… | |
… | |
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 | |
|
|
47 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
|
|
48 | |
49 | #define SB do { |
49 | #define SB do { |
50 | #define SE } while (0) |
50 | #define SE } while (0) |
51 | |
51 | |
52 | #if __GNUC__ >= 3 |
52 | #if __GNUC__ >= 3 |
53 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
53 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
54 | # define inline inline |
54 | # define INLINE static inline |
55 | #else |
55 | #else |
56 | # define expect(expr,value) (expr) |
56 | # define expect(expr,value) (expr) |
57 | # define inline static |
57 | # define INLINE static |
58 | #endif |
58 | #endif |
59 | |
59 | |
60 | #define expect_false(expr) expect ((expr) != 0, 0) |
60 | #define expect_false(expr) expect ((expr) != 0, 0) |
61 | #define expect_true(expr) expect ((expr) != 0, 1) |
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?)" |
62 | |
68 | |
63 | #ifdef USE_ITHREADS |
69 | #ifdef USE_ITHREADS |
64 | # define JSON_SLOW 1 |
70 | # define JSON_SLOW 1 |
65 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
66 | #else |
72 | #else |
… | |
… | |
69 | #endif |
75 | #endif |
70 | |
76 | |
71 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
77 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
72 | static SV *json_true, *json_false; |
78 | static SV *json_true, *json_false; |
73 | |
79 | |
|
|
80 | enum { |
|
|
81 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
82 | INCR_M_STR, // inside string |
|
|
83 | INCR_M_BS, // inside backslash |
|
|
84 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
85 | INCR_M_C1, // inside comment in other places |
|
|
86 | INCR_M_JSON // outside anything, count nesting |
|
|
87 | }; |
|
|
88 | |
|
|
89 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
90 | |
74 | typedef struct { |
91 | typedef struct { |
75 | U32 flags; |
92 | U32 flags; |
|
|
93 | U32 max_depth; |
|
|
94 | STRLEN max_size; |
|
|
95 | |
76 | SV *cb_object; |
96 | SV *cb_object; |
77 | HV *cb_sk_object; |
97 | HV *cb_sk_object; |
|
|
98 | |
|
|
99 | // for the incremental parser |
|
|
100 | SV *incr_text; // the source text so far |
|
|
101 | STRLEN incr_pos; // the current offset into the text |
|
|
102 | int incr_nest; // {[]}-nesting level |
|
|
103 | unsigned char incr_mode; |
78 | } JSON; |
104 | } JSON; |
|
|
105 | |
|
|
106 | INLINE void |
|
|
107 | json_init (JSON *json) |
|
|
108 | { |
|
|
109 | Zero (json, 1, JSON); |
|
|
110 | json->max_depth = 512; |
|
|
111 | } |
79 | |
112 | |
80 | ///////////////////////////////////////////////////////////////////////////// |
113 | ///////////////////////////////////////////////////////////////////////////// |
81 | // utility functions |
114 | // utility functions |
82 | |
115 | |
83 | inline void |
116 | INLINE SV * |
|
|
117 | get_bool (const char *name) |
|
|
118 | { |
|
|
119 | SV *sv = get_sv (name, 1); |
|
|
120 | |
|
|
121 | SvREADONLY_on (sv); |
|
|
122 | SvREADONLY_on (SvRV (sv)); |
|
|
123 | |
|
|
124 | return sv; |
|
|
125 | } |
|
|
126 | |
|
|
127 | INLINE void |
84 | shrink (SV *sv) |
128 | shrink (SV *sv) |
85 | { |
129 | { |
86 | sv_utf8_downgrade (sv, 1); |
130 | sv_utf8_downgrade (sv, 1); |
|
|
131 | |
87 | if (SvLEN (sv) > SvCUR (sv) + 1) |
132 | if (SvLEN (sv) > SvCUR (sv) + 1) |
88 | { |
133 | { |
89 | #ifdef SvPV_shrink_to_cur |
134 | #ifdef SvPV_shrink_to_cur |
90 | SvPV_shrink_to_cur (sv); |
135 | SvPV_shrink_to_cur (sv); |
91 | #elif defined (SvPV_renew) |
136 | #elif defined (SvPV_renew) |
… | |
… | |
97 | // decode an utf-8 character and return it, or (UV)-1 in |
142 | // decode an utf-8 character and return it, or (UV)-1 in |
98 | // case of an error. |
143 | // case of an error. |
99 | // we special-case "safe" characters from U+80 .. U+7FF, |
144 | // we special-case "safe" characters from U+80 .. U+7FF, |
100 | // but use the very good perl function to parse anything else. |
145 | // but use the very good perl function to parse anything else. |
101 | // note that we never call this function for a ascii codepoints |
146 | // note that we never call this function for a ascii codepoints |
102 | inline UV |
147 | INLINE UV |
103 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
148 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
104 | { |
149 | { |
105 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
150 | if (expect_true (len >= 2 |
106 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
151 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
107 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
152 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
108 | { |
153 | { |
109 | *clen = 2; |
154 | *clen = 2; |
110 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
155 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
111 | } |
156 | } |
112 | else |
157 | else |
113 | { |
158 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
114 | *clen = (STRLEN)-1; |
159 | } |
115 | return (UV)-1; |
160 | |
|
|
161 | // likewise for encoding, also never called for ascii codepoints |
|
|
162 | // this function takes advantage of this fact, although current gccs |
|
|
163 | // seem to optimise the check for >= 0x80 away anyways |
|
|
164 | INLINE unsigned char * |
|
|
165 | encode_utf8 (unsigned char *s, UV ch) |
|
|
166 | { |
|
|
167 | if (expect_false (ch < 0x000080)) |
|
|
168 | *s++ = ch; |
|
|
169 | else if (expect_true (ch < 0x000800)) |
|
|
170 | *s++ = 0xc0 | ( ch >> 6), |
|
|
171 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
172 | else if ( ch < 0x010000) |
|
|
173 | *s++ = 0xe0 | ( ch >> 12), |
|
|
174 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
175 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
176 | else if ( ch < 0x110000) |
|
|
177 | *s++ = 0xf0 | ( ch >> 18), |
|
|
178 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
179 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
180 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
181 | |
|
|
182 | return s; |
|
|
183 | } |
|
|
184 | |
|
|
185 | // convert offset pointer to character index, sv must be string |
|
|
186 | static STRLEN |
|
|
187 | ptr_to_index (SV *sv, char *offset) |
|
|
188 | { |
|
|
189 | return SvUTF8 (sv) |
|
|
190 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
191 | : offset - SvPVX (sv); |
|
|
192 | } |
|
|
193 | |
|
|
194 | ///////////////////////////////////////////////////////////////////////////// |
|
|
195 | // fp hell |
|
|
196 | |
|
|
197 | // scan a group of digits, and a trailing exponent |
|
|
198 | static void |
|
|
199 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
200 | { |
|
|
201 | UV uaccum = 0; |
|
|
202 | int eaccum = 0; |
|
|
203 | |
|
|
204 | // if we recurse too deep, skip all remaining digits |
|
|
205 | // to avoid a stack overflow attack |
|
|
206 | if (expect_false (--maxdepth <= 0)) |
|
|
207 | while (((U8)*s - '0') < 10) |
|
|
208 | ++s; |
|
|
209 | |
|
|
210 | for (;;) |
116 | } |
211 | { |
117 | } |
212 | U8 dig = (U8)*s - '0'; |
118 | |
213 | |
|
|
214 | if (expect_false (dig >= 10)) |
|
|
215 | { |
|
|
216 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
217 | { |
|
|
218 | ++s; |
|
|
219 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
220 | } |
|
|
221 | else if ((dig | ' ') == 'e' - '0') |
|
|
222 | { |
|
|
223 | int exp2 = 0; |
|
|
224 | int neg = 0; |
|
|
225 | |
|
|
226 | ++s; |
|
|
227 | |
|
|
228 | if (*s == '-') |
|
|
229 | { |
|
|
230 | ++s; |
|
|
231 | neg = 1; |
|
|
232 | } |
|
|
233 | else if (*s == '+') |
|
|
234 | ++s; |
|
|
235 | |
|
|
236 | while ((dig = (U8)*s - '0') < 10) |
|
|
237 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
238 | |
|
|
239 | *expo += neg ? -exp2 : exp2; |
|
|
240 | } |
|
|
241 | |
|
|
242 | break; |
|
|
243 | } |
|
|
244 | |
|
|
