… | |
… | |
27 | #define F_SPACE_AFTER 0x00000040UL |
27 | #define F_SPACE_AFTER 0x00000040UL |
28 | #define F_ALLOW_NONREF 0x00000100UL |
28 | #define F_ALLOW_NONREF 0x00000100UL |
29 | #define F_SHRINK 0x00000200UL |
29 | #define F_SHRINK 0x00000200UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
30 | #define F_ALLOW_BLESSED 0x00000400UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
31 | #define F_CONV_BLESSED 0x00000800UL |
|
|
32 | #define F_RELAXED 0x00001000UL |
|
|
33 | |
32 | #define F_MAXDEPTH 0xf8000000UL |
34 | #define F_MAXDEPTH 0xf8000000UL |
33 | #define S_MAXDEPTH 27 |
35 | #define S_MAXDEPTH 27 |
34 | #define F_MAXSIZE 0x01f00000UL |
36 | #define F_MAXSIZE 0x01f00000UL |
35 | #define S_MAXSIZE 20 |
37 | #define S_MAXSIZE 20 |
36 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
… | |
… | |
48 | |
50 | |
49 | #define SB do { |
51 | #define SB do { |
50 | #define SE } while (0) |
52 | #define SE } while (0) |
51 | |
53 | |
52 | #if __GNUC__ >= 3 |
54 | #if __GNUC__ >= 3 |
53 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
55 | # define expect(expr,value) __builtin_expect ((expr), (value)) |
54 | # define inline inline |
56 | # define INLINE static inline |
55 | #else |
57 | #else |
56 | # define expect(expr,value) (expr) |
58 | # define expect(expr,value) (expr) |
57 | # define inline static |
59 | # define INLINE static |
58 | #endif |
60 | #endif |
59 | |
61 | |
60 | #define expect_false(expr) expect ((expr) != 0, 0) |
62 | #define expect_false(expr) expect ((expr) != 0, 0) |
61 | #define expect_true(expr) expect ((expr) != 0, 1) |
63 | #define expect_true(expr) expect ((expr) != 0, 1) |
62 | |
64 | |
|
|
65 | #define IN_RANGE_INC(type,val,beg,end) \ |
|
|
66 | ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
|
|
67 | <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
|
|
68 | |
63 | #ifdef USE_ITHREADS |
69 | #ifdef USE_ITHREADS |
64 | # define JSON_SLOW 1 |
70 | # define JSON_SLOW 1 |
|
|
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
65 | #else |
72 | #else |
66 | # define JSON_SLOW 0 |
73 | # define JSON_SLOW 0 |
|
|
74 | # define JSON_STASH json_stash |
67 | #endif |
75 | #endif |
68 | |
76 | |
69 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
77 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
70 | static SV *json_true, *json_false; |
78 | static SV *json_true, *json_false; |
71 | |
79 | |
… | |
… | |
76 | } JSON; |
84 | } JSON; |
77 | |
85 | |
78 | ///////////////////////////////////////////////////////////////////////////// |
86 | ///////////////////////////////////////////////////////////////////////////// |
79 | // utility functions |
87 | // utility functions |
80 | |
88 | |
81 | inline void |
89 | INLINE void |
82 | shrink (SV *sv) |
90 | shrink (SV *sv) |
83 | { |
91 | { |
84 | sv_utf8_downgrade (sv, 1); |
92 | sv_utf8_downgrade (sv, 1); |
85 | if (SvLEN (sv) > SvCUR (sv) + 1) |
93 | if (SvLEN (sv) > SvCUR (sv) + 1) |
86 | { |
94 | { |
… | |
… | |
95 | // decode an utf-8 character and return it, or (UV)-1 in |
103 | // decode an utf-8 character and return it, or (UV)-1 in |
96 | // case of an error. |
104 | // case of an error. |
97 | // we special-case "safe" characters from U+80 .. U+7FF, |
105 | // we special-case "safe" characters from U+80 .. U+7FF, |
98 | // but use the very good perl function to parse anything else. |
106 | // but use the very good perl function to parse anything else. |
99 | // note that we never call this function for a ascii codepoints |
107 | // note that we never call this function for a ascii codepoints |
100 | inline UV |
108 | INLINE UV |
101 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
109 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
102 | { |
110 | { |
103 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
111 | if (expect_true (len >= 2 |
104 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
112 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
105 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
113 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
106 | { |
114 | { |
