1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
|
|
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
|
|
10 | #include <float.h> |
8 | |
11 | |
|
|
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
|
|
14 | #endif |
|
|
15 | |
|
|
16 | // some old perls do not have this, try to make it work, no |
|
|
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
|
|
18 | #ifndef UTF8_MAXBYTES |
|
|
19 | # define UTF8_MAXBYTES 13 |
|
|
20 | #endif |
|
|
21 | |
|
|
22 | // three extra for rounding, sign, and end of string |
|
|
23 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
|
|
24 | |
9 | #define F_ASCII 0x00000001 |
25 | #define F_ASCII 0x00000001UL |
|
|
26 | #define F_LATIN1 0x00000002UL |
10 | #define F_UTF8 0x00000002 |
27 | #define F_UTF8 0x00000004UL |
11 | #define F_INDENT 0x00000004 |
28 | #define F_INDENT 0x00000008UL |
12 | #define F_CANONICAL 0x00000008 |
29 | #define F_CANONICAL 0x00000010UL |
13 | #define F_SPACE_BEFORE 0x00000010 |
30 | #define F_SPACE_BEFORE 0x00000020UL |
14 | #define F_SPACE_AFTER 0x00000020 |
31 | #define F_SPACE_AFTER 0x00000040UL |
15 | #define F_ALLOW_NONREF 0x00000080 |
32 | #define F_ALLOW_NONREF 0x00000100UL |
16 | #define F_SHRINK 0x00000100 |
33 | #define F_SHRINK 0x00000200UL |
|
|
34 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
35 | #define F_CONV_BLESSED 0x00000800UL |
|
|
36 | #define F_RELAXED 0x00001000UL |
|
|
37 | #define F_ALLOW_UNKNOWN 0x00002000UL |
|
|
38 | #define F_ALLOW_TAGS 0x00004000UL |
|
|
39 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
17 | |
40 | |
18 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
41 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
19 | #define F_DEFAULT 0 |
|
|
20 | |
42 | |
21 | #define INIT_SIZE 32 // initial scalar size to be allocated |
43 | #define INIT_SIZE 32 // initial scalar size to be allocated |
22 | #define INDENT_STEP 3 // spaces per indentation level |
44 | #define INDENT_STEP 3 // spaces per indentation level |
23 | |
45 | |
24 | #define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character |
|
|
25 | #define SHORT_STRING_LEN 256 // special-case strings of up to this size |
46 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
|
|
47 | |
|
|
48 | #define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
26 | |
49 | |
27 | #define SB do { |
50 | #define SB do { |
28 | #define SE } while (0) |
51 | #define SE } while (0) |
29 | |
52 | |
|
|
53 | #if __GNUC__ >= 3 |
|
|
54 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
|
|
55 | # define INLINE static inline |
|
|
56 | #else |
|
|
57 | # define expect(expr,value) (expr) |
|
|
58 | # define INLINE static |
|
|
59 | #endif |
|
|
60 | |
|
|
61 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
62 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
63 | |
|
|
64 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
65 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
66 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
67 | |
|
|
68 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
69 | |
|
|
70 | #ifdef USE_ITHREADS |
|
|
71 | # define JSON_SLOW 1 |
|
|
72 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
73 | #else |
|
|
74 | # define JSON_SLOW 0 |
|
|
75 | # define JSON_STASH json_stash |
|
|
76 | #endif |
|
|
77 | |
|
|
78 | // the amount of HEs to allocate on the stack, when sorting keys |
|
|
79 | #define STACK_HES 64 |
|
|
80 | |
30 | static HV *json_stash; // JSON::XS:: |
81 | static HV *json_stash, *types_boolean_stash; // JSON::XS:: |
|
|
82 | static SV *types_true, *types_false, *sv_json; |
|
|
83 | |
|
|
84 | enum { |
|
|
85 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
86 | INCR_M_STR, // inside string |
|
|
87 | INCR_M_BS, // inside backslash |
|
|
88 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
89 | INCR_M_C1, // inside comment in other places |
|
|
90 | INCR_M_JSON // outside anything, count nesting |
|
|
91 | }; |
|
|
92 | |
|
|
93 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
94 | |
|
|
95 | typedef struct { |
|
|
96 | U32 flags; |
|
|
97 | U32 max_depth; |
|
|
98 | STRLEN max_size; |
|
|
99 | |
|
|
100 | SV *cb_object; |
|
|
101 | HV *cb_sk_object; |
|
|
102 | |
|
|
103 | // for the incremental parser |
|
|
104 | SV *incr_text; // the source text so far |
|
|
105 | STRLEN incr_pos; // the current offset into the text |
|
|
106 | int incr_nest; // {[]}-nesting level |
|
|
107 | unsigned char incr_mode; |
|
|
108 | } JSON; |
|
|
109 | |
|
|
110 | INLINE void |
|
|
111 | json_init (JSON *json) |
|
|
112 | { |
|
|
113 | Zero (json, 1, JSON); |
|
|
114 | json->max_depth = 512; |
|
|
115 | } |
31 | |
116 | |
32 | ///////////////////////////////////////////////////////////////////////////// |
117 | ///////////////////////////////////////////////////////////////////////////// |
33 | // utility functions |
118 | // utility functions |
34 | |
119 | |
35 | static UV * |
120 | INLINE SV * |
36 | SvJSON (SV *sv) |
121 | get_bool (const char *name) |
37 | { |
122 | { |
38 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
123 | SV *sv = get_sv (name, 1); |
39 | croak ("object is not of type JSON::XS"); |
|
|
40 | |
124 | |
41 | return &SvUVX (SvRV (sv)); |
125 | SvREADONLY_on (sv); |
42 | } |
126 | SvREADONLY_on (SvRV (sv)); |
43 | |
127 | |
44 | static void |
128 | return sv; |
|
|
129 | } |
|
|
130 | |
|
|
131 | INLINE void |
45 | shrink (SV *sv) |
132 | shrink (SV *sv) |
46 | { |
133 | { |
47 | sv_utf8_downgrade (sv, 1); |
134 | sv_utf8_downgrade (sv, 1); |
|
|
135 | |
48 | if (SvLEN (sv) > SvCUR (sv) + 1) |
136 | if (SvLEN (sv) > SvCUR (sv) + 1) |
49 | { |
137 | { |
50 | #ifdef SvPV_shrink_to_cur |
138 | #ifdef SvPV_shrink_to_cur |
51 | SvPV_shrink_to_cur (sv); |
139 | SvPV_shrink_to_cur (sv); |
52 | #elif defined (SvPV_renew) |
140 | #elif defined (SvPV_renew) |
53 | SvPV_renew (sv, SvCUR (sv) + 1); |
141 | SvPV_renew (sv, SvCUR (sv) + 1); |
54 | #endif |
142 | #endif |
55 | } |
143 | } |
56 | } |
144 | } |
57 | |
145 | |
|
|
146 | // decode an utf-8 character and return it, or (UV)-1 in |
|
|
147 | // case of an error. |
|
|
148 | // we special-case "safe" characters from U+80 .. U+7FF, |
|
|
149 | // but use the very good perl function to parse anything else. |
|
|
150 | // note that we never call this function for a ascii codepoints |
|
|
151 | INLINE UV |
|
|
152 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
|
|
153 | { |
|
|
154 | if (expect_true (len >= 2 |
|
|
155 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
|
|
156 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
|
|
157 | { |
|
|
158 | *clen = 2; |
|
|
159 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
|
|
160 | } |
|
|
161 | else |
|
|
162 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
|
|
163 | } |
|
|
164 | |
|
|
165 | // likewise for encoding, also never called for ascii codepoints |
|
|
166 | // this function takes advantage of this fact, although current gccs |
|
|
167 | // seem to optimise the check for >= 0x80 away anyways |
|
|
168 | INLINE unsigned char * |
|
|
169 | encode_utf8 (unsigned char *s, UV ch) |
|
|
170 | { |
|
|
171 | if (expect_false (ch < 0x000080)) |
|
|
172 | *s++ = ch; |
|
|
173 | else if (expect_true (ch < 0x000800)) |
|
|
174 | *s++ = 0xc0 | ( ch >> 6), |
|
|
175 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
176 | else if ( ch < 0x010000) |
|
|
177 | *s++ = 0xe0 | ( ch >> 12), |
|
|
178 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
179 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
180 | else if ( ch < 0x110000) |
|
|
181 | *s++ = 0xf0 | ( ch >> 18), |
|
|
182 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
183 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
184 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
185 | |
|
|
186 | return s; |
|
|
187 | } |
|
|
188 | |
|
|
189 | // convert offset pointer to character index, sv must be string |
|
|
190 | static STRLEN |
|
|
191 | ptr_to_index (SV *sv, char *offset) |
|
|
192 | { |
|
|
193 | return SvUTF8 (sv) |
|
|
194 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
195 | : offset - SvPVX (sv); |
|
|
196 | } |
|
|
197 | |
|
|
198 | ///////////////////////////////////////////////////////////////////////////// |
|
|
199 | // fp hell |
|
|
200 | |
|
|
201 | // scan a group of digits, and a trailing exponent |
|
|
202 | static void |
|
|
203 | json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
|
|
204 | { |
|
|
205 | UV uaccum = 0; |
|
|
206 | int eaccum = 0; |
|
|
207 | |
|
|
208 | // if we recurse too deep, skip all remaining digits |
|
|
209 | // to avoid a stack overflow attack |
|
|
210 | if (expect_false (--maxdepth <= 0)) |
|
|
211 | while (((U8)*s - '0') < 10) |
|
|
212 | ++s; |
|
|
213 | |
|
|
214 | for (;;) |
|
|
215 | { |
|
|
216 | U8 dig = (U8)*s - '0'; |
|
|
217 | |
|
|
218 | if (expect_false (dig >= 10)) |
|
|
219 | { |
|
|
220 | if (dig == (U8)((U8)'.' - (U8)'0')) |
|
|
221 | { |
|
|
222 | ++s; |
|
|
223 | json_atof_scan1 (s, accum, expo, 1, maxdepth); |
|
|
224 | } |
|
|
225 | else if ((dig | ' ') == 'e' - '0') |
|
|
226 | { |
|
|
227 | int exp2 = 0; |
|
|
228 | int neg = 0; |
|
|
229 | |
|
|
230 | ++s; |
|
|
231 | |
|
|
232 | if (*s == '-') |
|
|
233 | { |
|
|
234 | ++s; |
|
|
235 | neg = 1; |
|
|
236 | } |
|
|
237 | else if (*s == '+') |
|
|
238 | ++s; |
|
|
239 | |
|
|
240 | while ((dig = (U8)*s - '0') < 10) |
|
|
241 | exp2 = exp2 * 10 + *s++ - '0'; |
|
|
242 | |
|
|
243 | *expo += neg ? -exp2 : exp2; |
|
|
244 | } |
|
|
245 | |
|
|
246 | break; |
|
|
247 | } |
|
|
248 | |
|
|
249 | ++s; |
|
|
250 | |
|
|
251 | uaccum = uaccum * 10 + dig; |
|
|
252 | ++eaccum; |
|
|
253 | |
|
|
254 | // if we have too many digits, then recurse for more |
|
|
255 | // we actually do this for rather few digits |
|
|
256 | if (uaccum >= (UV_MAX - 9) / 10) |
|
|
257 | { |
|
|
258 | if (postdp) *expo -= eaccum; |
|
|
259 | json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
|
|
260 | if (postdp) *expo += eaccum; |
|
|
261 | |
|
|
262 | break; |
|
|
263 | } |
|
|
264 | } |
|
|
265 | |
|
|
266 | // this relies greatly on the quality of the pow () |
|
|
267 | // implementation of the platform, but a good |
|
