… | |
… | |
4 | |
4 | |
5 | #include <assert.h> |
5 | #include <assert.h> |
6 | #include <string.h> |
6 | #include <string.h> |
7 | #include <stdlib.h> |
7 | #include <stdlib.h> |
8 | #include <stdio.h> |
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
9 | #include <float.h> |
10 | #include <float.h> |
10 | |
11 | |
11 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | #endif |
14 | #endif |
14 | |
15 | |
15 | // some old perls do not have this, try to make it work, no |
16 | // some old perls do not have this, try to make it work, no |
16 | // guarentees, though. if it breaks, you get to keep the pieces. |
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
17 | #ifndef UTF8_MAXBYTES |
18 | #ifndef UTF8_MAXBYTES |
18 | # define UTF8_MAXBYTES 13 |
19 | # define UTF8_MAXBYTES 13 |
19 | #endif |
20 | #endif |
|
|
21 | |
|
|
22 | // three extra for rounding, sign, and end of string |
|
|
23 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
20 | |
24 | |
21 | #define F_ASCII 0x00000001UL |
25 | #define F_ASCII 0x00000001UL |
22 | #define F_LATIN1 0x00000002UL |
26 | #define F_LATIN1 0x00000002UL |
23 | #define F_UTF8 0x00000004UL |
27 | #define F_UTF8 0x00000004UL |
24 | #define F_INDENT 0x00000008UL |
28 | #define F_INDENT 0x00000008UL |
… | |
… | |
27 | #define F_SPACE_AFTER 0x00000040UL |
31 | #define F_SPACE_AFTER 0x00000040UL |
28 | #define F_ALLOW_NONREF 0x00000100UL |
32 | #define F_ALLOW_NONREF 0x00000100UL |
29 | #define F_SHRINK 0x00000200UL |
33 | #define F_SHRINK 0x00000200UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
34 | #define F_ALLOW_BLESSED 0x00000400UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
35 | #define F_CONV_BLESSED 0x00000800UL |
32 | #define F_MAXDEPTH 0xf8000000UL |
|
|
33 | #define S_MAXDEPTH 27 |
|
|
34 | #define F_MAXSIZE 0x01f00000UL |
36 | #define F_RELAXED 0x00001000UL |
35 | #define S_MAXSIZE 20 |
37 | #define F_ALLOW_UNKNOWN 0x00002000UL |
36 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
37 | |
39 | |
38 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
|
|
39 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
|
|
40 | |
|
|
41 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
42 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
43 | |
41 | |
44 | #define INIT_SIZE 32 // initial scalar size to be allocated |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
45 | #define INDENT_STEP 3 // spaces per indentation level |
43 | #define INDENT_STEP 3 // spaces per indentation level |
46 | |
44 | |
47 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
45 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
48 | |
46 | |
49 | #define SB do { |
47 | #define SB do { |
50 | #define SE } while (0) |
48 | #define SE } while (0) |
51 | |
49 | |
52 | #if __GNUC__ >= 3 |
50 | #if __GNUC__ >= 3 |
53 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
51 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
54 | # define inline inline |
52 | # define INLINE static inline |
55 | #else |
53 | #else |
56 | # define expect(expr,value) (expr) |
54 | # define expect(expr,value) (expr) |
57 | # define inline static |
55 | # define INLINE static |
58 | #endif |
56 | #endif |
59 | |
57 | |
60 | #define expect_false(expr) expect ((expr) != 0, 0) |
58 | #define expect_false(expr) expect ((expr) != 0, 0) |
61 | #define expect_true(expr) expect ((expr) != 0, 1) |
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?)" |
62 | |
66 | |
63 | #ifdef USE_ITHREADS |
67 | #ifdef USE_ITHREADS |
64 | # define JSON_SLOW 1 |
68 | # define JSON_SLOW 1 |
65 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
69 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
66 | #else |
70 | #else |
… | |
… | |
69 | #endif |
73 | #endif |
70 | |
74 | |
71 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
75 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
72 | static SV *json_true, *json_false; |
76 | static SV *json_true, *json_false; |
73 | |
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 | |
74 | typedef struct { |
89 | typedef struct { |
75 | U32 flags; |
90 | U32 flags; |
|
|
91 | U32 max_depth; |
|
|
92 | STRLEN max_size; |
|
|
93 | |
76 | SV *cb_object; |
94 | SV *cb_object; |
77 | HV *cb_sk_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; |
78 | } JSON; |
102 | } JSON; |
|
|
103 | |
|
|
104 | INLINE void |
|
|
105 | json_init (JSON *json) |
|
|
106 | { |
|
|
107 | Zero (json, 1, JSON); |
|
|
108 | json->max_depth = 512; |
|
|
109 | } |
79 | |
110 | |
80 | ///////////////////////////////////////////////////////////////////////////// |
111 | ///////////////////////////////////////////////////////////////////////////// |
81 | // utility functions |
112 | // utility functions |
82 | |
113 | |
83 | inline void |
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 |
84 | shrink (SV *sv) |
126 | shrink (SV *sv) |
85 | { |
127 | { |
86 | sv_utf8_downgrade (sv, 1); |
128 | sv_utf8_downgrade (sv, 1); |
|
|
129 | |
87 | if (SvLEN (sv) > SvCUR (sv) + 1) |
130 | if (SvLEN (sv) > SvCUR (sv) + 1) |
88 | { |
131 | { |
89 | #ifdef SvPV_shrink_to_cur |
132 | #ifdef SvPV_shrink_to_cur |
90 | SvPV_shrink_to_cur (sv); |
133 | SvPV_shrink_to_cur (sv); |
91 | #elif defined (SvPV_renew) |
134 | #elif defined (SvPV_renew) |
… | |
… | |
97 | // decode an utf-8 character and return it, or (UV)-1 in |
140 | // decode an utf-8 character and return it, or (UV)-1 in |
98 | // case of an error. |
141 | // case of an error. |
99 | // we special-case "safe" characters from U+80 .. U+7FF, |
142 | // we special-case "safe" characters from U+80 .. U+7FF, |
100 | // but use the very good perl function to parse anything else. |
143 | // but use the very good perl function to parse anything else. |
101 | // note that we never call this function for a ascii codepoints |
144 | // note that we never call this function for a ascii codepoints |
102 | inline UV |
145 | INLINE UV |
103 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
146 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
104 | { |
147 | { |
105 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
148 | if (expect_true (len >= 2 |
106 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
149 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
107 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
150 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
108 | { |
151 | { |
109 | *clen = 2; |
152 | *clen = 2; |
110 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
153 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
111 | } |
154 | } |
112 | else |
155 | else |
113 | { |
156 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
114 | *clen = (STRLEN)-1; |
157 | } |
115 | return (UV)-1; |
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) |
|
|
198 | { |
|
|
199 | UV uaccum = 0; |
|
|
200 | int eaccum = 0; |
|
|
201 | |
|
|
202 | for (;;) |
116 | } |
203 | { |
117 | } |
204 | U8 dig = (U8)*s - '0'; |
118 | |
205 | |
|
|
206 | if (expect_false (dig >= 10)) |
|
|
207 | { |
|
|
208 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
209 | { |
|
|
210 | ++s; |
|
|
211 | json_atof_scan1 (s, accum, expo, 1); |
|
|
212 | } |
|
|
213 | else if ((dig | ' ') == 'e' - '0') |
|
|
214 | { |
|
|
215 | int exp2 = 0; |
|
|
216 | int neg = 0; |
|
|
217 | |
|
|
218 | ++s; |