107 | *clen = 2; |
115 | *clen = 2; |
108 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
116 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
109 | } |
117 | } |
110 | else |
118 | else |
111 | { |
119 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
112 | *clen = (STRLEN)-1; |
120 | } |
113 | return (UV)-1; |
121 | |
|
|
122 | // likewise for encoding, also never called for ascii codepoints |
|
|
123 | // this function takes advantage of this fact, although current gccs |
|
|
124 | // seem to optimise the check for >= 0x80 away anyways |
|
|
125 | INLINE unsigned char * |
|
|
126 | encode_utf8 (unsigned char *s, UV ch) |
|
|
127 | { |
|
|
128 | if (ch <= 0x7FF) |
114 | } |
129 | { |
|
|
130 | *s++ = (ch >> 6) | 0xc0; |
|
|
131 | *s++ = (ch & 0x3f) | 0x80; |
|
|
132 | } |
|
|
133 | else |
|
|
134 | s = uvuni_to_utf8_flags (s, ch, 0); |
|
|
135 | |
|
|
136 | return s; |
115 | } |
137 | } |
116 | |
138 | |
117 | ///////////////////////////////////////////////////////////////////////////// |
139 | ///////////////////////////////////////////////////////////////////////////// |
118 | // encoder |
140 | // encoder |
119 | |
141 | |
… | |
… | |
124 | char *end; // SvEND (sv) |
146 | char *end; // SvEND (sv) |
125 | SV *sv; // result scalar |
147 | SV *sv; // result scalar |
126 | JSON json; |
148 | JSON json; |
127 | U32 indent; // indentation level |
149 | U32 indent; // indentation level |
128 | U32 maxdepth; // max. indentation/recursion level |
150 | U32 maxdepth; // max. indentation/recursion level |
|
|
151 | UV limit; // escape character values >= this value when encoding |
129 | } enc_t; |
152 | } enc_t; |
130 | |
153 | |
131 | inline void |
154 | INLINE void |
132 | need (enc_t *enc, STRLEN len) |
155 | need (enc_t *enc, STRLEN len) |
133 | { |
156 | { |
134 | if (expect_false (enc->cur + len >= enc->end)) |
157 | if (expect_false (enc->cur + len >= enc->end)) |
135 | { |
158 | { |
136 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
159 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
… | |
… | |
138 | enc->cur = SvPVX (enc->sv) + cur; |
161 | enc->cur = SvPVX (enc->sv) + cur; |
139 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
162 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
140 | } |
163 | } |
141 | } |
164 | } |
142 | |
165 | |
143 | inline void |
166 | INLINE void |
144 | encode_ch (enc_t *enc, char ch) |
167 | encode_ch (enc_t *enc, char ch) |
145 | { |
168 | { |
146 | need (enc, 1); |
169 | need (enc, 1); |
147 | *enc->cur++ = ch; |
170 | *enc->cur++ = ch; |
148 | } |
171 | } |
… | |
… | |
202 | { |
225 | { |
203 | uch = ch; |
226 | uch = ch; |
204 | clen = 1; |
227 | clen = 1; |
205 | } |
228 | } |
206 | |
229 | |
207 | if (uch > 0x10FFFFUL) |
230 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
208 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
209 | |
|
|
210 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
|
|
211 | { |
231 | { |
212 | if (uch > 0xFFFFUL) |
232 | if (uch > 0xFFFFUL) |
213 | { |
233 | { |
|
|
234 | if (uch > 0x10FFFFUL) |
|
|
235 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
236 | |
214 | need (enc, len += 11); |
237 | need (enc, len += 11); |
215 | sprintf (enc->cur, "\\u%04x\\u%04x", |
238 | sprintf (enc->cur, "\\u%04x\\u%04x", |
216 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
239 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
217 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
240 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
218 | enc->cur += 12; |
241 | enc->cur += 12; |
… | |
… | |
246 | while (--clen); |
269 | while (--clen); |
247 | } |
270 | } |
248 | else |
271 | else |
249 | { |
272 | { |
250 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
273 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
251 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
274 | enc->cur = encode_utf8 (enc->cur, uch); |
252 | ++str; |
275 | ++str; |
253 | } |
276 | } |
254 | } |
277 | } |
255 | } |
278 | } |
256 | } |
279 | } |
257 | |
280 | |
258 | --len; |
281 | --len; |
259 | } |
282 | } |
