc->indent; INDENT; encode_ch (enc, '}'); |
762 | if (len == 1 && *pv == '1') |
|
|
763 | encode_str (enc, "true", 4, 0); |
|
|
764 | else if (len == 1 && *pv == '0') |
|
|
765 | encode_str (enc, "false", 5, 0); |
|
|
766 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
767 | encode_str (enc, "null", 4, 0); |
|
|
768 | else |
|
|
769 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
|
|
770 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
771 | } |
|
|
772 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
773 | encode_str (enc, "null", 4, 0); |
|
|
774 | else |
|
|
775 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
776 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
357 | } |
777 | } |
358 | |
778 | |
359 | static void |
779 | static void |
360 | encode_sv (enc_t *enc, SV *sv) |
780 | encode_sv (enc_t *enc, SV *sv) |
361 | { |
781 | { |
… | |
… | |
369 | encode_str (enc, str, len, SvUTF8 (sv)); |
789 | encode_str (enc, str, len, SvUTF8 (sv)); |
370 | encode_ch (enc, '"'); |
790 | encode_ch (enc, '"'); |
371 | } |
791 | } |
372 | else if (SvNOKp (sv)) |
792 | else if (SvNOKp (sv)) |
373 | { |
793 | { |
|
|
794 | // trust that perl will do the right thing w.r.t. JSON syntax. |
374 | need (enc, NV_DIG + 32); |
795 | need (enc, NV_DIG + 32); |
375 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
796 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
376 | enc->cur += strlen (enc->cur); |
797 | enc->cur += strlen (enc->cur); |
377 | } |
798 | } |
378 | else if (SvIOKp (sv)) |
799 | else if (SvIOKp (sv)) |
379 | { |
800 | { |
|
|
801 | // we assume we can always read an IV as a UV and vice versa |
|
|
802 | // we assume two's complement |
|
|
803 | // we assume no aliasing issues in the union |
|
|
804 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
|
|
805 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
|
|
806 | { |
|
|
807 | // optimise the "small number case" |
|
|
808 | // code will likely be branchless and use only a single multiplication |
|
|
809 | // works for numbers up to 59074 |
|
|
810 | I32 i = SvIVX (sv); |
|
|
811 | U32 u; |
|
|
812 | char digit, nz = 0; |
|
|
813 | |
380 | need (enc, 64); |
814 | need (enc, 6); |
|
|
815 | |
|
|
816 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
817 | u = i < 0 ? -i : i; |
|
|
818 | |
|
|
819 | // convert to 4.28 fixed-point representation |
|
|
820 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
821 | |
|
|
822 | // now output digit by digit, each time masking out the integer part |
|
|
823 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
824 | // resulting in a net multiplication by 10. |
|
|
825 | // we always write the digit to memory but conditionally increment |
|
|
826 | // the pointer, to enable the use of conditional move instructions. |
|
|
827 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
|
|
828 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
|
|
829 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
|
|
830 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
|
|
831 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
832 | } |
|
|
833 | else |
|
|
834 | { |
|
|
835 | // large integer, use the (rather slow) snprintf way. |
|
|
836 | need (enc, IVUV_MAXCHARS); |
381 | enc->cur += |
837 | enc->cur += |
382 | SvIsUV(sv) |
838 | SvIsUV(sv) |
383 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
839 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
384 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
840 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
841 | } |
385 | } |
842 | } |
386 | else if (SvROK (sv)) |
843 | else if (SvROK (sv)) |
387 | { |
844 | encode_rv (enc, SvRV (sv)); |
388 | if (!--enc->max_recurse) |
845 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
389 | croak ("data structure too deep (hit recursion limit)"); |
|
|
390 | |
|
|
391 | sv = SvRV (sv); |
|
|
392 | |
|
|
393 | switch (SvTYPE (sv)) |
|
|
394 | { |
|
|
395 | case SVt_PVAV: encode_av (enc, (AV *)sv); break; |
|
|
396 | case SVt_PVHV: encode_hv (enc, (HV *)sv); break; |
|
|
397 | |
|
|
398 | default: |
|
|
399 | croak ("JSON can only represent references to arrays or hashes"); |
|
|
400 | } |
|
|
401 | } |
|
|
402 | else if (!SvOK (sv)) |
|
|
403 | encode_str (enc, "null", 4, 0); |
846 | encode_str (enc, "null", 4, 0); |
404 | else |
847 | else |
405 | croak ("encountered perl type that JSON cannot handle"); |
848 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data", |
|
|
849 | SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
406 | } |
850 | } |
407 | |
851 | |
408 | static SV * |
852 | static SV * |
409 | encode_json (SV *scalar, UV flags) |
853 | encode_json (SV *scalar, JSON *json) |
410 | { |
854 | { |
411 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
|
|
412 | croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)"); |
|
|
413 | |
|
|
414 | enc_t enc; |
855 | enc_t enc; |
415 | enc.flags = flags; |
856 | |
|
|
857 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
|
|
858 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
|
|
859 | |
|
|
860 | enc.json = *json; |
416 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
861 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
417 | enc.cur = SvPVX (enc.sv); |
862 | enc.cur = SvPVX (enc.sv); |
418 | enc.end = SvEND (enc.sv); |
863 | enc.end = SvEND (enc.sv); |
419 | enc.max_recurse = 0; |
|
|
420 | enc.indent = 0; |
864 | enc.indent = 0; |
|
|
865 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
866 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
867 | : 0x110000UL; |
421 | |
868 | |
422 | SvPOK_only (enc.sv); |
869 | SvPOK_only (enc.sv); |
423 | encode_sv (&enc, scalar); |
870 | encode_sv (&enc, scalar); |
|
|
871 | encode_nl (&enc); |
424 | |
872 | |
|
|
873 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
874 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
875 | |
425 | if (!(flags & (F_ASCII | F_UTF8))) |
876 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
426 | SvUTF8_on (enc.sv); |
877 | SvUTF8_on (enc.sv); |
427 | |
878 | |
428 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
429 | |
|
|
430 | #ifdef SvPV_shrink_to_cur |
|
|
431 | if (enc.flags & F_SHRINK) |
