… | |
… | |
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) |
|
|
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))) |
62 | |
68 | |
63 | #ifdef USE_ITHREADS |
69 | #ifdef USE_ITHREADS |
64 | # define JSON_SLOW 1 |
70 | # define JSON_SLOW 1 |
65 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
71 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
66 | #else |
72 | #else |
… | |
… | |
78 | } JSON; |
84 | } JSON; |
79 | |
85 | |
80 | ///////////////////////////////////////////////////////////////////////////// |
86 | ///////////////////////////////////////////////////////////////////////////// |
81 | // utility functions |
87 | // utility functions |
82 | |
88 | |
83 | inline void |
89 | INLINE void |
84 | shrink (SV *sv) |
90 | shrink (SV *sv) |
85 | { |
91 | { |
86 | sv_utf8_downgrade (sv, 1); |
92 | sv_utf8_downgrade (sv, 1); |
87 | if (SvLEN (sv) > SvCUR (sv) + 1) |
93 | if (SvLEN (sv) > SvCUR (sv) + 1) |
88 | { |
94 | { |
… | |
… | |
97 | // 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 |
98 | // case of an error. |
104 | // case of an error. |
99 | // we special-case "safe" characters from U+80 .. U+7FF, |
105 | // we special-case "safe" characters from U+80 .. U+7FF, |
100 | // but use the very good perl function to parse anything else. |
106 | // but use the very good perl function to parse anything else. |
101 | // note that we never call this function for a ascii codepoints |
107 | // note that we never call this function for a ascii codepoints |
102 | inline UV |
108 | INLINE UV |
103 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
109 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
104 | { |
110 | { |
105 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
111 | if (expect_true (len >= 2 |
106 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
112 | && IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
107 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
113 | && IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
108 | { |
114 | { |
109 | *clen = 2; |
115 | *clen = 2; |
110 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
116 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
111 | } |
117 | } |
112 | else |
118 | else |
113 | { |
119 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
114 | *clen = (STRLEN)-1; |
120 | } |
115 | return (UV)-1; |
121 | |
116 | } |
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 (expect_false (ch < 0x000080)) |
|
|
129 | *s++ = ch; |
|
|
130 | else if (expect_true (ch < 0x000800)) |
|
|
131 | *s++ = 0xc0 | ( ch >> 6), |
|
|
132 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
133 | else if ( ch < 0x010000) |
|
|
134 | *s++ = 0xe0 | ( ch >> 12), |
|
|
135 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
136 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
137 | else if ( ch < 0x110000) |
|
|
138 | *s++ = 0xf0 | ( ch >> 18), |
|
|
139 | *s++ = 0x80 | ((ch >> 12) & 0x3f), |
|
|
140 | *s++ = 0x80 | ((ch >> 6) & 0x3f), |
|
|
141 | *s++ = 0x80 | ( ch & 0x3f); |
|
|
142 | |
|
|
143 | return s; |
117 | } |
144 | } |
118 | |
145 | |
119 | ///////////////////////////////////////////////////////////////////////////// |
146 | ///////////////////////////////////////////////////////////////////////////// |
120 | // encoder |
147 | // encoder |
121 | |
148 | |
… | |
… | |
126 | char *end; // SvEND (sv) |
153 | char *end; // SvEND (sv) |
127 | SV *sv; // result scalar |
154 | SV *sv; // result scalar |
128 | JSON json; |
155 | JSON json; |
129 | U32 indent; // indentation level |
156 | U32 indent; // indentation level |
130 | U32 maxdepth; // max. indentation/recursion level |
157 | U32 maxdepth; // max. indentation/recursion level |
|
|
158 | UV limit; // escape character values >= this value when encoding |
131 | } enc_t; |
159 | } enc_t; |
132 | |
160 | |
133 | inline void |
161 | INLINE void |
134 | need (enc_t *enc, STRLEN len) |
162 | need (enc_t *enc, STRLEN len) |
135 | { |
163 | { |
136 | if (expect_false (enc->cur + len >= enc->end)) |
164 | if (expect_false (enc->cur + len >= enc->end)) |
137 | { |
165 | { |