|
268 | // implementation is hard to beat. |
|
|
269 | // (IEEE 754 conformant ones are required to be exact) |
|
|
270 | if (postdp) *expo -= eaccum; |
|
|
271 | *accum += uaccum * Perl_pow (10., *expo); |
|
|
272 | *expo += eaccum; |
|
|
273 | } |
|
|
274 | |
|
|
275 | static NV |
|
|
276 | json_atof (const char *s) |
|
|
277 | { |
|
|
278 | NV accum = 0.; |
|
|
279 | int expo = 0; |
|
|
280 | int neg = 0; |
|
|
281 | |
|
|
282 | if (*s == '-') |
|
|
283 | { |
|
|
284 | ++s; |
|
|
285 | neg = 1; |
|
|
286 | } |
|
|
287 | |
|
|
288 | // a recursion depth of ten gives us >>500 bits |
|
|
289 | json_atof_scan1 (s, &accum, &expo, 0, 10); |
|
|
290 | |
|
|
291 | return neg ? -accum : accum; |
|
|
292 | } |
58 | ///////////////////////////////////////////////////////////////////////////// |
293 | ///////////////////////////////////////////////////////////////////////////// |
59 | // encoder |
294 | // encoder |
60 | |
295 | |
61 | // structure used for encoding JSON |
296 | // structure used for encoding JSON |
62 | typedef struct |
297 | typedef struct |
63 | { |
298 | { |
64 | char *cur; // SvPVX (sv) + current output position |
299 | char *cur; // SvPVX (sv) + current output position |
65 | char *end; // SvEND (sv) |
300 | char *end; // SvEND (sv) |
66 | SV *sv; // result scalar |
301 | SV *sv; // result scalar |
67 | UV flags; // F_* |
302 | JSON json; |
68 | int indent; // indentation level |
303 | U32 indent; // indentation level |
69 | int max_depth; // max. recursion level |
304 | UV limit; // escape character values >= this value when encoding |
70 | } enc_t; |
305 | } enc_t; |
71 | |
306 | |
72 | static void |
307 | INLINE void |
73 | need (enc_t *enc, STRLEN len) |
308 | need (enc_t *enc, STRLEN len) |
74 | { |
309 | { |
75 | if (enc->cur + len >= enc->end) |
310 | if (expect_false (enc->cur + len >= enc->end)) |
76 | { |
311 | { |
77 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
312 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
78 | SvGROW (enc->sv, cur + len + 1); |
313 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
79 | enc->cur = SvPVX (enc->sv) + cur; |
314 | enc->cur = SvPVX (enc->sv) + cur; |
80 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); |
315 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
81 | } |
316 | } |
82 | } |
317 | } |
83 | |
318 | |
84 | static void |
319 | INLINE void |
85 | encode_ch (enc_t *enc, char ch) |
320 | encode_ch (enc_t *enc, char ch) |
86 | { |
321 | { |
87 | need (enc, 1); |
322 | need (enc, 1); |
88 | *enc->cur++ = ch; |
323 | *enc->cur++ = ch; |
89 | } |
324 | } |
… | |
… | |
97 | |
332 | |
98 | while (str < end) |
333 | while (str < end) |
99 | { |
334 | { |
100 | unsigned char ch = *(unsigned char *)str; |
335 | unsigned char ch = *(unsigned char *)str; |
101 | |
336 | |
102 | if (ch >= 0x20 && ch < 0x80) // most common case |
337 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
103 | { |
338 | { |
104 | if (ch == '"') // but with slow exceptions |
339 | if (expect_false (ch == '"')) // but with slow exceptions |
105 | { |
340 | { |
106 | need (enc, len += 1); |
341 | need (enc, len += 1); |
107 | *enc->cur++ = '\\'; |
342 | *enc->cur++ = '\\'; |
108 | *enc->cur++ = '"'; |
343 | *enc->cur++ = '"'; |
109 | } |
344 | } |
110 | else if (ch == '\\') |
345 | else if (expect_false (ch == '\\')) |
111 | { |
346 | { |
112 | need (enc, len += 1); |
347 | need (enc, len += 1); |
113 | *enc->cur++ = '\\'; |
348 | *enc->cur++ = '\\'; |
114 | *enc->cur++ = '\\'; |
349 | *enc->cur++ = '\\'; |
115 | } |
350 | } |
… | |
… | |
133 | STRLEN clen; |
368 | STRLEN clen; |
134 | UV uch; |
369 | UV uch; |
135 | |
370 | |
136 | if (is_utf8) |
371 | if (is_utf8) |
137 | { |
372 | { |
138 | uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
373 | uch = decode_utf8 (str, end - str, &clen); |
139 | if (clen == (STRLEN)-1) |
374 | if (clen == (STRLEN)-1) |
140 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
375 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
141 | } |
376 | } |
142 | else |
377 | else |
143 | { |
378 | { |
144 | uch = ch; |
379 | uch = ch; |
145 | clen = 1; |
380 | clen = 1; |
146 | } |
381 | } |
147 | |
382 | |
148 | if (uch > 0x10FFFFUL) |
383 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
149 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
150 | |
|
|
151 | if (uch < 0x80 || enc->flags & F_ASCII) |
|
|
152 | { |
384 | { |
153 | if (uch > 0xFFFFUL) |
385 | if (uch >= 0x10000UL) |
154 | { |
386 | { |
|
|
387 | if (uch >= 0x110000UL) |
|
|
388 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
389 | |
155 | need (enc, len += 11); |
390 | need (enc, len += 11); |
156 | sprintf (enc->cur, "\\u%04x\\u%04x", |
391 | sprintf (enc->cur, "\\u%04x\\u%04x", |
157 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
392 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
158 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
393 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
159 | enc->cur += 12; |
394 | enc->cur += 12; |
160 | } |
395 | } |
161 | else |
396 | else |
162 | { |
397 | { |
163 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
164 | need (enc, len += 5); |
398 | need (enc, len += 5); |
165 | *enc->cur++ = '\\'; |
399 | *enc->cur++ = '\\'; |
166 | *enc->cur++ = 'u'; |
400 | *enc->cur++ = 'u'; |
167 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
401 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
168 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
402 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
169 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
403 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
170 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
404 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
171 | } |
405 | } |
172 | |
406 | |
|
|
407 | str += clen; |
|
|
408 | } |
|
|
409 | else if (enc->json.flags & F_LATIN1) |
|
|
410 | { |
|
|
411 | *enc->cur++ = uch; |
173 | str += clen; |
412 | str += clen; |
174 | } |
413 | } |
175 | else if (is_utf8) |
414 | else if (is_utf8) |
176 | { |
415 | { |
177 | need (enc, len += clen); |
416 | need (enc, len += clen); |
… | |
… | |
181 | } |
420 | } |
182 | while (--clen); |
421 | while (--clen); |
183 | } |
422 | } |
184 | else |
423 | else |
185 | { |
424 | { |
186 | need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed |
425 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
187 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
426 | enc->cur = encode_utf8 (enc->cur, uch); |
188 | ++str; |
427 | ++str; |
189 | } |
428 | } |
190 | } |
429 | } |
191 | } |
430 | } |
192 | } |
431 | } |
193 | |
432 | |
194 | --len; |
433 | --len; |
195 | } |
434 | } |
196 | } |
435 | } |
197 | |
436 | |
198 | static void |
437 | INLINE void |
199 | encode_indent (enc_t *enc) |
438 | encode_indent (enc_t *enc) |
200 | { |
439 | { |
201 | if (enc->flags & F_INDENT) |
440 | if (enc->json.flags & F_INDENT) |
202 | { |
441 | { |
203 | int spaces = enc->indent * INDENT_STEP; |
442 | int spaces = enc->indent * INDENT_STEP; |
204 | |
443 | |
205 | need (enc, spaces); |
444 | need (enc, spaces); |
206 | memset (enc->cur, ' ', spaces); |
445 | memset (enc->cur, ' ', spaces); |
207 | enc->cur += spaces; |
446 | enc->cur += spaces; |
208 | } |
447 | } |
209 | } |
448 | } |
210 | |
449 | |
211 | static void |
450 | INLINE void |
212 | encode_space (enc_t *enc) |
451 | encode_space (enc_t *enc) |
213 | { |
452 | { |
214 | need (enc, 1); |
453 | need (enc, 1); |
215 | encode_ch (enc, ' '); |
454 | encode_ch (enc, ' '); |
216 | } |
455 | } |
217 | |
456 | |
218 | static void |
457 | INLINE void |
219 | encode_nl (enc_t *enc) |
458 | encode_nl (enc_t *enc) |
220 | { |
459 | { |
221 | if (enc->flags & F_INDENT) |
460 | if (enc->json.flags & F_INDENT) |
222 | { |
461 | { |
223 | need (enc, 1); |
462 | need (enc, 1); |
224 | encode_ch (enc, '\n'); |
463 | encode_ch (enc, '\n'); |
225 | } |
464 | } |
226 | } |
465 | } |
227 | |
466 | |
228 | static void |
467 | INLINE void |
229 | encode_comma (enc_t *enc) |
468 | encode_comma (enc_t *enc) |
230 | { |
469 | { |
231 | encode_ch (enc, ','); |
470 | encode_ch (enc, ','); |
232 | |
471 | |
233 | if (enc->flags & F_INDENT) |
472 | if (enc->json.flags & F_INDENT) |
234 | encode_nl (enc); |
473 | encode_nl (enc); |
235 | else if (enc->flags & F_SPACE_AFTER) |
474 | else if (enc->json.flags & F_SPACE_AFTER) |
236 | encode_space (enc); |
475 | encode_space (enc); |
237 | } |
476 | } |
238 | |
477 | |
239 | static void encode_sv (enc_t *enc, SV *sv); |
478 | static void encode_sv (enc_t *enc, SV *sv); |
240 | |
479 | |
241 | static void |
480 | static void |
242 | encode_av (enc_t *enc, AV *av) |
481 | encode_av (enc_t *enc, AV *av) |
243 | { |
482 | { |
244 | int i, len = av_len (av); |
483 | int i, len = av_len (av); |
245 | |
484 | |
246 | encode_ch (enc, '['); encode_nl (enc); |
485 | if (enc->indent >= enc->json.max_depth) |
247 | ++enc->indent; |
486 | croak (ERR_NESTING_EXCEEDED); |
248 | |
487 | |
|
|
488 | encode_ch (enc, '['); |
|
|
489 | |
|
|
490 | if (len >= 0) |
|
|
491 | { |
|
|
492 | encode_nl (enc); ++enc->indent; |
|
|
493 | |
249 | for (i = 0; i <= len; ++i) |
494 | for (i = 0; i <= len; ++i) |
250 | { |
495 | { |
|
|
496 | SV **svp = av_fetch (av, i, 0); |
|
|
497 | |
251 | encode_indent (enc); |
498 | encode_indent (enc); |
252 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
253 | |
499 | |
|
|
500 | if (svp) |
|
|
501 | encode_sv (enc, *svp); |
|
|
502 | else |
|
|
503 | encode_str (enc, "null", 4, 0); |
|
|
504 | |
254 | if (i < len) |
505 | if (i < len) |
255 | encode_comma (enc); |
506 | encode_comma (enc); |
256 | } |
507 | } |
257 | |
508 | |
|
|
509 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
510 | } |
|
|
511 | |
258 | encode_nl (enc); |
512 | encode_ch (enc, ']'); |
259 | |
|
|
260 | --enc->indent; |
|
|
261 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
262 | } |
513 | } |
263 | |
514 | |
264 | static void |
515 | static void |
265 | encode_he (enc_t *enc, HE *he) |
516 | encode_hk (enc_t *enc, HE *he) |
266 | { |
517 | { |
267 | encode_ch (enc, '"'); |
518 | encode_ch (enc, '"'); |
268 | |
519 | |
