1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
8 | #include "stdio.h" |
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
|
|
10 | #include <float.h> |
9 | |
11 | |
10 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
11 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
12 | #endif |
14 | #endif |
13 | |
15 | |
14 | // some old perls do not have this, try to make it work, no |
16 | // some old perls do not have this, try to make it work, no |
15 | // guarentees, though. if it breaks, you get to keep the pieces. |
17 | // guarentees, though. if it breaks, you get to keep the pieces. |
16 | #ifndef UTF8_MAXBYTES |
18 | #ifndef UTF8_MAXBYTES |
17 | # define UTF8_MAXBYTES 13 |
19 | # define UTF8_MAXBYTES 13 |
18 | #endif |
20 | #endif |
|
|
21 | |
|
|
22 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) |
19 | |
23 | |
20 | #define F_ASCII 0x00000001UL |
24 | #define F_ASCII 0x00000001UL |
21 | #define F_LATIN1 0x00000002UL |
25 | #define F_LATIN1 0x00000002UL |
22 | #define F_UTF8 0x00000004UL |
26 | #define F_UTF8 0x00000004UL |
23 | #define F_INDENT 0x00000008UL |
27 | #define F_INDENT 0x00000008UL |
… | |
… | |
25 | #define F_SPACE_BEFORE 0x00000020UL |
29 | #define F_SPACE_BEFORE 0x00000020UL |
26 | #define F_SPACE_AFTER 0x00000040UL |
30 | #define F_SPACE_AFTER 0x00000040UL |
27 | #define F_ALLOW_NONREF 0x00000100UL |
31 | #define F_ALLOW_NONREF 0x00000100UL |
28 | #define F_SHRINK 0x00000200UL |
32 | #define F_SHRINK 0x00000200UL |
29 | #define F_ALLOW_BLESSED 0x00000400UL |
33 | #define F_ALLOW_BLESSED 0x00000400UL |
30 | #define F_CONV_BLESSED 0x00000800UL // NYI |
34 | #define F_CONV_BLESSED 0x00000800UL |
31 | #define F_MAXDEPTH 0xf8000000UL |
35 | #define F_RELAXED 0x00001000UL |
32 | #define S_MAXDEPTH 27 |
36 | #define F_ALLOW_UNKNOWN 0x00002000UL |
33 | |
37 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
34 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
|
|
35 | |
38 | |
36 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
39 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
37 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
38 | |
40 | |
39 | #define INIT_SIZE 32 // initial scalar size to be allocated |
41 | #define INIT_SIZE 32 // initial scalar size to be allocated |
40 | #define INDENT_STEP 3 // spaces per indentation level |
42 | #define INDENT_STEP 3 // spaces per indentation level |
41 | |
43 | |
42 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
44 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
43 | |
45 | |
44 | #define SB do { |
46 | #define SB do { |
45 | #define SE } while (0) |
47 | #define SE } while (0) |
46 | |
48 | |
47 | #if __GNUC__ >= 3 |
49 | #if __GNUC__ >= 3 |
48 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
50 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
49 | # define inline inline |
51 | # define INLINE static inline |
50 | #else |
52 | #else |
51 | # define expect(expr,value) (expr) |
53 | # define expect(expr,value) (expr) |
52 | # define inline static |
54 | # define INLINE static |
53 | #endif |
55 | #endif |
54 | |
56 | |
55 | #define expect_false(expr) expect ((expr) != 0, 0) |
57 | #define expect_false(expr) expect ((expr) != 0, 0) |
56 | #define expect_true(expr) expect ((expr) != 0, 1) |
58 | #define expect_true(expr) expect ((expr) != 0, 1) |
57 | |
59 | |
|
|
60 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
61 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
62 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
63 | |
|
|
64 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
65 | |
|
|
66 | #ifdef USE_ITHREADS |
|
|
67 | # define JSON_SLOW 1 |
|
|
68 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
69 | #else |
|
|
70 | # define JSON_SLOW 0 |
|
|
71 | # define JSON_STASH json_stash |
|
|
72 | #endif |
|
|
73 | |
58 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
74 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
59 | static SV *json_true, *json_false; |
75 | static SV *json_true, *json_false; |
60 | |
76 | |
|
|
77 | enum { |
|
|
78 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
79 | INCR_M_STR, // inside string |
|
|
80 | INCR_M_BS, // inside backslash |
|
|
81 | INCR_M_JSON // outside anything, count nesting |
|
|
82 | }; |
|
|
83 | |
|
|
84 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
85 | |
|
|
86 | typedef struct { |
|
|
87 | U32 flags; |
|
|
88 | U32 max_depth; |
|
|
89 | STRLEN max_size; |
|
|
90 | |
|
|
91 | SV *cb_object; |
|
|
92 | HV *cb_sk_object; |
|
|
93 | |
|
|
94 | // for the incremental parser |
|
|
95 | SV *incr_text; // the source text so far |
|
|
96 | STRLEN incr_pos; // the current offset into the text |
|
|
97 | int incr_nest; // {[]}-nesting level |
|
|
98 | unsigned char incr_mode; |
|
|
99 | } JSON; |
|
|
100 | |
|
|
101 | INLINE void |
|
|
102 | json_init (JSON *json) |
|
|
103 | { |
|
|
104 | Zero (json, 1, JSON); |
|
|
105 | json->max_depth = 512; |
|
|
106 | } |
|
|
107 | |
61 | ///////////////////////////////////////////////////////////////////////////// |
108 | ///////////////////////////////////////////////////////////////////////////// |
62 | // utility functions |
109 | // utility functions |
63 | |
110 | |
64 | static UV * |
111 | INLINE SV * |
65 | SvJSON (SV *sv) |
112 | get_bool (const char *name) |
66 | { |
113 | { |
67 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
114 | SV *sv = get_sv (name, 1); |
68 | croak ("object is not of type JSON::XS"); |
|
|
69 | |
115 | |
70 | return &SvUVX (SvRV (sv)); |
116 | SvREADONLY_on (sv); |
71 | } |
117 | SvREADONLY_on (SvRV (sv)); |
72 | |
118 | |
73 | static void |
119 | return sv; |
|
|
120 | } |
|
|
121 | |
|
|
122 | INLINE void |
74 | shrink (SV *sv) |
123 | shrink (SV *sv) |
75 | { |
124 | { |
76 | sv_utf8_downgrade (sv, 1); |
125 | sv_utf8_downgrade (sv, 1); |
|
|
126 | |
77 | if (SvLEN (sv) > SvCUR (sv) + 1) |
127 | if (SvLEN (sv) > SvCUR (sv) + 1) |
78 | { |
128 | { |
79 | #ifdef SvPV_shrink_to_cur |
129 | #ifdef SvPV_shrink_to_cur |
80 | SvPV_shrink_to_cur (sv); |
130 | SvPV_shrink_to_cur (sv); |
81 | #elif defined (SvPV_renew) |
131 | #elif defined (SvPV_renew) |
… | |
… | |
87 | // decode an utf-8 character and return it, or (UV)-1 in |
137 | // decode an utf-8 character and return it, or (UV)-1 in |
88 | // case of an error. |
138 | // case of an error. |
89 | // we special-case "safe" characters from U+80 .. U+7FF, |
139 | // we special-case "safe" characters from U+80 .. U+7FF, |
90 | // but use the very good perl function to parse anything else. |
140 | // but use the very good perl function to parse anything else. |
91 | // note that we never call this function for a ascii codepoints |
141 | // note that we never call this function for a ascii codepoints |
92 | inline UV |
142 | INLINE UV |
93 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
143 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
94 | { |
144 | { |
95 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
145 | if (expect_true (len >= 2 |
96 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
146 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
97 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
147 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
98 | { |
148 | { |
99 | *clen = 2; |
149 | *clen = 2; |
100 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
150 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
101 | } |
151 | } |
102 | else |
152 | else |
103 | { |
153 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
104 | *clen = (STRLEN)-1; |
154 | } |
105 | return (UV)-1; |
155 | |
106 | } |
156 | // likewise for encoding, also never called for ascii codepoints |
|
|
157 | // this function takes advantage of this fact, although current gccs |
|
|
158 | // seem to optimise the check for >= 0x80 away anyways |
|
|
159 | INLINE unsigned char * |
|
|
160 | encode_utf8 (unsigned char *s, UV ch) |
|
|
161 | { |
|
|
162 | if (expect_false (ch < 0x000080)) |
|
|
163 | *s++ = ch; |
|
|
164 | else if (expect_true (ch < 0x000800)) |
|
|
165 | *s++ = 0xc0 | ( ch >> 6), |
|
|