879 | if (enc.json.flags & F_SHRINK) |
432 | SvPV_shrink_to_cur (enc.sv); |
880 | shrink (enc.sv); |
433 | #endif |
881 | |
434 | return enc.sv; |
882 | return enc.sv; |
435 | } |
883 | } |
436 | |
884 | |
437 | ///////////////////////////////////////////////////////////////////////////// |
885 | ///////////////////////////////////////////////////////////////////////////// |
|
|
886 | // decoder |
438 | |
887 | |
439 | #define WS \ |
888 | // structure used for decoding JSON |
|
|
889 | typedef struct |
|
|
890 | { |
|
|
891 | char *cur; // current parser pointer |
|
|
892 | char *end; // end of input string |
|
|
893 | const char *err; // parse error, if != 0 |
|
|
894 | JSON json; |
|
|
895 | U32 depth; // recursion depth |
|
|
896 | U32 maxdepth; // recursion depth limit |
|
|
897 | } dec_t; |
|
|
898 | |
|
|
899 | INLINE void |
|
|
900 | decode_comment (dec_t *dec) |
|
|
901 | { |
|
|
902 | // only '#'-style comments allowed a.t.m. |
|
|
903 | |
|
|
904 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
905 | ++dec->cur; |
|
|
906 | } |
|
|
907 | |
|
|
908 | INLINE void |
|
|
909 | decode_ws (dec_t *dec) |
|
|
910 | { |
440 | for (;;) \ |
911 | for (;;) |
441 | { \ |
912 | { |
442 | char ch = *dec->cur; \ |
913 | char ch = *dec->cur; |
|
|
914 | |
443 | if (ch > 0x20 \ |
915 | if (ch > 0x20) |
|
|
916 | { |
|
|
917 | if (expect_false (ch == '#')) |
|
|
918 | { |
|
|
919 | if (dec->json.flags & F_RELAXED) |
|
|
920 | decode_comment (dec); |
|
|
921 | else |
|
|
922 | break; |
|
|
923 | } |
|
|
924 | else |
|
|
925 | break; |
|
|
926 | } |
444 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ |
927 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
445 | break; \ |
928 | break; // parse error, but let higher level handle it, gives better error messages |
|
|
929 | |
446 | ++dec->cur; \ |
930 | ++dec->cur; |
447 | } |
931 | } |
|
|
932 | } |
448 | |
933 | |
449 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
934 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
|
|
935 | |
450 | #define EXPECT_CH(ch) SB \ |
936 | #define EXPECT_CH(ch) SB \ |
451 | if (*dec->cur != ch) \ |
937 | if (*dec->cur != ch) \ |
452 | ERR (# ch " expected"); \ |
938 | ERR (# ch " expected"); \ |
453 | ++dec->cur; \ |
939 | ++dec->cur; \ |
454 | SE |
940 | SE |
455 | |
941 | |
|
|
942 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
|
|
943 | #define DEC_DEC_DEPTH --dec->depth |
|
|
944 | |
456 | static SV *decode_sv (dec_t *dec); |
945 | static SV *decode_sv (dec_t *dec); |
457 | |
946 | |
458 | static signed char decode_hexdigit[256]; |
947 | static signed char decode_hexdigit[256]; |
459 | |
948 | |
460 | static UV |
949 | static UV |
461 | decode_4hex (dec_t *dec) |
950 | decode_4hex (dec_t *dec) |
462 | { |
951 | { |
463 | signed char d1, d2, d3, d4; |
952 | signed char d1, d2, d3, d4; |
|
|
953 | unsigned char *cur = (unsigned char *)dec->cur; |
464 | |
954 | |
465 | d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; |
955 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
466 | if (d1 < 0) ERR ("four hexadecimal digits expected"); |
956 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
467 | d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; |
957 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
468 | if (d2 < 0) ERR ("four hexadecimal digits expected"); |
958 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
469 | d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]]; |
|
|
470 | if (d3 < 0) ERR ("four hexadecimal digits expected"); |
|
|
471 | d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]]; |
|
|
472 | if (d4 < 0) ERR ("four hexadecimal digits expected"); |
|
|
473 | |
959 | |
474 | dec->cur += 4; |
960 | dec->cur += 4; |
475 | |
961 | |
476 | return ((UV)d1) << 12 |
962 | return ((UV)d1) << 12 |
477 | | ((UV)d2) << 8 |
963 | | ((UV)d2) << 8 |
… | |
… | |
480 | |
966 | |
481 | fail: |
967 | fail: |
482 | return (UV)-1; |
968 | return (UV)-1; |
483 | } |
969 | } |
484 | |
970 | |
485 | #define APPEND_GROW(n) SB \ |
|
|
486 | if (cur + (n) >= end) \ |
|
|
487 | { \ |
|
|
488 | STRLEN ofs = cur - SvPVX (sv); \ |
|
|
489 | SvGROW (sv, ofs + (n) + 1); \ |
|
|
490 | cur = SvPVX (sv) + ofs; \ |
|
|
491 | end = SvEND (sv); \ |
|
|
492 | } \ |
|
|
493 | SE |
|
|
494 | |
|
|
495 | #define APPEND_CH(ch) SB \ |
|
|
496 | APPEND_GROW (1); \ |
|
|
497 | *cur++ = (ch); \ |
|
|
498 | SE |
|
|
499 | |
|
|
500 | static SV * |
971 | static SV * |
501 | decode_str (dec_t *dec) |
972 | decode_str (dec_t *dec) |
502 | { |
973 | { |
503 | SV *sv = NEWSV (0,2); |
974 | SV *sv = 0; |
504 | int utf8 = 0; |
975 | int utf8 = 0; |
505 | char *cur = SvPVX (sv); |
976 | char *dec_cur = dec->cur; |
506 | char *end = SvEND (sv); |
|
|
507 | |
977 | |
508 | for (;;) |
978 | do |
509 | { |
979 | { |
510 | unsigned char ch = *(unsigned char *)dec->cur; |
980 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
|
|
981 | char *cur = buf; |
511 | |
982 | |
512 | if (ch == '"') |
983 | do |
513 | break; |
|
|
514 | else if (ch == '\\') |
|
|
515 | { |
984 | { |
516 | switch (*++dec->cur) |
985 | unsigned char ch = *(unsigned char *)dec_cur++; |
|
|
986 | |
|
|
987 | if (expect_false (ch == '"')) |
|
|
988 | { |
|
|
989 | --dec_cur; |
|
|
990 | break; |
517 | { |
991 | } |
518 | case '\\': |
992 | else if (expect_false (ch == '\\')) |
519 | case '/': |
993 | { |
520 | case '"': APPEND_CH (*dec->cur++); break; |
994 | switch (*dec_cur) |
521 | |
|
|
522 | case 'b': APPEND_CH ('\010'); ++dec->cur; break; |
|
|
523 | case 't': APPEND_CH ('\011'); ++dec->cur; break; |
|
|
524 | case 'n': APPEND_CH ('\012'); ++dec->cur; break; |
|
|
525 | case 'f': APPEND_CH ('\014'); ++dec->cur; break; |
|
|
526 | case 'r': APPEND_CH ('\015'); ++dec->cur; break; |
|
|
527 | |
|
|
528 | case 'u': |
|
|
529 | { |
995 | { |
530 | UV lo, hi; |
996 | case '\\': |
531 | ++dec->cur; |
997 | case '/': |
|
|
998 | case '"': *cur++ = *dec_cur++; break; |
532 | |
999 | |
533 | hi = decode_4hex (dec); |
1000 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
534 | if (hi == (UV)-1) |
1001 | case 't': ++dec_cur; *cur++ = '\011'; break; |
535 | goto fail; |
1002 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
|
|
1003 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
