1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
8 | #include "stdio.h" |
8 | #include <stdio.h> |
|
|
9 | #include <limits.h> |
|
|
10 | #include <float.h> |
9 | |
11 | |
10 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
12 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
11 | # define snprintf _snprintf // C compilers have this in stdio.h |
13 | # define snprintf _snprintf // C compilers have this in stdio.h |
12 | #endif |
14 | #endif |
13 | |
15 | |
14 | // some old perls do not have this, try to make it work, no |
16 | // some old perls do not have this, try to make it work, no |
15 | // guarentees, though. if it breaks, you get to keep the pieces. |
17 | // guarantees, though. if it breaks, you get to keep the pieces. |
16 | #ifndef UTF8_MAXBYTES |
18 | #ifndef UTF8_MAXBYTES |
17 | # define UTF8_MAXBYTES 13 |
19 | # define UTF8_MAXBYTES 13 |
18 | #endif |
20 | #endif |
|
|
21 | |
|
|
22 | // three extra for rounding, sign, and end of string |
|
|
23 | #define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
19 | |
24 | |
20 | #define F_ASCII 0x00000001UL |
25 | #define F_ASCII 0x00000001UL |
21 | #define F_LATIN1 0x00000002UL |
26 | #define F_LATIN1 0x00000002UL |
22 | #define F_UTF8 0x00000004UL |
27 | #define F_UTF8 0x00000004UL |
23 | #define F_INDENT 0x00000008UL |
28 | #define F_INDENT 0x00000008UL |
… | |
… | |
26 | #define F_SPACE_AFTER 0x00000040UL |
31 | #define F_SPACE_AFTER 0x00000040UL |
27 | #define F_ALLOW_NONREF 0x00000100UL |
32 | #define F_ALLOW_NONREF 0x00000100UL |
28 | #define F_SHRINK 0x00000200UL |
33 | #define F_SHRINK 0x00000200UL |
29 | #define F_ALLOW_BLESSED 0x00000400UL |
34 | #define F_ALLOW_BLESSED 0x00000400UL |
30 | #define F_CONV_BLESSED 0x00000800UL |
35 | #define F_CONV_BLESSED 0x00000800UL |
31 | #define F_MAXDEPTH 0xf8000000UL |
|
|
32 | #define S_MAXDEPTH 27 |
|
|
33 | #define F_MAXSIZE 0x01f00000UL |
36 | #define F_RELAXED 0x00001000UL |
34 | #define S_MAXSIZE 20 |
37 | #define F_ALLOW_UNKNOWN 0x00002000UL |
35 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
36 | |
39 | |
37 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
|
|
38 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
|
|
39 | |
|
|
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
40 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
41 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
|
|
42 | |
41 | |
43 | #define INIT_SIZE 32 // initial scalar size to be allocated |
42 | #define INIT_SIZE 32 // initial scalar size to be allocated |
44 | #define INDENT_STEP 3 // spaces per indentation level |
43 | #define INDENT_STEP 3 // spaces per indentation level |
45 | |
44 | |
46 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
45 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
47 | |
46 | |
48 | #define SB do { |
47 | #define SB do { |
49 | #define SE } while (0) |
48 | #define SE } while (0) |
50 | |
49 | |
51 | #if __GNUC__ >= 3 |
50 | #if __GNUC__ >= 3 |
52 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
51 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
53 | # define inline inline |
52 | # define INLINE static inline |
54 | #else |
53 | #else |
55 | # define expect(expr,value) (expr) |
54 | # define expect(expr,value) (expr) |
56 | # define inline static |
55 | # define INLINE static |
57 | #endif |
56 | #endif |
58 | |
57 | |
59 | #define expect_false(expr) expect ((expr) != 0, 0) |
58 | #define expect_false(expr) expect ((expr) != 0, 0) |
60 | #define expect_true(expr) expect ((expr) != 0, 1) |
59 | #define expect_true(expr) expect ((expr) != 0, 1) |
61 | |
60 | |
|
|
61 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
62 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
63 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
64 | |
|
|
65 | #define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
|
|
66 | |
|
|
67 | #ifdef USE_ITHREADS |
|
|
68 | # define JSON_SLOW 1 |
|
|
69 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
70 | #else |
|
|
71 | # define JSON_SLOW 0 |
|
|
72 | # define JSON_STASH json_stash |
|
|
73 | #endif |
|
|
74 | |
62 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
75 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
63 | static SV *json_true, *json_false; |
76 | static SV *json_true, *json_false; |
64 | |
77 | |
|
|
78 | enum { |
|
|
79 | INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
|
|
80 | INCR_M_STR, // inside string |
|
|
81 | INCR_M_BS, // inside backslash |
|
|
82 | INCR_M_C0, // inside comment in initial whitespace sequence |
|
|
83 | INCR_M_C1, // inside comment in other places |
|
|
84 | INCR_M_JSON // outside anything, count nesting |
|
|
85 | }; |
|
|
86 | |
|
|
87 | #define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
|
|
88 | |
65 | typedef struct { |
89 | typedef struct { |
66 | U32 flags; |
90 | U32 flags; |
|
|
91 | U32 max_depth; |
|
|
92 | STRLEN max_size; |
|
|
93 | |
67 | SV *cb_object; |
94 | SV *cb_object; |
68 | HV *cb_sk_object; |
95 | HV *cb_sk_object; |
|
|
96 | |
|
|
97 | // for the incremental parser |
|
|
98 | SV *incr_text; // the source text so far |
|
|
99 | STRLEN incr_pos; // the current offset into the text |
|
|
100 | int incr_nest; // {[]}-nesting level |
|
|
101 | unsigned char incr_mode; |
69 | } JSON; |
102 | } JSON; |
|
|
103 | |
|
|
104 | INLINE void |
|
|
105 | json_init (JSON *json) |
|
|
106 | { |
|
|
107 | Zero (json, 1, JSON); |
|
|
108 | json->max_depth = 512; |
|
|
109 | } |
70 | |
110 | |
71 | ///////////////////////////////////////////////////////////////////////////// |
111 | ///////////////////////////////////////////////////////////////////////////// |
72 | // utility functions |
112 | // utility functions |
73 | |
113 | |
74 | inline void |
114 | INLINE SV * |
|
|
115 | get_bool (const char *name) |
|
|
116 | { |
|
|
117 | SV *sv = get_sv (name, 1); |
|
|
118 | |
|
|
119 | SvREADONLY_on (sv); |
|
|
120 | SvREADONLY_on (SvRV (sv)); |
|
|
121 | |
|
|
122 | return sv; |
|
|
123 | } |
|
|
124 | |
|
|
125 | INLINE void |
75 | shrink (SV *sv) |
126 | shrink (SV *sv) |
76 | { |
127 | { |
77 | sv_utf8_downgrade (sv, 1); |
128 | sv_utf8_downgrade (sv, 1); |
|
|
129 | |
78 | if (SvLEN (sv) > SvCUR (sv) + 1) |
130 | if (SvLEN (sv) > SvCUR (sv) + 1) |
79 | { |
131 | { |
80 | #ifdef SvPV_shrink_to_cur |
132 | #ifdef SvPV_shrink_to_cur |
81 | SvPV_shrink_to_cur (sv); |
133 | SvPV_shrink_to_cur (sv); |
82 | #elif defined (SvPV_renew) |
134 | #elif defined (SvPV_renew) |
… | |
… | |
88 | // decode an utf-8 character and return it, or (UV)-1 in |
140 | // decode an utf-8 character and return it, or (UV)-1 in |
89 | // case of an error. |
141 | // case of an error. |
90 | // we special-case "safe" characters from U+80 .. U+7FF, |
142 | // we special-case "safe" characters from U+80 .. U+7FF, |
91 | // but use the very good perl function to parse anything else. |
143 | // but use the very good perl function to parse anything else. |
92 | // note that we never call this function for a ascii codepoints |
144 | // note that we never call this function for a ascii codepoints |
93 | inline UV |
145 | INLINE UV |
94 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
146 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
95 | { |
147 | { |
96 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
148 | if (expect_true (len >= 2 |
97 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
149 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
98 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
150 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
99 | { |
151 | { |
100 | *clen = 2; |
152 | *clen = 2; |
101 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
153 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
102 | } |
154 | } |
103 | else |
155 | else |
104 | { |
156 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
105 | *clen = (STRLEN)-1; |
157 | } |
106 | return (UV)-1; |
158 | |
107 | } |
159 | // likewise for encoding, also never called for ascii codepoints |
|
|
160 | // this function takes advantage of this fact, although current gccs |
|
|
161 | // seem to optimise the check for >= 0x80 away anyways |