|
|
219 | |
|
|
220 | if (*s == '-') |
|
|
221 | { |
|
|
222 | ++s; |
|
|
223 | neg = 1; |
|
|
224 | } |
|
|
225 | else if (*s == '+') |
|
|
226 | ++s; |
|
|
227 | |
|
|
228 | while ((dig = (U8)*s - '0') < 10) |
|
|
229 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
230 | |
|
|
231 | *expo += neg ? -exp2 : exp2; |
|
|
232 | } |
|
|
233 | |
|
|
234 | break; |
|
|
235 | } |
|
|
236 | |
|
|
237 | ++s; |
|
|
238 | |
|
|
239 | uaccum = uaccum * 10 + dig; |
|
|
240 | ++eaccum; |
|
|
241 | |
|
|
242 | // if we have too many digits, then recurse for more |
|
|
243 | // we actually do this for rather few digits |
|
|
244 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
245 | { |
|
|
246 | if (postdp) *expo -= eaccum; |
|
|
247 | json_atof_scan1 (s, accum, expo, postdp); |
|
|
248 | if (postdp) *expo += eaccum; |
|
|
249 | |
|
|
250 | break; |
|
|
251 | } |
|
|
252 | } |
|
|
253 | |
|
|
254 | // this relies greatly on the quality of the pow () |
|
|
255 | // implementation of the platform, but a good |
|
|
256 | // implementation is hard to beat. |
|
|
257 | if (postdp) *expo -= eaccum; |
|
|
258 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
259 | *expo += eaccum; |
|
|
260 | } |
|
|
261 | |
|
|
262 | static NV |
|
|
263 | json_atof (const char *s) |
|
|
264 | { |
|
|
265 | NV accum = 0.; |
|
|
266 | int expo = 0; |
|
|
267 | int neg = 0; |
|
|
268 | |
|
|
269 | if (*s == '-') |
|
|
270 | { |
|
|
271 | ++s; |
|
|
272 | neg = 1; |
|
|
273 | } |
|
|
274 | |
|
|
275 | json_atof_scan1 (s, &accum, &expo, 0); |
|
|
276 | |
|
|
277 | return neg ? -accum : accum; |
|
|
278 | } |
119 | ///////////////////////////////////////////////////////////////////////////// |
279 | ///////////////////////////////////////////////////////////////////////////// |
120 | // encoder |
280 | // encoder |
121 | |
281 | |
122 | // structure used for encoding JSON |
282 | // structure used for encoding JSON |
123 | typedef struct |
283 | typedef struct |
… | |
… | |
125 | char *cur; // SvPVX (sv) + current output position |
285 | char *cur; // SvPVX (sv) + current output position |
126 | char *end; // SvEND (sv) |
286 | char *end; // SvEND (sv) |
127 | SV *sv; // result scalar |
287 | SV *sv; // result scalar |
128 | JSON json; |
288 | JSON json; |
129 | U32 indent; // indentation level |
289 | U32 indent; // indentation level |
130 | U32 maxdepth; // max. indentation/recursion level |
290 | UV limit; // escape character values >= this value when encoding |
131 | } enc_t; |
291 | } enc_t; |
132 | |
292 | |
133 | inline void |
293 | INLINE void |
134 | need (enc_t *enc, STRLEN len) |
294 | need (enc_t *enc, STRLEN len) |
135 | { |
295 | { |
136 | if (expect_false (enc->cur + len >= enc->end)) |
296 | if (expect_false (enc->cur + len >= enc->end)) |
137 | { |
297 | { |
138 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
298 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
139 | SvGROW (enc->sv, cur + len + 1); |
299 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
140 | enc->cur = SvPVX (enc->sv) + cur; |
300 | enc->cur = SvPVX (enc->sv) + cur; |
141 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
301 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
142 | } |
302 | } |
143 | } |
303 | } |
144 | |
304 | |
145 | inline void |
305 | INLINE void |
146 | encode_ch (enc_t *enc, char ch) |
306 | encode_ch (enc_t *enc, char ch) |
147 | { |
307 | { |
148 | need (enc, 1); |
308 | need (enc, 1); |
149 | *enc->cur++ = ch; |
309 | *enc->cur++ = ch; |
150 | } |
310 | } |
… | |
… | |
204 | { |
364 | { |
205 | uch = ch; |
365 | uch = ch; |
206 | clen = 1; |
366 | clen = 1; |
207 | } |
367 | } |
208 | |
368 | |
209 | if (uch > 0x10FFFFUL) |
369 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
210 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
211 | |
|
|
212 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
|
|
213 | { |
370 | { |
214 | if (uch > 0xFFFFUL) |
371 | if (uch >= 0x10000UL) |
215 | { |
372 | { |
|
|
373 | if (uch >= 0x110000UL) |
|
|
374 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
375 | |
216 | need (enc, len += 11); |
376 | need (enc, len += 11); |
217 | sprintf (enc->cur, "\\u%04x\\u%04x", |
377 | sprintf (enc->cur, "\\u%04x\\u%04x", |
218 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
378 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
219 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
379 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
220 | enc->cur += 12; |
380 | enc->cur += 12; |
221 | } |
381 | } |
222 | else |
382 | else |
223 | { |
383 | { |
224 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
225 | need (enc, len += 5); |
384 | need (enc, len += 5); |
226 | *enc->cur++ = '\\'; |
385 | *enc->cur++ = '\\'; |
227 | *enc->cur++ = 'u'; |
386 | *enc->cur++ = 'u'; |
228 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
387 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
229 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
388 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
230 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
389 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
231 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
390 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
232 | } |
391 | } |
233 | |
392 | |
234 | str += clen; |
393 | str += clen; |
235 | } |
394 | } |
236 | else if (enc->json.flags & F_LATIN1) |
395 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
248 | while (--clen); |
407 | while (--clen); |
249 | } |
408 | } |
250 | else |
409 | else |
251 | { |
410 | { |
252 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
411 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
253 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
412 | enc->cur = encode_utf8 (enc->cur, uch); |
254 | ++str; |
413 | ++str; |
255 | } |
414 | } |
256 | } |
415 | } |
257 | } |
416 | } |
258 | } |
417 | } |
259 | |
418 | |
260 | --len; |
419 | --len; |
261 | } |
420 | } |
262 | } |
421 | } |
263 | |
422 | |
264 | inline void |
423 | INLINE void |
265 | encode_indent (enc_t *enc) |
424 | encode_indent (enc_t *enc) |
266 | { |
425 | { |
267 | if (enc->json.flags & F_INDENT) |
426 | if (enc->json.flags & F_INDENT) |
268 | { |
427 | { |
269 | int spaces = enc->indent * INDENT_STEP; |
428 | int spaces = enc->indent * INDENT_STEP; |
… | |
… | |
272 | memset (enc->cur, ' ', spaces); |
431 | memset (enc->cur, ' ', spaces); |
273 | enc->cur += spaces; |
432 | enc->cur += spaces; |
274 | } |
433 | } |
275 | } |
434 | } |
276 | |
435 | |
277 | inline void |
436 | INLINE void |
278 | encode_space (enc_t *enc) |
437 | encode_space (enc_t *enc) |
279 | { |
438 | { |
280 | need (enc, 1); |
439 | need (enc, 1); |
281 | encode_ch (enc, ' '); |
440 | encode_ch (enc, ' '); |
282 | } |
441 | } |
283 | |
442 | |
284 | inline void |
443 | INLINE void |
285 | encode_nl (enc_t *enc) |
444 | encode_nl (enc_t *enc) |
286 | { |
445 | { |
287 | if (enc->json.flags & F_INDENT) |
446 | if (enc->json.flags & F_INDENT) |
288 | { |
447 | { |
289 | need (enc, 1); |
448 | need (enc, 1); |
290 | encode_ch (enc, '\n'); |
449 | encode_ch (enc, '\n'); |
291 | } |
450 | } |
292 | } |
451 | } |
293 | |
452 | |
294 | inline void |
453 | INLINE void |
295 | encode_comma (enc_t *enc) |
454 | encode_comma (enc_t *enc) |
