1 |
#include "EXTERN.h" |
2 |
#include "perl.h" |
3 |
#include "XSUB.h" |
4 |
|
5 |
#include "assert.h" |
6 |
#include "string.h" |
7 |
#include "stdlib.h" |
8 |
#include "stdio.h" |
9 |
|
10 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
11 |
# define snprintf _snprintf // C compilers have this in stdio.h |
12 |
#endif |
13 |
|
14 |
#define F_ASCII 0x00000001UL |
15 |
#define F_LATIN1 0x00000002UL |
16 |
#define F_UTF8 0x00000004UL |
17 |
#define F_INDENT 0x00000008UL |
18 |
#define F_CANONICAL 0x00000010UL |
19 |
#define F_SPACE_BEFORE 0x00000020UL |
20 |
#define F_SPACE_AFTER 0x00000040UL |
21 |
#define F_ALLOW_NONREF 0x00000100UL |
22 |
#define F_SHRINK 0x00000200UL |
23 |
#define F_MAXDEPTH 0xf8000000UL |
24 |
#define S_MAXDEPTH 27 |
25 |
|
26 |
#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
27 |
|
28 |
// F_SELFCONVERT? <=> to_json/toJson |
29 |
// F_BLESSED? <=> { $__class__$ => } |
30 |
|
31 |
#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
32 |
#define F_DEFAULT (9UL << S_MAXDEPTH) |
33 |
|
34 |
#define INIT_SIZE 32 // initial scalar size to be allocated |
35 |
#define INDENT_STEP 3 // spaces per indentation level |
36 |
|
37 |
#define SHORT_STRING_LEN 512 // special-case strings of up to this size |
38 |
|
39 |
#define SB do { |
40 |
#define SE } while (0) |
41 |
|
42 |
static HV *json_stash; // JSON::XS:: |
43 |
|
44 |
///////////////////////////////////////////////////////////////////////////// |
45 |
// utility functions |
46 |
|
47 |
static UV * |
48 |
SvJSON (SV *sv) |
49 |
{ |
50 |
if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
51 |
croak ("object is not of type JSON::XS"); |
52 |
|
53 |
return &SvUVX (SvRV (sv)); |
54 |
} |
55 |
|
56 |
static void |
57 |
shrink (SV *sv) |
58 |
{ |
59 |
sv_utf8_downgrade (sv, 1); |
60 |
if (SvLEN (sv) > SvCUR (sv) + 1) |
61 |
{ |
62 |
#ifdef SvPV_shrink_to_cur |
63 |
SvPV_shrink_to_cur (sv); |
64 |
#elif defined (SvPV_renew) |
65 |
SvPV_renew (sv, SvCUR (sv) + 1); |
66 |
#endif |
67 |
} |
68 |
} |
69 |
|
70 |
// decode an utf-8 character and return it, or (UV)-1 in |
71 |
// case of an error. |
72 |
// we special-case "safe" characters from U+80 .. U+7FF, |
73 |
// but use the very good perl function to parse anything else. |
74 |
// note that we never call this function for a ascii codepoints |
75 |
static UV |
76 |
decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
77 |
{ |
78 |
if (s[0] > 0xdf || s[0] < 0xc2) |
79 |
return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
80 |
else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
81 |
{ |
82 |
*clen = 2; |
83 |
return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
84 |
} |
85 |
else |
86 |
{ |
87 |
*clen = (STRLEN)-1; |
88 |
return (UV)-1; |
89 |
} |
90 |
} |
91 |
|
92 |
///////////////////////////////////////////////////////////////////////////// |
93 |
// encoder |
94 |
|
95 |
// structure used for encoding JSON |
96 |
typedef struct |
97 |
{ |
98 |
char *cur; // SvPVX (sv) + current output position |
99 |
char *end; // SvEND (sv) |
100 |
SV *sv; // result scalar |
101 |
U32 flags; // F_* |
102 |
U32 indent; // indentation level |
103 |
U32 maxdepth; // max. indentation/recursion level |
104 |
} enc_t; |
105 |
|
106 |
static void |
107 |
need (enc_t *enc, STRLEN len) |
108 |
{ |
109 |
if (enc->cur + len >= enc->end) |
110 |
{ |
111 |
STRLEN cur = enc->cur - SvPVX (enc->sv); |
112 |
SvGROW (enc->sv, cur + len + 1); |
113 |
enc->cur = SvPVX (enc->sv) + cur; |
114 |
enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
115 |
} |
116 |
} |
117 |
|
118 |
static void |
119 |
encode_ch (enc_t *enc, char ch) |
120 |
{ |
121 |