|
|
162 | INLINE unsigned char * |
|
|
163 | encode_utf8 (unsigned char *s, UV ch) |
|
|
164 | { |
|
|
165 | if (expect_false (ch < 0x000080)) |
|
|
166 | *s++ = ch; |
|
|
167 | else if (expect_true (ch < 0x000800)) |
|
|
168 | *s++ = 0xc0 | ( ch >> 6), |
|
|
169 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
170 | else if ( ch < 0x010000) |
|
|
171 | *s++ = 0xe0 | ( ch >> 12), |
|
|
172 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
173 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
174 | else if ( ch < 0x110000) |
|
|
175 | *s++ = 0xf0 | ( ch >> 18), |
|
|
176 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
177 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
178 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
179 | |
|
|
180 | return s; |
|
|
181 | } |
|
|
182 | |
|
|
183 | // convert offset pointer to character index, sv must be string |
|
|
184 | static STRLEN |
|
|
185 | ptr_to_index (SV *sv, char *offset) |
|
|
186 | { |
|
|
187 | return SvUTF8 (sv) |
|
|
188 | ? utf8_distance (offset, SvPVX (sv)) |
|
|
189 | : offset - SvPVX (sv); |
108 | } |
190 | } |
109 | |
191 | |
110 | ///////////////////////////////////////////////////////////////////////////// |
192 | ///////////////////////////////////////////////////////////////////////////// |
111 | // encoder |
193 | // encoder |
112 | |
194 | |
… | |
… | |
116 | char *cur; // SvPVX (sv) + current output position |
198 | char *cur; // SvPVX (sv) + current output position |
117 | char *end; // SvEND (sv) |
199 | char *end; // SvEND (sv) |
118 | SV *sv; // result scalar |
200 | SV *sv; // result scalar |
119 | JSON json; |
201 | JSON json; |
120 | U32 indent; // indentation level |
202 | U32 indent; // indentation level |
121 | U32 maxdepth; // max. indentation/recursion level |
203 | UV limit; // escape character values >= this value when encoding |
122 | } enc_t; |
204 | } enc_t; |
123 | |
205 | |
124 | inline void |
206 | INLINE void |
125 | need (enc_t *enc, STRLEN len) |
207 | need (enc_t *enc, STRLEN len) |
126 | { |
208 | { |
127 | if (expect_false (enc->cur + len >= enc->end)) |
209 | if (expect_false (enc->cur + len >= enc->end)) |
128 | { |
210 | { |
129 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
211 | STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
130 | SvGROW (enc->sv, cur + len + 1); |
212 | SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
131 | enc->cur = SvPVX (enc->sv) + cur; |
213 | enc->cur = SvPVX (enc->sv) + cur; |
132 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
214 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
133 | } |
215 | } |
134 | } |
216 | } |
135 | |
217 | |
136 | inline void |
218 | INLINE void |
137 | encode_ch (enc_t *enc, char ch) |
219 | encode_ch (enc_t *enc, char ch) |
138 | { |
220 | { |
139 | need (enc, 1); |
221 | need (enc, 1); |
140 | *enc->cur++ = ch; |
222 | *enc->cur++ = ch; |
141 | } |
223 | } |
… | |
… | |
195 | { |
277 | { |
196 | uch = ch; |
278 | uch = ch; |
197 | clen = 1; |
279 | clen = 1; |
198 | } |
280 | } |
199 | |
281 | |
200 | if (uch > 0x10FFFFUL) |
282 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
201 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
202 | |
|
|
203 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
|
|
204 | { |
283 | { |
205 | if (uch > 0xFFFFUL) |
284 | if (uch >= 0x10000UL) |
206 | { |
285 | { |
|
|
286 | if (uch >= 0x110000UL) |
|
|
287 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
288 | |
207 | need (enc, len += 11); |
289 | need (enc, len += 11); |
208 | sprintf (enc->cur, "\\u%04x\\u%04x", |
290 | sprintf (enc->cur, "\\u%04x\\u%04x", |
209 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
291 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
210 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
292 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
211 | enc->cur += 12; |
293 | enc->cur += 12; |
212 | } |
294 | } |
213 | else |
295 | else |
214 | { |
296 | { |
215 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
216 | need (enc, len += 5); |
297 | need (enc, len += 5); |
217 | *enc->cur++ = '\\'; |
298 | *enc->cur++ = '\\'; |
218 | *enc->cur++ = 'u'; |
299 | *enc->cur++ = 'u'; |
219 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
300 | *enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
220 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
301 | *enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
221 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
302 | *enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
222 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
303 | *enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
223 | } |
304 | } |
224 | |
305 | |
225 | str += clen; |
306 | str += clen; |
226 | } |
307 | } |
227 | else if (enc->json.flags & F_LATIN1) |
308 | else if (enc->json.flags & F_LATIN1) |
… | |
… | |
239 | while (--clen); |
320 | while (--clen); |
240 | } |
321 | } |
241 | else |
322 | else |
242 | { |
323 | { |
243 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
324 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
244 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
325 | enc->cur = encode_utf8 (enc->cur, uch); |
245 | ++str; |
326 | ++str; |
246 | } |
327 | } |
247 | } |
328 | } |
248 | } |
329 | } |
249 | } |
330 | } |
250 | |
331 | |
251 | --len; |
332 | --len; |
252 | } |
333 | } |
253 | } |
334 | } |
254 | |
335 | |
255 | inline void |
336 | INLINE void |
256 | encode_indent (enc_t *enc) |
337 | encode_indent (enc_t *enc) |
257 | { |
338 | { |
258 | if (enc->json.flags & F_INDENT) |
339 | if (enc->json.flags & F_INDENT) |
259 | { |
340 | { |
260 | int spaces = enc->indent * INDENT_STEP; |
341 | int spaces = enc->indent * INDENT_STEP; |
… | |
… | |
263 | memset (enc->cur, ' ', spaces); |
344 | memset (enc->cur, ' ', spaces); |
264 | enc->cur += spaces; |
345 | enc->cur += spaces; |
265 | } |
346 | } |
266 | } |
347 | } |
267 | |
348 | |
268 | inline void |
349 | INLINE void |
269 | encode_space (enc_t *enc) |
350 | encode_space (enc_t *enc) |
270 | { |
351 | { |
271 | need (enc, 1); |
352 | need (enc, 1); |
272 | encode_ch (enc, ' '); |
353 | encode_ch (enc, ' '); |
273 | } |
354 | } |
274 | |
355 | |
275 | inline void |
356 | INLINE void |
276 | encode_nl (enc_t *enc) |
357 | encode_nl (enc_t *enc) |
277 | { |
358 | { |
278 | if (enc->json.flags & F_INDENT) |
359 | if (enc->json.flags & F_INDENT) |
279 | { |
360 | { |
280 | need (enc, 1); |
361 | need (enc, 1); |
281 | encode_ch (enc, '\n'); |
362 | encode_ch (enc, '\n'); |
282 | } |
363 | } |
283 | } |
364 | } |
284 | |
365 | |
285 | inline void |
366 | INLINE void |
286 | encode_comma (enc_t *enc) |
367 | encode_comma (enc_t *enc) |
287 | { |
368 | { |
288 | encode_ch (enc, ','); |
369 | encode_ch (enc, ','); |
289 | |
370 | |
290 | if (enc->json.flags & F_INDENT) |
371 | if (enc->json.flags & F_INDENT) |
… | |
… | |
298 | static void |
379 | static void |
299 | encode_av (enc_t *enc, AV *av) |
380 | encode_av (enc_t *enc, AV *av) |
300 | { |
381 | { |
301 | int i, len = av_len (av); |
382 | int i, len = av_len (av); |
302 | |
383 | |
303 | if (enc->indent >= enc->maxdepth) |
384 | if (enc->indent >= enc->json.max_depth) |
304 | croak ("data structure too deep (hit recursion limit)"); |
385 | croak (ERR_NESTING_EXCEEDED); |
305 | |
386 | |
306 | encode_ch (enc, '['); encode_nl (enc); |
387 | encode_ch (enc, '['); |
307 | ++enc->indent; |
388 | |
|
|
389 | if (len >= 0) |
|
|
390 | { |
|
|
391 | encode_nl (enc); ++enc->indent; |
308 | |
392 | |
309 | for (i = 0; i <= len; ++i) |
393 | for (i = 0; i <= len; ++i) |
310 | { |
394 | { |
|
|
395 | SV **svp = av_fetch (av, i, 0); |
|
|
396 | |
311 | encode_indent (enc); |
397 | encode_indent (enc); |
312 | encode_sv (enc, *av_fetch (av, i, 0)); |
|
|
313 | |
398 | |
|
|
399 | if (svp) |
|
|
400 | encode_sv (enc, *svp); |
|
|
401 | else |
|
|
402 | encode_str (enc, "null", 4, 0); |
|
|
403 | |
314 | if (i < len) |
404 | if (i < len) |
315 | encode_comma (enc); |
405 | encode_comma (enc); |
316 | } |
406 | } |
317 | |
407 | |
|
|
