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 |
#include <float.h> |
10 |
|
11 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
12 |
# define snprintf _snprintf // C compilers have this in stdio.h |
13 |
#endif |
14 |
|
15 |
// some old perls do not have this, try to make it work, no |
16 |
// guarentees, though. if it breaks, you get to keep the pieces. |
17 |
#ifndef UTF8_MAXBYTES |
18 |
# define UTF8_MAXBYTES 13 |
19 |
#endif |
20 |
|
21 |
#define F_ASCII 0x00000001UL |
22 |
#define F_LATIN1 0x00000002UL |
23 |
#define F_UTF8 0x00000004UL |
24 |
#define F_INDENT 0x00000008UL |
25 |
#define F_CANONICAL 0x00000010UL |
26 |
#define F_SPACE_BEFORE 0x00000020UL |
27 |
#define F_SPACE_AFTER 0x00000040UL |
28 |
#define F_ALLOW_NONREF 0x00000100UL |
29 |
#define F_SHRINK 0x00000200UL |
30 |
#define F_ALLOW_BLESSED 0x00000400UL |
31 |
#define F_CONV_BLESSED 0x00000800UL |
32 |
#define F_MAXDEPTH 0xf8000000UL |
33 |
#define S_MAXDEPTH 27 |
34 |
#define F_MAXSIZE 0x01f00000UL |
35 |
#define S_MAXSIZE 20 |
36 |
#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
37 |
|
38 |
#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
39 |
#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
40 |
|
41 |
#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
42 |
#define F_DEFAULT (9UL << S_MAXDEPTH) |
43 |
|
44 |
#define INIT_SIZE 32 // initial scalar size to be allocated |
45 |
#define INDENT_STEP 3 // spaces per indentation level |
46 |
|
47 |
#define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
48 |
|
49 |
#define SB do { |
50 |
#define SE } while (0) |
51 |
|
52 |
#if __GNUC__ >= 3 |
53 |
# define expect(expr,value) __builtin_expect ((expr),(value)) |
54 |
# define inline inline |
55 |
#else |
56 |
# define expect(expr,value) (expr) |
57 |
# define inline static |
58 |
#endif |
59 |
|
60 |
#define expect_false(expr) expect ((expr) != 0, 0) |
61 |
#define expect_true(expr) expect ((expr) != 0, 1) |
62 |
|
63 |
static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
64 |
static SV *json_true, *json_false; |
65 |
|
66 |
typedef struct { |
67 |
U32 flags; |
68 |
SV *cb_object; |
69 |
HV *cb_sk_object; |
70 |
} JSON; |
71 |
|
72 |
///////////////////////////////////////////////////////////////////////////// |
73 |
// utility functions |
74 |
|
75 |
inline void |
76 |
shrink (SV *sv) |
77 |
{ |
78 |
sv_utf8_downgrade (sv, 1); |
79 |
if (SvLEN (sv) > SvCUR (sv) + 1) |
80 |
{ |
81 |
#ifdef SvPV_shrink_to_cur |
82 |
SvPV_shrink_to_cur (sv); |
83 |
#elif defined (SvPV_renew) |
84 |
SvPV_renew (sv, SvCUR (sv) + 1); |
85 |
#endif |
86 |
} |
87 |
} |
88 |
|
89 |
// decode an utf-8 character and return it, or (UV)-1 in |
90 |
// case of an error. |
91 |
// we special-case "safe" characters from U+80 .. U+7FF, |
92 |
// but use the very good perl function to parse anything else. |
93 |
// note that we never call this function for a ascii codepoints |
94 |
inline UV |
95 |
decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
96 |
{ |
97 |
if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
98 |
return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
99 |
else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
100 |
{ |
101 |
*clen = 2; |
102 |
return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
103 |
} |
104 |
else |
105 |
{ |
106 |
*clen = (STRLEN)-1; |
107 |
return (UV)-1; |
108 |
} |
109 |
} |
110 |
|
111 |
///////////////////////////////////////////////////////////////////////////// |
112 |
// encoder |
113 |
|
114 |
// structure used for encoding JSON |
115 |
typedef struct |
116 |
{ |
117 |
char *cur; // SvPVX (sv) + current output position |
118 |
char *end; // SvEND (sv) |
119 |
SV *sv; // result scalar |
120 |
JSON json; |
121 |
U32 indent; // indentation level |
122 |
U32 maxdepth; // max. indentation/recursion level |
123 |
} enc_t; |
124 |
|
125 |
inline void |
126 |
need (enc_t *enc, STRLEN len) |
127 |
{ |
128 |
if (expect_false (enc->cur + len >= enc->end)) |
129 |
{ |
130 |
STRLEN cur = enc->cur - SvPVX (enc->sv); |
131 |
SvGROW (enc->sv, cur + len + 1); |
132 |
enc->cur = SvPVX (enc->sv) + cur; |
133 |
enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
134 |
} |
135 |
} |
136 |
|
137 |
inline void |
138 |
encode_ch (enc_t *enc, char ch) |
139 |
{ |
140 |
need (enc, 1); |
141 |
*enc->cur++ = ch; |
142 |
} |
143 |
|
144 |
static void |
145 |
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
146 |
{ |
147 |
char *end = str + len; |
148 |
|
149 |
need (enc, len); |
150 |
|
151 |
while (str < end) |
152 |
{ |
153 |
unsigned char ch = *(unsigned char *)str; |
154 |
|
155 |
if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
156 |
{ |
157 |
if (expect_false (ch == '"')) // but with slow exceptions |
158 |
{ |
159 |
need (enc, len += 1); |
160 |
*enc->cur++ = '\\'; |
161 |
*enc->cur++ = '"'; |
162 |
} |
163 |
else if (expect_false (ch == '\\')) |
164 |
{ |
165 |
need (enc, len += 1); |
166 |