269 | if (HeKLEN (he) == HEf_SVKEY) |
520 | if (HeKLEN (he) == HEf_SVKEY) |
270 | { |
521 | { |
271 | SV *sv = HeSVKEY (he); |
522 | SV *sv = HeSVKEY (he); |
272 | STRLEN len; |
523 | STRLEN len; |
273 | char *str; |
524 | char *str; |
274 | |
525 | |
275 | SvGETMAGIC (sv); |
526 | SvGETMAGIC (sv); |
276 | str = SvPV (sv, len); |
527 | str = SvPV (sv, len); |
277 | |
528 | |
278 | encode_str (enc, str, len, SvUTF8 (sv)); |
529 | encode_str (enc, str, len, SvUTF8 (sv)); |
279 | } |
530 | } |
280 | else |
531 | else |
281 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
532 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
282 | |
533 | |
283 | encode_ch (enc, '"'); |
534 | encode_ch (enc, '"'); |
284 | |
535 | |
285 | if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
536 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
286 | encode_ch (enc, ':'); |
537 | encode_ch (enc, ':'); |
287 | if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
538 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
288 | encode_sv (enc, HeVAL (he)); |
|
|
289 | } |
539 | } |
290 | |
540 | |
291 | // compare hash entries, used when all keys are bytestrings |
541 | // compare hash entries, used when all keys are bytestrings |
292 | static int |
542 | static int |
293 | he_cmp_fast (const void *a_, const void *b_) |
543 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
298 | HE *b = *(HE **)b_; |
548 | HE *b = *(HE **)b_; |
299 | |
549 | |
300 | STRLEN la = HeKLEN (a); |
550 | STRLEN la = HeKLEN (a); |
301 | STRLEN lb = HeKLEN (b); |
551 | STRLEN lb = HeKLEN (b); |
302 | |
552 | |
303 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
553 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
304 | cmp = la - lb; |
554 | cmp = lb - la; |
305 | |
555 | |
306 | return cmp; |
556 | return cmp; |
307 | } |
557 | } |
308 | |
558 | |
309 | // compare hash entries, used when some keys are sv's or utf-x |
559 | // compare hash entries, used when some keys are sv's or utf-x |
310 | static int |
560 | static int |
311 | he_cmp_slow (const void *a, const void *b) |
561 | he_cmp_slow (const void *a, const void *b) |
312 | { |
562 | { |
313 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
563 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
314 | } |
564 | } |
315 | |
565 | |
316 | static void |
566 | static void |
317 | encode_hv (enc_t *enc, HV *hv) |
567 | encode_hv (enc_t *enc, HV *hv) |
318 | { |
568 | { |
319 | int count, i; |
569 | HE *he; |
320 | |
570 | |
321 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
571 | if (enc->indent >= enc->json.max_depth) |
|
|
572 | croak (ERR_NESTING_EXCEEDED); |
322 | |
573 | |
323 | if ((count = hv_iterinit (hv))) |
574 | encode_ch (enc, '{'); |
324 | { |
575 | |
325 | // for canonical output we have to sort by keys first |
576 | // for canonical output we have to sort by keys first |
326 | // actually, this is mostly due to the stupid so-called |
577 | // actually, this is mostly due to the stupid so-called |
327 | // security workaround added somewhere in 5.8.x. |
578 | // security workaround added somewhere in 5.8.x |
328 | // that randomises hash orderings |
579 | // that randomises hash orderings |
329 | if (enc->flags & F_CANONICAL) |
580 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
581 | { |
|
|
582 | int count = hv_iterinit (hv); |
|
|
583 | |
|
|
584 | if (SvMAGICAL (hv)) |
330 | { |
585 | { |
331 | HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode |
586 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
587 | // but I don't care for the speed of this special case. |
|
|
588 | // note also that we will run into undefined behaviour when the two iterations |
|
|
589 | // do not result in the same count, something I might care for in some later release. |
|
|
590 | |
|
|
591 | count = 0; |
|
|
592 | while (hv_iternext (hv)) |
|
|
593 | ++count; |
|
|
594 | |
|
|
595 | hv_iterinit (hv); |
|
|
596 | } |
|
|
597 | |
|
|
598 | if (count) |
|
|
599 | { |
332 | int fast = 1; |
600 | int i, fast = 1; |
|
|
601 | HE *hes_stack [STACK_HES]; |
|
|
602 | HE **hes = hes_stack; |
|
|
603 | |
|
|
604 | // allocate larger arrays on the heap |
|
|
605 | if (count > STACK_HES) |
|
|
606 | { |
|
|
607 | SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes))); |
|
|
608 | hes = (HE **)SvPVX (sv); |
|
|
609 | } |
333 | |
610 | |
334 | i = 0; |
611 | i = 0; |
335 | while ((he = hv_iternext (hv))) |
612 | while ((he = hv_iternext (hv))) |
336 | { |
613 | { |
337 | hes [i++] = he; |
614 | hes [i++] = he; |
… | |
… | |
359 | |
636 | |
360 | FREETMPS; |
637 | FREETMPS; |
361 | LEAVE; |
638 | LEAVE; |
362 | } |
639 | } |
363 | |
640 | |
364 | for (i = 0; i < count; ++i) |
641 | encode_nl (enc); ++enc->indent; |
|
|
642 | |
|
|
643 | while (count--) |
365 | { |
644 | { |
366 | encode_indent (enc); |
645 | encode_indent (enc); |
|
|
646 | he = hes [count]; |
367 | encode_he (enc, hes [i]); |
647 | encode_hk (enc, he); |
|
|
648 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
368 | |
649 | |
369 | if (i < count - 1) |
650 | if (count) |
370 | encode_comma (enc); |
651 | encode_comma (enc); |
371 | } |
652 | } |
372 | |
653 | |
373 | encode_nl (enc); |
654 | encode_nl (enc); --enc->indent; encode_indent (enc); |
374 | } |
655 | } |
|
|
656 | } |
|
|
657 | else |
|
|
658 | { |
|
|
659 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
660 | if ((he = hv_iternext (hv))) |
|
|
661 | { |
|
|
662 | encode_nl (enc); ++enc->indent; |
|
|
663 | |
|
|
664 | for (;;) |
|
|
665 | { |
|
|
666 | encode_indent (enc); |
|
|
667 | encode_hk (enc, he); |
|
|
668 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
|
|
669 | |
|
|
670 | if (!(he = hv_iternext (hv))) |
|
|
671 | break; |
|
|
672 | |
|
|
673 | encode_comma (enc); |
|
|
674 | } |
|
|
675 | |
|
|
676 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
677 | } |
|
|
678 | } |
|
|
679 | |
|
|
680 | encode_ch (enc, '}'); |
|
|
681 | } |
|
|
682 | |
|
|
683 | // encode objects, arrays and special \0=false and \1=true values. |
|
|
684 | static void |
|
|
685 | encode_rv (enc_t *enc, SV *sv) |
|
|
686 | { |
|
|
687 | svtype svt; |
|
|
688 | GV *method; |
|
|
689 | |
|
|
690 | SvGETMAGIC (sv); |
|
|
691 | svt = SvTYPE (sv); |
|
|
692 | |
|
|
693 | if (expect_false (SvOBJECT (sv))) |
|
|
694 | { |
|
|
695 | HV *boolean_stash = !JSON_SLOW || types_boolean_stash |
|
|
696 | ? types_boolean_stash |
|
|
697 | : gv_stashpv ("Types::Serialiser::Boolean", 1); |
|
|
698 | HV *stash = SvSTASH (sv); |
|
|
699 | |
|
|
700 | if (stash == boolean_stash) |
|
|
701 | { |
|
|
702 | if (SvIV (sv)) |
|
|
703 | encode_str (enc, "true", 4, 0); |
|
|
704 | else |
|
|
705 | encode_str (enc, "false", 5, 0); |
|
|
706 | } |
|
|
707 | else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0))) |
|
|
708 | { |
|
|
709 | int count; |
|
|
710 | dSP; |
|
|
711 | |
|
|
712 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
713 | EXTEND (SP, 2); |
|
|
714 | // we re-bless the reference to get overload and other niceties right |
|
|
715 | PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
|
|
716 | PUSHs (sv_json); |
|
|
717 | |
|
|
718 | PUTBACK; |
|
|
719 | count = call_sv ((SV *)GvCV (method), G_ARRAY); |
|
|
720 | SPAGAIN; |
|
|
721 | |
|
|
722 | // catch this surprisingly common error |
|
|
723 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
724 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
725 | |
|
|
726 | encode_ch (enc, '('); |
|
|
727 | encode_ch (enc, '"'); |
|
|
728 | encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash)); |
|
|
729 | encode_ch (enc, '"'); |
|
|
730 | encode_ch (enc, ')'); |
|
|
731 | encode_ch (enc, '['); |
|
|
732 | |
|
|
733 | while (count) |
|
|
734 | { |
|
|
735 | encode_sv (enc, SP[1 - count--]); |
|
|
736 | |
|
|
737 | if (count) |
|
|
738 | encode_ch (enc, ','); |
|
|
739 | } |
|
|
740 | |
|
|
741 | encode_ch (enc, ']'); |
|
|
742 | |
|
|
743 | PUTBACK; |
|
|
744 | |
|
|
745 | FREETMPS; LEAVE; |
|
|
746 | } |
|
|
747 | else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0))) |
|
|
748 | { |
|
|
749 | dSP; |
|
|
750 | |
|
|
751 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
752 | // we re-bless the reference to get overload and other niceties right |
|
|
753 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
|
|
754 | |
|
|
755 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
756 | PUTBACK; |
|
|
757 | call_sv ((SV *)GvCV (method), G_SCALAR); |
|
|
758 | SPAGAIN; |
|
|
759 | |
|
|
760 | // catch this surprisingly common error |
|
|
761 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
762 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
763 | |
|
|
764 | sv = POPs; |
|
|
765 | PUTBACK; |
|
|
766 | |
|
|
767 | encode_sv (enc, sv); |
|
|
768 | |
|
|
769 | FREETMPS; LEAVE; |
|
|
770 | } |
|
|
771 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
772 | encode_str (enc, "null", 4, 0); |
375 | else |
773 | else |
376 | { |
774 | croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)", |
377 | SV *sv; |
775 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
378 | HE *he = hv_iternext (hv); |
776 | } |
379 | |
777 | else if (svt == SVt_PVHV) |
380 | for (;;) |
778 | encode_hv (enc, (HV *)sv); |
381 | { |
779 | else if (svt == SVt_PVAV) |
382 | encode_indent (enc); |
780 | encode_av (enc, (AV *)sv); |
383 | encode_he (enc, he); |
781 | else if (svt < SVt_PVAV) |
384 | |
|
|
385 | if (!(he = hv_iternext (hv))) |
|
|
386 | break; |
|
|
387 | |
|
|
388 | encode_comma (enc); |
|
|
389 | } |
|
|
390 | |
|
|
391 | encode_nl (enc); |
|
|
392 | } |
|
|
393 | } |
782 | { |
|
|
783 | STRLEN len = 0; |
|
|
784 | char *pv = svt ? SvPV (sv, len) : 0; |
394 | |
785 | |
395 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
786 | if (len == 1 && *pv == '1') |
|
|
787 | encode_str (enc, "true", 4, 0); |
|
|
788 | else if (len == 1 && *pv == '0') |
|
|
789 | encode_str (enc, "false", 5, 0); |
|
|
790 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
791 | encode_str (enc, "null", 4, 0); |
|
|
792 | else |
|
|
793 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
|
|
794 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
795 | } |
|