408 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
409 | } |
|
|
410 | |
318 | encode_nl (enc); |
411 | encode_ch (enc, ']'); |
319 | |
|
|
320 | --enc->indent; |
|
|
321 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
322 | } |
412 | } |
323 | |
413 | |
324 | static void |
414 | static void |
325 | encode_he (enc_t *enc, HE *he) |
415 | encode_hk (enc_t *enc, HE *he) |
326 | { |
416 | { |
327 | encode_ch (enc, '"'); |
417 | encode_ch (enc, '"'); |
328 | |
418 | |
329 | if (HeKLEN (he) == HEf_SVKEY) |
419 | if (HeKLEN (he) == HEf_SVKEY) |
330 | { |
420 | { |
… | |
… | |
343 | encode_ch (enc, '"'); |
433 | encode_ch (enc, '"'); |
344 | |
434 | |
345 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
435 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
346 | encode_ch (enc, ':'); |
436 | encode_ch (enc, ':'); |
347 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
437 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
348 | encode_sv (enc, HeVAL (he)); |
|
|
349 | } |
438 | } |
350 | |
439 | |
351 | // compare hash entries, used when all keys are bytestrings |
440 | // compare hash entries, used when all keys are bytestrings |
352 | static int |
441 | static int |
353 | he_cmp_fast (const void *a_, const void *b_) |
442 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
358 | HE *b = *(HE **)b_; |
447 | HE *b = *(HE **)b_; |
359 | |
448 | |
360 | STRLEN la = HeKLEN (a); |
449 | STRLEN la = HeKLEN (a); |
361 | STRLEN lb = HeKLEN (b); |
450 | STRLEN lb = HeKLEN (b); |
362 | |
451 | |
363 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
452 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
364 | cmp = la - lb; |
453 | cmp = lb - la; |
365 | |
454 | |
366 | return cmp; |
455 | return cmp; |
367 | } |
456 | } |
368 | |
457 | |
369 | // compare hash entries, used when some keys are sv's or utf-x |
458 | // compare hash entries, used when some keys are sv's or utf-x |
370 | static int |
459 | static int |
371 | he_cmp_slow (const void *a, const void *b) |
460 | he_cmp_slow (const void *a, const void *b) |
372 | { |
461 | { |
373 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
462 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
374 | } |
463 | } |
375 | |
464 | |
376 | static void |
465 | static void |
377 | encode_hv (enc_t *enc, HV *hv) |
466 | encode_hv (enc_t *enc, HV *hv) |
378 | { |
467 | { |
379 | int count, i; |
468 | HE *he; |
380 | |
469 | |
381 | if (enc->indent >= enc->maxdepth) |
470 | if (enc->indent >= enc->json.max_depth) |
382 | croak ("data structure too deep (hit recursion limit)"); |
471 | croak (ERR_NESTING_EXCEEDED); |
383 | |
472 | |
384 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
473 | encode_ch (enc, '{'); |
385 | |
474 | |
386 | if ((count = hv_iterinit (hv))) |
|
|
387 | { |
|
|
388 | // for canonical output we have to sort by keys first |
475 | // for canonical output we have to sort by keys first |
389 | // actually, this is mostly due to the stupid so-called |
476 | // actually, this is mostly due to the stupid so-called |
390 | // security workaround added somewhere in 5.8.x. |
477 | // security workaround added somewhere in 5.8.x |
391 | // that randomises hash orderings |
478 | // that randomises hash orderings |
392 | if (enc->json.flags & F_CANONICAL) |
479 | if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
|
|
480 | { |
|
|
481 | int count = hv_iterinit (hv); |
|
|
482 | |
|
|
483 | if (SvMAGICAL (hv)) |
393 | { |
484 | { |
|
|
485 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
486 | // but I don't care for the speed of this special case. |
|
|
487 | // note also that we will run into undefined behaviour when the two iterations |
|
|
488 | // do not result in the same count, something I might care for in some later release. |
|
|
489 | |
|
|
490 | count = 0; |
|
|
491 | while (hv_iternext (hv)) |
|
|
492 | ++count; |
|
|
493 | |
|
|
494 | hv_iterinit (hv); |
|
|
495 | } |
|
|
496 | |
|
|
497 | if (count) |
|
|
498 | { |
394 | int fast = 1; |
499 | int i, fast = 1; |
395 | HE *he; |
|
|
396 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
500 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
397 | HE **hes = _alloca (count * sizeof (HE)); |
501 | HE **hes = _alloca (count * sizeof (HE)); |
398 | #else |
502 | #else |
399 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
503 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
400 | #endif |
504 | #endif |
… | |
… | |
427 | |
531 | |
428 | FREETMPS; |
532 | FREETMPS; |
429 | LEAVE; |
533 | LEAVE; |
430 | } |
534 | } |
431 | |
535 | |
432 | for (i = 0; i < count; ++i) |
536 | encode_nl (enc); ++enc->indent; |
|
|
537 | |
|
|
538 | while (count--) |
433 | { |
539 | { |
434 | encode_indent (enc); |
540 | encode_indent (enc); |
|
|
541 | he = hes [count]; |
435 | encode_he (enc, hes [i]); |
542 | encode_hk (enc, he); |
|
|
543 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
436 | |
544 | |
437 | if (i < count - 1) |
545 | if (count) |
438 | encode_comma (enc); |
546 | encode_comma (enc); |
439 | } |
547 | } |
440 | |
548 | |
441 | encode_nl (enc); |
549 | encode_nl (enc); --enc->indent; encode_indent (enc); |
442 | } |
550 | } |
|
|
551 | } |
443 | else |
552 | else |
|
|
553 | { |
|
|
554 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
555 | if ((he = hv_iternext (hv))) |
444 | { |
556 | { |
445 | HE *he = hv_iternext (hv); |
557 | encode_nl (enc); ++enc->indent; |
446 | |
558 | |
447 | for (;;) |
559 | for (;;) |
448 | { |
560 | { |
449 | encode_indent (enc); |
561 | encode_indent (enc); |
450 | encode_he (enc, he); |
562 | encode_hk (enc, he); |
|
|
563 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
451 | |
564 | |
452 | if (!(he = hv_iternext (hv))) |
565 | if (!(he = hv_iternext (hv))) |
453 | break; |
566 | break; |
454 | |
567 | |
455 | encode_comma (enc); |
568 | encode_comma (enc); |
456 | } |
569 | } |
457 | |
570 | |
458 | encode_nl (enc); |
571 | encode_nl (enc); --enc->indent; encode_indent (enc); |
459 | } |
572 | } |
460 | } |
573 | } |
461 | |
574 | |
462 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
575 | encode_ch (enc, '}'); |
463 | } |
576 | } |
464 | |
577 | |
465 | // encode objects, arrays and special \0=false and \1=true values. |
578 | // encode objects, arrays and special \0=false and \1=true values. |
466 | static void |
579 | static void |
467 | encode_rv (enc_t *enc, SV *sv) |
580 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
471 | SvGETMAGIC (sv); |
584 | SvGETMAGIC (sv); |
472 | svt = SvTYPE (sv); |
585 | svt = SvTYPE (sv); |
473 | |
586 | |
474 | if (expect_false (SvOBJECT (sv))) |
587 | if (expect_false (SvOBJECT (sv))) |
475 | { |
588 | { |
|
|
589 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
590 | ? json_boolean_stash |
|
|
591 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
592 | |
476 | if (SvSTASH (sv) == json_boolean_stash) |
593 | if (SvSTASH (sv) == stash) |
477 | { |
594 | { |
478 | if (SvIV (sv)) |
595 | if (SvIV (sv)) |
479 | encode_str (enc, "true", 4, 0); |
596 | encode_str (enc, "true", 4, 0); |
480 | else |
597 | else |
481 | encode_str (enc, "false", 5, 0); |
598 | encode_str (enc, "false", 5, 0); |
… | |
… | |
489 | } |
606 | } |
490 | #endif |
607 | #endif |
491 | if (enc->json.flags & F_CONV_BLESSED) |
608 | if (enc->json.flags & F_CONV_BLESSED) |
492 | { |
609 | { |
493 | // we re-bless the reference to get overload and other niceties right |
610 | // we re-bless the reference to get overload and other niceties right |
494 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1); |
611 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
495 | |
612 | |
496 | if (to_json) |
613 | if (to_json) |
497 | { |
614 | { |
|
|
615 | dSP; |
|
|
616 | |
498 | dSP; ENTER; SAVETMPS; PUSHMARK (SP); |
617 | ENTER; SAVETMPS; PUSHMARK (SP); |
499 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
618 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
500 | |
619 | |
501 | // calling with G_SCALAR ensures that we always get a 1 reutrn value |
620 | // calling with G_SCALAR ensures that we always get a 1 return value |
502 | // check anyways. |