166 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
167 | else if ( ch < 0x010000) |
|
|
168 | *s++ = 0xe0 | ( ch >> 12), |
|
|
169 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
170 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
171 | else if ( ch < 0x110000) |
|
|
172 | *s++ = 0xf0 | ( ch >> 18), |
|
|
173 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
174 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
175 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
176 | |
|
|
177 | return s; |
107 | } |
178 | } |
108 | |
179 | |
109 | ///////////////////////////////////////////////////////////////////////////// |
180 | ///////////////////////////////////////////////////////////////////////////// |
110 | // encoder |
181 | // encoder |
111 | |
182 | |
… | |
… | |
113 | typedef struct |
184 | typedef struct |
114 | { |
185 | { |
115 | char *cur; // SvPVX (sv) + current output position |
186 | char *cur; // SvPVX (sv) + current output position |
116 | char *end; // SvEND (sv) |
187 | char *end; // SvEND (sv) |
117 | SV *sv; // result scalar |
188 | SV *sv; // result scalar |
118 | U32 flags; // F_* |
189 | JSON json; |
119 | U32 indent; // indentation level |
190 | U32 indent; // indentation level |
120 | U32 maxdepth; // max. indentation/recursion level |
191 | UV limit; // escape character values >= this value when encoding |
121 | } enc_t; |
192 | } enc_t; |
122 | |
193 | |
123 | inline void |
194 | INLINE void |
124 | need (enc_t *enc, STRLEN len) |
195 | need (enc_t *enc, STRLEN len) |
125 | { |
196 | { |
126 | if (expect_false (enc->cur + len >= enc->end)) |
197 | if (expect_false (enc->cur + len >= enc->end)) |
127 | { |
198 | { |
128 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
199 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
… | |
… | |
130 | enc->cur = SvPVX (enc->sv) + cur; |
201 | enc->cur = SvPVX (enc->sv) + cur; |
131 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
202 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
132 | } |
203 | } |
133 | } |
204 | } |
134 | |
205 | |
135 | inline void |
206 | INLINE void |
136 | encode_ch (enc_t *enc, char ch) |
207 | encode_ch (enc_t *enc, char ch) |
137 | { |
208 | { |
138 | need (enc, 1); |
209 | need (enc, 1); |
139 | *enc->cur++ = ch; |
210 | *enc->cur++ = ch; |
140 | } |
211 | } |
… | |
… | |
194 | { |
265 | { |
195 | uch = ch; |
266 | uch = ch; |
196 | clen = 1; |
267 | clen = 1; |
197 | } |
268 | } |
198 | |
269 | |
199 | if (uch > 0x10FFFFUL) |
270 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
200 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
201 | |
|
|
202 | if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF)) |
|
|
203 | { |
271 | { |
204 | if (uch > 0xFFFFUL) |
272 | if (uch >= 0x10000UL) |
205 | { |
273 | { |
|
|
274 | if (uch >= 0x110000UL) |
|
|
275 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
276 | |
206 | need (enc, len += 11); |
277 | need (enc, len += 11); |
207 | sprintf (enc->cur, "\\u%04x\\u%04x", |
278 | sprintf (enc->cur, "\\u%04x\\u%04x", |
208 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
279 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
209 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
280 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
210 | enc->cur += 12; |
281 | enc->cur += 12; |
… | |
… | |
221 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
292 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
222 | } |
293 | } |
223 | |
294 | |
224 | str += clen; |
295 | str += clen; |
225 | } |
296 | } |
226 | else if (enc->flags & F_LATIN1) |
297 | else if (enc->json.flags & F_LATIN1) |
227 | { |
298 | { |
228 | *enc->cur++ = uch; |
299 | *enc->cur++ = uch; |
229 | str += clen; |
300 | str += clen; |
230 | } |
301 | } |
231 | else if (is_utf8) |
302 | else if (is_utf8) |
… | |
… | |
238 | while (--clen); |
309 | while (--clen); |
239 | } |
310 | } |
240 | else |
311 | else |
241 | { |
312 | { |
242 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
313 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
243 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
314 | enc->cur = encode_utf8 (enc->cur, uch); |
244 | ++str; |
315 | ++str; |
245 | } |
316 | } |
246 | } |
317 | } |
247 | } |
318 | } |
248 | } |
319 | } |
249 | |
320 | |
250 | --len; |
321 | --len; |
251 | } |
322 | } |
252 | } |
323 | } |
253 | |
324 | |
254 | inline void |
325 | INLINE void |
255 | encode_indent (enc_t *enc) |
326 | encode_indent (enc_t *enc) |
256 | { |
327 | { |
257 | if (enc->flags & F_INDENT) |
328 | if (enc->json.flags & F_INDENT) |
258 | { |
329 | { |
259 | int spaces = enc->indent * INDENT_STEP; |
330 | int spaces = enc->indent * INDENT_STEP; |
260 | |
331 | |
261 | need (enc, spaces); |
332 | need (enc, spaces); |
262 | memset (enc->cur, ' ', spaces); |
333 | memset (enc->cur, ' ', spaces); |
263 | enc->cur += spaces; |
334 | enc->cur += spaces; |
264 | } |
335 | } |
265 | } |
336 | } |
266 | |
337 | |
267 | inline void |
338 | INLINE void |
268 | encode_space (enc_t *enc) |
339 | encode_space (enc_t *enc) |
269 | { |
340 | { |
270 | need (enc, 1); |
341 | need (enc, 1); |
271 | encode_ch (enc, ' '); |
342 | encode_ch (enc, ' '); |
272 | } |
343 | } |
273 | |
344 | |
274 | inline void |
345 | INLINE void |
275 | encode_nl (enc_t *enc) |
346 | encode_nl (enc_t *enc) |
276 | { |
347 | { |
277 | if (enc->flags & F_INDENT) |
348 | if (enc->json.flags & F_INDENT) |
278 | { |
349 | { |
279 | need (enc, 1); |
350 | need (enc, 1); |
280 | encode_ch (enc, '\n'); |
351 | encode_ch (enc, '\n'); |
281 | } |
352 | } |
282 | } |
353 | } |
283 | |
354 | |
284 | inline void |
355 | INLINE void |
285 | encode_comma (enc_t *enc) |
356 | encode_comma (enc_t *enc) |
286 | { |
357 | { |
287 | encode_ch (enc, ','); |
358 | encode_ch (enc, ','); |
288 | |
359 | |
289 | if (enc->flags & F_INDENT) |
360 | if (enc->json.flags & F_INDENT) |
290 | encode_nl (enc); |
361 | encode_nl (enc); |
291 | else if (enc->flags & F_SPACE_AFTER) |
362 | else if (enc->json.flags & F_SPACE_AFTER) |
292 | encode_space (enc); |
363 | encode_space (enc); |
293 | } |
364 | } |
294 | |
365 | |
295 | static void encode_sv (enc_t *enc, SV *sv); |
366 | static void encode_sv (enc_t *enc, SV *sv); |
296 | |
367 | |
297 | static void |
368 | static void |
298 | encode_av (enc_t *enc, AV *av) |
369 | encode_av (enc_t *enc, AV *av) |
299 | { |
370 | { |
300 | int i, len = av_len (av); |
371 | int i, len = av_len (av); |
301 | |
372 | |
302 | if (enc->indent >= enc->maxdepth) |
373 | if (enc->indent >= enc->json.max_depth) |
303 | croak ("data structure too deep (hit recursion limit)"); |
374 | croak (ERR_NESTING_EXCEEDED); |
304 | |
375 | |
305 | encode_ch (enc, '['); encode_nl (enc); |
376 | encode_ch (enc, '['); |
306 | ++enc->indent; |
377 | |
|
|
378 | if (len >= 0) |
|
|
379 | { |
|
|
380 | encode_nl (enc); ++enc->indent; |
307 | |
381 | |
308 | for (i = 0; i <= len; ++i) |
382 | for (i = 0; i <= len; ++i) |
309 | { |
383 | { |
|
|
384 | SV **svp = av_fetch (av, i, 0); |
|
|
385 | |
310 | encode_indent (enc); |
386 | encode_indent (enc); |
311 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
312 | |
387 | |
|
|
388 | if (svp) |
|
|
389 | encode_sv (enc, *svp); |
|
|
390 | else |
|
|
391 | encode_str (enc, "null", 4, 0); |
|
|
392 | |
313 | if (i < len) |
393 | if (i < len) |
314 | encode_comma (enc); |
394 | encode_comma (enc); |
315 | } |
395 | } |
316 | |
396 | |
|
|
397 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
398 | } |
|
|
399 | |
317 | encode_nl (enc); |
400 | encode_ch (enc, ']'); |
318 | |
|
|
319 | --enc->indent; |
|
|
320 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
321 | } |
401 | } |
322 | |
402 | |
323 | static void |
403 | static void |
324 | encode_he (enc_t *enc, HE *he) |
404 | encode_hk (enc_t *enc, HE *he) |
325 | { |
405 | { |
326 | encode_ch (enc, '"'); |
406 | encode_ch (enc, '"'); |
327 | |
407 | |
328 | if (HeKLEN (he) == HEf_SVKEY) |
408 | if (HeKLEN (he) == HEf_SVKEY) |
329 | { |
409 | { |
… | |
… | |
339 | else |
419 | else |
340 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