need (enc, 1); |
122 |
*enc->cur++ = ch; |
123 |
} |
124 |
|
125 |
static void |
126 |
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
127 |
{ |
128 |
char *end = str + len; |
129 |
|
130 |
need (enc, len); |
131 |
|
132 |
while (str < end) |
133 |
{ |
134 |
unsigned char ch = *(unsigned char *)str; |
135 |
|
136 |
if (ch >= 0x20 && ch < 0x80) // most common case |
137 |
{ |
138 |
if (ch == '"') // but with slow exceptions |
139 |
{ |
140 |
need (enc, len += 1); |
141 |
*enc->cur++ = '\\'; |
142 |
*enc->cur++ = '"'; |
143 |
} |
144 |
else if (ch == '\\') |
145 |
{ |
146 |
need (enc, len += 1); |
147 |
*enc->cur++ = '\\'; |
148 |
*enc->cur++ = '\\'; |
149 |
} |
150 |
else |
151 |
*enc->cur++ = ch; |
152 |
|
153 |
++str; |
154 |
} |
155 |
else |
156 |
{ |
157 |
switch (ch) |
158 |
{ |
159 |
case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
160 |
case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
161 |
case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
162 |
case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
163 |
case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
164 |
|
165 |
default: |
166 |
{ |
167 |
STRLEN clen; |
168 |
UV uch; |
169 |
|
170 |
if (is_utf8) |
171 |
{ |
172 |
//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
173 |
uch = decode_utf8 (str, end - str, &clen); |
174 |
if (clen == (STRLEN)-1) |
175 |
croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
176 |
} |
177 |
else |
178 |
{ |
179 |
uch = ch; |
180 |
clen = 1; |
181 |
} |
182 |
|
183 |
if (uch > 0x10FFFFUL) |
184 |
croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
185 |
|
186 |
if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF)) |
187 |
{ |
188 |
if (uch > 0xFFFFUL) |
189 |
{ |
190 |
need (enc, len += 11); |
191 |
sprintf (enc->cur, "\\u%04x\\u%04x", |
192 |
(int)((uch - 0x10000) / 0x400 + 0xD800), |
193 |
(int)((uch - 0x10000) % 0x400 + 0xDC00)); |
194 |
enc->cur += 12; |
195 |
} |
196 |
else |
197 |
{ |
198 |
static char hexdigit [16] = "0123456789abcdef"; |
199 |
need (enc, len += 5); |
200 |
*enc->cur++ = '\\'; |
201 |
*enc->cur++ = 'u'; |
202 |
*enc->cur++ = hexdigit [ uch >> 12 ]; |
203 |
*enc->cur++ = hexdigit [(uch >> 8) & 15]; |
204 |
*enc->cur++ = hexdigit [(uch >> 4) & 15]; |
205 |
*enc->cur++ = hexdigit [(uch >> 0) & 15]; |
206 |
} |
207 |
|
208 |
str += clen; |
209 |
} |
210 |
else if (enc->flags & F_LATIN1) |
211 |
{ |
212 |
*enc->cur++ = uch; |
213 |
str += clen; |
214 |
} |
215 |
else if (is_utf8) |
216 |
{ |
217 |
need (enc, len += clen); |
218 |
do |
219 |
{ |
220 |
*enc->cur++ = *str++; |
221 |
} |
222 |
while (--clen); |
223 |
} |
224 |
else |
225 |
{ |
226 |
need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
227 |
enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
228 |
++str; |
229 |
} |
230 |
} |
231 |
} |
232 |
} |
233 |
|
234 |
--len; |
235 |
} |
236 |
} |
237 |
|
238 |
static void |
239 |
encode_indent (enc_t *enc) |
240 |
{ |
241 |
if (enc->flags & F_INDENT) |
242 |
{ |
243 |
int spaces = enc->indent * INDENT_STEP; |
244 |
|
245 |
need (enc, spaces); |
246 |
memset (enc->cur, ' ', spaces); |
247 |
enc->cur += spaces; |
248 |
} |
249 |
} |
250 |
|
251 |
static void |
252 |
encode_space (enc_t *enc) |
253 |
{ |
254 |
need (enc, 1); |
255 |
encode_ch (enc, ' '); |
256 |
} |
257 |
|
258 |
static void |
259 |
encode_nl (enc_t *enc) |
260 |
{ |
261 |
if (enc->flags & F_INDENT) |
262 |
{ |
263 |
need (enc, 1); |
264 |
encode_ch (enc, '\n'); |
265 |
} |
266 |
} |
267 |
|
268 |
static void |
269 |
encode_comma (enc_t *enc) |
270 |
{ |
271 |
encode_ch (enc, ','); |
272 |
|
273 |
if (enc->flags & F_INDENT) |
274 |
