1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
|
|
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
|
|
10 | #include <float.h> |
8 | |
11 | |
|
|
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
|
|
14 | #endif |
|
|
15 | |
|
|
16 | // some old perls do not have this, try to make it work, no |
|
|
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
|
|
18 | #ifndef UTF8_MAXBYTES |
|
|
19 | # define UTF8_MAXBYTES 13 |
|
|
20 | #endif |
|
|
21 | |
|
|
22 | // compatibility with perl <5.18 |
|
|
23 | #ifndef HvNAMELEN_get |
|
|
24 | # define HvNAMELEN_get(hv) strlen (HvNAME (hv)) |
|
|
25 | #endif |
|
|
26 | #ifndef HvNAMELEN |
|
|
27 | # define HvNAMELEN(hv) HvNAMELEN_get (hv) |
|
|
28 | #endif |
|
|
29 | #ifndef HvNAMEUTF8 |
|
|
30 | # define HvNAMEUTF8(hv) 0 |
|
|
31 | #endif |
|
|
32 | |
|
|
33 | // three extra for rounding, sign, and end of string |
|
|
34 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
|
|
35 | |
9 | #define F_ASCII 0x00000001UL |
36 | #define F_ASCII 0x00000001UL |
|
|
37 | #define F_LATIN1 0x00000002UL |
10 | #define F_UTF8 0x00000002UL |
38 | #define F_UTF8 0x00000004UL |
11 | #define F_INDENT 0x00000004UL |
39 | #define F_INDENT 0x00000008UL |
12 | #define F_CANONICAL 0x00000008UL |
40 | #define F_CANONICAL 0x00000010UL |
13 | #define F_SPACE_BEFORE 0x00000010UL |
41 | #define F_SPACE_BEFORE 0x00000020UL |
14 | #define F_SPACE_AFTER 0x00000020UL |
42 | #define F_SPACE_AFTER 0x00000040UL |
15 | #define F_ALLOW_NONREF 0x00000080UL |
43 | #define F_ALLOW_NONREF 0x00000100UL |
16 | #define F_SHRINK 0x00000100UL |
44 | #define F_SHRINK 0x00000200UL |
|
|
45 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
46 | #define F_CONV_BLESSED 0x00000800UL |
17 | #define F_MAXDEPTH 0xf8000000UL |
47 | #define F_RELAXED 0x00001000UL |
18 | #define S_MAXDEPTH 27 |
48 | #define F_ALLOW_UNKNOWN 0x00002000UL |
19 | |
49 | #define F_ALLOW_TAGS 0x00004000UL |
20 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
50 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
21 | |
|
|
22 | // F_SELFCONVERT? <=> to_json/toJson |
|
|
23 | // F_BLESSED? <=> { $__class__$ => } |
|
|
24 | |
51 | |
25 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
52 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
26 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
27 | |
53 | |
28 | #define INIT_SIZE 32 // initial scalar size to be allocated |
54 | #define INIT_SIZE 32 // initial scalar size to be allocated |
29 | #define INDENT_STEP 3 // spaces per indentation level |
55 | #define INDENT_STEP 3 // spaces per indentation level |
30 | |
56 | |
31 | #define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character |
|
|
32 | #define SHORT_STRING_LEN 512 // special-case strings of up to this size |
57 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
|
|
58 | |
|
|
59 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
33 | |
60 | |
34 | #define SB do { |
61 | #define SB do { |
35 | #define SE } while (0) |
62 | #define SE } while (0) |
36 | |
63 | |
|
|
64 | #if __GNUC__ >= 3 |
|
|
65 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
|
|
66 | # define INLINE static inline |
|
|
67 | #else |
|
|
68 | # define expect(expr,value) (expr) |
|
|
69 | # define INLINE static |
|
|
70 | #endif |
|
|
71 | |
|
|
72 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
73 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
74 | |
|
|
75 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
76 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
77 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
78 | |
|
|
79 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
80 | |
|
|
81 | #ifdef USE_ITHREADS |
|
|
82 | # define JSON_SLOW 1 |
|
|
83 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
84 | #else |
|
|
85 | # define JSON_SLOW 0 |
|
|
86 | # define JSON_STASH json_stash |
|
|
87 | #endif |
|
|
88 | |
|
|
89 | // the amount of HEs to allocate on the stack, when sorting keys |
|
|
90 | #define STACK_HES 64 |
|
|
91 | |
37 | static HV *json_stash; // JSON::XS:: |
92 | static HV *json_stash, *types_boolean_stash; // JSON::XS:: |
|
|
93 | static SV *types_true, *types_false, *sv_json; |
|
|
94 | |
|
|
95 | enum { |
|
|
96 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
97 | INCR_M_STR, // inside string |
|
|
98 | INCR_M_BS, // inside backslash |
|
|
99 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
100 | INCR_M_C1, // inside comment in other places |
|
|
101 | INCR_M_JSON // outside anything, count nesting |
|
|
102 | }; |
|
|
103 | |
|
|
104 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
105 | |
|
|
106 | typedef struct { |
|
|
107 | U32 flags; |
|
|
108 | U32 max_depth; |
|
|
109 | STRLEN max_size; |
|
|
110 | |
|
|
111 | SV *cb_object; |
|
|
112 | HV *cb_sk_object; |
|
|
113 | |
|
|
114 | // for the incremental parser |
|
|
115 | SV *incr_text; // the source text so far |
|
|
116 | STRLEN incr_pos; // the current offset into the text |
|
|
117 | int incr_nest; // {[]}-nesting level |
|
|
118 | unsigned char incr_mode; |
|
|
119 | } JSON; |
|
|
120 | |
|
|
121 | INLINE void |
|
|
122 | json_init (JSON *json) |
|
|
123 | { |
|
|
124 | Zero (json, 1, JSON); |
|
|
125 | json->max_depth = 512; |
|
|
126 | } |
38 | |
127 | |
39 | ///////////////////////////////////////////////////////////////////////////// |
128 | ///////////////////////////////////////////////////////////////////////////// |
40 | // utility functions |
129 | // utility functions |
41 | |
130 | |
42 | static UV * |
131 | INLINE SV * |
43 | SvJSON (SV *sv) |
132 | get_bool (const char *name) |
44 | { |
133 | { |
45 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
134 | SV *sv = get_sv (name, 1); |
46 | croak ("object is not of type JSON::XS"); |
|
|
47 | |
135 | |
48 | return &SvUVX (SvRV (sv)); |
136 | SvREADONLY_on (sv); |
49 | } |
137 | SvREADONLY_on (SvRV (sv)); |
50 | |
138 | |
51 | static void |
139 | return sv; |
|
|
140 | } |
|
|
141 | |
|
|
142 | INLINE void |
52 | shrink (SV *sv) |
143 | shrink (SV *sv) |
53 | { |
144 | { |
54 | sv_utf8_downgrade (sv, 1); |
145 | sv_utf8_downgrade (sv, 1); |
|
|
146 | |
55 | if (SvLEN (sv) > SvCUR (sv) + 1) |
147 | if (SvLEN (sv) > SvCUR (sv) + 1) |
56 | { |
148 | { |
57 | #ifdef SvPV_shrink_to_cur |
149 | #ifdef SvPV_shrink_to_cur |
58 | SvPV_shrink_to_cur (sv); |
150 | SvPV_shrink_to_cur (sv); |
59 | #elif defined (SvPV_renew) |
151 | #elif defined (SvPV_renew) |
… | |
… | |
65 | // decode an utf-8 character and return it, or (UV)-1 in |
157 | // decode an utf-8 character and return it, or (UV)-1 in |
66 | // case of an error. |
158 | // case of an error. |
67 | // we special-case "safe" characters from U+80 .. U+7FF, |
159 | // we special-case "safe" characters from U+80 .. U+7FF, |
68 | // but use the very good perl function to parse anything else. |
160 | // but use the very good perl function to parse anything else. |
69 | // note that we never call this function for a ascii codepoints |
161 | // note that we never call this function for a ascii codepoints |
70 | static UV |
162 | INLINE UV |
71 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
163 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
72 | { |
164 | { |
73 | if (s[0] > 0xdf || s[0] < 0xc2) |
165 | if (expect_true (len >= 2 |
74 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
166 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
75 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
167 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
76 | { |
168 | { |
77 | *clen = 2; |
169 | *clen = 2; |
78 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
170 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
79 | } |
171 | } |
80 | else |
172 | else |
81 | { |
173 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
82 | *clen = (STRLEN)-1; |
174 | } |
83 | return (UV)-1; |
175 | |
|
|
176 | // likewise for encoding, also never called for ascii codepoints |
|
|
177 | // this function takes advantage of this fact, although current gccs |
|
|
178 | // seem to optimise the check for >= 0x80 away anyways |
|
|
179 | INLINE unsigned char * |
|
|
180 | encode_utf8 (unsigned char *s, UV ch) |
|
|
181 | { |
|
|
182 | if (expect_false (ch < 0x000080)) |
|
|
183 | *s++ = ch; |
|
|
184 | else if (expect_true (ch < 0x000800)) |
|
|
185 | *s++ = 0xc0 | ( ch >> 6), |
|
|
186 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
187 | else if ( ch < 0x010000) |
|
|
188 | *s++ = 0xe0 | ( ch >> 12), |
|
|
189 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
190 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
191 | else if ( ch < 0x110000) |
|
|
192 | *s++ = 0xf0 | ( ch >> 18), |
|
|
193 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
194 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
195 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
196 | |
|
|
197 | return s; |
|
|
198 | } |
|
|
199 | |
|
|
200 | // convert offset pointer to character index, sv must be string |
|
|
201 | static STRLEN |
|
|
202 | ptr_to_index (SV *sv, char *offset) |
|
|
203 | { |
|
|
204 | return SvUTF8 (sv) |
|
|
205 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
206 | : offset - SvPVX (sv); |
|
|
207 | } |
|
|
208 | |
|
|
209 | ///////////////////////////////////////////////////////////////////////////// |
|
|
210 | // fp hell |
|
|
211 | |
|
|
212 | // scan a group of digits, and a trailing exponent |
|
|
213 | static void |
|
|
214 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
215 | { |
|
|
216 | UV uaccum = 0; |
|
|
217 | int eaccum = 0; |
|
|
218 | |
|
|
219 | // if we recurse too deep, skip all remaining digits |
|
|
220 | // to avoid a stack overflow attack |
|
|
221 | if (expect_false (--maxdepth <= 0)) |
|
|
222 | while (((U8)*s - '0') < 10) |
|
|
223 | ++s; |
|
|
224 | |
|
|
225 | for (;;) |
84 | } |
226 | { |
85 | } |
227 | U8 dig = (U8)*s - '0'; |
86 | |
228 | |
|
|
229 | if (expect_false (dig >= 10)) |
|
|
230 | { |
|
|
231 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
232 | { |
|
|
233 | ++s; |
|
|
234 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
235 | } |
|
|
236 | else if ((dig | ' ') == 'e' - '0') |
|
|
237 | { |
|
|
238 | int exp2 = 0; |
|
|
239 | int neg = 0; |
|
|
240 | |
|
|
241 | ++s; |
|
|
242 | |
|
|
243 | if (*s == '-') |
|
|
244 | { |
|
|
245 | ++s; |
|
|
246 | neg = 1; |
|
|
247 | } |
|
|
248 | else if (*s == '+') |
|
|
249 | ++s; |
|
|
250 | |
|