296 | { |
455 | { |
297 | encode_ch (enc, ','); |
456 | encode_ch (enc, ','); |
298 | |
457 | |
299 | if (enc->json.flags & F_INDENT) |
458 | if (enc->json.flags & F_INDENT) |
… | |
… | |
307 | static void |
466 | static void |
308 | encode_av (enc_t *enc, AV *av) |
467 | encode_av (enc_t *enc, AV *av) |
309 | { |
468 | { |
310 | int i, len = av_len (av); |
469 | int i, len = av_len (av); |
311 | |
470 | |
312 | if (enc->indent >= enc->maxdepth) |
471 | if (enc->indent >= enc->json.max_depth) |
313 | croak ("data structure too deep (hit recursion limit)"); |
472 | croak (ERR_NESTING_EXCEEDED); |
314 | |
473 | |
315 | encode_ch (enc, '['); encode_nl (enc); |
474 | encode_ch (enc, '['); |
316 | ++enc->indent; |
475 | |
|
|
476 | if (len >= 0) |
|
|
477 | { |
|
|
478 | encode_nl (enc); ++enc->indent; |
317 | |
479 | |
318 | for (i = 0; i <= len; ++i) |
480 | for (i = 0; i <= len; ++i) |
319 | { |
481 | { |
320 | SV **svp = av_fetch (av, i, 0); |
482 | SV **svp = av_fetch (av, i, 0); |
321 | |
483 | |
322 | encode_indent (enc); |
484 | encode_indent (enc); |
323 | |
485 | |
324 | if (svp) |
486 | if (svp) |
325 | encode_sv (enc, *svp); |
487 | encode_sv (enc, *svp); |
326 | else |
488 | else |
327 | encode_str (enc, "null", 4, 0); |
489 | encode_str (enc, "null", 4, 0); |
328 | |
490 | |
329 | if (i < len) |
491 | if (i < len) |
330 | encode_comma (enc); |
492 | encode_comma (enc); |
331 | } |
493 | } |
332 | |
494 | |
|
|
495 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
496 | } |
|
|
497 | |
333 | encode_nl (enc); |
498 | encode_ch (enc, ']'); |
334 | |
|
|
335 | --enc->indent; |
|
|
336 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
337 | } |
499 | } |
338 | |
500 | |
339 | static void |
501 | static void |
340 | encode_hk (enc_t *enc, HE *he) |
502 | encode_hk (enc_t *enc, HE *he) |
341 | { |
503 | { |
… | |
… | |
389 | |
551 | |
390 | static void |
552 | static void |
391 | encode_hv (enc_t *enc, HV *hv) |
553 | encode_hv (enc_t *enc, HV *hv) |
392 | { |
554 | { |
393 | HE *he; |
555 | HE *he; |
394 | int count; |
|
|
395 | |
556 | |
396 | if (enc->indent >= enc->maxdepth) |
557 | if (enc->indent >= enc->json.max_depth) |
397 | croak ("data structure too deep (hit recursion limit)"); |
558 | croak (ERR_NESTING_EXCEEDED); |
398 | |
559 | |
399 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
560 | encode_ch (enc, '{'); |
400 | |
561 | |
401 | // for canonical output we have to sort by keys first |
562 | // for canonical output we have to sort by keys first |
402 | // actually, this is mostly due to the stupid so-called |
563 | // actually, this is mostly due to the stupid so-called |
403 | // security workaround added somewhere in 5.8.x. |
564 | // security workaround added somewhere in 5.8.x |
404 | // that randomises hash orderings |
565 | // that randomises hash orderings |
405 | if (enc->json.flags & F_CANONICAL) |
566 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
406 | { |
567 | { |
407 | int count = hv_iterinit (hv); |
568 | int count = hv_iterinit (hv); |
408 | |
569 | |
409 | if (SvMAGICAL (hv)) |
570 | if (SvMAGICAL (hv)) |
410 | { |
571 | { |
… | |
… | |
457 | |
618 | |
458 | FREETMPS; |
619 | FREETMPS; |
459 | LEAVE; |
620 | LEAVE; |
460 | } |
621 | } |
461 | |
622 | |
|
|
623 | encode_nl (enc); ++enc->indent; |
|
|
624 | |
462 | while (count--) |
625 | while (count--) |
463 | { |
626 | { |
464 | encode_indent (enc); |
627 | encode_indent (enc); |
465 | he = hes [count]; |
628 | he = hes [count]; |
466 | encode_hk (enc, he); |
629 | encode_hk (enc, he); |
467 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
630 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
468 | |
631 | |
469 | if (count) |
632 | if (count) |
470 | encode_comma (enc); |
633 | encode_comma (enc); |
471 | } |
634 | } |
|
|
635 | |
|
|
636 | encode_nl (enc); --enc->indent; encode_indent (enc); |
472 | } |
637 | } |
473 | } |
638 | } |
474 | else |
639 | else |
475 | { |
640 | { |
476 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
641 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
477 | if ((he = hv_iternext (hv))) |
642 | if ((he = hv_iternext (hv))) |
|
|
643 | { |
|
|
644 | encode_nl (enc); ++enc->indent; |
|
|
645 | |
478 | for (;;) |
646 | for (;;) |
479 | { |
647 | { |
480 | encode_indent (enc); |
648 | encode_indent (enc); |
481 | encode_hk (enc, he); |
649 | encode_hk (enc, he); |
482 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
650 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
483 | |
651 | |
484 | if (!(he = hv_iternext (hv))) |
652 | if (!(he = hv_iternext (hv))) |
485 | break; |
653 | break; |
486 | |
654 | |
487 | encode_comma (enc); |
655 | encode_comma (enc); |
488 | } |
656 | } |
489 | } |
|
|
490 | |
657 | |
|
|
658 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
659 | } |
|
|
660 | } |
|
|
661 | |
491 | encode_nl (enc); |
662 | encode_ch (enc, '}'); |
492 | |
|
|
493 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
|
|
494 | } |
663 | } |
495 | |
664 | |
496 | // encode objects, arrays and special \0=false and \1=true values. |
665 | // encode objects, arrays and special \0=false and \1=true values. |
497 | static void |
666 | static void |
498 | encode_rv (enc_t *enc, SV *sv) |
667 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
575 | |
744 | |
576 | if (len == 1 && *pv == '1') |
745 | if (len == 1 && *pv == '1') |
577 | encode_str (enc, "true", 4, 0); |
746 | encode_str (enc, "true", 4, 0); |
578 | else if (len == 1 && *pv == '0') |
747 | else if (len == 1 && *pv == '0') |
579 | encode_str (enc, "false", 5, 0); |
748 | encode_str (enc, "false", 5, 0); |
|
|
749 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
750 | encode_str (enc, "null", 4, 0); |
580 | else |
751 | else |
581 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
752 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
582 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
753 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
583 | } |
754 | } |
|
|
755 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
756 | encode_str (enc, "null", 4, 0); |
584 | else |
757 | else |
585 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
758 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
586 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
759 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
587 | } |
760 | } |
588 | |
761 | |
… | |
… | |
606 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
779 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
607 | enc->cur += strlen (enc->cur); |
780 | enc->cur += strlen (enc->cur); |
608 | } |
781 | } |
609 | else if (SvIOKp (sv)) |
782 | else if (SvIOKp (sv)) |
610 | { |
783 | { |
611 | // we assume we can always read an IV as a UV |
784 | // we assume we can always read an IV as a UV and vice versa |
612 | if (SvUV (sv) & ~(UV)0x7fff) |
785 | // we assume two's complement |
613 | { |
786 | // we assume no aliasing issues in the union |
614 | // large integer, use the (rather slow) snprintf way. |
787 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
615 | need (enc, sizeof (UV) * 3); |
788 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
616 | enc->cur += |
|
|
617 | SvIsUV(sv) |
|
|
618 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
|
|