encode_nl (enc); |
275 |
else if (enc->flags & F_SPACE_AFTER) |
276 |
encode_space (enc); |
277 |
} |
278 |
|
279 |
static void encode_sv (enc_t *enc, SV *sv); |
280 |
|
281 |
static void |
282 |
encode_av (enc_t *enc, AV *av) |
283 |
{ |
284 |
int i, len = av_len (av); |
285 |
|
286 |
if (enc->indent >= enc->maxdepth) |
287 |
croak ("data structure too deep (hit recursion limit)"); |
288 |
|
289 |
encode_ch (enc, '['); encode_nl (enc); |
290 |
++enc->indent; |
291 |
|
292 |
for (i = 0; i <= len; ++i) |
293 |
{ |
294 |
encode_indent (enc); |
295 |
encode_sv (enc, *av_fetch (av, i, 0)); |
296 |
|
297 |
if (i < len) |
298 |
encode_comma (enc); |
299 |
} |
300 |
|
301 |
encode_nl (enc); |
302 |
|
303 |
--enc->indent; |
304 |
encode_indent (enc); encode_ch (enc, ']'); |
305 |
} |
306 |
|
307 |
static void |
308 |
encode_he (enc_t *enc, HE *he) |
309 |
{ |
310 |
encode_ch (enc, '"'); |
311 |
|
312 |
if (HeKLEN (he) == HEf_SVKEY) |
313 |
{ |
314 |
SV *sv = HeSVKEY (he); |
315 |
STRLEN len; |
316 |
char *str; |
317 |
|
318 |
SvGETMAGIC (sv); |
319 |
str = SvPV (sv, len); |
320 |
|
321 |
encode_str (enc, str, len, SvUTF8 (sv)); |
322 |
} |
323 |
else |
324 |
encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
325 |
|
326 |
encode_ch (enc, '"'); |
327 |
|
328 |
if (enc->flags & F_SPACE_BEFORE) encode_space (enc); |
329 |
encode_ch (enc, ':'); |
330 |
if (enc->flags & F_SPACE_AFTER ) encode_space (enc); |
331 |
encode_sv (enc, HeVAL (he)); |
332 |
} |
333 |
|
334 |
// compare hash entries, used when all keys are bytestrings |
335 |
static int |
336 |
he_cmp_fast (const void *a_, const void *b_) |
337 |
{ |
338 |
int cmp; |
339 |
|
340 |
HE *a = *(HE **)a_; |
341 |
HE *b = *(HE **)b_; |
342 |
|
343 |
STRLEN la = HeKLEN (a); |
344 |
STRLEN lb = HeKLEN (b); |
345 |
|
346 |
if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
347 |
cmp = la - lb; |
348 |
|
349 |
return cmp; |
350 |
} |
351 |
|
352 |
// compare hash entries, used when some keys are sv's or utf-x |
353 |
static int |
354 |
he_cmp_slow (const void *a, const void *b) |
355 |
{ |
356 |
return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
357 |
} |
358 |
|
359 |
static void |
360 |
encode_hv (enc_t *enc, HV *hv) |
361 |
{ |
362 |
int count, i; |
363 |
|
364 |
if (enc->indent >= enc->maxdepth) |
365 |
croak ("data structure too deep (hit recursion limit)"); |
366 |
|
367 |
encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
368 |
|
369 |
if ((count = hv_iterinit (hv))) |
370 |
{ |
371 |
// for canonical output we have to sort by keys first |
372 |
// actually, this is mostly due to the stupid so-called |
373 |
// security workaround added somewhere in 5.8.x. |
374 |
// that randomises hash orderings |
375 |
if (enc->flags & F_CANONICAL) |
376 |
{ |
377 |
HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode |
378 |
int fast = 1; |
379 |
|
380 |
i = 0; |
381 |
while ((he = hv_iternext (hv))) |
382 |
{ |
383 |
hes [i++] = he; |
384 |
if (HeKLEN (he) < 0 || HeKUTF8 (he)) |
385 |
fast = 0; |
386 |
} |
387 |
|
388 |
assert (i == count); |
389 |
|
390 |
if (fast) |
391 |
qsort (hes, count, sizeof (HE *), he_cmp_fast); |
392 |
else |
393 |
{ |
394 |
// hack to forcefully disable "use bytes" |
395 |
COP cop = *PL_curcop; |
396 |
cop.op_private = 0; |
397 |
|
398 |
ENTER; |
399 |
SAVETMPS; |
400 |
|
401 |
SAVEVPTR (PL_curcop); |
402 |
PL_curcop = &cop; |
403 |
|
404 |
qsort (hes, count, sizeof (HE *), he_cmp_slow); |
405 |
|
406 |
FREETMPS; |
407 |
LEAVE; |
408 |
} |
409 |
|
410 |
for (i = 0; i < count; ++i) |
411 |
{ |
412 |
encode_indent (enc); |
413 |
encode_he (enc, hes [i]); |
414 |
|
415 |
if (i < count - 1) |
416 |