260 | } |
283 | } |
261 | |
284 | |
262 | inline void |
285 | INLINE void |
263 | encode_indent (enc_t *enc) |
286 | encode_indent (enc_t *enc) |
264 | { |
287 | { |
265 | if (enc->json.flags & F_INDENT) |
288 | if (enc->json.flags & F_INDENT) |
266 | { |
289 | { |
267 | int spaces = enc->indent * INDENT_STEP; |
290 | int spaces = enc->indent * INDENT_STEP; |
… | |
… | |
270 | memset (enc->cur, ' ', spaces); |
293 | memset (enc->cur, ' ', spaces); |
271 | enc->cur += spaces; |
294 | enc->cur += spaces; |
272 | } |
295 | } |
273 | } |
296 | } |
274 | |
297 | |
275 | inline void |
298 | INLINE void |
276 | encode_space (enc_t *enc) |
299 | encode_space (enc_t *enc) |
277 | { |
300 | { |
278 | need (enc, 1); |
301 | need (enc, 1); |
279 | encode_ch (enc, ' '); |
302 | encode_ch (enc, ' '); |
280 | } |
303 | } |
281 | |
304 | |
282 | inline void |
305 | INLINE void |
283 | encode_nl (enc_t *enc) |
306 | encode_nl (enc_t *enc) |
284 | { |
307 | { |
285 | if (enc->json.flags & F_INDENT) |
308 | if (enc->json.flags & F_INDENT) |
286 | { |
309 | { |
287 | need (enc, 1); |
310 | need (enc, 1); |
288 | encode_ch (enc, '\n'); |
311 | encode_ch (enc, '\n'); |
289 | } |
312 | } |
290 | } |
313 | } |
291 | |
314 | |
292 | inline void |
315 | INLINE void |
293 | encode_comma (enc_t *enc) |
316 | encode_comma (enc_t *enc) |
294 | { |
317 | { |
295 | encode_ch (enc, ','); |
318 | encode_ch (enc, ','); |
296 | |
319 | |
297 | if (enc->json.flags & F_INDENT) |
320 | if (enc->json.flags & F_INDENT) |
… | |
… | |
308 | int i, len = av_len (av); |
331 | int i, len = av_len (av); |
309 | |
332 | |
310 | if (enc->indent >= enc->maxdepth) |
333 | if (enc->indent >= enc->maxdepth) |
311 | croak ("data structure too deep (hit recursion limit)"); |
334 | croak ("data structure too deep (hit recursion limit)"); |
312 | |
335 | |
313 | encode_ch (enc, '['); encode_nl (enc); |
336 | encode_ch (enc, '['); |
314 | ++enc->indent; |
337 | |
|
|
338 | if (len >= 0) |
|
|
339 | { |
|
|
340 | encode_nl (enc); ++enc->indent; |
315 | |
341 | |
316 | for (i = 0; i <= len; ++i) |
342 | for (i = 0; i <= len; ++i) |
317 | { |
343 | { |
318 | SV **svp = av_fetch (av, i, 0); |
344 | SV **svp = av_fetch (av, i, 0); |
319 | |
345 | |
320 | encode_indent (enc); |
346 | encode_indent (enc); |
321 | |
347 | |
322 | if (svp) |
348 | if (svp) |
323 | encode_sv (enc, *svp); |
349 | encode_sv (enc, *svp); |
324 | else |
350 | else |
325 | encode_str (enc, "null", 4, 0); |
351 | encode_str (enc, "null", 4, 0); |
326 | |
352 | |
327 | if (i < len) |
353 | if (i < len) |
328 | encode_comma (enc); |
354 | encode_comma (enc); |
329 | } |
355 | } |
330 | |
356 | |
|
|
357 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
358 | } |
|
|
359 | |
331 | encode_nl (enc); |
360 | encode_ch (enc, ']'); |
332 | |
|
|
333 | --enc->indent; |
|
|
334 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
335 | } |
361 | } |
336 | |
362 | |
337 | static void |
363 | static void |
338 | encode_he (enc_t *enc, HE *he) |
364 | encode_hk (enc_t *enc, HE *he) |
339 | { |
365 | { |
340 | encode_ch (enc, '"'); |
366 | encode_ch (enc, '"'); |
341 | |
367 | |
342 | if (HeKLEN (he) == HEf_SVKEY) |
368 | if (HeKLEN (he) == HEf_SVKEY) |
343 | { |
369 | { |
… | |
… | |
356 | encode_ch (enc, '"'); |
382 | encode_ch (enc, '"'); |
357 | |
383 | |
358 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
384 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
359 | encode_ch (enc, ':'); |
385 | encode_ch (enc, ':'); |
360 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
386 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
361 | encode_sv (enc, HeVAL (he)); |
|
|
362 | } |
387 | } |
363 | |
388 | |
364 | // compare hash entries, used when all keys are bytestrings |
389 | // compare hash entries, used when all keys are bytestrings |
365 | static int |
390 | static int |
366 | he_cmp_fast (const void *a_, const void *b_) |
391 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
371 | HE *b = *(HE **)b_; |
396 | HE *b = *(HE **)b_; |
372 | |
397 | |
373 | STRLEN la = HeKLEN (a); |
398 | STRLEN la = HeKLEN (a); |