420 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
341 | |
421 | |
342 | encode_ch (enc, '"'); |
422 | encode_ch (enc, '"'); |
343 | |
423 | |
344 | if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
424 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
345 | encode_ch (enc, ':'); |
425 | encode_ch (enc, ':'); |
346 | if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
426 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
347 | encode_sv (enc, HeVAL (he)); |
|
|
348 | } |
427 | } |
349 | |
428 | |
350 | // compare hash entries, used when all keys are bytestrings |
429 | // compare hash entries, used when all keys are bytestrings |
351 | static int |
430 | static int |
352 | he_cmp_fast (const void *a_, const void *b_) |
431 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
357 | HE *b = *(HE **)b_; |
436 | HE *b = *(HE **)b_; |
358 | |
437 | |
359 | STRLEN la = HeKLEN (a); |
438 | STRLEN la = HeKLEN (a); |
360 | STRLEN lb = HeKLEN (b); |
439 | STRLEN lb = HeKLEN (b); |
361 | |
440 | |
362 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
441 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
363 | cmp = la - lb; |
442 | cmp = lb - la; |
364 | |
443 | |
365 | return cmp; |
444 | return cmp; |
366 | } |
445 | } |
367 | |
446 | |
368 | // compare hash entries, used when some keys are sv's or utf-x |
447 | // compare hash entries, used when some keys are sv's or utf-x |
369 | static int |
448 | static int |
370 | he_cmp_slow (const void *a, const void *b) |
449 | he_cmp_slow (const void *a, const void *b) |
371 | { |
450 | { |
372 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
451 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
373 | } |
452 | } |
374 | |
453 | |
375 | static void |
454 | static void |
376 | encode_hv (enc_t *enc, HV *hv) |
455 | encode_hv (enc_t *enc, HV *hv) |
377 | { |
456 | { |
378 | int count, i; |
457 | HE *he; |
379 | |
458 | |
380 | if (enc->indent >= enc->maxdepth) |
459 | if (enc->indent >= enc->json.max_depth) |
381 | croak ("data structure too deep (hit recursion limit)"); |
460 | croak (ERR_NESTING_EXCEEDED); |
382 | |
461 | |
383 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
462 | encode_ch (enc, '{'); |
384 | |
463 | |
385 | if ((count = hv_iterinit (hv))) |
|
|
386 | { |
|
|
387 | // for canonical output we have to sort by keys first |
464 | // for canonical output we have to sort by keys first |
388 | // actually, this is mostly due to the stupid so-called |
465 | // actually, this is mostly due to the stupid so-called |
389 | // security workaround added somewhere in 5.8.x. |
466 | // security workaround added somewhere in 5.8.x. |
390 | // that randomises hash orderings |
467 | // that randomises hash orderings |
391 | if (enc->flags & F_CANONICAL) |
468 | if (enc->json.flags & F_CANONICAL) |
|
|
469 | { |
|
|
470 | int count = hv_iterinit (hv); |
|
|
471 | |
|
|
472 | if (SvMAGICAL (hv)) |
392 | { |
473 | { |
|
|
474 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
475 | // but I don't care for the speed of this special case. |
|
|
476 | // note also that we will run into undefined behaviour when the two iterations |
|
|
477 | // do not result in the same count, something I might care for in some later release. |
|
|
478 | |
|
|
479 | count = 0; |
|
|
480 | while (hv_iternext (hv)) |
|
|
481 | ++count; |
|
|
482 | |
|
|
483 | hv_iterinit (hv); |
|
|
484 | } |
|
|
485 | |
|
|
486 | if (count) |
|
|
487 | { |
393 | int fast = 1; |
488 | int i, fast = 1; |
394 | HE *he; |
|
|
395 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
489 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
396 | HE **hes = _alloca (count * sizeof (HE)); |
490 | HE **hes = _alloca (count * sizeof (HE)); |
397 | #else |
491 | #else |
398 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
492 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
399 | #endif |
493 | #endif |
… | |
… | |
426 | |
520 | |
427 | FREETMPS; |
521 | FREETMPS; |
428 | LEAVE; |
522 | LEAVE; |
429 | } |
523 | } |
430 | |
524 | |
431 | for (i = 0; i < count; ++i) |
525 | encode_nl (enc); ++enc->indent; |
|
|
526 | |
|
|
527 | while (count--) |
432 | { |
528 | { |
433 | encode_indent (enc); |
529 | encode_indent (enc); |
|
|
530 | he = hes [count]; |
434 | encode_he (enc, hes [i]); |
531 | encode_hk (enc, he); |
|
|
532 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
435 | |
533 | |
436 | if (i < count - 1) |
534 | if (count) |
437 | encode_comma (enc); |
535 | encode_comma (enc); |
438 | } |
536 | } |
439 | |
537 | |
440 | encode_nl (enc); |
538 | encode_nl (enc); --enc->indent; encode_indent (enc); |
441 | } |
539 | } |
|
|
540 | } |
442 | else |
541 | else |
|
|
542 | { |
|
|
543 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
544 | if ((he = hv_iternext (hv))) |
443 | { |
545 | { |
444 | HE *he = hv_iternext (hv); |
546 | encode_nl (enc); ++enc->indent; |
445 | |
547 | |
446 | for (;;) |
548 | for (;;) |
447 | { |
549 | { |
448 | encode_indent (enc); |
550 | encode_indent (enc); |
449 | encode_he (enc, he); |
551 | encode_hk (enc, he); |
|
|
552 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
450 | |
553 | |
451 | if (!(he = hv_iternext (hv))) |
554 | if (!(he = hv_iternext (hv))) |
452 | break; |
555 | break; |
453 | |
556 | |
454 | encode_comma (enc); |
557 | encode_comma (enc); |
455 | } |
558 | } |
456 | |
559 | |
457 | encode_nl (enc); |
560 | encode_nl (enc); --enc->indent; encode_indent (enc); |
458 | } |
561 | } |
459 | } |
562 | } |
460 | |
563 | |
461 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
564 | encode_ch (enc, '}'); |
462 | } |
565 | } |
463 | |
566 | |
464 | // encode objects, arrays and special \0=false and \1=true values. |
567 | // encode objects, arrays and special \0=false and \1=true values. |
465 | static void |
568 | static void |
466 | encode_rv (enc_t *enc, SV *sv) |
569 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
470 | SvGETMAGIC (sv); |
573 | SvGETMAGIC (sv); |
471 | svt = SvTYPE (sv); |
574 | svt = SvTYPE (sv); |
472 | |
575 | |
473 | if (expect_false (SvOBJECT (sv))) |
576 | if (expect_false (SvOBJECT (sv))) |
474 | { |
577 | { |
|
|
578 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
579 | ? json_boolean_stash |
|
|
580 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
581 | |
475 | if (SvSTASH (sv) == json_boolean_stash) |
582 | if (SvSTASH (sv) == stash) |
476 | { |
583 | { |
477 | if (SvIV (sv) == 0) |
584 | if (SvIV (sv)) |
|
|
585 | encode_str (enc, "true", 4, 0); |
|
|
586 | else |
478 | encode_str (enc, "false", 5, 0); |
587 | encode_str (enc, "false", 5, 0); |
479 | else |
|
|
480 | encode_str (enc, "true", 4, 0); |
|
|
481 | } |
588 | } |
482 | else |
589 | else |
483 | { |
590 | { |
484 | #if 0 |
591 | #if 0 |
485 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
592 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
486 | { |
593 | { |
487 | // not yet |
594 | // not yet |
488 | } |
595 | } |
489 | #endif |
596 | #endif |
490 | if (enc->flags & F_CONV_BLESSED) |
597 | if (enc->json.flags & F_CONV_BLESSED) |
491 | { |
598 | { |
492 | // we re-bless the reference to get overload and other niceties right |
599 | // we re-bless the reference to get overload and other niceties right |
493 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1); |
600 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
494 | |
601 | |
495 | if (to_json) |
602 | if (to_json) |
496 | { |
603 | { |
497 | dSP; |
604 | dSP; |
498 | ENTER; |
605 | |
499 | SAVETMPS; |
|
|
500 | PUSHMARK (SP); |
606 | ENTER; SAVETMPS; PUSHMARK (SP); |
501 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
607 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
502 | |
608 | |
503 | // calling with G_SCALAR ensures that we always get a 1 reutrn value |
609 | // calling with G_SCALAR ensures that we always get a 1 return value |
504 | // check anyways. |
|
|
505 | PUTBACK; |
610 | PUTBACK; |
506 | assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR)); |
611 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
507 | SPAGAIN; |
