1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
8 | #include "stdio.h" |
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
|
|
10 | #include <float.h> |
9 | |
11 | |
10 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
11 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
12 | #endif |
14 | #endif |
13 | |
15 | |
14 | // 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 |
15 | // guarentees, though. if it breaks, you get to keep the pieces. |
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
16 | #ifndef UTF8_MAXBYTES |
18 | #ifndef UTF8_MAXBYTES |
17 | # define UTF8_MAXBYTES 13 |
19 | # define UTF8_MAXBYTES 13 |
18 | #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) |
19 | |
24 | |
20 | #define F_ASCII 0x00000001UL |
25 | #define F_ASCII 0x00000001UL |
21 | #define F_LATIN1 0x00000002UL |
26 | #define F_LATIN1 0x00000002UL |
22 | #define F_UTF8 0x00000004UL |
27 | #define F_UTF8 0x00000004UL |
23 | #define F_INDENT 0x00000008UL |
28 | #define F_INDENT 0x00000008UL |
… | |
… | |
26 | #define F_SPACE_AFTER 0x00000040UL |
31 | #define F_SPACE_AFTER 0x00000040UL |
27 | #define F_ALLOW_NONREF 0x00000100UL |
32 | #define F_ALLOW_NONREF 0x00000100UL |
28 | #define F_SHRINK 0x00000200UL |
33 | #define F_SHRINK 0x00000200UL |
29 | #define F_ALLOW_BLESSED 0x00000400UL |
34 | #define F_ALLOW_BLESSED 0x00000400UL |
30 | #define F_CONV_BLESSED 0x00000800UL |
35 | #define F_CONV_BLESSED 0x00000800UL |
31 | #define F_MAXDEPTH 0xf8000000UL |
|
|
32 | #define S_MAXDEPTH 27 |
|
|
33 | #define F_MAXSIZE 0x01f00000UL |
36 | #define F_RELAXED 0x00001000UL |
34 | #define S_MAXSIZE 20 |
37 | #define F_ALLOW_UNKNOWN 0x00002000UL |
35 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
36 | |
39 | |
37 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
|
|
38 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
|
|
39 | |
|
|
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
41 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
42 | |
41 | |
43 | #define INIT_SIZE 32 // initial scalar size to be allocated |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
44 | #define INDENT_STEP 3 // spaces per indentation level |
43 | #define INDENT_STEP 3 // spaces per indentation level |
45 | |
44 | |
46 | #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 |
47 | |
46 | |
48 | #define SB do { |
47 | #define SB do { |
49 | #define SE } while (0) |
48 | #define SE } while (0) |
50 | |
49 | |
51 | #if __GNUC__ >= 3 |
50 | #if __GNUC__ >= 3 |
52 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
51 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
53 | # define inline inline |
52 | # define INLINE static inline |
54 | #else |
53 | #else |
55 | # define expect(expr,value) (expr) |
54 | # define expect(expr,value) (expr) |
56 | # define inline static |
55 | # define INLINE static |
57 | #endif |
56 | #endif |
58 | |
57 | |
59 | #define expect_false(expr) expect ((expr) != 0, 0) |
58 | #define expect_false(expr) expect ((expr) != 0, 0) |
60 | #define expect_true(expr) expect ((expr) != 0, 1) |
59 | #define expect_true(expr) expect ((expr) != 0, 1) |
61 | |
60 | |
|
|
61 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
62 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
63 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
64 | |
|
|
65 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
66 | |
|
|
67 | #ifdef USE_ITHREADS |
|
|
68 | # define JSON_SLOW 1 |
|
|
69 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
70 | #else |
|
|
71 | # define JSON_SLOW 0 |
|
|
72 | # define JSON_STASH json_stash |
|
|
73 | #endif |
|
|
74 | |
62 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
75 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
63 | static SV *json_true, *json_false; |
76 | static SV *json_true, *json_false; |
64 | |
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 | |
65 | typedef struct { |
89 | typedef struct { |
66 | U32 flags; |
90 | U32 flags; |
|
|
91 | U32 max_depth; |
|
|
92 | STRLEN max_size; |
|
|
93 | |
67 | SV *cb_object; |
94 | SV *cb_object; |
68 | 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; |
69 | } JSON; |
102 | } JSON; |
|
|
103 | |
|
|
104 | INLINE void |
|
|
105 | json_init (JSON *json) |
|
|
106 | { |
|
|
107 | Zero (json, 1, JSON); |
|
|
108 | json->max_depth = 512; |
|
|
109 | } |
70 | |
110 | |
71 | ///////////////////////////////////////////////////////////////////////////// |
111 | ///////////////////////////////////////////////////////////////////////////// |
72 | // utility functions |
112 | // utility functions |
73 | |
113 | |
74 | 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 |
75 | shrink (SV *sv) |
126 | shrink (SV *sv) |
76 | { |
127 | { |
77 | sv_utf8_downgrade (sv, 1); |
128 | sv_utf8_downgrade (sv, 1); |
|
|
129 | |
78 | if (SvLEN (sv) > SvCUR (sv) + 1) |
130 | if (SvLEN (sv) > SvCUR (sv) + 1) |
79 | { |
131 | { |
80 | #ifdef SvPV_shrink_to_cur |
132 | #ifdef SvPV_shrink_to_cur |
81 | SvPV_shrink_to_cur (sv); |
133 | SvPV_shrink_to_cur (sv); |
82 | #elif defined (SvPV_renew) |
134 | #elif defined (SvPV_renew) |
… | |
… | |
88 | // 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 |
89 | // case of an error. |
141 | // case of an error. |
90 | // we special-case "safe" characters from U+80 .. U+7FF, |
142 | // we special-case "safe" characters from U+80 .. U+7FF, |
91 | // but use the very good perl function to parse anything else. |
143 | // but use the very good perl function to parse anything else. |
92 | // note that we never call this function for a ascii codepoints |
144 | // note that we never call this function for a ascii codepoints |
93 | inline UV |
145 | INLINE UV |
94 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
146 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
95 | { |
147 | { |
96 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
148 | if (expect_true (len >= 2 |
97 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
149 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
98 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
150 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
99 | { |
151 | { |
100 | *clen = 2; |
152 | *clen = 2; |
101 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
153 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
102 | } |
154 | } |
103 | else |
155 | else |
104 | { |
156 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
105 | *clen = (STRLEN)-1; |
157 | } |
106 | 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, int maxdepth) |
|
|
198 | { |
|
|
199 | UV uaccum = 0; |
|
|
200 | int eaccum = 0; |
|
|
201 | |
|
|
202 | // if we recurse too deep, skip all remaining digits |
|
|
203 | // to avoid a stack overflow attack |
|
|
204 | if (expect_false (--maxdepth <= 0)) |
|
|
205 | while (((U8)*s - '0') < 10) |
|
|
206 | ++s; |
|
|
207 | |
|
|
208 | for (;;) |
107 | } |
209 | { |
108 | } |
210 | U8 dig = (U8)*s - '0'; |
109 | |
211 | |
|
|
212 | if (expect_false (dig >= 10)) |
|
|
213 | { |
|
|
214 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
215 | { |
|
|
216 | ++s; |
|
|
217 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
218 | } |
|
|
219 | else if ((dig | ' ') == 'e' - '0') |
|
|
220 | { |
|
|
221 | int exp2 = 0; |
|
|
222 | int neg = 0; |
|
|
223 | |
|
|
224 | ++s; |
|
|
225 | |
|
|
226 | if (*s == '-') |
|
|
227 | { |
|
|
228 | ++s; |
|
|
229 | neg = 1; |
|
|
230 | } |
|
|
231 | else if (*s == '+') |
|
|
232 | ++s; |
|
|
233 | |
|
|
234 | while ((dig = (U8)*s - '0') < 10) |
|
|
235 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
236 | |
|
|
237 | *expo += neg ? -exp2 : exp2; |
|
|
238 | } |
|
|
239 | |
|
|
240 | break; |
|
|
241 | } |
|
|
242 | |
|
|
243 | ++s; |
|
|
244 | |
|
|
245 | uaccum = uaccum * 10 + dig; |
|
|
246 | ++eaccum; |
|
|
247 | |
|
|
248 | // if we have too many digits, then recurse for more |
|
|
249 | // we actually do this for rather few digits |
|
|
250 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
251 | { |
|
|
252 | if (postdp) *expo -= eaccum; |
|
|
253 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