*enc->cur++ = '\\'; |
167 |
*enc->cur++ = '\\'; |
168 |
} |
169 |
else |
170 |
*enc->cur++ = ch; |
171 |
|
172 |
++str; |
173 |
} |
174 |
else |
175 |
{ |
176 |
switch (ch) |
177 |
{ |
178 |
case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
179 |
case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
180 |
case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
181 |
case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
182 |
case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
183 |
|
184 |
default: |
185 |
{ |
186 |
STRLEN clen; |
187 |
UV uch; |
188 |
|
189 |
if (is_utf8) |
190 |
{ |
191 |
uch = decode_utf8 (str, end - str, &clen); |
192 |
if (clen == (STRLEN)-1) |
193 |
croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
194 |
} |
195 |
else |
196 |
{ |
197 |
uch = ch; |
198 |
clen = 1; |
199 |
} |
200 |
|
201 |
if (uch > 0x10FFFFUL) |
202 |
croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
203 |
|
204 |
if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
205 |
{ |
206 |
if (uch > 0xFFFFUL) |
207 |
{ |
208 |
need (enc, len += 11); |
209 |
sprintf (enc->cur, "\\u%04x\\u%04x", |
210 |
(int)((uch - 0x10000) / 0x400 + 0xD800), |
211 |
(int)((uch - 0x10000) % 0x400 + 0xDC00)); |
212 |
enc->cur += 12; |
213 |
} |
214 |
else |
215 |
{ |
216 |
static char hexdigit [16] = "0123456789abcdef"; |
217 |
need (enc, len += 5); |
218 |
*enc->cur++ = '\\'; |
219 |
*enc->cur++ = 'u'; |
220 |
*enc->cur++ = hexdigit [ uch >> 12 ]; |
221 |
*enc->cur++ = hexdigit [(uch >> 8) & 15]; |
222 |
*enc->cur++ = hexdigit [(uch >> 4) & 15]; |
223 |
*enc->cur++ = hexdigit [(uch >> 0) & 15]; |
224 |
} |
225 |
|
226 |
str += clen; |
227 |
} |
228 |
else if (enc->json.flags & F_LATIN1) |
229 |
{ |
230 |
*enc->cur++ = uch; |
231 |
str += clen; |
232 |
} |
233 |
else if (is_utf8) |
234 |
{ |
235 |
need (enc, len += clen); |
236 |
do |
237 |
{ |
238 |
*enc->cur++ = *str++; |
239 |
} |
240 |
while (--clen); |
241 |
} |
242 |
else |
243 |
{ |
244 |
need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
245 |
enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
246 |
++str; |
247 |
} |
248 |
} |
249 |
} |
250 |
} |
251 |
|
252 |
--len; |
253 |
} |
254 |
} |
255 |
|
256 |
inline void |
257 |
encode_indent (enc_t *enc) |
258 |
{ |
259 |
if (enc->json.flags & F_INDENT) |
260 |
{ |
261 |
int spaces = enc->indent * INDENT_STEP; |
262 |
|
263 |
need (enc, spaces); |
264 |
memset (enc->cur, ' ', spaces); |
265 |
enc->cur += spaces; |
266 |
} |
267 |
} |
268 |
|
269 |
inline void |
270 |
encode_space (enc_t *enc) |
271 |
{ |
272 |
need (enc, 1); |
273 |
encode_ch (enc, ' '); |
274 |
} |
275 |
|
276 |
inline void |
277 |
encode_nl (enc_t *enc) |
278 |
{ |
279 |
if (enc->json.flags & F_INDENT) |
280 |
{ |
281 |
need (enc, 1); |
282 |
encode_ch (enc, '\n'); |
283 |
} |
284 |
} |
285 |
|
286 |
inline void |
287 |
encode_comma (enc_t *enc) |
288 |
{ |
289 |
encode_ch (enc, ','); |
290 |
|
291 |
if (enc->json.flags & F_INDENT) |
292 |
encode_nl (enc); |
293 |
else if (enc->json.flags & F_SPACE_AFTER) |
294 |
encode_space (enc); |
295 |
} |
296 |
|
297 |
static void encode_sv (enc_t *enc, SV *sv); |
298 |
|
299 |
static void |
300 |
encode_av (enc_t *enc, AV *av) |
301 |
{ |
302 |
int i, len = av_len (av); |
303 |
|
304 |
if (enc->indent >= enc->maxdepth) |
305 |
croak ("data structure too deep (hit recursion limit)"); |
306 |
|
307 |
encode_ch (enc, '['); encode_nl (enc); |
308 |
++enc->indent; |
309 |
|
310 |
for (i = 0; i <= len; ++i) |
311 |
{ |
312 |
SV **svp = av_fetch (av, i, 0); |
313 |
|
314 |
encode_indent (enc); |
315 |
|
316 |
if (svp) |
317 |
encode_sv (enc, *svp); |
318 |
else |
319 |
encode_str (enc, "null", 4, 0); |
320 |
|
321 |
if (i < len) |
322 |
encode_comma (enc); |
323 |
} |
324 |
|
325 |
encode_nl (enc); |
326 |
|
327 |
--enc->indent; |
328 |
encode_indent (enc); encode_ch (enc, ']'); |
329 |
} |
330 |
|
331 |
static void |
332 |
encode_he (enc_t *enc, HE *he) |
333 |
{ |
334 |
encode_ch (enc, '"'); |
335 |
|
336 |
if (HeKLEN (he) == HEf_SVKEY) |
337 |
{ |
338 |
SV *sv = HeSVKEY (he); |
339 |
STRLEN len; |
340 |
char *str; |
341 |
|
342 |
SvGETMAGIC (sv); |
343 |
str = SvPV (sv, len); |
344 |
|
345 |
encode_str (enc, str, len, SvUTF8 (sv)); |
346 |
} |
347 |
else |
348 |
encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
349 |
|
350 |
encode_ch (enc, '"'); |
351 |
|
352 |
if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
353 |
encode_ch (enc, ':'); |
354 |
if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
355 |
encode_sv (enc, HeVAL (he)); |
356 |
} |
357 |
|
358 |
// compare hash entries, used when all keys are bytestrings |
359 |
static int |
360 |
he_cmp_fast (const void *a_, const void *b_) |
361 |
{ |
362 |
int cmp; |
363 |
|
364 |
HE *a = *(HE **)a_; |
365 |
HE *b = *(HE **)b_; |
366 |
|
367 |
STRLEN la = HeKLEN (a); |
368 |
STRLEN lb = HeKLEN (b); |
369 |
|
370 |
if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