|
|
1004 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
536 | |
1005 | |
537 | // possibly a surrogate pair |
1006 | case 'u': |
538 | if (hi >= 0xd800 && hi < 0xdc00) |
|
|
539 | { |
1007 | { |
540 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
1008 | UV lo, hi; |
541 | ERR ("missing low surrogate character in surrogate pair"); |
1009 | ++dec_cur; |
542 | |
1010 | |
543 | dec->cur += 2; |
1011 | dec->cur = dec_cur; |
544 | |
|
|
545 | lo = decode_4hex (dec); |
1012 | hi = decode_4hex (dec); |
|
|
1013 | dec_cur = dec->cur; |
546 | if (lo == (UV)-1) |
1014 | if (hi == (UV)-1) |
547 | goto fail; |
1015 | goto fail; |
548 | |
1016 | |
|
|
1017 | // possibly a surrogate pair |
|
|
1018 | if (hi >= 0xd800) |
|
|
1019 | if (hi < 0xdc00) |
|
|
1020 | { |
|
|
1021 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
|
|
1022 | ERR ("missing low surrogate character in surrogate pair"); |
|
|
1023 | |
|
|
1024 | dec_cur += 2; |
|
|
1025 | |
|
|
1026 | dec->cur = dec_cur; |
|
|
1027 | lo = decode_4hex (dec); |
|
|
1028 | dec_cur = dec->cur; |
|
|
1029 | if (lo == (UV)-1) |
|
|
1030 | goto fail; |
|
|
1031 | |
549 | if (lo < 0xdc00 || lo >= 0xe000) |
1032 | if (lo < 0xdc00 || lo >= 0xe000) |
550 | ERR ("surrogate pair expected"); |
1033 | ERR ("surrogate pair expected"); |
551 | |
1034 | |
552 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
1035 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
|
|
1036 | } |
|
|
1037 | else if (hi < 0xe000) |
|
|
1038 | ERR ("missing high surrogate character in surrogate pair"); |
|
|
1039 | |
|
|
1040 | if (hi >= 0x80) |
|
|
1041 | { |
|
|
1042 | utf8 = 1; |
|
|
1043 | |
|
|
1044 | cur = encode_utf8 (cur, hi); |
|
|
1045 | } |
|
|
1046 | else |
|
|
1047 | *cur++ = hi; |
553 | } |
1048 | } |
554 | else if (hi >= 0xdc00 && hi < 0xe000) |
|
|
555 | ERR ("missing high surrogate character in surrogate pair"); |
|
|
556 | |
|
|
557 | if (hi >= 0x80) |
|
|
558 | { |
1049 | break; |
559 | utf8 = 1; |
|
|
560 | |
1050 | |
561 | APPEND_GROW (4); // at most 4 bytes for 21 bits |
|
|
562 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
|
|
563 | } |
|
|
564 | else |
1051 | default: |
565 | APPEND_CH (hi); |
1052 | --dec_cur; |
|
|
1053 | ERR ("illegal backslash escape sequence in string"); |
566 | } |
1054 | } |
567 | break; |
1055 | } |
|
|
1056 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
|
|
1057 | *cur++ = ch; |
|
|
1058 | else if (ch >= 0x80) |
|
|
1059 | { |
|
|
1060 | STRLEN clen; |
568 | |
1061 | |
569 | default: |
|
|
570 | --dec->cur; |
1062 | --dec_cur; |
571 | ERR ("illegal backslash escape sequence in string"); |
1063 | |
|
|
1064 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
|
|
1065 | if (clen == (STRLEN)-1) |
|
|
1066 | ERR ("malformed UTF-8 character in JSON string"); |
|
|
1067 | |
|
|
1068 | do |
|
|
1069 | *cur++ = *dec_cur++; |
|
|
1070 | while (--clen); |
|
|
1071 | |
|
|
1072 | utf8 = 1; |
|
|
1073 | } |
|
|
1074 | else |
|
|
1075 | { |
|
|
1076 | --dec_cur; |
|
|
1077 | |
|
|
1078 | if (!ch) |
|
|
1079 | ERR ("unexpected end of string while parsing JSON string"); |
|
|
1080 | else |
|
|
1081 | ERR ("invalid character encountered while parsing JSON string"); |
572 | } |
1082 | } |
573 | } |
1083 | } |
574 | else if (ch >= 0x20 && ch <= 0x7f) |
1084 | while (cur < buf + SHORT_STRING_LEN); |
575 | APPEND_CH (*dec->cur++); |
1085 | |
576 | else if (ch >= 0x80) |
|
|
577 | { |
1086 | { |
578 | STRLEN clen; |
1087 | STRLEN len = cur - buf; |
579 | UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); |
|
|
580 | if (clen == (STRLEN)-1) |
|
|
581 | ERR ("malformed UTF-8 character in string, cannot convert to JSON"); |
|
|
582 | |
1088 | |
583 | APPEND_GROW (clen); |
1089 | if (sv) |
584 | do |
|
|
585 | { |
1090 | { |
586 | *cur++ = *dec->cur++; |
1091 | STRLEN cur = SvCUR (sv); |
|
|
1092 | |
|
|
1093 | if (SvLEN (sv) <= cur + len) |
|
|
1094 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1095 | |
|
|
1096 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
|
|
1097 | SvCUR_set (sv, SvCUR (sv) + len); |
587 | } |
1098 | } |
588 | while (--clen); |
|
|
589 | |
|
|
590 | utf8 = 1; |
|
|
591 | } |
|
|
592 | else if (dec->cur == dec->end) |
|
|
593 | ERR ("unexpected end of string while parsing json string"); |
|
|
594 | else |
1099 | else |
595 | ERR ("invalid character encountered"); |
1100 | sv = newSVpvn (buf, len); |
596 | } |
1101 | } |
|
|
1102 | } |
|
|
1103 | while (*dec_cur != '"'); |
597 | |
1104 | |
598 | ++dec->cur; |
1105 | ++dec_cur; |
599 | |
1106 | |
600 | SvCUR_set (sv, cur - SvPVX (sv)); |
1107 | if (sv) |
601 | |
1108 | { |
602 | SvPOK_only (sv); |
1109 | SvPOK_only (sv); |
603 | *SvEND (sv) = 0; |
1110 | *SvEND (sv) = 0; |
604 | |
1111 | |
605 | if (utf8) |
1112 | if (utf8) |
606 | SvUTF8_on (sv); |
1113 | SvUTF8_on (sv); |
|
|
1114 | } |
|
|
1115 | else |
|
|
1116 | sv = newSVpvn ("", 0); |
607 | |
1117 | |
608 | #ifdef SvPV_shrink_to_cur |
1118 | dec->cur = dec_cur; |
609 | if (dec->flags & F_SHRINK) |
|
|
610 | SvPV_shrink_to_cur (sv); |
|
|
611 | #endif |
|
|
612 | |
|
|
613 | return sv; |
1119 | return sv; |
614 | |
1120 | |
615 | fail: |
1121 | fail: |
616 | SvREFCNT_dec (sv); |
1122 | dec->cur = dec_cur; |
617 | return 0; |
1123 | return 0; |
618 | } |
1124 | } |
619 | |
1125 | |
620 | static SV * |
1126 | static SV * |
621 | decode_num (dec_t *dec) |
1127 | decode_num (dec_t *dec) |
… | |
… | |
679 | is_nv = 1; |
1185 | is_nv = 1; |
680 | } |
1186 | } |
681 | |
1187 | |
682 | if (!is_nv) |
1188 | if (!is_nv) |
683 | { |
1189 | { |
684 | UV uv; |
1190 | int len = dec->cur - start; |
685 | int numtype = grok_number (start, dec->cur - start, &uv); |
1191 | |
686 | if (numtype & IS_NUMBER_IN_UV) |
1192 | // special case the rather common 1..5-digit-int case |
687 | if (numtype & IS_NUMBER_NEG) |
1193 | if (*start == '-') |
|
|
1194 | switch (len) |
688 | { |
1195 | { |
689 | if (uv < (UV)IV_MIN) |
1196 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
690 | return newSViv (-(IV)uv); |
1197 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
|
|
1198 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
1199 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1200 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
691 | } |