138 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
166 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
… | |
… | |
140 | enc->cur = SvPVX (enc->sv) + cur; |
168 | enc->cur = SvPVX (enc->sv) + cur; |
141 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
169 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
142 | } |
170 | } |
143 | } |
171 | } |
144 | |
172 | |
145 | inline void |
173 | INLINE void |
146 | encode_ch (enc_t *enc, char ch) |
174 | encode_ch (enc_t *enc, char ch) |
147 | { |
175 | { |
148 | need (enc, 1); |
176 | need (enc, 1); |
149 | *enc->cur++ = ch; |
177 | *enc->cur++ = ch; |
150 | } |
178 | } |
… | |
… | |
204 | { |
232 | { |
205 | uch = ch; |
233 | uch = ch; |
206 | clen = 1; |
234 | clen = 1; |
207 | } |
235 | } |
208 | |
236 | |
209 | if (uch > 0x10FFFFUL) |
237 | if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
210 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
211 | |
|
|
212 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
|
|
213 | { |
238 | { |
214 | if (uch > 0xFFFFUL) |
239 | if (uch >= 0x10000UL) |
215 | { |
240 | { |
|
|
241 | if (uch >= 0x110000UL) |
|
|
242 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
243 | |
216 | need (enc, len += 11); |
244 | need (enc, len += 11); |
217 | sprintf (enc->cur, "\\u%04x\\u%04x", |
245 | sprintf (enc->cur, "\\u%04x\\u%04x", |
218 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
246 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
219 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
247 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
220 | enc->cur += 12; |
248 | enc->cur += 12; |
… | |
… | |
248 | while (--clen); |
276 | while (--clen); |
249 | } |
277 | } |
250 | else |
278 | else |
251 | { |
279 | { |
252 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
280 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
253 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
281 | enc->cur = encode_utf8 (enc->cur, uch); |
254 | ++str; |
282 | ++str; |
255 | } |
283 | } |
256 | } |
284 | } |
257 | } |
285 | } |
258 | } |
286 | } |
259 | |
287 | |
260 | --len; |
288 | --len; |
261 | } |
289 | } |
262 | } |
290 | } |
263 | |
291 | |
264 | inline void |
292 | INLINE void |
265 | encode_indent (enc_t *enc) |
293 | encode_indent (enc_t *enc) |
266 | { |
294 | { |
267 | if (enc->json.flags & F_INDENT) |
295 | if (enc->json.flags & F_INDENT) |
268 | { |
296 | { |
269 | int spaces = enc->indent * INDENT_STEP; |
297 | int spaces = enc->indent * INDENT_STEP; |
… | |
… | |
272 | memset (enc->cur, ' ', spaces); |
300 | memset (enc->cur, ' ', spaces); |
273 | enc->cur += spaces; |
301 | enc->cur += spaces; |
274 | } |
302 | } |
275 | } |
303 | } |
276 | |
304 | |
277 | inline void |
305 | INLINE void |
278 | encode_space (enc_t *enc) |
306 | encode_space (enc_t *enc) |
279 | { |
307 | { |
280 | need (enc, 1); |
308 | need (enc, 1); |
281 | encode_ch (enc, ' '); |
309 | encode_ch (enc, ' '); |
282 | } |
310 | } |
283 | |
311 | |
284 | inline void |
312 | INLINE void |
285 | encode_nl (enc_t *enc) |
313 | encode_nl (enc_t *enc) |
286 | { |
314 | { |
287 | if (enc->json.flags & F_INDENT) |
315 | if (enc->json.flags & F_INDENT) |
288 | { |
316 | { |
289 | need (enc, 1); |
317 | need (enc, 1); |
290 | encode_ch (enc, '\n'); |
318 | encode_ch (enc, '\n'); |
291 | } |
319 | } |
292 | } |
320 | } |
293 | |
321 | |
294 | inline void |
322 | INLINE void |
295 | encode_comma (enc_t *enc) |
323 | encode_comma (enc_t *enc) |
296 | { |
324 | { |
297 | encode_ch (enc, ','); |
325 | encode_ch (enc, ','); |
298 | |
326 | |
299 | if (enc->json.flags & F_INDENT) |
327 | if (enc->json.flags & F_INDENT) |
… | |
… | |
310 | int i, len = av_len (av); |
338 | int i, len = av_len (av); |
311 | |
339 | |
312 | if (enc->indent >= enc->maxdepth) |
340 | if (enc->indent >= enc->maxdepth) |
313 | croak ("data structure too deep (hit recursion limit)"); |
341 | croak ("data structure too deep (hit recursion limit)"); |
314 | |
342 | |
315 | encode_ch (enc, '['); encode_nl (enc); |
343 | encode_ch (enc, '['); |
316 | ++enc->indent; |
344 | |
|
|