254 | if (postdp) *expo += eaccum; |
|
|
255 | |
|
|
256 | break; |
|
|
257 | } |
|
|
258 | } |
|
|
259 | |
|
|
260 | // this relies greatly on the quality of the pow () |
|
|
261 | // implementation of the platform, but a good |
|
|
262 | // implementation is hard to beat. |
|
|
263 | if (postdp) *expo -= eaccum; |
|
|
264 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
265 | *expo += eaccum; |
|
|
266 | } |
|
|
267 | |
|
|
268 | static NV |
|
|
269 | json_atof (const char *s) |
|
|
270 | { |
|
|
271 | NV accum = 0.; |
|
|
272 | int expo = 0; |
|
|
273 | int neg = 0; |
|
|
274 | |
|
|
275 | if (*s == '-') |
|
|
276 | { |
|
|
277 | ++s; |
|
|
278 | neg = 1; |
|
|
279 | } |
|
|
280 | |
|
|
281 | // a recursion depth of ten gives us >>500 bits |
|
|
282 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
283 | |
|
|
284 | return neg ? -accum : accum; |
|
|
285 | } |
110 | ///////////////////////////////////////////////////////////////////////////// |
286 | ///////////////////////////////////////////////////////////////////////////// |
111 | // encoder |
287 | // encoder |
112 | |
288 | |
113 | // structure used for encoding JSON |
289 | // structure used for encoding JSON |
114 | typedef struct |
290 | typedef struct |
… | |
… | |
116 | char *cur; // SvPVX (sv) + current output position |
292 | char *cur; // SvPVX (sv) + current output position |
117 | char *end; // SvEND (sv) |
293 | char *end; // SvEND (sv) |
118 | SV *sv; // result scalar |
294 | SV *sv; // result scalar |
119 | JSON json; |
295 | JSON json; |
120 | U32 indent; // indentation level |
296 | U32 indent; // indentation level |
121 | U32 maxdepth; // max. indentation/recursion level |
297 | UV limit; // escape character values >= this value when encoding |
122 | } enc_t; |
298 | } enc_t; |
123 | |
299 | |
124 | inline void |
300 | INLINE void |
125 | need (enc_t *enc, STRLEN len) |
301 | need (enc_t *enc, STRLEN len) |
126 | { |
302 | { |
127 | if (expect_false (enc->cur + len >= enc->end)) |
303 | if (expect_false (enc->cur + len >= enc->end)) |
128 | { |
304 | { |
129 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
305 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
130 | SvGROW (enc->sv, cur + len + 1); |
306 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
131 | enc->cur = SvPVX (enc->sv) + cur; |
307 | enc->cur = SvPVX (enc->sv) + cur; |
132 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
308 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
133 | } |
309 | } |
134 | } |
310 | } |
135 | |
311 | |
136 | inline void |
312 | INLINE void |
137 | encode_ch (enc_t *enc, char ch) |
313 | encode_ch (enc_t *enc, char ch) |
138 | { |
314 | { |
139 | need (enc, 1); |
315 | need (enc, 1); |
140 | *enc->cur++ = ch; |
316 | *enc->cur++ = ch; |
141 | } |
317 | } |
… | |
… | |
195 | { |
371 | { |
196 | uch = ch; |
372 | uch = ch; |
197 | clen = 1; |
373 | clen = 1; |
198 | } |
374 | } |
199 | |
375 | |
200 | if (uch > 0x10FFFFUL) |
376 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
201 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
202 | |
|
|
203 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
|
|
204 | { |
377 | { |
205 | if (uch > 0xFFFFUL) |
378 | if (uch >= 0x10000UL) |
206 | { |
379 | { |
|
|
380 | if (uch >= 0x110000UL) |
|
|
381 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
382 | |
207 | need (enc, len += 11); |
383 | need (enc, len += 11); |
208 | sprintf (enc->cur, "\\u%04x\\u%04x", |
384 | sprintf (enc->cur, "\\u%04x\\u%04x", |
209 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
385 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
210 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
386 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
211 | enc->cur += 12; |
387 | enc->cur += 12; |
212 | } |
388 | } |
213 | else |
389 | else |
214 | { |
390 | { |
215 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
216 | need (enc, len += 5); |
391 | need (enc, len += 5); |
217 | *enc->cur++ = '\\'; |
392 | *enc->cur++ = '\\'; |
218 | *enc->cur++ = 'u'; |
393 | *enc->cur++ = 'u'; |
219 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
394 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
220 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
395 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
221 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
396 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
222 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
397 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
223 | } |
398 | } |
224 | |
399 | |
225 | str += clen; |
400 | str += clen; |
226 | } |
401 | } |
227 | else if (enc->json.flags & F_LATIN1) |
402 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
239 | while (--clen); |
414 | while (--clen); |
240 | } |
415 | } |
241 | else |
416 | else |
242 | { |
417 | { |
243 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
418 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
244 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
419 | enc->cur = encode_utf8 (enc->cur, uch); |
245 | ++str; |
420 | ++str; |
246 | } |
421 | } |
247 | } |
422 | } |
248 | } |
423 | } |
249 | } |
424 | } |
250 | |
425 | |
251 | --len; |
426 | --len; |
252 | } |
427 | } |
253 | } |
428 | } |
254 | |
429 | |
255 | inline void |
430 | INLINE void |
256 | encode_indent (enc_t *enc) |
431 | encode_indent (enc_t *enc) |
257 | { |
432 | { |
258 | if (enc->json.flags & F_INDENT) |
433 | if (enc->json.flags & F_INDENT) |
259 | { |
434 | { |
260 | int spaces = enc->indent * INDENT_STEP; |
435 | int spaces = enc->indent * INDENT_STEP; |
… | |
… | |
263 | memset (enc->cur, ' ', spaces); |
438 | memset (enc->cur, ' ', spaces); |
264 | enc->cur += spaces; |
439 | enc->cur += spaces; |
265 | } |
440 | } |
266 | } |
441 | } |
267 | |
442 | |
268 | inline void |
443 | INLINE void |
269 | encode_space (enc_t *enc) |
444 | encode_space (enc_t *enc) |
270 | { |
445 | { |
271 | need (enc, 1); |
446 | need (enc, 1); |
272 | encode_ch (enc, ' '); |
447 | encode_ch (enc, ' '); |
273 | } |
448 | } |
274 | |
449 | |
275 | inline void |
450 | INLINE void |
276 | encode_nl (enc_t *enc) |
451 | encode_nl (enc_t *enc) |
277 | { |
452 | { |
278 | if (enc->json.flags & F_INDENT) |
453 | if (enc->json.flags & F_INDENT) |
279 | { |
454 | { |
280 | need (enc, 1); |
455 | need (enc, 1); |
281 | encode_ch (enc, '\n'); |
456 | encode_ch (enc, '\n'); |
282 | } |
457 | } |
283 | } |
458 | } |
284 | |
459 | |
285 | inline void |
460 | INLINE void |
286 | encode_comma (enc_t *enc) |
461 | encode_comma (enc_t *enc) |
287 | { |
462 | { |
288 | encode_ch (enc, ','); |
463 | encode_ch (enc, ','); |
289 | |
464 | |
290 | if (enc->json.flags & F_INDENT) |
465 | if (enc->json.flags & F_INDENT) |
… | |
… | |
298 | static void |
473 | static void |
299 | encode_av (enc_t *enc, AV *av) |
474 | encode_av (enc_t *enc, AV *av) |
300 | { |
475 | { |
301 | int i, len = av_len (av); |
476 | int i, len = av_len (av); |
302 | |
477 | |
303 | if (enc->indent >= enc->maxdepth) |
478 | if (enc->indent >= enc->json.max_depth) |
304 | croak ("data structure too deep (hit recursion limit)"); |
479 | croak (ERR_NESTING_EXCEEDED); |
305 | |
480 | |
306 | encode_ch (enc, '['); encode_nl (enc); |
481 | encode_ch (enc, '['); |
307 | ++enc->indent; |
482 | |
|
|
483 | if (len >= 0) |
|
|
484 | { |
|
|
485 | encode_nl (enc); ++enc->indent; |
308 | |
486 | |
309 | for (i = 0; i <= len; ++i) |
487 | for (i = 0; i <= len; ++i) |
310 | { |
488 | { |
|
|
489 | SV **svp = av_fetch (av, i, 0); |
|
|
490 | |
311 | encode_indent (enc); |
491 | encode_indent (enc); |
312 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
313 | |
492 | |
|
|
493 | if (svp) |
|
|
494 | encode_sv (enc, *svp); |
|
|
495 | else |
|
|
496 | encode_str (enc, "null", 4, 0); |
|
|
497 | |
314 | if (i < len) |
498 | if (i < len) |
315 | encode_comma (enc); |
499 | encode_comma (enc); |
316 | } |
500 | } |
317 | |
501 | |
|
|
502 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
503 | } |
|
|
504 | |
318 | encode_nl (enc); |
505 | encode_ch (enc, ']'); |
319 | |
|
|
320 | --enc->indent; |
|