|
251 | while ((dig = (U8)*s - '0') < 10) |
|
|
252 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
253 | |
|
|
254 | *expo += neg ? -exp2 : exp2; |
|
|
255 | } |
|
|
256 | |
|
|
257 | break; |
|
|
258 | } |
|
|
259 | |
|
|
260 | ++s; |
|
|
261 | |
|
|
262 | uaccum = uaccum * 10 + dig; |
|
|
263 | ++eaccum; |
|
|
264 | |
|
|
265 | // if we have too many digits, then recurse for more |
|
|
266 | // we actually do this for rather few digits |
|
|
267 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
268 | { |
|
|
269 | if (postdp) *expo -= eaccum; |
|
|
270 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
271 | if (postdp) *expo += eaccum; |
|
|
272 | |
|
|
273 | break; |
|
|
274 | } |
|
|
275 | } |
|
|
276 | |
|
|
277 | // this relies greatly on the quality of the pow () |
|
|
278 | // implementation of the platform, but a good |
|
|
279 | // implementation is hard to beat. |
|
|
280 | // (IEEE 754 conformant ones are required to be exact) |
|
|
281 | if (postdp) *expo -= eaccum; |
|
|
282 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
283 | *expo += eaccum; |
|
|
284 | } |
|
|
285 | |
|
|
286 | static NV |
|
|
287 | json_atof (const char *s) |
|
|
288 | { |
|
|
289 | NV accum = 0.; |
|
|
290 | int expo = 0; |
|
|
291 | int neg = 0; |
|
|
292 | |
|
|
293 | if (*s == '-') |
|
|
294 | { |
|
|
295 | ++s; |
|
|
296 | neg = 1; |
|
|
297 | } |
|
|
298 | |
|
|
299 | // a recursion depth of ten gives us >>500 bits |
|
|
300 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
301 | |
|
|
302 | return neg ? -accum : accum; |
|
|
303 | } |
87 | ///////////////////////////////////////////////////////////////////////////// |
304 | ///////////////////////////////////////////////////////////////////////////// |
88 | // encoder |
305 | // encoder |
89 | |
306 | |
90 | // structure used for encoding JSON |
307 | // structure used for encoding JSON |
91 | typedef struct |
308 | typedef struct |
92 | { |
309 | { |
93 | char *cur; // SvPVX (sv) + current output position |
310 | char *cur; // SvPVX (sv) + current output position |
94 | char *end; // SvEND (sv) |
311 | char *end; // SvEND (sv) |
95 | SV *sv; // result scalar |
312 | SV *sv; // result scalar |
96 | U32 flags; // F_* |
313 | JSON json; |
97 | U32 indent; // indentation level |
314 | U32 indent; // indentation level |
98 | U32 maxdepth; // max. indentation/recursion level |
315 | UV limit; // escape character values >= this value when encoding |
99 | } enc_t; |
316 | } enc_t; |
100 | |
317 | |
101 | static void |
318 | INLINE void |
102 | need (enc_t *enc, STRLEN len) |
319 | need (enc_t *enc, STRLEN len) |
103 | { |
320 | { |
104 | if (enc->cur + len >= enc->end) |
321 | if (expect_false (enc->cur + len >= enc->end)) |
105 | { |
322 | { |
106 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
323 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
107 | SvGROW (enc->sv, cur + len + 1); |
324 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
108 | enc->cur = SvPVX (enc->sv) + cur; |
325 | enc->cur = SvPVX (enc->sv) + cur; |
109 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); |
326 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
110 | } |
327 | } |
111 | } |
328 | } |
112 | |
329 | |
113 | static void |
330 | INLINE void |
114 | encode_ch (enc_t *enc, char ch) |
331 | encode_ch (enc_t *enc, char ch) |
115 | { |
332 | { |
116 | need (enc, 1); |
333 | need (enc, 1); |
117 | *enc->cur++ = ch; |
334 | *enc->cur++ = ch; |
118 | } |
335 | } |
… | |
… | |
126 | |
343 | |
127 | while (str < end) |
344 | while (str < end) |
128 | { |
345 | { |
129 | unsigned char ch = *(unsigned char *)str; |
346 | unsigned char ch = *(unsigned char *)str; |
130 | |
347 | |
131 | if (ch >= 0x20 && ch < 0x80) // most common case |
348 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
132 | { |
349 | { |
133 | if (ch == '"') // but with slow exceptions |
350 | if (expect_false (ch == '"')) // but with slow exceptions |
134 | { |
351 | { |
135 | need (enc, len += 1); |
352 | need (enc, len += 1); |
136 | *enc->cur++ = '\\'; |
353 | *enc->cur++ = '\\'; |
137 | *enc->cur++ = '"'; |
354 | *enc->cur++ = '"'; |
138 | } |
355 | } |
139 | else if (ch == '\\') |
356 | else if (expect_false (ch == '\\')) |
140 | { |
357 | { |
141 | need (enc, len += 1); |
358 | need (enc, len += 1); |
142 | *enc->cur++ = '\\'; |
359 | *enc->cur++ = '\\'; |
143 | *enc->cur++ = '\\'; |
360 | *enc->cur++ = '\\'; |
144 | } |
361 | } |
… | |
… | |
162 | STRLEN clen; |
379 | STRLEN clen; |
163 | UV uch; |
380 | UV uch; |
164 | |
381 | |
165 | if (is_utf8) |
382 | if (is_utf8) |
166 | { |
383 | { |
167 | //uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
|
|
168 | uch = decode_utf8 (str, end - str, &clen); |
384 | uch = decode_utf8 (str, end - str, &clen); |
169 | if (clen == (STRLEN)-1) |
385 | if (clen == (STRLEN)-1) |
170 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
386 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
171 | } |
387 | } |
172 | else |
388 | else |
173 | { |
389 | { |
174 | uch = ch; |
390 | uch = ch; |
175 | clen = 1; |
391 | clen = 1; |
176 | } |
392 | } |
177 | |
393 | |
178 | if (uch > 0x10FFFFUL) |
394 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
179 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
180 | |
|
|
181 | if (uch < 0x80 || enc->flags & F_ASCII) |
|
|
182 | { |
395 | { |
183 | if (uch > 0xFFFFUL) |
396 | if (uch >= 0x10000UL) |
184 | { |
397 | { |
|
|
398 | if (uch >= 0x110000UL) |
|
|
399 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
400 | |
185 | need (enc, len += 11); |
401 | need (enc, len += 11); |
186 | sprintf (enc->cur, "\\u%04x\\u%04x", |
402 | sprintf (enc->cur, "\\u%04x\\u%04x", |
187 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
403 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
188 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
404 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
189 | enc->cur += 12; |
405 | enc->cur += 12; |
190 | } |
406 | } |
191 | else |
407 | else |
192 | { |
408 | { |
193 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
194 | need (enc, len += 5); |
409 | need (enc, len += 5); |
195 | *enc->cur++ = '\\'; |
410 | *enc->cur++ = '\\'; |
196 | *enc->cur++ = 'u'; |
411 | *enc->cur++ = 'u'; |
197 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
412 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
198 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
413 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
199 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
414 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
200 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
415 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
201 | } |
416 | } |
202 | |
417 | |
|
|
418 | str += clen; |
|
|
419 | } |
|
|
420 | else if (enc->json.flags & F_LATIN1) |
|
|
421 | { |
|
|
422 | *enc->cur++ = uch; |
203 | str += clen; |
423 | str += clen; |
204 | } |
424 | } |
205 | else if (is_utf8) |
425 | else if (is_utf8) |
206 | { |
426 | { |
207 | need (enc, len += clen); |
427 | need (enc, len += clen); |
… | |
… | |
211 | } |
431 | } |
212 | while (--clen); |
432 | while (--clen); |
213 | } |
433 | } |
214 | else |
434 | else |
215 | { |
435 | { |
216 | need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed |
436 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
217 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
437 | enc->cur = encode_utf8 (enc->cur, uch); |
218 | ++str; |
438 | ++str; |
219 | } |
439 | } |
220 | } |
440 | } |
221 | } |
441 | } |
222 | } |
442 | } |
223 | |
443 | |
224 | --len; |
444 | --len; |
225 | } |
445 | } |
226 | } |
446 | } |
227 | |
447 | |
228 | static void |
448 | INLINE void |
229 | encode_indent (enc_t *enc) |
449 | encode_indent (enc_t *enc) |
230 | { |
450 | { |
231 | if (enc->flags & F_INDENT) |
451 | if (enc->json.flags & F_INDENT) |
232 | { |
452 | { |
233 | int spaces = enc->indent * INDENT_STEP; |
453 | int spaces = enc->indent * INDENT_STEP; |
234 | |
454 | |
235 | need (enc, spaces); |
455 | need (enc, spaces); |
236 | memset (enc->cur, ' ', spaces); |
456 | memset (enc->cur, ' ', spaces); |
237 | enc->cur += spaces; |
457 | enc->cur += spaces; |
238 | } |
458 | } |
239 | } |
459 | } |
240 | |
460 | |
241 | static void |
461 | INLINE void |
242 | encode_space (enc_t *enc) |
462 | encode_space (enc_t *enc) |
243 | { |
463 | { |
244 | need (enc, 1); |
464 | need (enc, 1); |
245 | encode_ch (enc, ' '); |
465 | encode_ch (enc, ' '); |
246 | } |
466 | } |
247 | |
467 | |
248 | static void |
468 | INLINE void |
249 | encode_nl (enc_t *enc) |
469 | encode_nl (enc_t *enc) |
250 | { |
470 | { |
251 | if (enc->flags & F_INDENT) |
471 | if (enc->json.flags & F_INDENT) |
252 | { |
472 | { |
253 | need (enc, 1); |
473 | need (enc, 1); |
254 | encode_ch (enc, '\n'); |
474 | encode_ch (enc, '\n'); |
255 | } |
475 | } |
256 | } |
476 | } |
257 | |
477 | |
258 | static void |
478 | INLINE void |
259 | encode_comma (enc_t *enc) |
479 | encode_comma (enc_t *enc) |
260 | { |
480 | { |
261 | encode_ch (enc, ','); |
481 | encode_ch (enc, ','); |
262 | |
482 | |
263 | if (enc->flags & F_INDENT) |
483 | if (enc->json.flags & F_INDENT) |
264 | encode_nl (enc); |
484 | encode_nl (enc); |
265 | else if (enc->flags & F_SPACE_AFTER) |
485 | else if (enc->json.flags & F_SPACE_AFTER) |
266 | encode_space (enc); |
486 | encode_space (enc); |
267 | } |
487 | } |
268 | |
488 | |
269 | static void encode_sv (enc_t *enc, SV *sv); |
489 | static void encode_sv (enc_t *enc, SV *sv); |
270 | |
490 | |
271 | static void |
491 | static void |
272 | encode_av (enc_t *enc, AV *av) |
492 | encode_av (enc_t *enc, AV *av) |
273 | { |
493 | { |
274 | int i, len = av_len (av); |
494 | int i, len = av_len (av); |
275 | |
495 | |
276 | if (enc->indent >= enc->maxdepth) |
496 | if (enc->indent >= enc->json.max_depth) |
277 | croak ("data structure too deep (hit recursion limit)"); |
497 | croak (ERR_NESTING_EXCEEDED); |
278 | |
498 | |
279 | encode_ch (enc, '['); encode_nl (enc); |
499 | encode_ch (enc, '['); |
280 | ++enc->indent; |
|
|
281 | |
500 | |
|
|
501 | if (len >= 0) |
|
|
502 | { |
|
|
503 | encode_nl (enc); ++enc->indent; |
|
|
504 | |
282 | for (i = 0; i <= len; ++i) |
505 | for (i = 0; i <= len; ++i) |
283 | { |
506 | { |
|
|
507 | SV **svp = av_fetch (av, i, 0); |
|
|
508 | |
284 | encode_indent (enc); |
509 | encode_indent (enc); |