|
|
503 | PUTBACK; |
621 | PUTBACK; |
504 | assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR)); |
622 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
505 | SPAGAIN; |
623 | SPAGAIN; |
506 | |
624 | |
|
|
625 | // catch this surprisingly common error |
|
|
626 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
627 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
628 | |
|
|
629 | sv = POPs; |
|
|
630 | PUTBACK; |
|
|
631 | |
507 | encode_sv (enc, POPs); |
632 | encode_sv (enc, sv); |
508 | |
633 | |
509 | FREETMPS; LEAVE; |
634 | FREETMPS; LEAVE; |
510 | } |
635 | } |
511 | else if (enc->json.flags & F_ALLOW_BLESSED) |
636 | else if (enc->json.flags & F_ALLOW_BLESSED) |
512 | encode_str (enc, "null", 4, 0); |
637 | encode_str (enc, "null", 4, 0); |
… | |
… | |
532 | |
657 | |
533 | if (len == 1 && *pv == '1') |
658 | if (len == 1 && *pv == '1') |
534 | encode_str (enc, "true", 4, 0); |
659 | encode_str (enc, "true", 4, 0); |
535 | else if (len == 1 && *pv == '0') |
660 | else if (len == 1 && *pv == '0') |
536 | encode_str (enc, "false", 5, 0); |
661 | encode_str (enc, "false", 5, 0); |
|
|
662 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
663 | encode_str (enc, "null", 4, 0); |
537 | else |
664 | else |
538 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
665 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
539 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
666 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
540 | } |
667 | } |
|
|
668 | else if (enc->json.flags & F_ALLOW_UNKNOWN) |
|
|
669 | encode_str (enc, "null", 4, 0); |
541 | else |
670 | else |
542 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
671 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
543 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
672 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
544 | } |
673 | } |
545 | |
674 | |
… | |
… | |
563 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
692 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
564 | enc->cur += strlen (enc->cur); |
693 | enc->cur += strlen (enc->cur); |
565 | } |
694 | } |
566 | else if (SvIOKp (sv)) |
695 | else if (SvIOKp (sv)) |
567 | { |
696 | { |
568 | // we assume we can always read an IV as a UV |
697 | // we assume we can always read an IV as a UV and vice versa |
569 | if (SvUV (sv) & ~(UV)0x7fff) |
698 | // we assume two's complement |
570 | { |
699 | // we assume no aliasing issues in the union |
571 | // large integer, use the (rather slow) snprintf way. |
700 | if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
572 | need (enc, sizeof (UV) * 3); |
701 | : 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 | { |
702 | { |
580 | // optimise the "small number case" |
703 | // optimise the "small number case" |
581 | // code will likely be branchless and use only a single multiplication |
704 | // code will likely be branchless and use only a single multiplication |
|
|
705 | // works for numbers up to 59074 |
582 | I32 i = SvIV (sv); |
706 | I32 i = SvIVX (sv); |
583 | U32 u; |
707 | U32 u; |
584 | char digit, nz = 0; |
708 | char digit, nz = 0; |
585 | |
709 | |
586 | need (enc, 6); |
710 | need (enc, 6); |
587 | |
711 | |
… | |
… | |
593 | |
717 | |
594 | // now output digit by digit, each time masking out the integer part |
718 | // 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, |
719 | // and multiplying by 5 while moving the decimal point one to the right, |
596 | // resulting in a net multiplication by 10. |
720 | // resulting in a net multiplication by 10. |
597 | // we always write the digit to memory but conditionally increment |
721 | // we always write the digit to memory but conditionally increment |
598 | // the pointer, to ease the usage of conditional move instructions. |
722 | // 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; |
723 | 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; |
724 | 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; |
725 | 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; |
726 | 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' |
727 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
604 | } |
728 | } |
|
|
729 | else |
|
|
730 | { |
|
|
731 | // large integer, use the (rather slow) snprintf way. |
|
|
732 | need (enc, IVUV_MAXCHARS); |
|
|
733 | enc->cur += |
|
|
734 | SvIsUV(sv) |
|
|
735 | ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
|
|
736 | : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
|
|
737 | } |
605 | } |
738 | } |
606 | else if (SvROK (sv)) |
739 | else if (SvROK (sv)) |
607 | encode_rv (enc, SvRV (sv)); |
740 | encode_rv (enc, SvRV (sv)); |
608 | else if (!SvOK (sv)) |
741 | else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
609 | encode_str (enc, "null", 4, 0); |
742 | encode_str (enc, "null", 4, 0); |
610 | else |
743 | else |
611 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
744 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
612 | SvPV_nolen (sv), SvFLAGS (sv)); |
745 | SvPV_nolen (sv), SvFLAGS (sv)); |
613 | } |
746 | } |
… | |
… | |
623 | enc.json = *json; |
756 | enc.json = *json; |
624 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
757 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
625 | enc.cur = SvPVX (enc.sv); |
758 | enc.cur = SvPVX (enc.sv); |
626 | enc.end = SvEND (enc.sv); |
759 | enc.end = SvEND (enc.sv); |
627 | enc.indent = 0; |
760 | enc.indent = 0; |
628 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
761 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
762 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
763 | : 0x110000UL; |
629 | |
764 | |
630 | SvPOK_only (enc.sv); |
765 | SvPOK_only (enc.sv); |
631 | encode_sv (&enc, scalar); |
766 | encode_sv (&enc, scalar); |
|
|
767 | encode_nl (&enc); |
632 | |
768 | |
633 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
769 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
634 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
770 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
635 | |
771 | |
636 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
772 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
… | |
… | |
654 | JSON json; |
790 | JSON json; |
655 | U32 depth; // recursion depth |
791 | U32 depth; // recursion depth |
656 | U32 maxdepth; // recursion depth limit |
792 | U32 maxdepth; // recursion depth limit |
657 | } dec_t; |
793 | } dec_t; |
658 | |
794 | |
659 | inline void |
795 | INLINE void |
|
|
796 | decode_comment (dec_t *dec) |
|
|
797 | { |
|
|
798 | // only '#'-style comments allowed a.t.m. |
|
|
799 | |
|
|
800 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
801 | ++dec->cur; |
|
|
802 | } |
|
|
803 | |
|
|
804 | INLINE void |
660 | decode_ws (dec_t *dec) |
805 | decode_ws (dec_t *dec) |
661 | { |
806 | { |
662 | for (;;) |
807 | for (;;) |
663 | { |
808 | { |
664 | char ch = *dec->cur; |
809 | char ch = *dec->cur; |
665 | |
810 | |
666 | if (ch > 0x20 |
811 | if (ch > 0x20) |
|
|
812 | { |
|
|
813 | if (expect_false (ch == '#')) |
|
|
814 | { |
|
|
815 | if (dec->json.flags & F_RELAXED) |
|
|
816 | decode_comment (dec); |
|
|
817 | else |
|
|
818 | break; |
|
|
819 | } |
|
|
820 | else |
|
|
821 | break; |
|
|
822 | } |
667 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
823 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
668 | break; |
824 | break; // parse error, but let higher level handle it, gives better error messages |
669 | |
825 | |
670 | ++dec->cur; |
826 | ++dec->cur; |
671 | } |
827 | } |
672 | } |
828 | } |
673 | |
829 | |
… | |
… | |
677 | if (*dec->cur != ch) \ |
833 | if (*dec->cur != ch) \ |
678 | ERR (# ch " expected"); \ |
834 | ERR (# ch " expected"); \ |
679 | ++dec->cur; \ |
835 | ++dec->cur; \ |
680 | SE |
836 | SE |
681 | |
837 | |
682 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
838 | #define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
683 | #define DEC_DEC_DEPTH --dec->depth |
839 | #define DEC_DEC_DEPTH --dec->depth |
684 | |
840 | |
685 | static SV *decode_sv (dec_t *dec); |
841 | static SV *decode_sv (dec_t *dec); |