|
321 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
322 | } |
506 | } |
323 | |
507 | |
324 | static void |
508 | static void |
325 | encode_he (enc_t *enc, HE *he) |
509 | encode_hk (enc_t *enc, HE *he) |
326 | { |
510 | { |
327 | encode_ch (enc, '"'); |
511 | encode_ch (enc, '"'); |
328 | |
512 | |
329 | if (HeKLEN (he) == HEf_SVKEY) |
513 | if (HeKLEN (he) == HEf_SVKEY) |
330 | { |
514 | { |
… | |
… | |
343 | encode_ch (enc, '"'); |
527 | encode_ch (enc, '"'); |
344 | |
528 | |
345 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
529 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
346 | encode_ch (enc, ':'); |
530 | encode_ch (enc, ':'); |
347 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
531 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
348 | encode_sv (enc, HeVAL (he)); |
|
|
349 | } |
532 | } |
350 | |
533 | |
351 | // compare hash entries, used when all keys are bytestrings |
534 | // compare hash entries, used when all keys are bytestrings |
352 | static int |
535 | static int |
353 | he_cmp_fast (const void *a_, const void *b_) |
536 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
358 | HE *b = *(HE **)b_; |
541 | HE *b = *(HE **)b_; |
359 | |
542 | |
360 | STRLEN la = HeKLEN (a); |
543 | STRLEN la = HeKLEN (a); |
361 | STRLEN lb = HeKLEN (b); |
544 | STRLEN lb = HeKLEN (b); |
362 | |
545 | |
363 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
546 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
364 | cmp = la - lb; |
547 | cmp = lb - la; |
365 | |
548 | |
366 | return cmp; |
549 | return cmp; |
367 | } |
550 | } |
368 | |
551 | |
369 | // compare hash entries, used when some keys are sv's or utf-x |
552 | // compare hash entries, used when some keys are sv's or utf-x |
370 | static int |
553 | static int |
371 | he_cmp_slow (const void *a, const void *b) |
554 | he_cmp_slow (const void *a, const void *b) |
372 | { |
555 | { |
373 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
556 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
374 | } |
557 | } |
375 | |
558 | |
376 | static void |
559 | static void |
377 | encode_hv (enc_t *enc, HV *hv) |
560 | encode_hv (enc_t *enc, HV *hv) |
378 | { |
561 | { |
379 | int count, i; |
562 | HE *he; |
380 | |
563 | |
381 | if (enc->indent >= enc->maxdepth) |
564 | if (enc->indent >= enc->json.max_depth) |
382 | croak ("data structure too deep (hit recursion limit)"); |
565 | croak (ERR_NESTING_EXCEEDED); |
383 | |
566 | |
384 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
567 | encode_ch (enc, '{'); |
385 | |
568 | |
386 | if ((count = hv_iterinit (hv))) |
|
|
387 | { |
|
|
388 | // for canonical output we have to sort by keys first |
569 | // for canonical output we have to sort by keys first |
389 | // actually, this is mostly due to the stupid so-called |
570 | // actually, this is mostly due to the stupid so-called |
390 | // security workaround added somewhere in 5.8.x. |
571 | // security workaround added somewhere in 5.8.x |
391 | // that randomises hash orderings |
572 | // that randomises hash orderings |
392 | if (enc->json.flags & F_CANONICAL) |
573 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
574 | { |
|
|
575 | int count = hv_iterinit (hv); |
|
|
576 | |
|
|
577 | if (SvMAGICAL (hv)) |
393 | { |
578 | { |
|
|
579 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
580 | // but I don't care for the speed of this special case. |
|
|
581 | // note also that we will run into undefined behaviour when the two iterations |
|
|
582 | // do not result in the same count, something I might care for in some later release. |
|
|
583 | |
|
|
584 | count = 0; |
|
|
585 | while (hv_iternext (hv)) |
|
|
586 | ++count; |
|
|
587 | |
|
|
588 | hv_iterinit (hv); |
|
|
589 | } |
|
|
590 | |
|
|
591 | if (count) |
|
|
592 | { |
394 | int fast = 1; |
593 | int i, fast = 1; |
395 | HE *he; |
|
|
396 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
594 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
397 | HE **hes = _alloca (count * sizeof (HE)); |
595 | HE **hes = _alloca (count * sizeof (HE)); |
398 | #else |
596 | #else |
399 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
597 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
400 | #endif |
598 | #endif |
… | |
… | |
427 | |
625 | |
428 | FREETMPS; |
626 | FREETMPS; |
429 | LEAVE; |
627 | LEAVE; |
430 | } |
628 | } |
431 | |
629 | |
432 | for (i = 0; i < count; ++i) |
630 | encode_nl (enc); ++enc->indent; |
|
|
631 | |
|
|
632 | while (count--) |
433 | { |
633 | { |
434 | encode_indent (enc); |
634 | encode_indent (enc); |
|
|
635 | he = hes [count]; |
435 | encode_he (enc, hes [i]); |
636 | encode_hk (enc, he); |
|
|
637 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
436 | |
638 | |
437 | if (i < count - 1) |
639 | if (count) |
438 | encode_comma (enc); |
640 | encode_comma (enc); |
439 | } |
641 | } |
440 | |
642 | |
441 | encode_nl (enc); |
643 | encode_nl (enc); --enc->indent; encode_indent (enc); |
442 | } |
644 | } |
|
|
645 | } |
443 | else |
646 | else |
|
|
647 | { |
|
|
648 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
649 | if ((he = hv_iternext (hv))) |
444 | { |
650 | { |
445 | HE *he = hv_iternext (hv); |
651 | encode_nl (enc); ++enc->indent; |
446 | |
652 | |
447 | for (;;) |
653 | for (;;) |
448 | { |
654 | { |
449 | encode_indent (enc); |
655 | encode_indent (enc); |
450 | encode_he (enc, he); |
656 | encode_hk (enc, he); |
|
|
657 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
451 | |
658 | |
452 | if (!(he = hv_iternext (hv))) |
659 | if (!(he = hv_iternext (hv))) |
453 | break; |
660 | break; |
454 | |
661 | |
455 | encode_comma (enc); |
662 | encode_comma (enc); |
456 | } |
663 | } |
457 | |
664 | |
458 | encode_nl (enc); |
665 | encode_nl (enc); --enc->indent; encode_indent (enc); |
459 | } |
666 | } |
460 | } |
667 | } |
461 | |
668 | |
462 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
669 | encode_ch (enc, '}'); |
463 | } |
670 | } |
464 | |
671 | |
465 | // encode objects, arrays and special \0=false and \1=true values. |
672 | // encode objects, arrays and special \0=false and \1=true values. |
466 | static void |
673 | static void |
467 | encode_rv (enc_t *enc, SV *sv) |
674 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
471 | SvGETMAGIC (sv); |
678 | SvGETMAGIC (sv); |
472 | svt = SvTYPE (sv); |
679 | svt = SvTYPE (sv); |
473 | |
680 | |
474 | if (expect_false (SvOBJECT (sv))) |
681 | if (expect_false (SvOBJECT (sv))) |
475 | { |
682 | { |
|
|
683 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
684 | ? json_boolean_stash |
|
|
685 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
686 | |
476 | if (SvSTASH (sv) == json_boolean_stash) |
687 | if (SvSTASH (sv) == stash) |
477 | { |
688 | { |
478 | if (SvIV (sv)) |
689 | if (SvIV (sv)) |
479 | encode_str (enc, "true", 4, 0); |
690 | encode_str (enc, "true", 4, 0); |
480 | else |
691 | else |
481 | encode_str (enc, "false", 5, 0); |
692 | encode_str (enc, "false", 5, 0); |
… | |
… | |
489 | } |
700 | } |
490 | #endif |
701 | #endif |
491 | if (enc->json.flags & F_CONV_BLESSED) |
702 | if (enc->json.flags & F_CONV_BLESSED) |
492 | { |
703 | { |
493 | // we re-bless the reference to get overload and other niceties right |
704 | // we re-bless the reference to get overload and other niceties right |
494 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1); |
705 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
495 | |
706 | |
496 | if (to_json) |
707 | if (to_json) |
497 | { |
708 | { |
|
|
709 | dSP; |
|
|
710 | |
498 | dSP; ENTER; SAVETMPS; PUSHMARK (SP); |
711 | ENTER; SAVETMPS; PUSHMARK (SP); |
499 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
712 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
500 | |
713 | |
501 | // calling with G_SCALAR ensures that we always get a 1 reutrn value |
714 | // calling with G_SCALAR ensures that we always get a 1 return value |
502 | // check anyways. |
|
|
503 | PUTBACK; |
715 | PUTBACK; |
504 | assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR)); |
716 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