619 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
620 | } |
|
|
621 | else |
|
|
622 | { |
789 | { |
623 | // optimise the "small number case" |
790 | // optimise the "small number case" |
624 | // code will likely be branchless and use only a single multiplication |
791 | // code will likely be branchless and use only a single multiplication |
|
|
792 | // works for numbers up to 59074 |
625 | I32 i = SvIV (sv); |
793 | I32 i = SvIVX (sv); |
626 | U32 u; |
794 | U32 u; |
627 | char digit, nz = 0; |
795 | char digit, nz = 0; |
628 | |
796 | |
629 | need (enc, 6); |
797 | need (enc, 6); |
630 | |
798 | |
… | |
… | |
636 | |
804 | |
637 | // now output digit by digit, each time masking out the integer part |
805 | // now output digit by digit, each time masking out the integer part |
638 | // and multiplying by 5 while moving the decimal point one to the right, |
806 | // and multiplying by 5 while moving the decimal point one to the right, |
639 | // resulting in a net multiplication by 10. |
807 | // resulting in a net multiplication by 10. |
640 | // we always write the digit to memory but conditionally increment |
808 | // we always write the digit to memory but conditionally increment |
641 | // the pointer, to ease the usage of conditional move instructions. |
809 | // the pointer, to enable the use of conditional move instructions. |
642 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
810 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
643 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
811 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
644 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
812 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
645 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
813 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
646 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
814 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
647 | } |
815 | } |
|
|
816 | else |
|
|
817 | { |
|
|
818 | // large integer, use the (rather slow) snprintf way. |
|
|
819 | need (enc, IVUV_MAXCHARS); |
|
|
820 | enc->cur += |
|
|
821 | SvIsUV(sv) |
|
|
822 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
823 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
824 | } |
648 | } |
825 | } |
649 | else if (SvROK (sv)) |
826 | else if (SvROK (sv)) |
650 | encode_rv (enc, SvRV (sv)); |
827 | encode_rv (enc, SvRV (sv)); |
651 | else if (!SvOK (sv)) |
828 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
652 | encode_str (enc, "null", 4, 0); |
829 | encode_str (enc, "null", 4, 0); |
653 | else |
830 | else |
654 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
831 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
655 | SvPV_nolen (sv), SvFLAGS (sv)); |
832 | SvPV_nolen (sv), SvFLAGS (sv)); |
656 | } |
833 | } |
… | |
… | |
666 | enc.json = *json; |
843 | enc.json = *json; |
667 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
844 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
668 | enc.cur = SvPVX (enc.sv); |
845 | enc.cur = SvPVX (enc.sv); |
669 | enc.end = SvEND (enc.sv); |
846 | enc.end = SvEND (enc.sv); |
670 | enc.indent = 0; |
847 | enc.indent = 0; |
671 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
848 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
849 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
850 | : 0x110000UL; |
672 | |
851 | |
673 | SvPOK_only (enc.sv); |
852 | SvPOK_only (enc.sv); |
674 | encode_sv (&enc, scalar); |
853 | encode_sv (&enc, scalar); |
|
|
854 | encode_nl (&enc); |
675 | |
855 | |
676 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
856 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
677 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
857 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
678 | |
858 | |
679 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
859 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
697 | JSON json; |
877 | JSON json; |
698 | U32 depth; // recursion depth |
878 | U32 depth; // recursion depth |
699 | U32 maxdepth; // recursion depth limit |
879 | U32 maxdepth; // recursion depth limit |
700 | } dec_t; |
880 | } dec_t; |
701 | |
881 | |
702 | inline void |
882 | INLINE void |
|
|
883 | decode_comment (dec_t *dec) |
|
|
884 | { |
|
|
885 | // only '#'-style comments allowed a.t.m. |
|
|
886 | |
|
|
887 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
888 | ++dec->cur; |
|
|
889 | } |
|
|
890 | |
|
|
891 | INLINE void |
703 | decode_ws (dec_t *dec) |
892 | decode_ws (dec_t *dec) |
704 | { |
893 | { |
705 | for (;;) |
894 | for (;;) |
706 | { |
895 | { |
707 | char ch = *dec->cur; |
896 | char ch = *dec->cur; |
708 | |
897 | |
709 | if (ch > 0x20 |
898 | if (ch > 0x20) |
|
|
899 | { |
|
|
900 | if (expect_false (ch == '#')) |
|
|
901 | { |
|
|
902 | if (dec->json.flags & F_RELAXED) |
|
|
903 | decode_comment (dec); |
|
|
904 | else |
|
|
905 | break; |
|
|
906 | } |
|
|
907 | else |
|
|
908 | break; |
|
|
909 | } |
710 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
910 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
711 | break; |
911 | break; // parse error, but let higher level handle it, gives better error messages |
712 | |
912 | |
713 | ++dec->cur; |
913 | ++dec->cur; |
714 | } |
914 | } |
715 | } |
915 | } |
716 | |
916 | |
… | |
… | |
720 | if (*dec->cur != ch) \ |
920 | if (*dec->cur != ch) \ |
721 | ERR (# ch " expected"); \ |
921 | ERR (# ch " expected"); \ |
722 | ++dec->cur; \ |
922 | ++dec->cur; \ |
723 | SE |
923 | SE |
724 | |
924 | |
725 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
925 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
726 | #define DEC_DEC_DEPTH --dec->depth |
926 | #define DEC_DEC_DEPTH --dec->depth |
727 | |
927 | |
728 | static SV *decode_sv (dec_t *dec); |
928 | static SV *decode_sv (dec_t *dec); |
729 | |
929 | |
730 | static signed char decode_hexdigit[256]; |
930 | static signed char decode_hexdigit[256]; |
… | |
… | |
822 | |
1022 | |
823 | if (hi >= 0x80) |
1023 | if (hi >= 0x80) |
824 | { |
1024 | { |
825 | utf8 = 1; |
1025 | utf8 = 1; |
826 | |
1026 | |
827 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
1027 | cur = encode_utf8 (cur, hi); |
828 | } |
1028 | } |
829 | else |
1029 | else |
830 | *cur++ = hi; |
1030 | *cur++ = hi; |
831 | } |
1031 | } |
832 | break; |
1032 | break; |
… | |
… | |
834 | default: |
1034 | default: |
835 | --dec_cur; |
1035 | --dec_cur; |
836 | ERR ("illegal backslash escape sequence in string"); |
1036 | ERR ("illegal backslash escape sequence in string"); |
837 | } |
1037 | } |
838 | } |
1038 | } |
839 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
1039 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
840 | *cur++ = ch; |
1040 | *cur++ = ch; |
841 | else if (ch >= 0x80) |
1041 | else if (ch >= 0x80) |
842 | { |
1042 | { |
843 | STRLEN clen; |
1043 | STRLEN clen; |
844 | UV uch; |
|
|
845 | |
1044 | |
846 | --dec_cur; |
1045 | --dec_cur; |
847 | |
1046 | |
848 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1047 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
849 | if (clen == (STRLEN)-1) |
1048 | if (clen == (STRLEN)-1) |
850 | ERR ("malformed UTF-8 character in JSON string"); |
1049 | ERR ("malformed UTF-8 character in JSON string"); |
851 | |
1050 | |
