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