686 | |
842 | |
687 | static signed char decode_hexdigit[256]; |
843 | static signed char decode_hexdigit[256]; |
… | |
… | |
779 | |
935 | |
780 | if (hi >= 0x80) |
936 | if (hi >= 0x80) |
781 | { |
937 | { |
782 | utf8 = 1; |
938 | utf8 = 1; |
783 | |
939 | |
784 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
940 | cur = encode_utf8 (cur, hi); |
785 | } |
941 | } |
786 | else |
942 | else |
787 | *cur++ = hi; |
943 | *cur++ = hi; |
788 | } |
944 | } |
789 | break; |
945 | break; |
… | |
… | |
791 | default: |
947 | default: |
792 | --dec_cur; |
948 | --dec_cur; |
793 | ERR ("illegal backslash escape sequence in string"); |
949 | ERR ("illegal backslash escape sequence in string"); |
794 | } |
950 | } |
795 | } |
951 | } |
796 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
952 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
797 | *cur++ = ch; |
953 | *cur++ = ch; |
798 | else if (ch >= 0x80) |
954 | else if (ch >= 0x80) |
799 | { |
955 | { |
800 | STRLEN clen; |
956 | STRLEN clen; |
801 | UV uch; |
|
|
802 | |
957 | |
803 | --dec_cur; |
958 | --dec_cur; |
804 | |
959 | |
805 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
960 | decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
806 | if (clen == (STRLEN)-1) |
961 | if (clen == (STRLEN)-1) |
807 | ERR ("malformed UTF-8 character in JSON string"); |
962 | ERR ("malformed UTF-8 character in JSON string"); |
808 | |
963 | |
809 | do |
964 | do |
810 | *cur++ = *dec_cur++; |
965 | *cur++ = *dec_cur++; |
… | |
… | |
827 | { |
982 | { |
828 | STRLEN len = cur - buf; |
983 | STRLEN len = cur - buf; |
829 | |
984 | |
830 | if (sv) |
985 | if (sv) |
831 | { |
986 | { |
832 | SvGROW (sv, SvCUR (sv) + len + 1); |
987 | STRLEN cur = SvCUR (sv); |
|
|
988 | |
|
|
989 | if (SvLEN (sv) <= cur + len) |
|
|
990 | SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
991 | |
833 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
992 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
834 | SvCUR_set (sv, SvCUR (sv) + len); |
993 | SvCUR_set (sv, SvCUR (sv) + len); |
835 | } |
994 | } |
836 | else |
995 | else |
837 | sv = newSVpvn (buf, len); |
996 | sv = newSVpvn (buf, len); |
… | |
… | |
922 | is_nv = 1; |
1081 | is_nv = 1; |
923 | } |
1082 | } |
924 | |
1083 | |
925 | if (!is_nv) |
1084 | if (!is_nv) |
926 | { |
1085 | { |
|
|
1086 | int len = dec->cur - start; |
|
|
1087 | |
927 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1088 | // special case the rather common 1..5-digit-int case |
928 | if (*start == '-') |
1089 | if (*start == '-') |
929 | switch (dec->cur - start) |
1090 | switch (len) |
930 | { |
1091 | { |
931 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1092 | case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
932 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1093 | case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
933 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1094 | case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
934 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1095 | case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1096 | case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
935 | } |
1097 | } |
936 | else |
1098 | else |
937 | switch (dec->cur - start) |
1099 | switch (len) |
938 | { |
1100 | { |
939 | case 1: return newSViv ( start [0] - '0' * 1); |
1101 | case 1: return newSViv ( start [0] - '0' * 1); |
940 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1102 | 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); |
1103 | 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); |
1104 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1105 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
943 | } |
1106 | } |
944 | |
1107 | |
945 | { |
1108 | { |
946 | UV uv; |
1109 | UV uv; |
947 | int numtype = grok_number (start, dec->cur - start, &uv); |
1110 | int numtype = grok_number (start, len, &uv); |
948 | if (numtype & IS_NUMBER_IN_UV) |
1111 | if (numtype & IS_NUMBER_IN_UV) |
949 | if (numtype & IS_NUMBER_NEG) |
1112 | if (numtype & IS_NUMBER_NEG) |
950 | { |
1113 | { |
951 | if (uv < (UV)IV_MIN) |
1114 | if (uv < (UV)IV_MIN) |
952 | return newSViv (-(IV)uv); |
1115 | return newSViv (-(IV)uv); |
953 | } |
1116 | } |
954 | else |
1117 | else |
955 | return newSVuv (uv); |
1118 | return newSVuv (uv); |
956 | |
|
|
957 | // here would likely be the place for bigint support |
|
|
958 | } |
1119 | } |
959 | } |
|
|
960 | |
1120 | |
961 | // if we ever support bigint or bigfloat, this is the place for bigfloat |
1121 | len -= *start == '-' ? 1 : 0; |
|
|
1122 | |
|
|
1123 | // does not fit into IV or UV, try NV |
|
|
1124 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
|
|
1125 | #if defined (LDBL_DIG) |
|
|
1126 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
1127 | #endif |
|
|
1128 | ) |
|
|
1129 | // fits into NV without loss of precision |
|
|
1130 | return newSVnv (Atof (start)); |
|
|
1131 | |
|
|
1132 | // everything else fails, convert it to a string |
|
|
1133 | return newSVpvn (start, dec->cur - start); |
|
|
1134 | } |
|
|
1135 | |
|
|
1136 | // loss of precision here |
962 | return newSVnv (Atof (start)); |
1137 | return newSVnv (Atof (start)); |
963 | |
1138 | |
964 | fail: |
1139 | fail: |
965 | return 0; |
1140 | return 0; |
966 | } |
1141 | } |
… | |
… | |
996 | |
1171 | |
997 | if (*dec->cur != ',') |
1172 | if (*dec->cur != ',') |
998 | ERR (", or ] expected while parsing array"); |
1173 | ERR (", or ] expected while parsing array"); |
999 | |
1174 | |
1000 | ++dec->cur; |
1175 | ++dec->cur; |
|
|
1176 | |
|
|
1177 | decode_ws (dec); |
|
|
1178 | |
|
|
1179 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1180 | { |
|
|
1181 | ++dec->cur; |
|
|
1182 | break; |
|
|
1183 | } |
1001 | } |
1184 | } |
1002 | |
1185 | |
1003 | DEC_DEC_DEPTH; |
1186 | DEC_DEC_DEPTH; |
1004 | return newRV_noinc ((SV *)av); |
1187 | return newRV_noinc ((SV *)av); |
1005 | |
1188 | |
… | |
… | |
1021 | if (*dec->cur == '}') |
1204 | if (*dec->cur == '}') |
1022 | ++dec->cur; |
1205 | ++dec->cur; |
1023 | else |
1206 | else |
1024 | for (;;) |
1207 | for (;;) |
1025 | { |
1208 | { |
1026 | decode_ws (dec); EXPECT_CH ('"'); |
1209 | EXPECT_CH ('"'); |
1027 | |
1210 | |
1028 | // heuristic: assume that |
1211 | // heuristic: assume that |
1029 | // a) decode_str + hv_store_ent are abysmally slow. |
1212 | // a) decode_str + hv_store_ent are abysmally slow. |
1030 | // b) most hash keys are short, simple ascii text. |
1213 | // b) most hash keys are short, simple ascii text. |
1031 | // => try to "fast-match" such strings to avoid |
1214 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1035 | char *p = dec->cur; |
1218 | char *p = dec->cur; |
1036 | char *e = p + 24; // only try up to 24 bytes |
1219 | char *e = p + 24; // only try up to 24 bytes |
1037 | |
1220 | |
1038 | for (;;) |
1221 | for (;;) |
1039 | { |
1222 | { |
1040 | // the >= 0x80 is true on most architectures |
1223 | // the >= 0x80 is false on most architectures |
1041 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1224 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1042 | { |
1225 | { |
1043 | // slow path, back up and use decode_str |
1226 | // slow path, back up and use decode_str |
1044 | SV *key = decode_str (dec); |
1227 | SV *key = decode_str (dec); |
1045 | if (!key) |
1228 | if (!key) |
1046 | goto fail; |
1229 | goto fail; |
1047 | |
1230 | |
1048 | decode_ws (dec); EXPECT_CH (':'); |
1231 | decode_ws (dec); EXPECT_CH (':'); |
1049 | |
1232 | |
|
|
1233 | decode_ws (dec); |
1050 | value = decode_sv (dec); |
1234 | value = decode_sv (dec); |
1051 | if (!value) |
1235 | if (!value) |
1052 | { |
1236 | { |
1053 | SvREFCNT_dec (key); |
1237 | SvREFCNT_dec (key); |
1054 | goto fail; |
1238 | goto fail; |
… | |
… | |
1066 | int len = p - key; |
1250 | int len = p - key; |
1067 | dec->cur = p + 1; |
1251 | dec->cur = p + 1; |
1068 | |
1252 | |
1069 | decode_ws (dec); EXPECT_CH (':'); |
1253 | decode_ws (dec); EXPECT_CH (':'); |
1070 | |
1254 | |
|
|
1255 | decode_ws (dec); |
1071 | value = decode_sv (dec); |