852 | do |
1051 | do |
853 | *cur++ = *dec_cur++; |
1052 | *cur++ = *dec_cur++; |
… | |
… | |
870 | { |
1069 | { |
871 | STRLEN len = cur - buf; |
1070 | STRLEN len = cur - buf; |
872 | |
1071 | |
873 | if (sv) |
1072 | if (sv) |
874 | { |
1073 | { |
875 | SvGROW (sv, SvCUR (sv) + len + 1); |
1074 | STRLEN cur = SvCUR (sv); |
|
|
1075 | |
|
|
1076 | if (SvLEN (sv) <= cur + len) |
|
|
1077 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1078 | |
876 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1079 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
877 | SvCUR_set (sv, SvCUR (sv) + len); |
1080 | SvCUR_set (sv, SvCUR (sv) + len); |
878 | } |
1081 | } |
879 | else |
1082 | else |
880 | sv = newSVpvn (buf, len); |
1083 | sv = newSVpvn (buf, len); |
… | |
… | |
967 | |
1170 | |
968 | if (!is_nv) |
1171 | if (!is_nv) |
969 | { |
1172 | { |
970 | int len = dec->cur - start; |
1173 | int len = dec->cur - start; |
971 | |
1174 | |
972 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1175 | // special case the rather common 1..5-digit-int case |
973 | if (*start == '-') |
1176 | if (*start == '-') |
974 | switch (len) |
1177 | switch (len) |
975 | { |
1178 | { |
976 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1179 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
977 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1180 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
978 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1181 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
979 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1182 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1183 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
980 | } |
1184 | } |
981 | else |
1185 | else |
982 | switch (len) |
1186 | switch (len) |
983 | { |
1187 | { |
984 | case 1: return newSViv ( start [0] - '0' * 1); |
1188 | case 1: return newSViv ( start [0] - '0' * 1); |
985 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1189 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
986 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1190 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
987 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1191 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1192 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
988 | } |
1193 | } |
989 | |
1194 | |
990 | { |
1195 | { |
991 | UV uv; |
1196 | UV uv; |
992 | int numtype = grok_number (start, len, &uv); |
1197 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1001 | } |
1206 | } |
1002 | |
1207 | |
1003 | len -= *start == '-' ? 1 : 0; |
1208 | len -= *start == '-' ? 1 : 0; |
1004 | |
1209 | |
1005 | // does not fit into IV or UV, try NV |
1210 | // does not fit into IV or UV, try NV |
1006 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
1211 | if (len <= NV_DIG) |
1007 | #if defined (LDBL_DIG) |
|
|
1008 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
1009 | #endif |
|
|
1010 | ) |
|
|
1011 | // fits into NV without loss of precision |
1212 | // fits into NV without loss of precision |
1012 | return newSVnv (Atof (start)); |
1213 | return newSVnv (json_atof (start)); |
1013 | |
1214 | |
1014 | // everything else fails, convert it to a string |
1215 | // everything else fails, convert it to a string |
1015 | return newSVpvn (start, dec->cur - start); |
1216 | return newSVpvn (start, dec->cur - start); |
1016 | } |
1217 | } |
1017 | |
1218 | |
1018 | // loss of precision here |
1219 | // loss of precision here |
1019 | return newSVnv (Atof (start)); |
1220 | return newSVnv (json_atof (start)); |
1020 | |
1221 | |
1021 | fail: |
1222 | fail: |
1022 | return 0; |
1223 | return 0; |
1023 | } |
1224 | } |
1024 | |
1225 | |
… | |
… | |
1053 | |
1254 | |
1054 | if (*dec->cur != ',') |
1255 | if (*dec->cur != ',') |
1055 | ERR (", or ] expected while parsing array"); |
1256 | ERR (", or ] expected while parsing array"); |
1056 | |
1257 | |
1057 | ++dec->cur; |
1258 | ++dec->cur; |
|
|
1259 | |
|
|
1260 | decode_ws (dec); |
|
|
1261 | |
|
|
1262 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1263 | { |
|
|
1264 | ++dec->cur; |
|
|
1265 | break; |
|
|
1266 | } |
1058 | } |
1267 | } |
1059 | |
1268 | |
1060 | DEC_DEC_DEPTH; |
1269 | DEC_DEC_DEPTH; |
1061 | return newRV_noinc ((SV *)av); |
1270 | return newRV_noinc ((SV *)av); |
1062 | |
1271 | |
… | |
… | |
1078 | if (*dec->cur == '}') |
1287 | if (*dec->cur == '}') |
1079 | ++dec->cur; |
1288 | ++dec->cur; |
1080 | else |
1289 | else |
1081 | for (;;) |
1290 | for (;;) |
1082 | { |
1291 | { |
1083 | decode_ws (dec); EXPECT_CH ('"'); |
1292 | EXPECT_CH ('"'); |
1084 | |
1293 | |
1085 | // heuristic: assume that |
1294 | // heuristic: assume that |
1086 | // a) decode_str + hv_store_ent are abysmally slow. |
1295 | // a) decode_str + hv_store_ent are abysmally slow. |
1087 | // b) most hash keys are short, simple ascii text. |
1296 | // b) most hash keys are short, simple ascii text. |
1088 | // => try to "fast-match" such strings to avoid |
1297 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1092 | char *p = dec->cur; |
1301 | char *p = dec->cur; |
1093 | char *e = p + 24; // only try up to 24 bytes |
1302 | char *e = p + 24; // only try up to 24 bytes |
1094 | |
1303 | |
1095 | for (;;) |
1304 | for (;;) |
1096 | { |
1305 | { |
1097 | // the >= 0x80 is true on most architectures |
1306 | // the >= 0x80 is false on most architectures |
1098 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1307 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1099 | { |
1308 | { |
1100 | // slow path, back up and use decode_str |
1309 | // slow path, back up and use decode_str |
1101 | SV *key = decode_str (dec); |
1310 | SV *key = decode_str (dec); |
1102 | if (!key) |
1311 | if (!key) |
1103 | goto fail; |
1312 | goto fail; |
1104 | |
1313 | |
1105 | decode_ws (dec); EXPECT_CH (':'); |
1314 | decode_ws (dec); EXPECT_CH (':'); |
1106 | |
1315 | |
|
|
1316 | decode_ws (dec); |
1107 | value = decode_sv (dec); |
1317 | value = decode_sv (dec); |
1108 | if (!value) |
1318 | if (!value) |
1109 | { |
1319 | { |
1110 | SvREFCNT_dec (key); |
1320 | SvREFCNT_dec (key); |
1111 | goto fail; |
1321 | goto fail; |
… | |
… | |
1123 | int len = p - key; |
1333 | int len = p - key; |
1124 | dec->cur = p + 1; |
1334 | dec->cur = p + 1; |
1125 | |
1335 | |
1126 | decode_ws (dec); EXPECT_CH (':'); |
1336 | decode_ws (dec); EXPECT_CH (':'); |
1127 | |
1337 | |
|
|
1338 | decode_ws (dec); |
1128 | value = decode_sv (dec); |
1339 | value = decode_sv (dec); |
1129 | if (!value) |
1340 | if (!value) |
1130 | goto fail; |
1341 | goto fail; |
1131 | |
1342 | |
1132 | hv_store (hv, key, len, value, 0); |
1343 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1148 | |
1359 | |
1149 | if (*dec->cur != ',') |
1360 | if (*dec->cur != ',') |
1150 | ERR (", or } expected while parsing object/hash"); |
1361 | ERR (", or } expected while parsing object/hash"); |
1151 | |
1362 | |
1152 | ++dec->cur; |
1363 | ++dec->cur; |
|
|
1364 | |
|
|
1365 | decode_ws (dec); |
|
|
1366 | |
|
|
1367 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1368 | { |
|
|
1369 | ++dec->cur; |
|
|
1370 | break; |
|
|
1371 | } |
1153 | } |
1372 | } |
1154 | |
1373 | |
1155 | DEC_DEC_DEPTH; |
1374 | DEC_DEC_DEPTH; |
1156 | sv = newRV_noinc ((SV *)hv); |