245 | ++s; |
|
|
246 | |
|
|
247 | uaccum = uaccum * 10 + dig; |
|
|
248 | ++eaccum; |
|
|
249 | |
|
|
250 | // if we have too many digits, then recurse for more |
|
|
251 | // we actually do this for rather few digits |
|
|
252 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
253 | { |
|
|
254 | if (postdp) *expo -= eaccum; |
|
|
255 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
256 | if (postdp) *expo += eaccum; |
|
|
257 | |
|
|
258 | break; |
|
|
259 | } |
|
|
260 | } |
|
|
261 | |
|
|
262 | // this relies greatly on the quality of the pow () |
|
|
263 | // implementation of the platform, but a good |
|
|
264 | // implementation is hard to beat. |
|
|
265 | if (postdp) *expo -= eaccum; |
|
|
266 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
267 | *expo += eaccum; |
|
|
268 | } |
|
|
269 | |
|
|
270 | static NV |
|
|
271 | json_atof (const char *s) |
|
|
272 | { |
|
|
273 | NV accum = 0.; |
|
|
274 | int expo = 0; |
|
|
275 | int neg = 0; |
|
|
276 | |
|
|
277 | if (*s == '-') |
|
|
278 | { |
|
|
279 | ++s; |
|
|
280 | neg = 1; |
|
|
281 | } |
|
|
282 | |
|
|
283 | // a recursion depth of ten gives us >>500 bits |
|
|
284 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
285 | |
|
|
286 | return neg ? -accum : accum; |
|
|
287 | } |
119 | ///////////////////////////////////////////////////////////////////////////// |
288 | ///////////////////////////////////////////////////////////////////////////// |
120 | // encoder |
289 | // encoder |
121 | |
290 | |
122 | // structure used for encoding JSON |
291 | // structure used for encoding JSON |
123 | typedef struct |
292 | typedef struct |
… | |
… | |
125 | char *cur; // SvPVX (sv) + current output position |
294 | char *cur; // SvPVX (sv) + current output position |
126 | char *end; // SvEND (sv) |
295 | char *end; // SvEND (sv) |
127 | SV *sv; // result scalar |
296 | SV *sv; // result scalar |
128 | JSON json; |
297 | JSON json; |
129 | U32 indent; // indentation level |
298 | U32 indent; // indentation level |
130 | U32 maxdepth; // max. indentation/recursion level |
299 | UV limit; // escape character values >= this value when encoding |
131 | } enc_t; |
300 | } enc_t; |
132 | |
301 | |
133 | inline void |
302 | INLINE void |
134 | need (enc_t *enc, STRLEN len) |
303 | need (enc_t *enc, STRLEN len) |
135 | { |
304 | { |
136 | if (expect_false (enc->cur + len >= enc->end)) |
305 | if (expect_false (enc->cur + len >= enc->end)) |
137 | { |
306 | { |
138 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
307 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
139 | SvGROW (enc->sv, cur + len + 1); |
308 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
140 | enc->cur = SvPVX (enc->sv) + cur; |
309 | enc->cur = SvPVX (enc->sv) + cur; |
141 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
310 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
142 | } |
311 | } |
143 | } |
312 | } |
144 | |
313 | |
145 | inline void |
314 | INLINE void |
146 | encode_ch (enc_t *enc, char ch) |
315 | encode_ch (enc_t *enc, char ch) |
147 | { |
316 | { |
148 | need (enc, 1); |
317 | need (enc, 1); |
149 | *enc->cur++ = ch; |
318 | *enc->cur++ = ch; |
150 | } |
319 | } |
… | |
… | |
204 | { |
373 | { |
205 | uch = ch; |
374 | uch = ch; |
206 | clen = 1; |
375 | clen = 1; |
207 | } |
376 | } |
208 | |
377 | |
209 | if (uch > 0x10FFFFUL) |
378 | 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 | { |
379 | { |
214 | if (uch > 0xFFFFUL) |
380 | if (uch >= 0x10000UL) |
215 | { |
381 | { |
|
|
382 | if (uch >= 0x110000UL) |
|
|
383 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
384 | |
216 | need (enc, len += 11); |
385 | need (enc, len += 11); |
217 | sprintf (enc->cur, "\\u%04x\\u%04x", |
386 | sprintf (enc->cur, "\\u%04x\\u%04x", |
218 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
387 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
219 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
388 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
220 | enc->cur += 12; |
389 | enc->cur += 12; |
221 | } |
390 | } |
222 | else |
391 | else |
223 | { |
392 | { |
224 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
225 | need (enc, len += 5); |
393 | need (enc, len += 5); |
226 | *enc->cur++ = '\\'; |
394 | *enc->cur++ = '\\'; |
227 | *enc->cur++ = 'u'; |
395 | *enc->cur++ = 'u'; |
228 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
396 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
229 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
397 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
230 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
398 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
231 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
399 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
232 | } |
400 | } |
233 | |
401 | |
234 | str += clen; |
402 | str += clen; |
235 | } |
403 | } |
236 | else if (enc->json.flags & F_LATIN1) |
404 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
248 | while (--clen); |
416 | while (--clen); |
249 | } |
417 | } |
250 | else |
418 | else |
251 | { |
419 | { |
252 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
420 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
253 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
421 | enc->cur = encode_utf8 (enc->cur, uch); |
254 | ++str; |
422 | ++str; |
255 | } |
423 | } |
256 | } |
424 | } |
257 | } |
425 | } |
258 | } |
426 | } |
259 | |
427 | |
260 | --len; |
428 | --len; |
261 | } |
429 | } |
262 | } |
430 | } |
263 | |
431 | |
264 | inline void |
432 | INLINE void |
265 | encode_indent (enc_t *enc) |
433 | encode_indent (enc_t *enc) |
266 | { |
434 | { |
267 | if (enc->json.flags & F_INDENT) |
435 | if (enc->json.flags & F_INDENT) |
268 | { |
436 | { |
269 | int spaces = enc->indent * INDENT_STEP; |
437 | int spaces = enc->indent * INDENT_STEP; |
… | |
… | |
272 | memset (enc->cur, ' ', spaces); |
440 | memset (enc->cur, ' ', spaces); |
273 | enc->cur += spaces; |
441 | enc->cur += spaces; |
274 | } |
442 | } |
275 | } |
443 | } |
276 | |
444 | |
277 | inline void |
445 | INLINE void |
278 | encode_space (enc_t *enc) |
446 | encode_space (enc_t *enc) |
279 | { |
447 | { |
280 | need (enc, 1); |
448 | need (enc, 1); |
281 | encode_ch (enc, ' '); |
449 | encode_ch (enc, ' '); |
282 | } |
450 | } |
283 | |
451 | |
284 | inline void |
452 | INLINE void |
285 | encode_nl (enc_t *enc) |
453 | encode_nl (enc_t *enc) |
286 | { |
454 | { |
287 | if (enc->json.flags & F_INDENT) |
455 | if (enc->json.flags & F_INDENT) |
288 | { |
456 | { |
289 | need (enc, 1); |
457 | need (enc, 1); |
290 | encode_ch (enc, '\n'); |
458 | encode_ch (enc, '\n'); |
291 | } |
459 | } |
292 | } |
460 | } |
293 | |
461 | |
294 | inline void |
462 | INLINE void |
295 | encode_comma (enc_t *enc) |
463 | encode_comma (enc_t *enc) |
296 | { |
464 | { |
297 | encode_ch (enc, ','); |
465 | encode_ch (enc, ','); |
298 | |
466 | |
299 | if (enc->json.flags & F_INDENT) |
467 | if (enc->json.flags & F_INDENT) |
… | |
… | |
307 | static void |
475 | static void |
308 | encode_av (enc_t *enc, AV *av) |
476 | encode_av (enc_t *enc, AV *av) |
309 | { |
477 | { |
310 | int i, len = av_len (av); |
478 | int i, len = av_len (av); |
311 | |
479 | |
312 | if (enc->indent >= enc->maxdepth) |
480 | if (enc->indent >= enc->json.max_depth) |
313 | croak ("data structure too deep (hit recursion limit)"); |
481 | croak (ERR_NESTING_EXCEEDED); |
314 | |
482 | |
315 | encode_ch (enc, '['); encode_nl (enc); |
483 | encode_ch (enc, '['); |
316 | ++enc->indent; |
484 | |
|
|
485 | if (len >= 0) |
|
|
486 | { |
|
|
487 | encode_nl (enc); ++enc->indent; |
317 | |
488 | |
318 | for (i = 0; i <= len; ++i) |
489 | for (i = 0; i <= len; ++i) |
319 | { |
490 | { |
320 | SV **svp = av_fetch (av, i, 0); |