1256 | value = decode_sv (dec); |
1072 | if (!value) |
1257 | if (!value) |
1073 | goto fail; |
1258 | goto fail; |
1074 | |
1259 | |
1075 | hv_store (hv, key, len, value, 0); |
1260 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1091 | |
1276 | |
1092 | if (*dec->cur != ',') |
1277 | if (*dec->cur != ',') |
1093 | ERR (", or } expected while parsing object/hash"); |
1278 | ERR (", or } expected while parsing object/hash"); |
1094 | |
1279 | |
1095 | ++dec->cur; |
1280 | ++dec->cur; |
|
|
1281 | |
|
|
1282 | decode_ws (dec); |
|
|
1283 | |
|
|
1284 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1285 | { |
|
|
1286 | ++dec->cur; |
|
|
1287 | break; |
|
|
1288 | } |
1096 | } |
1289 | } |
1097 | |
1290 | |
1098 | DEC_DEC_DEPTH; |
1291 | DEC_DEC_DEPTH; |
1099 | sv = newRV_noinc ((SV *)hv); |
1292 | sv = newRV_noinc ((SV *)hv); |
1100 | |
1293 | |
… | |
… | |
1113 | // might want to optimise this for common cases. |
1306 | // might want to optimise this for common cases. |
1114 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1307 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1115 | |
1308 | |
1116 | if (cb) |
1309 | if (cb) |
1117 | { |
1310 | { |
|
|
1311 | dSP; |
1118 | int count; |
1312 | int count; |
1119 | ENTER; SAVETMPS; |
|
|
1120 | |
1313 | |
1121 | dSP; PUSHMARK (SP); |
1314 | ENTER; SAVETMPS; PUSHMARK (SP); |
1122 | XPUSHs (HeVAL (he)); |
1315 | XPUSHs (HeVAL (he)); |
1123 | |
1316 | |
1124 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1317 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1125 | |
1318 | |
1126 | if (count == 1) |
1319 | if (count == 1) |
… | |
… | |
1134 | } |
1327 | } |
1135 | } |
1328 | } |
1136 | |
1329 | |
1137 | if (dec->json.cb_object) |
1330 | if (dec->json.cb_object) |
1138 | { |
1331 | { |
|
|
1332 | dSP; |
1139 | int count; |
1333 | int count; |
1140 | ENTER; SAVETMPS; |
|
|
1141 | |
1334 | |
1142 | dSP; ENTER; SAVETMPS; PUSHMARK (SP); |
1335 | ENTER; SAVETMPS; PUSHMARK (SP); |
1143 | XPUSHs (sv_2mortal (sv)); |
1336 | XPUSHs (sv_2mortal (sv)); |
1144 | |
1337 | |
1145 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1338 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1146 | |
1339 | |
1147 | if (count == 1) |
1340 | if (count == 1) |
… | |
… | |
1165 | } |
1358 | } |
1166 | |
1359 | |
1167 | static SV * |
1360 | static SV * |
1168 | decode_sv (dec_t *dec) |
1361 | decode_sv (dec_t *dec) |
1169 | { |
1362 | { |
1170 | decode_ws (dec); |
|
|
1171 | |
|
|
1172 | // the beauty of JSON: you need exactly one character lookahead |
1363 | // the beauty of JSON: you need exactly one character lookahead |
1173 | // to parse anything. |
1364 | // to parse everything. |
1174 | switch (*dec->cur) |
1365 | switch (*dec->cur) |
1175 | { |
1366 | { |
1176 | case '"': ++dec->cur; return decode_str (dec); |
1367 | case '"': ++dec->cur; return decode_str (dec); |
1177 | case '[': ++dec->cur; return decode_av (dec); |
1368 | case '[': ++dec->cur; return decode_av (dec); |
1178 | case '{': ++dec->cur; return decode_hv (dec); |
1369 | case '{': ++dec->cur; return decode_hv (dec); |
1179 | |
1370 | |
1180 | case '-': |
1371 | case '-': |
1181 | case '0': case '1': case '2': case '3': case '4': |
1372 | case '0': case '1': case '2': case '3': case '4': |
1182 | case '5': case '6': case '7': case '8': case '9': |
1373 | case '5': case '6': case '7': case '8': case '9': |
1183 | return decode_num (dec); |
1374 | return decode_num (dec); |
1184 | |
1375 | |
1185 | case 't': |
1376 | case 't': |
1186 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1377 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1187 | { |
1378 | { |
1188 | dec->cur += 4; |
1379 | dec->cur += 4; |
|
|
1380 | #if JSON_SLOW |
|
|
1381 | json_true = get_bool ("JSON::XS::true"); |
|
|
1382 | #endif |
1189 | return SvREFCNT_inc (json_true); |
1383 | return newSVsv (json_true); |
1190 | } |
1384 | } |
1191 | else |
1385 | else |
1192 | ERR ("'true' expected"); |
1386 | ERR ("'true' expected"); |
1193 | |
1387 | |
1194 | break; |
1388 | break; |
1195 | |
1389 | |
1196 | case 'f': |
1390 | case 'f': |
1197 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1391 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1198 | { |
1392 | { |
1199 | dec->cur += 5; |
1393 | dec->cur += 5; |
|
|
1394 | #if JSON_SLOW |
|
|
1395 | json_false = get_bool ("JSON::XS::false"); |
|
|
1396 | #endif |
1200 | return SvREFCNT_inc (json_false); |
1397 | return newSVsv (json_false); |
1201 | } |
1398 | } |
1202 | else |
1399 | else |
1203 | ERR ("'false' expected"); |
1400 | ERR ("'false' expected"); |
1204 | |
1401 | |
1205 | break; |
1402 | break; |
… | |
… | |
1223 | fail: |
1420 | fail: |
1224 | return 0; |
1421 | return 0; |
1225 | } |
1422 | } |
1226 | |
1423 | |
1227 | static SV * |
1424 | static SV * |
1228 | decode_json (SV *string, JSON *json, UV *offset_return) |
1425 | decode_json (SV *string, JSON *json, char **offset_return) |
1229 | { |
1426 | { |
1230 | dec_t dec; |
1427 | dec_t dec; |
1231 | UV offset; |
|
|
1232 | SV *sv; |
1428 | SV *sv; |
1233 | |
1429 | |
|
|
1430 | /* work around bugs in 5.10 where manipulating magic values |
|
|
1431 | * will perl ignore the magic in subsequent accesses |
|
|
1432 | */ |
1234 | SvGETMAGIC (string); |
1433 | /*SvGETMAGIC (string);*/ |
|
|
1434 | if (SvMAGICAL (string)) |
|
|
1435 | string = sv_2mortal (newSVsv (string)); |
|
|
1436 | |
1235 | SvUPGRADE (string, SVt_PV); |
1437 | SvUPGRADE (string, SVt_PV); |
1236 | |
1438 | |
1237 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1439 | /* work around a bug in perl 5.10, which causes SvCUR to fail an |
|
|
1440 | * assertion with -DDEBUGGING, although SvCUR is documented to |
|
|
1441 | * return the xpv_cur field which certainly exists after upgrading. |
|
|
1442 | * according to nicholas clark, calling SvPOK fixes this. |
|
|
1443 | * But it doesn't fix it, so try another workaround, call SvPV_nolen |
|
|
1444 | * and hope for the best. |
|
|
1445 | * Damnit, SvPV_nolen still trips over yet another assertion. This |
|
|
1446 | * assertion business is seriously broken, try yet another workaround |
|
|
1447 | * for the broken -DDEBUGGING. |
|
|
1448 | */ |
|
|
1449 | { |
|
|
1450 | #ifdef DEBUGGING |
|
|
1451 | STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
|
|
1452 | #else |
|
|
1453 | STRLEN offset = SvCUR (string); |
|
|
1454 | #endif |
|
|
1455 | |
|
|
1456 | if (offset > json->max_size && json->max_size) |
1238 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1457 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1239 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1458 | (unsigned long)SvCUR (string), (unsigned long)json->max_size); |
|
|
1459 | } |
1240 | |
1460 | |
1241 | if (json->flags & F_UTF8) |
1461 | if (json->flags & F_UTF8) |
1242 | sv_utf8_downgrade (string, 0); |
1462 | sv_utf8_downgrade (string, 0); |
1243 | else |
1463 | else |
1244 | sv_utf8_upgrade (string); |
1464 | sv_utf8_upgrade (string); |
1245 | |
1465 | |
1246 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1466 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1247 | |
1467 | |
1248 | dec.json = *json; |
1468 | dec.json = *json; |
1249 | dec.cur = SvPVX (string); |
1469 | dec.cur = SvPVX (string); |
1250 | dec.end = SvEND (string); |
1470 | dec.end = SvEND (string); |
1251 | dec.err = 0; |
1471 | dec.err = 0; |
1252 | dec.depth = 0; |
1472 | dec.depth = 0; |
1253 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
|
|
1254 | |
1473 | |
1255 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1474 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1256 | dec.json.flags |= F_HOOK; |
1475 | dec.json.flags |= F_HOOK; |
1257 | |
1476 | |
1258 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1477 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1478 | |
|
|
1479 | decode_ws (&dec); |
1259 | sv = decode_sv (&dec); |
1480 | sv = decode_sv (&dec); |
|
|
1481 | |
|
|
1482 | if (offset_return) |
|
|
1483 | *offset_return = dec.cur; |
1260 | |
1484 | |
1261 | if (!(offset_return || !sv)) |
1485 | if (!(offset_return || !sv)) |
1262 | { |
1486 | { |
1263 | // check for trailing garbage |