345 | if (len >= 0) |
|
|
346 | { |
|
|
347 | encode_nl (enc); ++enc->indent; |
317 | |
348 | |
318 | for (i = 0; i <= len; ++i) |
349 | for (i = 0; i <= len; ++i) |
319 | { |
350 | { |
320 | SV **svp = av_fetch (av, i, 0); |
351 | SV **svp = av_fetch (av, i, 0); |
321 | |
352 | |
322 | encode_indent (enc); |
353 | encode_indent (enc); |
323 | |
354 | |
324 | if (svp) |
355 | if (svp) |
325 | encode_sv (enc, *svp); |
356 | encode_sv (enc, *svp); |
326 | else |
357 | else |
327 | encode_str (enc, "null", 4, 0); |
358 | encode_str (enc, "null", 4, 0); |
328 | |
359 | |
329 | if (i < len) |
360 | if (i < len) |
330 | encode_comma (enc); |
361 | encode_comma (enc); |
331 | } |
362 | } |
332 | |
363 | |
|
|
364 | encode_nl (enc); --enc->indent; encode_indent (enc); |
|
|
365 | } |
|
|
366 | |
333 | encode_nl (enc); |
367 | encode_ch (enc, ']'); |
334 | |
|
|
335 | --enc->indent; |
|
|
336 | encode_indent (enc); encode_ch (enc, ']'); |
|
|
337 | } |
368 | } |
338 | |
369 | |
339 | static void |
370 | static void |
340 | encode_he (enc_t *enc, HE *he) |
371 | encode_hk (enc_t *enc, HE *he) |
341 | { |
372 | { |
342 | encode_ch (enc, '"'); |
373 | encode_ch (enc, '"'); |
343 | |
374 | |
344 | if (HeKLEN (he) == HEf_SVKEY) |
375 | if (HeKLEN (he) == HEf_SVKEY) |
345 | { |
376 | { |
… | |
… | |
358 | encode_ch (enc, '"'); |
389 | encode_ch (enc, '"'); |
359 | |
390 | |
360 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
391 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
361 | encode_ch (enc, ':'); |
392 | encode_ch (enc, ':'); |
362 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
393 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
363 | encode_sv (enc, HeVAL (he)); |
|
|
364 | } |
394 | } |
365 | |
395 | |
366 | // compare hash entries, used when all keys are bytestrings |
396 | // compare hash entries, used when all keys are bytestrings |
367 | static int |
397 | static int |
368 | he_cmp_fast (const void *a_, const void *b_) |
398 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
373 | HE *b = *(HE **)b_; |
403 | HE *b = *(HE **)b_; |
374 | |
404 | |
375 | STRLEN la = HeKLEN (a); |
405 | STRLEN la = HeKLEN (a); |
376 | STRLEN lb = HeKLEN (b); |
406 | STRLEN lb = HeKLEN (b); |
377 | |
407 | |
378 | if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
408 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
379 | cmp = la - lb; |
409 | cmp = lb - la; |
380 | |
410 | |
381 | return cmp; |
411 | return cmp; |
382 | } |
412 | } |
383 | |
413 | |
384 | // compare hash entries, used when some keys are sv's or utf-x |
414 | // compare hash entries, used when some keys are sv's or utf-x |
385 | static int |
415 | static int |
386 | he_cmp_slow (const void *a, const void *b) |
416 | he_cmp_slow (const void *a, const void *b) |
387 | { |
417 | { |
388 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
418 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
389 | } |
419 | } |
390 | |
420 | |
391 | static void |
421 | static void |
392 | encode_hv (enc_t *enc, HV *hv) |
422 | encode_hv (enc_t *enc, HV *hv) |
393 | { |
423 | { |
|
|
424 | HE *he; |
394 | int count, i; |
425 | int count; |
395 | |
426 | |
396 | if (enc->indent >= enc->maxdepth) |
427 | if (enc->indent >= enc->maxdepth) |
397 | croak ("data structure too deep (hit recursion limit)"); |
428 | croak ("data structure too deep (hit recursion limit)"); |
398 | |
429 | |
399 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
430 | encode_ch (enc, '{'); |
400 | |
431 | |
401 | if ((count = hv_iterinit (hv))) |
|
|
402 | { |
|
|
403 | // for canonical output we have to sort by keys first |
432 | // for canonical output we have to sort by keys first |
404 | // actually, this is mostly due to the stupid so-called |
433 | // actually, this is mostly due to the stupid so-called |
405 | // security workaround added somewhere in 5.8.x. |
434 | // security workaround added somewhere in 5.8.x. |
406 | // that randomises hash orderings |
435 | // that randomises hash orderings |
407 | if (enc->json.flags & F_CANONICAL) |
436 | if (enc->json.flags & F_CANONICAL) |
|
|
437 | { |
|