491 | SV **svp = av_fetch (av, i, 0); |
321 | |
492 | |
322 | encode_indent (enc); |
493 | encode_indent (enc); |
323 | |
494 | |
324 | if (svp) |
495 | if (svp) |
325 | encode_sv (enc, *svp); |
496 | encode_sv (enc, *svp); |
326 | else |
497 | else |
327 | encode_str (enc, "null", 4, 0); |
498 | encode_str (enc, "null", 4, 0); |
328 | |
499 | |
329 | if (i < len) |
500 | if (i < len) |
330 | encode_comma (enc); |
501 | encode_comma (enc); |
331 | } |
502 | } |
332 | |
503 | |
|
|
504 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
505 | } |
|
|
506 | |
333 | encode_nl (enc); |
507 | encode_ch (enc, ']'); |
334 | |
|
|
335 | --enc->indent; |
|
|
336 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
337 | } |
508 | } |
338 | |
509 | |
339 | static void |
510 | static void |
340 | encode_he (enc_t *enc, HE *he) |
511 | encode_hk (enc_t *enc, HE *he) |
341 | { |
512 | { |
342 | encode_ch (enc, '"'); |
513 | encode_ch (enc, '"'); |
343 | |
514 | |
344 | if (HeKLEN (he) == HEf_SVKEY) |
515 | if (HeKLEN (he) == HEf_SVKEY) |
345 | { |
516 | { |
… | |
… | |
358 | encode_ch (enc, '"'); |
529 | encode_ch (enc, '"'); |
359 | |
530 | |
360 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
531 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
361 | encode_ch (enc, ':'); |
532 | encode_ch (enc, ':'); |
362 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
533 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
363 | encode_sv (enc, HeVAL (he)); |
|
|
364 | } |
534 | } |
365 | |
535 | |
366 | // compare hash entries, used when all keys are bytestrings |
536 | // compare hash entries, used when all keys are bytestrings |
367 | static int |
537 | static int |
368 | he_cmp_fast (const void *a_, const void *b_) |
538 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
373 | HE *b = *(HE **)b_; |
543 | HE *b = *(HE **)b_; |
374 | |
544 | |
375 | STRLEN la = HeKLEN (a); |
545 | STRLEN la = HeKLEN (a); |
376 | STRLEN lb = HeKLEN (b); |
546 | STRLEN lb = HeKLEN (b); |
377 | |
547 | |
378 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
548 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
379 | cmp = la - lb; |
549 | cmp = lb - la; |
380 | |
550 | |
381 | return cmp; |
551 | return cmp; |
382 | } |
552 | } |
383 | |
553 | |
384 | // compare hash entries, used when some keys are sv's or utf-x |
554 | // compare hash entries, used when some keys are sv's or utf-x |
385 | static int |
555 | static int |
386 | he_cmp_slow (const void *a, const void *b) |
556 | he_cmp_slow (const void *a, const void *b) |
387 | { |
557 | { |
388 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
558 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
389 | } |
559 | } |
390 | |
560 | |
391 | static void |
561 | static void |
392 | encode_hv (enc_t *enc, HV *hv) |
562 | encode_hv (enc_t *enc, HV *hv) |
393 | { |
563 | { |
394 | int count, i; |
564 | HE *he; |
395 | |
565 | |
396 | if (enc->indent >= enc->maxdepth) |
566 | if (enc->indent >= enc->json.max_depth) |
397 | croak ("data structure too deep (hit recursion limit)"); |
567 | croak (ERR_NESTING_EXCEEDED); |
398 | |
568 | |
399 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
569 | encode_ch (enc, '{'); |
400 | |
570 | |
401 | if ((count = hv_iterinit (hv))) |
|
|
402 | { |
|
|
403 | // for canonical output we have to sort by keys first |
571 | // for canonical output we have to sort by keys first |
404 | // actually, this is mostly due to the stupid so-called |
572 | // actually, this is mostly due to the stupid so-called |
405 | // security workaround added somewhere in 5.8.x. |
573 | // security workaround added somewhere in 5.8.x |
406 | // that randomises hash orderings |
574 | // that randomises hash orderings |
407 | if (enc->json.flags & F_CANONICAL) |
575 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
576 | { |
|
|
577 | int count = hv_iterinit (hv); |
|
|
578 | |
|
|
579 | if (SvMAGICAL (hv)) |
408 | { |
580 | { |
|
|
581 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
582 | // but I don't care for the speed of this special case. |
|
|
583 | // note also that we will run into undefined behaviour when the two iterations |
|
|
584 | // do not result in the same count, something I might care for in some later release. |
|
|
585 | |
|
|
586 | count = 0; |
|
|
587 | while (hv_iternext (hv)) |
|
|
588 | ++count; |
|
|
589 | |
|
|
590 | hv_iterinit (hv); |
|
|
591 | } |
|
|
592 | |
|
|
593 | if (count) |
|
|
594 | { |
409 | int fast = 1; |
595 | int i, fast = 1; |
410 | HE *he; |
|
|
411 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
596 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
412 | HE **hes = _alloca (count * sizeof (HE)); |
597 | HE **hes = _alloca (count * sizeof (HE)); |
413 | #else |
598 | #else |
414 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
599 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
415 | #endif |
600 | #endif |
… | |
… | |
442 | |
627 | |
443 | FREETMPS; |
628 | FREETMPS; |
444 | LEAVE; |
629 | LEAVE; |
445 | } |
630 | } |
446 | |
631 | |
447 | for (i = 0; i < count; ++i) |
632 | encode_nl (enc); ++enc->indent; |
|
|
633 | |
|
|
634 | while (count--) |
448 | { |
635 | { |
449 | encode_indent (enc); |
636 | encode_indent (enc); |
|
|
637 | he = hes [count]; |
450 | encode_he (enc, hes [i]); |
638 | encode_hk (enc, he); |
|
|
639 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
451 | |
640 | |
452 | if (i < count - 1) |
641 | if (count) |
453 | encode_comma (enc); |
642 | encode_comma (enc); |
454 | } |
643 | } |
455 | |
644 | |
456 | encode_nl (enc); |
645 | encode_nl (enc); --enc->indent; encode_indent (enc); |
457 | } |
646 | } |
|
|
647 | } |
458 | else |
648 | else |
|
|
649 | { |
|
|
650 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
651 | if ((he = hv_iternext (hv))) |
459 | { |
652 | { |
460 | HE *he = hv_iternext (hv); |
653 | encode_nl (enc); ++enc->indent; |
461 | |
654 | |
462 | for (;;) |
655 | for (;;) |
463 | { |
656 | { |
464 | encode_indent (enc); |
657 | encode_indent (enc); |
465 | encode_he (enc, he); |
658 | encode_hk (enc, he); |
|
|
659 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
466 | |
660 | |
467 | if (!(he = hv_iternext (hv))) |
661 | if (!(he = hv_iternext (hv))) |
468 | break; |
662 | break; |
469 | |
663 | |
470 | encode_comma (enc); |
664 | encode_comma (enc); |
471 | } |
665 | } |
472 | |
666 | |
473 | encode_nl (enc); |
667 | encode_nl (enc); --enc->indent; encode_indent (enc); |
474 | } |
668 | } |
475 | } |
669 | } |
476 | |
670 | |
477 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
671 | encode_ch (enc, '}'); |
478 | } |
672 | } |
479 | |
673 | |
480 | // encode objects, arrays and special \0=false and \1=true values. |
674 | // encode objects, arrays and special \0=false and \1=true values. |
481 | static void |
675 | static void |
482 | encode_rv (enc_t *enc, SV *sv) |
676 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
559 | |
753 | |
560 | if (len == 1 && *pv == '1') |
754 | if (len == 1 && *pv == '1') |
561 | encode_str (enc, "true", 4, 0); |
755 | encode_str (enc, "true", 4, 0); |
562 | else if (len == 1 && *pv == '0') |
756 | else if (len == 1 && *pv == '0') |
563 | encode_str (enc, "false", 5, 0); |
757 | encode_str (enc, "false", 5, 0); |
|
|
758 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
759 | encode_str (enc, "null", 4, 0); |
564 | else |
760 | else |
565 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
761 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
566 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
762 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
567 | } |
763 | } |
|
|
764 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
765 | encode_str (enc, "null", 4, 0); |
568 | else |
766 | else |