1201 | } |
|
|
1202 | else |
|
|
1203 | switch (len) |
|
|
1204 | { |
|
|
1205 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
1206 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
1207 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
1208 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1209 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
|
|
1210 | } |
|
|
1211 | |
|
|
1212 | { |
|
|
1213 | UV uv; |
|
|
1214 | int numtype = grok_number (start, len, &uv); |
|
|
1215 | if (numtype & IS_NUMBER_IN_UV) |
|
|
1216 | if (numtype & IS_NUMBER_NEG) |
|
|
1217 | { |
|
|
1218 | if (uv < (UV)IV_MIN) |
|
|
1219 | return newSViv (-(IV)uv); |
|
|
1220 | } |
692 | else |
1221 | else |
693 | return newSVuv (uv); |
1222 | return newSVuv (uv); |
694 | } |
1223 | } |
695 | |
1224 | |
|
|
1225 | len -= *start == '-' ? 1 : 0; |
|
|
1226 | |
|
|
1227 | // does not fit into IV or UV, try NV |
|
|
1228 | if (len <= NV_DIG) |
|
|
1229 | // fits into NV without loss of precision |
|
|
1230 | return newSVnv (json_atof (start)); |
|
|
1231 | |
|
|
1232 | // everything else fails, convert it to a string |
|
|
1233 | return newSVpvn (start, dec->cur - start); |
|
|
1234 | } |
|
|
1235 | |
|
|
1236 | // loss of precision here |
696 | return newSVnv (Atof (start)); |
1237 | return newSVnv (json_atof (start)); |
697 | |
1238 | |
698 | fail: |
1239 | fail: |
699 | return 0; |
1240 | return 0; |
700 | } |
1241 | } |
701 | |
1242 | |
702 | static SV * |
1243 | static SV * |
703 | decode_av (dec_t *dec) |
1244 | decode_av (dec_t *dec) |
704 | { |
1245 | { |
705 | AV *av = newAV (); |
1246 | AV *av = newAV (); |
706 | |
1247 | |
707 | WS; |
1248 | DEC_INC_DEPTH; |
|
|
1249 | decode_ws (dec); |
|
|
1250 | |
708 | if (*dec->cur == ']') |
1251 | if (*dec->cur == ']') |
709 | ++dec->cur; |
1252 | ++dec->cur; |
710 | else |
1253 | else |
711 | for (;;) |
1254 | for (;;) |
712 | { |
1255 | { |
… | |
… | |
716 | if (!value) |
1259 | if (!value) |
717 | goto fail; |
1260 | goto fail; |
718 | |
1261 | |
719 | av_push (av, value); |
1262 | av_push (av, value); |
720 | |
1263 | |
721 | WS; |
1264 | decode_ws (dec); |
722 | |
1265 | |
723 | if (*dec->cur == ']') |
1266 | if (*dec->cur == ']') |
724 | { |
1267 | { |
725 | ++dec->cur; |
1268 | ++dec->cur; |
726 | break; |
1269 | break; |
727 | } |
1270 | } |
728 | |
1271 | |
729 | if (*dec->cur != ',') |
1272 | if (*dec->cur != ',') |
730 | ERR (", or ] expected while parsing array"); |
1273 | ERR (", or ] expected while parsing array"); |
731 | |
1274 | |
732 | ++dec->cur; |
1275 | ++dec->cur; |
|
|
1276 | |
|
|
1277 | decode_ws (dec); |
|
|
1278 | |
|
|
1279 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1280 | { |
|
|
1281 | ++dec->cur; |
|
|
1282 | break; |
|
|
1283 | } |
733 | } |
1284 | } |
734 | |
1285 | |
|
|
1286 | DEC_DEC_DEPTH; |
735 | return newRV_noinc ((SV *)av); |
1287 | return newRV_noinc ((SV *)av); |
736 | |
1288 | |
737 | fail: |
1289 | fail: |
738 | SvREFCNT_dec (av); |
1290 | SvREFCNT_dec (av); |
|
|
1291 | DEC_DEC_DEPTH; |
739 | return 0; |
1292 | return 0; |
740 | } |
1293 | } |
741 | |
1294 | |
742 | static SV * |
1295 | static SV * |
743 | decode_hv (dec_t *dec) |
1296 | decode_hv (dec_t *dec) |
744 | { |
1297 | { |
|
|
1298 | SV *sv; |
745 | HV *hv = newHV (); |
1299 | HV *hv = newHV (); |
746 | |
1300 | |
747 | WS; |
1301 | DEC_INC_DEPTH; |
|
|
1302 | decode_ws (dec); |
|
|
1303 | |
748 | if (*dec->cur == '}') |
1304 | if (*dec->cur == '}') |
749 | ++dec->cur; |
1305 | ++dec->cur; |
750 | else |
1306 | else |
751 | for (;;) |
1307 | for (;;) |
752 | { |
1308 | { |
753 | SV *key, *value; |
|
|
754 | |
|
|
755 | WS; EXPECT_CH ('"'); |
1309 | EXPECT_CH ('"'); |
756 | |
1310 | |
757 | key = decode_str (dec); |
1311 | // heuristic: assume that |
758 | if (!key) |
1312 | // a) decode_str + hv_store_ent are abysmally slow. |
759 | goto fail; |
1313 | // b) most hash keys are short, simple ascii text. |
|
|
1314 | // => try to "fast-match" such strings to avoid |
|
|
1315 | // the overhead of decode_str + hv_store_ent. |
|
|
1316 | { |
|
|
1317 | SV *value; |
|
|
1318 | char *p = dec->cur; |
|
|
1319 | char *e = p + 24; // only try up to 24 bytes |
760 | |
1320 | |
761 | WS; EXPECT_CH (':'); |
1321 | for (;;) |
762 | |
|
|
763 | value = decode_sv (dec); |
|
|
764 | if (!value) |
|
|
765 | { |
1322 | { |
|
|
1323 | // the >= 0x80 is false on most architectures |
|
|
1324 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1325 | { |
|
|
1326 | // slow path, back up and use decode_str |
|
|
1327 | SV *key = decode_str (dec); |
|
|
1328 | if (!key) |
|
|
1329 | goto fail; |
|
|
1330 | |
|
|
1331 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1332 | |
|
|
1333 | decode_ws (dec); |
|
|
1334 | value = decode_sv (dec); |
|
|
1335 | if (!value) |
|
|
1336 | { |
|
|
1337 | SvREFCNT_dec (key); |
|
|
1338 | goto fail; |
|
|
1339 | } |
|
|
1340 | |
|
|
1341 | hv_store_ent (hv, key, value, 0); |
766 | SvREFCNT_dec (key); |
1342 | SvREFCNT_dec (key); |
|
|
1343 | |
|
|
1344 | break; |
|
|
1345 | } |
|
|
1346 | else if (*p == '"') |
|
|
1347 | { |
|
|
1348 | // fast path, got a simple key |
|
|
1349 | char *key = dec->cur; |
|
|
1350 | int len = p - key; |
|
|
1351 | dec->cur = p + 1; |
|
|
1352 | |
|
|
1353 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1354 | |
|
|
1355 | decode_ws (dec); |
|
|
1356 | value = decode_sv (dec); |
|
|
1357 | if (!value) |
767 | goto fail; |
1358 | goto fail; |
|
|
1359 | |
|
|
1360 | hv_store (hv, key, len, value, 0); |
|
|
1361 | |
|
|
1362 | break; |
|
|
1363 | } |
|
|
1364 | |
|
|
1365 | ++p; |
768 | } |
1366 | } |
769 | |
|
|
770 | //TODO: optimise |
|
|
771 | hv_store_ent (hv, key, value, 0); |
|
|
772 | |
|
|
773 | WS; |
1367 | } |
|
|
1368 | |
|
|
1369 | decode_ws (dec); |
774 | |
1370 | |
775 | if (*dec->cur == '}') |
1371 | if (*dec->cur == '}') |
776 | { |
1372 | { |
777 | ++dec->cur; |
1373 | ++dec->cur; |
778 | break; |
1374 | break; |
… | |
… | |
780 | |
1376 | |
781 | if (*dec->cur != ',') |
1377 | if (*dec->cur != ',') |
782 | ERR (", or } expected while parsing object/hash"); |
1378 | ERR (", or } expected while parsing object/hash"); |
783 | |
1379 | |
784 | ++dec->cur; |
1380 | ++dec->cur; |
|
|
1381 | |
|
|
1382 | decode_ws (dec); |
|