1487 | // check for trailing garbage |
1264 | decode_ws (&dec); |
1488 | decode_ws (&dec); |
… | |
… | |
1267 | { |
1491 | { |
1268 | dec.err = "garbage after JSON object"; |
1492 | dec.err = "garbage after JSON object"; |
1269 | SvREFCNT_dec (sv); |
1493 | SvREFCNT_dec (sv); |
1270 | sv = 0; |
1494 | sv = 0; |
1271 | } |
1495 | } |
1272 | } |
|
|
1273 | |
|
|
1274 | if (offset_return || !sv) |
|
|
1275 | { |
|
|
1276 | offset = dec.json.flags & F_UTF8 |
|
|
1277 | ? dec.cur - SvPVX (string) |
|
|
1278 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1279 | |
|
|
1280 | if (offset_return) |
|
|
1281 | *offset_return = offset; |
|
|
1282 | } |
1496 | } |
1283 | |
1497 | |
1284 | if (!sv) |
1498 | if (!sv) |
1285 | { |
1499 | { |
1286 | SV *uni = sv_newmortal (); |
1500 | SV *uni = sv_newmortal (); |
… | |
… | |
1292 | SAVEVPTR (PL_curcop); |
1506 | SAVEVPTR (PL_curcop); |
1293 | PL_curcop = &cop; |
1507 | PL_curcop = &cop; |
1294 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1508 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1295 | LEAVE; |
1509 | LEAVE; |
1296 | |
1510 | |
1297 | croak ("%s, at character offset %d [\"%s\"]", |
1511 | croak ("%s, at character offset %d (before \"%s\")", |
1298 | dec.err, |
1512 | dec.err, |
1299 | (int)offset, |
1513 | ptr_to_index (string, dec.cur), |
1300 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1514 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1301 | } |
1515 | } |
1302 | |
1516 | |
1303 | sv = sv_2mortal (sv); |
1517 | sv = sv_2mortal (sv); |
1304 | |
1518 | |
1305 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1519 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1306 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1520 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1307 | |
1521 | |
1308 | return sv; |
1522 | return sv; |
|
|
1523 | } |
|
|
1524 | |
|
|
1525 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1526 | // incremental parser |
|
|
1527 | |
|
|
1528 | static void |
|
|
1529 | incr_parse (JSON *self) |
|
|
1530 | { |
|
|
1531 | const char *p = SvPVX (self->incr_text) + self->incr_pos; |
|
|
1532 | |
|
|
1533 | // the state machine here is a bit convoluted and could be simplified a lot |
|
|
1534 | // but this would make it slower, so... |
|
|
1535 | |
|
|
1536 | for (;;) |
|
|
1537 | { |
|
|
1538 | //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 |
|
|
1539 | switch (self->incr_mode) |
|
|
1540 | { |
|
|
1541 | // only used for initial whitespace skipping |
|
|
1542 | case INCR_M_WS: |
|
|
1543 | for (;;) |
|
|
1544 | { |
|
|
1545 | if (*p > 0x20) |
|
|
1546 | { |
|
|
1547 | if (*p == '#') |
|
|
1548 | { |
|
|
1549 | self->incr_mode = INCR_M_C0; |
|
|
1550 | goto incr_m_c; |
|
|
1551 | } |
|
|
1552 | else |
|
|
1553 | { |
|
|
1554 | self->incr_mode = INCR_M_JSON; |
|
|
1555 | goto incr_m_json; |
|
|
1556 | } |
|
|
1557 | } |
|
|
1558 | else if (!*p) |
|
|
1559 | goto interrupt; |
|
|
1560 | |
|
|
1561 | ++p; |
|
|
1562 | } |
|
|
1563 | |
|
|
1564 | // skip a single char inside a string (for \\-processing) |
|
|
1565 | case INCR_M_BS: |
|
|
1566 | if (!*p) |
|
|
1567 | goto interrupt; |
|
|
1568 | |
|
|
1569 | ++p; |
|
|
1570 | self->incr_mode = INCR_M_STR; |
|
|
1571 | goto incr_m_str; |
|
|
1572 | |
|
|
1573 | // inside #-style comments |
|
|
1574 | case INCR_M_C0: |
|
|
1575 | case INCR_M_C1: |
|
|
1576 | incr_m_c: |
|
|
1577 | for (;;) |
|
|
1578 | { |
|
|
1579 | if (*p == '\n') |
|
|
1580 | { |
|
|
1581 | self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
|
|
1582 | break; |
|
|
1583 | } |
|
|
1584 | else if (!*p) |
|
|
1585 | goto interrupt; |
|
|
1586 | |
|
|
1587 | ++p; |
|
|
1588 | } |
|
|
1589 | |
|
|
1590 | break; |
|
|
1591 | |
|
|
1592 | // inside a string |
|
|
1593 | case INCR_M_STR: |
|
|
1594 | incr_m_str: |
|
|
1595 | for (;;) |
|
|
1596 | { |
|
|
1597 | if (*p == '"') |
|
|
1598 | { |
|
|
1599 | ++p; |
|
|
1600 | self->incr_mode = INCR_M_JSON; |
|
|
1601 | |
|
|
1602 | if (!self->incr_nest) |
|
|
1603 | goto interrupt; |
|
|
1604 | |
|
|
1605 | goto incr_m_json; |
|
|
1606 | } |
|
|
1607 | else if (*p == '\\') |
|
|
1608 | { |
|
|
1609 | ++p; // "virtually" consumes character after \ |
|
|
1610 | |
|
|
1611 | if (!*p) // if at end of string we have to switch modes |
|
|
1612 | { |
|
|
1613 | self->incr_mode = INCR_M_BS; |
|
|
1614 | goto interrupt; |
|
|
1615 | } |
|
|
1616 | } |
|
|
1617 | else if (!*p) |
|
|
1618 | goto interrupt; |
|
|
1619 | |
|
|
1620 | ++p; |
|
|
1621 | } |
|
|
1622 | |
|
|
1623 | // after initial ws, outside string |
|
|
1624 | case INCR_M_JSON: |
|
|
1625 | incr_m_json: |
|
|
1626 | for (;;) |
|
|
1627 | { |
|
|
1628 | switch (*p++) |
|
|
1629 | { |
|
|
1630 | case 0: |
|
|
1631 | --p; |
|
|
1632 | goto interrupt; |
|
|
1633 | |
|
|
1634 | case 0x09: |
|
|
1635 | case 0x0a: |
|
|
1636 | case 0x0d: |
|
|
1637 | case 0x20: |
|
|
1638 | if (!self->incr_nest) |
|
|
1639 | { |
|
|
1640 | --p; // do not eat the whitespace, let the next round do it |
|
|
1641 | goto interrupt; |
|
|
1642 | } |
|
|
1643 | break; |
|
|
1644 | |
|
|
1645 | case '"': |
|
|
1646 | self->incr_mode = INCR_M_STR; |
|
|
1647 | goto incr_m_str; |
|
|
1648 | |
|
|
1649 | case '[': |
|
|
1650 | case '{': |
|
|
1651 | if (++self->incr_nest > self->max_depth) |
|
|
1652 | croak (ERR_NESTING_EXCEEDED); |
|
|
1653 | break; |
|
|
1654 | |
|
|
1655 | case ']': |
|
|
1656 | case '}': |
|
|
1657 | if (--self->incr_nest <= 0) |
|
|
1658 | goto interrupt; |
|
|
1659 | break; |
|
|
1660 | |
|
|
1661 | case '#': |
|
|
1662 | self->incr_mode = INCR_M_C1; |
|
|
1663 | goto incr_m_c; |
|
|
1664 | } |
|
|
1665 | } |
|
|
1666 | } |
|
|
1667 | |
|
|
1668 | modechange: |
|
|
1669 | ; |
|
|
1670 | } |
|
|
1671 | |
|
|
1672 | interrupt: |
|
|
1673 | self->incr_pos = p - SvPVX (self->incr_text); |
|
|
1674 | //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
|
|
1675 | //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1309 | } |
1676 | } |
1310 | |
1677 | |
1311 | ///////////////////////////////////////////////////////////////////////////// |
1678 | ///////////////////////////////////////////////////////////////////////////// |
1312 | // XS interface functions |
1679 | // XS interface functions |
1313 | |
1680 | |
… | |
… | |
1325 | : -1; |
1692 | : -1; |
1326 | |
1693 | |
1327 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1694 | json_stash = gv_stashpv ("JSON::XS" , 1); |
1328 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1695 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1329 | |
1696 | |
1330 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1697 | json_true = get_bool ("JSON::XS::true"); |
1331 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1698 | json_false = get_bool ("JSON::XS::false"); |
|
|
1699 | |
|
|
1700 | CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1332 | } |
1701 | } |
1333 | |
1702 | |
1334 | PROTOTYPES: DISABLE |
1703 | PROTOTYPES: DISABLE |
|
|
1704 | |
|
|
1705 | void CLONE (...) |
|
|
1706 | CODE: |
|
|
1707 | json_stash = 0; |
|
|
1708 | json_boolean_stash = 0; |
1335 | |
1709 | |
1336 | void new (char *klass) |
1710 | void new (char *klass) |
1337 | PPCODE: |
1711 | PPCODE: |
1338 | { |
1712 | { |
1339 | SV *pv = NEWSV (0, sizeof (JSON)); |
1713 | SV *pv = NEWSV (0, sizeof (JSON)); |
1340 | SvPOK_only (pv); |
1714 | SvPOK_only (pv); |
1341 | Zero (SvPVX (pv), 1, JSON); |
1715 | json_init ((JSON *)SvPVX (pv)); |
1342 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1716 | XPUSHs (sv_2mortal (sv_bless ( |
1343 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash))); |
1717 | newRV_noinc (pv), |
|
|
1718 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1719 | ))); |
1344 | } |
1720 | } |
1345 | |
1721 | |
1346 | void ascii (JSON *self, int enable = 1) |
1722 | void ascii (JSON *self, int enable = 1) |
1347 | ALIAS: |
1723 | ALIAS: |
1348 | ascii = F_ASCII |
1724 | ascii = F_ASCII |
… | |
… | |
1355 | pretty = F_PRETTY |
1731 | pretty = F_PRETTY |
1356 | allow_nonref = F_ALLOW_NONREF |
1732 | allow_nonref = F_ALLOW_NONREF |
1357 | shrink = F_SHRINK |
1733 | shrink = F_SHRINK |