encode_comma (enc); |
417 |
} |
418 |
|
419 |
encode_nl (enc); |
420 |
} |
421 |
else |
422 |
{ |
423 |
HE *he = hv_iternext (hv); |
424 |
|
425 |
for (;;) |
426 |
{ |
427 |
encode_indent (enc); |
428 |
encode_he (enc, he); |
429 |
|
430 |
if (!(he = hv_iternext (hv))) |
431 |
break; |
432 |
|
433 |
encode_comma (enc); |
434 |
} |
435 |
|
436 |
encode_nl (enc); |
437 |
} |
438 |
} |
439 |
|
440 |
--enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
441 |
} |
442 |
|
443 |
// encode objects, arrays and special \0=false and \1=true values. |
444 |
static void |
445 |
encode_rv (enc_t *enc, SV *sv) |
446 |
{ |
447 |
svtype svt; |
448 |
|
449 |
SvGETMAGIC (sv); |
450 |
svt = SvTYPE (sv); |
451 |
|
452 |
if (svt == SVt_PVHV) |
453 |
encode_hv (enc, (HV *)sv); |
454 |
else if (svt == SVt_PVAV) |
455 |
encode_av (enc, (AV *)sv); |
456 |
else if (svt < SVt_PVAV) |
457 |
{ |
458 |
if (SvNIOK (sv) && SvIV (sv) == 0) |
459 |
encode_str (enc, "false", 5, 0); |
460 |
else if (SvNIOK (sv) && SvIV (sv) == 1) |
461 |
encode_str (enc, "true", 4, 0); |
462 |
else |
463 |
croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
464 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
465 |
} |
466 |
else |
467 |
croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
468 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
469 |
} |
470 |
|
471 |
static void |
472 |
encode_sv (enc_t *enc, SV *sv) |
473 |
{ |
474 |
SvGETMAGIC (sv); |
475 |
|
476 |
if (SvPOKp (sv)) |
477 |
{ |
478 |
STRLEN len; |
479 |
char *str = SvPV (sv, len); |
480 |
encode_ch (enc, '"'); |
481 |
encode_str (enc, str, len, SvUTF8 (sv)); |
482 |
encode_ch (enc, '"'); |
483 |
} |
484 |
else if (SvNOKp (sv)) |
485 |
{ |
486 |
need (enc, NV_DIG + 32); |
487 |
Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
488 |
enc->cur += strlen (enc->cur); |
489 |
} |
490 |
else if (SvIOKp (sv)) |
491 |
{ |
492 |
need (enc, 64); |
493 |
enc->cur += |
494 |
SvIsUV(sv) |
495 |
? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
496 |
: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
497 |
} |
498 |
else if (SvROK (sv)) |
499 |
encode_rv (enc, SvRV (sv)); |
500 |
else if (!SvOK (sv)) |
501 |
encode_str (enc, "null", 4, 0); |
502 |
else |
503 |
croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
504 |
SvPV_nolen (sv), SvFLAGS (sv)); |
505 |
} |
506 |
|
507 |
static SV * |
508 |
encode_json (SV *scalar, U32 flags) |
509 |
{ |
510 |
enc_t enc; |
511 |
|
512 |
if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
513 |
croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
514 |
|
515 |
enc.flags = flags; |
516 |
enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
517 |
enc.cur = SvPVX (enc.sv); |
518 |
enc.end = SvEND (enc.sv); |
519 |
enc.indent = 0; |
520 |
enc.maxdepth = DEC_DEPTH (flags); |
521 |
|
522 |
SvPOK_only (enc.sv); |
523 |
encode_sv (&enc, scalar); |
524 |
|
525 |
SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
526 |
*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
527 |
|
528 |
if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
529 |
SvUTF8_on (enc.sv); |
530 |
|
531 |
if (enc.flags & F_SHRINK) |
532 |
shrink (enc.sv); |
533 |
|
534 |
return enc.sv; |
535 |
} |
536 |
|
537 |
///////////////////////////////////////////////////////////////////////////// |
538 |
// decoder |
539 |
|
540 |
// structure used for decoding JSON |
541 |
typedef struct |
542 |
{ |
543 |
char *cur; // current parser pointer |
544 |
char *end; // end of input string |
545 |
const char *err; // parse error, if != 0 |
546 |
U32 flags; // F_* |
547 |
U32 depth; // recursion depth |
548 |
U32 maxdepth; // recursion depth limit |