|
1383 | |
|
|
1384 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1385 | { |
|
|
1386 | ++dec->cur; |
|
|
1387 | break; |
|
|
1388 | } |
785 | } |
1389 | } |
786 | |
1390 | |
|
|
1391 | DEC_DEC_DEPTH; |
787 | return newRV_noinc ((SV *)hv); |
1392 | sv = newRV_noinc ((SV *)hv); |
|
|
1393 | |
|
|
1394 | // check filter callbacks |
|
|
1395 | if (dec->json.flags & F_HOOK) |
|
|
1396 | { |
|
|
1397 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1398 | { |
|
|
1399 | HE *cb, *he; |
|
|
1400 | |
|
|
1401 | hv_iterinit (hv); |
|
|
1402 | he = hv_iternext (hv); |
|
|
1403 | hv_iterinit (hv); |
|
|
1404 | |
|
|
1405 | // the next line creates a mortal sv each time its called. |
|
|
1406 | // might want to optimise this for common cases. |
|
|
1407 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1408 | |
|
|
1409 | if (cb) |
|
|
1410 | { |
|
|
1411 | dSP; |
|
|
1412 | int count; |
|
|
1413 | |
|
|
1414 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1415 | XPUSHs (HeVAL (he)); |
|
|
1416 | sv_2mortal (sv); |
|
|
1417 | |
|
|
1418 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1419 | |
|
|
1420 | if (count == 1) |
|
|
1421 | { |
|
|
1422 | sv = newSVsv (POPs); |
|
|
1423 | FREETMPS; LEAVE; |
|
|
1424 | return sv; |
|
|
1425 | } |
|
|
1426 | |
|
|
1427 | SvREFCNT_inc (sv); |
|
|
1428 | FREETMPS; LEAVE; |
|
|
1429 | } |
|
|
1430 | } |
|
|
1431 | |
|
|
1432 | if (dec->json.cb_object) |
|
|
1433 | { |
|
|
1434 | dSP; |
|
|
1435 | int count; |
|
|
1436 | |
|
|
1437 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1438 | XPUSHs (sv_2mortal (sv)); |
|
|
1439 | |
|
|
1440 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1441 | |
|
|
1442 | if (count == 1) |
|
|
1443 | { |
|
|
1444 | sv = newSVsv (POPs); |
|
|
1445 | FREETMPS; LEAVE; |
|
|
1446 | return sv; |
|
|
1447 | } |
|
|
1448 | |
|
|
1449 | SvREFCNT_inc (sv); |
|
|
1450 | FREETMPS; LEAVE; |
|
|
1451 | } |
|
|
1452 | } |
|
|
1453 | |
|
|
1454 | return sv; |
788 | |
1455 | |
789 | fail: |
1456 | fail: |
790 | SvREFCNT_dec (hv); |
1457 | SvREFCNT_dec (hv); |
|
|
1458 | DEC_DEC_DEPTH; |
791 | return 0; |
1459 | return 0; |
792 | } |
1460 | } |
793 | |
1461 | |
794 | static SV * |
1462 | static SV * |
795 | decode_sv (dec_t *dec) |
1463 | decode_sv (dec_t *dec) |
796 | { |
1464 | { |
797 | WS; |
1465 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1466 | // to parse everything. |
798 | switch (*dec->cur) |
1467 | switch (*dec->cur) |
799 | { |
1468 | { |
800 | case '"': ++dec->cur; return decode_str (dec); |
1469 | case '"': ++dec->cur; return decode_str (dec); |
801 | case '[': ++dec->cur; return decode_av (dec); |
1470 | case '[': ++dec->cur; return decode_av (dec); |
802 | case '{': ++dec->cur; return decode_hv (dec); |
1471 | case '{': ++dec->cur; return decode_hv (dec); |
803 | |
1472 | |
804 | case '-': |
1473 | case '-': |
805 | case '0': case '1': case '2': case '3': case '4': |
1474 | case '0': case '1': case '2': case '3': case '4': |
806 | case '5': case '6': case '7': case '8': case '9': |
1475 | case '5': case '6': case '7': case '8': case '9': |
807 | return decode_num (dec); |
1476 | return decode_num (dec); |
808 | |
1477 | |
809 | case 't': |
1478 | case 't': |
810 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1479 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
811 | { |
1480 | { |
812 | dec->cur += 4; |
1481 | dec->cur += 4; |
|
|
1482 | #if JSON_SLOW |
|
|
1483 | json_true = get_bool ("JSON::XS::true"); |
|
|
1484 | #endif |
813 | return newSViv (1); |
1485 | return newSVsv (json_true); |
814 | } |
1486 | } |
815 | else |
1487 | else |
816 | ERR ("'true' expected"); |
1488 | ERR ("'true' expected"); |
817 | |
1489 | |
818 | break; |
1490 | break; |
819 | |
1491 | |
820 | case 'f': |
1492 | case 'f': |
821 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1493 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
822 | { |
1494 | { |
823 | dec->cur += 5; |
1495 | dec->cur += 5; |
|
|
1496 | #if JSON_SLOW |
|
|
1497 | json_false = get_bool ("JSON::XS::false"); |
|
|
1498 | #endif |
824 | return newSViv (0); |
1499 | return newSVsv (json_false); |
825 | } |
1500 | } |
826 | else |
1501 | else |
827 | ERR ("'false' expected"); |
1502 | ERR ("'false' expected"); |
828 | |
1503 | |
829 | break; |
1504 | break; |
… | |
… | |
838 | ERR ("'null' expected"); |
1513 | ERR ("'null' expected"); |
839 | |
1514 | |
840 | break; |
1515 | break; |
841 | |
1516 | |
842 | default: |
1517 | default: |
843 | ERR ("malformed json string"); |
1518 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
844 | break; |
1519 | break; |
845 | } |
1520 | } |
846 | |
1521 | |
847 | fail: |
1522 | fail: |
848 | return 0; |
1523 | return 0; |
849 | } |
1524 | } |
850 | |
1525 | |
851 | static SV * |
1526 | static SV * |
852 | decode_json (SV *string, UV flags) |
1527 | decode_json (SV *string, JSON *json, char **offset_return) |
853 | { |
1528 | { |
|
|
1529 | dec_t dec; |
854 | SV *sv; |
1530 | SV *sv; |
855 | |
1531 | |
856 | if (flags & F_UTF8) |
1532 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1533 | * makes perl ignore the magic in subsequent accesses. |
|
|
1534 | * also make a copy of non-PV values, to get them into a clean |
|
|
1535 | * state (SvPV should do that, but it's buggy, see below). |
|
|
1536 | */ |
|
|
1537 | /*SvGETMAGIC (string);*/ |
|
|
1538 | if (SvMAGICAL (string) || !SvPOK (string)) |
|
|
1539 | string = sv_2mortal (newSVsv (string)); |
|
|
1540 | |
|
|
1541 | SvUPGRADE (string, SVt_PV); |
|
|
1542 | |
|
|
1543 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1544 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1545 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1546 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1547 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1548 | * and hope for the best. |
|
|
1549 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1550 | * assertion business is seriously broken, try yet another workaround |
|
|
1551 | * for the broken -DDEBUGGING. |
|
|
1552 | */ |
|
|
1553 | { |
|
|
1554 | #ifdef DEBUGGING |
|
|
1555 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1556 | #else |
|
|
1557 | STRLEN offset = SvCUR (string); |
|
|
1558 | #endif |
|
|
1559 | |
|
|