371 |
cmp = la - lb; |
372 |
|
373 |
return cmp; |
374 |
} |
375 |
|
376 |
// compare hash entries, used when some keys are sv's or utf-x |
377 |
static int |
378 |
he_cmp_slow (const void *a, const void *b) |
379 |
{ |
380 |
return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
381 |
} |
382 |
|
383 |
static void |
384 |
encode_hv (enc_t *enc, HV *hv) |
385 |
{ |
386 |
int count, i; |
387 |
|
388 |
if (enc->indent >= enc->maxdepth) |
389 |
croak ("data structure too deep (hit recursion limit)"); |
390 |
|
391 |
encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
392 |
|
393 |
if ((count = hv_iterinit (hv))) |
394 |
{ |
395 |
// for canonical output we have to sort by keys first |
396 |
// actually, this is mostly due to the stupid so-called |
397 |
// security workaround added somewhere in 5.8.x. |
398 |
// that randomises hash orderings |
399 |
if (enc->json.flags & F_CANONICAL) |
400 |
{ |
401 |
int fast = 1; |
402 |
HE *he; |
403 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
404 |
HE **hes = _alloca (count * sizeof (HE)); |
405 |
#else |
406 |
HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
407 |
#endif |
408 |
|
409 |
i = 0; |
410 |
while ((he = hv_iternext (hv))) |
411 |
{ |
412 |
hes [i++] = he; |
413 |
if (HeKLEN (he) < 0 || HeKUTF8 (he)) |
414 |
fast = 0; |
415 |
} |
416 |
|
417 |
assert (i == count); |
418 |
|
419 |
if (fast) |
420 |
qsort (hes, count, sizeof (HE *), he_cmp_fast); |
421 |
else |
422 |
{ |
423 |
// hack to forcefully disable "use bytes" |
424 |
COP cop = *PL_curcop; |
425 |
cop.op_private = 0; |
426 |
|
427 |
ENTER; |
428 |
SAVETMPS; |
429 |
|
430 |
SAVEVPTR (PL_curcop); |
431 |
PL_curcop = &cop; |
432 |
|
433 |
qsort (hes, count, sizeof (HE *), he_cmp_slow); |
434 |
|
435 |
FREETMPS; |
436 |
LEAVE; |
437 |
} |
438 |
|
439 |
for (i = 0; i < count; ++i) |
440 |
{ |
441 |
encode_indent (enc); |
442 |
encode_he (enc, hes [i]); |
443 |
|
444 |
if (i < count - 1) |
445 |
encode_comma (enc); |
446 |
} |
447 |
|
448 |
encode_nl (enc); |
449 |
} |
450 |
else |
451 |
{ |
452 |
HE *he = hv_iternext (hv); |
453 |
|
454 |
for (;;) |
455 |
{ |
456 |
encode_indent (enc); |
457 |
encode_he (enc, he); |
458 |
|
459 |
if (!(he = hv_iternext (hv))) |
460 |
break; |
461 |
|
462 |
encode_comma (enc); |
463 |
} |
464 |
|
465 |
encode_nl (enc); |
466 |
} |
467 |
} |
468 |
|
469 |
--enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
470 |
} |
471 |
|
472 |
// encode objects, arrays and special \0=false and \1=true values. |
473 |
static void |
474 |
encode_rv (enc_t *enc, SV *sv) |
475 |
{ |
476 |
svtype svt; |
477 |
|
478 |
SvGETMAGIC (sv); |
479 |
svt = SvTYPE (sv); |
480 |
|
481 |
if (expect_false (SvOBJECT (sv))) |
482 |
{ |
483 |
if (SvSTASH (sv) == json_boolean_stash) |
484 |
{ |
485 |
if (SvIV (sv)) |
486 |
encode_str (enc, "true", 4, 0); |
487 |
else |
488 |
encode_str (enc, "false", 5, 0); |
489 |
} |
490 |
else |
491 |
{ |
492 |
#if 0 |
493 |
if (0 && sv_derived_from (rv, "JSON::Literal")) |
494 |
{ |
495 |
// not yet |
496 |
} |
497 |
#endif |
498 |
if (enc->json.flags & F_CONV_BLESSED) |
499 |
{ |
500 |
// we re-bless the reference to get overload and other niceties right |
501 |
GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
502 |
|
503 |
if (to_json) |
504 |
{ |
505 |
int count; |
506 |
dSP; |
507 |
|
508 |
ENTER; SAVETMPS; PUSHMARK (SP); |
509 |
XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
510 |
|
511 |
// calling with G_SCALAR ensures that we always get a 1 return value |
512 |
PUTBACK; |
513 |
call_sv ((SV *)GvCV (to_json), G_SCALAR); |
514 |
SPAGAIN; |
515 |
|
516 |
// catch this surprisingly common error |
517 |
if (SvROK (TOPs) && SvRV (TOPs) == sv) |
518 |
croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
519 |
|
520 |
sv = POPs; |
521 |
PUTBACK; |
522 |
|
523 |
encode_sv (enc, sv); |
524 |
|
525 |
FREETMPS; LEAVE; |
526 |
} |
527 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
528 |
encode_str (enc, "null", 4, 0); |
529 |
else |
530 |
croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
531 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
532 |
} |
533 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
534 |
encode_str (enc, "null", 4, 0); |
535 |
else |
536 |
croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
537 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
538 |
} |
539 |
} |
540 |
else if (svt == SVt_PVHV) |
541 |
encode_hv (enc, (HV *)sv); |
542 |
else if (svt == SVt_PVAV) |
543 |
encode_av (enc, (AV *)sv); |
544 |
else if (svt < SVt_PVAV) |
545 |
{ |
546 |
STRLEN len = 0; |
547 |
char *pv = svt ? SvPV (sv, len) : 0; |
548 |
|
549 |
if (len == 1 && *pv == '1') |
550 |
encode_str (enc, "true", 4, 0); |
551 |
else if (len == 1 && *pv == '0') |
552 |
encode_str (enc, "false", 5, 0); |
553 |
else |
554 |
croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
555 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