549 |
} dec_t; |
550 |
|
551 |
static void |
552 |
decode_ws (dec_t *dec) |
553 |
{ |
554 |
for (;;) |
555 |
{ |
556 |
char ch = *dec->cur; |
557 |
|
558 |
if (ch > 0x20 |
559 |
|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
560 |
break; |
561 |
|
562 |
++dec->cur; |
563 |
} |
564 |
} |
565 |
|
566 |
#define ERR(reason) SB dec->err = reason; goto fail; SE |
567 |
|
568 |
#define EXPECT_CH(ch) SB \ |
569 |
if (*dec->cur != ch) \ |
570 |
ERR (# ch " expected"); \ |
571 |
++dec->cur; \ |
572 |
SE |
573 |
|
574 |
#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
575 |
#define DEC_DEC_DEPTH --dec->depth |
576 |
|
577 |
static SV *decode_sv (dec_t *dec); |
578 |
|
579 |
static signed char decode_hexdigit[256]; |
580 |
|
581 |
static UV |
582 |
decode_4hex (dec_t *dec) |
583 |
{ |
584 |
signed char d1, d2, d3, d4; |
585 |
unsigned char *cur = (unsigned char *)dec->cur; |
586 |
|
587 |
d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected"); |
588 |
d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected"); |
589 |
d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected"); |
590 |
d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected"); |
591 |
|
592 |
dec->cur += 4; |
593 |
|
594 |
return ((UV)d1) << 12 |
595 |
| ((UV)d2) << 8 |
596 |
| ((UV)d3) << 4 |
597 |
| ((UV)d4); |
598 |
|
599 |
fail: |
600 |
return (UV)-1; |
601 |
} |
602 |
|
603 |
static SV * |
604 |
decode_str (dec_t *dec) |
605 |
{ |
606 |
SV *sv = 0; |
607 |
int utf8 = 0; |
608 |
|
609 |
do |
610 |
{ |
611 |
char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
612 |
char *cur = buf; |
613 |
|
614 |
do |
615 |
{ |
616 |
unsigned char ch = *(unsigned char *)dec->cur++; |
617 |
|
618 |
if (ch == '"') |
619 |
{ |
620 |
--dec->cur; |
621 |
break; |
622 |
} |
623 |
else if (ch == '\\') |
624 |
{ |
625 |
switch (*dec->cur) |
626 |
{ |
627 |
case '\\': |
628 |
case '/': |
629 |
case '"': *cur++ = *dec->cur++; break; |
630 |
|
631 |
case 'b': ++dec->cur; *cur++ = '\010'; break; |
632 |
case 't': ++dec->cur; *cur++ = '\011'; break; |
633 |
case 'n': ++dec->cur; *cur++ = '\012'; break; |
634 |
case 'f': ++dec->cur; *cur++ = '\014'; break; |
635 |
case 'r': ++dec->cur; *cur++ = '\015'; break; |
636 |
|
637 |
case 'u': |
638 |
{ |
639 |
UV lo, hi; |
640 |
++dec->cur; |
641 |
|
642 |
hi = decode_4hex (dec); |
643 |
if (hi == (UV)-1) |
644 |
goto fail; |
645 |
|
646 |
// possibly a surrogate pair |
647 |
if (hi >= 0xd800) |
648 |
if (hi < 0xdc00) |
649 |
{ |
650 |
if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
651 |
ERR ("missing low surrogate character in surrogate pair"); |
652 |
|
653 |
dec->cur += 2; |
654 |
|
655 |
lo = decode_4hex (dec); |
656 |
if (lo == (UV)-1) |
657 |
goto fail; |
658 |
|
659 |
if (lo < 0xdc00 || lo >= 0xe000) |
660 |
ERR ("surrogate pair expected"); |
661 |
|
662 |
hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
663 |
} |
664 |
else if (hi < 0xe000) |
665 |
ERR ("missing high surrogate character in surrogate pair"); |
666 |
|
667 |
if (hi >= 0x80) |
668 |
{ |
669 |
utf8 = 1; |
670 |
|
671 |
cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
672 |
} |
673 |
else |
674 |
*cur++ = hi; |
675 |
} |
676 |
break; |
677 |
|
678 |
default: |
679 |
--dec->cur; |
680 |
ERR ("illegal backslash escape sequence in string"); |
681 |
} |
682 |
} |
683 |
else if (ch >= 0x20 && ch <= 0x7f) |
684 |
*cur++ = ch; |
685 |
else if (ch >= 0x80) |
686 |
{ |
687 |
STRLEN clen; |
688 |
UV uch; |
689 |
|
690 |
--dec->cur; |
691 |
|
692 |
uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen); |
693 |
if (clen == (STRLEN)-1) |
694 |
ERR ("malformed UTF-8 character in JSON string"); |