1375 | sv = newRV_noinc ((SV *)hv); |
1157 | |
1376 | |
… | |
… | |
1222 | } |
1441 | } |
1223 | |
1442 | |
1224 | static SV * |
1443 | static SV * |
1225 | decode_sv (dec_t *dec) |
1444 | decode_sv (dec_t *dec) |
1226 | { |
1445 | { |
1227 | decode_ws (dec); |
|
|
1228 | |
|
|
1229 | // the beauty of JSON: you need exactly one character lookahead |
1446 | // the beauty of JSON: you need exactly one character lookahead |
1230 | // to parse anything. |
1447 | // to parse everything. |
1231 | switch (*dec->cur) |
1448 | switch (*dec->cur) |
1232 | { |
1449 | { |
1233 | case '"': ++dec->cur; return decode_str (dec); |
1450 | case '"': ++dec->cur; return decode_str (dec); |
1234 | case '[': ++dec->cur; return decode_av (dec); |
1451 | case '[': ++dec->cur; return decode_av (dec); |
1235 | case '{': ++dec->cur; return decode_hv (dec); |
1452 | case '{': ++dec->cur; return decode_hv (dec); |
1236 | |
1453 | |
1237 | case '-': |
1454 | case '-': |
1238 | case '0': case '1': case '2': case '3': case '4': |
1455 | case '0': case '1': case '2': case '3': case '4': |
1239 | case '5': case '6': case '7': case '8': case '9': |
1456 | case '5': case '6': case '7': case '8': case '9': |
1240 | return decode_num (dec); |
1457 | return decode_num (dec); |
… | |
… | |
1242 | case 't': |
1459 | case 't': |
1243 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1460 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1244 | { |
1461 | { |
1245 | dec->cur += 4; |
1462 | dec->cur += 4; |
1246 | #if JSON_SLOW |
1463 | #if JSON_SLOW |
1247 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
1464 | json_true = get_bool ("JSON::XS::true"); |
1248 | #endif |
1465 | #endif |
1249 | return SvREFCNT_inc (json_true); |
1466 | return newSVsv (json_true); |
1250 | } |
1467 | } |
1251 | else |
1468 | else |
1252 | ERR ("'true' expected"); |
1469 | ERR ("'true' expected"); |
1253 | |
1470 | |
1254 | break; |
1471 | break; |
… | |
… | |
1256 | case 'f': |
1473 | case 'f': |
1257 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1474 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1258 | { |
1475 | { |
1259 | dec->cur += 5; |
1476 | dec->cur += 5; |
1260 | #if JSON_SLOW |
1477 | #if JSON_SLOW |
1261 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1478 | json_false = get_bool ("JSON::XS::false"); |
1262 | #endif |
1479 | #endif |
1263 | return SvREFCNT_inc (json_false); |
1480 | return newSVsv (json_false); |
1264 | } |
1481 | } |
1265 | else |
1482 | else |
1266 | ERR ("'false' expected"); |
1483 | ERR ("'false' expected"); |
1267 | |
1484 | |
1268 | break; |
1485 | break; |
… | |
… | |
1286 | fail: |
1503 | fail: |
1287 | return 0; |
1504 | return 0; |
1288 | } |
1505 | } |
1289 | |
1506 | |
1290 | static SV * |
1507 | static SV * |
1291 | decode_json (SV *string, JSON *json, UV *offset_return) |
1508 | decode_json (SV *string, JSON *json, char **offset_return) |
1292 | { |
1509 | { |
1293 | dec_t dec; |
1510 | dec_t dec; |
1294 | UV offset; |
|
|
1295 | SV *sv; |
1511 | SV *sv; |
1296 | |
1512 | |
|
|
1513 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1514 | * will perl ignore the magic in subsequent accesses |
|
|
1515 | */ |
1297 | SvGETMAGIC (string); |
1516 | /*SvGETMAGIC (string);*/ |
|
|
1517 | if (SvMAGICAL (string)) |
|
|
1518 | string = sv_2mortal (newSVsv (string)); |
|
|
1519 | |
1298 | SvUPGRADE (string, SVt_PV); |
1520 | SvUPGRADE (string, SVt_PV); |
1299 | |
1521 | |
1300 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1522 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1523 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1524 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1525 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1526 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1527 | * and hope for the best. |
|
|
1528 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1529 | * assertion business is seriously broken, try yet another workaround |
|
|
1530 | * for the broken -DDEBUGGING. |
|
|
1531 | */ |
|
|
1532 | { |
|
|
1533 | #ifdef DEBUGGING |
|
|
1534 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1535 | #else |
|
|
1536 | STRLEN offset = SvCUR (string); |
|
|
1537 | #endif |
|
|
1538 | |
|
|
1539 | if (offset > json->max_size && json->max_size) |
1301 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1540 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1302 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1541 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1542 | } |
1303 | |
1543 | |
1304 | if (json->flags & F_UTF8) |
1544 | if (json->flags & F_UTF8) |
1305 | sv_utf8_downgrade (string, 0); |
1545 | sv_utf8_downgrade (string, 0); |
1306 | else |
1546 | else |
1307 | sv_utf8_upgrade (string); |
1547 | sv_utf8_upgrade (string); |
1308 | |
1548 | |
1309 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1549 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1310 | |
1550 | |
1311 | dec.json = *json; |
1551 | dec.json = *json; |
1312 | dec.cur = SvPVX (string); |
1552 | dec.cur = SvPVX (string); |
1313 | dec.end = SvEND (string); |
1553 | dec.end = SvEND (string); |
1314 | dec.err = 0; |
1554 | dec.err = 0; |
1315 | dec.depth = 0; |
1555 | dec.depth = 0; |
1316 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1317 | |
1556 | |
1318 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1557 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1319 | dec.json.flags |= F_HOOK; |
1558 | dec.json.flags |= F_HOOK; |
1320 | |
1559 | |
1321 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1560 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1561 | |
|
|
1562 | decode_ws (&dec); |
1322 | sv = decode_sv (&dec); |
1563 | sv = decode_sv (&dec); |
|
|
1564 | |
|
|
1565 | if (offset_return) |
|
|
1566 | *offset_return = dec.cur; |
1323 | |
1567 | |
1324 | if (!(offset_return || !sv)) |
1568 | if (!(offset_return || !sv)) |
1325 | { |
1569 | { |
1326 | // check for trailing garbage |
1570 | // check for trailing garbage |
1327 | decode_ws (&dec); |
1571 | decode_ws (&dec); |
… | |
… | |
1330 | { |
1574 | { |
1331 | dec.err = "garbage after JSON object"; |
1575 | dec.err = "garbage after JSON object"; |
1332 | SvREFCNT_dec (sv); |
1576 | SvREFCNT_dec (sv); |
1333 | sv = 0; |
1577 | sv = 0; |
1334 | } |
1578 | } |
1335 | } |
|
|
1336 | |
|
|
1337 | if (offset_return || !sv) |
|
|
1338 | { |
|
|
1339 | offset = dec.json.flags & F_UTF8 |
|
|
1340 | ? dec.cur - SvPVX (string) |
|
|
1341 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1342 | |
|
|
1343 | if (offset_return) |
|
|
1344 | *offset_return = offset; |
|
|
1345 | } |
1579 | } |
1346 | |
1580 | |
1347 | if (!sv) |
1581 | if (!sv) |
1348 | { |
1582 | { |
1349 | SV *uni = sv_newmortal (); |
1583 | SV *uni = sv_newmortal (); |
… | |
… | |
1355 | SAVEVPTR (PL_curcop); |
1589 | SAVEVPTR (PL_curcop); |
1356 | PL_curcop = &cop; |
1590 | PL_curcop = &cop; |
1357 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1591 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1358 | LEAVE; |
1592 | LEAVE; |
1359 | |
1593 | |
1360 | croak ("%s, at character offset %d [\"%s\"]", |
1594 | croak ("%s, at character offset %d (before \"%s\")", |
1361 | dec.err, |
1595 | dec.err, |
1362 | (int)offset, |
1596 | ptr_to_index (string, dec.cur), |