285 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
286 | |
510 | |
|
|
511 | if (svp) |
|
|
512 | encode_sv (enc, *svp); |
|
|
513 | else |
|
|
514 | encode_str (enc, "null", 4, 0); |
|
|
515 | |
287 | if (i < len) |
516 | if (i < len) |
288 | encode_comma (enc); |
517 | encode_comma (enc); |
289 | } |
518 | } |
290 | |
519 | |
|
|
520 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
521 | } |
|
|
522 | |
291 | encode_nl (enc); |
523 | encode_ch (enc, ']'); |
292 | |
|
|
293 | --enc->indent; |
|
|
294 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
295 | } |
524 | } |
296 | |
525 | |
297 | static void |
526 | static void |
298 | encode_he (enc_t *enc, HE *he) |
527 | encode_hk (enc_t *enc, HE *he) |
299 | { |
528 | { |
300 | encode_ch (enc, '"'); |
529 | encode_ch (enc, '"'); |
301 | |
530 | |
302 | if (HeKLEN (he) == HEf_SVKEY) |
531 | if (HeKLEN (he) == HEf_SVKEY) |
303 | { |
532 | { |
304 | SV *sv = HeSVKEY (he); |
533 | SV *sv = HeSVKEY (he); |
305 | STRLEN len; |
534 | STRLEN len; |
306 | char *str; |
535 | char *str; |
307 | |
536 | |
308 | SvGETMAGIC (sv); |
537 | SvGETMAGIC (sv); |
309 | str = SvPV (sv, len); |
538 | str = SvPV (sv, len); |
310 | |
539 | |
311 | encode_str (enc, str, len, SvUTF8 (sv)); |
540 | encode_str (enc, str, len, SvUTF8 (sv)); |
312 | } |
541 | } |
313 | else |
542 | else |
314 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
543 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
315 | |
544 | |
316 | encode_ch (enc, '"'); |
545 | encode_ch (enc, '"'); |
317 | |
546 | |
318 | if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
547 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
319 | encode_ch (enc, ':'); |
548 | encode_ch (enc, ':'); |
320 | if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
549 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
321 | encode_sv (enc, HeVAL (he)); |
|
|
322 | } |
550 | } |
323 | |
551 | |
324 | // compare hash entries, used when all keys are bytestrings |
552 | // compare hash entries, used when all keys are bytestrings |
325 | static int |
553 | static int |
326 | he_cmp_fast (const void *a_, const void *b_) |
554 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
331 | HE *b = *(HE **)b_; |
559 | HE *b = *(HE **)b_; |
332 | |
560 | |
333 | STRLEN la = HeKLEN (a); |
561 | STRLEN la = HeKLEN (a); |
334 | STRLEN lb = HeKLEN (b); |
562 | STRLEN lb = HeKLEN (b); |
335 | |
563 | |
336 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
564 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
337 | cmp = la - lb; |
565 | cmp = lb - la; |
338 | |
566 | |
339 | return cmp; |
567 | return cmp; |
340 | } |
568 | } |
341 | |
569 | |
342 | // compare hash entries, used when some keys are sv's or utf-x |
570 | // compare hash entries, used when some keys are sv's or utf-x |
343 | static int |
571 | static int |
344 | he_cmp_slow (const void *a, const void *b) |
572 | he_cmp_slow (const void *a, const void *b) |
345 | { |
573 | { |
346 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
574 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
347 | } |
575 | } |
348 | |
576 | |
349 | static void |
577 | static void |
350 | encode_hv (enc_t *enc, HV *hv) |
578 | encode_hv (enc_t *enc, HV *hv) |
351 | { |
579 | { |
352 | int count, i; |
580 | HE *he; |
353 | |
581 | |
354 | if (enc->indent >= enc->maxdepth) |
582 | if (enc->indent >= enc->json.max_depth) |
355 | croak ("data structure too deep (hit recursion limit)"); |
583 | croak (ERR_NESTING_EXCEEDED); |
356 | |
584 | |
357 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
585 | encode_ch (enc, '{'); |
358 | |
586 | |
359 | if ((count = hv_iterinit (hv))) |
|
|
360 | { |
|
|
361 | // for canonical output we have to sort by keys first |
587 | // for canonical output we have to sort by keys first |
362 | // actually, this is mostly due to the stupid so-called |
588 | // actually, this is mostly due to the stupid so-called |
363 | // security workaround added somewhere in 5.8.x. |
589 | // security workaround added somewhere in 5.8.x |
364 | // that randomises hash orderings |
590 | // that randomises hash orderings |
365 | if (enc->flags & F_CANONICAL) |
591 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
592 | { |
|
|
593 | int count = hv_iterinit (hv); |
|
|
594 | |
|
|
595 | if (SvMAGICAL (hv)) |
366 | { |
596 | { |
367 | HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode |
597 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
598 | // but I don't care for the speed of this special case. |
|
|
599 | // note also that we will run into undefined behaviour when the two iterations |
|
|
600 | // do not result in the same count, something I might care for in some later release. |
|
|
601 | |
|
|
602 | count = 0; |
|
|
603 | while (hv_iternext (hv)) |
|
|
604 | ++count; |
|
|
605 | |
|
|
606 | hv_iterinit (hv); |
|
|
607 | } |
|
|
608 | |
|
|
609 | if (count) |
|
|
610 | { |
368 | int fast = 1; |
611 | int i, fast = 1; |
|
|
612 | HE *hes_stack [STACK_HES]; |
|
|
613 | HE **hes = hes_stack; |
|
|
614 | |
|
|
615 | // allocate larger arrays on the heap |
|
|
616 | if (count > STACK_HES) |
|
|
617 | { |
|
|
618 | SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes))); |
|
|
619 | hes = (HE **)SvPVX (sv); |
|
|
620 | } |
369 | |
621 | |
370 | i = 0; |
622 | i = 0; |
371 | while ((he = hv_iternext (hv))) |
623 | while ((he = hv_iternext (hv))) |
372 | { |
624 | { |
373 | hes [i++] = he; |
625 | hes [i++] = he; |
… | |
… | |
395 | |
647 | |
396 | FREETMPS; |
648 | FREETMPS; |
397 | LEAVE; |
649 | LEAVE; |
398 | } |
650 | } |
399 | |
651 | |
400 | for (i = 0; i < count; ++i) |
652 | encode_nl (enc); ++enc->indent; |
|
|
653 | |
|
|
654 | while (count--) |
401 | { |
655 | { |
402 | encode_indent (enc); |
656 | encode_indent (enc); |
|
|
657 | he = hes [count]; |
403 | encode_he (enc, hes [i]); |
658 | encode_hk (enc, he); |
|
|
659 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
404 | |
660 | |
405 | if (i < count - 1) |
661 | if (count) |
406 | encode_comma (enc); |
662 | encode_comma (enc); |
407 | } |
663 | } |
408 | |
664 | |
409 | encode_nl (enc); |
665 | encode_nl (enc); --enc->indent; encode_indent (enc); |
410 | } |
666 | } |
|
|
667 | } |
411 | else |
668 | else |
|
|
669 | { |
|
|
670 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
671 | if ((he = hv_iternext (hv))) |
412 | { |
672 | { |
413 | SV *sv; |
673 | encode_nl (enc); ++enc->indent; |
414 | HE *he = hv_iternext (hv); |
|
|
415 | |
674 | |
416 | for (;;) |
675 | for (;;) |
417 | { |
676 | { |
418 | encode_indent (enc); |
677 | encode_indent (enc); |
419 | encode_he (enc, he); |
678 | encode_hk (enc, he); |
|
|
679 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
420 | |
680 | |
421 | if (!(he = hv_iternext (hv))) |
681 | if (!(he = hv_iternext (hv))) |
422 | break; |
682 | break; |
423 | |
683 | |
424 | encode_comma (enc); |
684 | encode_comma (enc); |
425 | } |
685 | } |
426 | |
686 | |
427 | encode_nl (enc); |
687 | encode_nl (enc); --enc->indent; encode_indent (enc); |
428 | } |
688 | } |
429 | } |
689 | } |
430 | |
690 | |
431 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
691 | encode_ch (enc, '}'); |
432 | } |
692 | } |
433 | |
693 | |
434 | // encode objects, arrays and special \0=false and \1=true values. |
694 | // encode objects, arrays and special \0=false and \1=true values. |
435 | static void |
695 | static void |
436 | encode_rv (enc_t *enc, SV *sv) |
696 | encode_rv (enc_t *enc, SV *sv) |
437 | { |
697 | { |
|
|
698 | svtype svt; |
|
|
699 | GV *method; |
|
|
700 | |
438 | SvGETMAGIC (sv); |
701 | SvGETMAGIC (sv); |
439 | |
|
|
440 | svtype svt = SvTYPE (sv); |
702 | svt = SvTYPE (sv); |
441 | |
703 | |
|
|
704 | if (expect_false (SvOBJECT (sv))) |
|
|
705 | { |
|
|
706 | HV *boolean_stash = !JSON_SLOW || types_boolean_stash |
|
|
707 | ? types_boolean_stash |
|
|
708 | : gv_stashpv ("Types::Serialiser::Boolean", 1); |
|
|
709 | HV *stash = SvSTASH (sv); |
|
|
710 | |
|
|
711 | if (stash == boolean_stash) |
|
|
712 | { |
|
|
713 | if (SvIV (sv)) |
|
|
714 | encode_str (enc, "true", 4, 0); |
|
|
715 | else |
|
|
716 | encode_str (enc, "false", 5, 0); |
|
|
717 | } |
|
|
718 | else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0))) |
|
|
719 | { |
|
|
720 | int count; |
|
|
721 | dSP; |
|
|
722 | |
|
|
723 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
724 | EXTEND (SP, 2); |
|
|
725 | // we re-bless the reference to get overload and other niceties right |
|
|
726 | PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
|
|
727 | PUSHs (sv_json); |
|
|
728 | |
|
|
729 | PUTBACK; |
|
|
730 | count = call_sv ((SV *)GvCV (method), G_ARRAY); |
|
|
731 | SPAGAIN; |
|
|
732 | |
|
|
733 | // catch this surprisingly common error |
|
|
734 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
735 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
736 | |
|
|
737 | encode_ch (enc, '('); |
|
|
738 | encode_ch (enc, '"'); |
|
|
739 | encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash)); |
|
|
740 | encode_ch (enc, '"'); |
|
|
741 | encode_ch (enc, ')'); |
|
|
742 | encode_ch (enc, '['); |
|
|
743 | |
|
|
744 | while (count) |
|
|
745 | { |
|
|
746 | encode_sv (enc, SP[1 - count--]); |
|
|
747 | |
|
|
748 | if (count) |
|
|
749 | encode_ch (enc, ','); |
|
|
750 | } |
|
|
751 | |
|
|
752 | encode_ch (enc, ']'); |
|
|
753 | |
|
|
754 | PUTBACK; |
|
|
755 | |
|
|
756 | FREETMPS; LEAVE; |
|
|
757 | } |
|
|
758 | else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0))) |
|
|
759 | { |
|
|
760 | dSP; |
|
|
761 | |
|
|
762 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
763 | // we re-bless the reference to get overload and other niceties right |
|
|
764 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
|
|
765 | |
|
|
766 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
767 | PUTBACK; |
|
|
768 | call_sv ((SV *)GvCV (method), G_SCALAR); |
|
|
769 | SPAGAIN; |
|
|
770 | |
|
|
771 | // catch this surprisingly common error |
|
|
772 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
773 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
774 | |
|
|
775 | sv = POPs; |
|
|
776 | PUTBACK; |
|
|
777 | |
|
|
778 | encode_sv (enc, sv); |
|
|
779 | |
|
|
780 | FREETMPS; LEAVE; |
|
|
781 | } |
|
|
782 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
783 | encode_str (enc, "null", 4, 0); |
|
|
784 | else |
|
|
785 | croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)", |