695 |
|
696 |
do |
697 |
*cur++ = *dec->cur++; |
698 |
while (--clen); |
699 |
|
700 |
utf8 = 1; |
701 |
} |
702 |
else |
703 |
{ |
704 |
--dec->cur; |
705 |
|
706 |
if (!ch) |
707 |
ERR ("unexpected end of string while parsing JSON string"); |
708 |
else |
709 |
ERR ("invalid character encountered while parsing JSON string"); |
710 |
} |
711 |
} |
712 |
while (cur < buf + SHORT_STRING_LEN); |
713 |
|
714 |
{ |
715 |
STRLEN len = cur - buf; |
716 |
|
717 |
if (sv) |
718 |
{ |
719 |
SvGROW (sv, SvCUR (sv) + len + 1); |
720 |
memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
721 |
SvCUR_set (sv, SvCUR (sv) + len); |
722 |
} |
723 |
else |
724 |
sv = newSVpvn (buf, len); |
725 |
} |
726 |
} |
727 |
while (*dec->cur != '"'); |
728 |
|
729 |
++dec->cur; |
730 |
|
731 |
if (sv) |
732 |
{ |
733 |
SvPOK_only (sv); |
734 |
*SvEND (sv) = 0; |
735 |
|
736 |
if (utf8) |
737 |
SvUTF8_on (sv); |
738 |
} |
739 |
else |
740 |
sv = newSVpvn ("", 0); |
741 |
|
742 |
return sv; |
743 |
|
744 |
fail: |
745 |
return 0; |
746 |
} |
747 |
|
748 |
static SV * |
749 |
decode_num (dec_t *dec) |
750 |
{ |
751 |
int is_nv = 0; |
752 |
char *start = dec->cur; |
753 |
|
754 |
// [minus] |
755 |
if (*dec->cur == '-') |
756 |
++dec->cur; |
757 |
|
758 |
if (*dec->cur == '0') |
759 |
{ |
760 |
++dec->cur; |
761 |
if (*dec->cur >= '0' && *dec->cur <= '9') |
762 |
ERR ("malformed number (leading zero must not be followed by another digit)"); |
763 |
} |
764 |
else if (*dec->cur < '0' || *dec->cur > '9') |
765 |
ERR ("malformed number (no digits after initial minus)"); |
766 |
else |
767 |
do |
768 |
{ |
769 |
++dec->cur; |
770 |
} |
771 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
772 |
|
773 |
// [frac] |
774 |
if (*dec->cur == '.') |
775 |
{ |
776 |
++dec->cur; |
777 |
|
778 |
if (*dec->cur < '0' || *dec->cur > '9') |
779 |
ERR ("malformed number (no digits after decimal point)"); |
780 |
|
781 |
do |
782 |
{ |
783 |
++dec->cur; |
784 |
} |
785 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
786 |
|
787 |
is_nv = 1; |
788 |
} |
789 |
|
790 |
// [exp] |
791 |
if (*dec->cur == 'e' || *dec->cur == 'E') |
792 |
{ |
793 |
++dec->cur; |
794 |
|
795 |
if (*dec->cur == '-' || *dec->cur == '+') |
796 |
++dec->cur; |
797 |
|
798 |
if (*dec->cur < '0' || *dec->cur > '9') |
799 |
ERR ("malformed number (no digits after exp sign)"); |
800 |
|
801 |
do |
802 |
{ |
803 |
++dec->cur; |
804 |
} |
805 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
806 |
|
807 |
is_nv = 1; |
808 |
} |
809 |
|
810 |
if (!is_nv) |
811 |
{ |
812 |
UV uv; |
813 |
int numtype = grok_number (start, dec->cur - start, &uv); |
814 |
if (numtype & IS_NUMBER_IN_UV) |
815 |
if (numtype & IS_NUMBER_NEG) |
816 |
{ |
817 |
if (uv < (UV)IV_MIN) |
818 |
return newSViv (-(IV)uv); |
819 |
} |
820 |
else |
821 |
return newSVuv (uv); |
822 |
} |
823 |
|
824 |
return newSVnv (Atof (start)); |
825 |
|
826 |
fail: |
827 |
return 0; |
828 |
} |
829 |
|
830 |
static SV * |
831 |
decode_av (dec_t *dec) |
832 |
{ |
833 |
AV *av = newAV (); |
834 |
|
835 |
DEC_INC_DEPTH; |
836 |
decode_ws (dec); |
837 |
|
838 |
if (*dec->cur == ']') |
839 |
++dec->cur; |
840 |
else |
841 |
for (;;) |
842 |
{ |
843 |
SV *value; |
844 |
|
845 |
value = decode_sv (dec); |
846 |
if (!value) |
847 |
goto fail; |
848 |
|
849 |
av_push (av, value); |
850 |
|
851 |
decode_ws (dec); |
852 |
|
853 |
if (*dec->cur == ']') |
854 |
{ |
855 |
++dec->cur; |
856 |
break; |
857 |
} |
858 |
|
859 |
if (*dec->cur != ',') |
860 |
ERR (", or ] expected while parsing array"); |
861 |
|
862 |
++dec->cur; |
863 |
} |
864 |
|
865 |
DEC_DEC_DEPTH; |
866 |
return newRV_noinc ((SV *)av); |