|
438 | int count = hv_iterinit (hv); |
|
|
439 | |
|
|
440 | if (SvMAGICAL (hv)) |
408 | { |
441 | { |
|
|
442 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
443 | // but I don't care for the speed of this special case. |
|
|
444 | // note also that we will run into undefined behaviour when the two iterations |
|
|
445 | // do not result in the same count, something I might care for in some later release. |
|
|
446 | |
|
|
447 | count = 0; |
|
|
448 | while (hv_iternext (hv)) |
|
|
449 | ++count; |
|
|
450 | |
|
|
451 | hv_iterinit (hv); |
|
|
452 | } |
|
|
453 | |
|
|
454 | if (count) |
|
|
455 | { |
409 | int fast = 1; |
456 | int i, fast = 1; |
410 | HE *he; |
|
|
411 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
457 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
412 | HE **hes = _alloca (count * sizeof (HE)); |
458 | HE **hes = _alloca (count * sizeof (HE)); |
413 | #else |
459 | #else |
414 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
460 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
415 | #endif |
461 | #endif |
… | |
… | |
442 | |
488 | |
443 | FREETMPS; |
489 | FREETMPS; |
444 | LEAVE; |
490 | LEAVE; |
445 | } |
491 | } |
446 | |
492 | |
447 | for (i = 0; i < count; ++i) |
493 | encode_nl (enc); ++enc->indent; |
|
|
494 | |
|
|
495 | while (count--) |
448 | { |
496 | { |
449 | encode_indent (enc); |
497 | encode_indent (enc); |
|
|
498 | he = hes [count]; |
450 | encode_he (enc, hes [i]); |
499 | encode_hk (enc, he); |
|
|
500 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
451 | |
501 | |
452 | if (i < count - 1) |
502 | if (count) |
453 | encode_comma (enc); |
503 | encode_comma (enc); |
454 | } |
504 | } |
455 | |
505 | |
456 | encode_nl (enc); |
506 | encode_nl (enc); --enc->indent; encode_indent (enc); |
457 | } |
507 | } |
|
|
508 | } |
458 | else |
509 | else |
|
|
510 | { |
|
|
511 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
512 | if ((he = hv_iternext (hv))) |
459 | { |
513 | { |
460 | HE *he = hv_iternext (hv); |
514 | encode_nl (enc); ++enc->indent; |
461 | |
515 | |
462 | for (;;) |
516 | for (;;) |
463 | { |
517 | { |
464 | encode_indent (enc); |
518 | encode_indent (enc); |
465 | encode_he (enc, he); |
519 | encode_hk (enc, he); |
|
|
520 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
466 | |
521 | |
467 | if (!(he = hv_iternext (hv))) |
522 | if (!(he = hv_iternext (hv))) |
468 | break; |
523 | break; |
469 | |
524 | |
470 | encode_comma (enc); |
525 | encode_comma (enc); |
471 | } |
526 | } |
472 | |
527 | |
473 | encode_nl (enc); |
528 | encode_nl (enc); --enc->indent; encode_indent (enc); |
474 | } |
529 | } |
475 | } |
530 | } |
476 | |
531 | |
477 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
532 | encode_ch (enc, '}'); |
478 | } |
533 | } |
479 | |
534 | |
480 | // encode objects, arrays and special \0=false and \1=true values. |
535 | // encode objects, arrays and special \0=false and \1=true values. |
481 | static void |
536 | static void |
482 | encode_rv (enc_t *enc, SV *sv) |
537 | encode_rv (enc_t *enc, SV *sv) |
… | |
… | |
590 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
645 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
591 | enc->cur += strlen (enc->cur); |
646 | enc->cur += strlen (enc->cur); |
592 | } |
647 | } |
593 | else if (SvIOKp (sv)) |
648 | else if (SvIOKp (sv)) |
594 | { |
649 | { |
595 | // we assume we can always read an IV as a UV |
650 | // we assume we can always read an IV as a UV and vice versa |
596 | if (SvUV (sv) & ~(UV)0x7fff) |
651 | // we assume two's complement |
|
|
652 | // we assume no aliasing issues in the union |
|
|
653 | if (SvIsUV (sv) ? SvUVX (sv) > 59000 |
|
|
654 | : SvIVX (sv) > 59000 || SvIVX (sv) < -59000) |
597 | { |
655 | { |
598 | // large integer, use the (rather slow) snprintf way. |
656 | // large integer, use the (rather slow) snprintf way. |
599 | need (enc, sizeof (UV) * 3); |
657 | need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8 |
600 | enc->cur += |
658 | enc->cur += |
601 | SvIsUV(sv) |
659 | SvIsUV(sv) |
602 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