1560 | if (offset > json->max_size && json->max_size) |
|
|
1561 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1562 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1563 | } |
|
|
1564 | |
|
|
1565 | if (DECODE_WANTS_OCTETS (json)) |
857 | sv_utf8_downgrade (string, 0); |
1566 | sv_utf8_downgrade (string, 0); |
858 | else |
1567 | else |
859 | sv_utf8_upgrade (string); |
1568 | sv_utf8_upgrade (string); |
860 | |
1569 | |
861 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1570 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
862 | |
1571 | |
863 | dec_t dec; |
1572 | dec.json = *json; |
864 | dec.flags = flags; |
|
|
865 | dec.cur = SvPVX (string); |
1573 | dec.cur = SvPVX (string); |
866 | dec.end = SvEND (string); |
1574 | dec.end = SvEND (string); |
867 | dec.err = 0; |
1575 | dec.err = 0; |
|
|
1576 | dec.depth = 0; |
868 | |
1577 | |
|
|
1578 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1579 | dec.json.flags |= F_HOOK; |
|
|
1580 | |
|
|
1581 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1582 | |
|
|
1583 | decode_ws (&dec); |
869 | sv = decode_sv (&dec); |
1584 | sv = decode_sv (&dec); |
870 | |
1585 | |
|
|
1586 | if (offset_return) |
|
|
1587 | *offset_return = dec.cur; |
|
|
1588 | |
|
|
1589 | if (!(offset_return || !sv)) |
|
|
1590 | { |
|
|
1591 | // check for trailing garbage |
|
|
1592 | decode_ws (&dec); |
|
|
1593 | |
|
|
1594 | if (*dec.cur) |
|
|
1595 | { |
|
|
1596 | dec.err = "garbage after JSON object"; |
|
|
1597 | SvREFCNT_dec (sv); |
|
|
1598 | sv = 0; |
|
|
1599 | } |
|
|
1600 | } |
|
|
1601 | |
871 | if (!sv) |
1602 | if (!sv) |
872 | { |
1603 | { |
873 | IV offset = utf8_distance (dec.cur, SvPVX (string)); |
|
|
874 | SV *uni = sv_newmortal (); |
1604 | SV *uni = sv_newmortal (); |
|
|
1605 | |
875 | // horrible hack to silence warning inside pv_uni_display |
1606 | // horrible hack to silence warning inside pv_uni_display |
876 | COP cop; |
1607 | COP cop = *PL_curcop; |
877 | memset (&cop, 0, sizeof (cop)); |
|
|
878 | cop.cop_warnings = pWARN_NONE; |
1608 | cop.cop_warnings = pWARN_NONE; |
|
|
1609 | ENTER; |
879 | SAVEVPTR (PL_curcop); |
1610 | SAVEVPTR (PL_curcop); |
880 | PL_curcop = &cop; |
1611 | PL_curcop = &cop; |
881 | |
|
|
882 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1612 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
|
|
1613 | LEAVE; |
|
|
1614 | |
883 | croak ("%s, at character offset %d (%s)", |
1615 | croak ("%s, at character offset %d (before \"%s\")", |
884 | dec.err, |
1616 | dec.err, |
885 | (int)offset, |
1617 | (int)ptr_to_index (string, dec.cur), |
886 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1618 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
887 | } |
1619 | } |
888 | |
1620 | |
889 | sv = sv_2mortal (sv); |
1621 | sv = sv_2mortal (sv); |
890 | |
1622 | |
891 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1623 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
892 | croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); |
1624 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
893 | |
1625 | |
894 | return sv; |
1626 | return sv; |
895 | } |
1627 | } |
896 | |
1628 | |
|
|
1629 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1630 | // incremental parser |
|
|
1631 | |
|
|
1632 | static void |
|
|
1633 | incr_parse (JSON *self) |
|
|
1634 | { |
|
|
1635 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1636 | |
|
|
1637 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1638 | // but this would make it slower, so... |
|
|
1639 | |
|
|
1640 | for (;;) |
|
|
1641 | { |
|
|
1642 | //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 |
|
|
1643 | switch (self->incr_mode) |
|
|
1644 | { |
|
|
1645 | // only used for initial whitespace skipping |
|
|
1646 | case INCR_M_WS: |
|
|
1647 | for (;;) |
|
|
1648 | { |
|
|
1649 | if (*p > 0x20) |
|
|
1650 | { |
|
|
1651 | if (*p == '#') |
|
|
1652 | { |
|
|
1653 | self->incr_mode = INCR_M_C0; |
|
|
1654 | goto incr_m_c; |
|
|
1655 | } |
|
|
1656 | else |
|
|
1657 | { |
|
|
1658 | self->incr_mode = INCR_M_JSON; |
|
|
1659 | goto incr_m_json; |
|
|
1660 | } |
|
|
1661 | } |
|
|
1662 | else if (!*p) |
|
|
1663 | goto interrupt; |
|
|
1664 | |
|
|
1665 | ++p; |
|
|
1666 | } |
|
|
1667 | |
|
|
1668 | // skip a single char inside a string (for \\-processing) |
|
|
1669 | case INCR_M_BS: |
|
|
1670 | if (!*p) |
|
|
1671 | goto interrupt; |
|
|
1672 | |
|
|
1673 | ++p; |
|
|
1674 | self->incr_mode = INCR_M_STR; |
|
|
1675 | goto incr_m_str; |
|
|
1676 | |
|
|
1677 | // inside #-style comments |
|
|
1678 | case INCR_M_C0: |
|
|
1679 | case INCR_M_C1: |
|
|
1680 | incr_m_c: |
|
|
1681 | for (;;) |
|
|
1682 | { |
|
|
1683 | if (*p == '\n') |
|
|
1684 | { |
|
|
1685 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1686 | break; |
|
|
1687 | } |
|
|
1688 | else if (!*p) |
|
|
1689 | goto interrupt; |
|
|
1690 | |
|
|
1691 | ++p; |
|
|
1692 | } |
|
|
1693 | |
|
|
1694 | break; |
|
|
1695 | |
|
|
1696 | // inside a string |
|
|
1697 | case INCR_M_STR: |
|
|
1698 | incr_m_str: |
|
|
1699 | for (;;) |
|
|
1700 | { |
|
|
1701 | if (*p == '"') |
|
|
1702 | { |
|
|
1703 | ++p; |
|
|
1704 | self->incr_mode = INCR_M_JSON; |
|
|
1705 | |
|
|
1706 | if (!self->incr_nest) |
|
|
1707 | goto interrupt; |
|
|
1708 | |
|
|
1709 | goto incr_m_json; |
|
|
1710 | } |
|
|
1711 | else if (*p == '\\') |
|
|
1712 | { |
|
|
1713 | ++p; // "virtually" consumes character after \ |
|
|
1714 | |
|
|
1715 | if (!*p) // if at end of string we have to switch modes |
|
|
1716 | { |
|
|
1717 | self->incr_mode = INCR_M_BS; |
|
|
1718 | goto interrupt; |
|
|
1719 | } |
|
|
1720 | } |
|
|
1721 | else if (!*p) |
|
|
1722 | goto interrupt; |
|
|
1723 | |
|
|
1724 | ++p; |
|
|
1725 | } |
|
|
1726 | |
|
|
1727 | // after initial ws, outside string |
|
|
1728 | case INCR_M_JSON: |
|
|
1729 | incr_m_json: |
|
|
1730 | for (;;) |
|
|
1731 | { |
|
|
1732 | switch (*p++) |
|
|
1733 | { |
|
|
1734 | case 0: |
|
|
1735 | --p; |
|
|
1736 | goto interrupt; |
|
|
1737 | |
|
|
1738 | case 0x09: |
|
|
1739 | case 0x0a: |
|
|
1740 | case 0x0d: |
|
|
1741 | case 0x20: |
|
|
1742 | if (!self->incr_nest) |
|
|
1743 | { |
|
|
1744 | --p; // do not eat the whitespace, let the next round do it |
|
|
1745 | goto interrupt; |