1363 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1597 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1364 | } |
1598 | } |
1365 | |
1599 | |
1366 | sv = sv_2mortal (sv); |
1600 | sv = sv_2mortal (sv); |
1367 | |
1601 | |
1368 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1602 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1369 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1603 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1370 | |
1604 | |
1371 | return sv; |
1605 | return sv; |
|
|
1606 | } |
|
|
1607 | |
|
|
1608 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1609 | // incremental parser |
|
|
1610 | |
|
|
1611 | static void |
|
|
1612 | incr_parse (JSON *self) |
|
|
1613 | { |
|
|
1614 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1615 | |
|
|
1616 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1617 | // but this would make it slower, so... |
|
|
1618 | |
|
|
1619 | for (;;) |
|
|
1620 | { |
|
|
1621 | //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 |
|
|
1622 | switch (self->incr_mode) |
|
|
1623 | { |
|
|
1624 | // only used for initial whitespace skipping |
|
|
1625 | case INCR_M_WS: |
|
|
1626 | for (;;) |
|
|
1627 | { |
|
|
1628 | if (*p > 0x20) |
|
|
1629 | { |
|
|
1630 | if (*p == '#') |
|
|
1631 | { |
|
|
1632 | self->incr_mode = INCR_M_C0; |
|
|
1633 | goto incr_m_c; |
|
|
1634 | } |
|
|
1635 | else |
|
|
1636 | { |
|
|
1637 | self->incr_mode = INCR_M_JSON; |
|
|
1638 | goto incr_m_json; |
|
|
1639 | } |
|
|
1640 | } |
|
|
1641 | else if (!*p) |
|
|
1642 | goto interrupt; |
|
|
1643 | |
|
|
1644 | ++p; |
|
|
1645 | } |
|
|
1646 | |
|
|
1647 | // skip a single char inside a string (for \\-processing) |
|
|
1648 | case INCR_M_BS: |
|
|
1649 | if (!*p) |
|
|
1650 | goto interrupt; |
|
|
1651 | |
|
|
1652 | ++p; |
|
|
1653 | self->incr_mode = INCR_M_STR; |
|
|
1654 | goto incr_m_str; |
|
|
1655 | |
|
|
1656 | // inside #-style comments |
|
|
1657 | case INCR_M_C0: |
|
|
1658 | case INCR_M_C1: |
|
|
1659 | incr_m_c: |
|
|
1660 | for (;;) |
|
|
1661 | { |
|
|
1662 | if (*p == '\n') |
|
|
1663 | { |
|
|
1664 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1665 | break; |
|
|
1666 | } |
|
|
1667 | else if (!*p) |
|
|
1668 | goto interrupt; |
|
|
1669 | |
|
|
1670 | ++p; |
|
|
1671 | } |
|
|
1672 | |
|
|
1673 | break; |
|
|
1674 | |
|
|
1675 | // inside a string |
|
|
1676 | case INCR_M_STR: |
|
|
1677 | incr_m_str: |
|
|
1678 | for (;;) |
|
|
1679 | { |
|
|
1680 | if (*p == '"') |
|
|
1681 | { |
|
|
1682 | ++p; |
|
|
1683 | self->incr_mode = INCR_M_JSON; |
|
|
1684 | |
|
|
1685 | if (!self->incr_nest) |
|
|
1686 | goto interrupt; |
|
|
1687 | |
|
|
1688 | goto incr_m_json; |
|
|
1689 | } |
|
|
1690 | else if (*p == '\\') |
|
|
1691 | { |
|
|
1692 | ++p; // "virtually" consumes character after \ |
|
|
1693 | |
|
|
1694 | if (!*p) // if at end of string we have to switch modes |
|
|
1695 | { |
|
|
1696 | self->incr_mode = INCR_M_BS; |
|
|
1697 | goto interrupt; |
|
|
1698 | } |
|
|
1699 | } |
|
|
1700 | else if (!*p) |
|
|
1701 | goto interrupt; |
|
|
1702 | |
|
|
1703 | ++p; |
|
|
1704 | } |
|
|
1705 | |
|
|
1706 | // after initial ws, outside string |
|
|
1707 | case INCR_M_JSON: |
|
|
1708 | incr_m_json: |
|
|
1709 | for (;;) |
|
|
1710 | { |
|
|
1711 | switch (*p++) |
|
|
1712 | { |
|
|
1713 | case 0: |
|
|
1714 | --p; |
|
|
1715 | goto interrupt; |
|
|
1716 | |
|
|
1717 | case 0x09: |
|
|
1718 | case 0x0a: |
|
|
1719 | case 0x0d: |
|
|
1720 | case 0x20: |
|
|
1721 | if (!self->incr_nest) |
|
|
1722 | { |
|
|
1723 | --p; // do not eat the whitespace, let the next round do it |
|
|
1724 | goto interrupt; |
|
|
1725 | } |
|
|
1726 | break; |
|
|
1727 | |
|
|
1728 | case '"': |
|
|
1729 | self->incr_mode = INCR_M_STR; |
|
|
1730 | goto incr_m_str; |
|
|
1731 | |
|
|
1732 | case '[': |
|
|
1733 | case '{': |
|
|
1734 | if (++self->incr_nest > self->max_depth) |
|
|
1735 | croak (ERR_NESTING_EXCEEDED); |
|
|
1736 | break; |
|
|
1737 | |
|
|
1738 | case ']': |
|
|
1739 | case '}': |
|
|
1740 | if (--self->incr_nest <= 0) |
|
|
1741 | goto interrupt; |
|
|
1742 | break; |
|
|
1743 | |
|
|
1744 | case '#': |
|
|
1745 | self->incr_mode = INCR_M_C1; |
|
|
1746 | goto incr_m_c; |
|
|
1747 | } |
|
|
1748 | } |
|
|
1749 | } |
|
|
1750 | |
|
|
1751 | modechange: |
|
|
1752 | ; |
|
|
1753 | } |
|
|
1754 | |
|
|
1755 | interrupt: |
|
|
1756 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1757 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1758 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1372 | } |
1759 | } |
1373 | |
1760 | |
1374 | ///////////////////////////////////////////////////////////////////////////// |
1761 | ///////////////////////////////////////////////////////////////////////////// |
1375 | // XS interface functions |
1762 | // XS interface functions |
1376 | |
1763 | |
… | |
… | |
1388 | : -1; |
1775 | : -1; |
1389 | |
1776 | |
1390 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1777 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1391 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1778 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1392 | |
1779 | |
1393 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1780 | json_true = get_bool ("JSON::XS::true"); |
1394 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1781 | json_false = get_bool ("JSON::XS::false"); |
|
|
1782 | |
|
|
1783 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1395 | } |
1784 | } |
1396 | |
1785 | |
1397 | PROTOTYPES: DISABLE |
1786 | PROTOTYPES: DISABLE |
1398 | |
1787 | |
1399 | void CLONE (...) |
1788 | void CLONE (...) |
… | |
… | |
1404 | void new (char *klass) |
1793 | void new (char *klass) |
1405 | PPCODE: |
1794 | PPCODE: |
1406 | { |
1795 | { |
1407 | SV *pv = NEWSV (0, sizeof (JSON)); |
1796 | SV *pv = NEWSV (0, sizeof (JSON)); |
1408 | SvPOK_only (pv); |
1797 | SvPOK_only (pv); |
1409 | Zero (SvPVX (pv), 1, JSON); |
1798 | json_init ((JSON *)SvPVX (pv)); |
1410 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1799 | XPUSHs (sv_2mortal (sv_bless ( |
1411 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); |
1800 | newRV_noinc (pv), |
|
|
1801 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1802 | ))); |
1412 | } |
1803 | } |
1413 | |
1804 | |
1414 | void ascii (JSON *self, int enable = 1) |
1805 | void ascii (JSON *self, int enable = 1) |
1415 | ALIAS: |
1806 | ALIAS: |
1416 | ascii = F_ASCII |
1807 | ascii = F_ASCII |
… | |
… | |
1423 | pretty = F_PRETTY |
1814 | pretty = F_PRETTY |
1424 | allow_nonref = F_ALLOW_NONREF |
1815 | allow_nonref = F_ALLOW_NONREF |
1425 | shrink = F_SHRINK |
1816 | shrink = F_SHRINK |
1426 | allow_blessed = F_ALLOW_BLESSED |
1817 | allow_blessed = F_ALLOW_BLESSED |
1427 | convert_blessed = F_CONV_BLESSED |
1818 | convert_blessed = F_CONV_BLESSED |
|
|
1819 | relaxed = F_RELAXED |
|
|
1820 | allow_unknown = F_ALLOW_UNKNOWN |
1428 | PPCODE: |
1821 | PPCODE: |
1429 | { |
1822 | { |
1430 | if (enable) |
1823 | if (enable) |
1431 | self->flags |= ix; |
1824 | self->flags |= ix; |
1432 | else |
1825 | else |
1433 | self->flags &= ~ix; |
1826 | self->flags &= ~ix; |
1434 | |
1827 | |
1435 | XPUSHs (ST (0)); |
1828 | XPUSHs (ST (0)); |
1436 | } |
1829 | } |
1437 | |
1830 | |
1438 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1831 | void get_ascii (JSON *self) |