612 | SPAGAIN; |
508 | |
613 | |
|
|
614 | // catch this surprisingly common error |
|
|
615 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
616 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
617 | |
|
|
618 | sv = POPs; |
|
|
619 | PUTBACK; |
|
|
620 | |
509 | encode_sv (enc, POPs); |
621 | encode_sv (enc, sv); |
510 | |
622 | |
511 | FREETMPS; |
623 | FREETMPS; LEAVE; |
512 | LEAVE; |
|
|
513 | } |
624 | } |
514 | else if (enc->flags & F_ALLOW_BLESSED) |
625 | else if (enc->json.flags & F_ALLOW_BLESSED) |
515 | encode_str (enc, "null", 4, 0); |
626 | encode_str (enc, "null", 4, 0); |
516 | else |
627 | else |
517 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
628 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
518 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
629 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
519 | } |
630 | } |
520 | else if (enc->flags & F_ALLOW_BLESSED) |
631 | else if (enc->json.flags & F_ALLOW_BLESSED) |
521 | encode_str (enc, "null", 4, 0); |
632 | encode_str (enc, "null", 4, 0); |
522 | else |
633 | else |
523 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
634 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
524 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
635 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
525 | } |
636 | } |
… | |
… | |
528 | encode_hv (enc, (HV *)sv); |
639 | encode_hv (enc, (HV *)sv); |
529 | else if (svt == SVt_PVAV) |
640 | else if (svt == SVt_PVAV) |
530 | encode_av (enc, (AV *)sv); |
641 | encode_av (enc, (AV *)sv); |
531 | else if (svt < SVt_PVAV) |
642 | else if (svt < SVt_PVAV) |
532 | { |
643 | { |
533 | if (SvNIOK (sv) && SvIV (sv) == 0) |
644 | STRLEN len = 0; |
|
|
645 | char *pv = svt ? SvPV (sv, len) : 0; |
|
|
646 | |
|
|
647 | if (len == 1 && *pv == '1') |
|
|
648 | encode_str (enc, "true", 4, 0); |
|
|
649 | else if (len == 1 && *pv == '0') |
534 | encode_str (enc, "false", 5, 0); |
650 | encode_str (enc, "false", 5, 0); |
535 | else if (SvNIOK (sv) && SvIV (sv) == 1) |
651 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
536 | encode_str (enc, "true", 4, 0); |
652 | encode_str (enc, "null", 4, 0); |
537 | else |
653 | else |
538 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
654 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
539 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
655 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
540 | } |
656 | } |
|
|
657 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
658 | encode_str (enc, "null", 4, 0); |
541 | else |
659 | else |
542 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
660 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
543 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
661 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
544 | } |
662 | } |
545 | |
663 | |
… | |
… | |
563 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
681 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
564 | enc->cur += strlen (enc->cur); |
682 | enc->cur += strlen (enc->cur); |
565 | } |
683 | } |
566 | else if (SvIOKp (sv)) |
684 | else if (SvIOKp (sv)) |
567 | { |
685 | { |
568 | // we assume we can always read an IV as a UV |
686 | // we assume we can always read an IV as a UV and vice versa |
569 | if (SvUV (sv) & ~(UV)0x7fff) |
687 | // we assume two's complement |
570 | { |
688 | // we assume no aliasing issues in the union |
571 | // large integer, use the (rather slow) snprintf way. |
689 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
572 | need (enc, sizeof (UV) * 3); |
690 | : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
573 | enc->cur += |
|
|
574 | SvIsUV(sv) |
|
|
575 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
|
|
576 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
577 | } |
|
|
578 | else |
|
|
579 | { |
691 | { |
580 | // optimise the "small number case" |
692 | // optimise the "small number case" |
581 | // code will likely be branchless and use only a single multiplication |
693 | // code will likely be branchless and use only a single multiplication |
|
|
694 | // works for numbers up to 59074 |
582 | I32 i = SvIV (sv); |
695 | I32 i = SvIVX (sv); |
583 | U32 u; |
696 | U32 u; |
584 | char digit, nz = 0; |
697 | char digit, nz = 0; |
585 | |
698 | |
586 | need (enc, 6); |
699 | need (enc, 6); |
587 | |
700 | |
… | |
… | |
593 | |
706 | |
594 | // now output digit by digit, each time masking out the integer part |
707 | // now output digit by digit, each time masking out the integer part |
595 | // and multiplying by 5 while moving the decimal point one to the right, |
708 | // and multiplying by 5 while moving the decimal point one to the right, |
596 | // resulting in a net multiplication by 10. |
709 | // resulting in a net multiplication by 10. |
597 | // we always write the digit to memory but conditionally increment |
710 | // we always write the digit to memory but conditionally increment |
598 | // the pointer, to ease the usage of conditional move instructions. |
711 | // the pointer, to enable the use of conditional move instructions. |
599 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
712 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
600 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
713 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
601 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
714 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
602 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
715 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
603 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
716 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
604 | } |
717 | } |
|
|
718 | else |
|
|
719 | { |
|
|
720 | // large integer, use the (rather slow) snprintf way. |
|
|
721 | need (enc, IVUV_MAXCHARS); |
|
|
722 | enc->cur += |
|
|
723 | SvIsUV(sv) |
|
|
724 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
725 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
726 | } |
605 | } |
727 | } |
606 | else if (SvROK (sv)) |
728 | else if (SvROK (sv)) |
607 | encode_rv (enc, SvRV (sv)); |
729 | encode_rv (enc, SvRV (sv)); |
608 | else if (!SvOK (sv)) |
730 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
609 | encode_str (enc, "null", 4, 0); |
731 | encode_str (enc, "null", 4, 0); |
610 | else |
732 | else |
611 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
733 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
612 | SvPV_nolen (sv), SvFLAGS (sv)); |
734 | SvPV_nolen (sv), SvFLAGS (sv)); |
613 | } |
735 | } |
614 | |
736 | |
615 | static SV * |
737 | static SV * |
616 | encode_json (SV *scalar, U32 flags) |
738 | encode_json (SV *scalar, JSON *json) |
617 | { |
739 | { |
618 | enc_t enc; |
740 | enc_t enc; |
619 | |
741 | |
620 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
742 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
621 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
743 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
622 | |
744 | |
623 | enc.flags = flags; |
745 | enc.json = *json; |
624 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
746 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
625 | enc.cur = SvPVX (enc.sv); |
747 | enc.cur = SvPVX (enc.sv); |
626 | enc.end = SvEND (enc.sv); |
748 | enc.end = SvEND (enc.sv); |
627 | enc.indent = 0; |
749 | enc.indent = 0; |
628 | enc.maxdepth = DEC_DEPTH (flags); |
750 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
751 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
752 | : 0x110000UL; |
629 | |
753 | |
630 | SvPOK_only (enc.sv); |
754 | SvPOK_only (enc.sv); |
631 | encode_sv (&enc, scalar); |
755 | encode_sv (&enc, scalar); |
632 | |
756 | |
633 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
757 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
634 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
758 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
635 | |
759 | |
636 | if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
760 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