|
|
1746 | } |
|
|
1747 | break; |
|
|
1748 | |
|
|
1749 | case '"': |
|
|
1750 | self->incr_mode = INCR_M_STR; |
|
|
1751 | goto incr_m_str; |
|
|
1752 | |
|
|
1753 | case '[': |
|
|
1754 | case '{': |
|
|
1755 | if (++self->incr_nest > self->max_depth) |
|
|
1756 | croak (ERR_NESTING_EXCEEDED); |
|
|
1757 | break; |
|
|
1758 | |
|
|
1759 | case ']': |
|
|
1760 | case '}': |
|
|
1761 | if (--self->incr_nest <= 0) |
|
|
1762 | goto interrupt; |
|
|
1763 | break; |
|
|
1764 | |
|
|
1765 | case '#': |
|
|
1766 | self->incr_mode = INCR_M_C1; |
|
|
1767 | goto incr_m_c; |
|
|
1768 | } |
|
|
1769 | } |
|
|
1770 | } |
|
|
1771 | |
|
|
1772 | modechange: |
|
|
1773 | ; |
|
|
1774 | } |
|
|
1775 | |
|
|
1776 | interrupt: |
|
|
1777 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1778 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1779 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
|
|
1780 | } |
|
|
1781 | |
|
|
1782 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1783 | // XS interface functions |
|
|
1784 | |
897 | MODULE = JSON::XS PACKAGE = JSON::XS |
1785 | MODULE = JSON::XS PACKAGE = JSON::XS |
898 | |
1786 | |
899 | BOOT: |
1787 | BOOT: |
900 | { |
1788 | { |
901 | int i; |
1789 | int i; |
902 | |
1790 | |
903 | memset (decode_hexdigit, 0xff, 256); |
|
|
904 | for (i = 10; i--; ) |
1791 | for (i = 0; i < 256; ++i) |
905 | decode_hexdigit ['0' + i] = i; |
1792 | decode_hexdigit [i] = |
|
|
1793 | i >= '0' && i <= '9' ? i - '0' |
|
|
1794 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
|
|
1795 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
|
|
1796 | : -1; |
906 | |
1797 | |
907 | for (i = 7; i--; ) |
|
|
908 | { |
|
|
909 | decode_hexdigit ['a' + i] = 10 + i; |
|
|
910 | decode_hexdigit ['A' + i] = 10 + i; |
|
|
911 | } |
|
|
912 | |
|
|
913 | json_stash = gv_stashpv ("JSON::XS", 1); |
1798 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1799 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
1800 | |
|
|
1801 | json_true = get_bool ("JSON::XS::true"); |
|
|
1802 | json_false = get_bool ("JSON::XS::false"); |
|
|
1803 | |
|
|
1804 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
914 | } |
1805 | } |
915 | |
1806 | |
916 | PROTOTYPES: DISABLE |
1807 | PROTOTYPES: DISABLE |
917 | |
1808 | |
918 | SV *new (char *dummy) |
1809 | void CLONE (...) |
919 | CODE: |
1810 | CODE: |
920 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1811 | json_stash = 0; |
|
|
1812 | json_boolean_stash = 0; |
|
|
1813 | |
|
|
1814 | void new (char *klass) |
|
|
1815 | PPCODE: |
|
|
1816 | { |
|
|
1817 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1818 | SvPOK_only (pv); |
|
|
1819 | json_init ((JSON *)SvPVX (pv)); |
|
|
1820 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1821 | newRV_noinc (pv), |
|
|
1822 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1823 | ))); |
|
|
1824 | } |
|
|
1825 | |
|
|
1826 | void ascii (JSON *self, int enable = 1) |
|
|
1827 | ALIAS: |
|
|
1828 | ascii = F_ASCII |
|
|
1829 | latin1 = F_LATIN1 |
|
|
1830 | utf8 = F_UTF8 |
|
|
1831 | indent = F_INDENT |
|
|
1832 | canonical = F_CANONICAL |
|
|
1833 | space_before = F_SPACE_BEFORE |
|
|
1834 | space_after = F_SPACE_AFTER |
|
|
1835 | pretty = F_PRETTY |
|
|
1836 | allow_nonref = F_ALLOW_NONREF |
|
|
1837 | shrink = F_SHRINK |
|
|
1838 | allow_blessed = F_ALLOW_BLESSED |
|
|
1839 | convert_blessed = F_CONV_BLESSED |
|
|
1840 | relaxed = F_RELAXED |
|
|
1841 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
1842 | PPCODE: |
|
|
1843 | { |
|
|
1844 | if (enable) |
|
|
1845 | self->flags |= ix; |
|
|
1846 | else |
|
|
1847 | self->flags &= ~ix; |
|
|
1848 | |
|
|
1849 | XPUSHs (ST (0)); |
|
|
1850 | } |
|
|
1851 | |
|
|
1852 | void get_ascii (JSON *self) |
|
|
1853 | ALIAS: |
|
|
1854 | get_ascii = F_ASCII |
|
|
1855 | get_latin1 = F_LATIN1 |
|
|
1856 | get_utf8 = F_UTF8 |
|
|
1857 | get_indent = F_INDENT |
|
|
1858 | get_canonical = F_CANONICAL |
|
|
1859 | get_space_before = F_SPACE_BEFORE |
|
|
1860 | get_space_after = F_SPACE_AFTER |
|
|
1861 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1862 | get_shrink = F_SHRINK |
|
|
1863 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1864 | get_convert_blessed = F_CONV_BLESSED |
|
|
1865 | get_relaxed = F_RELAXED |
|
|
1866 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
1867 | PPCODE: |
|
|
1868 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1869 | |
|
|
1870 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
|
|
1871 | PPCODE: |
|
|
1872 | self->max_depth = max_depth; |
|
|
1873 | XPUSHs (ST (0)); |
|
|
1874 | |
|
|
1875 | U32 get_max_depth (JSON *self) |
|
|
1876 | CODE: |
|
|
1877 | RETVAL = self->max_depth; |
921 | OUTPUT: |
1878 | OUTPUT: |
922 | RETVAL |
1879 | RETVAL |
923 | |
1880 | |
924 | SV *ascii (SV *self, int enable = 1) |
1881 | void max_size (JSON *self, U32 max_size = 0) |
925 | ALIAS: |
1882 | PPCODE: |
926 | ascii = F_ASCII |
1883 | self->max_size = max_size; |
927 | utf8 = F_UTF8 |
1884 | XPUSHs (ST (0)); |
928 | indent = F_INDENT |
1885 | |
929 | canonical = F_CANONICAL |
1886 | int get_max_size (JSON *self) |
930 | space_before = F_SPACE_BEFORE |
|
|
931 | space_after = F_SPACE_AFTER |
|
|
932 | json_rpc = F_JSON_RPC |
|
|
933 | pretty = F_PRETTY |
|
|
934 | allow_nonref = F_ALLOW_NONREF |
|
|
935 | shrink = F_SHRINK |
|
|
936 | CODE: |
1887 | CODE: |
937 | { |
1888 | RETVAL = self->max_size; |
938 | UV *uv = SvJSON (self); |
|
|
939 | if (enable) |
|
|
940 | *uv |= ix; |
|
|
941 | else |
|
|
942 | *uv &= ~ix; |
|
|
943 | |
|
|
944 | RETVAL = newSVsv (self); |
|
|
945 | } |
|
|
946 | OUTPUT: |
1889 | OUTPUT: |
947 | RETVAL |
1890 | RETVAL |
948 | |
1891 | |
949 | void encode (SV *self, SV *scalar) |
1892 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
950 | PPCODE: |
1893 | PPCODE: |
951 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
1894 | { |
|
|
1895 | SvREFCNT_dec (self->cb_object); |
|
|
1896 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
952 | |
1897 | |
953 | void decode (SV *self, SV *jsonstr) |
1898 | XPUSHs (ST (0)); |
|
|
1899 | } |
|
|
1900 | |
|
|
1901 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
954 | PPCODE: |
1902 | PPCODE: |
955 | XPUSHs (decode_json (jsonstr, *SvJSON (self))); |