556 |
} |
557 |
else |
558 |
croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
559 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
560 |
} |
561 |
|
562 |
static void |
563 |
encode_sv (enc_t *enc, SV *sv) |
564 |
{ |
565 |
SvGETMAGIC (sv); |
566 |
|
567 |
if (SvPOKp (sv)) |
568 |
{ |
569 |
STRLEN len; |
570 |
char *str = SvPV (sv, len); |
571 |
encode_ch (enc, '"'); |
572 |
encode_str (enc, str, len, SvUTF8 (sv)); |
573 |
encode_ch (enc, '"'); |
574 |
} |
575 |
else if (SvNOKp (sv)) |
576 |
{ |
577 |
// trust that perl will do the right thing w.r.t. JSON syntax. |
578 |
need (enc, NV_DIG + 32); |
579 |
Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
580 |
enc->cur += strlen (enc->cur); |
581 |
} |
582 |
else if (SvIOKp (sv)) |
583 |
{ |
584 |
// we assume we can always read an IV as a UV |
585 |
if (SvUV (sv) & ~(UV)0x7fff) |
586 |
{ |
587 |
// large integer, use the (rather slow) snprintf way. |
588 |
need (enc, sizeof (UV) * 3); |
589 |
enc->cur += |
590 |
SvIsUV(sv) |
591 |
? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
592 |
: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
593 |
} |
594 |
else |
595 |
{ |
596 |
// optimise the "small number case" |
597 |
// code will likely be branchless and use only a single multiplication |
598 |
I32 i = SvIV (sv); |
599 |
U32 u; |
600 |
char digit, nz = 0; |
601 |
|
602 |
need (enc, 6); |
603 |
|
604 |
*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
605 |
u = i < 0 ? -i : i; |
606 |
|
607 |
// convert to 4.28 fixed-point representation |
608 |
u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
609 |
|
610 |
// now output digit by digit, each time masking out the integer part |
611 |
// and multiplying by 5 while moving the decimal point one to the right, |
612 |
// resulting in a net multiplication by 10. |
613 |
// we always write the digit to memory but conditionally increment |
614 |
// the pointer, to ease the usage of conditional move instructions. |
615 |
digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
616 |
digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
617 |
digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
618 |
digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
619 |
digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
620 |
} |
621 |
} |
622 |
else if (SvROK (sv)) |
623 |
encode_rv (enc, SvRV (sv)); |
624 |
else if (!SvOK (sv)) |
625 |
encode_str (enc, "null", 4, 0); |
626 |
else |
627 |
croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
628 |
SvPV_nolen (sv), SvFLAGS (sv)); |
629 |
} |
630 |
|
631 |
static SV * |
632 |
encode_json (SV *scalar, JSON *json) |
633 |
{ |
634 |
enc_t enc; |
635 |
|
636 |
if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
637 |
croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
638 |
|
639 |
enc.json = *json; |
640 |
enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
641 |
enc.cur = SvPVX (enc.sv); |
642 |
enc.end = SvEND (enc.sv); |
643 |
enc.indent = 0; |
644 |
enc.maxdepth = DEC_DEPTH (enc.json.flags); |
645 |
|
646 |
SvPOK_only (enc.sv); |
647 |
encode_sv (&enc, scalar); |
648 |
|
649 |
SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
650 |
*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
651 |
|
652 |
if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
653 |
SvUTF8_on (enc.sv); |
654 |
|
655 |
if (enc.json.flags & F_SHRINK) |
656 |
shrink (enc.sv); |
657 |
|
658 |
return enc.sv; |
659 |
} |
660 |
|
661 |
///////////////////////////////////////////////////////////////////////////// |
662 |
// decoder |
663 |
|
664 |
// structure used for decoding JSON |
665 |
typedef struct |
666 |
{ |
667 |
char *cur; // current parser pointer |
668 |
char *end; // end of input string |
669 |
const char *err; // parse error, if != 0 |
670 |
JSON json; |
671 |
U32 depth; // recursion depth |
672 |
U32 maxdepth; // recursion depth limit |
673 |
} dec_t; |
674 |
|
675 |
inline void |
676 |
decode_ws (dec_t *dec) |
677 |
{ |
678 |
for (;;) |
679 |
{ |
680 |
char ch = *dec->cur; |
681 |
|
682 |
if (ch > 0x20 |
683 |
|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
684 |
break; |
685 |
|
686 |
++dec->cur; |
687 |
} |
688 |
} |
689 |
|
690 |
#define ERR(reason) SB dec->err = reason; goto fail; SE |
691 |
|
692 |
#define EXPECT_CH(ch) SB \ |
693 |
if (*dec->cur != ch) \ |
694 |
ERR (# ch " expected"); \ |
695 |
++dec->cur; \ |
696 |
SE |
697 |
|
698 |
#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
699 |
#define DEC_DEC_DEPTH --dec->depth |
700 |
|
701 |
static SV *decode_sv (dec_t *dec); |
702 |
|
703 |
static signed char decode_hexdigit[256]; |
704 |
|
705 |
static UV |
706 |
decode_4hex (dec_t *dec) |
707 |
{ |
708 |
signed char d1, d2, d3, d4; |
709 |
unsigned char *cur = (unsigned char *)dec->cur; |
710 |
|
711 |
d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
712 |
d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
713 |