637 | SvUTF8_on (enc.sv); |
761 | SvUTF8_on (enc.sv); |
638 | |
762 | |
639 | if (enc.flags & F_SHRINK) |
763 | if (enc.json.flags & F_SHRINK) |
640 | shrink (enc.sv); |
764 | shrink (enc.sv); |
641 | |
765 | |
642 | return enc.sv; |
766 | return enc.sv; |
643 | } |
767 | } |
644 | |
768 | |
… | |
… | |
649 | typedef struct |
773 | typedef struct |
650 | { |
774 | { |
651 | char *cur; // current parser pointer |
775 | char *cur; // current parser pointer |
652 | char *end; // end of input string |
776 | char *end; // end of input string |
653 | const char *err; // parse error, if != 0 |
777 | const char *err; // parse error, if != 0 |
654 | U32 flags; // F_* |
778 | JSON json; |
655 | U32 depth; // recursion depth |
779 | U32 depth; // recursion depth |
656 | U32 maxdepth; // recursion depth limit |
780 | U32 maxdepth; // recursion depth limit |
657 | } dec_t; |
781 | } dec_t; |
658 | |
782 | |
659 | inline void |
783 | INLINE void |
|
|
784 | decode_comment (dec_t *dec) |
|
|
785 | { |
|
|
786 | // only '#'-style comments allowed a.t.m. |
|
|
787 | |
|
|
788 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
789 | ++dec->cur; |
|
|
790 | } |
|
|
791 | |
|
|
792 | INLINE void |
660 | decode_ws (dec_t *dec) |
793 | decode_ws (dec_t *dec) |
661 | { |
794 | { |
662 | for (;;) |
795 | for (;;) |
663 | { |
796 | { |
664 | char ch = *dec->cur; |
797 | char ch = *dec->cur; |
665 | |
798 | |
666 | if (ch > 0x20 |
799 | if (ch > 0x20) |
|
|
800 | { |
|
|
801 | if (expect_false (ch == '#')) |
|
|
802 | { |
|
|
803 | if (dec->json.flags & F_RELAXED) |
|
|
804 | decode_comment (dec); |
|
|
805 | else |
|
|
806 | break; |
|
|
807 | } |
|
|
808 | else |
|
|
809 | break; |
|
|
810 | } |
667 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
811 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
668 | break; |
812 | break; // parse error, but let higher level handle it, gives better error messages |
669 | |
813 | |
670 | ++dec->cur; |
814 | ++dec->cur; |
671 | } |
815 | } |
672 | } |
816 | } |
673 | |
817 | |
… | |
… | |
677 | if (*dec->cur != ch) \ |
821 | if (*dec->cur != ch) \ |
678 | ERR (# ch " expected"); \ |
822 | ERR (# ch " expected"); \ |
679 | ++dec->cur; \ |
823 | ++dec->cur; \ |
680 | SE |
824 | SE |
681 | |
825 | |
682 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
826 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
683 | #define DEC_DEC_DEPTH --dec->depth |
827 | #define DEC_DEC_DEPTH --dec->depth |
684 | |
828 | |
685 | static SV *decode_sv (dec_t *dec); |
829 | static SV *decode_sv (dec_t *dec); |
686 | |
830 | |
687 | static signed char decode_hexdigit[256]; |
831 | static signed char decode_hexdigit[256]; |
… | |
… | |
779 | |
923 | |
780 | if (hi >= 0x80) |
924 | if (hi >= 0x80) |
781 | { |
925 | { |
782 | utf8 = 1; |
926 | utf8 = 1; |
783 | |
927 | |
784 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
928 | cur = encode_utf8 (cur, hi); |
785 | } |
929 | } |
786 | else |
930 | else |
787 | *cur++ = hi; |
931 | *cur++ = hi; |
788 | } |
932 | } |
789 | break; |
933 | break; |
… | |
… | |
791 | default: |
935 | default: |
792 | --dec_cur; |
936 | --dec_cur; |
793 | ERR ("illegal backslash escape sequence in string"); |
937 | ERR ("illegal backslash escape sequence in string"); |
794 | } |
938 | } |
795 | } |
939 | } |
796 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
940 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
797 | *cur++ = ch; |
941 | *cur++ = ch; |
798 | else if (ch >= 0x80) |
942 | else if (ch >= 0x80) |
799 | { |
943 | { |
800 | STRLEN clen; |
944 | STRLEN clen; |
801 | UV uch; |
945 | UV uch; |
… | |
… | |
922 | is_nv = 1; |
1066 | is_nv = 1; |
923 | } |
1067 | } |
924 | |
1068 | |
925 | if (!is_nv) |
1069 | if (!is_nv) |
926 | { |
1070 | { |
|
|
1071 | int len = dec->cur - start; |
|
|
1072 | |
927 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1073 | // special case the rather common 1..5-digit-int case |
928 | if (*start == '-') |
1074 | if (*start == '-') |
929 | switch (dec->cur - start) |
1075 | switch (len) |
930 | { |
1076 | { |
931 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1077 | case 2: return newSViv (-( start [1] - '0' * 1)); |
932 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1078 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
933 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1079 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
934 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1080 | case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1081 | case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
935 | } |
1082 | } |
936 | else |
1083 | else |
937 | switch (dec->cur - start) |
1084 | switch (len) |
938 | { |
1085 | { |
939 | case 1: return newSViv ( start [0] - '0' * 1); |
1086 | case 1: return newSViv ( start [0] - '0' * 1); |
940 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1087 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
941 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1088 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
942 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1089 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1090 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
943 | } |
1091 | } |
944 | |
1092 | |
945 | { |
1093 | { |
946 | UV uv; |
1094 | UV uv; |
947 | int numtype = grok_number (start, dec->cur - start, &uv); |
1095 | int numtype = grok_number (start, len, &uv); |
948 | if (numtype & IS_NUMBER_IN_UV) |
1096 | if (numtype & IS_NUMBER_IN_UV) |
949 | if (numtype & IS_NUMBER_NEG) |
1097 | if (numtype & IS_NUMBER_NEG) |
950 | { |
1098 | { |
951 | if (uv < (UV)IV_MIN) |
1099 | if (uv < (UV)IV_MIN) |
952 | return newSViv (-(IV)uv); |
1100 | return newSViv (-(IV)uv); |
953 | } |
1101 | } |
954 | else |
1102 | else |
955 | return newSVuv (uv); |
1103 | return newSVuv (uv); |
956 | |
|
|
957 | // here would likely be the place for bigint support |
|
|
958 | } |
1104 | } |
959 | } |
|
|
960 | |
1105 | |
961 | // if we ever support bigint or bigfloat, this is the place for bigfloat |
1106 | len -= *start == '-' ? 1 : 0; |
|
|
1107 | |
|
|
1108 | // does not fit into IV or UV, try NV |
|
|
1109 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
|
|
1110 | #if defined (LDBL_DIG) |
|
|
1111 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
1112 | #endif |
|
|
1113 | ) |
|
|
1114 | // fits into NV without loss of precision |
|
|
1115 | return newSVnv (Atof (start)); |
|
|
1116 | |
|
|
1117 | // everything else fails, convert it to a string |
|
|
1118 | return newSVpvn (start, dec->cur - start); |
|
|
1119 | } |
|
|
1120 | |
|
|
1121 | // loss of precision here |
962 | return newSVnv (Atof (start)); |
1122 | return newSVnv (Atof (start)); |
963 | |
1123 | |
964 | fail: |
1124 | fail: |
965 | return 0; |
1125 | return 0; |
966 | } |
1126 | } |
… | |
… | |
996 | |
1156 | |
997 | if (*dec->cur != ',') |
1157 | if (*dec->cur != ',') |
998 | ERR (", or ] expected while parsing array"); |
1158 | ERR (", or ] expected while parsing array"); |
999 | |
1159 | |
1000 | ++dec->cur; |
1160 | ++dec->cur; |
|
|
1161 | |
|
|
1162 | decode_ws (dec); |
|
|
1163 | |
|
|
1164 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1165 | { |
|
|
1166 | ++dec->cur; |
|
|
1167 | break; |
|
|
1168 | } |
1001 | } |
1169 | } |
1002 | |
1170 | |
1003 | DEC_DEC_DEPTH; |
1171 | DEC_DEC_DEPTH; |
1004 | return newRV_noinc ((SV *)av); |
1172 | return newRV_noinc ((SV *)av); |
1005 | |
1173 | |
… | |
… | |
1010 | } |
1178 | } |
1011 | |
1179 | |
1012 | static SV * |
1180 | static SV * |
1013 | decode_hv (dec_t *dec) |
1181 | decode_hv (dec_t *dec) |
1014 | { |
1182 | { |
|
|
1183 | SV *sv; |
1015 | HV *hv = newHV (); |
1184 | HV *hv = newHV (); |
1016 | |
1185 | |
1017 | DEC_INC_DEPTH; |
1186 | DEC_INC_DEPTH; |
1018 | decode_ws (dec); |
1187 | decode_ws (dec); |
1019 | |
1188 | |
1020 | if (*dec->cur == '}') |
1189 | if (*dec->cur == '}') |
1021 | ++dec->cur; |
1190 | ++dec->cur; |
1022 | else |
1191 | else |
1023 | for (;;) |
1192 | for (;;) |
1024 | { |
1193 | { |
|
|
1194 | EXPECT_CH ('"'); |