660 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
603 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
661 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
604 | } |
662 | } |
605 | else |
663 | else |
606 | { |
664 | { |
607 | // optimise the "small number case" |
665 | // optimise the "small number case" |
608 | // code will likely be branchless and use only a single multiplication |
666 | // code will likely be branchless and use only a single multiplication |
|
|
667 | // works for numbers up to 59074 |
609 | I32 i = SvIV (sv); |
668 | I32 i = SvIVX (sv); |
610 | U32 u; |
669 | U32 u; |
611 | char digit, nz = 0; |
670 | char digit, nz = 0; |
612 | |
671 | |
613 | need (enc, 6); |
672 | need (enc, 6); |
614 | |
673 | |
… | |
… | |
651 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
710 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
652 | enc.cur = SvPVX (enc.sv); |
711 | enc.cur = SvPVX (enc.sv); |
653 | enc.end = SvEND (enc.sv); |
712 | enc.end = SvEND (enc.sv); |
654 | enc.indent = 0; |
713 | enc.indent = 0; |
655 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
714 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
|
|
715 | enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
|
|
716 | : enc.json.flags & F_LATIN1 ? 0x000100UL |
|
|
717 | : 0x110000UL; |
656 | |
718 | |
657 | SvPOK_only (enc.sv); |
719 | SvPOK_only (enc.sv); |
658 | encode_sv (&enc, scalar); |
720 | encode_sv (&enc, scalar); |
659 | |
721 | |
660 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
722 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
… | |
… | |
681 | JSON json; |
743 | JSON json; |
682 | U32 depth; // recursion depth |
744 | U32 depth; // recursion depth |
683 | U32 maxdepth; // recursion depth limit |
745 | U32 maxdepth; // recursion depth limit |
684 | } dec_t; |
746 | } dec_t; |
685 | |
747 | |
686 | inline void |
748 | INLINE void |
|
|
749 | decode_comment (dec_t *dec) |
|
|
750 | { |
|
|
751 | // only '#'-style comments allowed a.t.m. |
|
|
752 | |
|
|
753 | while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
|
|
754 | ++dec->cur; |
|
|
755 | } |
|
|
756 | |
|
|
757 | INLINE void |
687 | decode_ws (dec_t *dec) |
758 | decode_ws (dec_t *dec) |
688 | { |
759 | { |
689 | for (;;) |
760 | for (;;) |
690 | { |
761 | { |
691 | char ch = *dec->cur; |
762 | char ch = *dec->cur; |
692 | |
763 | |
693 | if (ch > 0x20 |
764 | if (ch > 0x20) |
|
|
765 | { |
|
|
766 | if (expect_false (ch == '#')) |
|
|
767 | { |
|
|
768 | if (dec->json.flags & F_RELAXED) |
|
|
769 | decode_comment (dec); |
|
|
770 | else |
|
|
771 | break; |
|
|
772 | } |
|
|
773 | else |
|
|
774 | break; |
|
|
775 | } |
694 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
776 | else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
695 | break; |
777 | break; // parse error, but let higher level handle it, gives better error messages |
696 | |
778 | |
697 | ++dec->cur; |
779 | ++dec->cur; |
698 | } |
780 | } |
699 | } |
781 | } |
700 | |
782 | |
… | |
… | |
806 | |
888 | |
807 | if (hi >= 0x80) |
889 | if (hi >= 0x80) |
808 | { |
890 | { |
809 | utf8 = 1; |
891 | utf8 = 1; |
810 | |
892 | |
811 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
893 | cur = encode_utf8 (cur, hi); |
812 | } |
894 | } |
813 | else |
895 | else |
814 | *cur++ = hi; |
896 | *cur++ = hi; |
815 | } |
897 | } |
816 | break; |
898 | break; |
… | |
… | |
818 | default: |
900 | default: |
819 | --dec_cur; |
901 | --dec_cur; |
820 | ERR ("illegal backslash escape sequence in string"); |
902 | ERR ("illegal backslash escape sequence in string"); |
821 | } |
903 | } |
822 | } |
904 | } |
823 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
905 | else if (expect_true (ch >= 0x20 && ch < 0x80)) |
824 | *cur++ = ch; |
906 | *cur++ = ch; |
825 | else if (ch >= 0x80) |
907 | else if (ch >= 0x80) |
826 | { |
908 | { |
827 | STRLEN clen; |
909 | STRLEN clen; |
828 | UV uch; |
910 | UV uch; |
… | |
… | |
951 | |
1033 | |
952 | if (!is_nv) |
1034 | if (!is_nv) |
953 | { |
1035 | { |
954 | int len = dec->cur - start; |
1036 | int len = dec->cur - start; |