|
|
1832 | ALIAS: |
|
|
1833 | get_ascii = F_ASCII |
|
|
1834 | get_latin1 = F_LATIN1 |
|
|
1835 | get_utf8 = F_UTF8 |
|
|
1836 | get_indent = F_INDENT |
|
|
1837 | get_canonical = F_CANONICAL |
|
|
1838 | get_space_before = F_SPACE_BEFORE |
|
|
1839 | get_space_after = F_SPACE_AFTER |
|
|
1840 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1841 | get_shrink = F_SHRINK |
|
|
1842 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1843 | get_convert_blessed = F_CONV_BLESSED |
|
|
1844 | get_relaxed = F_RELAXED |
|
|
1845 | get_allow_unknown = F_ALLOW_UNKNOWN |
1439 | PPCODE: |
1846 | PPCODE: |
1440 | { |
1847 | XPUSHs (boolSV (self->flags & ix)); |
1441 | UV log2 = 0; |
|
|
1442 | |
1848 | |
1443 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1849 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1444 | |
1850 | PPCODE: |
1445 | while ((1UL << log2) < max_depth) |
1851 | self->max_depth = max_depth; |
1446 | ++log2; |
|
|
1447 | |
|
|
1448 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1449 | |
|
|
1450 | XPUSHs (ST (0)); |
1852 | XPUSHs (ST (0)); |
1451 | } |
|
|
1452 | |
1853 | |
|
|
1854 | U32 get_max_depth (JSON *self) |
|
|
1855 | CODE: |
|
|
1856 | RETVAL = self->max_depth; |
|
|
1857 | OUTPUT: |
|
|
1858 | RETVAL |
|
|
1859 | |
1453 | void max_size (JSON *self, UV max_size = 0) |
1860 | void max_size (JSON *self, U32 max_size = 0) |
1454 | PPCODE: |
1861 | PPCODE: |
1455 | { |
1862 | self->max_size = max_size; |
1456 | UV log2 = 0; |
|
|
1457 | |
|
|
1458 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1459 | if (max_size == 1) max_size = 2; |
|
|
1460 | |
|
|
1461 | while ((1UL << log2) < max_size) |
|
|
1462 | ++log2; |
|
|
1463 | |
|
|
1464 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1465 | |
|
|
1466 | XPUSHs (ST (0)); |
1863 | XPUSHs (ST (0)); |
1467 | } |
1864 | |
|
|
1865 | int get_max_size (JSON *self) |
|
|
1866 | CODE: |
|
|
1867 | RETVAL = self->max_size; |
|
|
1868 | OUTPUT: |
|
|
1869 | RETVAL |
1468 | |
1870 | |
1469 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1871 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1470 | PPCODE: |
1872 | PPCODE: |
1471 | { |
1873 | { |
1472 | SvREFCNT_dec (self->cb_object); |
1874 | SvREFCNT_dec (self->cb_object); |
… | |
… | |
1506 | XPUSHs (decode_json (jsonstr, self, 0)); |
1908 | XPUSHs (decode_json (jsonstr, self, 0)); |
1507 | |
1909 | |
1508 | void decode_prefix (JSON *self, SV *jsonstr) |
1910 | void decode_prefix (JSON *self, SV *jsonstr) |
1509 | PPCODE: |
1911 | PPCODE: |
1510 | { |
1912 | { |
1511 | UV offset; |
1913 | char *offset; |
1512 | EXTEND (SP, 2); |
1914 | EXTEND (SP, 2); |
1513 | PUSHs (decode_json (jsonstr, self, &offset)); |
1915 | PUSHs (decode_json (jsonstr, self, &offset)); |
1514 | PUSHs (sv_2mortal (newSVuv (offset))); |
1916 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1917 | } |
|
|
1918 | |
|
|
1919 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1920 | PPCODE: |
|
|
1921 | { |
|
|
1922 | if (!self->incr_text) |
|
|
1923 | self->incr_text = newSVpvn ("", 0); |
|
|
1924 | |
|
|
1925 | // append data, if any |
|
|
1926 | if (jsonstr) |
|
|
1927 | { |
|
|
1928 | if (SvUTF8 (jsonstr)) |
|
|
1929 | { |
|
|
1930 | if (!SvUTF8 (self->incr_text)) |
|
|
1931 | { |
|
|
1932 | /* utf-8-ness differs, need to upgrade */ |
|
|
1933 | sv_utf8_upgrade (self->incr_text); |
|
|
1934 | |
|
|
1935 | if (self->incr_pos) |
|
|
1936 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1937 | - (U8 *)SvPVX (self->incr_text); |
|
|
1938 | } |
|
|
1939 | } |
|
|
1940 | else if (SvUTF8 (self->incr_text)) |
|
|
1941 | sv_utf8_upgrade (jsonstr); |
|
|
1942 | |
|
|
1943 | { |
|
|
1944 | STRLEN len; |
|
|
1945 | const char *str = SvPV (jsonstr, len); |
|
|
1946 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1947 | |
|
|
1948 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1949 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1950 | |
|
|
1951 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1952 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1953 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1954 | } |
|
|
1955 | } |
|
|
1956 | |
|
|
1957 | if (GIMME_V != G_VOID) |
|
|
1958 | do |
|
|
1959 | { |
|
|
1960 | char *offset; |
|
|
1961 | |
|
|
1962 | if (!INCR_DONE (self)) |
|
|
1963 | { |
|
|
1964 | incr_parse (self); |
|
|
1965 | |
|
|
1966 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1967 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1968 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1969 | |
|
|
1970 | if (!INCR_DONE (self)) |
|
|
1971 | break; |
|
|
1972 | } |
|
|
1973 | |
|
|
1974 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1975 | |
|
|
1976 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
1977 | self->incr_nest = 0; |
|
|
1978 | self->incr_mode = 0; |
|
|
1979 | |
|
|
1980 | sv_chop (self->incr_text, offset); |
|
|
1981 | } |
|
|
1982 | while (GIMME_V == G_ARRAY); |
|
|
1983 | } |
|
|
1984 | |
|
|
1985 | SV *incr_text (JSON *self) |
|
|
1986 | ATTRS: lvalue |
|
|
1987 | CODE: |
|
|
1988 | { |
|
|
1989 | if (self->incr_pos) |
|
|
1990 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
1991 | |
|
|
1992 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
1993 | } |
|
|
1994 | OUTPUT: |
|
|
1995 | RETVAL |
|
|
1996 | |
|
|
1997 | void incr_skip (JSON *self) |
|
|
1998 | CODE: |
|
|
1999 | { |
|
|
2000 | if (self->incr_pos) |
|
|
2001 | { |
|
|
2002 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2003 | self->incr_pos = 0; |
|
|
2004 | self->incr_nest = 0; |
|
|
2005 | self->incr_mode = 0; |
|
|
2006 | } |
|
|
2007 | } |
|
|
2008 | |
|
|
2009 | void incr_reset (JSON *self) |
|
|
2010 | CODE: |
|
|
2011 | { |
|
|
2012 | SvREFCNT_dec (self->incr_text); |
|
|
2013 | self->incr_text = 0; |
|
|
2014 | self->incr_pos = 0; |
|
|
2015 | self->incr_nest = 0; |
|
|
2016 | self->incr_mode = 0; |
1515 | } |
2017 | } |
1516 | |
2018 | |
1517 | void DESTROY (JSON *self) |
2019 | void DESTROY (JSON *self) |
1518 | CODE: |
2020 | CODE: |
1519 | SvREFCNT_dec (self->cb_sk_object); |
2021 | SvREFCNT_dec (self->cb_sk_object); |
1520 | SvREFCNT_dec (self->cb_object); |
2022 | SvREFCNT_dec (self->cb_object); |
|
|
2023 | SvREFCNT_dec (self->incr_text); |
1521 | |
2024 | |
1522 | PROTOTYPES: ENABLE |
2025 | PROTOTYPES: ENABLE |
1523 | |
2026 | |
1524 | void to_json (SV *scalar) |
2027 | void encode_json (SV *scalar) |
|
|
2028 | ALIAS: |
|
|
2029 | to_json_ = 0 |
|
|
2030 | encode_json = F_UTF8 |
1525 | PPCODE: |
2031 | PPCODE: |
1526 | { |
2032 | { |
1527 | JSON json = { F_DEFAULT | F_UTF8 }; |
2033 | JSON json; |
|
|
2034 | json_init (&json); |
|
|
2035 | json.flags |= ix; |
1528 | XPUSHs (encode_json (scalar, &json)); |
2036 | XPUSHs (encode_json (scalar, &json)); |
1529 | } |
2037 | } |
1530 | |
2038 | |
1531 | void from_json (SV *jsonstr) |
2039 | void decode_json (SV *jsonstr) |
|
|
2040 | ALIAS: |
|
|
2041 | from_json_ = 0 |
|
|
2042 | decode_json = F_UTF8 |
1532 | PPCODE: |
2043 | PPCODE: |
1533 | { |
2044 | { |
1534 | JSON json = { F_DEFAULT | F_UTF8 }; |
2045 | JSON json; |
|
|
2046 | json_init (&json); |
|
|
2047 | json.flags |= ix; |
1535 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2048 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1536 | } |
2049 | } |
1537 | |
2050 | |