1903 | { |
|
|
1904 | if (!self->cb_sk_object) |
|
|
1905 | self->cb_sk_object = newHV (); |
|
|
1906 | |
|
|
1907 | if (SvOK (cb)) |
|
|
1908 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
1909 | else |
|
|
1910 | { |
|
|
1911 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
1912 | |
|
|
1913 | if (!HvKEYS (self->cb_sk_object)) |
|
|
1914 | { |
|
|
1915 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1916 | self->cb_sk_object = 0; |
|
|
1917 | } |
|
|
1918 | } |
|
|
1919 | |
|
|
1920 | XPUSHs (ST (0)); |
|
|
1921 | } |
|
|
1922 | |
|
|
1923 | void encode (JSON *self, SV *scalar) |
|
|
1924 | PPCODE: |
|
|
1925 | PUTBACK; scalar = encode_json (scalar, self); SPAGAIN; |
|
|
1926 | XPUSHs (scalar); |
|
|
1927 | |
|
|
1928 | void decode (JSON *self, SV *jsonstr) |
|
|
1929 | PPCODE: |
|
|
1930 | PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN; |
|
|
1931 | XPUSHs (jsonstr); |
|
|
1932 | |
|
|
1933 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
1934 | PPCODE: |
|
|
1935 | { |
|
|
1936 | SV *sv; |
|
|
1937 | char *offset; |
|
|
1938 | PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN; |
|
|
1939 | EXTEND (SP, 2); |
|
|
1940 | PUSHs (sv); |
|
|
1941 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1942 | } |
|
|
1943 | |
|
|
1944 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1945 | PPCODE: |
|
|
1946 | { |
|
|
1947 | if (!self->incr_text) |
|
|
1948 | self->incr_text = newSVpvn ("", 0); |
|
|
1949 | |
|
|
1950 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
1951 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
1952 | if (DECODE_WANTS_OCTETS (self)) |
|
|
1953 | { |
|
|
1954 | if (self->incr_pos) |
|
|
1955 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
1956 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
1957 | |
|
|
1958 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
1959 | } |
|
|
1960 | else |
|
|
1961 | { |
|
|
1962 | sv_utf8_upgrade (self->incr_text); |
|
|
1963 | |
|
|
1964 | if (self->incr_pos) |
|
|
1965 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1966 | - (U8 *)SvPVX (self->incr_text); |
|
|
1967 | } |
|
|
1968 | |
|
|
1969 | // append data, if any |
|
|
1970 | if (jsonstr) |
|
|
1971 | { |
|
|
1972 | /* make sure both strings have same encoding */ |
|
|
1973 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
|
|
1974 | if (SvUTF8 (jsonstr)) |
|
|
1975 | sv_utf8_downgrade (jsonstr, 0); |
|
|
1976 | else |
|
|
1977 | sv_utf8_upgrade (jsonstr); |
|
|
1978 | |
|
|
1979 | /* and then just blindly append */ |
|
|
1980 | { |
|
|
1981 | STRLEN len; |
|
|
1982 | const char *str = SvPV (jsonstr, len); |
|
|
1983 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1984 | |
|
|
1985 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1986 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1987 | |
|
|
1988 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1989 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1990 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1991 | } |
|
|
1992 | } |
|
|
1993 | |
|
|
1994 | if (GIMME_V != G_VOID) |
|
|
1995 | do |
|
|
1996 | { |
|
|
1997 | SV *sv; |
|
|
1998 | char *offset; |
|
|
1999 | |
|
|
2000 | if (!INCR_DONE (self)) |
|
|
2001 | { |
|
|
2002 | incr_parse (self); |
|
|
2003 | |
|
|
2004 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
2005 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
2006 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
2007 | |
|
|
2008 | if (!INCR_DONE (self)) |
|
|
2009 | { |
|
|
2010 | // as an optimisation, do not accumulate white space in the incr buffer |
|
|
2011 | if (self->incr_mode == INCR_M_WS && self->incr_pos) |
|
|
2012 | { |
|
|
2013 | self->incr_pos = 0; |
|
|
2014 | SvCUR_set (self->incr_text, 0); |
|
|
2015 | } |
|
|
2016 | |
|
|
2017 | break; |
|
|
2018 | } |
|
|
2019 | } |
|
|
2020 | |
|
|
2021 | PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN; |
|
|
2022 | XPUSHs (sv); |
|
|
2023 | |
|
|
2024 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
2025 | self->incr_nest = 0; |
|
|
2026 | self->incr_mode = 0; |
|
|
2027 | |
|
|
2028 | sv_chop (self->incr_text, offset); |
|
|
2029 | } |
|
|
2030 | while (GIMME_V == G_ARRAY); |
|
|
2031 | } |
|
|
2032 | |
|
|
2033 | SV *incr_text (JSON *self) |
|
|
2034 | ATTRS: lvalue |
|
|
2035 | CODE: |
|
|
2036 | { |
|
|
2037 | if (self->incr_pos) |
|
|
2038 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
2039 | |
|
|
2040 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2041 | } |
|
|
2042 | OUTPUT: |
|
|
2043 | RETVAL |
|
|
2044 | |
|
|
2045 | void incr_skip (JSON *self) |
|
|
2046 | CODE: |
|
|
2047 | { |
|
|
2048 | if (self->incr_pos) |
|
|
2049 | { |
|
|
2050 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2051 | self->incr_pos = 0; |
|
|
2052 | self->incr_nest = 0; |
|
|
2053 | self->incr_mode = 0; |
|
|
2054 | } |
|
|
2055 | } |
|
|
2056 | |
|
|
2057 | void incr_reset (JSON *self) |
|
|
2058 | CODE: |
|
|
2059 | { |
|
|
2060 | SvREFCNT_dec (self->incr_text); |
|
|
2061 | self->incr_text = 0; |
|
|
2062 | self->incr_pos = 0; |
|
|
2063 | self->incr_nest = 0; |
|
|
2064 | self->incr_mode = 0; |
|
|
2065 | } |
|
|
2066 | |
|
|
2067 | void DESTROY (JSON *self) |
|
|
2068 | CODE: |
|
|
2069 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2070 | SvREFCNT_dec (self->cb_object); |
|
|
2071 | SvREFCNT_dec (self->incr_text); |
956 | |
2072 | |
957 | PROTOTYPES: ENABLE |
2073 | PROTOTYPES: ENABLE |
958 | |
2074 | |
959 | void to_json (SV *scalar) |
2075 | void encode_json (SV *scalar) |
960 | PPCODE: |
2076 | PPCODE: |
961 | XPUSHs (encode_json (scalar, F_UTF8)); |
2077 | { |
|
|
2078 | JSON json; |
|
|
2079 | json_init (&json); |
|
|
2080 | json.flags |= F_UTF8; |
|
|
2081 | PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN; |
|
|
2082 | XPUSHs (scalar); |
|
|
2083 | } |
962 | |
2084 | |
963 | void from_json (SV *jsonstr) |
2085 | void decode_json (SV *jsonstr) |
964 | PPCODE: |
2086 | PPCODE: |
965 | XPUSHs (decode_json (jsonstr, F_UTF8)); |
2087 | { |
|
|
2088 | JSON json; |
|
|
2089 | json_init (&json); |
|
|
2090 | json.flags |= F_UTF8; |
|
|
2091 | PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN; |
|
|
2092 | XPUSHs (jsonstr); |
|
|
2093 | } |
966 | |
2094 | |