569 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
767 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
570 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
768 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
571 | } |
769 | } |
572 | |
770 | |
… | |
… | |
590 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
788 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
591 | enc->cur += strlen (enc->cur); |
789 | enc->cur += strlen (enc->cur); |
592 | } |
790 | } |
593 | else if (SvIOKp (sv)) |
791 | else if (SvIOKp (sv)) |
594 | { |
792 | { |
595 | // we assume we can always read an IV as a UV |
793 | // we assume we can always read an IV as a UV and vice versa |
596 | if (SvUV (sv) & ~(UV)0x7fff) |
794 | // we assume two's complement |
597 | { |
795 | // we assume no aliasing issues in the union |
598 | // large integer, use the (rather slow) snprintf way. |
796 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
599 | need (enc, sizeof (UV) * 3); |
797 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
600 | enc->cur += |
|
|
601 | SvIsUV(sv) |
|
|
602 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
|
|
603 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
604 | } |
|
|
605 | else |
|
|
606 | { |
798 | { |
607 | // optimise the "small number case" |
799 | // optimise the "small number case" |
608 | // code will likely be branchless and use only a single multiplication |
800 | // code will likely be branchless and use only a single multiplication |
|
|
801 | // works for numbers up to 59074 |
609 | I32 i = SvIV (sv); |
802 | I32 i = SvIVX (sv); |
610 | U32 u; |
803 | U32 u; |
611 | char digit, nz = 0; |
804 | char digit, nz = 0; |
612 | |
805 | |
613 | need (enc, 6); |
806 | need (enc, 6); |
614 | |
807 | |
… | |
… | |
620 | |
813 | |
621 | // now output digit by digit, each time masking out the integer part |
814 | // now output digit by digit, each time masking out the integer part |
622 | // and multiplying by 5 while moving the decimal point one to the right, |
815 | // and multiplying by 5 while moving the decimal point one to the right, |
623 | // resulting in a net multiplication by 10. |
816 | // resulting in a net multiplication by 10. |
624 | // we always write the digit to memory but conditionally increment |
817 | // we always write the digit to memory but conditionally increment |
625 | // the pointer, to ease the usage of conditional move instructions. |
818 | // the pointer, to enable the use of conditional move instructions. |
626 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
819 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
627 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
820 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
628 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
821 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
629 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
822 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
630 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
823 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
631 | } |
824 | } |
|
|
825 | else |
|
|
826 | { |
|
|
827 | // large integer, use the (rather slow) snprintf way. |
|
|
828 | need (enc, IVUV_MAXCHARS); |
|
|
829 | enc->cur += |
|
|
830 | SvIsUV(sv) |
|
|
831 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
832 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
833 | } |
632 | } |
834 | } |
633 | else if (SvROK (sv)) |
835 | else if (SvROK (sv)) |
634 | encode_rv (enc, SvRV (sv)); |
836 | encode_rv (enc, SvRV (sv)); |
635 | else if (!SvOK (sv)) |
837 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
636 | encode_str (enc, "null", 4, 0); |
838 | encode_str (enc, "null", 4, 0); |
637 | else |
839 | else |
638 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
840 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
639 | SvPV_nolen (sv), SvFLAGS (sv)); |
841 | SvPV_nolen (sv), SvFLAGS (sv)); |
640 | } |
842 | } |
… | |
… | |
650 | enc.json = *json; |
852 | enc.json = *json; |
651 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
853 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
652 | enc.cur = SvPVX (enc.sv); |
854 | enc.cur = SvPVX (enc.sv); |
653 | enc.end = SvEND (enc.sv); |
855 | enc.end = SvEND (enc.sv); |
654 | enc.indent = 0; |
856 | enc.indent = 0; |
655 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
857 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
858 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
859 | : 0x110000UL; |
656 | |
860 | |
657 | SvPOK_only (enc.sv); |
861 | SvPOK_only (enc.sv); |
658 | encode_sv (&enc, scalar); |
862 | encode_sv (&enc, scalar); |
|
|
863 | encode_nl (&enc); |
659 | |
864 | |
660 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
865 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
661 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
866 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
662 | |
867 | |
663 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
868 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
681 | JSON json; |
886 | JSON json; |
682 | U32 depth; // recursion depth |
887 | U32 depth; // recursion depth |
683 | U32 maxdepth; // recursion depth limit |
888 | U32 maxdepth; // recursion depth limit |
684 | } dec_t; |
889 | } dec_t; |
685 | |
890 | |
686 | inline void |
891 | INLINE void |
|
|
892 | decode_comment (dec_t *dec) |
|
|
893 | { |
|
|
894 | // only '#'-style comments allowed a.t.m. |
|
|
895 | |
|
|
896 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
897 | ++dec->cur; |
|
|
898 | } |
|
|
899 | |
|
|
900 | INLINE void |
687 | decode_ws (dec_t *dec) |
901 | decode_ws (dec_t *dec) |
688 | { |
902 | { |
689 | for (;;) |
903 | for (;;) |
690 | { |
904 | { |
691 | char ch = *dec->cur; |
905 | char ch = *dec->cur; |
692 | |
906 | |
693 | if (ch > 0x20 |
907 | if (ch > 0x20) |
|
|
908 | { |
|
|
909 | if (expect_false (ch == '#')) |
|
|
910 | { |
|
|
911 | if (dec->json.flags & F_RELAXED) |
|
|
912 | decode_comment (dec); |
|
|
913 | else |
|
|
914 | break; |
|
|
915 | } |
|
|
916 | else |
|
|
917 | break; |
|
|
918 | } |
694 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
919 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
695 | break; |
920 | break; // parse error, but let higher level handle it, gives better error messages |
696 | |
921 | |
697 | ++dec->cur; |
922 | ++dec->cur; |
698 | } |
923 | } |
699 | } |
924 | } |
700 | |
925 | |
… | |
… | |
704 | if (*dec->cur != ch) \ |
929 | if (*dec->cur != ch) \ |
705 | ERR (# ch " expected"); \ |
930 | ERR (# ch " expected"); \ |
706 | ++dec->cur; \ |
931 | ++dec->cur; \ |
707 | SE |
932 | SE |
708 | |
933 | |
709 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
934 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
710 | #define DEC_DEC_DEPTH --dec->depth |
935 | #define DEC_DEC_DEPTH --dec->depth |
711 | |
936 | |
712 | static SV *decode_sv (dec_t *dec); |
937 | static SV *decode_sv (dec_t *dec); |
713 | |
938 | |
714 | static signed char decode_hexdigit[256]; |
939 | static signed char decode_hexdigit[256]; |
… | |
… | |
806 | |
1031 | |
807 | if (hi >= 0x80) |
1032 | if (hi >= 0x80) |
808 | { |
1033 | { |
809 | utf8 = 1; |
1034 | utf8 = 1; |
810 | |
1035 | |
811 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
1036 | cur = encode_utf8 (cur, hi); |
812 | } |
1037 | } |
813 | else |
1038 | else |
814 | *cur++ = hi; |
1039 | *cur++ = hi; |
815 | } |
1040 | } |
816 | break; |
1041 | break; |
… | |
… | |
818 | default: |
1043 | default: |
819 | --dec_cur; |
1044 | --dec_cur; |
820 | ERR ("illegal backslash escape sequence in string"); |
1045 | ERR ("illegal backslash escape sequence in string"); |
821 | } |
1046 | } |
822 | } |
1047 | } |