|
|
786 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
787 | } |
442 | if (svt == SVt_PVHV) |
788 | else if (svt == SVt_PVHV) |
443 | encode_hv (enc, (HV *)sv); |
789 | encode_hv (enc, (HV *)sv); |
444 | else if (svt == SVt_PVAV) |
790 | else if (svt == SVt_PVAV) |
445 | encode_av (enc, (AV *)sv); |
791 | encode_av (enc, (AV *)sv); |
446 | else if (svt < SVt_PVAV) |
792 | else if (svt < SVt_PVAV) |
447 | { |
793 | { |
448 | if (SvNIOK (sv) && SvIV (sv) == 0) |
794 | STRLEN len = 0; |
|
|
795 | char *pv = svt ? SvPV (sv, len) : 0; |
|
|
796 | |
|
|
797 | if (len == 1 && *pv == '1') |
|
|
798 | encode_str (enc, "true", 4, 0); |
|
|
799 | else if (len == 1 && *pv == '0') |
449 | encode_str (enc, "false", 5, 0); |
800 | encode_str (enc, "false", 5, 0); |
450 | else if (SvNIOK (sv) && SvIV (sv) == 1) |
801 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
451 | encode_str (enc, "true", 4, 0); |
802 | encode_str (enc, "null", 4, 0); |
452 | else |
803 | else |
453 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
804 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
454 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
805 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
455 | } |
806 | } |
|
|
807 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
808 | encode_str (enc, "null", 4, 0); |
456 | else |
809 | else |
457 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
810 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
458 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
811 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
459 | } |
812 | } |
460 | |
813 | |
… | |
… | |
471 | encode_str (enc, str, len, SvUTF8 (sv)); |
824 | encode_str (enc, str, len, SvUTF8 (sv)); |
472 | encode_ch (enc, '"'); |
825 | encode_ch (enc, '"'); |
473 | } |
826 | } |
474 | else if (SvNOKp (sv)) |
827 | else if (SvNOKp (sv)) |
475 | { |
828 | { |
|
|
829 | // trust that perl will do the right thing w.r.t. JSON syntax. |
476 | need (enc, NV_DIG + 32); |
830 | need (enc, NV_DIG + 32); |
477 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
831 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
478 | enc->cur += strlen (enc->cur); |
832 | enc->cur += strlen (enc->cur); |
479 | } |
833 | } |
480 | else if (SvIOKp (sv)) |
834 | else if (SvIOKp (sv)) |
481 | { |
835 | { |
|
|
836 | // we assume we can always read an IV as a UV and vice versa |
|
|
837 | // we assume two's complement |
|
|
838 | // we assume no aliasing issues in the union |
|
|
839 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
|
|
840 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
|
|
841 | { |
|
|
842 | // optimise the "small number case" |
|
|
843 | // code will likely be branchless and use only a single multiplication |
|
|
844 | // works for numbers up to 59074 |
|
|
845 | I32 i = SvIVX (sv); |
|
|
846 | U32 u; |
|
|
847 | char digit, nz = 0; |
|
|
848 | |
482 | need (enc, 64); |
849 | need (enc, 6); |
|
|
850 | |
|
|
851 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
852 | u = i < 0 ? -i : i; |
|
|
853 | |
|
|
854 | // convert to 4.28 fixed-point representation |
|
|
855 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
856 | |
|
|
857 | // now output digit by digit, each time masking out the integer part |
|
|
858 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
859 | // resulting in a net multiplication by 10. |
|
|
860 | // we always write the digit to memory but conditionally increment |
|
|
861 | // the pointer, to enable the use of conditional move instructions. |
|
|
862 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
|
|
863 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
|
|
864 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
|
|
865 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
|
|
866 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
867 | } |
|
|
868 | else |
|
|
869 | { |
|
|
870 | // large integer, use the (rather slow) snprintf way. |
|
|
871 | need (enc, IVUV_MAXCHARS); |
483 | enc->cur += |
872 | enc->cur += |
484 | SvIsUV(sv) |
873 | SvIsUV(sv) |
485 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
874 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
486 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
875 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
876 | } |
487 | } |
877 | } |
488 | else if (SvROK (sv)) |
878 | else if (SvROK (sv)) |
489 | encode_rv (enc, SvRV (sv)); |
879 | encode_rv (enc, SvRV (sv)); |
490 | else if (!SvOK (sv)) |
880 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
491 | encode_str (enc, "null", 4, 0); |
881 | encode_str (enc, "null", 4, 0); |
492 | else |
882 | else |
493 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
883 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data", |
494 | SvPV_nolen (sv), SvFLAGS (sv)); |
884 | SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
495 | } |
885 | } |
496 | |
886 | |
497 | static SV * |
887 | static SV * |
498 | encode_json (SV *scalar, U32 flags) |
888 | encode_json (SV *scalar, JSON *json) |
499 | { |
889 | { |
|
|
890 | enc_t enc; |
|
|
891 | |
500 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
892 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
501 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
893 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
502 | |
894 | |
503 | enc_t enc; |
895 | enc.json = *json; |
504 | enc.flags = flags; |
|
|
505 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
896 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
506 | enc.cur = SvPVX (enc.sv); |
897 | enc.cur = SvPVX (enc.sv); |
507 | enc.end = SvEND (enc.sv); |
898 | enc.end = SvEND (enc.sv); |
508 | enc.indent = 0; |
899 | enc.indent = 0; |
509 | enc.maxdepth = DEC_DEPTH (flags); |
900 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
901 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
902 | : 0x110000UL; |
510 | |
903 | |
511 | SvPOK_only (enc.sv); |
904 | SvPOK_only (enc.sv); |
512 | encode_sv (&enc, scalar); |
905 | encode_sv (&enc, scalar); |
|
|
906 | encode_nl (&enc); |
513 | |
907 | |
|
|
908 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
909 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
910 | |
514 | if (!(flags & (F_ASCII | F_UTF8))) |
911 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
515 | SvUTF8_on (enc.sv); |
912 | SvUTF8_on (enc.sv); |
516 | |
913 | |
517 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
518 | |
|
|
519 | if (enc.flags & F_SHRINK) |
914 | if (enc.json.flags & F_SHRINK) |
520 | shrink (enc.sv); |
915 | shrink (enc.sv); |
521 | |
916 | |
522 | return enc.sv; |
917 | return enc.sv; |
523 | } |
918 | } |
524 | |
919 | |
… | |
… | |
529 | typedef struct |
924 | typedef struct |
530 | { |
925 | { |
531 | char *cur; // current parser pointer |
926 | char *cur; // current parser pointer |
532 | char *end; // end of input string |
927 | char *end; // end of input string |
533 | const char *err; // parse error, if != 0 |
928 | const char *err; // parse error, if != 0 |
534 | U32 flags; // F_* |
929 | JSON json; |
535 | U32 depth; // recursion depth |
930 | U32 depth; // recursion depth |
536 | U32 maxdepth; // recursion depth limit |
931 | U32 maxdepth; // recursion depth limit |
537 | } dec_t; |
932 | } dec_t; |
538 | |
933 | |
539 | static void |
934 | INLINE void |
|
|
935 | decode_comment (dec_t *dec) |
|
|
936 | { |
|
|
937 | // only '#'-style comments allowed a.t.m. |
|
|
938 | |
|
|
939 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
940 | ++dec->cur; |
|
|
941 | } |
|
|
942 | |
|
|
943 | INLINE void |
540 | decode_ws (dec_t *dec) |
944 | decode_ws (dec_t *dec) |
541 | { |
945 | { |
542 | for (;;) |
946 | for (;;) |
543 | { |
947 | { |
544 | char ch = *dec->cur; |
948 | char ch = *dec->cur; |
545 | |
949 | |
546 | if (ch > 0x20 |
950 | if (ch > 0x20) |
|
|
951 | { |
|
|
952 | if (expect_false (ch == '#')) |
|
|
953 | { |
|
|
954 | if (dec->json.flags & F_RELAXED) |
|
|
955 | decode_comment (dec); |
|
|
956 | else |
|
|
957 | break; |
|
|
958 | } |
|
|
959 | else |
|
|
960 | break; |
|
|
961 | } |
547 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
962 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
548 | break; |
963 | break; // parse error, but let higher level handle it, gives better error messages |
549 | |
964 | |
550 | ++dec->cur; |
965 | ++dec->cur; |
551 | } |
966 | } |
552 | } |
967 | } |
553 | |
968 | |
… | |
… | |
557 | if (*dec->cur != ch) \ |
972 | if (*dec->cur != ch) \ |
558 | ERR (# ch " expected"); \ |
973 | ERR (# ch " expected"); \ |
559 | ++dec->cur; \ |
974 | ++dec->cur; \ |
560 | SE |
975 | SE |
561 | |
976 | |
562 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
977 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
563 | #define DEC_DEC_DEPTH --dec->depth |
978 | #define DEC_DEC_DEPTH --dec->depth |
564 | |
979 | |
565 | static SV *decode_sv (dec_t *dec); |
980 | static SV *decode_sv (dec_t *dec); |
566 | |
981 | |
567 | static signed char decode_hexdigit[256]; |
982 | static signed char decode_hexdigit[256]; |
… | |
… | |
570 | decode_4hex (dec_t *dec) |
985 | decode_4hex (dec_t *dec) |
571 | { |
986 | { |
572 | signed char d1, d2, d3, d4; |
987 | signed char d1, d2, d3, d4; |
573 | unsigned char *cur = (unsigned char *)dec->cur; |
988 | unsigned char *cur = (unsigned char *)dec->cur; |
574 | |
989 | |
575 | d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); |
990 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
576 | d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); |
991 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
577 | d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); |
992 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
578 | d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); |
993 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
579 | |
994 | |
580 | dec->cur += 4; |
995 | dec->cur += 4; |
581 | |
996 | |
582 | return ((UV)d1) << 12 |
997 | return ((UV)d1) << 12 |
583 | | ((UV)d2) << 8 |
998 | | ((UV)d2) << 8 |
… | |
… | |
591 | static SV * |
1006 | static SV * |
592 | decode_str (dec_t *dec) |
1007 | decode_str (dec_t *dec) |
593 | { |
1008 | { |
594 | SV *sv = 0; |
1009 | SV *sv = 0; |
595 | int utf8 = 0; |
1010 | int utf8 = 0; |
|
|
1011 | char *dec_cur = dec->cur; |
596 | |
1012 | |
597 | do |
1013 | do |
598 | { |
1014 | { |