1358 | allow_blessed = F_ALLOW_BLESSED |
1734 | allow_blessed = F_ALLOW_BLESSED |
1359 | convert_blessed = F_CONV_BLESSED |
1735 | convert_blessed = F_CONV_BLESSED |
|
|
1736 | relaxed = F_RELAXED |
|
|
1737 | allow_unknown = F_ALLOW_UNKNOWN |
1360 | PPCODE: |
1738 | PPCODE: |
1361 | { |
1739 | { |
1362 | if (enable) |
1740 | if (enable) |
1363 | self->flags |= ix; |
1741 | self->flags |= ix; |
1364 | else |
1742 | else |
1365 | self->flags &= ~ix; |
1743 | self->flags &= ~ix; |
1366 | |
1744 | |
1367 | XPUSHs (ST (0)); |
1745 | XPUSHs (ST (0)); |
1368 | } |
1746 | } |
1369 | |
1747 | |
1370 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1748 | void get_ascii (JSON *self) |
|
|
1749 | ALIAS: |
|
|
1750 | get_ascii = F_ASCII |
|
|
1751 | get_latin1 = F_LATIN1 |
|
|
1752 | get_utf8 = F_UTF8 |
|
|
1753 | get_indent = F_INDENT |
|
|
1754 | get_canonical = F_CANONICAL |
|
|
1755 | get_space_before = F_SPACE_BEFORE |
|
|
1756 | get_space_after = F_SPACE_AFTER |
|
|
1757 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1758 | get_shrink = F_SHRINK |
|
|
1759 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1760 | get_convert_blessed = F_CONV_BLESSED |
|
|
1761 | get_relaxed = F_RELAXED |
|
|
1762 | get_allow_unknown = F_ALLOW_UNKNOWN |
1371 | PPCODE: |
1763 | PPCODE: |
1372 | { |
1764 | XPUSHs (boolSV (self->flags & ix)); |
1373 | UV log2 = 0; |
|
|
1374 | |
1765 | |
1375 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1766 | void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1376 | |
1767 | PPCODE: |
1377 | while ((1UL << log2) < max_depth) |
1768 | self->max_depth = max_depth; |
1378 | ++log2; |
|
|
1379 | |
|
|
1380 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1381 | |
|
|
1382 | XPUSHs (ST (0)); |
1769 | XPUSHs (ST (0)); |
1383 | } |
|
|
1384 | |
1770 | |
|
|
1771 | U32 get_max_depth (JSON *self) |
|
|
1772 | CODE: |
|
|
1773 | RETVAL = self->max_depth; |
|
|
1774 | OUTPUT: |
|
|
1775 | RETVAL |
|
|
1776 | |
1385 | void max_size (JSON *self, UV max_size = 0) |
1777 | void max_size (JSON *self, U32 max_size = 0) |
1386 | PPCODE: |
1778 | PPCODE: |
1387 | { |
1779 | self->max_size = max_size; |
1388 | UV log2 = 0; |
|
|
1389 | |
|
|
1390 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1391 | if (max_size == 1) max_size = 2; |
|
|
1392 | |
|
|
1393 | while ((1UL << log2) < max_size) |
|
|
1394 | ++log2; |
|
|
1395 | |
|
|
1396 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1397 | |
|
|
1398 | XPUSHs (ST (0)); |
1780 | XPUSHs (ST (0)); |
1399 | } |
1781 | |
|
|
1782 | int get_max_size (JSON *self) |
|
|
1783 | CODE: |
|
|
1784 | RETVAL = self->max_size; |
|
|
1785 | OUTPUT: |
|
|
1786 | RETVAL |
1400 | |
1787 | |
1401 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1788 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1402 | PPCODE: |
1789 | PPCODE: |
1403 | { |
1790 | { |
1404 | SvREFCNT_dec (self->cb_object); |
1791 | SvREFCNT_dec (self->cb_object); |
… | |
… | |
1438 | XPUSHs (decode_json (jsonstr, self, 0)); |
1825 | XPUSHs (decode_json (jsonstr, self, 0)); |
1439 | |
1826 | |
1440 | void decode_prefix (JSON *self, SV *jsonstr) |
1827 | void decode_prefix (JSON *self, SV *jsonstr) |
1441 | PPCODE: |
1828 | PPCODE: |
1442 | { |
1829 | { |
1443 | UV offset; |
1830 | char *offset; |
1444 | EXTEND (SP, 2); |
1831 | EXTEND (SP, 2); |
1445 | PUSHs (decode_json (jsonstr, self, &offset)); |
1832 | PUSHs (decode_json (jsonstr, self, &offset)); |
1446 | PUSHs (sv_2mortal (newSVuv (offset))); |
1833 | PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
|
|
1834 | } |
|
|
1835 | |
|
|
1836 | void incr_parse (JSON *self, SV *jsonstr = 0) |
|
|
1837 | PPCODE: |
|
|
1838 | { |
|
|
1839 | if (!self->incr_text) |
|
|
1840 | self->incr_text = newSVpvn ("", 0); |
|
|
1841 | |
|
|
1842 | // append data, if any |
|
|
1843 | if (jsonstr) |
|
|
1844 | { |
|
|
1845 | if (SvUTF8 (jsonstr)) |
|
|
1846 | { |
|
|
1847 | if (!SvUTF8 (self->incr_text)) |
|
|
1848 | { |
|
|
1849 | /* utf-8-ness differs, need to upgrade */ |
|
|
1850 | sv_utf8_upgrade (self->incr_text); |
|
|
1851 | |
|
|
1852 | if (self->incr_pos) |
|
|
1853 | self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
|
|
1854 | - (U8 *)SvPVX (self->incr_text); |
|
|
1855 | } |
|
|
1856 | } |
|
|
1857 | else if (SvUTF8 (self->incr_text)) |
|
|
1858 | sv_utf8_upgrade (jsonstr); |
|
|
1859 | |
|
|
1860 | { |
|
|
1861 | STRLEN len; |
|
|
1862 | const char *str = SvPV (jsonstr, len); |
|
|
1863 | STRLEN cur = SvCUR (self->incr_text); |
|
|
1864 | |
|
|
1865 | if (SvLEN (self->incr_text) <= cur + len) |
|
|
1866 | SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
|
|
1867 | |
|
|
1868 | Move (str, SvEND (self->incr_text), len, char); |
|
|
1869 | SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
|
|
1870 | *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1871 | } |
|
|
1872 | } |
|
|
1873 | |
|
|
1874 | if (GIMME_V != G_VOID) |
|
|
1875 | do |
|
|
1876 | { |
|
|
1877 | char *offset; |
|
|
1878 | |
|
|
1879 | if (!INCR_DONE (self)) |
|
|
1880 | { |
|
|
1881 | incr_parse (self); |
|
|
1882 | |
|
|
1883 | if (self->incr_pos > self->max_size && self->max_size) |
|
|
1884 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1885 | (unsigned long)self->incr_pos, (unsigned long)self->max_size); |
|
|
1886 | |
|
|
1887 | if (!INCR_DONE (self)) |
|
|
1888 | break; |
|
|
1889 | } |
|
|
1890 | |
|
|
1891 | XPUSHs (decode_json (self->incr_text, self, &offset)); |
|
|
1892 | |
|
|
1893 | self->incr_pos -= offset - SvPVX (self->incr_text); |
|
|
1894 | self->incr_nest = 0; |
|
|
1895 | self->incr_mode = 0; |
|
|
1896 | |
|
|
1897 | sv_chop (self->incr_text, offset); |
|
|
1898 | } |
|
|
1899 | while (GIMME_V == G_ARRAY); |
|
|
1900 | } |
|
|
1901 | |
|
|
1902 | SV *incr_text (JSON *self) |
|
|
1903 | ATTRS: lvalue |
|
|
1904 | CODE: |
|
|
1905 | { |
|
|
1906 | if (self->incr_pos) |
|
|
1907 | croak ("incr_text can not be called when the incremental parser already started parsing"); |
|
|
1908 | |
|
|
1909 | RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
|
|
1910 | } |
|
|
1911 | OUTPUT: |
|
|
1912 | RETVAL |
|
|
1913 | |
|
|
1914 | void incr_skip (JSON *self) |
|
|
1915 | CODE: |
|
|
1916 | { |
|
|
1917 | if (self->incr_pos) |
|
|
1918 | { |
|
|
1919 | sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
|
|
1920 | self->incr_pos = 0; |
|
|
1921 | self->incr_nest = 0; |
|
|
1922 | self->incr_mode = 0; |
|
|
1923 | } |
|
|
1924 | } |
|
|
1925 | |
|
|
1926 | void incr_reset (JSON *self) |
|
|
1927 | CODE: |
|
|
1928 | { |
|
|
1929 | SvREFCNT_dec (self->incr_text); |
|
|
1930 | self->incr_text = 0; |
|
|
1931 | self->incr_pos = 0; |
|
|
1932 | self->incr_nest = 0; |
|
|
1933 | self->incr_mode = 0; |
1447 | } |
1934 | } |
1448 | |
1935 | |
1449 | void DESTROY (JSON *self) |
1936 | void DESTROY (JSON *self) |
1450 | CODE: |
1937 | CODE: |
1451 | SvREFCNT_dec (self->cb_sk_object); |
1938 | SvREFCNT_dec (self->cb_sk_object); |
1452 | SvREFCNT_dec (self->cb_object); |
1939 | SvREFCNT_dec (self->cb_object); |
|
|
1940 | SvREFCNT_dec (self->incr_text); |
1453 | |
1941 | |
1454 | PROTOTYPES: ENABLE |
1942 | PROTOTYPES: ENABLE |
1455 | |
1943 | |
1456 | void to_json (SV *scalar) |
1944 | void encode_json (SV *scalar) |
|
|
1945 | ALIAS: |
|
|
1946 | to_json_ = 0 |
|
|
1947 | encode_json = F_UTF8 |
1457 | PPCODE: |
1948 | PPCODE: |
1458 | { |
1949 | { |
1459 | JSON json = { F_DEFAULT | F_UTF8 }; |
1950 | JSON json; |
|
|
1951 | json_init (&json); |
|
|
1952 | json.flags |= ix; |
1460 | XPUSHs (encode_json (scalar, &json)); |
1953 | XPUSHs (encode_json (scalar, &json)); |
1461 | } |
1954 | } |
1462 | |
1955 | |
1463 | void from_json (SV *jsonstr) |
1956 | void decode_json (SV *jsonstr) |
|
|
1957 | ALIAS: |
|
|
1958 | from_json_ = 0 |
|
|
1959 | decode_json = F_UTF8 |
1464 | PPCODE: |
1960 | PPCODE: |
1465 | { |
1961 | { |
1466 | JSON json = { F_DEFAULT | F_UTF8 }; |
1962 | JSON json; |
|
|
1963 | json_init (&json); |
|
|
1964 | json.flags |= ix; |
1467 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1965 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1468 | } |
1966 | } |
1469 | |
1967 | |
|
|
1968 | |