|
796 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
797 | encode_str (enc, "null", 4, 0); |
|
|
798 | else |
|
|
799 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
800 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
396 | } |
801 | } |
397 | |
802 | |
398 | static void |
803 | static void |
399 | encode_sv (enc_t *enc, SV *sv) |
804 | encode_sv (enc_t *enc, SV *sv) |
400 | { |
805 | { |
… | |
… | |
408 | encode_str (enc, str, len, SvUTF8 (sv)); |
813 | encode_str (enc, str, len, SvUTF8 (sv)); |
409 | encode_ch (enc, '"'); |
814 | encode_ch (enc, '"'); |
410 | } |
815 | } |
411 | else if (SvNOKp (sv)) |
816 | else if (SvNOKp (sv)) |
412 | { |
817 | { |
|
|
818 | // trust that perl will do the right thing w.r.t. JSON syntax. |
413 | need (enc, NV_DIG + 32); |
819 | need (enc, NV_DIG + 32); |
414 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
820 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
415 | enc->cur += strlen (enc->cur); |
821 | enc->cur += strlen (enc->cur); |
416 | } |
822 | } |
417 | else if (SvIOKp (sv)) |
823 | else if (SvIOKp (sv)) |
418 | { |
824 | { |
|
|
825 | // we assume we can always read an IV as a UV and vice versa |
|
|
826 | // we assume two's complement |
|
|
827 | // we assume no aliasing issues in the union |
|
|
828 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
|
|
829 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
|
|
830 | { |
|
|
831 | // optimise the "small number case" |
|
|
832 | // code will likely be branchless and use only a single multiplication |
|
|
833 | // works for numbers up to 59074 |
|
|
834 | I32 i = SvIVX (sv); |
|
|
835 | U32 u; |
|
|
836 | char digit, nz = 0; |
|
|
837 | |
419 | need (enc, 64); |
838 | need (enc, 6); |
|
|
839 | |
|
|
840 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
841 | u = i < 0 ? -i : i; |
|
|
842 | |
|
|
843 | // convert to 4.28 fixed-point representation |
|
|
844 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
845 | |
|
|
846 | // now output digit by digit, each time masking out the integer part |
|
|
847 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
848 | // resulting in a net multiplication by 10. |
|
|
849 | // we always write the digit to memory but conditionally increment |
|
|
850 | // the pointer, to enable the use of conditional move instructions. |
|
|
851 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
|
|
852 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
|
|
853 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
|
|
854 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
|
|
855 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
856 | } |
|
|
857 | else |
|
|
858 | { |
|
|
859 | // large integer, use the (rather slow) snprintf way. |
|
|
860 | need (enc, IVUV_MAXCHARS); |
420 | enc->cur += |
861 | enc->cur += |
421 | SvIsUV(sv) |
862 | SvIsUV(sv) |
422 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
863 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
423 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
864 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
865 | } |
424 | } |
866 | } |
425 | else if (SvROK (sv)) |
867 | else if (SvROK (sv)) |
426 | { |
868 | encode_rv (enc, SvRV (sv)); |
427 | SV *rv = SvRV (sv); |
869 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
428 | |
|
|
429 | if (enc->indent >= enc->max_depth) |
|
|
430 | croak ("data structure too deep (hit recursion limit)"); |
|
|
431 | |
|
|
432 | switch (SvTYPE (rv)) |
|
|
433 | { |
|
|
434 | case SVt_PVAV: encode_av (enc, (AV *)rv); break; |
|
|
435 | case SVt_PVHV: encode_hv (enc, (HV *)rv); break; |
|
|
436 | |
|
|
437 | default: |
|
|
438 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
439 | SvPV_nolen (sv)); |
|
|
440 | } |
|
|
441 | } |
|
|
442 | else if (!SvOK (sv)) |
|
|
443 | encode_str (enc, "null", 4, 0); |
870 | encode_str (enc, "null", 4, 0); |
444 | else |
871 | else |
445 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
872 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data", |
446 | SvPV_nolen (sv), SvFLAGS (sv)); |
873 | SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
447 | } |
874 | } |
448 | |
875 | |
449 | static SV * |
876 | static SV * |
450 | encode_json (SV *scalar, UV flags) |
877 | encode_json (SV *scalar, JSON *json) |
451 | { |
878 | { |
|
|
879 | enc_t enc; |
|
|
880 | |
452 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
881 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
453 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
882 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
454 | |
883 | |
455 | enc_t enc; |
884 | enc.json = *json; |
456 | enc.flags = flags; |
|
|
457 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
885 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
458 | enc.cur = SvPVX (enc.sv); |
886 | enc.cur = SvPVX (enc.sv); |
459 | enc.end = SvEND (enc.sv); |
887 | enc.end = SvEND (enc.sv); |
460 | enc.indent = 0; |
888 | enc.indent = 0; |
461 | enc.max_depth = 0x7fffffffUL; |
889 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
890 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
891 | : 0x110000UL; |
462 | |
892 | |
463 | SvPOK_only (enc.sv); |
893 | SvPOK_only (enc.sv); |
464 | encode_sv (&enc, scalar); |
894 | encode_sv (&enc, scalar); |
|
|
895 | encode_nl (&enc); |
465 | |
896 | |
|
|
897 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
898 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
899 | |
466 | if (!(flags & (F_ASCII | F_UTF8))) |
900 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
467 | SvUTF8_on (enc.sv); |
901 | SvUTF8_on (enc.sv); |
468 | |
902 | |
469 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
470 | |
|
|
471 | if (enc.flags & F_SHRINK) |
903 | if (enc.json.flags & F_SHRINK) |
472 | shrink (enc.sv); |
904 | shrink (enc.sv); |
473 | |
905 | |
474 | return enc.sv; |
906 | return enc.sv; |
475 | } |
907 | } |
476 | |
908 | |
… | |
… | |
481 | typedef struct |
913 | typedef struct |
482 | { |
914 | { |
483 | char *cur; // current parser pointer |
915 | char *cur; // current parser pointer |
484 | char *end; // end of input string |
916 | char *end; // end of input string |
485 | const char *err; // parse error, if != 0 |
917 | const char *err; // parse error, if != 0 |
486 | UV flags; // F_* |
918 | JSON json; |
|
|
919 | U32 depth; // recursion depth |
|
|
920 | U32 maxdepth; // recursion depth limit |
487 | } dec_t; |
921 | } dec_t; |
488 | |
922 | |
489 | static void |
923 | INLINE void |
|
|
924 | decode_comment (dec_t *dec) |
|
|
925 | { |
|
|
926 | // only '#'-style comments allowed a.t.m. |
|
|
927 | |
|
|
928 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
929 | ++dec->cur; |
|
|
930 | } |
|
|
931 | |
|
|
932 | INLINE void |
490 | decode_ws (dec_t *dec) |
933 | decode_ws (dec_t *dec) |
491 | { |
934 | { |
492 | for (;;) |
935 | for (;;) |
493 | { |
936 | { |
494 | char ch = *dec->cur; |
937 | char ch = *dec->cur; |
495 | |
938 | |
496 | if (ch > 0x20 |
939 | if (ch > 0x20) |
|
|
940 | { |
|
|
941 | if (expect_false (ch == '#')) |
|
|
942 | { |
|
|
943 | if (dec->json.flags & F_RELAXED) |
|
|
944 | decode_comment (dec); |
|
|
945 | else |
|
|
946 | break; |
|
|
947 | } |
|
|
948 | else |
|
|
949 | break; |
|
|
950 | } |
497 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
951 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
498 | break; |
952 | break; // parse error, but let higher level handle it, gives better error messages |
499 | |
953 | |
500 | ++dec->cur; |
954 | ++dec->cur; |
501 | } |
955 | } |
502 | } |
956 | } |
503 | |
957 | |
504 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
958 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
|
|
959 | |
505 | #define EXPECT_CH(ch) SB \ |
960 | #define EXPECT_CH(ch) SB \ |
506 | if (*dec->cur != ch) \ |
961 | if (*dec->cur != ch) \ |
507 | ERR (# ch " expected"); \ |
962 | ERR (# ch " expected"); \ |
508 | ++dec->cur; \ |
963 | ++dec->cur; \ |
509 | SE |
964 | SE |
510 | |
965 | |
|
|
966 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
|
|
967 | #define DEC_DEC_DEPTH --dec->depth |
|
|
968 | |
511 | static SV *decode_sv (dec_t *dec); |
969 | static SV *decode_sv (dec_t *dec); |
512 | |
970 | |
513 | static signed char decode_hexdigit[256]; |
971 | static signed char decode_hexdigit[256]; |
514 | |
972 | |
515 | static UV |
973 | static UV |
516 | decode_4hex (dec_t *dec) |
974 | decode_4hex (dec_t *dec) |
517 | { |
975 | { |
518 | signed char d1, d2, d3, d4; |
976 | signed char d1, d2, d3, d4; |
519 | unsigned char *cur = (unsigned char *)dec->cur; |
977 | unsigned char *cur = (unsigned char *)dec->cur; |
520 | |
978 | |
521 | d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); |
979 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
522 | d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); |
980 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
523 | d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); |
981 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
524 | d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); |
982 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
525 | |
983 | |
526 | dec->cur += 4; |
984 | dec->cur += 4; |
527 | |
985 | |
528 | return ((UV)d1) << 12 |
986 | return ((UV)d1) << 12 |
529 | | ((UV)d2) << 8 |
987 | | ((UV)d2) << 8 |
… | |
… | |
537 | static SV * |
995 | static SV * |
538 | decode_str (dec_t *dec) |
996 | decode_str (dec_t *dec) |
539 | { |
997 | { |
540 | SV *sv = 0; |
998 | SV *sv = 0; |
541 | int utf8 = 0; |
999 | int utf8 = 0; |
|
|
1000 | char *dec_cur = dec->cur; |
542 | |
1001 | |
543 | do |
1002 | do |
544 | { |
1003 | { |
545 | char buf [SHORT_STRING_LEN + UTF8_MAX_LEN]; |
1004 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
546 | char *cur = buf; |
1005 | char *cur = buf; |
547 | |
1006 | |
548 | do |
1007 | do |
549 | { |
1008 | { |
550 | unsigned char ch = *(unsigned char *)dec->cur++; |
1009 | unsigned char ch = *(unsigned char *)dec_cur++; |
551 | |
1010 | |
552 | if (ch == '"') |
1011 | if (expect_false (ch == '"')) |
553 | { |
1012 | { |
554 | --dec->cur; |
1013 | --dec_cur; |
555 | break; |
1014 | break; |