823 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
1048 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
824 | *cur++ = ch; |
1049 | *cur++ = ch; |
825 | else if (ch >= 0x80) |
1050 | else if (ch >= 0x80) |
826 | { |
1051 | { |
827 | STRLEN clen; |
1052 | STRLEN clen; |
828 | UV uch; |
|
|
829 | |
1053 | |
830 | --dec_cur; |
1054 | --dec_cur; |
831 | |
1055 | |
832 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1056 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
833 | if (clen == (STRLEN)-1) |
1057 | if (clen == (STRLEN)-1) |
834 | ERR ("malformed UTF-8 character in JSON string"); |
1058 | ERR ("malformed UTF-8 character in JSON string"); |
835 | |
1059 | |
836 | do |
1060 | do |
837 | *cur++ = *dec_cur++; |
1061 | *cur++ = *dec_cur++; |
… | |
… | |
854 | { |
1078 | { |
855 | STRLEN len = cur - buf; |
1079 | STRLEN len = cur - buf; |
856 | |
1080 | |
857 | if (sv) |
1081 | if (sv) |
858 | { |
1082 | { |
859 | SvGROW (sv, SvCUR (sv) + len + 1); |
1083 | STRLEN cur = SvCUR (sv); |
|
|
1084 | |
|
|
1085 | if (SvLEN (sv) <= cur + len) |
|
|
1086 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1087 | |
860 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1088 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
861 | SvCUR_set (sv, SvCUR (sv) + len); |
1089 | SvCUR_set (sv, SvCUR (sv) + len); |
862 | } |
1090 | } |
863 | else |
1091 | else |
864 | sv = newSVpvn (buf, len); |
1092 | sv = newSVpvn (buf, len); |
… | |
… | |
951 | |
1179 | |
952 | if (!is_nv) |
1180 | if (!is_nv) |
953 | { |
1181 | { |
954 | int len = dec->cur - start; |
1182 | int len = dec->cur - start; |
955 | |
1183 | |
956 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1184 | // special case the rather common 1..5-digit-int case |
957 | if (*start == '-') |
1185 | if (*start == '-') |
958 | switch (len) |
1186 | switch (len) |
959 | { |
1187 | { |
960 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1188 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
961 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1189 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
962 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1190 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
963 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1191 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1192 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
964 | } |
1193 | } |
965 | else |
1194 | else |
966 | switch (len) |
1195 | switch (len) |
967 | { |
1196 | { |
968 | case 1: return newSViv ( start [0] - '0' * 1); |
1197 | case 1: return newSViv ( start [0] - '0' * 1); |
969 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1198 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
970 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1199 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
971 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1200 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1201 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
972 | } |
1202 | } |
973 | |
1203 | |
974 | { |
1204 | { |
975 | UV uv; |
1205 | UV uv; |
976 | int numtype = grok_number (start, len, &uv); |
1206 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
985 | } |
1215 | } |
986 | |
1216 | |
987 | len -= *start == '-' ? 1 : 0; |
1217 | len -= *start == '-' ? 1 : 0; |
988 | |
1218 | |
989 | // does not fit into IV or UV, try NV |
1219 | // does not fit into IV or UV, try NV |
990 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
1220 | if (len <= NV_DIG) |
991 | #if defined (LDBL_DIG) |
|
|
992 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
993 | #endif |
|
|
994 | ) |
|
|
995 | // fits into NV without loss of precision |
1221 | // fits into NV without loss of precision |
996 | return newSVnv (Atof (start)); |
1222 | return newSVnv (json_atof (start)); |
997 | |
1223 | |
998 | // everything else fails, convert it to a string |
1224 | // everything else fails, convert it to a string |
999 | return newSVpvn (start, dec->cur - start); |
1225 | return newSVpvn (start, dec->cur - start); |
1000 | } |
1226 | } |
1001 | |
1227 | |
1002 | // loss of precision here |
1228 | // loss of precision here |
1003 | return newSVnv (Atof (start)); |
1229 | return newSVnv (json_atof (start)); |
1004 | |
1230 | |
1005 | fail: |
1231 | fail: |
1006 | return 0; |
1232 | return 0; |
1007 | } |
1233 | } |
1008 | |
1234 | |
… | |
… | |
1037 | |
1263 | |
1038 | if (*dec->cur != ',') |
1264 | if (*dec->cur != ',') |
1039 | ERR (", or ] expected while parsing array"); |
1265 | ERR (", or ] expected while parsing array"); |
1040 | |
1266 | |
1041 | ++dec->cur; |
1267 | ++dec->cur; |
|
|
1268 | |
|
|
1269 | decode_ws (dec); |
|
|
1270 | |
|
|
1271 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1272 | { |
|
|
1273 | ++dec->cur; |
|
|
1274 | break; |
|
|
1275 | } |
1042 | } |
1276 | } |
1043 | |
1277 | |
1044 | DEC_DEC_DEPTH; |
1278 | DEC_DEC_DEPTH; |
1045 | return newRV_noinc ((SV *)av); |
1279 | return newRV_noinc ((SV *)av); |
1046 | |
1280 | |
… | |
… | |
1062 | if (*dec->cur == '}') |
1296 | if (*dec->cur == '}') |
1063 | ++dec->cur; |
1297 | ++dec->cur; |
1064 | else |
1298 | else |
1065 | for (;;) |
1299 | for (;;) |
1066 | { |
1300 | { |
1067 | decode_ws (dec); EXPECT_CH ('"'); |
1301 | EXPECT_CH ('"'); |
1068 | |
1302 | |
1069 | // heuristic: assume that |
1303 | // heuristic: assume that |
1070 | // a) decode_str + hv_store_ent are abysmally slow. |
1304 | // a) decode_str + hv_store_ent are abysmally slow. |
1071 | // b) most hash keys are short, simple ascii text. |
1305 | // b) most hash keys are short, simple ascii text. |
1072 | // => try to "fast-match" such strings to avoid |
1306 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1076 | char *p = dec->cur; |
1310 | char *p = dec->cur; |
1077 | char *e = p + 24; // only try up to 24 bytes |
1311 | char *e = p + 24; // only try up to 24 bytes |
1078 | |
1312 | |
1079 | for (;;) |
1313 | for (;;) |
1080 | { |
1314 | { |
1081 | // the >= 0x80 is true on most architectures |
1315 | // the >= 0x80 is false on most architectures |
1082 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1316 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1083 | { |
1317 | { |
1084 | // slow path, back up and use decode_str |
1318 | // slow path, back up and use decode_str |
1085 | SV *key = decode_str (dec); |
1319 | SV *key = decode_str (dec); |
1086 | if (!key) |
1320 | if (!key) |
1087 | goto fail; |
1321 | goto fail; |
1088 | |
1322 | |
1089 | decode_ws (dec); EXPECT_CH (':'); |
1323 | decode_ws (dec); EXPECT_CH (':'); |
1090 | |
1324 | |
|
|
1325 | decode_ws (dec); |
1091 | value = decode_sv (dec); |
1326 | value = decode_sv (dec); |
1092 | if (!value) |
1327 | if (!value) |
1093 | { |
1328 | { |
1094 | SvREFCNT_dec (key); |
1329 | SvREFCNT_dec (key); |
1095 | goto fail; |
1330 | goto fail; |
… | |
… | |
1107 | int len = p - key; |
1342 | int len = p - key; |
1108 | dec->cur = p + 1; |
1343 | dec->cur = p + 1; |
1109 | |
1344 | |
1110 | decode_ws (dec); EXPECT_CH (':'); |
1345 | decode_ws (dec); EXPECT_CH (':'); |
1111 | |
1346 | |
|
|
1347 | decode_ws (dec); |
1112 | value = decode_sv (dec); |
1348 | value = decode_sv (dec); |
1113 | if (!value) |
1349 | if (!value) |
1114 | goto fail; |
1350 | goto fail; |
1115 | |
1351 | |
1116 | hv_store (hv, key, len, value, 0); |
1352 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1132 | |
1368 | |
1133 | if (*dec->cur != ',') |
1369 | if (*dec->cur != ',') |
1134 | ERR (", or } expected while parsing object/hash"); |
1370 | ERR (", or } expected while parsing object/hash"); |
1135 | |
1371 | |
1136 | ++dec->cur; |
1372 | ++dec->cur; |
|
|
1373 | |
|
|
1374 | decode_ws (dec); |
|
|
1375 | |
|
|
1376 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1377 | { |