d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
714 |
d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
715 |
|
716 |
dec->cur += 4; |
717 |
|
718 |
return ((UV)d1) << 12 |
719 |
| ((UV)d2) << 8 |
720 |
| ((UV)d3) << 4 |
721 |
| ((UV)d4); |
722 |
|
723 |
fail: |
724 |
return (UV)-1; |
725 |
} |
726 |
|
727 |
static SV * |
728 |
decode_str (dec_t *dec) |
729 |
{ |
730 |
SV *sv = 0; |
731 |
int utf8 = 0; |
732 |
char *dec_cur = dec->cur; |
733 |
|
734 |
do |
735 |
{ |
736 |
char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
737 |
char *cur = buf; |
738 |
|
739 |
do |
740 |
{ |
741 |
unsigned char ch = *(unsigned char *)dec_cur++; |
742 |
|
743 |
if (expect_false (ch == '"')) |
744 |
{ |
745 |
--dec_cur; |
746 |
break; |
747 |
} |
748 |
else if (expect_false (ch == '\\')) |
749 |
{ |
750 |
switch (*dec_cur) |
751 |
{ |
752 |
case '\\': |
753 |
case '/': |
754 |
case '"': *cur++ = *dec_cur++; break; |
755 |
|
756 |
case 'b': ++dec_cur; *cur++ = '\010'; break; |
757 |
case 't': ++dec_cur; *cur++ = '\011'; break; |
758 |
case 'n': ++dec_cur; *cur++ = '\012'; break; |
759 |
case 'f': ++dec_cur; *cur++ = '\014'; break; |
760 |
case 'r': ++dec_cur; *cur++ = '\015'; break; |
761 |
|
762 |
case 'u': |
763 |
{ |
764 |
UV lo, hi; |
765 |
++dec_cur; |
766 |
|
767 |
dec->cur = dec_cur; |
768 |
hi = decode_4hex (dec); |
769 |
dec_cur = dec->cur; |
770 |
if (hi == (UV)-1) |
771 |
goto fail; |
772 |
|
773 |
// possibly a surrogate pair |
774 |
if (hi >= 0xd800) |
775 |
if (hi < 0xdc00) |
776 |
{ |
777 |
if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
778 |
ERR ("missing low surrogate character in surrogate pair"); |
779 |
|
780 |
dec_cur += 2; |
781 |
|
782 |
dec->cur = dec_cur; |
783 |
lo = decode_4hex (dec); |
784 |
dec_cur = dec->cur; |
785 |
if (lo == (UV)-1) |
786 |
goto fail; |
787 |
|
788 |
if (lo < 0xdc00 || lo >= 0xe000) |
789 |
ERR ("surrogate pair expected"); |
790 |
|
791 |
hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
792 |
} |
793 |
else if (hi < 0xe000) |
794 |
ERR ("missing high surrogate character in surrogate pair"); |
795 |
|
796 |
if (hi >= 0x80) |
797 |
{ |
798 |
utf8 = 1; |
799 |
|
800 |
cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
801 |
} |
802 |
else |
803 |
*cur++ = hi; |
804 |
} |
805 |
break; |
806 |
|
807 |
default: |
808 |
--dec_cur; |
809 |
ERR ("illegal backslash escape sequence in string"); |
810 |
} |
811 |
} |
812 |
else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
813 |
*cur++ = ch; |
814 |
else if (ch >= 0x80) |
815 |
{ |
816 |
STRLEN clen; |
817 |
UV uch; |
818 |
|
819 |
--dec_cur; |
820 |
|
821 |
uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
822 |
if (clen == (STRLEN)-1) |
823 |
ERR ("malformed UTF-8 character in JSON string"); |
824 |
|
825 |
do |
826 |
*cur++ = *dec_cur++; |
827 |
while (--clen); |
828 |
|
829 |
utf8 = 1; |
830 |
} |
831 |
else |
832 |
{ |
833 |
--dec_cur; |
834 |
|
835 |
if (!ch) |
836 |
ERR ("unexpected end of string while parsing JSON string"); |
837 |
else |
838 |
ERR ("invalid character encountered while parsing JSON string"); |
839 |
} |
840 |
} |
841 |
while (cur < buf + SHORT_STRING_LEN); |
842 |
|
843 |
{ |
844 |
STRLEN len = cur - buf; |
845 |
|
846 |
if (sv) |
847 |
{ |
848 |
SvGROW (sv, SvCUR (sv) + len + 1); |
849 |
memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
850 |
SvCUR_set (sv, SvCUR (sv) + len); |
851 |
} |
852 |
else |
853 |
sv = newSVpvn (buf, len); |
854 |
} |
855 |
} |
856 |
while (*dec_cur != '"'); |
857 |
|
858 |
++dec_cur; |
859 |
|
860 |
if (sv) |
861 |
{ |
862 |
SvPOK_only (sv); |
863 |
*SvEND (sv) = 0; |
864 |
|
865 |
if (utf8) |
866 |
SvUTF8_on (sv); |
867 |
} |
868 |
else |
869 |
sv = newSVpvn ("", 0); |
870 |
|
871 |
dec->cur = dec_cur; |
872 |
return sv; |
873 |
|
874 |
fail: |
875 |
dec->cur = dec_cur; |
876 |
return 0; |
877 |
} |
878 |
|
879 |
static SV * |
880 |
decode_num (dec_t *dec) |
881 |
{ |
882 |
int is_nv = 0; |
883 |
char *start = dec->cur; |
884 |
|
885 |
// [minus] |
886 |
if (*dec->cur == '-') |
887 |
++dec->cur; |
888 |
|
889 |
if (*dec->cur == '0') |
890 |
{ |
891 |
++dec->cur; |
892 |
if (*dec->cur >= '0' && *dec->cur <= '9') |
893 |
ERR ("malformed number (leading zero must not be followed by another digit)"); |
894 |
} |
895 |
else if (*dec->cur < '0' || *dec->cur > '9') |
896 |
ERR ("malformed number (no digits after initial minus)"); |
897 |
else |
898 |
do |
899 |
{ |
900 |
++dec->cur; |
901 |
} |
902 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
903 |
|
904 |
// [frac] |
905 |
if (*dec->cur == '.') |
906 |
{ |
907 |
++dec->cur; |
908 |
|
909 |
if (*dec->cur < '0' || *dec->cur > '9') |
910 |
ERR ("malformed number (no digits after decimal point)"); |
911 |
|
912 |
do |
913 |
{ |
914 |
++dec->cur; |
915 |
} |
916 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
917 |
|
918 |
is_nv = 1; |
919 |
} |
920 |
|
921 |
// [exp] |
922 |
if (*dec->cur == 'e' || *dec->cur == 'E') |