374 | STRLEN lb = HeKLEN (b); |
399 | STRLEN lb = HeKLEN (b); |
375 | |
400 | |
376 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
401 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
377 | cmp = la - lb; |
402 | cmp = lb - la; |
378 | |
403 | |
379 | return cmp; |
404 | return cmp; |
380 | } |
405 | } |
381 | |
406 | |
382 | // compare hash entries, used when some keys are sv's or utf-x |
407 | // compare hash entries, used when some keys are sv's or utf-x |
383 | static int |
408 | static int |
384 | he_cmp_slow (const void *a, const void *b) |
409 | he_cmp_slow (const void *a, const void *b) |
385 | { |
410 | { |
386 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
411 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
387 | } |
412 | } |
388 | |
413 | |
389 | static void |
414 | static void |
390 | encode_hv (enc_t *enc, HV *hv) |
415 | encode_hv (enc_t *enc, HV *hv) |
391 | { |
416 | { |
|
|
417 | HE *he; |
392 | int count, i; |
418 | int count; |
393 | |
419 | |
394 | if (enc->indent >= enc->maxdepth) |
420 | if (enc->indent >= enc->maxdepth) |
395 | croak ("data structure too deep (hit recursion limit)"); |
421 | croak ("data structure too deep (hit recursion limit)"); |
396 | |
422 | |
397 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
423 | encode_ch (enc, '{'); |
398 | |
424 | |
399 | if ((count = hv_iterinit (hv))) |
|
|
400 | { |
|
|
401 | // for canonical output we have to sort by keys first |
425 | // for canonical output we have to sort by keys first |
402 | // actually, this is mostly due to the stupid so-called |
426 | // actually, this is mostly due to the stupid so-called |
403 | // security workaround added somewhere in 5.8.x. |
427 | // security workaround added somewhere in 5.8.x. |
404 | // that randomises hash orderings |
428 | // that randomises hash orderings |
405 | if (enc->json.flags & F_CANONICAL) |
429 | if (enc->json.flags & F_CANONICAL) |
|
|
430 | { |
|
|
431 | int count = hv_iterinit (hv); |
|
|
432 | |
|
|
433 | if (SvMAGICAL (hv)) |
406 | { |
434 | { |
|
|
435 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
436 | // but I don't care for the speed of this special case. |
|
|
437 | // note also that we will run into undefined behaviour when the two iterations |
|
|
438 | // do not result in the same count, something I might care for in some later release. |
|
|
439 | |
|
|
440 | count = 0; |
|
|
441 | while (hv_iternext (hv)) |
|
|
442 | ++count; |
|
|
443 | |
|
|
444 | hv_iterinit (hv); |
|
|
445 | } |
|
|
446 | |
|
|
447 | if (count) |
|
|
448 | { |
407 | int fast = 1; |
449 | int i, fast = 1; |
408 | HE *he; |
|
|
409 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
450 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
410 | HE **hes = _alloca (count * sizeof (HE)); |
451 | HE **hes = _alloca (count * sizeof (HE)); |
411 | #else |
452 | #else |
412 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
453 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
413 | #endif |
454 | #endif |
… | |
… | |
440 | |
481 | |
441 | FREETMPS; |
482 | FREETMPS; |
442 | LEAVE; |
483 | LEAVE; |
443 | } |
484 | } |
444 | |
485 | |
445 | for (i = 0; i < count; ++i) |
486 | encode_nl (enc); ++enc->indent; |
|
|
487 | |
|
|
488 | while (count--) |
446 | { |
489 | { |
447 | encode_indent (enc); |
490 | encode_indent (enc); |
|
|
491 | he = hes [count]; |
448 | encode_he (enc, hes [i]); |
492 | encode_hk (enc, he); |
|
|
493 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
449 | |
494 | |
450 | if (i < count - 1) |
495 | if (count) |
451 | encode_comma (enc); |
496 | encode_comma (enc); |
452 | } |
497 | } |
453 | |
498 | |
454 | encode_nl (enc); |
499 | encode_nl (enc); --enc->indent; encode_indent (enc); |
455 | } |
500 | } |
|
|
501 | } |
456 | else |
502 | else |
|
|
503 | { |
|
|
504 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
505 | if ((he = hv_iternext (hv))) |
457 | { |
506 | { |
458 | HE *he = hv_iternext (hv); |
507 | encode_nl (enc); ++enc->indent; |
459 | |
508 | |
460 | for (;;) |
509 | for (;;) |
461 | { |
510 | { |