|
|
1195 | |
|
|
1196 | // heuristic: assume that |
|
|
1197 | // a) decode_str + hv_store_ent are abysmally slow. |
|
|
1198 | // b) most hash keys are short, simple ascii text. |
|
|
1199 | // => try to "fast-match" such strings to avoid |
|
|
1200 | // the overhead of decode_str + hv_store_ent. |
|
|
1201 | { |
1025 | SV *key, *value; |
1202 | SV *value; |
|
|
1203 | char *p = dec->cur; |
|
|
1204 | char *e = p + 24; // only try up to 24 bytes |
1026 | |
1205 | |
1027 | decode_ws (dec); EXPECT_CH ('"'); |
1206 | for (;;) |
1028 | |
|
|
1029 | key = decode_str (dec); |
|
|
1030 | if (!key) |
|
|
1031 | goto fail; |
|
|
1032 | |
|
|
1033 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1034 | |
|
|
1035 | value = decode_sv (dec); |
|
|
1036 | if (!value) |
|
|
1037 | { |
1207 | { |
|
|
1208 | // the >= 0x80 is false on most architectures |
|
|
1209 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1210 | { |
|
|
1211 | // slow path, back up and use decode_str |
|
|
1212 | SV *key = decode_str (dec); |
|
|
1213 | if (!key) |
|
|
1214 | goto fail; |
|
|
1215 | |
|
|
1216 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1217 | |
|
|
1218 | decode_ws (dec); |
|
|
1219 | value = decode_sv (dec); |
|
|
1220 | if (!value) |
|
|
1221 | { |
|
|
1222 | SvREFCNT_dec (key); |
|
|
1223 | goto fail; |
|
|
1224 | } |
|
|
1225 | |
|
|
1226 | hv_store_ent (hv, key, value, 0); |
1038 | SvREFCNT_dec (key); |
1227 | SvREFCNT_dec (key); |
|
|
1228 | |
|
|
1229 | break; |
|
|
1230 | } |
|
|
1231 | else if (*p == '"') |
|
|
1232 | { |
|
|
1233 | // fast path, got a simple key |
|
|
1234 | char *key = dec->cur; |
|
|
1235 | int len = p - key; |
|
|
1236 | dec->cur = p + 1; |
|
|
1237 | |
|
|
1238 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1239 | |
|
|
1240 | decode_ws (dec); |
|
|
1241 | value = decode_sv (dec); |
|
|
1242 | if (!value) |
1039 | goto fail; |
1243 | goto fail; |
|
|
1244 | |
|
|
1245 | hv_store (hv, key, len, value, 0); |
|
|
1246 | |
|
|
1247 | break; |
|
|
1248 | } |
|
|
1249 | |
|
|
1250 | ++p; |
1040 | } |
1251 | } |
1041 | |
1252 | } |
1042 | hv_store_ent (hv, key, value, 0); |
|
|
1043 | SvREFCNT_dec (key); |
|
|
1044 | |
1253 | |
1045 | decode_ws (dec); |
1254 | decode_ws (dec); |
1046 | |
1255 | |
1047 | if (*dec->cur == '}') |
1256 | if (*dec->cur == '}') |
1048 | { |
1257 | { |
… | |
… | |
1052 | |
1261 | |
1053 | if (*dec->cur != ',') |
1262 | if (*dec->cur != ',') |
1054 | ERR (", or } expected while parsing object/hash"); |
1263 | ERR (", or } expected while parsing object/hash"); |
1055 | |
1264 | |
1056 | ++dec->cur; |
1265 | ++dec->cur; |
|
|
1266 | |
|
|
1267 | decode_ws (dec); |
|
|
1268 | |
|
|
1269 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1270 | { |
|
|
1271 | ++dec->cur; |
|
|
1272 | break; |
|
|
1273 | } |
1057 | } |
1274 | } |
1058 | |
1275 | |
1059 | DEC_DEC_DEPTH; |
1276 | DEC_DEC_DEPTH; |
1060 | return newRV_noinc ((SV *)hv); |
1277 | sv = newRV_noinc ((SV *)hv); |
|
|
1278 | |
|
|
1279 | // check filter callbacks |
|
|
1280 | if (dec->json.flags & F_HOOK) |
|
|
1281 | { |
|
|
1282 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1283 | { |
|
|
1284 | HE *cb, *he; |
|
|
1285 | |
|
|
1286 | hv_iterinit (hv); |
|
|
1287 | he = hv_iternext (hv); |
|
|
1288 | hv_iterinit (hv); |
|
|
1289 | |
|
|
1290 | // the next line creates a mortal sv each time its called. |
|
|
1291 | // might want to optimise this for common cases. |
|
|
1292 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1293 | |
|
|
1294 | if (cb) |
|
|
1295 | { |
|
|
1296 | dSP; |
|
|
1297 | int count; |
|
|
1298 | |
|
|
1299 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1300 | XPUSHs (HeVAL (he)); |
|
|
1301 | |
|
|
1302 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1303 | |
|
|
1304 | if (count == 1) |
|
|
1305 | { |
|
|
1306 | sv = newSVsv (POPs); |
|
|
1307 | FREETMPS; LEAVE; |
|
|
1308 | return sv; |
|
|
1309 | } |
|
|
1310 | |
|
|
1311 | FREETMPS; LEAVE; |
|
|
1312 | } |
|
|
1313 | } |
|
|
1314 | |
|
|
1315 | if (dec->json.cb_object) |
|
|
1316 | { |
|
|
1317 | dSP; |
|
|
1318 | int count; |
|
|
1319 | |
|
|
1320 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1321 | XPUSHs (sv_2mortal (sv)); |
|
|
1322 | |
|
|
1323 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1324 | |
|
|
1325 | if (count == 1) |
|
|
1326 | { |
|
|
1327 | sv = newSVsv (POPs); |
|
|
1328 | FREETMPS; LEAVE; |
|
|
1329 | return sv; |
|
|
1330 | } |
|
|
1331 | |
|
|
1332 | SvREFCNT_inc (sv); |
|
|
1333 | FREETMPS; LEAVE; |
|
|
1334 | } |
|
|
1335 | } |
|
|
1336 | |
|
|
1337 | return sv; |
1061 | |
1338 | |
1062 | fail: |
1339 | fail: |
1063 | SvREFCNT_dec (hv); |
1340 | SvREFCNT_dec (hv); |
1064 | DEC_DEC_DEPTH; |
1341 | DEC_DEC_DEPTH; |
1065 | return 0; |
1342 | return 0; |
1066 | } |
1343 | } |
1067 | |
1344 | |
1068 | static SV * |
1345 | static SV * |
1069 | decode_sv (dec_t *dec) |
1346 | decode_sv (dec_t *dec) |
1070 | { |
1347 | { |
1071 | decode_ws (dec); |
|
|
1072 | |
|
|
1073 | // the beauty of JSON: you need exactly one character lookahead |
1348 | // the beauty of JSON: you need exactly one character lookahead |
1074 | // to parse anything. |
1349 | // to parse everything. |
1075 | switch (*dec->cur) |
1350 | switch (*dec->cur) |
1076 | { |
1351 | { |
1077 | case '"': ++dec->cur; return decode_str (dec); |
1352 | case '"': ++dec->cur; return decode_str (dec); |
1078 | case '[': ++dec->cur; return decode_av (dec); |
1353 | case '[': ++dec->cur; return decode_av (dec); |
1079 | case '{': ++dec->cur; return decode_hv (dec); |
1354 | case '{': ++dec->cur; return decode_hv (dec); |
1080 | |
1355 | |
1081 | case '-': |
1356 | case '-': |
1082 | case '0': case '1': case '2': case '3': case '4': |
1357 | case '0': case '1': case '2': case '3': case '4': |
1083 | case '5': case '6': case '7': case '8': case '9': |
1358 | case '5': case '6': case '7': case '8': case '9': |
1084 | return decode_num (dec); |
1359 | return decode_num (dec); |
1085 | |
1360 | |
1086 | case 't': |
1361 | case 't': |
1087 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1362 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1088 | { |
1363 | { |
1089 | dec->cur += 4; |
1364 | dec->cur += 4; |
|
|
1365 | #if JSON_SLOW |
|
|
1366 | json_true = get_bool ("JSON::XS::true"); |
|
|
1367 | #endif |
1090 | return SvREFCNT_inc (json_true); |
1368 | return newSVsv (json_true); |
1091 | } |
1369 | } |
1092 | else |
1370 | else |
1093 | ERR ("'true' expected"); |
1371 | ERR ("'true' expected"); |
1094 | |
1372 | |
1095 | break; |
1373 | break; |
1096 | |
1374 | |
1097 | case 'f': |
1375 | case 'f': |
1098 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1376 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1099 | { |
1377 | { |
1100 | dec->cur += 5; |
1378 | dec->cur += 5; |
|
|
1379 | #if JSON_SLOW |
|
|
1380 | json_false = get_bool ("JSON::XS::false"); |
|
|
1381 | #endif |
1101 | return SvREFCNT_inc (json_false); |
1382 | return newSVsv (json_false); |
1102 | } |
1383 | } |
1103 | else |
1384 | else |
1104 | ERR ("'false' expected"); |
1385 | ERR ("'false' expected"); |
1105 | |
1386 | |
1106 | break; |
1387 | break; |
… | |
… | |
1124 | fail: |
1405 | fail: |
1125 | return 0; |
1406 | return 0; |
1126 | } |
1407 | } |
1127 | |
1408 | |
1128 | static SV * |
1409 | static SV * |
1129 | decode_json (SV *string, U32 flags, UV *offset_return) |
1410 | decode_json (SV *string, JSON *json, STRLEN *offset_return) |
1130 | { |
1411 | { |
1131 | dec_t dec; |
1412 | dec_t dec; |
1132 | UV offset; |
1413 | STRLEN offset; |
1133 | SV *sv; |
1414 | SV *sv; |
1134 | |
1415 | |
1135 | SvGETMAGIC (string); |
1416 | SvGETMAGIC (string); |
1136 | SvUPGRADE (string, SVt_PV); |
1417 | SvUPGRADE (string, SVt_PV); |
1137 | |
1418 | |
|
|
1419 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1420 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1421 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1422 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1423 | */ |
|
|
1424 | SvPOK (string); |
|
|
1425 | |
|
|
1426 | if (SvCUR (string) > json->max_size && json->max_size) |