556 | } |
1015 | } |
557 | else if (ch == '\\') |
1016 | else if (expect_false (ch == '\\')) |
558 | { |
1017 | { |
559 | switch (*dec->cur) |
1018 | switch (*dec_cur) |
560 | { |
1019 | { |
561 | case '\\': |
1020 | case '\\': |
562 | case '/': |
1021 | case '/': |
563 | case '"': *cur++ = *dec->cur++; break; |
1022 | case '"': *cur++ = *dec_cur++; break; |
564 | |
1023 | |
565 | case 'b': ++dec->cur; *cur++ = '\010'; break; |
1024 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
566 | case 't': ++dec->cur; *cur++ = '\011'; break; |
1025 | case 't': ++dec_cur; *cur++ = '\011'; break; |
567 | case 'n': ++dec->cur; *cur++ = '\012'; break; |
1026 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
568 | case 'f': ++dec->cur; *cur++ = '\014'; break; |
1027 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
569 | case 'r': ++dec->cur; *cur++ = '\015'; break; |
1028 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
570 | |
1029 | |
571 | case 'u': |
1030 | case 'u': |
572 | { |
1031 | { |
573 | UV lo, hi; |
1032 | UV lo, hi; |
574 | ++dec->cur; |
1033 | ++dec_cur; |
575 | |
1034 | |
|
|
1035 | dec->cur = dec_cur; |
576 | hi = decode_4hex (dec); |
1036 | hi = decode_4hex (dec); |
|
|
1037 | dec_cur = dec->cur; |
577 | if (hi == (UV)-1) |
1038 | if (hi == (UV)-1) |
578 | goto fail; |
1039 | goto fail; |
579 | |
1040 | |
580 | // possibly a surrogate pair |
1041 | // possibly a surrogate pair |
581 | if (hi >= 0xd800) |
1042 | if (hi >= 0xd800) |
582 | if (hi < 0xdc00) |
1043 | if (hi < 0xdc00) |
583 | { |
1044 | { |
584 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
1045 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
585 | ERR ("missing low surrogate character in surrogate pair"); |
1046 | ERR ("missing low surrogate character in surrogate pair"); |
586 | |
1047 | |
587 | dec->cur += 2; |
1048 | dec_cur += 2; |
588 | |
1049 | |
|
|
1050 | dec->cur = dec_cur; |
589 | lo = decode_4hex (dec); |
1051 | lo = decode_4hex (dec); |
|
|
1052 | dec_cur = dec->cur; |
590 | if (lo == (UV)-1) |
1053 | if (lo == (UV)-1) |
591 | goto fail; |
1054 | goto fail; |
592 | |
1055 | |
593 | if (lo < 0xdc00 || lo >= 0xe000) |
1056 | if (lo < 0xdc00 || lo >= 0xe000) |
594 | ERR ("surrogate pair expected"); |
1057 | ERR ("surrogate pair expected"); |
… | |
… | |
600 | |
1063 | |
601 | if (hi >= 0x80) |
1064 | if (hi >= 0x80) |
602 | { |
1065 | { |
603 | utf8 = 1; |
1066 | utf8 = 1; |
604 | |
1067 | |
605 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
1068 | cur = encode_utf8 (cur, hi); |
606 | } |
1069 | } |
607 | else |
1070 | else |
608 | *cur++ = hi; |
1071 | *cur++ = hi; |
609 | } |
1072 | } |
610 | break; |
1073 | break; |
611 | |
1074 | |
612 | default: |
1075 | default: |
613 | --dec->cur; |
1076 | --dec_cur; |
614 | ERR ("illegal backslash escape sequence in string"); |
1077 | ERR ("illegal backslash escape sequence in string"); |
615 | } |
1078 | } |
616 | } |
1079 | } |
617 | else if (ch >= 0x20 && ch <= 0x7f) |
1080 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
618 | *cur++ = ch; |
1081 | *cur++ = ch; |
619 | else if (ch >= 0x80) |
1082 | else if (ch >= 0x80) |
620 | { |
1083 | { |
621 | --dec->cur; |
|
|
622 | |
|
|
623 | STRLEN clen; |
1084 | STRLEN clen; |
624 | UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); |
1085 | |
|
|
1086 | --dec_cur; |
|
|
1087 | |
|
|
1088 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
625 | if (clen == (STRLEN)-1) |
1089 | if (clen == (STRLEN)-1) |
626 | ERR ("malformed UTF-8 character in JSON string"); |
1090 | ERR ("malformed UTF-8 character in JSON string"); |
627 | |
1091 | |
628 | do |
1092 | do |
629 | { |
|
|
630 | *cur++ = *dec->cur++; |
1093 | *cur++ = *dec_cur++; |
631 | } |
|
|
632 | while (--clen); |
1094 | while (--clen); |
633 | |
1095 | |
634 | utf8 = 1; |
1096 | utf8 = 1; |
635 | } |
1097 | } |
636 | else if (!ch) |
|
|
637 | ERR ("unexpected end of string while parsing json string"); |
|
|
638 | else |
1098 | else |
|
|
1099 | { |
|
|
1100 | --dec_cur; |
|
|
1101 | |
|
|
1102 | if (!ch) |
|
|
1103 | ERR ("unexpected end of string while parsing JSON string"); |
|
|
1104 | else |
639 | ERR ("invalid character encountered"); |
1105 | ERR ("invalid character encountered while parsing JSON string"); |
640 | |
1106 | } |
641 | } |
1107 | } |
642 | while (cur < buf + SHORT_STRING_LEN); |
1108 | while (cur < buf + SHORT_STRING_LEN); |
643 | |
1109 | |
|
|
1110 | { |
644 | STRLEN len = cur - buf; |
1111 | STRLEN len = cur - buf; |
645 | |
1112 | |
646 | if (sv) |
1113 | if (sv) |
647 | { |
1114 | { |
648 | SvGROW (sv, SvCUR (sv) + len + 1); |
1115 | STRLEN cur = SvCUR (sv); |
|
|
1116 | |
|
|
1117 | if (SvLEN (sv) <= cur + len) |
|
|
1118 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1119 | |
649 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1120 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
650 | SvCUR_set (sv, SvCUR (sv) + len); |
1121 | SvCUR_set (sv, SvCUR (sv) + len); |
651 | } |
1122 | } |
652 | else |
1123 | else |
653 | sv = newSVpvn (buf, len); |
1124 | sv = newSVpvn (buf, len); |
654 | } |
1125 | } |
|
|
1126 | } |
655 | while (*dec->cur != '"'); |
1127 | while (*dec_cur != '"'); |
656 | |
1128 | |
657 | ++dec->cur; |
1129 | ++dec_cur; |
658 | |
1130 | |
659 | if (sv) |
1131 | if (sv) |
660 | { |
1132 | { |
661 | SvPOK_only (sv); |
1133 | SvPOK_only (sv); |
662 | *SvEND (sv) = 0; |
1134 | *SvEND (sv) = 0; |
… | |
… | |
665 | SvUTF8_on (sv); |
1137 | SvUTF8_on (sv); |
666 | } |
1138 | } |
667 | else |
1139 | else |
668 | sv = newSVpvn ("", 0); |
1140 | sv = newSVpvn ("", 0); |
669 | |
1141 | |
|
|
1142 | dec->cur = dec_cur; |
670 | return sv; |
1143 | return sv; |
671 | |
1144 | |
672 | fail: |
1145 | fail: |
|
|
1146 | dec->cur = dec_cur; |
673 | return 0; |
1147 | return 0; |
674 | } |
1148 | } |
675 | |
1149 | |
676 | static SV * |
1150 | static SV * |
677 | decode_num (dec_t *dec) |
1151 | decode_num (dec_t *dec) |
… | |
… | |
735 | is_nv = 1; |
1209 | is_nv = 1; |
736 | } |
1210 | } |
737 | |
1211 | |
738 | if (!is_nv) |
1212 | if (!is_nv) |
739 | { |
1213 | { |
740 | UV uv; |
1214 | int len = dec->cur - start; |
741 | int numtype = grok_number (start, dec->cur - start, &uv); |
1215 | |
742 | if (numtype & IS_NUMBER_IN_UV) |
1216 | // special case the rather common 1..5-digit-int case |
743 | if (numtype & IS_NUMBER_NEG) |
1217 | if (*start == '-') |
|
|
1218 | switch (len) |
744 | { |
1219 | { |
745 | if (uv < (UV)IV_MIN) |
1220 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
746 | return newSViv (-(IV)uv); |
1221 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
|
|
1222 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
1223 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1224 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
747 | } |
1225 | } |
|
|
1226 | else |
|
|
1227 | switch (len) |
|
|
1228 | { |
|
|
1229 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
1230 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
1231 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
1232 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1233 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
|
|
1234 | } |
|
|
1235 | |
|
|
1236 | { |
|
|
1237 | UV uv; |
|
|
1238 | int numtype = grok_number (start, len, &uv); |
|
|
1239 | if (numtype & IS_NUMBER_IN_UV) |
|
|
1240 | if (numtype & IS_NUMBER_NEG) |
|
|
1241 | { |
|
|
1242 | if (uv < (UV)IV_MIN) |
|
|
1243 | return newSViv (-(IV)uv); |
|
|
1244 | } |
748 | else |
1245 | else |
749 | return newSVuv (uv); |
1246 | return newSVuv (uv); |
750 | } |
1247 | } |
751 | |
1248 | |
|
|
1249 | len -= *start == '-' ? 1 : 0; |
|
|
1250 | |
|
|
1251 | // does not fit into IV or UV, try NV |
|
|
1252 | if (len <= NV_DIG) |
|
|
1253 | // fits into NV without loss of precision |
|
|
1254 | return newSVnv (json_atof (start)); |
|
|
1255 | |
|
|
1256 | // everything else fails, convert it to a string |
|
|
1257 | return newSVpvn (start, dec->cur - start); |
|
|
1258 | } |
|
|
1259 | |
|
|
1260 | // loss of precision here |
752 | return newSVnv (Atof (start)); |
1261 | return newSVnv (json_atof (start)); |
753 | |
1262 | |
754 | fail: |
1263 | fail: |
755 | return 0; |
1264 | return 0; |
756 | } |
1265 | } |
757 | |
1266 | |
758 | static SV * |
1267 | static SV * |
759 | decode_av (dec_t *dec) |
1268 | decode_av (dec_t *dec) |
760 | { |
1269 | { |
761 | AV *av = newAV (); |
1270 | AV *av = newAV (); |
762 | |
1271 | |
|
|
1272 | DEC_INC_DEPTH; |
763 | decode_ws (dec); |
1273 | decode_ws (dec); |
|
|
1274 | |
764 | if (*dec->cur == ']') |
1275 | if (*dec->cur == ']') |
765 | ++dec->cur; |
1276 | ++dec->cur; |
766 | else |
1277 | else |
767 | for (;;) |
1278 | for (;;) |
768 | { |
1279 | { |
… | |
… | |
779 | if (*dec->cur == ']') |
1290 | if (*dec->cur == ']') |
780 | { |
1291 | { |
781 | ++dec->cur; |
1292 | ++dec->cur; |
782 | break; |
1293 | break; |
783 | } |
1294 | } |
784 | |
1295 | |
785 | if (*dec->cur != ',') |
1296 | if (*dec->cur != ',') |
786 | ERR (", or ] expected while parsing array"); |
1297 | ERR (", or ] expected while parsing array"); |
787 | |
1298 | |
788 | ++dec->cur; |
1299 | ++dec->cur; |
|
|
1300 | |
|
|
1301 | decode_ws (dec); |
|
|
1302 | |
|
|
1303 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1304 | { |
|
|
1305 | ++dec->cur; |
|
|
1306 | break; |
|
|
1307 | } |
789 | } |
1308 | } |
790 | |
1309 | |
|
|
1310 | DEC_DEC_DEPTH; |
791 | return newRV_noinc ((SV *)av); |
1311 | return newRV_noinc ((SV *)av); |
792 | |
1312 | |
793 | fail: |
1313 | fail: |
794 | SvREFCNT_dec (av); |
1314 | SvREFCNT_dec (av); |
|
|
1315 | DEC_DEC_DEPTH; |
795 | return 0; |
1316 | return 0; |
796 | } |
1317 | } |
797 | |
1318 | |
798 | static SV * |
1319 | static SV * |
799 | decode_hv (dec_t *dec) |
1320 | decode_hv (dec_t *dec) |
800 | { |
1321 | { |
|
|
1322 | SV *sv; |
801 | HV *hv = newHV (); |
1323 | HV *hv = newHV (); |
802 | |
1324 | |
|
|
1325 | DEC_INC_DEPTH; |
803 | decode_ws (dec); |
1326 | decode_ws (dec); |
|
|
1327 | |
804 | if (*dec->cur == '}') |
1328 | if (*dec->cur == '}') |
805 | ++dec->cur; |
1329 | ++dec->cur; |
806 | else |
1330 | else |
807 | for (;;) |
1331 | for (;;) |
808 | { |
