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