|
|
1378 | ++dec->cur; |
|
|
1379 | break; |
|
|
1380 | } |
1137 | } |
1381 | } |
1138 | |
1382 | |
1139 | DEC_DEC_DEPTH; |
1383 | DEC_DEC_DEPTH; |
1140 | sv = newRV_noinc ((SV *)hv); |
1384 | sv = newRV_noinc ((SV *)hv); |
1141 | |
1385 | |
… | |
… | |
1159 | dSP; |
1403 | dSP; |
1160 | int count; |
1404 | int count; |
1161 | |
1405 | |
1162 | ENTER; SAVETMPS; PUSHMARK (SP); |
1406 | ENTER; SAVETMPS; PUSHMARK (SP); |
1163 | XPUSHs (HeVAL (he)); |
1407 | XPUSHs (HeVAL (he)); |
|
|
1408 | sv_2mortal (sv); |
1164 | |
1409 | |
1165 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1410 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1166 | |
1411 | |
1167 | if (count == 1) |
1412 | if (count == 1) |
1168 | { |
1413 | { |
1169 | sv = newSVsv (POPs); |
1414 | sv = newSVsv (POPs); |
1170 | FREETMPS; LEAVE; |
1415 | FREETMPS; LEAVE; |
1171 | return sv; |
1416 | return sv; |
1172 | } |
1417 | } |
1173 | |
1418 | |
|
|
1419 | SvREFCNT_inc (sv); |
1174 | FREETMPS; LEAVE; |
1420 | FREETMPS; LEAVE; |
1175 | } |
1421 | } |
1176 | } |
1422 | } |
1177 | |
1423 | |
1178 | if (dec->json.cb_object) |
1424 | if (dec->json.cb_object) |
… | |
… | |
1206 | } |
1452 | } |
1207 | |
1453 | |
1208 | static SV * |
1454 | static SV * |
1209 | decode_sv (dec_t *dec) |
1455 | decode_sv (dec_t *dec) |
1210 | { |
1456 | { |
1211 | decode_ws (dec); |
|
|
1212 | |
|
|
1213 | // the beauty of JSON: you need exactly one character lookahead |
1457 | // the beauty of JSON: you need exactly one character lookahead |
1214 | // to parse anything. |
1458 | // to parse everything. |
1215 | switch (*dec->cur) |
1459 | switch (*dec->cur) |
1216 | { |
1460 | { |
1217 | case '"': ++dec->cur; return decode_str (dec); |
1461 | case '"': ++dec->cur; return decode_str (dec); |
1218 | case '[': ++dec->cur; return decode_av (dec); |
1462 | case '[': ++dec->cur; return decode_av (dec); |
1219 | case '{': ++dec->cur; return decode_hv (dec); |
1463 | case '{': ++dec->cur; return decode_hv (dec); |
1220 | |
1464 | |
1221 | case '-': |
1465 | case '-': |
1222 | case '0': case '1': case '2': case '3': case '4': |
1466 | case '0': case '1': case '2': case '3': case '4': |
1223 | case '5': case '6': case '7': case '8': case '9': |
1467 | case '5': case '6': case '7': case '8': case '9': |
1224 | return decode_num (dec); |
1468 | return decode_num (dec); |
1225 | |
1469 | |
1226 | case 't': |
1470 | case 't': |
1227 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1471 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1228 | { |
1472 | { |
1229 | dec->cur += 4; |
1473 | dec->cur += 4; |
|
|
1474 | #if JSON_SLOW |
|
|
1475 | json_true = get_bool ("JSON::XS::true"); |
|
|
1476 | #endif |
1230 | return SvREFCNT_inc (json_true); |
1477 | return newSVsv (json_true); |
1231 | } |
1478 | } |
1232 | else |
1479 | else |
1233 | ERR ("'true' expected"); |
1480 | ERR ("'true' expected"); |
1234 | |
1481 | |
1235 | break; |
1482 | break; |
1236 | |
1483 | |
1237 | case 'f': |
1484 | case 'f': |
1238 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1485 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1239 | { |
1486 | { |
1240 | dec->cur += 5; |
1487 | dec->cur += 5; |
|
|
1488 | #if JSON_SLOW |
|
|
1489 | json_false = get_bool ("JSON::XS::false"); |
|
|
1490 | #endif |
1241 | return SvREFCNT_inc (json_false); |
1491 | return newSVsv (json_false); |
1242 | } |
1492 | } |
1243 | else |
1493 | else |
1244 | ERR ("'false' expected"); |
1494 | ERR ("'false' expected"); |
1245 | |
1495 | |
1246 | break; |
1496 | break; |
… | |
… | |
1264 | fail: |
1514 | fail: |
1265 | return 0; |
1515 | return 0; |
1266 | } |
1516 | } |
1267 | |
1517 | |
1268 | static SV * |
1518 | static SV * |
1269 | decode_json (SV *string, JSON *json, UV *offset_return) |
1519 | decode_json (SV *string, JSON *json, char **offset_return) |
1270 | { |
1520 | { |
1271 | dec_t dec; |
1521 | dec_t dec; |
1272 | UV offset; |
|
|
1273 | SV *sv; |
1522 | SV *sv; |
1274 | |
1523 | |
|
|
1524 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1525 | * will perl ignore the magic in subsequent accesses |
|
|
1526 | */ |
1275 | SvGETMAGIC (string); |
1527 | /*SvGETMAGIC (string);*/ |
|
|
1528 | if (SvMAGICAL (string)) |
|
|
1529 | string = sv_2mortal (newSVsv (string)); |
|
|
1530 | |
1276 | SvUPGRADE (string, SVt_PV); |
1531 | SvUPGRADE (string, SVt_PV); |
1277 | |
1532 | |
1278 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1533 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1534 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1535 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1536 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1537 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1538 | * and hope for the best. |
|
|
1539 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1540 | * assertion business is seriously broken, try yet another workaround |
|
|
1541 | * for the broken -DDEBUGGING. |
|
|
1542 | */ |
|
|
1543 | { |
|
|
1544 | #ifdef DEBUGGING |
|
|
1545 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1546 | #else |
|
|
1547 | STRLEN offset = SvCUR (string); |
|
|
1548 | #endif |
|
|
1549 | |
|
|
1550 | if (offset > json->max_size && json->max_size) |
1279 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1551 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1280 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1552 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1553 | } |
1281 | |
1554 | |
1282 | if (json->flags & F_UTF8) |
1555 | if (DECODE_WANTS_OCTETS (json)) |
1283 | sv_utf8_downgrade (string, 0); |
1556 | sv_utf8_downgrade (string, 0); |
1284 | else |
1557 | else |
1285 | sv_utf8_upgrade (string); |
1558 | sv_utf8_upgrade (string); |
1286 | |
1559 | |
1287 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1560 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1288 | |
1561 | |
1289 | dec.json = *json; |
1562 | dec.json = *json; |
1290 | dec.cur = SvPVX (string); |
1563 | dec.cur = SvPVX (string); |
1291 | dec.end = SvEND (string); |
1564 | dec.end = SvEND (string); |
1292 | dec.err = 0; |
1565 | dec.err = 0; |
1293 | dec.depth = 0; |
1566 | dec.depth = 0; |
1294 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1295 | |
1567 | |
1296 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1568 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1297 | dec.json.flags |= F_HOOK; |
1569 | dec.json.flags |= F_HOOK; |
1298 | |
1570 | |
1299 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1571 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1572 | |
|
|
1573 | decode_ws (&dec); |
1300 | sv = decode_sv (&dec); |
1574 | sv = decode_sv (&dec); |
|
|
1575 | |
|
|
1576 | if (offset_return) |
|
|
1577 | *offset_return = dec.cur; |
1301 | |
1578 | |
1302 | if (!(offset_return || !sv)) |
1579 | if (!(offset_return || !sv)) |
1303 | { |
1580 | { |
1304 | // check for trailing garbage |
1581 | // check for trailing garbage |
1305 | decode_ws (&dec); |
1582 | decode_ws (&dec); |
… | |
… | |
1308 | { |
1585 | { |
1309 | dec.err = "garbage after JSON object"; |
1586 | dec.err = "garbage after JSON object"; |
1310 | SvREFCNT_dec (sv); |
1587 | SvREFCNT_dec (sv); |
1311 | sv = 0; |
1588 | sv = 0; |
1312 | } |
1589 | } |
1313 | } |
|
|
1314 | |
|
|
1315 | if (offset_return || !sv) |
|
|
1316 | { |
|
|
1317 | offset = dec.json.flags & F_UTF8 |
|
|
1318 | ? dec.cur - SvPVX (string) |
|
|
1319 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1320 | |
|
|
1321 | if (offset_return) |
|
|
1322 | *offset_return = offset; |
|
|
1323 | } |
1590 | } |
1324 | |
1591 | |
1325 | if (!sv) |
1592 | if (!sv) |
1326 | { |
1593 | { |
1327 | SV *uni = sv_newmortal (); |
1594 | SV *uni = sv_newmortal (); |
… | |
… | |
1333 | SAVEVPTR (PL_curcop); |
1600 | SAVEVPTR (PL_curcop); |