505 | SPAGAIN; |
717 | SPAGAIN; |
506 | |
718 | |
|
|
719 | // catch this surprisingly common error |
|
|
720 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
721 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
722 | |
|
|
723 | sv = POPs; |
|
|
724 | PUTBACK; |
|
|
725 | |
507 | encode_sv (enc, POPs); |
726 | encode_sv (enc, sv); |
508 | |
727 | |
509 | FREETMPS; LEAVE; |
728 | FREETMPS; LEAVE; |
510 | } |
729 | } |
511 | else if (enc->json.flags & F_ALLOW_BLESSED) |
730 | else if (enc->json.flags & F_ALLOW_BLESSED) |
512 | encode_str (enc, "null", 4, 0); |
731 | encode_str (enc, "null", 4, 0); |
… | |
… | |
532 | |
751 | |
533 | if (len == 1 && *pv == '1') |
752 | if (len == 1 && *pv == '1') |
534 | encode_str (enc, "true", 4, 0); |
753 | encode_str (enc, "true", 4, 0); |
535 | else if (len == 1 && *pv == '0') |
754 | else if (len == 1 && *pv == '0') |
536 | encode_str (enc, "false", 5, 0); |
755 | encode_str (enc, "false", 5, 0); |
|
|
756 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
757 | encode_str (enc, "null", 4, 0); |
537 | else |
758 | else |
538 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
759 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
539 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
760 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
540 | } |
761 | } |
|
|
762 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
763 | encode_str (enc, "null", 4, 0); |
541 | else |
764 | else |
542 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
765 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
543 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
766 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
544 | } |
767 | } |
545 | |
768 | |
… | |
… | |
563 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
786 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
564 | enc->cur += strlen (enc->cur); |
787 | enc->cur += strlen (enc->cur); |
565 | } |
788 | } |
566 | else if (SvIOKp (sv)) |
789 | else if (SvIOKp (sv)) |
567 | { |
790 | { |
568 | // we assume we can always read an IV as a UV |
791 | // we assume we can always read an IV as a UV and vice versa |
569 | if (SvUV (sv) & ~(UV)0x7fff) |
792 | // we assume two's complement |
570 | { |
793 | // we assume no aliasing issues in the union |
571 | // large integer, use the (rather slow) snprintf way. |
794 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
572 | need (enc, sizeof (UV) * 3); |
795 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
573 | enc->cur += |
|
|
574 | SvIsUV(sv) |
|
|
575 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
|
|
576 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
577 | } |
|
|
578 | else |
|
|
579 | { |
796 | { |
580 | // optimise the "small number case" |
797 | // optimise the "small number case" |
581 | // code will likely be branchless and use only a single multiplication |
798 | // code will likely be branchless and use only a single multiplication |
|
|
799 | // works for numbers up to 59074 |
582 | I32 i = SvIV (sv); |
800 | I32 i = SvIVX (sv); |
583 | U32 u; |
801 | U32 u; |
584 | char digit, nz = 0; |
802 | char digit, nz = 0; |
585 | |
803 | |
586 | need (enc, 6); |
804 | need (enc, 6); |
587 | |
805 | |
… | |
… | |
593 | |
811 | |
594 | // now output digit by digit, each time masking out the integer part |
812 | // now output digit by digit, each time masking out the integer part |
595 | // and multiplying by 5 while moving the decimal point one to the right, |
813 | // and multiplying by 5 while moving the decimal point one to the right, |
596 | // resulting in a net multiplication by 10. |
814 | // resulting in a net multiplication by 10. |
597 | // we always write the digit to memory but conditionally increment |
815 | // we always write the digit to memory but conditionally increment |
598 | // the pointer, to ease the usage of conditional move instructions. |
816 | // the pointer, to enable the use of conditional move instructions. |
599 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
817 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
600 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
818 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
601 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
819 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
602 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
820 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
603 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
821 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
604 | } |
822 | } |
|
|
823 | else |
|
|
824 | { |
|
|
825 | // large integer, use the (rather slow) snprintf way. |
|
|
826 | need (enc, IVUV_MAXCHARS); |
|
|
827 | enc->cur += |
|
|
828 | SvIsUV(sv) |
|
|
829 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
830 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
831 | } |
605 | } |
832 | } |
606 | else if (SvROK (sv)) |
833 | else if (SvROK (sv)) |
607 | encode_rv (enc, SvRV (sv)); |
834 | encode_rv (enc, SvRV (sv)); |
608 | else if (!SvOK (sv)) |
835 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
609 | encode_str (enc, "null", 4, 0); |
836 | encode_str (enc, "null", 4, 0); |
610 | else |
837 | else |
611 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
838 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
612 | SvPV_nolen (sv), SvFLAGS (sv)); |
839 | SvPV_nolen (sv), SvFLAGS (sv)); |
613 | } |
840 | } |
… | |
… | |
623 | enc.json = *json; |
850 | enc.json = *json; |
624 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
851 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
625 | enc.cur = SvPVX (enc.sv); |
852 | enc.cur = SvPVX (enc.sv); |
626 | enc.end = SvEND (enc.sv); |
853 | enc.end = SvEND (enc.sv); |
627 | enc.indent = 0; |
854 | enc.indent = 0; |
628 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
855 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
856 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
857 | : 0x110000UL; |
629 | |
858 | |
630 | SvPOK_only (enc.sv); |
859 | SvPOK_only (enc.sv); |
631 | encode_sv (&enc, scalar); |
860 | encode_sv (&enc, scalar); |
|
|
861 | encode_nl (&enc); |
632 | |
862 | |
633 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
863 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
634 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
864 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
635 | |
865 | |
636 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
866 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
654 | JSON json; |
884 | JSON json; |
655 | U32 depth; // recursion depth |
885 | U32 depth; // recursion depth |
656 | U32 maxdepth; // recursion depth limit |
886 | U32 maxdepth; // recursion depth limit |
657 | } dec_t; |
887 | } dec_t; |
658 | |
888 | |
659 | inline void |
889 | INLINE void |
|
|
890 | decode_comment (dec_t *dec) |
|
|
891 | { |
|
|
892 | // only '#'-style comments allowed a.t.m. |
|
|
893 | |
|
|
894 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
895 | ++dec->cur; |
|
|
896 | } |
|
|
897 | |
|
|
898 | INLINE void |
660 | decode_ws (dec_t *dec) |
899 | decode_ws (dec_t *dec) |
661 | { |
900 | { |
662 | for (;;) |
901 | for (;;) |
663 | { |
902 | { |
664 | char ch = *dec->cur; |
903 | char ch = *dec->cur; |
665 | |
904 | |
666 | if (ch > 0x20 |
905 | if (ch > 0x20) |
|
|
906 | { |
|
|
907 | if (expect_false (ch == '#')) |
|
|
908 | { |
|
|
909 | if (dec->json.flags & F_RELAXED) |
|
|
910 | decode_comment (dec); |
|
|
911 | else |
|
|
912 | break; |
|
|
913 | } |
|
|
914 | else |
|
|
915 | break; |
|
|
916 | } |
667 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
917 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
668 | break; |
918 | break; // parse error, but let higher level handle it, gives better error messages |
669 | |
919 | |
670 | ++dec->cur; |
920 | ++dec->cur; |
671 | } |
921 | } |
672 | } |
922 | } |
673 | |
923 | |
… | |
… | |
677 | if (*dec->cur != ch) \ |
927 | if (*dec->cur != ch) \ |
678 | ERR (# ch " expected"); \ |
928 | ERR (# ch " expected"); \ |