1332 | { |
|
|
1333 | EXPECT_CH ('"'); |
|
|
1334 | |
|
|
1335 | // heuristic: assume that |
|
|
1336 | // a) decode_str + hv_store_ent are abysmally slow. |
|
|
1337 | // b) most hash keys are short, simple ascii text. |
|
|
1338 | // => try to "fast-match" such strings to avoid |
|
|
1339 | // the overhead of decode_str + hv_store_ent. |
|
|
1340 | { |
809 | SV *key, *value; |
1341 | SV *value; |
|
|
1342 | char *p = dec->cur; |
|
|
1343 | char *e = p + 24; // only try up to 24 bytes |
810 | |
1344 | |
811 | decode_ws (dec); EXPECT_CH ('"'); |
1345 | for (;;) |
812 | |
|
|
813 | key = decode_str (dec); |
|
|
814 | if (!key) |
|
|
815 | goto fail; |
|
|
816 | |
|
|
817 | decode_ws (dec); EXPECT_CH (':'); |
|
|
818 | |
|
|
819 | value = decode_sv (dec); |
|
|
820 | if (!value) |
|
|
821 | { |
1346 | { |
|
|
1347 | // the >= 0x80 is false on most architectures |
|
|
1348 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1349 | { |
|
|
1350 | // slow path, back up and use decode_str |
|
|
1351 | SV *key = decode_str (dec); |
|
|
1352 | if (!key) |
|
|
1353 | goto fail; |
|
|
1354 | |
|
|
1355 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1356 | |
|
|
1357 | decode_ws (dec); |
|
|
1358 | value = decode_sv (dec); |
|
|
1359 | if (!value) |
|
|
1360 | { |
|
|
1361 | SvREFCNT_dec (key); |
|
|
1362 | goto fail; |
|
|
1363 | } |
|
|
1364 | |
|
|
1365 | hv_store_ent (hv, key, value, 0); |
822 | SvREFCNT_dec (key); |
1366 | SvREFCNT_dec (key); |
|
|
1367 | |
|
|
1368 | break; |
|
|
1369 | } |
|
|
1370 | else if (*p == '"') |
|
|
1371 | { |
|
|
1372 | // fast path, got a simple key |
|
|
1373 | char *key = dec->cur; |
|
|
1374 | int len = p - key; |
|
|
1375 | dec->cur = p + 1; |
|
|
1376 | |
|
|
1377 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1378 | |
|
|
1379 | decode_ws (dec); |
|
|
1380 | value = decode_sv (dec); |
|
|
1381 | if (!value) |
823 | goto fail; |
1382 | goto fail; |
|
|
1383 | |
|
|
1384 | hv_store (hv, key, len, value, 0); |
|
|
1385 | |
|
|
1386 | break; |
|
|
1387 | } |
|
|
1388 | |
|
|
1389 | ++p; |
824 | } |
1390 | } |
825 | |
1391 | } |
826 | //TODO: optimise |
|
|
827 | hv_store_ent (hv, key, value, 0); |
|
|
828 | |
1392 | |
829 | decode_ws (dec); |
1393 | decode_ws (dec); |
830 | |
1394 | |
831 | if (*dec->cur == '}') |
1395 | if (*dec->cur == '}') |
832 | { |
1396 | { |
… | |
… | |
836 | |
1400 | |
837 | if (*dec->cur != ',') |
1401 | if (*dec->cur != ',') |
838 | ERR (", or } expected while parsing object/hash"); |
1402 | ERR (", or } expected while parsing object/hash"); |
839 | |
1403 | |
840 | ++dec->cur; |
1404 | ++dec->cur; |
|
|
1405 | |
|
|
1406 | decode_ws (dec); |
|
|
1407 | |
|
|
1408 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1409 | { |
|
|
1410 | ++dec->cur; |
|
|
1411 | break; |
|
|
1412 | } |
841 | } |
1413 | } |
842 | |
1414 | |
|
|
1415 | DEC_DEC_DEPTH; |
843 | return newRV_noinc ((SV *)hv); |
1416 | sv = newRV_noinc ((SV *)hv); |
|
|
1417 | |
|
|
1418 | // check filter callbacks |
|
|
1419 | if (dec->json.flags & F_HOOK) |
|
|
1420 | { |
|
|
1421 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1422 | { |
|
|
1423 | HE *cb, *he; |
|
|
1424 | |
|
|
1425 | hv_iterinit (hv); |
|
|
1426 | he = hv_iternext (hv); |
|
|
1427 | hv_iterinit (hv); |
|
|
1428 | |
|
|
1429 | // the next line creates a mortal sv each time its called. |
|
|
1430 | // might want to optimise this for common cases. |
|
|
1431 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1432 | |
|
|
1433 | if (cb) |
|
|
1434 | { |
|
|
1435 | dSP; |
|
|
1436 | int count; |
|
|
1437 | |
|
|
1438 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1439 | XPUSHs (HeVAL (he)); |
|
|
1440 | sv_2mortal (sv); |
|
|
1441 | |
|
|
1442 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1443 | |
|
|
1444 | if (count == 1) |
|
|
1445 | { |
|
|
1446 | sv = newSVsv (POPs); |
|
|
1447 | FREETMPS; LEAVE; |
|
|
1448 | return sv; |
|
|
1449 | } |
|
|
1450 | |
|
|
1451 | SvREFCNT_inc (sv); |
|
|
1452 | FREETMPS; LEAVE; |
|
|
1453 | } |
|
|
1454 | } |
|
|
1455 | |
|
|
1456 | if (dec->json.cb_object) |
|
|
1457 | { |
|
|
1458 | dSP; |
|
|
1459 | int count; |
|
|
1460 | |
|
|
1461 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1462 | XPUSHs (sv_2mortal (sv)); |
|
|
1463 | |
|
|
1464 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1465 | |
|
|
1466 | if (count == 1) |
|
|
1467 | { |
|
|
1468 | sv = newSVsv (POPs); |
|
|
1469 | FREETMPS; LEAVE; |
|
|
1470 | return sv; |
|
|
1471 | } |
|
|
1472 | |
|
|
1473 | SvREFCNT_inc (sv); |
|
|
1474 | FREETMPS; LEAVE; |
|
|
1475 | } |
|
|
1476 | } |
|
|
1477 | |
|
|
1478 | return sv; |
844 | |
1479 | |
845 | fail: |
1480 | fail: |
846 | SvREFCNT_dec (hv); |
1481 | SvREFCNT_dec (hv); |
|
|
1482 | DEC_DEC_DEPTH; |
|
|
1483 | return 0; |
|
|
1484 | } |
|
|
1485 | |
|
|
1486 | static SV * |
|
|
1487 | decode_tag (dec_t *dec) |
|
|
1488 | { |
|
|
1489 | SV *tag = 0; |
|
|
1490 | SV *val = 0; |
|
|
1491 | |
|
|
1492 | if (!(dec->json.flags & F_ALLOW_TAGS)) |
|
|
1493 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
|
|
1494 | |
|
|
1495 | ++dec->cur; |
|
|
1496 | |
|
|
1497 | tag = decode_sv (dec); |
|
|
1498 | if (!tag) |
|
|
1499 | goto fail; |
|
|
1500 | |
|
|
1501 | if (!SvPOK (tag)) |
|
|
1502 | ERR ("malformed JSON string, (tag) must be a string"); |
|
|
1503 | |
|
|
1504 | if (*dec->cur != ')') |
|
|
1505 | ERR (") expected after tag"); |
|
|
1506 | |
|
|
1507 | ++dec->cur; |
|
|
1508 | |
|
|
1509 | val = decode_sv (dec); |
|
|
1510 | if (!val) |
|
|
1511 | goto fail; |
|
|
1512 | |
|
|
1513 | if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV) |
|
|
1514 | ERR ("malformed JSON string, tag value must be an array"); |
|
|
1515 | |
|
|
1516 | { |
|
|
1517 | AV *av = (AV *)SvRV (val); |
|
|
1518 | int i, len = av_len (av) + 1; |
|
|
1519 | HV *stash = gv_stashsv (tag, 0); |
|
|
1520 | SV *sv; |
|
|
1521 | |
|
|
1522 | if (!stash) |
|
|
1523 | ERR ("cannot decode perl-object (package does not exist)"); |
|
|
1524 | |
|
|
1525 | GV *method = gv_fetchmethod_autoload (stash, "THAW", 0); |
|
|
1526 | |
|
|
1527 | if (!method) |
|
|
1528 | ERR ("cannot decode perl-object (package does not have a THAW method)"); |
|
|
1529 | |
|
|
1530 | dSP; |
|
|
1531 | |
|
|
1532 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1533 | EXTEND (SP, len + 2); |
|
|
1534 | // we re-bless the reference to get overload and other niceties right |
|
|
1535 | PUSHs (tag); |
|
|
1536 | PUSHs (sv_json); |
|
|
1537 | |
|
|
1538 | for (i = 0; i < len; ++i) |
|
|
1539 | PUSHs (*av_fetch (av, i, 1)); |
|
|
1540 | |
|
|
1541 | PUTBACK; |
|
|
1542 | call_sv ((SV *)GvCV (method), G_SCALAR); |
|
|
1543 | SPAGAIN; |
|
|
1544 | |
|
|
1545 | SvREFCNT_dec (tag); |
|
|
1546 | SvREFCNT_dec (val); |
|
|
1547 | sv = SvREFCNT_inc (POPs); |
|
|
1548 | |
|
|
1549 | PUTBACK; |
|
|
1550 | |
|
|
1551 | FREETMPS; LEAVE; |
|
|
1552 | |
|
|
1553 | return sv; |
|
|
1554 | } |
|
|
1555 | |
|
|
1556 | fail: |
|
|
1557 | SvREFCNT_dec (tag); |
|
|
1558 | SvREFCNT_dec (val); |
847 | return 0; |
1559 | return 0; |
848 | } |
1560 | } |
849 | |
1561 | |
850 | static SV * |
1562 | static SV * |
851 | decode_sv (dec_t *dec) |
1563 | decode_sv (dec_t *dec) |
852 | { |
1564 | { |
853 | decode_ws (dec); |
1565 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1566 | // to parse everything. |
854 | switch (*dec->cur) |
1567 | switch (*dec->cur) |
855 | { |
1568 | { |
856 | case '"': ++dec->cur; return decode_str (dec); |
1569 | case '"': ++dec->cur; return decode_str (dec); |
857 | case '[': ++dec->cur; return decode_av (dec); |
1570 | case '[': ++dec->cur; return decode_av (dec); |
858 | case '{': ++dec->cur; return decode_hv (dec); |
1571 | case '{': ++dec->cur; return decode_hv (dec); |
|
|
1572 | case '(': return decode_tag (dec); |
859 | |
1573 | |
860 | case '-': |
1574 | case '-': |
861 | case '0': case '1': case '2': case '3': case '4': |
1575 | case '0': case '1': case '2': case '3': case '4': |
862 | case '5': case '6': case '7': case '8': case '9': |
1576 | case '5': case '6': case '7': case '8': case '9': |
863 | return decode_num (dec); |
1577 | return decode_num (dec); |
864 | |
1578 | |
865 | case 't': |
1579 | case 't': |
866 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1580 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
867 | { |
1581 | { |
868 | dec->cur += 4; |
1582 | dec->cur += 4; |
|
|
1583 | #if JSON_SLOW |
|
|
1584 | types_true = get_bool ("Types::Serialiser::true"); |
|
|
1585 | #endif |
869 | return newSViv (1); |
1586 | return newSVsv (types_true); |
870 | } |
1587 | } |
871 | else |
1588 | else |
872 | ERR ("'true' expected"); |
1589 | ERR ("'true' expected"); |
873 | |
1590 | |
874 | break; |
1591 | break; |
875 | |
1592 | |
876 | case 'f': |
1593 | case 'f': |
877 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1594 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
878 | { |
1595 | { |
879 | dec->cur += 5; |
1596 | dec->cur += 5; |
|
|
1597 | #if JSON_SLOW |
|
|
1598 | types_false = get_bool ("Types::Serialiser::false"); |
|
|
1599 | #endif |
880 | return newSViv (0); |
1600 | return newSVsv (types_false); |
881 | } |
1601 | } |
882 | else |
1602 | else |
883 | ERR ("'false' expected"); |
1603 | ERR ("'false' expected"); |
884 | |
1604 | |
885 | break; |
1605 | break; |
… | |
… | |
894 | ERR ("'null' expected"); |
1614 | ERR ("'null' expected"); |
895 | |
1615 | |
896 | break; |
1616 | break; |
897 | |
1617 | |
898 | default: |
1618 | default: |
899 | ERR ("malformed json string, neither array, object, number, string or atom"); |
1619 | ERR ("malformed JSON string, neither tag, array, object, number, string or atom"); |
900 | break; |
1620 | break; |
901 | } |
1621 | } |
902 | |
1622 | |
903 | fail: |
1623 | fail: |
904 | return 0; |
1624 | return 0; |
905 | } |
1625 | } |
906 | |
1626 | |
907 | static SV * |
1627 | static SV * |
908 | decode_json (SV *string, UV flags) |
1628 | decode_json (SV *string, JSON *json, char **offset_return) |
909 | { |
1629 | { |
|
|
1630 | dec_t dec; |
910 | SV *sv; |
1631 | SV *sv; |
911 | |
1632 | |
912 | if (flags & F_UTF8) |
1633 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1634 | * makes perl ignore the magic in subsequent accesses. |