599 | char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; |
1015 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
600 | char *cur = buf; |
1016 | char *cur = buf; |
601 | |
1017 | |
602 | do |
1018 | do |
603 | { |
1019 | { |
604 | unsigned char ch = *(unsigned char *)dec->cur++; |
1020 | unsigned char ch = *(unsigned char *)dec_cur++; |
605 | |
1021 | |
606 | if (ch == '"') |
1022 | if (expect_false (ch == '"')) |
607 | { |
1023 | { |
608 | --dec->cur; |
1024 | --dec_cur; |
609 | break; |
1025 | break; |
610 | } |
1026 | } |
611 | else if (ch == '\\') |
1027 | else if (expect_false (ch == '\\')) |
612 | { |
1028 | { |
613 | switch (*dec->cur) |
1029 | switch (*dec_cur) |
614 | { |
1030 | { |
615 | case '\\': |
1031 | case '\\': |
616 | case '/': |
1032 | case '/': |
617 | case '"': *cur++ = *dec->cur++; break; |
1033 | case '"': *cur++ = *dec_cur++; break; |
618 | |
1034 | |
619 | case 'b': ++dec->cur; *cur++ = '\010'; break; |
1035 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
620 | case 't': ++dec->cur; *cur++ = '\011'; break; |
1036 | case 't': ++dec_cur; *cur++ = '\011'; break; |
621 | case 'n': ++dec->cur; *cur++ = '\012'; break; |
1037 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
622 | case 'f': ++dec->cur; *cur++ = '\014'; break; |
1038 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
623 | case 'r': ++dec->cur; *cur++ = '\015'; break; |
1039 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
624 | |
1040 | |
625 | case 'u': |
1041 | case 'u': |
626 | { |
1042 | { |
627 | UV lo, hi; |
1043 | UV lo, hi; |
628 | ++dec->cur; |
1044 | ++dec_cur; |
629 | |
1045 | |
|
|
1046 | dec->cur = dec_cur; |
630 | hi = decode_4hex (dec); |
1047 | hi = decode_4hex (dec); |
|
|
1048 | dec_cur = dec->cur; |
631 | if (hi == (UV)-1) |
1049 | if (hi == (UV)-1) |
632 | goto fail; |
1050 | goto fail; |
633 | |
1051 | |
634 | // possibly a surrogate pair |
1052 | // possibly a surrogate pair |
635 | if (hi >= 0xd800) |
1053 | if (hi >= 0xd800) |
636 | if (hi < 0xdc00) |
1054 | if (hi < 0xdc00) |
637 | { |
1055 | { |
638 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
1056 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
639 | ERR ("missing low surrogate character in surrogate pair"); |
1057 | ERR ("missing low surrogate character in surrogate pair"); |
640 | |
1058 | |
641 | dec->cur += 2; |
1059 | dec_cur += 2; |
642 | |
1060 | |
|
|
1061 | dec->cur = dec_cur; |
643 | lo = decode_4hex (dec); |
1062 | lo = decode_4hex (dec); |
|
|
1063 | dec_cur = dec->cur; |
644 | if (lo == (UV)-1) |
1064 | if (lo == (UV)-1) |
645 | goto fail; |
1065 | goto fail; |
646 | |
1066 | |
647 | if (lo < 0xdc00 || lo >= 0xe000) |
1067 | if (lo < 0xdc00 || lo >= 0xe000) |
648 | ERR ("surrogate pair expected"); |
1068 | ERR ("surrogate pair expected"); |
… | |
… | |
654 | |
1074 | |
655 | if (hi >= 0x80) |
1075 | if (hi >= 0x80) |
656 | { |
1076 | { |
657 | utf8 = 1; |
1077 | utf8 = 1; |
658 | |
1078 | |
659 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
1079 | cur = encode_utf8 (cur, hi); |
660 | } |
1080 | } |
661 | else |
1081 | else |
662 | *cur++ = hi; |
1082 | *cur++ = hi; |
663 | } |
1083 | } |
664 | break; |
1084 | break; |
665 | |
1085 | |
666 | default: |
1086 | default: |
667 | --dec->cur; |
1087 | --dec_cur; |
668 | ERR ("illegal backslash escape sequence in string"); |
1088 | ERR ("illegal backslash escape sequence in string"); |
669 | } |
1089 | } |
670 | } |
1090 | } |
671 | else if (ch >= 0x20 && ch <= 0x7f) |
1091 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
672 | *cur++ = ch; |
1092 | *cur++ = ch; |
673 | else if (ch >= 0x80) |
1093 | else if (ch >= 0x80) |
674 | { |
1094 | { |
675 | --dec->cur; |
|
|
676 | |
|
|
677 | STRLEN clen; |
1095 | STRLEN clen; |
|
|
1096 | |
|
|
1097 | --dec_cur; |
|
|
1098 | |
678 | UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); |
1099 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
679 | if (clen == (STRLEN)-1) |
1100 | if (clen == (STRLEN)-1) |
680 | ERR ("malformed UTF-8 character in JSON string"); |
1101 | ERR ("malformed UTF-8 character in JSON string"); |
681 | |
1102 | |
682 | do |
1103 | do |
683 | *cur++ = *dec->cur++; |
1104 | *cur++ = *dec_cur++; |
684 | while (--clen); |
1105 | while (--clen); |
685 | |
1106 | |
686 | utf8 = 1; |
1107 | utf8 = 1; |
687 | } |
1108 | } |
688 | else |
1109 | else |
689 | { |
1110 | { |
690 | --dec->cur; |
1111 | --dec_cur; |
691 | |
1112 | |
692 | if (!ch) |
1113 | if (!ch) |
693 | ERR ("unexpected end of string while parsing JSON string"); |
1114 | ERR ("unexpected end of string while parsing JSON string"); |
694 | else |
1115 | else |
695 | ERR ("invalid character encountered while parsing JSON string"); |
1116 | ERR ("invalid character encountered while parsing JSON string"); |
696 | } |
1117 | } |
697 | } |
1118 | } |
698 | while (cur < buf + SHORT_STRING_LEN); |
1119 | while (cur < buf + SHORT_STRING_LEN); |
699 | |
1120 | |
|
|
1121 | { |
700 | STRLEN len = cur - buf; |
1122 | STRLEN len = cur - buf; |
701 | |
1123 | |
702 | if (sv) |
1124 | if (sv) |
703 | { |
1125 | { |
704 | SvGROW (sv, SvCUR (sv) + len + 1); |
1126 | STRLEN cur = SvCUR (sv); |
|
|
1127 | |
|
|
1128 | if (SvLEN (sv) <= cur + len) |
|
|
1129 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1130 | |
705 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1131 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
706 | SvCUR_set (sv, SvCUR (sv) + len); |
1132 | SvCUR_set (sv, SvCUR (sv) + len); |
707 | } |
1133 | } |
708 | else |
1134 | else |
709 | sv = newSVpvn (buf, len); |
1135 | sv = newSVpvn (buf, len); |
710 | } |
1136 | } |
|
|
1137 | } |
711 | while (*dec->cur != '"'); |
1138 | while (*dec_cur != '"'); |
712 | |
1139 | |
713 | ++dec->cur; |
1140 | ++dec_cur; |
714 | |
1141 | |
715 | if (sv) |
1142 | if (sv) |
716 | { |
1143 | { |
717 | SvPOK_only (sv); |
1144 | SvPOK_only (sv); |
718 | *SvEND (sv) = 0; |
1145 | *SvEND (sv) = 0; |
… | |
… | |
721 | SvUTF8_on (sv); |
1148 | SvUTF8_on (sv); |
722 | } |
1149 | } |
723 | else |
1150 | else |
724 | sv = newSVpvn ("", 0); |
1151 | sv = newSVpvn ("", 0); |
725 | |
1152 | |
|
|
1153 | dec->cur = dec_cur; |
726 | return sv; |
1154 | return sv; |
727 | |
1155 | |
728 | fail: |
1156 | fail: |
|
|
1157 | dec->cur = dec_cur; |
729 | return 0; |
1158 | return 0; |
730 | } |
1159 | } |
731 | |
1160 | |
732 | static SV * |
1161 | static SV * |
733 | decode_num (dec_t *dec) |
1162 | decode_num (dec_t *dec) |
… | |
… | |
791 | is_nv = 1; |
1220 | is_nv = 1; |
792 | } |
1221 | } |
793 | |
1222 | |
794 | if (!is_nv) |
1223 | if (!is_nv) |
795 | { |
1224 | { |
796 | UV uv; |
1225 | int len = dec->cur - start; |
797 | int numtype = grok_number (start, dec->cur - start, &uv); |
1226 | |
798 | if (numtype & IS_NUMBER_IN_UV) |
1227 | // special case the rather common 1..5-digit-int case |
799 | if (numtype & IS_NUMBER_NEG) |
1228 | if (*start == '-') |
|
|
1229 | switch (len) |
800 | { |
1230 | { |
801 | if (uv < (UV)IV_MIN) |
1231 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
802 | return newSViv (-(IV)uv); |
1232 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
|
|
1233 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
1234 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1235 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
803 | } |
1236 | } |
|
|
1237 | else |
|
|
1238 | switch (len) |
|
|
1239 | { |
|
|
1240 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
1241 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
1242 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
1243 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1244 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
|
|
1245 | } |
|
|
1246 | |
|
|
1247 | { |
|
|
1248 | UV uv; |
|
|
1249 | int numtype = grok_number (start, len, &uv); |
|
|
1250 | if (numtype & IS_NUMBER_IN_UV) |
|
|
1251 | if (numtype & IS_NUMBER_NEG) |
|
|
1252 | { |
|
|
1253 | if (uv < (UV)IV_MIN) |
|
|
1254 | return newSViv (-(IV)uv); |
|
|
1255 | } |
804 | else |
1256 | else |
805 | return newSVuv (uv); |
1257 | return newSVuv (uv); |
806 | } |
1258 | } |
807 | |
1259 | |
|
|
1260 | len -= *start == '-' ? 1 : 0; |
|
|
1261 | |
|
|
1262 | // does not fit into IV or UV, try NV |
|
|
1263 | if (len <= NV_DIG) |
|
|
1264 | // fits into NV without loss of precision |
|
|
1265 | return newSVnv (json_atof (start)); |
|
|
1266 | |
|
|
1267 | // everything else fails, convert it to a string |
|
|
1268 | return newSVpvn (start, dec->cur - start); |
|
|
1269 | } |
|
|
1270 | |
|
|
1271 | // loss of precision here |
808 | return newSVnv (Atof (start)); |
1272 | return newSVnv (json_atof (start)); |
809 | |
1273 | |
810 | fail: |
1274 | fail: |
811 | return 0; |
1275 | return 0; |
812 | } |
1276 | } |
813 | |
1277 | |
… | |
… | |
837 | if (*dec->cur == ']') |
1301 | if (*dec->cur == ']') |
838 | { |
1302 | { |
839 | ++dec->cur; |
1303 | ++dec->cur; |
840 | break; |
1304 | break; |
841 | } |
1305 | } |
842 | |
1306 | |
843 | if (*dec->cur != ',') |
1307 | if (*dec->cur != ',') |
844 | ERR (", or ] expected while parsing array"); |
1308 | ERR (", or ] expected while parsing array"); |
845 | |
1309 | |
846 | ++dec->cur; |
1310 | ++dec->cur; |
|
|
1311 | |
|
|
1312 | decode_ws (dec); |
|
|
1313 | |
|
|
1314 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1315 | { |
|
|
1316 | ++dec->cur; |
|
|
1317 | break; |
|
|
1318 | } |
847 | } |
1319 | } |
848 | |
1320 | |
849 | DEC_DEC_DEPTH; |
1321 | DEC_DEC_DEPTH; |
850 | return newRV_noinc ((SV *)av); |
1322 | return newRV_noinc ((SV *)av); |
851 | |
1323 | |
… | |
… | |
856 | } |
1328 | } |
857 | |
1329 | |
858 | static SV * |
1330 | static SV * |
859 | decode_hv (dec_t *dec) |
1331 | decode_hv (dec_t *dec) |
860 | { |
1332 | { |
|
|
1333 | SV *sv; |
861 | HV *hv = newHV (); |
1334 | HV *hv = newHV (); |
862 | |
1335 | |
863 | DEC_INC_DEPTH; |
1336 | DEC_INC_DEPTH; |
864 | decode_ws (dec); |
1337 | decode_ws (dec); |
865 | |
1338 | |
866 | if (*dec->cur == '}') |
1339 | if (*dec->cur == '}') |
867 | ++dec->cur; |
1340 | ++dec->cur; |
868 | else |
1341 | else |
869 | for (;;) |
1342 | for (;;) |
870 | { |
1343 | { |
|
|
1344 | EXPECT_CH ('"'); |
|
|
1345 | |
|
|
1346 | // heuristic: assume that |
|
|
1347 | // a) decode_str + hv_store_ent are abysmally slow. |
|
|
1348 | // b) most hash keys are short, simple ascii text. |
|
|
1349 | // => try to "fast-match" such strings to avoid |
|
|
1350 | // the overhead of decode_str + hv_store_ent. |