462 | encode_indent (enc); |
511 | encode_indent (enc); |
463 | encode_he (enc, he); |
512 | encode_hk (enc, he); |
|
|
513 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
464 | |
514 | |
465 | if (!(he = hv_iternext (hv))) |
515 | if (!(he = hv_iternext (hv))) |
466 | break; |
516 | break; |
467 | |
517 | |
468 | encode_comma (enc); |
518 | encode_comma (enc); |
469 | } |
519 | } |
470 | |
520 | |
471 | encode_nl (enc); |
521 | encode_nl (enc); --enc->indent; encode_indent (enc); |
472 | } |
522 | } |
473 | } |
523 | } |
474 | |
524 | |
475 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
525 | encode_ch (enc, '}'); |
476 | } |
526 | } |
477 | |
527 | |
478 | // encode objects, arrays and special \0=false and \1=true values. |
528 | // encode objects, arrays and special \0=false and \1=true values. |
479 | static void |
529 | static void |
480 | encode_rv (enc_t *enc, SV *sv) |
530 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
649 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
699 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
650 | enc.cur = SvPVX (enc.sv); |
700 | enc.cur = SvPVX (enc.sv); |
651 | enc.end = SvEND (enc.sv); |
701 | enc.end = SvEND (enc.sv); |
652 | enc.indent = 0; |
702 | enc.indent = 0; |
653 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
703 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
|
|
704 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
705 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
706 | : 0x10FFFFUL; |
654 | |
707 | |
655 | SvPOK_only (enc.sv); |
708 | SvPOK_only (enc.sv); |
656 | encode_sv (&enc, scalar); |
709 | encode_sv (&enc, scalar); |
657 | |
710 | |
658 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
711 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
… | |
… | |
679 | JSON json; |
732 | JSON json; |
680 | U32 depth; // recursion depth |
733 | U32 depth; // recursion depth |
681 | U32 maxdepth; // recursion depth limit |
734 | U32 maxdepth; // recursion depth limit |
682 | } dec_t; |
735 | } dec_t; |
683 | |
736 | |
684 | inline void |
737 | INLINE void |
|
|
738 | decode_comment (dec_t *dec) |
|
|
739 | { |
|
|
740 | // only '#'-style comments allowed a.t.m. |
|
|
741 | |
|
|
742 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
743 | ++dec->cur; |
|
|
744 | } |
|
|
745 | |
|
|
746 | INLINE void |
685 | decode_ws (dec_t *dec) |
747 | decode_ws (dec_t *dec) |
686 | { |
748 | { |
687 | for (;;) |
749 | for (;;) |
688 | { |
750 | { |
689 | char ch = *dec->cur; |
751 | char ch = *dec->cur; |
690 | |
752 | |
691 | if (ch > 0x20 |
753 | if (ch > 0x20) |
|
|
754 | { |
|
|
755 | if (expect_false (ch == '#')) |
|
|
756 | { |
|
|
757 | if (dec->json.flags & F_RELAXED) |
|
|
758 | decode_comment (dec); |
|
|
759 | else |
|
|
760 | break; |
|
|
761 | } |
|
|
762 | else |
|
|
763 | break; |
|
|
764 | } |
692 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
765 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
693 | break; |
766 | break; // parse error, but let higher level handle it, gives better error messages |
694 | |
767 | |
695 | ++dec->cur; |
768 | ++dec->cur; |
696 | } |
769 | } |
697 | } |
770 | } |
698 | |
771 | |
… | |
… | |
804 | |
877 | |
805 | if (hi >= 0x80) |
878 | if (hi >= 0x80) |
806 | { |
879 | { |
807 | utf8 = 1; |
880 | utf8 = 1; |
808 | |
881 | |
809 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
882 | cur = encode_utf8 (cur, hi); |
810 | } |
883 | } |
811 | else |
884 | else |
812 | *cur++ = hi; |
885 | *cur++ = hi; |
813 | } |
886 | } |
814 | break; |
887 | break; |
… | |
… | |
949 | |
1022 | |
950 | if (!is_nv) |
1023 | if (!is_nv) |
951 | { |
1024 | { |
952 | int len = dec->cur - start; |
1025 | int len = dec->cur - start; |
953 | |
1026 | |
954 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1027 | // special case the rather common 1..5-digit-int case |
955 | if (*start == '-') |
1028 | if (*start == '-') |
956 | switch (len) |
1029 | switch (len) |
957 | { |
1030 | { |
958 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1031 | case 2: return newSViv (-( start [1] - '0' * 1)); |