1334 | PL_curcop = &cop; |
1601 | PL_curcop = &cop; |
1335 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1602 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1336 | LEAVE; |
1603 | LEAVE; |
1337 | |
1604 | |
1338 | croak ("%s, at character offset %d [\"%s\"]", |
1605 | croak ("%s, at character offset %d (before \"%s\")", |
1339 | dec.err, |
1606 | dec.err, |
1340 | (int)offset, |
1607 | ptr_to_index (string, dec.cur), |
1341 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1608 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1342 | } |
1609 | } |
1343 | |
1610 | |
1344 | sv = sv_2mortal (sv); |
1611 | sv = sv_2mortal (sv); |
1345 | |
1612 | |
1346 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1613 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1347 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1614 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1348 | |
1615 | |
1349 | return sv; |
1616 | return sv; |
|
|
1617 | } |
|
|
1618 | |
|
|
1619 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1620 | // incremental parser |
|
|
1621 | |
|
|
1622 | static void |
|
|
1623 | incr_parse (JSON *self) |
|
|
1624 | { |
|
|
1625 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1626 | |
|
|
1627 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1628 | // but this would make it slower, so... |
|
|
1629 | |
|
|
1630 | for (;;) |
|
|
1631 | { |
|
|
1632 | //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 |
|
|
1633 | switch (self->incr_mode) |
|
|
1634 | { |
|
|
1635 | // only used for initial whitespace skipping |
|
|
1636 | case INCR_M_WS: |
|
|
1637 | for (;;) |
|
|
1638 | { |
|
|
1639 | if (*p > 0x20) |
|
|
1640 | { |
|
|
1641 | if (*p == '#') |
|
|
1642 | { |
|
|
1643 | self->incr_mode = INCR_M_C0; |
|
|
1644 | goto incr_m_c; |
|
|
1645 | } |
|
|
1646 | else |
|
|
1647 | { |
|
|
1648 | self->incr_mode = INCR_M_JSON; |
|
|
1649 | goto incr_m_json; |
|
|
1650 | } |
|
|
1651 | } |
|
|
1652 | else if (!*p) |
|
|
1653 | goto interrupt; |
|
|
1654 | |
|
|
1655 | ++p; |
|
|
1656 | } |
|
|
1657 | |
|
|
1658 | // skip a single char inside a string (for \\-processing) |
|
|
1659 | case INCR_M_BS: |
|
|
1660 | if (!*p) |
|
|
1661 | goto interrupt; |
|
|
1662 | |
|
|
1663 | ++p; |
|
|
1664 | self->incr_mode = INCR_M_STR; |
|
|
1665 | goto incr_m_str; |
|
|
1666 | |
|
|
1667 | // inside #-style comments |
|
|
1668 | case INCR_M_C0: |
|
|
1669 | case INCR_M_C1: |
|
|
1670 | incr_m_c: |
|
|
1671 | for (;;) |
|
|
1672 | { |
|
|
1673 | if (*p == '\n') |
|
|
1674 | { |
|
|
1675 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1676 | break; |
|
|
1677 | } |
|
|
1678 | else if (!*p) |
|
|
1679 | goto interrupt; |
|
|
1680 | |
|
|
1681 | ++p; |
|
|
1682 | } |
|
|
1683 | |
|
|
1684 | break; |
|
|
1685 | |
|
|
1686 | // inside a string |
|
|
1687 | case INCR_M_STR: |
|
|
1688 | incr_m_str: |
|
|
1689 | for (;;) |
|
|
1690 | { |
|
|
1691 | if (*p == '"') |
|
|
1692 | { |
|
|
1693 | ++p; |
|
|
1694 | self->incr_mode = INCR_M_JSON; |
|
|
1695 | |
|
|
1696 | if (!self->incr_nest) |
|
|
1697 | goto interrupt; |
|
|
1698 | |
|
|
1699 | goto incr_m_json; |
|
|
1700 | } |
|
|
1701 | else if (*p == '\\') |
|
|
1702 | { |
|
|
1703 | ++p; // "virtually" consumes character after \ |
|
|
1704 | |
|
|
1705 | if (!*p) // if at end of string we have to switch modes |
|
|
1706 | { |
|
|
1707 | self->incr_mode = INCR_M_BS; |
|
|
1708 | goto interrupt; |
|
|
1709 | } |
|
|
1710 | } |
|
|
1711 | else if (!*p) |
|
|
1712 | goto interrupt; |
|
|
1713 | |
|
|
1714 | ++p; |
|
|
1715 | } |
|
|
1716 | |
|
|
1717 | // after initial ws, outside string |
|
|
1718 | case INCR_M_JSON: |
|
|
1719 | incr_m_json: |
|
|
1720 | for (;;) |
|
|
1721 | { |
|
|
1722 | switch (*p++) |
|
|
1723 | { |
|
|
1724 | case 0: |
|
|
1725 | --p; |
|
|
1726 | goto interrupt; |
|
|
1727 | |
|
|
1728 | case 0x09: |
|
|
1729 | case 0x0a: |
|
|
1730 | case 0x0d: |
|
|
1731 | case 0x20: |
|
|
1732 | if (!self->incr_nest) |
|
|
1733 | { |
|
|
1734 | --p; // do not eat the whitespace, let the next round do it |
|
|
1735 | goto interrupt; |
|
|
1736 | } |
|
|
1737 | break; |
|
|
1738 | |
|
|
1739 | case '"': |
|
|
1740 | self->incr_mode = INCR_M_STR; |
|
|
1741 | goto incr_m_str; |
|
|
1742 | |
|
|
1743 | case '[': |
|
|
1744 | case '{': |
|
|
1745 | if (++self->incr_nest > self->max_depth) |
|
|
1746 | croak (ERR_NESTING_EXCEEDED); |
|
|
1747 | break; |
|
|
1748 | |
|
|
1749 | case ']': |
|
|
1750 | case '}': |
|
|
1751 | if (--self->incr_nest <= 0) |
|
|
1752 | goto interrupt; |
|
|
1753 | break; |
|
|
1754 | |
|
|
1755 | case '#': |
|
|
1756 | self->incr_mode = INCR_M_C1; |
|
|
1757 | goto incr_m_c; |
|
|
1758 | } |
|
|
1759 | } |
|
|
1760 | } |
|
|
1761 | |
|
|
1762 | modechange: |
|
|
1763 | ; |
|
|
1764 | } |
|
|
1765 | |
|
|
1766 | interrupt: |
|
|
1767 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1768 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1769 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1350 | } |
1770 | } |
1351 | |
1771 | |
1352 | ///////////////////////////////////////////////////////////////////////////// |
1772 | ///////////////////////////////////////////////////////////////////////////// |
1353 | // XS interface functions |
1773 | // XS interface functions |
1354 | |
1774 | |
… | |
… | |
1366 | : -1; |
1786 | : -1; |
1367 | |
1787 | |
1368 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1788 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1369 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1789 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1370 | |
1790 | |
1371 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1791 | json_true = get_bool ("JSON::XS::true"); |
1372 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1792 | json_false = get_bool ("JSON::XS::false"); |
|
|
1793 | |
|
|
1794 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1373 | } |
1795 | } |
1374 | |
1796 | |
1375 | PROTOTYPES: DISABLE |
1797 | PROTOTYPES: DISABLE |
1376 | |
1798 | |
1377 | void CLONE (...) |
1799 | void CLONE (...) |
… | |
… | |
1382 | void new (char *klass) |
1804 | void new (char *klass) |
1383 | PPCODE: |
1805 | PPCODE: |
1384 | { |
1806 | { |
1385 | SV *pv = NEWSV (0, sizeof (JSON)); |
1807 | SV *pv = NEWSV (0, sizeof (JSON)); |
1386 | SvPOK_only (pv); |
1808 | SvPOK_only (pv); |
1387 | Zero (SvPVX (pv), 1, JSON); |
1809 | json_init ((JSON *)SvPVX (pv)); |
1388 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1810 | XPUSHs (sv_2mortal (sv_bless ( |
1389 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); |
1811 | newRV_noinc (pv), |
|
|
1812 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1813 | ))); |
1390 | } |
1814 | } |
1391 | |
1815 | |
1392 | void ascii (JSON *self, int enable = 1) |
1816 | void ascii (JSON *self, int enable = 1) |
1393 | ALIAS: |
1817 | ALIAS: |
1394 | ascii = F_ASCII |
1818 | ascii = F_ASCII |
… | |
… | |
1401 | pretty = F_PRETTY |
1825 | pretty = F_PRETTY |
1402 | allow_nonref = F_ALLOW_NONREF |
1826 | allow_nonref = F_ALLOW_NONREF |
1403 | shrink = F_SHRINK |
1827 | shrink = F_SHRINK |
1404 | allow_blessed = F_ALLOW_BLESSED |
1828 | allow_blessed = F_ALLOW_BLESSED |
1405 | convert_blessed = F_CONV_BLESSED |
1829 | convert_blessed = F_CONV_BLESSED |
|
|
1830 | relaxed = F_RELAXED |
|
|
1831 | allow_unknown = F_ALLOW_UNKNOWN |
1406 | PPCODE: |
1832 | PPCODE: |
1407 | { |
1833 | { |
1408 | if (enable) |
1834 | if (enable) |
1409 | self->flags |= ix; |
1835 | self->flags |= ix; |
1410 | else |
1836 | else |
1411 | self->flags &= ~ix; |
1837 | self->flags &= ~ix; |
1412 | |
1838 | |
1413 | XPUSHs (ST (0)); |
1839 | XPUSHs (ST (0)); |
1414 | } |
1840 | } |
1415 | |
1841 | |
1416 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1842 | void get_ascii (JSON *self) |
|
|