679 | ++dec->cur; \ |
929 | ++dec->cur; \ |
680 | SE |
930 | SE |
681 | |
931 | |
682 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
932 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
683 | #define DEC_DEC_DEPTH --dec->depth |
933 | #define DEC_DEC_DEPTH --dec->depth |
684 | |
934 | |
685 | static SV *decode_sv (dec_t *dec); |
935 | static SV *decode_sv (dec_t *dec); |
686 | |
936 | |
687 | static signed char decode_hexdigit[256]; |
937 | static signed char decode_hexdigit[256]; |
… | |
… | |
779 | |
1029 | |
780 | if (hi >= 0x80) |
1030 | if (hi >= 0x80) |
781 | { |
1031 | { |
782 | utf8 = 1; |
1032 | utf8 = 1; |
783 | |
1033 | |
784 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
1034 | cur = encode_utf8 (cur, hi); |
785 | } |
1035 | } |
786 | else |
1036 | else |
787 | *cur++ = hi; |
1037 | *cur++ = hi; |
788 | } |
1038 | } |
789 | break; |
1039 | break; |
… | |
… | |
791 | default: |
1041 | default: |
792 | --dec_cur; |
1042 | --dec_cur; |
793 | ERR ("illegal backslash escape sequence in string"); |
1043 | ERR ("illegal backslash escape sequence in string"); |
794 | } |
1044 | } |
795 | } |
1045 | } |
796 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
1046 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
797 | *cur++ = ch; |
1047 | *cur++ = ch; |
798 | else if (ch >= 0x80) |
1048 | else if (ch >= 0x80) |
799 | { |
1049 | { |
800 | STRLEN clen; |
1050 | STRLEN clen; |
801 | UV uch; |
|
|
802 | |
1051 | |
803 | --dec_cur; |
1052 | --dec_cur; |
804 | |
1053 | |
805 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1054 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
806 | if (clen == (STRLEN)-1) |
1055 | if (clen == (STRLEN)-1) |
807 | ERR ("malformed UTF-8 character in JSON string"); |
1056 | ERR ("malformed UTF-8 character in JSON string"); |
808 | |
1057 | |
809 | do |
1058 | do |
810 | *cur++ = *dec_cur++; |
1059 | *cur++ = *dec_cur++; |
… | |
… | |
827 | { |
1076 | { |
828 | STRLEN len = cur - buf; |
1077 | STRLEN len = cur - buf; |
829 | |
1078 | |
830 | if (sv) |
1079 | if (sv) |
831 | { |
1080 | { |
832 | SvGROW (sv, SvCUR (sv) + len + 1); |
1081 | STRLEN cur = SvCUR (sv); |
|
|
1082 | |
|
|
1083 | if (SvLEN (sv) <= cur + len) |
|
|
1084 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1085 | |
833 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1086 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
834 | SvCUR_set (sv, SvCUR (sv) + len); |
1087 | SvCUR_set (sv, SvCUR (sv) + len); |
835 | } |
1088 | } |
836 | else |
1089 | else |
837 | sv = newSVpvn (buf, len); |
1090 | sv = newSVpvn (buf, len); |
… | |
… | |
922 | is_nv = 1; |
1175 | is_nv = 1; |
923 | } |
1176 | } |
924 | |
1177 | |
925 | if (!is_nv) |
1178 | if (!is_nv) |
926 | { |
1179 | { |
|
|
1180 | int len = dec->cur - start; |
|
|
1181 | |
927 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1182 | // special case the rather common 1..5-digit-int case |
928 | if (*start == '-') |
1183 | if (*start == '-') |
929 | switch (dec->cur - start) |
1184 | switch (len) |
930 | { |
1185 | { |
931 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1186 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
932 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1187 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
933 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1188 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
934 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1189 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1190 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
935 | } |
1191 | } |
936 | else |
1192 | else |
937 | switch (dec->cur - start) |
1193 | switch (len) |
938 | { |
1194 | { |
939 | case 1: return newSViv ( start [0] - '0' * 1); |
1195 | case 1: return newSViv ( start [0] - '0' * 1); |
940 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1196 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
941 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1197 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
942 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1198 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1199 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
943 | } |
1200 | } |
944 | |
1201 | |
945 | { |
1202 | { |
946 | UV uv; |
1203 | UV uv; |
947 | int numtype = grok_number (start, dec->cur - start, &uv); |
1204 | int numtype = grok_number (start, len, &uv); |
948 | if (numtype & IS_NUMBER_IN_UV) |
1205 | if (numtype & IS_NUMBER_IN_UV) |
949 | if (numtype & IS_NUMBER_NEG) |
1206 | if (numtype & IS_NUMBER_NEG) |
950 | { |
1207 | { |
951 | if (uv < (UV)IV_MIN) |
1208 | if (uv < (UV)IV_MIN) |
952 | return newSViv (-(IV)uv); |
1209 | return newSViv (-(IV)uv); |
953 | } |
1210 | } |
954 | else |
1211 | else |
955 | return newSVuv (uv); |
1212 | return newSVuv (uv); |
956 | |
|
|
957 | // here would likely be the place for bigint support |
|
|
958 | } |
1213 | } |
959 | } |
|
|
960 | |
1214 | |
961 | // if we ever support bigint or bigfloat, this is the place for bigfloat |
1215 | len -= *start == '-' ? 1 : 0; |
|
|
1216 | |
|
|
1217 | // does not fit into IV or UV, try NV |
|
|
1218 | if (len <= NV_DIG) |
|
|
1219 | // fits into NV without loss of precision |
|
|
1220 | return newSVnv (json_atof (start)); |
|
|
1221 | |
|
|
1222 | // everything else fails, convert it to a string |
|
|
1223 | return newSVpvn (start, dec->cur - start); |
|
|
1224 | } |
|
|
1225 | |
|
|
1226 | // loss of precision here |
962 | return newSVnv (Atof (start)); |
1227 | return newSVnv (json_atof (start)); |
963 | |
1228 | |
964 | fail: |
1229 | fail: |
965 | return 0; |
1230 | return 0; |
966 | } |
1231 | } |
967 | |
1232 | |
… | |
… | |
996 | |
1261 | |
997 | if (*dec->cur != ',') |
1262 | if (*dec->cur != ',') |
998 | ERR (", or ] expected while parsing array"); |
1263 | ERR (", or ] expected while parsing array"); |
999 | |
1264 | |
1000 | ++dec->cur; |
1265 | ++dec->cur; |
|
|
1266 | |
|
|
1267 | decode_ws (dec); |
|
|
1268 | |
|
|
1269 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1270 | { |
|
|
1271 | ++dec->cur; |
|
|
1272 | break; |
|
|
1273 | } |
1001 | } |
1274 | } |
1002 | |
1275 | |
1003 | DEC_DEC_DEPTH; |
1276 | DEC_DEC_DEPTH; |
1004 | return newRV_noinc ((SV *)av); |
1277 | return newRV_noinc ((SV *)av); |
1005 | |
1278 | |
… | |
… | |
1021 | if (*dec->cur == '}') |
1294 | if (*dec->cur == '}') |
1022 | ++dec->cur; |
1295 | ++dec->cur; |
1023 | else |
1296 | else |
1024 | for (;;) |
1297 | for (;;) |
1025 | { |
1298 | { |
1026 | decode_ws (dec); EXPECT_CH ('"'); |
1299 | EXPECT_CH ('"'); |
1027 | |
1300 | |
1028 | // heuristic: assume that |
1301 | // heuristic: assume that |
1029 | // a) decode_str + hv_store_ent are abysmally slow. |
1302 | // a) decode_str + hv_store_ent are abysmally slow. |
1030 | // b) most hash keys are short, simple ascii text. |
1303 | // b) most hash keys are short, simple ascii text. |
1031 | // => try to "fast-match" such strings to avoid |
1304 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1035 | char *p = dec->cur; |
1308 | char *p = dec->cur; |
1036 | char *e = p + 24; // only try up to 24 bytes |
1309 | char *e = p + 24; // only try up to 24 bytes |
1037 | |
1310 | |
1038 | for (;;) |
1311 | for (;;) |
1039 | { |
1312 | { |
1040 | // the >= 0x80 is true on most architectures |
1313 | // the >= 0x80 is false on most architectures |
1041 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1314 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1042 | { |
1315 | { |
1043 | // slow path, back up and use decode_str |
1316 | // slow path, back up and use decode_str |
1044 | SV *key = decode_str (dec); |
1317 | SV *key = decode_str (dec); |
1045 | if (!key) |
1318 | if (!key) |
1046 | goto fail; |
1319 | goto fail; |
1047 | |
1320 | |
1048 | decode_ws (dec); EXPECT_CH (':'); |
1321 | decode_ws (dec); EXPECT_CH (':'); |
1049 | |
1322 | |
|
|
1323 | decode_ws (dec); |
1050 | value = decode_sv (dec); |
1324 | value = decode_sv (dec); |
1051 | if (!value) |
1325 | if (!value) |
1052 | { |
1326 | { |
1053 | SvREFCNT_dec (key); |
1327 | SvREFCNT_dec (key); |