|
|
1427 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1428 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1429 | |
1138 | if (flags & F_UTF8) |
1430 | if (json->flags & F_UTF8) |
1139 | sv_utf8_downgrade (string, 0); |
1431 | sv_utf8_downgrade (string, 0); |
1140 | else |
1432 | else |
1141 | sv_utf8_upgrade (string); |
1433 | sv_utf8_upgrade (string); |
1142 | |
1434 | |
1143 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1435 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1144 | |
1436 | |
1145 | dec.flags = flags; |
1437 | dec.json = *json; |
1146 | dec.cur = SvPVX (string); |
1438 | dec.cur = SvPVX (string); |
1147 | dec.end = SvEND (string); |
1439 | dec.end = SvEND (string); |
1148 | dec.err = 0; |
1440 | dec.err = 0; |
1149 | dec.depth = 0; |
1441 | dec.depth = 0; |
1150 | dec.maxdepth = DEC_DEPTH (dec.flags); |
1442 | |
|
|
1443 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1444 | dec.json.flags |= F_HOOK; |
1151 | |
1445 | |
1152 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1446 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1447 | |
|
|
1448 | decode_ws (&dec); |
1153 | sv = decode_sv (&dec); |
1449 | sv = decode_sv (&dec); |
1154 | |
1450 | |
1155 | if (!(offset_return || !sv)) |
1451 | if (!(offset_return || !sv)) |
1156 | { |
1452 | { |
1157 | // check for trailing garbage |
1453 | // check for trailing garbage |
… | |
… | |
1165 | } |
1461 | } |
1166 | } |
1462 | } |
1167 | |
1463 | |
1168 | if (offset_return || !sv) |
1464 | if (offset_return || !sv) |
1169 | { |
1465 | { |
1170 | offset = dec.flags & F_UTF8 |
1466 | offset = dec.json.flags & F_UTF8 |
1171 | ? dec.cur - SvPVX (string) |
1467 | ? dec.cur - SvPVX (string) |
1172 | : utf8_distance (dec.cur, SvPVX (string)); |
1468 | : utf8_distance (dec.cur, SvPVX (string)); |
1173 | |
1469 | |
1174 | if (offset_return) |
1470 | if (offset_return) |
1175 | *offset_return = offset; |
1471 | *offset_return = offset; |
… | |
… | |
1194 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1490 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1195 | } |
1491 | } |
1196 | |
1492 | |
1197 | sv = sv_2mortal (sv); |
1493 | sv = sv_2mortal (sv); |
1198 | |
1494 | |
1199 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1495 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1200 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1496 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1201 | |
1497 | |
1202 | return sv; |
1498 | return sv; |
|
|
1499 | } |
|
|
1500 | |
|
|
1501 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1502 | // incremental parser |
|
|
1503 | |
|
|
1504 | static void |
|
|
1505 | incr_parse (JSON *self) |
|
|
1506 | { |
|
|
1507 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1508 | |
|
|
1509 | for (;;) |
|
|
1510 | { |
|
|
1511 | //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 |
|
|
1512 | switch (self->incr_mode) |
|
|
1513 | { |
|
|
1514 | // only used for intiial whitespace skipping |
|
|
1515 | case INCR_M_WS: |
|
|
1516 | for (;;) |
|
|
1517 | { |
|
|
1518 | if (*p > 0x20) |
|
|
1519 | { |
|
|
1520 | self->incr_mode = INCR_M_JSON; |
|
|
1521 | goto incr_m_json; |
|
|
1522 | } |
|
|
1523 | else if (!*p) |
|
|
1524 | goto interrupt; |
|
|
1525 | |
|
|
1526 | ++p; |
|
|
1527 | } |
|
|
1528 | |
|
|
1529 | // skip a single char inside a string (for \\-processing) |
|
|
1530 | case INCR_M_BS: |
|
|
1531 | if (!*p) |
|
|
1532 | goto interrupt; |
|
|
1533 | |
|
|
1534 | ++p; |
|
|
1535 | self->incr_mode = INCR_M_STR; |
|
|
1536 | goto incr_m_str; |
|
|
1537 | |
|
|
1538 | // inside a string |
|
|
1539 | case INCR_M_STR: |
|
|
1540 | incr_m_str: |
|
|
1541 | for (;;) |
|
|
1542 | { |
|
|
1543 | if (*p == '"') |
|
|
1544 | { |
|
|
1545 | ++p; |
|
|
1546 | self->incr_mode = INCR_M_JSON; |
|
|
1547 | |
|
|
1548 | if (!self->incr_nest) |
|
|
1549 | goto interrupt; |
|
|
1550 | |
|
|
1551 | goto incr_m_json; |
|
|
1552 | } |
|
|
1553 | else if (*p == '\\') |
|
|
1554 | { |
|
|
1555 | ++p; // "virtually" consumes character after \ |
|
|
1556 | |
|
|
1557 | if (!*p) // if at end of string we have to switch modes |
|
|
1558 | { |
|
|
1559 | self->incr_mode = INCR_M_BS; |
|
|
1560 | goto interrupt; |
|
|
1561 | } |
|
|
1562 | } |
|
|
1563 | else if (!*p) |
|
|
1564 | goto interrupt; |
|
|
1565 | |
|
|
1566 | ++p; |
|
|
1567 | } |
|
|
1568 | |
|
|
1569 | // after initial ws, outside string |
|
|
1570 | case INCR_M_JSON: |
|
|
1571 | incr_m_json: |
|
|
1572 | for (;;) |
|
|
1573 | { |
|
|
1574 | switch (*p++) |
|
|
1575 | { |
|
|
1576 | case 0: |
|
|
1577 | --p; |
|
|
1578 | goto interrupt; |
|
|
1579 | |
|
|
1580 | case 0x09: |
|
|
1581 | case 0x0a: |
|
|
1582 | case 0x0d: |
|
|
1583 | case 0x20: |
|
|
1584 | if (!self->incr_nest) |
|
|
1585 | { |
|
|
1586 | --p; // do not eat the whitespace, let the next round do it |
|
|
1587 | goto interrupt; |
|
|
1588 | } |
|
|
1589 | break; |
|
|
1590 | |
|
|
1591 | case '"': |
|
|
1592 | self->incr_mode = INCR_M_STR; |
|
|
1593 | goto incr_m_str; |
|
|
1594 | |
|
|
1595 | case '[': |
|
|
1596 | case '{': |
|
|
1597 | if (++self->incr_nest > self->max_depth) |
|
|
1598 | croak (ERR_NESTING_EXCEEDED); |
|
|
1599 | break; |
|
|
1600 | |
|
|
1601 | case ']': |
|
|
1602 | case '}': |
|
|
1603 | if (--self->incr_nest <= 0) |
|
|
1604 | goto interrupt; |
|
|
1605 | } |
|
|
1606 | } |
|
|
1607 | } |
|
|
1608 | |
|
|
1609 | modechange: |
|
|
1610 | ; |
|
|
1611 | } |
|
|
1612 | |
|
|
1613 | interrupt: |
|
|
1614 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1615 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1203 | } |
1616 | } |
1204 | |
1617 | |
1205 | ///////////////////////////////////////////////////////////////////////////// |
1618 | ///////////////////////////////////////////////////////////////////////////// |
1206 | // XS interface functions |
1619 | // XS interface functions |
1207 | |
1620 | |
… | |
… | |
1219 | : -1; |
1632 | : -1; |
1220 | |
1633 | |
1221 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1634 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1222 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1635 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1223 | |
1636 | |
1224 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1637 | json_true = get_bool ("JSON::XS::true"); |
1225 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1638 | json_false = get_bool ("JSON::XS::false"); |
1226 | } |
1639 | } |
1227 | |
1640 | |
1228 | PROTOTYPES: DISABLE |
1641 | PROTOTYPES: DISABLE |
1229 | |
1642 | |
1230 | SV *new (char *dummy) |
1643 | void CLONE (...) |
1231 | CODE: |
1644 | CODE: |
1232 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1645 | json_stash = 0; |
1233 | OUTPUT: |
1646 | json_boolean_stash = 0; |
1234 | RETVAL |
|
|
1235 | |
1647 | |
|
|
1648 | void new (char *klass) |
|
|
1649 | PPCODE: |
|
|
1650 | { |
|
|
1651 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1652 | SvPOK_only (pv); |
|
|
1653 | json_init ((JSON *)SvPVX (pv)); |
|
|
1654 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1655 | newRV_noinc (pv), |
|
|
1656 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1657 | ))); |
|
|
1658 | } |
|
|
1659 | |
1236 | SV *ascii (SV *self, int enable = 1) |
1660 | void ascii (JSON *self, int enable = 1) |
1237 | ALIAS: |
1661 | ALIAS: |
1238 | ascii = F_ASCII |
1662 | ascii = F_ASCII |
1239 | latin1 = F_LATIN1 |
1663 | latin1 = F_LATIN1 |
1240 | utf8 = F_UTF8 |
1664 | utf8 = F_UTF8 |
1241 | indent = F_INDENT |
1665 | indent = F_INDENT |
… | |
… | |
1245 | pretty = F_PRETTY |
1669 | pretty = F_PRETTY |
1246 | allow_nonref = F_ALLOW_NONREF |
1670 | allow_nonref = F_ALLOW_NONREF |
1247 | shrink = F_SHRINK |
1671 | shrink = F_SHRINK |
1248 | allow_blessed = F_ALLOW_BLESSED |
1672 | allow_blessed = F_ALLOW_BLESSED |
1249 | convert_blessed = F_CONV_BLESSED |
1673 | convert_blessed = F_CONV_BLESSED |
|
|
1674 | relaxed = F_RELAXED |
|
|
1675 | allow_unknown = F_ALLOW_UNKNOWN |
|
|
1676 | PPCODE: |
|
|
1677 | { |
|
|
1678 | if (enable) |
|
|
1679 | self->flags |= ix; |
|
|
1680 | else |
|
|