|
|
1635 | * also make a copy of non-PV values, to get them into a clean |
|
|
1636 | * state (SvPV should do that, but it's buggy, see below). |
|
|
1637 | */ |
|
|
1638 | /*SvGETMAGIC (string);*/ |
|
|
1639 | if (SvMAGICAL (string) || !SvPOK (string)) |
|
|
1640 | string = sv_2mortal (newSVsv (string)); |
|
|
1641 | |
|
|
1642 | SvUPGRADE (string, SVt_PV); |
|
|
1643 | |
|
|
1644 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1645 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1646 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1647 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1648 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1649 | * and hope for the best. |
|
|
1650 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1651 | * assertion business is seriously broken, try yet another workaround |
|
|
1652 | * for the broken -DDEBUGGING. |
|
|
1653 | */ |
|
|
1654 | { |
|
|
1655 | #ifdef DEBUGGING |
|
|
1656 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1657 | #else |
|
|
1658 | STRLEN offset = SvCUR (string); |
|
|
1659 | #endif |
|
|
1660 | |
|
|
1661 | if (offset > json->max_size && json->max_size) |
|
|
1662 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1663 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1664 | } |
|
|
1665 | |
|
|
1666 | if (DECODE_WANTS_OCTETS (json)) |
913 | sv_utf8_downgrade (string, 0); |
1667 | sv_utf8_downgrade (string, 0); |
914 | else |
1668 | else |
915 | sv_utf8_upgrade (string); |
1669 | sv_utf8_upgrade (string); |
916 | |
1670 | |
917 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1671 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
918 | |
1672 | |
919 | dec_t dec; |
1673 | dec.json = *json; |
920 | dec.flags = flags; |
|
|
921 | dec.cur = SvPVX (string); |
1674 | dec.cur = SvPVX (string); |
922 | dec.end = SvEND (string); |
1675 | dec.end = SvEND (string); |
923 | dec.err = 0; |
1676 | dec.err = 0; |
|
|
1677 | dec.depth = 0; |
924 | |
1678 | |
|
|
1679 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1680 | dec.json.flags |= F_HOOK; |
|
|
1681 | |
|
|
1682 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1683 | |
|
|
1684 | decode_ws (&dec); |
925 | sv = decode_sv (&dec); |
1685 | sv = decode_sv (&dec); |
926 | |
1686 | |
|
|
1687 | if (offset_return) |
|
|
1688 | *offset_return = dec.cur; |
|
|
1689 | |
|
|
1690 | if (!(offset_return || !sv)) |
|
|
1691 | { |
|
|
1692 | // check for trailing garbage |
|
|
1693 | decode_ws (&dec); |
|
|
1694 | |
|
|
1695 | if (*dec.cur) |
|
|
1696 | { |
|
|
1697 | dec.err = "garbage after JSON object"; |
|
|
1698 | SvREFCNT_dec (sv); |
|
|
1699 | sv = 0; |
|
|
1700 | } |
|
|
1701 | } |
|
|
1702 | |
927 | if (!sv) |
1703 | if (!sv) |
928 | { |
1704 | { |
929 | IV offset = dec.flags & F_UTF8 |
|
|
930 | ? dec.cur - SvPVX (string) |
|
|
931 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
932 | SV *uni = sv_newmortal (); |
1705 | SV *uni = sv_newmortal (); |
933 | |
1706 | |
934 | // horrible hack to silence warning inside pv_uni_display |
1707 | // horrible hack to silence warning inside pv_uni_display |
935 | COP cop = *PL_curcop; |
1708 | COP cop = *PL_curcop; |
936 | cop.cop_warnings = pWARN_NONE; |
1709 | cop.cop_warnings = pWARN_NONE; |
… | |
… | |
938 | SAVEVPTR (PL_curcop); |
1711 | SAVEVPTR (PL_curcop); |
939 | PL_curcop = &cop; |
1712 | PL_curcop = &cop; |
940 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1713 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
941 | LEAVE; |
1714 | LEAVE; |
942 | |
1715 | |
943 | croak ("%s, at character offset %d (%s)", |
1716 | croak ("%s, at character offset %d (before \"%s\")", |
944 | dec.err, |
1717 | dec.err, |
945 | (int)offset, |
1718 | (int)ptr_to_index (string, dec.cur), |
946 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1719 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
947 | } |
1720 | } |
948 | |
1721 | |
949 | sv = sv_2mortal (sv); |
1722 | sv = sv_2mortal (sv); |
950 | |
1723 | |
951 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1724 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
952 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1725 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
953 | |
1726 | |
954 | return sv; |
1727 | return sv; |
|
|
1728 | } |
|
|
1729 | |
|
|
1730 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1731 | // incremental parser |
|
|
1732 | |
|
|
1733 | static void |
|
|
1734 | incr_parse (JSON *self) |
|
|
1735 | { |
|
|
1736 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1737 | |
|
|
1738 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1739 | // but this would make it slower, so... |
|
|
1740 | |
|
|
1741 | for (;;) |
|
|
1742 | { |
|
|
1743 | //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 |
|
|
1744 | switch (self->incr_mode) |
|
|
1745 | { |
|
|
1746 | // only used for initial whitespace skipping |
|
|
1747 | case INCR_M_WS: |
|
|
1748 | for (;;) |
|
|
1749 | { |
|
|
1750 | if (*p > 0x20) |
|
|
1751 | { |
|
|
1752 | if (*p == '#') |
|
|
1753 | { |
|
|
1754 | self->incr_mode = INCR_M_C0; |
|
|
1755 | goto incr_m_c; |
|
|
1756 | } |
|
|
1757 | else |
|
|
1758 | { |
|
|
1759 | self->incr_mode = INCR_M_JSON; |
|
|
1760 | goto incr_m_json; |
|
|
1761 | } |
|
|
1762 | } |
|
|
1763 | else if (!*p) |
|
|
1764 | goto interrupt; |
|
|
1765 | |
|
|
1766 | ++p; |
|
|
1767 | } |
|
|
1768 | |
|
|
1769 | // skip a single char inside a string (for \\-processing) |
|
|
1770 | case INCR_M_BS: |
|
|
1771 | if (!*p) |
|
|
1772 | goto interrupt; |
|
|
1773 | |
|
|
1774 | ++p; |
|
|
1775 | self->incr_mode = INCR_M_STR; |
|
|
1776 | goto incr_m_str; |
|
|
1777 | |
|
|
1778 | // inside #-style comments |
|
|
1779 | case INCR_M_C0: |
|
|
1780 | case INCR_M_C1: |
|
|
1781 | incr_m_c: |
|
|
1782 | for (;;) |
|
|
1783 | { |
|
|
1784 | if (*p == '\n') |
|
|
1785 | { |
|
|
1786 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1787 | break; |
|
|
1788 | } |
|
|
1789 | else if (!*p) |
|
|
1790 | goto interrupt; |
|
|
1791 | |
|
|
1792 | ++p; |
|
|
1793 | } |
|
|
1794 | |
|
|
1795 | break; |
|
|
1796 | |
|
|
1797 | // inside a string |
|
|
1798 | case INCR_M_STR: |
|
|
1799 | incr_m_str: |
|
|
1800 | for (;;) |
|
|
1801 | { |
|
|
1802 | if (*p == '"') |
|
|
1803 | { |
|
|
1804 | ++p; |
|
|
1805 | self->incr_mode = INCR_M_JSON; |
|
|
1806 | |
|
|
1807 | if (!self->incr_nest) |
|
|
1808 | goto interrupt; |
|
|
1809 | |
|
|
1810 | goto incr_m_json; |
|
|
1811 | } |
|
|
1812 | else if (*p == '\\') |
|
|
1813 | { |
|
|
1814 | ++p; // "virtually" consumes character after \ |
|
|
1815 | |
|
|
1816 | if (!*p) // if at end of string we have to switch modes |
|
|
1817 | { |
|
|
1818 | self->incr_mode = INCR_M_BS; |
|
|
1819 | goto interrupt; |
|
|
1820 | } |
|
|
1821 | } |
|
|
1822 | else if (!*p) |
|
|
1823 | goto interrupt; |
|
|
1824 | |
|
|
1825 | ++p; |
|
|
1826 | } |
|
|
1827 | |
|
|
1828 | // after initial ws, outside string |
|
|
1829 | case INCR_M_JSON: |
|
|
1830 | incr_m_json: |
|
|
1831 | for (;;) |
|
|
1832 | { |
|
|
1833 | switch (*p++) |
|
|
1834 | { |
|
|
1835 | case 0: |
|
|
1836 | --p; |
|
|
1837 | goto interrupt; |
|
|
1838 | |
|
|
1839 | case 0x09: |
|
|
1840 | case 0x0a: |
|
|
1841 | case 0x0d: |
|
|
1842 | case 0x20: |
|
|
1843 | if (!self->incr_nest) |
|
|
1844 | { |
|
|
1845 | --p; // do not eat the whitespace, let the next round do it |
|
|
1846 | goto interrupt; |
|
|
1847 | } |
|
|
1848 | break; |
|
|
1849 | |
|
|
1850 | case '"': |
|
|
1851 | self->incr_mode = INCR_M_STR; |
|
|
1852 | goto incr_m_str; |
|
|
1853 | |
|
|
1854 | case '[': |
|
|
1855 | case '{': |
|
|
1856 | case '(': |
|
|
1857 | if (++self->incr_nest > self->max_depth) |
|
|
1858 | croak (ERR_NESTING_EXCEEDED); |
|
|
1859 | break; |
|
|
1860 | |
|
|
1861 | case ']': |
|
|
1862 | case '}': |
|
|
1863 | if (--self->incr_nest <= 0) |
|
|
1864 | goto interrupt; |
|
|
1865 | break; |
|
|
1866 | |
|
|
1867 | case ')': |
|
|
1868 | --self->incr_nest; |
|
|
1869 | break; |
|
|
1870 | |
|
|
1871 | case '#': |
|
|
1872 | self->incr_mode = INCR_M_C1; |
|
|
1873 | goto incr_m_c; |
|
|
1874 | } |
|
|
1875 | } |
|
|
1876 | } |
|
|
1877 | |
|
|
1878 | modechange: |
|
|
1879 | ; |
|
|
1880 | } |
|
|
1881 | |
|
|
1882 | interrupt: |
|
|
1883 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1884 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1885 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
955 | } |
1886 | } |
956 | |
1887 | |
957 | ///////////////////////////////////////////////////////////////////////////// |
1888 | ///////////////////////////////////////////////////////////////////////////// |
958 | // XS interface functions |
1889 | // XS interface functions |
959 | |
1890 | |
… | |
… | |
961 | |
1892 | |
962 | BOOT: |
1893 | BOOT: |
963 | { |
1894 | { |
964 | int i; |
1895 | int i; |
965 | |
1896 | |
966 | memset (decode_hexdigit, 0xff, 256); |
|
|
967 | for (i = 10; i--; ) |
1897 | for (i = 0; i < 256; ++i) |
968 | decode_hexdigit ['0' + i] = i; |
1898 | decode_hexdigit [i] = |
|
|
1899 | i >= '0' && i <= '9' ? i - '0' |
|
|
1900 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
|
|
1901 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
|
|
1902 | : -1; |
969 | |
1903 | |
970 | for (i = 7; i--; ) |
|
|
971 | { |
|
|
972 | decode_hexdigit ['a' + i] = 10 + i; |
|
|
973 | decode_hexdigit ['A' + i] = 10 + i; |
|
|
974 | } |
|
|
975 | |
|
|
976 | json_stash = gv_stashpv ("JSON::XS", 1); |
1904 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1905 | types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1); |
|
|
1906 | |
|
|
1907 | types_true = get_bool ("Types::Serialiser::true"); |
|
|
1908 | types_false = get_bool ("Types::Serialiser::false"); |
|
|
1909 | |
|
|
1910 | sv_json = newSVpv ("JSON", 0); |
|
|
1911 | SvREADONLY_on (sv_json); |
|
|
1912 | |
|
|
1913 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
977 | } |
1914 | } |
978 | |
1915 | |
979 | PROTOTYPES: DISABLE |
1916 | PROTOTYPES: DISABLE |
980 | |
1917 | |
981 | SV *new (char *dummy) |
1918 | void CLONE (...) |
982 | CODE: |
1919 | CODE: |
983 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1920 | json_stash = 0; |