959 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1032 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
960 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1033 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
961 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1034 | case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1035 | case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
962 | } |
1036 | } |
963 | else |
1037 | else |
964 | switch (len) |
1038 | switch (len) |
965 | { |
1039 | { |
966 | case 1: return newSViv ( start [0] - '0' * 1); |
1040 | case 1: return newSViv ( start [0] - '0' * 1); |
967 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1041 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
968 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1042 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
969 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1043 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1044 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
970 | } |
1045 | } |
971 | |
1046 | |
972 | { |
1047 | { |
973 | UV uv; |
1048 | UV uv; |
974 | int numtype = grok_number (start, len, &uv); |
1049 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1035 | |
1110 | |
1036 | if (*dec->cur != ',') |
1111 | if (*dec->cur != ',') |
1037 | ERR (", or ] expected while parsing array"); |
1112 | ERR (", or ] expected while parsing array"); |
1038 | |
1113 | |
1039 | ++dec->cur; |
1114 | ++dec->cur; |
|
|
1115 | |
|
|
1116 | decode_ws (dec); |
|
|
1117 | |
|
|
1118 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1119 | { |
|
|
1120 | ++dec->cur; |
|
|
1121 | break; |
|
|
1122 | } |
1040 | } |
1123 | } |
1041 | |
1124 | |
1042 | DEC_DEC_DEPTH; |
1125 | DEC_DEC_DEPTH; |
1043 | return newRV_noinc ((SV *)av); |
1126 | return newRV_noinc ((SV *)av); |
1044 | |
1127 | |
… | |
… | |
1060 | if (*dec->cur == '}') |
1143 | if (*dec->cur == '}') |
1061 | ++dec->cur; |
1144 | ++dec->cur; |
1062 | else |
1145 | else |
1063 | for (;;) |
1146 | for (;;) |
1064 | { |
1147 | { |
1065 | decode_ws (dec); EXPECT_CH ('"'); |
1148 | EXPECT_CH ('"'); |
1066 | |
1149 | |
1067 | // heuristic: assume that |
1150 | // heuristic: assume that |
1068 | // a) decode_str + hv_store_ent are abysmally slow. |
1151 | // a) decode_str + hv_store_ent are abysmally slow. |
1069 | // b) most hash keys are short, simple ascii text. |
1152 | // b) most hash keys are short, simple ascii text. |
1070 | // => try to "fast-match" such strings to avoid |
1153 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1084 | if (!key) |
1167 | if (!key) |
1085 | goto fail; |
1168 | goto fail; |
1086 | |
1169 | |
1087 | decode_ws (dec); EXPECT_CH (':'); |
1170 | decode_ws (dec); EXPECT_CH (':'); |
1088 | |
1171 | |
|
|
1172 | decode_ws (dec); |
1089 | value = decode_sv (dec); |
1173 | value = decode_sv (dec); |
1090 | if (!value) |
1174 | if (!value) |
1091 | { |
1175 | { |
1092 | SvREFCNT_dec (key); |
1176 | SvREFCNT_dec (key); |
1093 | goto fail; |
1177 | goto fail; |
… | |
… | |
1105 | int len = p - key; |
1189 | int len = p - key; |
1106 | dec->cur = p + 1; |
1190 | dec->cur = p + 1; |
1107 | |
1191 | |
1108 | decode_ws (dec); EXPECT_CH (':'); |
1192 | decode_ws (dec); EXPECT_CH (':'); |
1109 | |
1193 | |
|
|
1194 | decode_ws (dec); |
1110 | value = decode_sv (dec); |
1195 | value = decode_sv (dec); |
1111 | if (!value) |
1196 | if (!value) |
1112 | goto fail; |
1197 | goto fail; |
1113 | |
1198 | |
1114 | hv_store (hv, key, len, value, 0); |
1199 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1130 | |
1215 | |
1131 | if (*dec->cur != ',') |
1216 | if (*dec->cur != ',') |
1132 | ERR (", or } expected while parsing object/hash"); |
1217 | ERR (", or } expected while parsing object/hash"); |
1133 | |
1218 | |
1134 | ++dec->cur; |
1219 | ++dec->cur; |
|
|
1220 | |
|
|
1221 | decode_ws (dec); |
|
|
1222 | |
|
|
1223 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1224 | { |
|
|
1225 | ++dec->cur; |
|
|
1226 | break; |
|
|
1227 | } |
1135 | } |
1228 | } |
1136 | |
1229 | |