955 | |
1037 | |
956 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
1038 | // special case the rather common 1..5-digit-int case |
957 | if (*start == '-') |
1039 | if (*start == '-') |
958 | switch (len) |
1040 | switch (len) |
959 | { |
1041 | { |
960 | case 2: return newSViv (-( start [1] - '0' * 1)); |
1042 | case 2: return newSViv (-( start [1] - '0' * 1)); |
961 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
1043 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
962 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1044 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
963 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1045 | case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
|
|
1046 | case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
964 | } |
1047 | } |
965 | else |
1048 | else |
966 | switch (len) |
1049 | switch (len) |
967 | { |
1050 | { |
968 | case 1: return newSViv ( start [0] - '0' * 1); |
1051 | case 1: return newSViv ( start [0] - '0' * 1); |
969 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1052 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
970 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1053 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
971 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1054 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
1055 | case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
972 | } |
1056 | } |
973 | |
1057 | |
974 | { |
1058 | { |
975 | UV uv; |
1059 | UV uv; |
976 | int numtype = grok_number (start, len, &uv); |
1060 | int numtype = grok_number (start, len, &uv); |
… | |
… | |
1037 | |
1121 | |
1038 | if (*dec->cur != ',') |
1122 | if (*dec->cur != ',') |
1039 | ERR (", or ] expected while parsing array"); |
1123 | ERR (", or ] expected while parsing array"); |
1040 | |
1124 | |
1041 | ++dec->cur; |
1125 | ++dec->cur; |
|
|
1126 | |
|
|
1127 | decode_ws (dec); |
|
|
1128 | |
|
|
1129 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1130 | { |
|
|
1131 | ++dec->cur; |
|
|
1132 | break; |
|
|
1133 | } |
1042 | } |
1134 | } |
1043 | |
1135 | |
1044 | DEC_DEC_DEPTH; |
1136 | DEC_DEC_DEPTH; |
1045 | return newRV_noinc ((SV *)av); |
1137 | return newRV_noinc ((SV *)av); |
1046 | |
1138 | |
… | |
… | |
1062 | if (*dec->cur == '}') |
1154 | if (*dec->cur == '}') |
1063 | ++dec->cur; |
1155 | ++dec->cur; |
1064 | else |
1156 | else |
1065 | for (;;) |
1157 | for (;;) |
1066 | { |
1158 | { |
1067 | decode_ws (dec); EXPECT_CH ('"'); |
1159 | EXPECT_CH ('"'); |
1068 | |
1160 | |
1069 | // heuristic: assume that |
1161 | // heuristic: assume that |
1070 | // a) decode_str + hv_store_ent are abysmally slow. |
1162 | // a) decode_str + hv_store_ent are abysmally slow. |
1071 | // b) most hash keys are short, simple ascii text. |
1163 | // b) most hash keys are short, simple ascii text. |
1072 | // => try to "fast-match" such strings to avoid |
1164 | // => try to "fast-match" such strings to avoid |
… | |
… | |
1086 | if (!key) |
1178 | if (!key) |
1087 | goto fail; |
1179 | goto fail; |
1088 | |
1180 | |
1089 | decode_ws (dec); EXPECT_CH (':'); |
1181 | decode_ws (dec); EXPECT_CH (':'); |
1090 | |
1182 | |
|
|
1183 | decode_ws (dec); |
1091 | value = decode_sv (dec); |
1184 | value = decode_sv (dec); |
1092 | if (!value) |
1185 | if (!value) |
1093 | { |
1186 | { |
1094 | SvREFCNT_dec (key); |
1187 | SvREFCNT_dec (key); |
1095 | goto fail; |
1188 | goto fail; |
… | |
… | |
1107 | int len = p - key; |
1200 | int len = p - key; |
1108 | dec->cur = p + 1; |
1201 | dec->cur = p + 1; |
1109 | |
1202 | |
1110 | decode_ws (dec); EXPECT_CH (':'); |
1203 | decode_ws (dec); EXPECT_CH (':'); |
1111 | |
1204 | |
|
|
1205 | decode_ws (dec); |
1112 | value = decode_sv (dec); |
1206 | value = decode_sv (dec); |
1113 | if (!value) |
1207 | if (!value) |
1114 | goto fail; |
1208 | goto fail; |
1115 | |
1209 | |
1116 | hv_store (hv, key, len, value, 0); |
1210 | hv_store (hv, key, len, value, 0); |
… | |
… | |
1132 | |
1226 | |
1133 | if (*dec->cur != ',') |
1227 | if (*dec->cur != ',') |
1134 | ERR (", or } expected while parsing object/hash"); |