1054 | goto fail; |
1328 | goto fail; |
… | |
… | |
1066 | int len = p - key; |
1340 | int len = p - key; |
1067 | dec->cur = p + 1; |
1341 | dec->cur = p + 1; |
1068 | |
1342 | |
1069 | decode_ws (dec); EXPECT_CH (':'); |
1343 | decode_ws (dec); EXPECT_CH (':'); |
1070 | |
1344 | |
|
|
1345 | decode_ws (dec); |
1071 | value = decode_sv (dec); |
1346 | value = decode_sv (dec); |
1072 | if (!value) |
1347 | if (!value) |
1073 | goto fail; |
1348 | goto fail; |
1074 | |
1349 | |
1075 | hv_store (hv, key, len, value, 0); |
1350 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1091 | |
1366 | |
1092 | if (*dec->cur != ',') |
1367 | if (*dec->cur != ',') |
1093 | ERR (", or } expected while parsing object/hash"); |
1368 | ERR (", or } expected while parsing object/hash"); |
1094 | |
1369 | |
1095 | ++dec->cur; |
1370 | ++dec->cur; |
|
|
1371 | |
|
|
1372 | decode_ws (dec); |
|
|
1373 | |
|
|
1374 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1375 | { |
|
|
1376 | ++dec->cur; |
|
|
1377 | break; |
|
|
1378 | } |
1096 | } |
1379 | } |
1097 | |
1380 | |
1098 | DEC_DEC_DEPTH; |
1381 | DEC_DEC_DEPTH; |
1099 | sv = newRV_noinc ((SV *)hv); |
1382 | sv = newRV_noinc ((SV *)hv); |
1100 | |
1383 | |
… | |
… | |
1113 | // might want to optimise this for common cases. |
1396 | // might want to optimise this for common cases. |
1114 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1397 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1115 | |
1398 | |
1116 | if (cb) |
1399 | if (cb) |
1117 | { |
1400 | { |
|
|
1401 | dSP; |
1118 | int count; |
1402 | int count; |
1119 | ENTER; SAVETMPS; |
|
|
1120 | |
1403 | |
1121 | dSP; PUSHMARK (SP); |
1404 | ENTER; SAVETMPS; PUSHMARK (SP); |
1122 | XPUSHs (HeVAL (he)); |
1405 | XPUSHs (HeVAL (he)); |
1123 | |
1406 | |
1124 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1407 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1125 | |
1408 | |
1126 | if (count == 1) |
1409 | if (count == 1) |
… | |
… | |
1134 | } |
1417 | } |
1135 | } |
1418 | } |
1136 | |
1419 | |
1137 | if (dec->json.cb_object) |
1420 | if (dec->json.cb_object) |
1138 | { |
1421 | { |
|
|
1422 | dSP; |
1139 | int count; |
1423 | int count; |
1140 | ENTER; SAVETMPS; |
|
|
1141 | |
1424 | |
1142 | dSP; ENTER; SAVETMPS; PUSHMARK (SP); |
1425 | ENTER; SAVETMPS; PUSHMARK (SP); |
1143 | XPUSHs (sv_2mortal (sv)); |
1426 | XPUSHs (sv_2mortal (sv)); |
1144 | |
1427 | |
1145 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1428 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1146 | |
1429 | |
1147 | if (count == 1) |
1430 | if (count == 1) |
… | |
… | |
1165 | } |
1448 | } |
1166 | |
1449 | |
1167 | static SV * |
1450 | static SV * |
1168 | decode_sv (dec_t *dec) |
1451 | decode_sv (dec_t *dec) |
1169 | { |
1452 | { |
1170 | decode_ws (dec); |
|
|
1171 | |
|
|
1172 | // the beauty of JSON: you need exactly one character lookahead |
1453 | // the beauty of JSON: you need exactly one character lookahead |
1173 | // to parse anything. |
1454 | // to parse everything. |
1174 | switch (*dec->cur) |
1455 | switch (*dec->cur) |
1175 | { |
1456 | { |
1176 | case '"': ++dec->cur; return decode_str (dec); |
1457 | case '"': ++dec->cur; return decode_str (dec); |
1177 | case '[': ++dec->cur; return decode_av (dec); |
1458 | case '[': ++dec->cur; return decode_av (dec); |
1178 | case '{': ++dec->cur; return decode_hv (dec); |
1459 | case '{': ++dec->cur; return decode_hv (dec); |
1179 | |
1460 | |
1180 | case '-': |
1461 | case '-': |
1181 | case '0': case '1': case '2': case '3': case '4': |
1462 | case '0': case '1': case '2': case '3': case '4': |
1182 | case '5': case '6': case '7': case '8': case '9': |
1463 | case '5': case '6': case '7': case '8': case '9': |
1183 | return decode_num (dec); |
1464 | return decode_num (dec); |
1184 | |
1465 | |
1185 | case 't': |
1466 | case 't': |
1186 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1467 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1187 | { |
1468 | { |
1188 | dec->cur += 4; |
1469 | dec->cur += 4; |
|
|
1470 | #if JSON_SLOW |
|
|
1471 | json_true = get_bool ("JSON::XS::true"); |
|
|
1472 | #endif |
1189 | return SvREFCNT_inc (json_true); |
1473 | return newSVsv (json_true); |
1190 | } |
1474 | } |
1191 | else |
1475 | else |
1192 | ERR ("'true' expected"); |
1476 | ERR ("'true' expected"); |
1193 | |
1477 | |
1194 | break; |
1478 | break; |
1195 | |
1479 | |
1196 | case 'f': |
1480 | case 'f': |
1197 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1481 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1198 | { |
1482 | { |
1199 | dec->cur += 5; |
1483 | dec->cur += 5; |
|
|
1484 | #if JSON_SLOW |
|
|
1485 | json_false = get_bool ("JSON::XS::false"); |
|
|
1486 | #endif |
1200 | return SvREFCNT_inc (json_false); |
1487 | return newSVsv (json_false); |
1201 | } |
1488 | } |
1202 | else |
1489 | else |
1203 | ERR ("'false' expected"); |
1490 | ERR ("'false' expected"); |
1204 | |
1491 | |
1205 | break; |
1492 | break; |
… | |
… | |
1223 | fail: |
1510 | fail: |
1224 | return 0; |
1511 | return 0; |
1225 | } |
1512 | } |
1226 | |
1513 | |
1227 | static SV * |
1514 | static SV * |
1228 | decode_json (SV *string, JSON *json, UV *offset_return) |
1515 | decode_json (SV *string, JSON *json, char **offset_return) |
1229 | { |
1516 | { |
1230 | dec_t dec; |
1517 | dec_t dec; |
1231 | UV offset; |
|
|
1232 | SV *sv; |
1518 | SV *sv; |
1233 | |
1519 | |
|
|
1520 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1521 | * will perl ignore the magic in subsequent accesses |
|
|
1522 | */ |
1234 | SvGETMAGIC (string); |
1523 | /*SvGETMAGIC (string);*/ |
|
|
1524 | if (SvMAGICAL (string)) |
|
|
1525 | string = sv_2mortal (newSVsv (string)); |
|
|
1526 | |
1235 | SvUPGRADE (string, SVt_PV); |
1527 | SvUPGRADE (string, SVt_PV); |
1236 | |
1528 | |
1237 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1529 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1530 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1531 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1532 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1533 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1534 | * and hope for the best. |
|
|
1535 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1536 | * assertion business is seriously broken, try yet another workaround |
|
|
1537 | * for the broken -DDEBUGGING. |
|
|
1538 | */ |
|
|
1539 | { |
|
|
1540 | #ifdef DEBUGGING |
|
|
1541 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1542 | #else |
|
|
1543 | STRLEN offset = SvCUR (string); |
|
|
1544 | #endif |
|
|
1545 | |
|
|
1546 | if (offset > json->max_size && json->max_size) |
1238 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1547 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1239 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1548 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1549 | } |
1240 | |
1550 | |
1241 | if (json->flags & F_UTF8) |
1551 | if (json->flags & F_UTF8) |
1242 | sv_utf8_downgrade (string, 0); |
1552 | sv_utf8_downgrade (string, 0); |
1243 | else |
1553 | else |
1244 | sv_utf8_upgrade (string); |
1554 | sv_utf8_upgrade (string); |
1245 | |
1555 | |
1246 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1556 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1247 | |
1557 | |
1248 | dec.json = *json; |
1558 | dec.json = *json; |
1249 | dec.cur = SvPVX (string); |
1559 | dec.cur = SvPVX (string); |
1250 | dec.end = SvEND (string); |
1560 | dec.end = SvEND (string); |
1251 | dec.err = 0; |
1561 | dec.err = 0; |
1252 | dec.depth = 0; |
1562 | dec.depth = 0; |
1253 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1254 | |
1563 | |
1255 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1564 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1256 | dec.json.flags |= F_HOOK; |
1565 | dec.json.flags |= F_HOOK; |
1257 | |
1566 | |
1258 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1567 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1568 | |
|
|
1569 | decode_ws (&dec); |
1259 | sv = decode_sv (&dec); |
1570 | sv = decode_sv (&dec); |
|
|
1571 | |
|
|