1137 | DEC_DEC_DEPTH; |
1230 | DEC_DEC_DEPTH; |
1138 | sv = newRV_noinc ((SV *)hv); |
1231 | sv = newRV_noinc ((SV *)hv); |
1139 | |
1232 | |
… | |
… | |
1204 | } |
1297 | } |
1205 | |
1298 | |
1206 | static SV * |
1299 | static SV * |
1207 | decode_sv (dec_t *dec) |
1300 | decode_sv (dec_t *dec) |
1208 | { |
1301 | { |
1209 | decode_ws (dec); |
|
|
1210 | |
|
|
1211 | // the beauty of JSON: you need exactly one character lookahead |
1302 | // the beauty of JSON: you need exactly one character lookahead |
1212 | // to parse anything. |
1303 | // to parse everything. |
1213 | switch (*dec->cur) |
1304 | switch (*dec->cur) |
1214 | { |
1305 | { |
1215 | case '"': ++dec->cur; return decode_str (dec); |
1306 | case '"': ++dec->cur; return decode_str (dec); |
1216 | case '[': ++dec->cur; return decode_av (dec); |
1307 | case '[': ++dec->cur; return decode_av (dec); |
1217 | case '{': ++dec->cur; return decode_hv (dec); |
1308 | case '{': ++dec->cur; return decode_hv (dec); |
1218 | |
1309 | |
1219 | case '-': |
1310 | case '-': |
1220 | case '0': case '1': case '2': case '3': case '4': |
1311 | case '0': case '1': case '2': case '3': case '4': |
1221 | case '5': case '6': case '7': case '8': case '9': |
1312 | case '5': case '6': case '7': case '8': case '9': |
1222 | return decode_num (dec); |
1313 | return decode_num (dec); |
1223 | |
1314 | |
1224 | case 't': |
1315 | case 't': |
1225 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1316 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1226 | { |
1317 | { |
1227 | dec->cur += 4; |
1318 | dec->cur += 4; |
|
|
1319 | #if JSON_SLOW |
|
|
1320 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
|
|
1321 | #endif |
1228 | return SvREFCNT_inc (json_true); |
1322 | return SvREFCNT_inc (json_true); |
1229 | } |
1323 | } |
1230 | else |
1324 | else |
1231 | ERR ("'true' expected"); |
1325 | ERR ("'true' expected"); |
1232 | |
1326 | |
… | |
… | |
1234 | |
1328 | |
1235 | case 'f': |
1329 | case 'f': |
1236 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1330 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1237 | { |
1331 | { |
1238 | dec->cur += 5; |
1332 | dec->cur += 5; |
|
|
1333 | #if JSON_SLOW |
|
|
1334 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
|
|
1335 | #endif |
1239 | return SvREFCNT_inc (json_false); |
1336 | return SvREFCNT_inc (json_false); |
1240 | } |
1337 | } |
1241 | else |
1338 | else |
1242 | ERR ("'false' expected"); |
1339 | ERR ("'false' expected"); |
1243 | |
1340 | |
… | |
… | |
1293 | |
1390 | |
1294 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1391 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1295 | dec.json.flags |= F_HOOK; |
1392 | dec.json.flags |= F_HOOK; |
1296 | |
1393 | |
1297 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1394 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1395 | |
|
|
1396 | decode_ws (&dec); |
1298 | sv = decode_sv (&dec); |
1397 | sv = decode_sv (&dec); |
1299 | |
1398 | |
1300 | if (!(offset_return || !sv)) |
1399 | if (!(offset_return || !sv)) |
1301 | { |
1400 | { |
1302 | // check for trailing garbage |
1401 | // check for trailing garbage |
… | |
… | |
1378 | json_boolean_stash = 0; |
1477 | json_boolean_stash = 0; |
1379 | |
1478 | |
1380 | void new (char *klass) |
1479 | void new (char *klass) |
1381 | PPCODE: |
1480 | PPCODE: |
1382 | { |
1481 | { |
1383 | HV *stash = !JSON_SLOW || json_stash |
|
|
1384 | ? json_stash |
|
|
1385 | : gv_stashpv ("JSON::XS", 1); |
|
|
1386 | SV *pv = NEWSV (0, sizeof (JSON)); |
1482 | SV *pv = NEWSV (0, sizeof (JSON)); |
1387 | SvPOK_only (pv); |
1483 | SvPOK_only (pv); |
1388 | Zero (SvPVX (pv), 1, JSON); |
1484 | Zero (SvPVX (pv), 1, JSON); |
1389 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1485 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1390 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash))); |
1486 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1487 | newRV_noinc (pv), |
|
|
1488 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1489 | ))); |
1391 | } |
1490 | } |
1392 | |
1491 | |