1228 | ERR (", or } expected while parsing object/hash"); |
1135 | |
1229 | |
1136 | ++dec->cur; |
1230 | ++dec->cur; |
|
|
1231 | |
|
|
1232 | decode_ws (dec); |
|
|
1233 | |
|
|
1234 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1235 | { |
|
|
1236 | ++dec->cur; |
|
|
1237 | break; |
|
|
1238 | } |
1137 | } |
1239 | } |
1138 | |
1240 | |
1139 | DEC_DEC_DEPTH; |
1241 | DEC_DEC_DEPTH; |
1140 | sv = newRV_noinc ((SV *)hv); |
1242 | sv = newRV_noinc ((SV *)hv); |
1141 | |
1243 | |
… | |
… | |
1206 | } |
1308 | } |
1207 | |
1309 | |
1208 | static SV * |
1310 | static SV * |
1209 | decode_sv (dec_t *dec) |
1311 | decode_sv (dec_t *dec) |
1210 | { |
1312 | { |
1211 | decode_ws (dec); |
|
|
1212 | |
|
|
1213 | // the beauty of JSON: you need exactly one character lookahead |
1313 | // the beauty of JSON: you need exactly one character lookahead |
1214 | // to parse anything. |
1314 | // to parse everything. |
1215 | switch (*dec->cur) |
1315 | switch (*dec->cur) |
1216 | { |
1316 | { |
1217 | case '"': ++dec->cur; return decode_str (dec); |
1317 | case '"': ++dec->cur; return decode_str (dec); |
1218 | case '[': ++dec->cur; return decode_av (dec); |
1318 | case '[': ++dec->cur; return decode_av (dec); |
1219 | case '{': ++dec->cur; return decode_hv (dec); |
1319 | case '{': ++dec->cur; return decode_hv (dec); |
1220 | |
1320 | |
1221 | case '-': |
1321 | case '-': |
1222 | case '0': case '1': case '2': case '3': case '4': |
1322 | case '0': case '1': case '2': case '3': case '4': |
1223 | case '5': case '6': case '7': case '8': case '9': |
1323 | case '5': case '6': case '7': case '8': case '9': |
1224 | return decode_num (dec); |
1324 | return decode_num (dec); |
1225 | |
1325 | |
1226 | case 't': |
1326 | case 't': |
1227 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1327 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1228 | { |
1328 | { |
1229 | dec->cur += 4; |
1329 | dec->cur += 4; |
|
|
1330 | #if JSON_SLOW |
|
|
1331 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
|
|
1332 | #endif |
1230 | return SvREFCNT_inc (json_true); |
1333 | return SvREFCNT_inc (json_true); |
1231 | } |
1334 | } |
1232 | else |
1335 | else |
1233 | ERR ("'true' expected"); |
1336 | ERR ("'true' expected"); |
1234 | |
1337 | |
… | |
… | |
1236 | |
1339 | |
1237 | case 'f': |
1340 | case 'f': |
1238 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1341 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1239 | { |
1342 | { |
1240 | dec->cur += 5; |
1343 | dec->cur += 5; |
|
|
1344 | #if JSON_SLOW |
|
|
1345 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
|
|
1346 | #endif |
1241 | return SvREFCNT_inc (json_false); |
1347 | return SvREFCNT_inc (json_false); |
1242 | } |
1348 | } |
1243 | else |
1349 | else |
1244 | ERR ("'false' expected"); |
1350 | ERR ("'false' expected"); |
1245 | |
1351 | |
… | |
… | |
1295 | |
1401 | |
1296 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1402 | if (dec.json.cb_object || dec.json.cb_sk_object) |
1297 | dec.json.flags |= F_HOOK; |
1403 | dec.json.flags |= F_HOOK; |
1298 | |
1404 | |
1299 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1405 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1406 | |
|
|
1407 | decode_ws (&dec); |
1300 | sv = decode_sv (&dec); |
1408 | sv = decode_sv (&dec); |
1301 | |
1409 | |
1302 | if (!(offset_return || !sv)) |
1410 | if (!(offset_return || !sv)) |
1303 | { |
1411 | { |
1304 | // check for trailing garbage |
1412 | // check for trailing garbage |
… | |
… | |
1384 | { |
1492 | { |
1385 | SV *pv = NEWSV (0, sizeof (JSON)); |
1493 | SV *pv = NEWSV (0, sizeof (JSON)); |
1386 | SvPOK_only (pv); |
1494 | SvPOK_only (pv); |
1387 | Zero (SvPVX (pv), 1, JSON); |
1495 | Zero (SvPVX (pv), 1, JSON); |
1388 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1496 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1389 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); |
1497 | XPUSHs (sv_2mortal (sv_bless ( |
|
|
1498 | newRV_noinc (pv), |
|
|
1499 | strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
|
|
1500 | ))); |