923 |
{ |
924 |
++dec->cur; |
925 |
|
926 |
if (*dec->cur == '-' || *dec->cur == '+') |
927 |
++dec->cur; |
928 |
|
929 |
if (*dec->cur < '0' || *dec->cur > '9') |
930 |
ERR ("malformed number (no digits after exp sign)"); |
931 |
|
932 |
do |
933 |
{ |
934 |
++dec->cur; |
935 |
} |
936 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
937 |
|
938 |
is_nv = 1; |
939 |
} |
940 |
|
941 |
if (!is_nv) |
942 |
{ |
943 |
int len = dec->cur - start; |
944 |
|
945 |
// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
946 |
if (*start == '-') |
947 |
switch (len) |
948 |
{ |
949 |
case 2: return newSViv (-( start [1] - '0' * 1)); |
950 |
case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
951 |
case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
952 |
case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
953 |
} |
954 |
else |
955 |
switch (len) |
956 |
{ |
957 |
case 1: return newSViv ( start [0] - '0' * 1); |
958 |
case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
959 |
case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
960 |
case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
961 |
} |
962 |
|
963 |
{ |
964 |
UV uv; |
965 |
int numtype = grok_number (start, len, &uv); |
966 |
if (numtype & IS_NUMBER_IN_UV) |
967 |
if (numtype & IS_NUMBER_NEG) |
968 |
{ |
969 |
if (uv < (UV)IV_MIN) |
970 |
return newSViv (-(IV)uv); |
971 |
} |
972 |
else |
973 |
return newSVuv (uv); |
974 |
} |
975 |
|
976 |
len -= *start == '-' ? 1 : 0; |
977 |
|
978 |
// does not fit into IV or UV, try NV |
979 |
if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
980 |
#if defined (LDBL_DIG) |
981 |
|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
982 |
#endif |
983 |
) |
984 |
// fits into NV without loss of precision |
985 |
return newSVnv (Atof (start)); |
986 |
|
987 |
// everything else fails, convert it to a string |
988 |
return newSVpvn (start, dec->cur - start); |
989 |
} |
990 |
|
991 |
// loss of precision here |
992 |
return newSVnv (Atof (start)); |
993 |
|
994 |
fail: |
995 |
return 0; |
996 |
} |
997 |
|
998 |
static SV * |
999 |
decode_av (dec_t *dec) |
1000 |
{ |
1001 |
AV *av = newAV (); |
1002 |
|
1003 |
DEC_INC_DEPTH; |
1004 |
decode_ws (dec); |
1005 |
|
1006 |
if (*dec->cur == ']') |
1007 |
++dec->cur; |
1008 |
else |
1009 |
for (;;) |
1010 |
{ |
1011 |
SV *value; |
1012 |
|
1013 |
value = decode_sv (dec); |
1014 |
if (!value) |
1015 |
goto fail; |
1016 |
|
1017 |
av_push (av, value); |
1018 |
|
1019 |
decode_ws (dec); |
1020 |
|
1021 |
if (*dec->cur == ']') |
1022 |
{ |
1023 |
++dec->cur; |
1024 |
break; |
1025 |
} |
1026 |
|
1027 |
if (*dec->cur != ',') |
1028 |
ERR (", or ] expected while parsing array"); |
1029 |
|
1030 |
++dec->cur; |
1031 |
} |
1032 |
|
1033 |
DEC_DEC_DEPTH; |
1034 |
return newRV_noinc ((SV *)av); |
1035 |
|
1036 |
fail: |
1037 |
SvREFCNT_dec (av); |
1038 |
DEC_DEC_DEPTH; |
1039 |
return 0; |
1040 |
} |
1041 |
|
1042 |
static SV * |
1043 |
decode_hv (dec_t *dec) |
1044 |
{ |
1045 |
SV *sv; |
1046 |
HV *hv = newHV (); |
1047 |
|
1048 |
DEC_INC_DEPTH; |
1049 |
decode_ws (dec); |
1050 |
|
1051 |
if (*dec->cur == '}') |
1052 |
++dec->cur; |
1053 |
else |
1054 |
for (;;) |
1055 |
{ |
1056 |
decode_ws (dec); EXPECT_CH ('"'); |
1057 |
|
1058 |
// heuristic: assume that |
1059 |
// a) decode_str + hv_store_ent are abysmally slow. |
1060 |
// b) most hash keys are short, simple ascii text. |
1061 |
// => try to "fast-match" such strings to avoid |
1062 |
// the overhead of decode_str + hv_store_ent. |
1063 |
{ |
1064 |
SV *value; |
1065 |
char *p = dec->cur; |
1066 |
char *e = p + 24; // only try up to 24 bytes |
1067 |
|
1068 |
for (;;) |
1069 |
{ |
1070 |
// the >= 0x80 is true on most architectures |
1071 |
if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1072 |
{ |
1073 |
// slow path, back up and use decode_str |
1074 |
SV *key = decode_str (dec); |
1075 |
if (!key) |
1076 |
goto fail; |
1077 |
|
1078 |
decode_ws (dec); EXPECT_CH (':'); |
1079 |
|
1080 |
value = decode_sv (dec); |
1081 |
if (!value) |
1082 |
{ |
1083 |
SvREFCNT_dec (key); |
1084 |
goto fail; |
1085 |
} |
1086 |
|
1087 |
hv_store_ent (hv, key, value, 0); |
1088 |
SvREFCNT_dec (key); |
1089 |
|
1090 |
break; |
1091 |
} |
1092 |
else if (*p == '"') |
1093 |
{ |
1094 |
// fast path, got a simple key |
1095 |
char *key = dec->cur; |
1096 |
int len = p - key; |
1097 |
dec->cur = p + 1; |
1098 |
|
1099 |
decode_ws (dec); EXPECT_CH (':'); |
1100 |
|
1101 |
value = decode_sv (dec); |
1102 |
if (!value) |
1103 |
goto fail; |
1104 |
|
1105 |
hv_store (hv, key, len, value, 0); |
1106 |
|
1107 |
break; |
1108 |
} |
1109 |
|
1110 |
++p; |
1111 |
} |
1112 |
} |
1113 |
|
1114 |
decode_ws (dec); |
1115 |
|
1116 |
if (*dec->cur == '}') |
1117 |
{ |
1118 |
++dec->cur; |
1119 |
break; |
1120 |
} |
1121 |
|
1122 |
if (*dec->cur != ',') |
1123 |
ERR (", or } expected while parsing object/hash"); |
1124 |
|
1125 |
++dec->cur; |
1126 |