867 |
|
868 |
fail: |
869 |
SvREFCNT_dec (av); |
870 |
DEC_DEC_DEPTH; |
871 |
return 0; |
872 |
} |
873 |
|
874 |
static SV * |
875 |
decode_hv (dec_t *dec) |
876 |
{ |
877 |
HV *hv = newHV (); |
878 |
|
879 |
DEC_INC_DEPTH; |
880 |
decode_ws (dec); |
881 |
|
882 |
if (*dec->cur == '}') |
883 |
++dec->cur; |
884 |
else |
885 |
for (;;) |
886 |
{ |
887 |
SV *key, *value; |
888 |
|
889 |
decode_ws (dec); EXPECT_CH ('"'); |
890 |
|
891 |
key = decode_str (dec); |
892 |
if (!key) |
893 |
goto fail; |
894 |
|
895 |
decode_ws (dec); EXPECT_CH (':'); |
896 |
|
897 |
value = decode_sv (dec); |
898 |
if (!value) |
899 |
{ |
900 |
SvREFCNT_dec (key); |
901 |
goto fail; |
902 |
} |
903 |
|
904 |
hv_store_ent (hv, key, value, 0); |
905 |
SvREFCNT_dec (key); |
906 |
|
907 |
decode_ws (dec); |
908 |
|
909 |
if (*dec->cur == '}') |
910 |
{ |
911 |
++dec->cur; |
912 |
break; |
913 |
} |
914 |
|
915 |
if (*dec->cur != ',') |
916 |
ERR (", or } expected while parsing object/hash"); |
917 |
|
918 |
++dec->cur; |
919 |
} |
920 |
|
921 |
DEC_DEC_DEPTH; |
922 |
return newRV_noinc ((SV *)hv); |
923 |
|
924 |
fail: |
925 |
SvREFCNT_dec (hv); |
926 |
DEC_DEC_DEPTH; |
927 |
return 0; |
928 |
} |
929 |
|
930 |
static SV * |
931 |
decode_sv (dec_t *dec) |
932 |
{ |
933 |
decode_ws (dec); |
934 |
switch (*dec->cur) |
935 |
{ |
936 |
case '"': ++dec->cur; return decode_str (dec); |
937 |
case '[': ++dec->cur; return decode_av (dec); |
938 |
case '{': ++dec->cur; return decode_hv (dec); |
939 |
|
940 |
case '-': |
941 |
case '0': case '1': case '2': case '3': case '4': |
942 |
case '5': case '6': case '7': case '8': case '9': |
943 |
return decode_num (dec); |
944 |
|
945 |
case 't': |
946 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
947 |
{ |
948 |
dec->cur += 4; |
949 |
return newSViv (1); |
950 |
} |
951 |
else |
952 |
ERR ("'true' expected"); |
953 |
|
954 |
break; |
955 |
|
956 |
case 'f': |
957 |
if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
958 |
{ |
959 |
dec->cur += 5; |
960 |
return newSViv (0); |
961 |
} |
962 |
else |
963 |
ERR ("'false' expected"); |
964 |
|
965 |
break; |
966 |
|
967 |
case 'n': |
968 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
969 |
{ |
970 |
dec->cur += 4; |
971 |
return newSVsv (&PL_sv_undef); |
972 |
} |
973 |
else |
974 |
ERR ("'null' expected"); |
975 |
|
976 |
break; |
977 |
|
978 |
default: |
979 |
ERR ("malformed JSON string, neither array, object, number, string or atom"); |
980 |
break; |
981 |
} |
982 |
|
983 |
fail: |
984 |
return 0; |
985 |
} |
986 |
|
987 |
static SV * |
988 |
decode_json (SV *string, U32 flags, UV *offset_return) |
989 |
{ |
990 |
dec_t dec; |
991 |
UV offset; |
992 |
SV *sv; |
993 |
|
994 |
SvGETMAGIC (string); |
995 |
SvUPGRADE (string, SVt_PV); |
996 |
|
997 |
if (flags & F_UTF8) |
998 |
sv_utf8_downgrade (string, 0); |
999 |
else |
1000 |
sv_utf8_upgrade (string); |
1001 |
|
1002 |
SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1003 |
|
1004 |
dec.flags = flags; |
1005 |
dec.cur = SvPVX (string); |
1006 |
dec.end = SvEND (string); |
1007 |
dec.err = 0; |
1008 |
dec.depth = 0; |
1009 |
dec.maxdepth = DEC_DEPTH (dec.flags); |
1010 |
|
1011 |
*dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1012 |
sv = decode_sv (&dec); |
1013 |
|
1014 |
if (offset_return || !sv) |
1015 |
{ |
1016 |
offset = dec.flags & F_UTF8 |
1017 |
? dec.cur - SvPVX (string) |
1018 |
: utf8_distance (dec.cur, SvPVX (string)); |
1019 |
|
1020 |
if (offset_return) |
1021 |
*offset_return = offset; |
1022 |
} |
1023 |
else |
1024 |
{ |
1025 |
// check for trailing garbage |
1026 |
decode_ws (&dec); |
1027 |
|
1028 |
if (*dec.cur) |