1390 | } |
1501 | } |
1391 | |
1502 | |
1392 | void ascii (JSON *self, int enable = 1) |
1503 | void ascii (JSON *self, int enable = 1) |
1393 | ALIAS: |
1504 | ALIAS: |
1394 | ascii = F_ASCII |
1505 | ascii = F_ASCII |
… | |
… | |
1401 | pretty = F_PRETTY |
1512 | pretty = F_PRETTY |
1402 | allow_nonref = F_ALLOW_NONREF |
1513 | allow_nonref = F_ALLOW_NONREF |
1403 | shrink = F_SHRINK |
1514 | shrink = F_SHRINK |
1404 | allow_blessed = F_ALLOW_BLESSED |
1515 | allow_blessed = F_ALLOW_BLESSED |
1405 | convert_blessed = F_CONV_BLESSED |
1516 | convert_blessed = F_CONV_BLESSED |
|
|
1517 | relaxed = F_RELAXED |
1406 | PPCODE: |
1518 | PPCODE: |
1407 | { |
1519 | { |
1408 | if (enable) |
1520 | if (enable) |
1409 | self->flags |= ix; |
1521 | self->flags |= ix; |
1410 | else |
1522 | else |
1411 | self->flags &= ~ix; |
1523 | self->flags &= ~ix; |
1412 | |
1524 | |
1413 | XPUSHs (ST (0)); |
1525 | XPUSHs (ST (0)); |
1414 | } |
1526 | } |
1415 | |
1527 | |
|
|
1528 | void get_ascii (JSON *self) |
|
|
1529 | ALIAS: |
|
|
1530 | get_ascii = F_ASCII |
|
|
1531 | get_latin1 = F_LATIN1 |
|
|
1532 | get_utf8 = F_UTF8 |
|
|
1533 | get_indent = F_INDENT |
|
|
1534 | get_canonical = F_CANONICAL |
|
|
1535 | get_space_before = F_SPACE_BEFORE |
|
|
1536 | get_space_after = F_SPACE_AFTER |
|
|
1537 | get_allow_nonref = F_ALLOW_NONREF |
|
|
1538 | get_shrink = F_SHRINK |
|
|
1539 | get_allow_blessed = F_ALLOW_BLESSED |
|
|
1540 | get_convert_blessed = F_CONV_BLESSED |
|
|
1541 | get_relaxed = F_RELAXED |
|
|
1542 | PPCODE: |
|
|
1543 | XPUSHs (boolSV (self->flags & ix)); |
|
|
1544 | |
1416 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1545 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1417 | PPCODE: |
1546 | PPCODE: |
1418 | { |
1547 | { |
1419 | UV log2 = 0; |
1548 | UV log2 = 0; |
1420 | |
1549 | |
… | |
… | |
1426 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1555 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1427 | |
1556 | |
1428 | XPUSHs (ST (0)); |
1557 | XPUSHs (ST (0)); |
1429 | } |
1558 | } |
1430 | |
1559 | |
|
|
1560 | U32 get_max_depth (JSON *self) |
|
|
1561 | CODE: |
|
|
1562 | RETVAL = DEC_DEPTH (self->flags); |
|
|
1563 | OUTPUT: |
|
|
1564 | RETVAL |
|
|
1565 | |
1431 | void max_size (JSON *self, UV max_size = 0) |
1566 | void max_size (JSON *self, UV max_size = 0) |
1432 | PPCODE: |
1567 | PPCODE: |
1433 | { |
1568 | { |
1434 | UV log2 = 0; |
1569 | UV log2 = 0; |
1435 | |
1570 | |
… | |
… | |
1441 | |
1576 | |
1442 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
1577 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
1443 | |
1578 | |
1444 | XPUSHs (ST (0)); |
1579 | XPUSHs (ST (0)); |
1445 | } |
1580 | } |
|
|
1581 | |
|
|
1582 | int get_max_size (JSON *self) |
|
|
1583 | CODE: |
|
|
1584 | RETVAL = DEC_SIZE (self->flags); |
|
|
1585 | OUTPUT: |
|
|
1586 | RETVAL |
1446 | |
1587 | |
1447 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1588 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1448 | PPCODE: |
1589 | PPCODE: |
1449 | { |
1590 | { |
1450 | SvREFCNT_dec (self->cb_object); |
1591 | SvREFCNT_dec (self->cb_object); |
… | |
… | |
1497 | SvREFCNT_dec (self->cb_sk_object); |
1638 | SvREFCNT_dec (self->cb_sk_object); |
1498 | SvREFCNT_dec (self->cb_object); |
1639 | SvREFCNT_dec (self->cb_object); |
1499 | |
1640 | |
1500 | PROTOTYPES: ENABLE |
1641 | PROTOTYPES: ENABLE |
1501 | |
1642 | |
1502 | void to_json (SV *scalar) |
1643 | void encode_json (SV *scalar) |
1503 | PPCODE: |
1644 | PPCODE: |
1504 | { |
1645 | { |
1505 | JSON json = { F_DEFAULT | F_UTF8 }; |
1646 | JSON json = { F_DEFAULT | F_UTF8 }; |
1506 | XPUSHs (encode_json (scalar, &json)); |
1647 | XPUSHs (encode_json (scalar, &json)); |
1507 | } |
1648 | } |
1508 | |
1649 | |
1509 | void from_json (SV *jsonstr) |
1650 | void decode_json (SV *jsonstr) |
1510 | PPCODE: |
1651 | PPCODE: |
1511 | { |
1652 | { |
1512 | JSON json = { F_DEFAULT | F_UTF8 }; |
1653 | JSON json = { F_DEFAULT | F_UTF8 }; |
1513 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1654 | XPUSHs (decode_json (jsonstr, &json, 0)); |
1514 | } |
1655 | } |