1681 | self->flags &= ~ix; |
|
|
1682 | |
|
|
1683 | XPUSHs (ST (0)); |
|
|
1684 | } |
|
|
1685 | |
|
|
1686 | void get_ascii (JSON *self) |
|
|
1687 | ALIAS: |
|
|
1688 | get_ascii = F_ASCII |
|
|
1689 | get_latin1 = F_LATIN1 |
|
|
1690 | get_utf8 = F_UTF8 |
|
|
1691 | get_indent = F_INDENT |
|
|
1692 | get_canonical = F_CANONICAL |
|
|
1693 | get_space_before = F_SPACE_BEFORE |
|
|
1694 | get_space_after = F_SPACE_AFTER |
|
|
1695 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1696 | get_shrink = F_SHRINK |
|
|
1697 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1698 | get_convert_blessed = F_CONV_BLESSED |
|
|
1699 | get_relaxed = F_RELAXED |
|
|
1700 | get_allow_unknown = F_ALLOW_UNKNOWN |
|
|
1701 | PPCODE: |
|
|
1702 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1703 | |
|
|
1704 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
|
|
1705 | PPCODE: |
|
|
1706 | self->max_depth = max_depth; |
|
|
1707 | XPUSHs (ST (0)); |
|
|
1708 | |
|
|
1709 | U32 get_max_depth (JSON *self) |
1250 | CODE: |
1710 | CODE: |
1251 | { |
1711 | RETVAL = self->max_depth; |
1252 | UV *uv = SvJSON (self); |
|
|
1253 | if (enable) |
|
|
1254 | *uv |= ix; |
|
|
1255 | else |
|
|
1256 | *uv &= ~ix; |
|
|
1257 | |
|
|
1258 | RETVAL = newSVsv (self); |
|
|
1259 | } |
|
|
1260 | OUTPUT: |
1712 | OUTPUT: |
1261 | RETVAL |
1713 | RETVAL |
1262 | |
1714 | |
1263 | SV *max_depth (SV *self, UV max_depth = 0x80000000UL) |
1715 | void max_size (JSON *self, U32 max_size = 0) |
|
|
1716 | PPCODE: |
|
|
1717 | self->max_size = max_size; |
|
|
1718 | XPUSHs (ST (0)); |
|
|
1719 | |
|
|
1720 | int get_max_size (JSON *self) |
1264 | CODE: |
1721 | CODE: |
1265 | { |
1722 | RETVAL = self->max_size; |
1266 | UV *uv = SvJSON (self); |
|
|
1267 | UV log2 = 0; |
|
|
1268 | |
|
|
1269 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
|
|
1270 | |
|
|
1271 | while ((1UL << log2) < max_depth) |
|
|
1272 | ++log2; |
|
|
1273 | |
|
|
1274 | *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1275 | |
|
|
1276 | RETVAL = newSVsv (self); |
|
|
1277 | } |
|
|
1278 | OUTPUT: |
1723 | OUTPUT: |
1279 | RETVAL |
1724 | RETVAL |
1280 | |
1725 | |
1281 | void encode (SV *self, SV *scalar) |
1726 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1282 | PPCODE: |
1727 | PPCODE: |
1283 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
1728 | { |
|
|
1729 | SvREFCNT_dec (self->cb_object); |
|
|
1730 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
1284 | |
1731 | |
1285 | void decode (SV *self, SV *jsonstr) |
1732 | XPUSHs (ST (0)); |
|
|
1733 | } |
|
|
1734 | |
|
|
1735 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
1286 | PPCODE: |
1736 | PPCODE: |
1287 | XPUSHs (decode_json (jsonstr, *SvJSON (self), 0)); |
1737 | { |
|
|
1738 | if (!self->cb_sk_object) |
|
|
1739 | self->cb_sk_object = newHV (); |
1288 | |
1740 | |
1289 | void decode_prefix (SV *self, SV *jsonstr) |
1741 | if (SvOK (cb)) |
|
|
1742 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
1743 | else |
|
|
1744 | { |
|
|
1745 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
|
|
1746 | |
|
|
1747 | if (!HvKEYS (self->cb_sk_object)) |
|
|
1748 | { |
|
|
1749 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1750 | self->cb_sk_object = 0; |
|
|
1751 | } |
|
|
1752 | } |
|
|
1753 | |
|
|
1754 | XPUSHs (ST (0)); |
|
|
1755 | } |
|
|
1756 | |
|
|
1757 | void encode (JSON *self, SV *scalar) |
1290 | PPCODE: |
1758 | PPCODE: |
|
|
1759 | XPUSHs (encode_json (scalar, self)); |
|
|
1760 | |
|
|
1761 | void decode (JSON *self, SV *jsonstr) |
|
|
1762 | PPCODE: |
|
|
1763 | XPUSHs (decode_json (jsonstr, self, 0)); |
|
|
1764 | |
|
|
1765 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
1766 | PPCODE: |
1291 | { |
1767 | { |
1292 | UV offset; |
1768 | STRLEN offset; |
1293 | EXTEND (SP, 2); |
1769 | EXTEND (SP, 2); |
1294 | PUSHs (decode_json (jsonstr, *SvJSON (self), &offset)); |
1770 | PUSHs (decode_json (jsonstr, self, &offset)); |
1295 | PUSHs (sv_2mortal (newSVuv (offset))); |
1771 | PUSHs (sv_2mortal (newSVuv (offset))); |
1296 | } |
1772 | } |
1297 | |
1773 | |
|
|
1774 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1775 | PPCODE: |
|
|
1776 | { |
|
|
1777 | if (!self->incr_text) |
|
|
1778 | self->incr_text = newSVpvn ("", 0); |
|
|
1779 | |
|
|
1780 | // append data, if any |
|
|
1781 | if (jsonstr) |
|
|
1782 | { |
|
|
1783 | if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) |
|
|
1784 | { |
|
|
1785 | /* utf-8-ness differs, need to upgrade */ |
|
|
1786 | sv_utf8_upgrade (self->incr_text); |
|
|
1787 | |
|
|
1788 | if (self->incr_pos) |
|
|
1789 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1790 | - (U8 *)SvPVX (self->incr_text); |
|
|
1791 | } |
|
|
1792 | |
|
|
1793 | { |
|
|
1794 | STRLEN len; |
|
|
1795 | const char *str = SvPV (jsonstr, len); |
|
|
1796 | SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); |
|
|
1797 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1798 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1799 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1800 | } |
|
|
1801 | } |
|
|
1802 | |
|
|
1803 | if (GIMME_V != G_VOID) |
|
|
1804 | do |
|
|
1805 | { |
|
|
1806 | STRLEN offset; |
|
|
1807 | |
|
|
1808 | if (!INCR_DONE (self)) |
|
|
1809 | { |
|
|
1810 | incr_parse (self); |
|
|
1811 | |
|
|
1812 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1813 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1814 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1815 | |
|
|
1816 | if (!INCR_DONE (self)) |
|
|
1817 | break; |
|
|
1818 | } |
|
|
1819 | |
|
|
1820 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1821 | |
|
|
1822 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset); |
|
|
1823 | self->incr_pos -= offset; |
|
|
1824 | self->incr_nest = 0; |
|
|
1825 | self->incr_mode = 0; |
|
|
1826 | } |
|
|
1827 | while (GIMME_V == G_ARRAY); |
|
|
1828 | } |
|
|
1829 | |
|
|
1830 | SV *incr_text (JSON *self) |
|
|
1831 | ATTRS: lvalue |
|
|
1832 | CODE: |
|
|
1833 | { |
|
|
1834 | if (self->incr_pos) |
|
|
1835 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
1836 | |
|
|
1837 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
1838 | } |
|
|
1839 | OUTPUT: |
|
|
1840 | RETVAL |
|
|
1841 | |
|
|
1842 | void incr_skip (JSON *self) |
|
|
1843 | CODE: |
|
|
1844 | { |
|
|
1845 | if (self->incr_pos) |
|
|
1846 | { |
|
|
1847 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
1848 | self->incr_pos = 0; |
|
|
1849 | self->incr_nest = 0; |
|
|
1850 | self->incr_mode = 0; |
|
|
1851 | } |
|
|
1852 | } |
|
|
1853 | |
|
|
1854 | void incr_reset (JSON *self) |
|
|
1855 | CODE: |
|
|
1856 | { |
|
|
1857 | SvREFCNT_dec (self->incr_text); |
|
|
1858 | self->incr_text = 0; |
|
|
1859 | self->incr_pos = 0; |
|
|
1860 | self->incr_nest = 0; |
|
|
1861 | self->incr_mode = 0; |
|
|
1862 | } |
|
|
1863 | |
|
|
1864 | void DESTROY (JSON *self) |
|
|
1865 | CODE: |
|
|
1866 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1867 | SvREFCNT_dec (self->cb_object); |
|
|
1868 | SvREFCNT_dec (self->incr_text); |
|
|
1869 | |
1298 | PROTOTYPES: ENABLE |
1870 | PROTOTYPES: ENABLE |
1299 | |
1871 | |
1300 | void to_json (SV *scalar) |
1872 | void encode_json (SV *scalar) |
1301 | ALIAS: |
1873 | ALIAS: |
1302 | objToJson = 0 |
1874 | to_json_ = 0 |
|
|
1875 | encode_json = F_UTF8 |
1303 | PPCODE: |
1876 | PPCODE: |
|
|
1877 | { |
|
|
1878 | JSON json; |
|
|
1879 | json_init (&json); |
|
|
1880 | json.flags |= ix; |
1304 | XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); |
1881 | XPUSHs (encode_json (scalar, &json)); |
|
|
1882 | } |
1305 | |
1883 | |
1306 | void from_json (SV *jsonstr) |
1884 | void decode_json (SV *jsonstr) |
1307 | ALIAS: |
1885 | ALIAS: |
1308 | jsonToObj = 0 |
1886 | from_json_ = 0 |
|
|
1887 | decode_json = F_UTF8 |
1309 | PPCODE: |
1888 | PPCODE: |
|
|
1889 | { |
|
|
1890 | JSON json; |
|
|
1891 | json_init (&json); |
|
|
1892 | json.flags |= ix; |
1310 | XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0)); |
1893 | XPUSHs (decode_json (jsonstr, &json, 0)); |
|
|
1894 | } |
1311 | |
1895 | |
|
|
1896 | |