1029 |
{ |
1030 |
dec.err = "garbage after JSON object"; |
1031 |
SvREFCNT_dec (sv); |
1032 |
sv = 0; |
1033 |
} |
1034 |
} |
1035 |
|
1036 |
if (!sv) |
1037 |
{ |
1038 |
SV *uni = sv_newmortal (); |
1039 |
|
1040 |
// horrible hack to silence warning inside pv_uni_display |
1041 |
COP cop = *PL_curcop; |
1042 |
cop.cop_warnings = pWARN_NONE; |
1043 |
ENTER; |
1044 |
SAVEVPTR (PL_curcop); |
1045 |
PL_curcop = &cop; |
1046 |
pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1047 |
LEAVE; |
1048 |
|
1049 |
croak ("%s, at character offset %d [\"%s\"]", |
1050 |
dec.err, |
1051 |
(int)offset, |
1052 |
dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1053 |
} |
1054 |
|
1055 |
sv = sv_2mortal (sv); |
1056 |
|
1057 |
if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1058 |
croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1059 |
|
1060 |
return sv; |
1061 |
} |
1062 |
|
1063 |
///////////////////////////////////////////////////////////////////////////// |
1064 |
// XS interface functions |
1065 |
|
1066 |
MODULE = JSON::XS PACKAGE = JSON::XS |
1067 |
|
1068 |
BOOT: |
1069 |
{ |
1070 |
int i; |
1071 |
|
1072 |
memset (decode_hexdigit, 0xff, 256); |
1073 |
|
1074 |
for (i = 0; i < 256; ++i) |
1075 |
decode_hexdigit [i] = |
1076 |
i >= '0' && i <= '9' ? i - '0' |
1077 |
: i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1078 |
: i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1079 |
: -1; |
1080 |
|
1081 |
json_stash = gv_stashpv ("JSON::XS", 1); |
1082 |
} |
1083 |
|
1084 |
PROTOTYPES: DISABLE |
1085 |
|
1086 |
SV *new (char *dummy) |
1087 |
CODE: |
1088 |
RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1089 |
OUTPUT: |
1090 |
RETVAL |
1091 |
|
1092 |
SV *ascii (SV *self, int enable = 1) |
1093 |
ALIAS: |
1094 |
ascii = F_ASCII |
1095 |
latin1 = F_LATIN1 |
1096 |
utf8 = F_UTF8 |
1097 |
indent = F_INDENT |
1098 |
canonical = F_CANONICAL |
1099 |
space_before = F_SPACE_BEFORE |
1100 |
space_after = F_SPACE_AFTER |
1101 |
pretty = F_PRETTY |
1102 |
allow_nonref = F_ALLOW_NONREF |
1103 |
shrink = F_SHRINK |
1104 |
CODE: |
1105 |
{ |
1106 |
UV *uv = SvJSON (self); |
1107 |
if (enable) |
1108 |
*uv |= ix; |
1109 |
else |
1110 |
*uv &= ~ix; |
1111 |
|
1112 |
RETVAL = newSVsv (self); |
1113 |
} |
1114 |
OUTPUT: |
1115 |
RETVAL |
1116 |
|
1117 |
SV *max_depth (SV *self, UV max_depth = 0x80000000UL) |
1118 |
CODE: |
1119 |
{ |
1120 |
UV *uv = SvJSON (self); |
1121 |
UV log2 = 0; |
1122 |
|
1123 |
if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1124 |
|
1125 |
while ((1UL << log2) < max_depth) |
1126 |
++log2; |
1127 |
|
1128 |
*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1129 |
|
1130 |
RETVAL = newSVsv (self); |
1131 |
} |
1132 |
OUTPUT: |
1133 |
RETVAL |
1134 |
|
1135 |
void encode (SV *self, SV *scalar) |
1136 |
PPCODE: |
1137 |
XPUSHs (encode_json (scalar, *SvJSON (self))); |
1138 |
|
1139 |
void decode (SV *self, SV *jsonstr) |
1140 |
PPCODE: |
1141 |
XPUSHs (decode_json (jsonstr, *SvJSON (self), 0)); |
1142 |
|
1143 |
void decode_prefix (SV *self, SV *jsonstr) |
1144 |
PPCODE: |
1145 |
{ |
1146 |
UV offset; |
1147 |
EXTEND (SP, 2); |
1148 |
PUSHs (decode_json (jsonstr, *SvJSON (self), &offset)); |
1149 |
PUSHs (sv_2mortal (newSVuv (offset))); |
1150 |
} |
1151 |
|
1152 |
PROTOTYPES: ENABLE |
1153 |
|
1154 |
void to_json (SV *scalar) |
1155 |
ALIAS: |
1156 |
objToJson = 0 |
1157 |
PPCODE: |
1158 |
XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8)); |
1159 |
|
1160 |
void from_json (SV *jsonstr) |
1161 |
ALIAS: |
1162 |
jsonToObj = 0 |
1163 |
PPCODE: |
1164 |
XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0)); |
1165 |
|