1572 | if (offset_return) |
|
|
1573 | *offset_return = dec.cur; |
1260 | |
1574 | |
1261 | if (!(offset_return || !sv)) |
1575 | if (!(offset_return || !sv)) |
1262 | { |
1576 | { |
1263 | // check for trailing garbage |
1577 | // check for trailing garbage |
1264 | decode_ws (&dec); |
1578 | decode_ws (&dec); |
… | |
… | |
1267 | { |
1581 | { |
1268 | dec.err = "garbage after JSON object"; |
1582 | dec.err = "garbage after JSON object"; |
1269 | SvREFCNT_dec (sv); |
1583 | SvREFCNT_dec (sv); |
1270 | sv = 0; |
1584 | sv = 0; |
1271 | } |
1585 | } |
1272 | } |
|
|
1273 | |
|
|
1274 | if (offset_return || !sv) |
|
|
1275 | { |
|
|
1276 | offset = dec.json.flags & F_UTF8 |
|
|
1277 | ? dec.cur - SvPVX (string) |
|
|
1278 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1279 | |
|
|
1280 | if (offset_return) |
|
|
1281 | *offset_return = offset; |
|
|
1282 | } |
1586 | } |
1283 | |
1587 | |
1284 | if (!sv) |
1588 | if (!sv) |
1285 | { |
1589 | { |
1286 | SV *uni = sv_newmortal (); |
1590 | SV *uni = sv_newmortal (); |
… | |
… | |
1292 | SAVEVPTR (PL_curcop); |
1596 | SAVEVPTR (PL_curcop); |
1293 | PL_curcop = &cop; |
1597 | PL_curcop = &cop; |
1294 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1598 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1295 | LEAVE; |
1599 | LEAVE; |
1296 | |
1600 | |
1297 | croak ("%s, at character offset %d [\"%s\"]", |
1601 | croak ("%s, at character offset %d (before \"%s\")", |
1298 | dec.err, |
1602 | dec.err, |
1299 | (int)offset, |
1603 | ptr_to_index (string, dec.cur), |
1300 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1604 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1301 | } |
1605 | } |
1302 | |
1606 | |
1303 | sv = sv_2mortal (sv); |
1607 | sv = sv_2mortal (sv); |
1304 | |
1608 | |
1305 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1609 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1306 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1610 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1307 | |
1611 | |
1308 | return sv; |
1612 | return sv; |
|
|
1613 | } |
|
|
1614 | |
|
|
1615 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1616 | // incremental parser |
|
|
1617 | |
|
|
1618 | static void |
|
|
1619 | incr_parse (JSON *self) |
|
|
1620 | { |
|
|
1621 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1622 | |
|
|
1623 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1624 | // but this would make it slower, so... |
|
|
1625 | |
|
|
1626 | for (;;) |
|
|
1627 | { |
|
|
1628 | //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 |
|
|
1629 | switch (self->incr_mode) |
|
|
1630 | { |
|
|
1631 | // only used for initial whitespace skipping |
|
|
1632 | case INCR_M_WS: |
|
|
1633 | for (;;) |
|
|
1634 | { |
|
|
1635 | if (*p > 0x20) |
|
|
1636 | { |
|
|
1637 | if (*p == '#') |
|
|
1638 | { |
|
|
1639 | self->incr_mode = INCR_M_C0; |
|
|
1640 | goto incr_m_c; |
|
|
1641 | } |
|
|
1642 | else |
|
|
1643 | { |
|
|
1644 | self->incr_mode = INCR_M_JSON; |
|
|
1645 | goto incr_m_json; |
|
|
1646 | } |
|
|
1647 | } |
|
|
1648 | else if (!*p) |
|
|
1649 | goto interrupt; |
|
|
1650 | |
|
|
1651 | ++p; |
|
|
1652 | } |
|
|
1653 | |
|
|
1654 | // skip a single char inside a string (for \\-processing) |
|
|
1655 | case INCR_M_BS: |
|
|
1656 | if (!*p) |
|
|
1657 | goto interrupt; |
|
|
1658 | |
|
|
1659 | ++p; |
|
|
1660 | self->incr_mode = INCR_M_STR; |
|
|
1661 | goto incr_m_str; |
|
|
1662 | |
|
|
1663 | // inside #-style comments |
|
|
1664 | case INCR_M_C0: |
|
|
1665 | case INCR_M_C1: |
|
|
1666 | incr_m_c: |
|
|
1667 | for (;;) |
|
|
1668 | { |
|
|
1669 | if (*p == '\n') |
|
|
1670 | { |
|
|
1671 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1672 | break; |
|
|
1673 | } |
|
|
1674 | else if (!*p) |
|
|
1675 | goto interrupt; |
|
|
1676 | |
|
|
1677 | ++p; |
|
|
1678 | } |
|
|
1679 | |
|
|
1680 | break; |
|
|
1681 | |
|
|
1682 | // inside a string |
|
|
1683 | case INCR_M_STR: |
|
|
1684 | incr_m_str: |
|
|
1685 | for (;;) |
|
|
1686 | { |
|
|
1687 | if (*p == '"') |
|
|
1688 | { |
|
|
1689 | ++p; |
|
|
1690 | self->incr_mode = INCR_M_JSON; |
|
|
1691 | |
|
|
1692 | if (!self->incr_nest) |
|
|
1693 | goto interrupt; |
|
|
1694 | |
|
|
1695 | goto incr_m_json; |
|
|
1696 | } |
|
|
1697 | else if (*p == '\\') |
|
|
1698 | { |
|
|
1699 | ++p; // "virtually" consumes character after \ |
|
|
1700 | |
|
|
1701 | if (!*p) // if at end of string we have to switch modes |
|
|
1702 | { |
|
|
1703 | self->incr_mode = INCR_M_BS; |
|
|
1704 | goto interrupt; |
|
|
1705 | } |
|
|
1706 | } |
|
|
1707 | else if (!*p) |
|
|
1708 | goto interrupt; |
|
|
1709 | |
|
|
1710 | ++p; |
|
|
1711 | } |
|
|
1712 | |
|
|
1713 | // after initial ws, outside string |
|
|
1714 | case INCR_M_JSON: |
|
|
1715 | incr_m_json: |
|
|
1716 | for (;;) |
|
|
1717 | { |
|
|
1718 | switch (*p++) |
|
|
1719 | { |
|
|
1720 | case 0: |
|
|
1721 | --p; |
|
|
1722 | goto interrupt; |
|
|
1723 | |
|
|
1724 | case 0x09: |
|
|
1725 | case 0x0a: |
|
|
1726 | case 0x0d: |
|
|
1727 | case 0x20: |
|
|
1728 | if (!self->incr_nest) |
|
|
1729 | { |
|
|
1730 | --p; // do not eat the whitespace, let the next round do it |
|
|
1731 | goto interrupt; |
|
|
1732 | } |
|
|
1733 | break; |
|
|
1734 | |
|
|
1735 | case '"': |
|
|
1736 | self->incr_mode = INCR_M_STR; |
|
|
1737 | goto incr_m_str; |
|
|
1738 | |
|
|
1739 | case '[': |
|
|
1740 | case '{': |
|
|
1741 | if (++self->incr_nest > self->max_depth) |
|
|
1742 | croak (ERR_NESTING_EXCEEDED); |
|
|
1743 | break; |
|
|
1744 | |
|
|
1745 | case ']': |
|
|
1746 | case '}': |
|
|
1747 | if (--self->incr_nest <= 0) |
|
|
1748 | goto interrupt; |
|
|
1749 | break; |
|
|
1750 | |
|
|
1751 | case '#': |
|
|
1752 | self->incr_mode = INCR_M_C1; |
|
|
1753 | goto incr_m_c; |
|
|
1754 | } |
|
|
1755 | } |
|
|
1756 | } |
|
|
1757 | |
|
|
1758 | modechange: |
|
|
1759 | ; |
|
|
1760 | } |
|
|
1761 | |
|
|
1762 | interrupt: |
|
|
1763 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1764 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1765 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1309 | } |
1766 | } |
1310 | |
1767 | |
1311 | ///////////////////////////////////////////////////////////////////////////// |
1768 | ///////////////////////////////////////////////////////////////////////////// |
1312 | // XS interface functions |
1769 | // XS interface functions |
1313 | |
1770 | |
… | |
… | |
1325 | : -1; |
1782 | : -1; |
1326 | |
1783 | |
1327 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1784 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1328 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1785 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1329 | |
1786 | |
1330 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1787 | json_true = get_bool ("JSON::XS::true"); |
1331 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1788 | json_false = get_bool ("JSON::XS::false"); |
|
|
1789 | |
|
|
1790 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1332 | } |
1791 | } |
1333 | |
1792 | |
1334 | PROTOTYPES: DISABLE |
1793 | PROTOTYPES: DISABLE |
|
|
1794 | |
|
|
1795 | void CLONE (...) |
|
|
1796 | CODE: |
|
|
1797 | json_stash = 0; |
|
|
1798 | json_boolean_stash = 0; |
1335 | |
1799 | |
1336 | void new (char *klass) |
1800 | void new (char *klass) |
1337 | PPCODE: |
1801 | PPCODE: |
1338 | { |
1802 | { |
1339 | SV *pv = NEWSV (0, sizeof (JSON)); |
1803 | SV *pv = NEWSV (0, sizeof (JSON)); |
1340 | SvPOK_only (pv); |
1804 | SvPOK_only (pv); |
1341 | Zero (SvPVX (pv), 1, JSON); |
1805 | json_init ((JSON *)SvPVX (pv)); |
1342 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1806 | XPUSHs (sv_2mortal (sv_bless ( |
1343 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash))); |
1807 | newRV_noinc (pv), |
|
|
1808 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1809 | ))); |
1344 | } |
1810 | } |
1345 | |
1811 | |
1346 | void ascii (JSON *self, int enable = 1) |
1812 | void ascii (JSON *self, int enable = 1) |
1347 | ALIAS: |
1813 | ALIAS: |
1348 | ascii = F_ASCII |
1814 | ascii = F_ASCII |
… | |
… | |
1355 | pretty = F_PRETTY |
1821 | pretty = F_PRETTY |