|
|
1351 | { |
871 | SV *key, *value; |
1352 | SV *value; |
|
|
1353 | char *p = dec->cur; |
|
|
1354 | char *e = p + 24; // only try up to 24 bytes |
872 | |
1355 | |
873 | decode_ws (dec); EXPECT_CH ('"'); |
1356 | for (;;) |
874 | |
|
|
875 | key = decode_str (dec); |
|
|
876 | if (!key) |
|
|
877 | goto fail; |
|
|
878 | |
|
|
879 | decode_ws (dec); EXPECT_CH (':'); |
|
|
880 | |
|
|
881 | value = decode_sv (dec); |
|
|
882 | if (!value) |
|
|
883 | { |
1357 | { |
|
|
1358 | // the >= 0x80 is false on most architectures |
|
|
1359 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1360 | { |
|
|
1361 | // slow path, back up and use decode_str |
|
|
1362 | SV *key = decode_str (dec); |
|
|
1363 | if (!key) |
|
|
1364 | goto fail; |
|
|
1365 | |
|
|
1366 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1367 | |
|
|
1368 | decode_ws (dec); |
|
|
1369 | value = decode_sv (dec); |
|
|
1370 | if (!value) |
|
|
1371 | { |
|
|
1372 | SvREFCNT_dec (key); |
|
|
1373 | goto fail; |
|
|
1374 | } |
|
|
1375 | |
|
|
1376 | hv_store_ent (hv, key, value, 0); |
884 | SvREFCNT_dec (key); |
1377 | SvREFCNT_dec (key); |
|
|
1378 | |
|
|
1379 | break; |
|
|
1380 | } |
|
|
1381 | else if (*p == '"') |
|
|
1382 | { |
|
|
1383 | // fast path, got a simple key |
|
|
1384 | char *key = dec->cur; |
|
|
1385 | int len = p - key; |
|
|
1386 | dec->cur = p + 1; |
|
|
1387 | |
|
|
1388 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1389 | |
|
|
1390 | decode_ws (dec); |
|
|
1391 | value = decode_sv (dec); |
|
|
1392 | if (!value) |
885 | goto fail; |
1393 | goto fail; |
|
|
1394 | |
|
|
1395 | hv_store (hv, key, len, value, 0); |
|
|
1396 | |
|
|
1397 | break; |
|
|
1398 | } |
|
|
1399 | |
|
|
1400 | ++p; |
886 | } |
1401 | } |
887 | |
1402 | } |
888 | hv_store_ent (hv, key, value, 0); |
|
|
889 | SvREFCNT_dec (key); |
|
|
890 | |
1403 | |
891 | decode_ws (dec); |
1404 | decode_ws (dec); |
892 | |
1405 | |
893 | if (*dec->cur == '}') |
1406 | if (*dec->cur == '}') |
894 | { |
1407 | { |
… | |
… | |
898 | |
1411 | |
899 | if (*dec->cur != ',') |
1412 | if (*dec->cur != ',') |
900 | ERR (", or } expected while parsing object/hash"); |
1413 | ERR (", or } expected while parsing object/hash"); |
901 | |
1414 | |
902 | ++dec->cur; |
1415 | ++dec->cur; |
|
|
1416 | |
|
|
1417 | decode_ws (dec); |
|
|
1418 | |
|
|
1419 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1420 | { |
|
|
1421 | ++dec->cur; |
|
|
1422 | break; |
|
|
1423 | } |
903 | } |
1424 | } |
904 | |
1425 | |
905 | DEC_DEC_DEPTH; |
1426 | DEC_DEC_DEPTH; |
906 | return newRV_noinc ((SV *)hv); |
1427 | sv = newRV_noinc ((SV *)hv); |
|
|
1428 | |
|
|
1429 | // check filter callbacks |
|
|
1430 | if (dec->json.flags & F_HOOK) |
|
|
1431 | { |
|
|
1432 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1433 | { |
|
|
1434 | HE *cb, *he; |
|
|
1435 | |
|
|
1436 | hv_iterinit (hv); |
|
|
1437 | he = hv_iternext (hv); |
|
|
1438 | hv_iterinit (hv); |
|
|
1439 | |
|
|
1440 | // the next line creates a mortal sv each time its called. |
|
|
1441 | // might want to optimise this for common cases. |
|
|
1442 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1443 | |
|
|
1444 | if (cb) |
|
|
1445 | { |
|
|
1446 | dSP; |
|
|
1447 | int count; |
|
|
1448 | |
|
|
1449 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1450 | XPUSHs (HeVAL (he)); |
|
|
1451 | sv_2mortal (sv); |
|
|
1452 | |
|
|
1453 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1454 | |
|
|
1455 | if (count == 1) |
|
|
1456 | { |
|
|
1457 | sv = newSVsv (POPs); |
|
|
1458 | FREETMPS; LEAVE; |
|
|
1459 | return sv; |
|
|
1460 | } |
|
|
1461 | |
|
|
1462 | SvREFCNT_inc (sv); |
|
|
1463 | FREETMPS; LEAVE; |
|
|
1464 | } |
|
|
1465 | } |
|
|
1466 | |
|
|
1467 | if (dec->json.cb_object) |
|
|
1468 | { |
|
|
1469 | dSP; |
|
|
1470 | int count; |
|
|
1471 | |
|
|
1472 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1473 | XPUSHs (sv_2mortal (sv)); |
|
|
1474 | |
|
|
1475 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1476 | |
|
|
1477 | if (count == 1) |
|
|
1478 | { |
|
|
1479 | sv = newSVsv (POPs); |
|
|
1480 | FREETMPS; LEAVE; |
|
|
1481 | return sv; |
|
|
1482 | } |
|
|
1483 | |
|
|
1484 | SvREFCNT_inc (sv); |
|
|
1485 | FREETMPS; LEAVE; |
|
|
1486 | } |
|
|
1487 | } |
|
|
1488 | |
|
|
1489 | return sv; |
907 | |
1490 | |
908 | fail: |
1491 | fail: |
909 | SvREFCNT_dec (hv); |
1492 | SvREFCNT_dec (hv); |
910 | DEC_DEC_DEPTH; |
1493 | DEC_DEC_DEPTH; |
911 | return 0; |
1494 | return 0; |
912 | } |
1495 | } |
913 | |
1496 | |
914 | static SV * |
1497 | static SV * |
|
|
1498 | decode_tag (dec_t *dec) |
|
|
1499 | { |
|
|
1500 | SV *tag = 0; |
|
|
1501 | SV *val = 0; |
|
|
1502 | |
|
|
1503 | if (!(dec->json.flags & F_ALLOW_TAGS)) |
|
|
1504 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
|
|
1505 | |
|
|
1506 | ++dec->cur; |
|
|
1507 | |
|
|
1508 | decode_ws (dec); |
|
|
1509 | |
|
|
1510 | tag = decode_sv (dec); |
|
|
1511 | if (!tag) |
|
|
1512 | goto fail; |
|
|
1513 | |
|
|
1514 | if (!SvPOK (tag)) |
|
|
1515 | ERR ("malformed JSON string, (tag) must be a string"); |
|
|
1516 | |
|
|
1517 | decode_ws (dec); |
|
|
1518 | |
|
|
1519 | if (*dec->cur != ')') |
|
|
1520 | ERR (") expected after tag"); |
|
|
1521 | |
|
|
1522 | ++dec->cur; |
|
|
1523 | |
|
|
1524 | decode_ws (dec); |
|
|
1525 | |
|
|
1526 | val = decode_sv (dec); |
|
|
1527 | if (!val) |
|
|
1528 | goto fail; |
|
|
1529 | |
|
|
1530 | if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV) |
|
|
1531 | ERR ("malformed JSON string, tag value must be an array"); |
|
|
1532 | |
|
|
1533 | { |
|
|
1534 | AV *av = (AV *)SvRV (val); |
|
|
1535 | int i, len = av_len (av) + 1; |
|
|
1536 | HV *stash = gv_stashsv (tag, 0); |
|
|
1537 | SV *sv; |
|
|
1538 | |
|
|
1539 | if (!stash) |
|
|
1540 | ERR ("cannot decode perl-object (package does not exist)"); |
|
|
1541 | |
|
|
1542 | GV *method = gv_fetchmethod_autoload (stash, "THAW", 0); |
|
|
1543 | |
|
|
1544 | if (!method) |
|
|
1545 | ERR ("cannot decode perl-object (package does not have a THAW method)"); |
|
|
1546 | |
|
|
1547 | dSP; |
|
|
1548 | |
|
|
1549 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1550 | EXTEND (SP, len + 2); |
|
|
1551 | // we re-bless the reference to get overload and other niceties right |
|
|
1552 | PUSHs (tag); |
|
|
1553 | PUSHs (sv_json); |
|
|
1554 | |
|
|
1555 | for (i = 0; i < len; ++i) |
|
|
1556 | PUSHs (*av_fetch (av, i, 1)); |
|
|
1557 | |
|
|
1558 | PUTBACK; |
|
|
1559 | call_sv ((SV *)GvCV (method), G_SCALAR); |
|
|
1560 | SPAGAIN; |
|
|
1561 | |
|
|
1562 | SvREFCNT_dec (tag); |
|
|
1563 | SvREFCNT_dec (val); |
|
|
1564 | sv = SvREFCNT_inc (POPs); |
|
|
1565 | |
|
|
1566 | PUTBACK; |
|
|
1567 | |
|
|
1568 | FREETMPS; LEAVE; |
|
|
1569 | |
|
|
1570 | return sv; |
|
|
1571 | } |
|
|
1572 | |
|
|
1573 | fail: |
|
|
1574 | SvREFCNT_dec (tag); |
|
|
1575 | SvREFCNT_dec (val); |
|
|
1576 | return 0; |
|
|
1577 | } |
|
|
1578 | |
|
|
1579 | static SV * |
915 | decode_sv (dec_t *dec) |
1580 | decode_sv (dec_t *dec) |
916 | { |
1581 | { |
917 | decode_ws (dec); |
1582 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1583 | // to parse everything. |
918 | switch (*dec->cur) |
1584 | switch (*dec->cur) |
919 | { |
1585 | { |
920 | case '"': ++dec->cur; return decode_str (dec); |
1586 | case '"': ++dec->cur; return decode_str (dec); |
921 | case '[': ++dec->cur; return decode_av (dec); |
1587 | case '[': ++dec->cur; return decode_av (dec); |
922 | case '{': ++dec->cur; return decode_hv (dec); |
1588 | case '{': ++dec->cur; return decode_hv (dec); |
|
|
1589 | case '(': return decode_tag (dec); |
923 | |
1590 | |
924 | case '-': |
1591 | case '-': |
925 | case '0': case '1': case '2': case '3': case '4': |
1592 | case '0': case '1': case '2': case '3': case '4': |
926 | case '5': case '6': case '7': case '8': case '9': |
1593 | case '5': case '6': case '7': case '8': case '9': |
927 | return decode_num (dec); |
1594 | return decode_num (dec); |
928 | |
1595 | |
929 | case 't': |
1596 | case 't': |
930 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1597 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
931 | { |
1598 | { |
932 | dec->cur += 4; |
1599 | dec->cur += 4; |
|
|
1600 | #if JSON_SLOW |
|
|
1601 | types_true = get_bool ("Types::Serialiser::true"); |
|
|
1602 | #endif |
933 | return newSViv (1); |
1603 | return newSVsv (types_true); |
934 | } |
1604 | } |
935 | else |
1605 | else |
936 | ERR ("'true' expected"); |
1606 | ERR ("'true' expected"); |
937 | |
1607 | |
938 | break; |
1608 | break; |
939 | |
1609 | |
940 | case 'f': |
1610 | case 'f': |
941 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1611 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
942 | { |
1612 | { |
943 | dec->cur += 5; |
1613 | dec->cur += 5; |
|
|
1614 | #if JSON_SLOW |
|
|
1615 | types_false = get_bool ("Types::Serialiser::false"); |
|
|
1616 | #endif |
944 | return newSViv (0); |
1617 | return newSVsv (types_false); |
945 | } |
1618 | } |
946 | else |
1619 | else |
947 | ERR ("'false' expected"); |
1620 | ERR ("'false' expected"); |
948 | |
1621 | |
949 | break; |
1622 | break; |
… | |
… | |
958 | ERR ("'null' expected"); |
1631 | ERR ("'null' expected"); |
959 | |
1632 | |
960 | break; |
1633 | break; |
961 | |
1634 | |
962 | default: |
1635 | default: |
963 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
1636 | ERR ("malformed JSON string, neither tag, array, object, number, string or atom"); |
964 | break; |
1637 | break; |
965 | } |
1638 | } |
966 | |
1639 | |
967 | fail: |
1640 | fail: |
968 | return 0; |
1641 | return 0; |
969 | } |
1642 | } |
970 | |
1643 | |
971 | static SV * |
1644 | static SV * |
972 | decode_json (SV *string, U32 flags) |
1645 | decode_json (SV *string, JSON *json, char **offset_return) |
973 | { |
1646 | { |
|
|
1647 | dec_t dec; |
974 | SV *sv; |
1648 | SV *sv; |
975 | |
1649 | |
|
|
1650 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1651 | * makes perl ignore the magic in subsequent accesses. |
|
|
1652 | * also make a copy of non-PV values, to get them into a clean |
|
|
1653 | * state (SvPV should do that, but it's buggy, see below). |
|
|
1654 | */ |
|
|
1655 | /*SvGETMAGIC (string);*/ |
|
|
1656 | if (SvMAGICAL (string) || !SvPOK (string)) |
|
|
1657 | string = sv_2mortal (newSVsv (string)); |
|
|
1658 | |
976 | SvUPGRADE (string, SVt_PV); |
1659 | SvUPGRADE (string, SVt_PV); |
977 | |
1660 | |