|
|
1921 | types_boolean_stash = 0; |
|
|
1922 | |
|
|
1923 | void new (char *klass) |
|
|
1924 | PPCODE: |
|
|
1925 | { |
|
|
1926 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1927 | SvPOK_only (pv); |
|
|
1928 | json_init ((JSON *)SvPVX (pv)); |
|
|
1929 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1930 | newRV_noinc (pv), |
|
|
1931 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1932 | ))); |
|
|
1933 | } |
|
|
1934 | |
|
|
1935 | void ascii (JSON *self, int enable = 1) |
|
|
1936 | ALIAS: |
|
|
1937 | ascii = F_ASCII |
|
|
1938 | latin1 = F_LATIN1 |
|
|
1939 | utf8 = F_UTF8 |
|
|
1940 | indent = F_INDENT |
|
|
1941 | canonical = F_CANONICAL |
|
|
1942 | space_before = F_SPACE_BEFORE |
|
|
1943 | space_after = F_SPACE_AFTER |
|
|
1944 | pretty = F_PRETTY |
|
|
1945 | allow_nonref = F_ALLOW_NONREF |
|
|
1946 | shrink = F_SHRINK |
|
|
1947 | allow_blessed = F_ALLOW_BLESSED |
|
|
1948 | convert_blessed = F_CONV_BLESSED |
|
|
1949 | relaxed = F_RELAXED |
|
|
1950 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
1951 | allow_tags = F_ALLOW_TAGS |
|
|
1952 | PPCODE: |
|
|
1953 | { |
|
|
1954 | if (enable) |
|
|
1955 | self->flags |= ix; |
|
|
1956 | else |
|
|
1957 | self->flags &= ~ix; |
|
|
1958 | |
|
|
1959 | XPUSHs (ST (0)); |
|
|
1960 | } |
|
|
1961 | |
|
|
1962 | void get_ascii (JSON *self) |
|
|
1963 | ALIAS: |
|
|
1964 | get_ascii = F_ASCII |
|
|
1965 | get_latin1 = F_LATIN1 |
|
|
1966 | get_utf8 = F_UTF8 |
|
|
1967 | get_indent = F_INDENT |
|
|
1968 | get_canonical = F_CANONICAL |
|
|
1969 | get_space_before = F_SPACE_BEFORE |
|
|
1970 | get_space_after = F_SPACE_AFTER |
|
|
1971 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1972 | get_shrink = F_SHRINK |
|
|
1973 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1974 | get_convert_blessed = F_CONV_BLESSED |
|
|
1975 | get_relaxed = F_RELAXED |
|
|
1976 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
1977 | get_allow_tags = F_ALLOW_TAGS |
|
|
1978 | PPCODE: |
|
|
1979 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1980 | |
|
|
1981 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
|
|
1982 | PPCODE: |
|
|
1983 | self->max_depth = max_depth; |
|
|
1984 | XPUSHs (ST (0)); |
|
|
1985 | |
|
|
1986 | U32 get_max_depth (JSON *self) |
|
|
1987 | CODE: |
|
|
1988 | RETVAL = self->max_depth; |
984 | OUTPUT: |
1989 | OUTPUT: |
985 | RETVAL |
1990 | RETVAL |
986 | |
1991 | |
987 | SV *ascii (SV *self, int enable = 1) |
1992 | void max_size (JSON *self, U32 max_size = 0) |
988 | ALIAS: |
1993 | PPCODE: |
989 | ascii = F_ASCII |
1994 | self->max_size = max_size; |
990 | utf8 = F_UTF8 |
1995 | XPUSHs (ST (0)); |
991 | indent = F_INDENT |
1996 | |
992 | canonical = F_CANONICAL |
1997 | int get_max_size (JSON *self) |
993 | space_before = F_SPACE_BEFORE |
|
|
994 | space_after = F_SPACE_AFTER |
|
|
995 | pretty = F_PRETTY |
|
|
996 | allow_nonref = F_ALLOW_NONREF |
|
|
997 | shrink = F_SHRINK |
|
|
998 | CODE: |
1998 | CODE: |
999 | { |
1999 | RETVAL = self->max_size; |
1000 | UV *uv = SvJSON (self); |
|
|
1001 | if (enable) |
|
|
1002 | *uv |= ix; |
|
|
1003 | else |
|
|
1004 | *uv &= ~ix; |
|
|
1005 | |
|
|
1006 | RETVAL = newSVsv (self); |
|
|
1007 | } |
|
|
1008 | OUTPUT: |
2000 | OUTPUT: |
1009 | RETVAL |
2001 | RETVAL |
1010 | |
2002 | |
1011 | void encode (SV *self, SV *scalar) |
2003 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1012 | PPCODE: |
2004 | PPCODE: |
1013 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
2005 | { |
|
|
2006 | SvREFCNT_dec (self->cb_object); |
|
|
2007 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
1014 | |
2008 | |
1015 | void decode (SV *self, SV *jsonstr) |
2009 | XPUSHs (ST (0)); |
|
|
2010 | } |
|
|
2011 | |
|
|
2012 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
1016 | PPCODE: |
2013 | PPCODE: |
1017 | XPUSHs (decode_json (jsonstr, *SvJSON (self))); |
2014 | { |
|
|
2015 | if (!self->cb_sk_object) |
|
|
2016 | self->cb_sk_object = newHV (); |
|
|
2017 | |
|
|
2018 | if (SvOK (cb)) |
|
|
2019 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
2020 | else |
|
|
2021 | { |
|
|
2022 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
2023 | |
|
|
2024 | if (!HvKEYS (self->cb_sk_object)) |
|
|
2025 | { |
|
|
2026 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2027 | self->cb_sk_object = 0; |
|
|
2028 | } |
|
|
2029 | } |
|
|
2030 | |
|
|
2031 | XPUSHs (ST (0)); |
|
|
2032 | } |
|
|
2033 | |
|
|
2034 | void encode (JSON *self, SV *scalar) |
|
|
2035 | PPCODE: |
|
|
2036 | PUTBACK; scalar = encode_json (scalar, self); SPAGAIN; |
|
|
2037 | XPUSHs (scalar); |
|
|
2038 | |
|
|
2039 | void decode (JSON *self, SV *jsonstr) |
|
|
2040 | PPCODE: |
|
|
2041 | PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN; |
|
|
2042 | XPUSHs (jsonstr); |
|
|
2043 | |
|
|
2044 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
2045 | PPCODE: |
|
|
2046 | { |
|
|
2047 | SV *sv; |
|
|
2048 | char *offset; |
|
|
2049 | PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN; |
|
|
2050 | EXTEND (SP, 2); |
|
|
2051 | PUSHs (sv); |
|
|
2052 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
2053 | } |
|
|
2054 | |
|
|
2055 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
2056 | PPCODE: |
|
|
2057 | { |
|
|
2058 | if (!self->incr_text) |
|
|
2059 | self->incr_text = newSVpvn ("", 0); |
|
|
2060 | |
|
|
2061 | /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
|
|
2062 | if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
|
|
2063 | if (DECODE_WANTS_OCTETS (self)) |
|
|
2064 | { |
|
|
2065 | if (self->incr_pos) |
|
|
2066 | self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
|
|
2067 | (U8 *)SvPVX (self->incr_text) + self->incr_pos); |
|
|
2068 | |
|
|
2069 | sv_utf8_downgrade (self->incr_text, 0); |
|
|
2070 | } |
|
|
2071 | else |
|
|
2072 | { |
|
|
2073 | sv_utf8_upgrade (self->incr_text); |
|
|
2074 | |
|
|
2075 | if (self->incr_pos) |
|
|
2076 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
2077 | - (U8 *)SvPVX (self->incr_text); |
|
|
2078 | } |
|
|
2079 | |
|
|
2080 | // append data, if any |
|
|
2081 | if (jsonstr) |
|
|
2082 | { |
|
|
2083 | /* make sure both strings have same encoding */ |
|
|
2084 | if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
|
|
2085 | if (SvUTF8 (jsonstr)) |
|
|
2086 | sv_utf8_downgrade (jsonstr, 0); |
|
|
2087 | else |
|
|
2088 | sv_utf8_upgrade (jsonstr); |
|
|
2089 | |
|
|
2090 | /* and then just blindly append */ |
|
|
2091 | { |
|
|
2092 | STRLEN len; |
|
|
2093 | const char *str = SvPV (jsonstr, len); |
|
|
2094 | STRLEN cur = SvCUR (self->incr_text); |
|
|
2095 | |
|
|
2096 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
2097 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
2098 | |
|
|
2099 | Move (str, SvEND (self->incr_text), len, char); |
|
|
2100 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
2101 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
2102 | } |
|
|
2103 | } |
|
|
2104 | |
|
|
2105 | if (GIMME_V != G_VOID) |
|
|
2106 | do |
|
|
2107 | { |
|
|
2108 | SV *sv; |
|
|
2109 | char *offset; |
|
|
2110 | |
|
|
2111 | if (!INCR_DONE (self)) |
|
|
2112 | { |
|
|
2113 | incr_parse (self); |
|
|
2114 | |
|
|
2115 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
2116 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
2117 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
2118 | |
|
|
2119 | if (!INCR_DONE (self)) |
|
|
2120 | { |
|
|
2121 | // as an optimisation, do not accumulate white space in the incr buffer |
|
|
2122 | if (self->incr_mode == INCR_M_WS && self->incr_pos) |
|
|
2123 | { |
|
|
2124 | self->incr_pos = 0; |
|
|
2125 | SvCUR_set (self->incr_text, 0); |
|
|
2126 | } |
|
|
2127 | |
|
|
2128 | break; |
|
|
2129 | } |
|
|
2130 | } |
|
|
2131 | |
|
|
2132 | PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN; |
|
|
2133 | XPUSHs (sv); |
|
|
2134 | |
|
|
2135 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
2136 | self->incr_nest = 0; |
|
|
2137 | self->incr_mode = 0; |
|
|
2138 | |
|
|
2139 | sv_chop (self->incr_text, offset); |
|
|
2140 | } |
|
|
2141 | while (GIMME_V == G_ARRAY); |
|
|
2142 | } |
|
|
2143 | |
|
|
2144 | SV *incr_text (JSON *self) |
|
|
2145 | ATTRS: lvalue |
|
|
2146 | CODE: |
|
|
2147 | { |
|
|
2148 | if (self->incr_pos) |
|
|
2149 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
2150 | |
|
|
2151 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
2152 | } |
|
|
2153 | OUTPUT: |
|
|
2154 | RETVAL |
|
|
2155 | |
|
|
2156 | void incr_skip (JSON *self) |
|
|
2157 | CODE: |
|
|
2158 | { |
|
|
2159 | if (self->incr_pos) |
|
|
2160 | { |
|
|
2161 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
2162 | self->incr_pos = 0; |
|
|
2163 | self->incr_nest = 0; |
|
|
2164 | self->incr_mode = 0; |
|
|
2165 | } |
|
|
2166 | } |
|
|
2167 | |
|
|
2168 | void incr_reset (JSON *self) |
|
|
2169 | CODE: |
|
|
2170 | { |
|
|
2171 | SvREFCNT_dec (self->incr_text); |
|
|
2172 | self->incr_text = 0; |
|
|
2173 | self->incr_pos = 0; |
|
|
2174 | self->incr_nest = 0; |
|
|
2175 | self->incr_mode = 0; |
|
|
2176 | } |
|
|
2177 | |
|
|
2178 | void DESTROY (JSON *self) |
|
|
2179 | CODE: |
|
|
2180 | SvREFCNT_dec (self->cb_sk_object); |
|
|
2181 | SvREFCNT_dec (self->cb_object); |
|
|
2182 | SvREFCNT_dec (self->incr_text); |
1018 | |
2183 | |
1019 | PROTOTYPES: ENABLE |
2184 | PROTOTYPES: ENABLE |
1020 | |
2185 | |
1021 | void to_json (SV *scalar) |
2186 | void encode_json (SV *scalar) |
1022 | PPCODE: |
2187 | PPCODE: |
1023 | XPUSHs (encode_json (scalar, F_UTF8)); |
2188 | { |
|
|
2189 | JSON json; |
|
|
2190 | json_init (&json); |
|
|
2191 | json.flags |= F_UTF8; |
|
|
2192 | PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN; |
|
|
2193 | XPUSHs (scalar); |
|
|
2194 | } |
1024 | |
2195 | |
1025 | void from_json (SV *jsonstr) |
2196 | void decode_json (SV *jsonstr) |
1026 | PPCODE: |
2197 | PPCODE: |
1027 | XPUSHs (decode_json (jsonstr, F_UTF8)); |
2198 | { |
|
|
2199 | JSON json; |
|
|
2200 | json_init (&json); |
|
|
2201 | json.flags |= F_UTF8; |
|
|
2202 | PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN; |
|
|
2203 | XPUSHs (jsonstr); |
|
|
2204 | } |
1028 | |
2205 | |