1356 | allow_nonref = F_ALLOW_NONREF |
1822 | allow_nonref = F_ALLOW_NONREF |
1357 | shrink = F_SHRINK |
1823 | shrink = F_SHRINK |
1358 | allow_blessed = F_ALLOW_BLESSED |
1824 | allow_blessed = F_ALLOW_BLESSED |
1359 | convert_blessed = F_CONV_BLESSED |
1825 | convert_blessed = F_CONV_BLESSED |
|
|
1826 | relaxed = F_RELAXED |
|
|
1827 | allow_unknown = F_ALLOW_UNKNOWN |
1360 | PPCODE: |
1828 | PPCODE: |
1361 | { |
1829 | { |
1362 | if (enable) |
1830 | if (enable) |
1363 | self->flags |= ix; |
1831 | self->flags |= ix; |
1364 | else |
1832 | else |
1365 | self->flags &= ~ix; |
1833 | self->flags &= ~ix; |
1366 | |
1834 | |
1367 | XPUSHs (ST (0)); |
1835 | XPUSHs (ST (0)); |
1368 | } |
1836 | } |
1369 | |
1837 | |
1370 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1838 | void get_ascii (JSON *self) |
|
|
1839 | ALIAS: |
|
|
1840 | get_ascii = F_ASCII |
|
|
1841 | get_latin1 = F_LATIN1 |
|
|
1842 | get_utf8 = F_UTF8 |
|
|
1843 | get_indent = F_INDENT |
|
|
1844 | get_canonical = F_CANONICAL |
|
|
1845 | get_space_before = F_SPACE_BEFORE |
|
|
1846 | get_space_after = F_SPACE_AFTER |
|
|
1847 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1848 | get_shrink = F_SHRINK |
|
|
1849 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1850 | get_convert_blessed = F_CONV_BLESSED |
|
|
1851 | get_relaxed = F_RELAXED |
|
|
1852 | get_allow_unknown = F_ALLOW_UNKNOWN |
1371 | PPCODE: |
1853 | PPCODE: |
1372 | { |
1854 | XPUSHs (boolSV (self->flags & ix)); |
1373 | UV log2 = 0; |
|
|
1374 | |
1855 | |
1375 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1856 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1376 | |
1857 | PPCODE: |
1377 | while ((1UL << log2) < max_depth) |
1858 | self->max_depth = max_depth; |
1378 | ++log2; |
|
|
1379 | |
|
|
1380 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1381 | |
|
|
1382 | XPUSHs (ST (0)); |
1859 | XPUSHs (ST (0)); |
1383 | } |
|
|
1384 | |
1860 | |
|
|
1861 | U32 get_max_depth (JSON *self) |
|
|
1862 | CODE: |
|
|
1863 | RETVAL = self->max_depth; |
|
|
1864 | OUTPUT: |
|
|
1865 | RETVAL |
|
|
1866 | |
1385 | void max_size (JSON *self, UV max_size = 0) |
1867 | void max_size (JSON *self, U32 max_size = 0) |
1386 | PPCODE: |
1868 | PPCODE: |
1387 | { |
1869 | self->max_size = max_size; |
1388 | UV log2 = 0; |
|
|
1389 | |
|
|
1390 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1391 | if (max_size == 1) max_size = 2; |
|
|
1392 | |
|
|
1393 | while ((1UL << log2) < max_size) |
|
|
1394 | ++log2; |
|
|
1395 | |
|
|
1396 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1397 | |
|
|
1398 | XPUSHs (ST (0)); |
1870 | XPUSHs (ST (0)); |
1399 | } |
1871 | |
|
|
1872 | int get_max_size (JSON *self) |
|
|
1873 | CODE: |
|
|
1874 | RETVAL = self->max_size; |
|
|
1875 | OUTPUT: |
|
|
1876 | RETVAL |
1400 | |
1877 | |
1401 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1878 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1402 | PPCODE: |
1879 | PPCODE: |
1403 | { |
1880 | { |
1404 | SvREFCNT_dec (self->cb_object); |
1881 | SvREFCNT_dec (self->cb_object); |
… | |
… | |
1438 | XPUSHs (decode_json (jsonstr, self, 0)); |
1915 | XPUSHs (decode_json (jsonstr, self, 0)); |
1439 | |
1916 | |
1440 | void decode_prefix (JSON *self, SV *jsonstr) |
1917 | void decode_prefix (JSON *self, SV *jsonstr) |
1441 | PPCODE: |
1918 | PPCODE: |
1442 | { |
1919 | { |
1443 | UV offset; |
1920 | char *offset; |
1444 | EXTEND (SP, 2); |
1921 | EXTEND (SP, 2); |
1445 | PUSHs (decode_json (jsonstr, self, &offset)); |
1922 | PUSHs (decode_json (jsonstr, self, &offset)); |
1446 | PUSHs (sv_2mortal (newSVuv (offset))); |
1923 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1924 | } |
|
|
1925 | |
|
|
1926 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1927 | PPCODE: |
|
|
1928 | { |
|
|
1929 | if (!self->incr_text) |
|
|
1930 | self->incr_text = newSVpvn ("", 0); |
|
|
1931 | |
|
|
1932 | // append data, if any |
|
|
1933 | if (jsonstr) |
|
|
1934 | { |
|
|
1935 | if (SvUTF8 (jsonstr)) |
|
|
1936 | { |
|
|
1937 | if (!SvUTF8 (self->incr_text)) |
|
|
1938 | { |
|
|
1939 | /* utf-8-ness differs, need to upgrade */ |
|
|
1940 | sv_utf8_upgrade (self->incr_text); |
|
|
1941 | |
|
|
1942 | if (self->incr_pos) |
|
|
1943 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1944 | - (U8 *)SvPVX (self->incr_text); |
|
|
1945 | } |
|
|
1946 | } |
|
|
1947 | else if (SvUTF8 (self->incr_text)) |
|
|
1948 | sv_utf8_upgrade (jsonstr); |
|
|
1949 | |
|
|
1950 | { |
|
|
1951 | STRLEN len; |
|
|
1952 | const char *str = SvPV (jsonstr, len); |
|
|
1953 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1954 | |
|
|
1955 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1956 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1957 | |
|
|
1958 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1959 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1960 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1961 | } |
|
|
1962 | } |
|
|
1963 | |
|
|
1964 | if (GIMME_V != G_VOID) |
|
|
1965 | do |
|
|
1966 | { |
|
|
1967 | char *offset; |
|
|
1968 | |
|
|
1969 | if (!INCR_DONE (self)) |
|
|
1970 | { |
|
|
1971 | incr_parse (self); |
|
|
1972 | |
|
|
1973 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1974 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1975 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1976 | |
|
|
1977 | if (!INCR_DONE (self)) |
|
|
1978 | break; |
|
|
1979 | } |
|
|
1980 | |
|
|
1981 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1982 | |
|
|
1983 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
1984 | self->incr_nest = 0; |
|
|
1985 | self->incr_mode = 0; |
|
|
1986 | |
|
|
1987 | sv_chop (self->incr_text, offset); |
|
|
1988 | } |
|
|
1989 | while (GIMME_V == G_ARRAY); |
|
|
1990 | } |
|
|
1991 | |
|
|
1992 | SV *incr_text (JSON *self) |
|
|
1993 | ATTRS: lvalue |
|
|
1994 | CODE: |
|
|
1995 | { |
|
|
1996 | if (self->incr_pos) |
|
|
1997 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
1998 | |
|
|
1999 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2000 | } |
|
|
2001 | OUTPUT: |
|
|
2002 | RETVAL |
|
|
2003 | |
|
|
2004 | void incr_skip (JSON *self) |
|
|
2005 | CODE: |
|
|
2006 | { |
|
|
2007 | if (self->incr_pos) |
|
|
2008 | { |
|
|
2009 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2010 | self->incr_pos = 0; |
|
|
2011 | self->incr_nest = 0; |
|
|
2012 | self->incr_mode = 0; |
|
|
2013 | } |
|
|
2014 | } |
|
|
2015 | |
|
|
2016 | void incr_reset (JSON *self) |
|
|
2017 | CODE: |
|
|
2018 | { |
|
|
2019 | SvREFCNT_dec (self->incr_text); |
|
|
2020 | self->incr_text = 0; |
|
|
2021 | self->incr_pos = 0; |
|
|
2022 | self->incr_nest = 0; |
|
|
2023 | self->incr_mode = 0; |
1447 | } |
2024 | } |
1448 | |
2025 | |
1449 | void DESTROY (JSON *self) |
2026 | void DESTROY (JSON *self) |
1450 | CODE: |
2027 | CODE: |
1451 | SvREFCNT_dec (self->cb_sk_object); |
2028 | SvREFCNT_dec (self->cb_sk_object); |
1452 | SvREFCNT_dec (self->cb_object); |
2029 | SvREFCNT_dec (self->cb_object); |
|
|
2030 | SvREFCNT_dec (self->incr_text); |
1453 | |
2031 | |
1454 | PROTOTYPES: ENABLE |
2032 | PROTOTYPES: ENABLE |
1455 | |
2033 | |
1456 | void to_json (SV *scalar) |
2034 | void encode_json (SV *scalar) |
|
|
2035 | ALIAS: |
|
|
2036 | to_json_ = 0 |
|
|
2037 | encode_json = F_UTF8 |
1457 | PPCODE: |
2038 | PPCODE: |
1458 | { |
2039 | { |
1459 | JSON json = { F_DEFAULT | F_UTF8 }; |
2040 | JSON json; |
|
|
2041 | json_init (&json); |
|
|
2042 | json.flags |= ix; |
1460 | XPUSHs (encode_json (scalar, &json)); |
2043 | XPUSHs (encode_json (scalar, &json)); |
1461 | } |
2044 | } |
1462 | |
2045 | |
1463 | void from_json (SV *jsonstr) |
2046 | void decode_json (SV *jsonstr) |
|
|
2047 | ALIAS: |
|
|
2048 | from_json_ = 0 |
|
|
2049 | decode_json = F_UTF8 |
1464 | PPCODE: |
2050 | PPCODE: |
1465 | { |
2051 | { |
1466 | JSON json = { F_DEFAULT | F_UTF8 }; |
2052 | JSON json; |
|
|
2053 | json_init (&json); |
|
|
2054 | json.flags |= ix; |
1467 | XPUSHs (decode_json (jsonstr, &json, 0)); |
2055 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1468 | } |
2056 | } |
1469 | |
2057 | |