} |
1127 |
|
1128 |
DEC_DEC_DEPTH; |
1129 |
sv = newRV_noinc ((SV *)hv); |
1130 |
|
1131 |
// check filter callbacks |
1132 |
if (dec->json.flags & F_HOOK) |
1133 |
{ |
1134 |
if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
1135 |
{ |
1136 |
HE *cb, *he; |
1137 |
|
1138 |
hv_iterinit (hv); |
1139 |
he = hv_iternext (hv); |
1140 |
hv_iterinit (hv); |
1141 |
|
1142 |
// the next line creates a mortal sv each time its called. |
1143 |
// might want to optimise this for common cases. |
1144 |
cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1145 |
|
1146 |
if (cb) |
1147 |
{ |
1148 |
dSP; |
1149 |
int count; |
1150 |
|
1151 |
ENTER; SAVETMPS; PUSHMARK (SP); |
1152 |
XPUSHs (HeVAL (he)); |
1153 |
|
1154 |
PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1155 |
|
1156 |
if (count == 1) |
1157 |
{ |
1158 |
sv = newSVsv (POPs); |
1159 |
FREETMPS; LEAVE; |
1160 |
return sv; |
1161 |
} |
1162 |
|
1163 |
FREETMPS; LEAVE; |
1164 |
} |
1165 |
} |
1166 |
|
1167 |
if (dec->json.cb_object) |
1168 |
{ |
1169 |
dSP; |
1170 |
int count; |
1171 |
|
1172 |
ENTER; SAVETMPS; PUSHMARK (SP); |
1173 |
XPUSHs (sv_2mortal (sv)); |
1174 |
|
1175 |
PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1176 |
|
1177 |
if (count == 1) |
1178 |
{ |
1179 |
sv = newSVsv (POPs); |
1180 |
FREETMPS; LEAVE; |
1181 |
return sv; |
1182 |
} |
1183 |
|
1184 |
SvREFCNT_inc (sv); |
1185 |
FREETMPS; LEAVE; |
1186 |
} |
1187 |
} |
1188 |
|
1189 |
return sv; |
1190 |
|
1191 |
fail: |
1192 |
SvREFCNT_dec (hv); |
1193 |
DEC_DEC_DEPTH; |
1194 |
return 0; |
1195 |
} |
1196 |
|
1197 |
static SV * |
1198 |
decode_sv (dec_t *dec) |
1199 |
{ |
1200 |
decode_ws (dec); |
1201 |
|
1202 |
// the beauty of JSON: you need exactly one character lookahead |
1203 |
// to parse anything. |
1204 |
switch (*dec->cur) |
1205 |
{ |
1206 |
case '"': ++dec->cur; return decode_str (dec); |
1207 |
case '[': ++dec->cur; return decode_av (dec); |
1208 |
case '{': ++dec->cur; return decode_hv (dec); |
1209 |
|
1210 |
case '-': |
1211 |
case '0': case '1': case '2': case '3': case '4': |
1212 |
case '5': case '6': case '7': case '8': case '9': |
1213 |
return decode_num (dec); |
1214 |
|
1215 |
case 't': |
1216 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1217 |
{ |
1218 |
dec->cur += 4; |
1219 |
return SvREFCNT_inc (json_true); |
1220 |
} |
1221 |
else |
1222 |
ERR ("'true' expected"); |
1223 |
|
1224 |
break; |
1225 |
|
1226 |
case 'f': |
1227 |
if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1228 |
{ |
1229 |
dec->cur += 5; |
1230 |
return SvREFCNT_inc (json_false); |
1231 |
} |
1232 |
else |
1233 |
ERR ("'false' expected"); |
1234 |
|
1235 |
break; |
1236 |
|
1237 |
case 'n': |
1238 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
1239 |
{ |
1240 |
dec->cur += 4; |
1241 |
return newSVsv (&PL_sv_undef); |
1242 |
} |
1243 |
else |
1244 |
ERR ("'null' expected"); |
1245 |
|
1246 |
break; |
1247 |
|
1248 |
default: |
1249 |
ERR ("malformed JSON string, neither array, object, number, string or atom"); |
1250 |
break; |
1251 |
} |
1252 |
|
1253 |
fail: |
1254 |
return 0; |
1255 |
} |
1256 |
|
1257 |
static SV * |
1258 |
decode_json (SV *string, JSON *json, UV *offset_return) |
1259 |
{ |
1260 |
dec_t dec; |
1261 |
UV offset; |
1262 |
SV *sv; |
1263 |
|
1264 |
SvGETMAGIC (string); |
1265 |
SvUPGRADE (string, SVt_PV); |
1266 |
|
1267 |
if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
1268 |
croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1269 |
(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
1270 |
|
1271 |
if (json->flags & F_UTF8) |
1272 |
sv_utf8_downgrade (string, 0); |
1273 |
else |
1274 |
sv_utf8_upgrade (string); |
1275 |
|
1276 |
SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1277 |
|
1278 |
dec.json = *json; |
1279 |
dec.cur = SvPVX (string); |
1280 |
dec.end = SvEND (string); |
1281 |
dec.err = 0; |
1282 |
dec.depth = 0; |
1283 |
dec.maxdepth = DEC_DEPTH (dec.json.flags); |
1284 |
|
1285 |
if (dec.json.cb_object || dec.json.cb_sk_object) |
1286 |
dec.json.flags |= F_HOOK; |
1287 |
|
1288 |
*dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1289 |
sv = decode_sv (&dec); |
1290 |
|
1291 |
if (!(offset_return || !sv)) |
1292 |
{ |
1293 |
// check for trailing garbage |
1294 |
decode_ws (&dec); |
1295 |
|
1296 |
if (*dec.cur) |
1297 |
{ |
1298 |
dec.err = "garbage after JSON object"; |
1299 |
SvREFCNT_dec (sv); |
1300 |
sv = 0; |
1301 |
} |
1302 |
} |
1303 |
|
1304 |
if (offset_return || !sv) |
1305 |
{ |
1306 |
offset = dec.json.flags & F_UTF8 |
1307 |
? dec.cur - SvPVX (string) |
1308 |
: utf8_distance (dec.cur, SvPVX (string)); |
1309 |
|
1310 |
if (offset_return) |
1311 |
*offset_return = offset; |
1312 |
} |
1313 |
|
1314 |
if (!sv) |
1315 |
{ |
1316 |
SV *uni = sv_newmortal (); |
1317 |
|
1318 |
// horrible hack to silence warning inside pv_uni_display |
1319 |
COP cop = *PL_curcop; |
1320 |
cop.cop_warnings = pWARN_NONE; |
1321 |
ENTER; |