1843 | ALIAS: |
|
|
1844 | get_ascii = F_ASCII |
|
|
1845 | get_latin1 = F_LATIN1 |
|
|
1846 | get_utf8 = F_UTF8 |
|
|
1847 | get_indent = F_INDENT |
|
|
1848 | get_canonical = F_CANONICAL |
|
|
1849 | get_space_before = F_SPACE_BEFORE |
|
|
1850 | get_space_after = F_SPACE_AFTER |
|
|
1851 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1852 | get_shrink = F_SHRINK |
|
|
1853 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1854 | get_convert_blessed = F_CONV_BLESSED |
|
|
1855 | get_relaxed = F_RELAXED |
|
|
1856 | get_allow_unknown = F_ALLOW_UNKNOWN |
1417 | PPCODE: |
1857 | PPCODE: |
1418 | { |
1858 | XPUSHs (boolSV (self->flags & ix)); |
1419 | UV log2 = 0; |
|
|
1420 | |
1859 | |
1421 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1860 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1422 | |
1861 | PPCODE: |
1423 | while ((1UL << log2) < max_depth) |
1862 | self->max_depth = max_depth; |
1424 | ++log2; |
|
|
1425 | |
|
|
1426 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1427 | |
|
|
1428 | XPUSHs (ST (0)); |
1863 | XPUSHs (ST (0)); |
1429 | } |
|
|
1430 | |
1864 | |
|
|
1865 | U32 get_max_depth (JSON *self) |
|
|
1866 | CODE: |
|
|
1867 | RETVAL = self->max_depth; |
|
|
1868 | OUTPUT: |
|
|
1869 | RETVAL |
|
|
1870 | |
1431 | void max_size (JSON *self, UV max_size = 0) |
1871 | void max_size (JSON *self, U32 max_size = 0) |
1432 | PPCODE: |
1872 | PPCODE: |
1433 | { |
1873 | self->max_size = max_size; |
1434 | UV log2 = 0; |
|
|
1435 | |
|
|
1436 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1437 | if (max_size == 1) max_size = 2; |
|
|
1438 | |
|
|
1439 | while ((1UL << log2) < max_size) |
|
|
1440 | ++log2; |
|
|
1441 | |
|
|
1442 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1443 | |
|
|
1444 | XPUSHs (ST (0)); |
1874 | XPUSHs (ST (0)); |
1445 | } |
1875 | |
|
|
1876 | int get_max_size (JSON *self) |
|
|
1877 | CODE: |
|
|
1878 | RETVAL = self->max_size; |
|
|
1879 | OUTPUT: |
|
|
1880 | RETVAL |
1446 | |
1881 | |
1447 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1882 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1448 | PPCODE: |
1883 | PPCODE: |
1449 | { |
1884 | { |
1450 | SvREFCNT_dec (self->cb_object); |
1885 | SvREFCNT_dec (self->cb_object); |
… | |
… | |
1484 | XPUSHs (decode_json (jsonstr, self, 0)); |
1919 | XPUSHs (decode_json (jsonstr, self, 0)); |
1485 | |
1920 | |
1486 | void decode_prefix (JSON *self, SV *jsonstr) |
1921 | void decode_prefix (JSON *self, SV *jsonstr) |
1487 | PPCODE: |
1922 | PPCODE: |
1488 | { |
1923 | { |
1489 | UV offset; |
1924 | char *offset; |
1490 | EXTEND (SP, 2); |
1925 | EXTEND (SP, 2); |
1491 | PUSHs (decode_json (jsonstr, self, &offset)); |
1926 | PUSHs (decode_json (jsonstr, self, &offset)); |
1492 | PUSHs (sv_2mortal (newSVuv (offset))); |
1927 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1928 | } |
|
|
1929 | |
|
|
1930 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1931 | PPCODE: |
|
|
1932 | { |
|
|
1933 | if (!self->incr_text) |
|
|
1934 | self->incr_text = newSVpvn ("", 0); |
|
|
1935 | |
|
|
1936 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
1937 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
1938 | if (DECODE_WANTS_OCTETS (self)) |
|
|
1939 | { |
|
|
1940 | if (self->incr_pos) |
|
|
1941 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
1942 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
1943 | |
|
|
1944 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
1945 | } |
|
|
1946 | else |
|
|
1947 | { |
|
|
1948 | sv_utf8_upgrade (self->incr_text); |
|
|
1949 | |
|
|
1950 | if (self->incr_pos) |
|
|
1951 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1952 | - (U8 *)SvPVX (self->incr_text); |
|
|
1953 | } |
|
|
1954 | |
|
|
1955 | // append data, if any |
|
|
1956 | if (jsonstr) |
|
|
1957 | { |
|
|
1958 | /* make sure both strings have same encoding */ |
|
|
1959 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
|
|
1960 | if (SvUTF8 (jsonstr)) |
|
|
1961 | sv_utf8_downgrade (jsonstr, 0); |
|
|
1962 | else |
|
|
1963 | sv_utf8_upgrade (jsonstr); |
|
|
1964 | |
|
|
1965 | /* and then just blindly append */ |
|
|
1966 | { |
|
|
1967 | STRLEN len; |
|
|
1968 | const char *str = SvPV (jsonstr, len); |
|
|
1969 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1970 | |
|
|
1971 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1972 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1973 | |
|
|
1974 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1975 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1976 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1977 | } |
|
|
1978 | } |
|
|
1979 | |
|
|
1980 | if (GIMME_V != G_VOID) |
|
|
1981 | do |
|
|
1982 | { |
|
|
1983 | char *offset; |
|
|
1984 | |
|
|
1985 | if (!INCR_DONE (self)) |
|
|
1986 | { |
|
|
1987 | incr_parse (self); |
|
|
1988 | |
|
|
1989 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1990 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1991 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1992 | |
|
|
1993 | if (!INCR_DONE (self)) |
|
|
1994 | break; |
|
|
1995 | } |
|
|
1996 | |
|
|
1997 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1998 | |
|
|
1999 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
2000 | self->incr_nest = 0; |
|
|
2001 | self->incr_mode = 0; |
|
|
2002 | |
|
|
2003 | sv_chop (self->incr_text, offset); |
|
|
2004 | } |
|
|
2005 | while (GIMME_V == G_ARRAY); |
|
|
2006 | } |
|
|
2007 | |
|
|
2008 | SV *incr_text (JSON *self) |
|
|
2009 | ATTRS: lvalue |
|
|
2010 | CODE: |
|
|
2011 | { |
|
|
2012 | if (self->incr_pos) |
|
|
2013 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
2014 | |
|
|
2015 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2016 | } |
|
|
2017 | OUTPUT: |
|
|
2018 | RETVAL |
|
|
2019 | |
|
|
2020 | void incr_skip (JSON *self) |
|
|
2021 | CODE: |
|
|
2022 | { |
|
|
2023 | if (self->incr_pos) |
|
|
2024 | { |
|
|
2025 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2026 | self->incr_pos = 0; |
|
|
2027 | self->incr_nest = 0; |
|
|
2028 | self->incr_mode = 0; |
|
|
2029 | } |
|
|
2030 | } |
|
|
2031 | |
|
|
2032 | void incr_reset (JSON *self) |
|
|
2033 | CODE: |
|
|
2034 | { |
|
|
2035 | SvREFCNT_dec (self->incr_text); |
|
|
2036 | self->incr_text = 0; |
|
|
2037 | self->incr_pos = 0; |
|
|
2038 | self->incr_nest = 0; |
|
|
2039 | self->incr_mode = 0; |
1493 | } |
2040 | } |
1494 | |
2041 | |
1495 | void DESTROY (JSON *self) |
2042 | void DESTROY (JSON *self) |
1496 | CODE: |
2043 | CODE: |
1497 | SvREFCNT_dec (self->cb_sk_object); |
2044 | SvREFCNT_dec (self->cb_sk_object); |
1498 | SvREFCNT_dec (self->cb_object); |
2045 | SvREFCNT_dec (self->cb_object); |
|
|
2046 | SvREFCNT_dec (self->incr_text); |
1499 | |
2047 | |
1500 | PROTOTYPES: ENABLE |
2048 | PROTOTYPES: ENABLE |
1501 | |
2049 | |
1502 | void to_json (SV *scalar) |
2050 | void encode_json (SV *scalar) |
|
|
2051 | ALIAS: |
|
|
2052 | to_json_ = 0 |
|
|
2053 | encode_json = F_UTF8 |
1503 | PPCODE: |
2054 | PPCODE: |
1504 | { |
2055 | { |
1505 | JSON json = { F_DEFAULT | F_UTF8 }; |
2056 | JSON json; |
|
|
2057 | json_init (&json); |
|
|
2058 | json.flags |= ix; |
1506 | XPUSHs (encode_json (scalar, &json)); |
2059 | XPUSHs (encode_json (scalar, &json)); |
1507 | } |
2060 | } |
1508 | |
2061 | |
1509 | void from_json (SV *jsonstr) |
2062 | void decode_json (SV *jsonstr) |
|
|
2063 | ALIAS: |
|
|
2064 | from_json_ = 0 |
|
|
2065 | decode_json = F_UTF8 |
1510 | PPCODE: |
2066 | PPCODE: |
1511 | { |
2067 | { |
1512 | JSON json = { F_DEFAULT | F_UTF8 }; |
2068 | JSON json; |
|
|
2069 | json_init (&json); |
|
|
2070 | json.flags |= ix; |
1513 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2071 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1514 | } |
2072 | } |
1515 | |
2073 | |