978 | if (flags & F_UTF8) |
1661 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1662 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1663 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1664 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1665 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1666 | * and hope for the best. |
|
|
1667 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1668 | * assertion business is seriously broken, try yet another workaround |
|
|
1669 | * for the broken -DDEBUGGING. |
|
|
1670 | */ |
|
|
1671 | { |
|
|
1672 | #ifdef DEBUGGING |
|
|
1673 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1674 | #else |
|
|
1675 | STRLEN offset = SvCUR (string); |
|
|
1676 | #endif |
|
|
1677 | |
|
|
1678 | if (offset > json->max_size && json->max_size) |
|
|
1679 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1680 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1681 | } |
|
|
1682 | |
|
|
1683 | if (DECODE_WANTS_OCTETS (json)) |
979 | sv_utf8_downgrade (string, 0); |
1684 | sv_utf8_downgrade (string, 0); |
980 | else |
1685 | else |
981 | sv_utf8_upgrade (string); |
1686 | sv_utf8_upgrade (string); |
982 | |
1687 | |
983 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1688 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
984 | |
1689 | |
985 | dec_t dec; |
1690 | dec.json = *json; |
986 | dec.flags = flags; |
|
|
987 | dec.cur = SvPVX (string); |
1691 | dec.cur = SvPVX (string); |
988 | dec.end = SvEND (string); |
1692 | dec.end = SvEND (string); |
989 | dec.err = 0; |
1693 | dec.err = 0; |
990 | dec.depth = 0; |
1694 | dec.depth = 0; |
991 | dec.maxdepth = DEC_DEPTH (dec.flags); |
|
|
992 | |
1695 | |
|
|
1696 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1697 | dec.json.flags |= F_HOOK; |
|
|
1698 | |
993 | *dec.end = 0; // this should basically be a nop, too, but make sure its there |
1699 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1700 | |
|
|
1701 | decode_ws (&dec); |
994 | sv = decode_sv (&dec); |
1702 | sv = decode_sv (&dec); |
995 | |
1703 | |
|
|
1704 | if (offset_return) |
|
|
1705 | *offset_return = dec.cur; |
|
|
1706 | |
|
|
1707 | if (!(offset_return || !sv)) |
|
|
1708 | { |
|
|
1709 | // check for trailing garbage |
|
|
1710 | decode_ws (&dec); |
|
|
1711 | |
|
|
1712 | if (*dec.cur) |
|
|
1713 | { |
|
|
1714 | dec.err = "garbage after JSON object"; |
|
|
1715 | SvREFCNT_dec (sv); |
|
|
1716 | sv = 0; |
|
|
1717 | } |
|
|
1718 | } |
|
|
1719 | |
996 | if (!sv) |
1720 | if (!sv) |
997 | { |
1721 | { |
998 | IV offset = dec.flags & F_UTF8 |
|
|
999 | ? dec.cur - SvPVX (string) |
|
|
1000 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1001 | SV *uni = sv_newmortal (); |
1722 | SV *uni = sv_newmortal (); |
1002 | |
1723 | |
1003 | // horrible hack to silence warning inside pv_uni_display |
1724 | // horrible hack to silence warning inside pv_uni_display |
1004 | COP cop = *PL_curcop; |
1725 | COP cop = *PL_curcop; |
1005 | cop.cop_warnings = pWARN_NONE; |
1726 | cop.cop_warnings = pWARN_NONE; |
… | |
… | |
1007 | SAVEVPTR (PL_curcop); |
1728 | SAVEVPTR (PL_curcop); |
1008 | PL_curcop = &cop; |
1729 | PL_curcop = &cop; |
1009 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1730 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1010 | LEAVE; |
1731 | LEAVE; |
1011 | |
1732 | |
1012 | croak ("%s, at character offset %d [\"%s\"]", |
1733 | croak ("%s, at character offset %d (before \"%s\")", |
1013 | dec.err, |
1734 | dec.err, |
1014 | (int)offset, |
1735 | (int)ptr_to_index (string, dec.cur), |
1015 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1736 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1016 | } |
1737 | } |
1017 | |
1738 | |
1018 | sv = sv_2mortal (sv); |
1739 | sv = sv_2mortal (sv); |
1019 | |
1740 | |
1020 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1741 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1021 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1742 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1022 | |
1743 | |
1023 | return sv; |
1744 | return sv; |
|
|
1745 | } |
|
|
1746 | |
|
|
1747 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1748 | // incremental parser |
|
|
1749 | |
|
|
1750 | static void |
|
|
1751 | incr_parse (JSON *self) |
|
|
1752 | { |
|
|
1753 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1754 | |
|
|
1755 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1756 | // but this would make it slower, so... |
|
|
1757 | |
|
|
1758 | for (;;) |
|
|
1759 | { |
|
|
1760 | //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 |
|
|
1761 | switch (self->incr_mode) |
|
|
1762 | { |
|
|
1763 | // only used for initial whitespace skipping |
|
|
1764 | case INCR_M_WS: |
|
|
1765 | for (;;) |
|
|
1766 | { |
|
|
1767 | if (*p > 0x20) |
|
|
1768 | { |
|
|
1769 | if (*p == '#') |
|
|
1770 | { |
|
|
1771 | self->incr_mode = INCR_M_C0; |
|
|
1772 | goto incr_m_c; |
|
|
1773 | } |
|
|
1774 | else |
|
|
1775 | { |
|
|
1776 | self->incr_mode = INCR_M_JSON; |
|
|
1777 | goto incr_m_json; |
|
|
1778 | } |
|
|
1779 | } |
|
|
1780 | else if (!*p) |
|
|
1781 | goto interrupt; |
|
|
1782 | |
|
|
1783 | ++p; |
|
|
1784 | } |
|
|
1785 | |
|
|
1786 | // skip a single char inside a string (for \\-processing) |
|
|
1787 | case INCR_M_BS: |
|
|
1788 | if (!*p) |
|
|
1789 | goto interrupt; |
|
|
1790 | |
|
|
1791 | ++p; |
|
|
1792 | self->incr_mode = INCR_M_STR; |
|
|
1793 | goto incr_m_str; |
|
|
1794 | |
|
|
1795 | // inside #-style comments |
|
|
1796 | case INCR_M_C0: |
|
|
1797 | case INCR_M_C1: |
|
|
1798 | incr_m_c: |
|
|
1799 | for (;;) |
|
|
1800 | { |
|
|
1801 | if (*p == '\n') |
|
|
1802 | { |
|
|
1803 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1804 | break; |
|
|
1805 | } |
|
|
1806 | else if (!*p) |
|
|
1807 | goto interrupt; |
|
|
1808 | |
|
|
1809 | ++p; |
|
|
1810 | } |
|
|
1811 | |
|
|
1812 | break; |
|
|
1813 | |
|
|
1814 | // inside a string |
|
|
1815 | case INCR_M_STR: |
|
|
1816 | incr_m_str: |
|
|
1817 | for (;;) |
|
|
1818 | { |
|
|
1819 | if (*p == '"') |
|
|
1820 | { |
|
|
1821 | ++p; |
|
|
1822 | self->incr_mode = INCR_M_JSON; |
|
|
1823 | |
|
|
1824 | if (!self->incr_nest) |
|
|
1825 | goto interrupt; |
|
|
1826 | |
|
|
1827 | goto incr_m_json; |
|
|
1828 | } |
|
|
1829 | else if (*p == '\\') |
|
|
1830 | { |
|
|
1831 | ++p; // "virtually" consumes character after \ |
|
|
1832 | |
|
|
1833 | if (!*p) // if at end of string we have to switch modes |
|
|
1834 | { |
|
|
1835 | self->incr_mode = INCR_M_BS; |
|
|
1836 | goto interrupt; |
|
|
1837 | } |
|
|
1838 | } |
|
|
1839 | else if (!*p) |
|
|
1840 | goto interrupt; |
|
|
1841 | |
|
|
1842 | ++p; |
|
|
1843 | } |
|
|
1844 | |
|
|
1845 | // after initial ws, outside string |
|
|
1846 | case INCR_M_JSON: |
|
|
1847 | incr_m_json: |
|
|
1848 | for (;;) |
|
|
1849 | { |
|
|
1850 | switch (*p++) |
|
|
1851 | { |
|
|
1852 | case 0: |
|
|
1853 | --p; |
|
|
1854 | goto interrupt; |
|
|
1855 | |
|
|
1856 | case 0x09: |
|
|
1857 | case 0x0a: |
|
|
1858 | case 0x0d: |
|
|
1859 | case 0x20: |
|
|
1860 | if (!self->incr_nest) |
|
|
1861 | { |
|
|
1862 | --p; // do not eat the whitespace, let the next round do it |
|
|
1863 | goto interrupt; |
|
|
1864 | } |
|
|
1865 | break; |
|
|
1866 | |
|
|
1867 | case '"': |
|
|
1868 | self->incr_mode = INCR_M_STR; |
|
|
1869 | goto incr_m_str; |
|
|
1870 | |
|
|
1871 | case '[': |
|
|
1872 | case '{': |
|
|
1873 | case '(': |
|
|
1874 | if (++self->incr_nest > self->max_depth) |
|
|
1875 | croak (ERR_NESTING_EXCEEDED); |
|
|
1876 | break; |
|
|
1877 | |
|
|
1878 | case ']': |
|
|
1879 | case '}': |
|
|
1880 | if (--self->incr_nest <= 0) |
|
|
1881 | goto interrupt; |
|
|
1882 | break; |
|
|
1883 | |
|
|
1884 | case ')': |
|
|
1885 | --self->incr_nest; |
|
|
1886 | break; |
|
|
1887 | |
|
|
1888 | case '#': |
|
|
1889 | self->incr_mode = INCR_M_C1; |
|
|
1890 | goto incr_m_c; |
|
|
1891 | } |
|
|
1892 | } |
|
|
1893 | } |
|
|
1894 | |
|
|
1895 | modechange: |
|
|
1896 | ; |
|
|
1897 | } |
|
|
1898 | |
|
|
1899 | interrupt: |
|
|
1900 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1901 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1902 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1024 | } |
1903 | } |
1025 | |
1904 | |
1026 | ///////////////////////////////////////////////////////////////////////////// |
1905 | ///////////////////////////////////////////////////////////////////////////// |
1027 | // XS interface functions |
1906 | // XS interface functions |
1028 | |
1907 | |
1029 | MODULE = JSON::XS PACKAGE = JSON::XS |
1908 | MODULE = JSON::XS PACKAGE = JSON::XS |
1030 | |
1909 | |
1031 | BOOT: |
1910 | BOOT: |
1032 | { |
1911 | { |
1033 | int i; |
1912 | int i; |
1034 | |
|
|
1035 | memset (decode_hexdigit, 0xff, 256); |
|
|
1036 | |
1913 | |
1037 | for (i = 0; i < 256; ++i) |
1914 | for (i = 0; i < 256; ++i) |
1038 | decode_hexdigit [i] = |
1915 | decode_hexdigit [i] = |
1039 | i >= '0' && i <= '9' ? i - '0' |
1916 | i >= '0' && i <= '9' ? i - '0' |
1040 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1917 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1041 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1918 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1042 | : -1; |
1919 | : -1; |
1043 | |
1920 | |
1044 | json_stash = gv_stashpv ("JSON::XS", 1); |
1921 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1922 | types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1); |
|
|
1923 | |
|
|
1924 | types_true = get_bool ("Types::Serialiser::true"); |
|
|
1925 | types_false = get_bool ("Types::Serialiser::false"); |
|
|
1926 | |
|
|
1927 | sv_json = newSVpv ("JSON", 0); |
|
|
1928 | SvREADONLY_on (sv_json); |
|
|
1929 | |
|
|
1930 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1045 | } |
1931 | } |
1046 | |
1932 | |
1047 | PROTOTYPES: DISABLE |
1933 | PROTOTYPES: DISABLE |
1048 | |
1934 | |
1049 | SV *new (char *dummy) |
1935 | void CLONE (...) |
1050 | CODE: |
1936 | CODE: |
1051 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1937 | json_stash = 0; |
|
|
1938 | types_boolean_stash = 0; |
|
|
1939 | |
|
|
1940 | void new (char *klass) |
|
|
1941 | PPCODE: |
|
|
1942 | { |
|
|
1943 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1944 | SvPOK_only (pv); |
|
|
1945 | json_init ((JSON *)SvPVX (pv)); |
|
|
1946 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1947 | newRV_noinc (pv), |
|
|