1322 |
SAVEVPTR (PL_curcop); |
1323 |
PL_curcop = &cop; |
1324 |
pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1325 |
LEAVE; |
1326 |
|
1327 |
croak ("%s, at character offset %d [\"%s\"]", |
1328 |
dec.err, |
1329 |
(int)offset, |
1330 |
dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1331 |
} |
1332 |
|
1333 |
sv = sv_2mortal (sv); |
1334 |
|
1335 |
if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1336 |
croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1337 |
|
1338 |
return sv; |
1339 |
} |
1340 |
|
1341 |
///////////////////////////////////////////////////////////////////////////// |
1342 |
// XS interface functions |
1343 |
|
1344 |
MODULE = JSON::XS PACKAGE = JSON::XS |
1345 |
|
1346 |
BOOT: |
1347 |
{ |
1348 |
int i; |
1349 |
|
1350 |
for (i = 0; i < 256; ++i) |
1351 |
decode_hexdigit [i] = |
1352 |
i >= '0' && i <= '9' ? i - '0' |
1353 |
: i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1354 |
: i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1355 |
: -1; |
1356 |
|
1357 |
json_stash = gv_stashpv ("JSON::XS" , 1); |
1358 |
json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1359 |
|
1360 |
json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
1361 |
json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
1362 |
} |
1363 |
|
1364 |
PROTOTYPES: DISABLE |
1365 |
|
1366 |
void new (char *klass) |
1367 |
PPCODE: |
1368 |
{ |
1369 |
SV *pv = NEWSV (0, sizeof (JSON)); |
1370 |
SvPOK_only (pv); |
1371 |
Zero (SvPVX (pv), 1, JSON); |
1372 |
((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
1373 |
XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash))); |
1374 |
} |
1375 |
|
1376 |
void ascii (JSON *self, int enable = 1) |
1377 |
ALIAS: |
1378 |
ascii = F_ASCII |
1379 |
latin1 = F_LATIN1 |
1380 |
utf8 = F_UTF8 |
1381 |
indent = F_INDENT |
1382 |
canonical = F_CANONICAL |
1383 |
space_before = F_SPACE_BEFORE |
1384 |
space_after = F_SPACE_AFTER |
1385 |
pretty = F_PRETTY |
1386 |
allow_nonref = F_ALLOW_NONREF |
1387 |
shrink = F_SHRINK |
1388 |
allow_blessed = F_ALLOW_BLESSED |
1389 |
convert_blessed = F_CONV_BLESSED |
1390 |
PPCODE: |
1391 |
{ |
1392 |
if (enable) |
1393 |
self->flags |= ix; |
1394 |
else |
1395 |
self->flags &= ~ix; |
1396 |
|
1397 |
XPUSHs (ST (0)); |
1398 |
} |
1399 |
|
1400 |
void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
1401 |
PPCODE: |
1402 |
{ |
1403 |
UV log2 = 0; |
1404 |
|
1405 |
if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
1406 |
|
1407 |
while ((1UL << log2) < max_depth) |
1408 |
++log2; |
1409 |
|
1410 |
self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
1411 |
|
1412 |
XPUSHs (ST (0)); |
1413 |
} |
1414 |
|
1415 |
void max_size (JSON *self, UV max_size = 0) |
1416 |
PPCODE: |
1417 |
{ |
1418 |
UV log2 = 0; |
1419 |
|
1420 |
if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
1421 |
if (max_size == 1) max_size = 2; |
1422 |
|
1423 |
while ((1UL << log2) < max_size) |
1424 |
++log2; |
1425 |
|
1426 |
self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
1427 |
|
1428 |
XPUSHs (ST (0)); |
1429 |
} |
1430 |
|
1431 |
void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1432 |
PPCODE: |
1433 |
{ |
1434 |
SvREFCNT_dec (self->cb_object); |
1435 |
self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
1436 |
|
1437 |
XPUSHs (ST (0)); |
1438 |
} |
1439 |
|
1440 |
void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
1441 |
PPCODE: |
1442 |
{ |
1443 |
if (!self->cb_sk_object) |
1444 |
self->cb_sk_object = newHV (); |
1445 |
|
1446 |
if (SvOK (cb)) |
1447 |
hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
1448 |
else |
1449 |
{ |
1450 |
hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
1451 |
|
1452 |
if (!HvKEYS (self->cb_sk_object)) |
1453 |
{ |
1454 |
SvREFCNT_dec (self->cb_sk_object); |
1455 |
self->cb_sk_object = 0; |
1456 |
} |
1457 |
} |
1458 |
|
1459 |
XPUSHs (ST (0)); |
1460 |
} |
1461 |
|
1462 |
void encode (JSON *self, SV *scalar) |
1463 |
PPCODE: |
1464 |
XPUSHs (encode_json (scalar, self)); |
1465 |
|
1466 |
void decode (JSON *self, SV *jsonstr) |
1467 |
PPCODE: |
1468 |
XPUSHs (decode_json (jsonstr, self, 0)); |
1469 |
|
1470 |
void decode_prefix (JSON *self, SV *jsonstr) |
1471 |
PPCODE: |
1472 |
{ |
1473 |
UV offset; |
1474 |
EXTEND (SP, 2); |
1475 |
PUSHs (decode_json (jsonstr, self, &offset)); |
1476 |
PUSHs (sv_2mortal (newSVuv (offset))); |
1477 |
} |
1478 |
|
1479 |
void DESTROY (JSON *self) |
1480 |
CODE: |
1481 |
SvREFCNT_dec (self->cb_sk_object); |
1482 |
SvREFCNT_dec (self->cb_object); |
1483 |
|
1484 |
PROTOTYPES: ENABLE |
1485 |
|
1486 |
void to_json (SV *scalar) |
1487 |
PPCODE: |
1488 |
{ |
1489 |
JSON json = { F_DEFAULT | F_UTF8 }; |
1490 |
XPUSHs (encode_json (scalar, &json)); |
1491 |
} |
1492 |
|
1493 |
void from_json (SV *jsonstr) |
1494 |
PPCODE: |
1495 |
{ |
1496 |
JSON json = { F_DEFAULT | F_UTF8 }; |
1497 |
XPUSHs (decode_json (jsonstr, &json, 0)); |
1498 |
} |
1499 |
|