1393 | void ascii (JSON *self, int enable = 1) |
1492 | void ascii (JSON *self, int enable = 1) |
1394 | ALIAS: |
1493 | ALIAS: |
1395 | ascii = F_ASCII |
1494 | ascii = F_ASCII |
… | |
… | |
1402 | pretty = F_PRETTY |
1501 | pretty = F_PRETTY |
1403 | allow_nonref = F_ALLOW_NONREF |
1502 | allow_nonref = F_ALLOW_NONREF |
1404 | shrink = F_SHRINK |
1503 | shrink = F_SHRINK |
1405 | allow_blessed = F_ALLOW_BLESSED |
1504 | allow_blessed = F_ALLOW_BLESSED |
1406 | convert_blessed = F_CONV_BLESSED |
1505 | convert_blessed = F_CONV_BLESSED |
|
|
1506 | relaxed = F_RELAXED |
1407 | PPCODE: |
1507 | PPCODE: |
1408 | { |
1508 | { |
1409 | if (enable) |
1509 | if (enable) |
1410 | self->flags |= ix; |
1510 | self->flags |= ix; |
1411 | else |
1511 | else |
1412 | self->flags &= ~ix; |
1512 | self->flags &= ~ix; |
1413 | |
1513 | |
1414 | XPUSHs (ST (0)); |
1514 | XPUSHs (ST (0)); |
1415 | } |
1515 | } |
1416 | |
1516 | |
|
|
1517 | void get_ascii (JSON *self) |
|
|
1518 | ALIAS: |
|
|
1519 | get_ascii = F_ASCII |
|
|
1520 | get_latin1 = F_LATIN1 |
|
|
1521 | get_utf8 = F_UTF8 |
|
|
1522 | get_indent = F_INDENT |
|
|
1523 | get_canonical = F_CANONICAL |
|
|
1524 | get_space_before = F_SPACE_BEFORE |
|
|
1525 | get_space_after = F_SPACE_AFTER |
|
|
1526 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1527 | get_shrink = F_SHRINK |
|
|
1528 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1529 | get_convert_blessed = F_CONV_BLESSED |
|
|
1530 | get_relaxed = F_RELAXED |
|
|
1531 | PPCODE: |
|
|
1532 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1533 | |
1417 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1534 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1418 | PPCODE: |
1535 | PPCODE: |
1419 | { |
1536 | { |
1420 | UV log2 = 0; |
1537 | UV log2 = 0; |
1421 | |
1538 | |
… | |
… | |
1427 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1544 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1428 | |
1545 | |
1429 | XPUSHs (ST (0)); |
1546 | XPUSHs (ST (0)); |
1430 | } |
1547 | } |
1431 | |
1548 | |
|
|
1549 | U32 get_max_depth (JSON *self) |
|
|
1550 | CODE: |
|
|
1551 | RETVAL = DEC_DEPTH (self->flags); |
|
|
1552 | OUTPUT: |
|
|
1553 | RETVAL |
|
|
1554 | |
1432 | void max_size (JSON *self, UV max_size = 0) |
1555 | void max_size (JSON *self, UV max_size = 0) |
1433 | PPCODE: |
1556 | PPCODE: |
1434 | { |
1557 | { |
1435 | UV log2 = 0; |
1558 | UV log2 = 0; |
1436 | |
1559 | |
… | |
… | |
1442 | |
1565 | |
1443 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
1566 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
1444 | |
1567 | |
1445 | XPUSHs (ST (0)); |
1568 | XPUSHs (ST (0)); |
1446 | } |
1569 | } |
|
|
1570 | |
|
|
1571 | int get_max_size (JSON *self) |
|
|
1572 | CODE: |
|
|
1573 | RETVAL = DEC_SIZE (self->flags); |
|
|
1574 | OUTPUT: |
|
|
1575 | RETVAL |
1447 | |
1576 | |
1448 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1577 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1449 | PPCODE: |
1578 | PPCODE: |
1450 | { |
1579 | { |
1451 | SvREFCNT_dec (self->cb_object); |
1580 | SvREFCNT_dec (self->cb_object); |
… | |
… | |
1498 | SvREFCNT_dec (self->cb_sk_object); |
1627 | SvREFCNT_dec (self->cb_sk_object); |
1499 | SvREFCNT_dec (self->cb_object); |
1628 | SvREFCNT_dec (self->cb_object); |
1500 | |
1629 | |
1501 | PROTOTYPES: ENABLE |
1630 | PROTOTYPES: ENABLE |
1502 | |
1631 | |
1503 | void to_json (SV *scalar) |
1632 | void encode_json (SV *scalar) |
1504 | PPCODE: |
1633 | PPCODE: |
1505 | { |
1634 | { |
1506 | JSON json = { F_DEFAULT | F_UTF8 }; |
1635 | JSON json = { F_DEFAULT | F_UTF8 }; |
1507 | XPUSHs (encode_json (scalar, &json)); |
1636 | XPUSHs (encode_json (scalar, &json)); |
1508 | } |
1637 | } |
1509 | |
1638 | |
1510 | void from_json (SV *jsonstr) |
1639 | void decode_json (SV *jsonstr) |
1511 | PPCODE: |
1640 | PPCODE: |
1512 | { |
1641 | { |
1513 | JSON json = { F_DEFAULT | F_UTF8 }; |
1642 | JSON json = { F_DEFAULT | F_UTF8 }; |
1514 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1643 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1515 | } |
1644 | } |