1948 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1949 | ))); |
|
|
1950 | } |
|
|
1951 | |
|
|
1952 | void ascii (JSON *self, int enable = 1) |
|
|
1953 | ALIAS: |
|
|
1954 | ascii = F_ASCII |
|
|
1955 | latin1 = F_LATIN1 |
|
|
1956 | utf8 = F_UTF8 |
|
|
1957 | indent = F_INDENT |
|
|
1958 | canonical = F_CANONICAL |
|
|
1959 | space_before = F_SPACE_BEFORE |
|
|
1960 | space_after = F_SPACE_AFTER |
|
|
1961 | pretty = F_PRETTY |
|
|
1962 | allow_nonref = F_ALLOW_NONREF |
|
|
1963 | shrink = F_SHRINK |
|
|
1964 | allow_blessed = F_ALLOW_BLESSED |
|
|
1965 | convert_blessed = F_CONV_BLESSED |
|
|
1966 | relaxed = F_RELAXED |
|
|
1967 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
1968 | allow_tags = F_ALLOW_TAGS |
|
|
1969 | PPCODE: |
|
|
1970 | { |
|
|
1971 | if (enable) |
|
|
1972 | self->flags |= ix; |
|
|
1973 | else |
|
|
1974 | self->flags &= ~ix; |
|
|
1975 | |
|
|
1976 | XPUSHs (ST (0)); |
|
|
1977 | } |
|
|
1978 | |
|
|
1979 | void get_ascii (JSON *self) |
|
|
1980 | ALIAS: |
|
|
1981 | get_ascii = F_ASCII |
|
|
1982 | get_latin1 = F_LATIN1 |
|
|
1983 | get_utf8 = F_UTF8 |
|
|
1984 | get_indent = F_INDENT |
|
|
1985 | get_canonical = F_CANONICAL |
|
|
1986 | get_space_before = F_SPACE_BEFORE |
|
|
1987 | get_space_after = F_SPACE_AFTER |
|
|
1988 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1989 | get_shrink = F_SHRINK |
|
|
1990 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1991 | get_convert_blessed = F_CONV_BLESSED |
|
|
1992 | get_relaxed = F_RELAXED |
|
|
1993 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
1994 | get_allow_tags = F_ALLOW_TAGS |
|
|
1995 | PPCODE: |
|
|
1996 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1997 | |
|
|
1998 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
|
|
1999 | PPCODE: |
|
|
2000 | self->max_depth = max_depth; |
|
|
2001 | XPUSHs (ST (0)); |
|
|
2002 | |
|
|
2003 | U32 get_max_depth (JSON *self) |
|
|
2004 | CODE: |
|
|
2005 | RETVAL = self->max_depth; |
1052 | OUTPUT: |
2006 | OUTPUT: |
1053 | RETVAL |
2007 | RETVAL |
1054 | |
2008 | |
1055 | SV *ascii (SV *self, int enable = 1) |
2009 | void max_size (JSON *self, U32 max_size = 0) |
1056 | ALIAS: |
2010 | PPCODE: |
1057 | ascii = F_ASCII |
2011 | self->max_size = max_size; |
1058 | utf8 = F_UTF8 |
2012 | XPUSHs (ST (0)); |
1059 | indent = F_INDENT |
2013 | |
1060 | canonical = F_CANONICAL |
2014 | int get_max_size (JSON *self) |
1061 | space_before = F_SPACE_BEFORE |
|
|
1062 | space_after = F_SPACE_AFTER |
|
|
1063 | pretty = F_PRETTY |
|
|
1064 | allow_nonref = F_ALLOW_NONREF |
|
|
1065 | shrink = F_SHRINK |
|
|
1066 | CODE: |
2015 | CODE: |
1067 | { |
2016 | RETVAL = self->max_size; |
1068 | UV *uv = SvJSON (self); |
|
|
1069 | if (enable) |
|
|
1070 | *uv |= ix; |
|
|
1071 | else |
|
|
1072 | *uv &= ~ix; |
|
|
1073 | |
|
|
1074 | RETVAL = newSVsv (self); |
|
|
1075 | } |
|
|
1076 | OUTPUT: |
2017 | OUTPUT: |
1077 | RETVAL |
2018 | RETVAL |
1078 | |
2019 | |
1079 | SV *max_depth (SV *self, int max_depth = 0x80000000UL) |
2020 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
|
|
2021 | PPCODE: |
|
|
2022 | { |
|
|
2023 | SvREFCNT_dec (self->cb_object); |
|
|
2024 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
|
|
2025 | |
|
|
2026 | XPUSHs (ST (0)); |
|
|
2027 | } |
|
|
2028 | |
|
|
2029 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
|
|
2030 | PPCODE: |
|
|
2031 | { |
|
|
2032 | if (!self->cb_sk_object) |
|
|
2033 | self->cb_sk_object = newHV (); |
|
|
2034 | |
|
|
2035 | if (SvOK (cb)) |
|
|
2036 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
2037 | else |
|
|
2038 | { |
|
|
2039 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
2040 | |
|
|
2041 | if (!HvKEYS (self->cb_sk_object)) |
|
|
2042 | { |
|
|
2043 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2044 | self->cb_sk_object = 0; |
|
|
2045 | } |
|
|
2046 | } |
|
|
2047 | |
|
|
2048 | XPUSHs (ST (0)); |
|
|
2049 | } |
|
|
2050 | |
|
|
2051 | void encode (JSON *self, SV *scalar) |
|
|
2052 | PPCODE: |
|
|
2053 | PUTBACK; scalar = encode_json (scalar, self); SPAGAIN; |
|
|
2054 | XPUSHs (scalar); |
|
|
2055 | |
|
|
2056 | void decode (JSON *self, SV *jsonstr) |
|
|
2057 | PPCODE: |
|
|
2058 | PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN; |
|
|
2059 | XPUSHs (jsonstr); |
|
|
2060 | |
|
|
2061 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
2062 | PPCODE: |
|
|
2063 | { |
|
|
2064 | SV *sv; |
|
|
2065 | char *offset; |
|
|
2066 | PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN; |
|
|
2067 | EXTEND (SP, 2); |
|
|
2068 | PUSHs (sv); |
|
|
2069 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
2070 | } |
|
|
2071 | |
|
|
2072 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
2073 | PPCODE: |
|
|
2074 | { |
|
|
2075 | if (!self->incr_text) |
|
|
2076 | self->incr_text = newSVpvn ("", 0); |
|
|
2077 | |
|
|
2078 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
2079 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
2080 | if (DECODE_WANTS_OCTETS (self)) |
|
|
2081 | { |
|
|
2082 | if (self->incr_pos) |
|
|
2083 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
2084 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
2085 | |
|
|
2086 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
2087 | } |
|
|
2088 | else |
|
|
2089 | { |
|
|
2090 | sv_utf8_upgrade (self->incr_text); |
|
|
2091 | |
|
|
2092 | if (self->incr_pos) |
|
|
2093 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
2094 | - (U8 *)SvPVX (self->incr_text); |
|
|
2095 | } |
|
|
2096 | |
|
|
2097 | // append data, if any |
|
|
2098 | if (jsonstr) |
|
|
2099 | { |
|
|
2100 | /* make sure both strings have same encoding */ |
|
|
2101 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
|
|
2102 | if (SvUTF8 (jsonstr)) |
|
|
2103 | sv_utf8_downgrade (jsonstr, 0); |
|
|
2104 | else |
|
|
2105 | sv_utf8_upgrade (jsonstr); |
|
|
2106 | |
|
|
2107 | /* and then just blindly append */ |
|
|
2108 | { |
|
|
2109 | STRLEN len; |
|
|
2110 | const char *str = SvPV (jsonstr, len); |
|
|
2111 | STRLEN cur = SvCUR (self->incr_text); |
|
|
2112 | |
|
|
2113 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
2114 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
2115 | |
|
|
2116 | Move (str, SvEND (self->incr_text), len, char); |
|
|
2117 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
2118 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
2119 | } |
|
|
2120 | } |
|
|
2121 | |
|
|
2122 | if (GIMME_V != G_VOID) |
|
|
2123 | do |
|
|
2124 | { |
|
|
2125 | SV *sv; |
|
|
2126 | char *offset; |
|
|
2127 | |
|
|
2128 | if (!INCR_DONE (self)) |
|
|
2129 | { |
|
|
2130 | incr_parse (self); |
|
|
2131 | |
|
|
2132 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
2133 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
2134 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
2135 | |
|
|
2136 | if (!INCR_DONE (self)) |
|
|
2137 | { |
|
|
2138 | // as an optimisation, do not accumulate white space in the incr buffer |
|
|
2139 | if (self->incr_mode == INCR_M_WS && self->incr_pos) |
|
|
2140 | { |
|
|
2141 | self->incr_pos = 0; |
|
|
2142 | SvCUR_set (self->incr_text, 0); |
|
|
2143 | } |
|
|
2144 | |
|
|
2145 | break; |
|
|
2146 | } |
|
|
2147 | } |
|
|
2148 | |
|
|
2149 | PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN; |
|
|
2150 | XPUSHs (sv); |
|
|
2151 | |
|
|
2152 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
2153 | self->incr_nest = 0; |
|
|
2154 | self->incr_mode = 0; |
|
|
2155 | |
|
|
2156 | sv_chop (self->incr_text, offset); |
|
|
2157 | } |
|
|
2158 | while (GIMME_V == G_ARRAY); |
|
|
2159 | } |
|
|
2160 | |
|
|
2161 | SV *incr_text (JSON *self) |
|
|
2162 | ATTRS: lvalue |
1080 | CODE: |
2163 | CODE: |
1081 | { |
2164 | { |
1082 | UV *uv = SvJSON (self); |
2165 | if (self->incr_pos) |
1083 | UV log2 = 0; |
2166 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
1084 | |
2167 | |
1085 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
2168 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
1086 | |
|
|
1087 | while ((1UL << log2) < max_depth) |
|
|
1088 | ++log2; |
|
|
1089 | |
|
|
1090 | *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1091 | |
|
|
1092 | RETVAL = newSVsv (self); |
|
|
1093 | } |
2169 | } |
1094 | OUTPUT: |
2170 | OUTPUT: |
1095 | RETVAL |
2171 | RETVAL |
1096 | |
2172 | |
|
|
2173 | void incr_skip (JSON *self) |
|
|
2174 | CODE: |
|
|
2175 | { |
|
|
2176 | if (self->incr_pos) |
|
|
2177 | { |
|
|
2178 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2179 | self->incr_pos = 0; |
|
|
2180 | self->incr_nest = 0; |
|
|
2181 | self->incr_mode = 0; |
|
|
2182 | } |
|
|
2183 | } |
|
|
2184 | |
|
|
2185 | void incr_reset (JSON *self) |
|
|
2186 | CODE: |
|
|
2187 | { |
|
|
2188 | SvREFCNT_dec (self->incr_text); |
|
|
2189 | self->incr_text = 0; |
|
|
2190 | self->incr_pos = 0; |
|
|
2191 | self->incr_nest = 0; |
|
|
2192 | self->incr_mode = 0; |
|
|
2193 | } |
|
|
2194 | |
|
|
2195 | void DESTROY (JSON *self) |
|
|
2196 | CODE: |
|
|
2197 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2198 | SvREFCNT_dec (self->cb_object); |
|
|
2199 | SvREFCNT_dec (self->incr_text); |
|
|
2200 | |
|
|
2201 | PROTOTYPES: ENABLE |
|
|
2202 | |
1097 | void encode (SV *self, SV *scalar) |
2203 | void encode_json (SV *scalar) |
1098 | PPCODE: |
2204 | PPCODE: |
1099 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
2205 | { |
|
|
2206 | JSON json; |
|
|
2207 | json_init (&json); |
|
|
2208 | json.flags |= F_UTF8; |
|
|
2209 | PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN; |
|
|
2210 | XPUSHs (scalar); |
|
|
2211 | } |
1100 | |
2212 | |
1101 | void decode (SV *self, SV *jsonstr) |
2213 | void decode_json (SV *jsonstr) |
1102 | PPCODE: |
2214 | PPCODE: |
1103 | XPUSHs (decode_json (jsonstr, *SvJSON (self))); |
2215 | { |
|
|
2216 | JSON json; |
|
|
2217 | json_init (&json); |
|
|
2218 | json.flags |= F_UTF8; |
|
|
2219 | PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN; |
|
|
2220 | XPUSHs (jsonstr); |
|
|
2221 | } |
1104 | |
2222 | |
1105 | PROTOTYPES: ENABLE |
|
|
1106 | |
|
|
1107 | void to_json (SV *scalar) |
|
|
1108 | ALIAS: |
|
|
1109 | objToJson = 0 |
|
|
1110 | PPCODE: |
|
|
1111 | XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); |
|
|
1112 | |
|
|
1113 | void from_json (SV *jsonstr) |
|
|
1114 | ALIAS: |
|
|
1115 | jsonToObj = 0 |
|
|
1116 | PPCODE: |
|
|
1117 | XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8)); |
|
|
1118 | |
|
|