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 <limits.h> |
10 |
#include <float.h> |
11 |
#include <inttypes.h> |
12 |
|
13 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
14 |
# define snprintf _snprintf // C compilers have this in stdio.h |
15 |
#endif |
16 |
|
17 |
// some old perls do not have this, try to make it work, no |
18 |
// guarantees, though. if it breaks, you get to keep the pieces. |
19 |
#ifndef UTF8_MAXBYTES |
20 |
# define UTF8_MAXBYTES 13 |
21 |
#endif |
22 |
|
23 |
// compatibility with perl <5.18 |
24 |
#ifndef HvNAMELEN_get |
25 |
# define HvNAMELEN_get(hv) strlen (HvNAME (hv)) |
26 |
#endif |
27 |
#ifndef HvNAMELEN |
28 |
# define HvNAMELEN(hv) HvNAMELEN_get (hv) |
29 |
#endif |
30 |
#ifndef HvNAMEUTF8 |
31 |
# define HvNAMEUTF8(hv) 0 |
32 |
#endif |
33 |
|
34 |
// three extra for rounding, sign, and end of string |
35 |
#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
36 |
|
37 |
#define F_ASCII 0x00000001UL |
38 |
#define F_LATIN1 0x00000002UL |
39 |
#define F_UTF8 0x00000004UL |
40 |
#define F_INDENT 0x00000008UL |
41 |
#define F_CANONICAL 0x00000010UL |
42 |
#define F_SPACE_BEFORE 0x00000020UL |
43 |
#define F_SPACE_AFTER 0x00000040UL |
44 |
#define F_ALLOW_NONREF 0x00000100UL |
45 |
#define F_SHRINK 0x00000200UL |
46 |
#define F_ALLOW_BLESSED 0x00000400UL |
47 |
#define F_CONV_BLESSED 0x00000800UL |
48 |
#define F_RELAXED 0x00001000UL |
49 |
#define F_ALLOW_UNKNOWN 0x00002000UL |
50 |
#define F_ALLOW_TAGS 0x00004000UL |
51 |
#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
52 |
|
53 |
#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
54 |
|
55 |
#define INIT_SIZE 32 // initial scalar size to be allocated |
56 |
#define INDENT_STEP 3 // spaces per indentation level |
57 |
|
58 |
#define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
59 |
|
60 |
#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8) |
61 |
|
62 |
#define SB do { |
63 |
#define SE } while (0) |
64 |
|
65 |
#if __GNUC__ >= 3 |
66 |
# define expect(expr,value) __builtin_expect ((expr), (value)) |
67 |
# define INLINE static inline |
68 |
#else |
69 |
# define expect(expr,value) (expr) |
70 |
# define INLINE static |
71 |
#endif |
72 |
|
73 |
#define expect_false(expr) expect ((expr) != 0, 0) |
74 |
#define expect_true(expr) expect ((expr) != 0, 1) |
75 |
|
76 |
#define IN_RANGE_INC(type,val,beg,end) \ |
77 |
((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
78 |
<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
79 |
|
80 |
#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
81 |
|
82 |
#ifdef USE_ITHREADS |
83 |
# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1)) |
84 |
# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1)) |
85 |
# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value)) |
86 |
#else |
87 |
# define JSON_STASH json_stash |
88 |
# define BOOL_STASH bool_stash |
89 |
# define GET_BOOL(value) bool_ ## value |
90 |
#endif |
91 |
|
92 |
// the amount of HEs to allocate on the stack, when sorting keys |
93 |
#define STACK_HES 64 |
94 |
|
95 |
static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean:: |
96 |
static SV *bool_false, *bool_true; |
97 |
static SV *sv_json; |
98 |
|
99 |
enum { |
100 |
INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
101 |
INCR_M_STR, // inside string |
102 |
INCR_M_BS, // inside backslash |
103 |
INCR_M_C0, // inside comment in initial whitespace sequence |
104 |
INCR_M_C1, // inside comment in other places |
105 |
INCR_M_JSON // outside anything, count nesting |
106 |
}; |
107 |
|
108 |
#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
109 |
|
110 |
typedef struct { |
111 |
U32 flags; |
112 |
U32 max_depth; |
113 |
STRLEN max_size; |
114 |
|
115 |
SV *cb_object; |
116 |
HV *cb_sk_object; |
117 |
|
118 |
// for the incremental parser |
119 |
SV *incr_text; // the source text so far |
120 |
STRLEN incr_pos; // the current offset into the text |
121 |
int incr_nest; // {[]}-nesting level |
122 |
unsigned char incr_mode; |
123 |
|
124 |
SV *v_false, *v_true; |
125 |
} JSON; |
126 |
|
127 |
INLINE void |
128 |
json_init (JSON *json) |
129 |
{ |
130 |
Zero (json, 1, JSON); |
131 |
json->max_depth = 512; |
132 |
} |
133 |
|
134 |
///////////////////////////////////////////////////////////////////////////// |
135 |
// utility functions |
136 |
|
137 |
INLINE SV * |
138 |
get_bool (const char *name) |
139 |
{ |
140 |
SV *sv = get_sv (name, 1); |
141 |
|
142 |
SvREADONLY_on (sv); |
143 |
SvREADONLY_on (SvRV (sv)); |
144 |
|
145 |
return sv; |
146 |
} |
147 |
|
148 |
INLINE void |
149 |
shrink (SV *sv) |
150 |
{ |
151 |
sv_utf8_downgrade (sv, 1); |
152 |
|
153 |
if (SvLEN (sv) > SvCUR (sv) + 1) |
154 |
{ |
155 |
#ifdef SvPV_shrink_to_cur |
156 |
SvPV_shrink_to_cur (sv); |
157 |
#elif defined (SvPV_renew) |
158 |
SvPV_renew (sv, SvCUR (sv) + 1); |
159 |
#endif |
160 |
} |
161 |
} |
162 |
|
163 |
/* adds two STRLENs together, slow, and with paranoia */ |
164 |
STRLEN |
165 |
strlen_sum (STRLEN l1, STRLEN l2) |
166 |
{ |
167 |
size_t sum = l1 + l2; |
168 |
|
169 |
if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum) |
170 |
croak ("JSON::XS: string size overflow"); |
171 |
|
172 |
return sum; |
173 |
} |
174 |
|
175 |
/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */ |
176 |
static char * |
177 |
json_sv_grow (SV *sv, size_t len1, size_t len2) |
178 |
{ |
179 |
len1 = strlen_sum (len1, len2); |
180 |
len1 = strlen_sum (len1, len1 >> 1); |
181 |
|
182 |
if (len1 > 4096 - 24) |
183 |
len1 = (len1 | 4095) - 24; |
184 |
|
185 |
return SvGROW (sv, len1); |
186 |
} |
187 |
|
188 |
// decode a utf-8 character and return it, or (UV)-1 in |
189 |
// case of an error. |
190 |
// we special-case "safe" characters from U+80 .. U+7FF, |
191 |
// but use the very good perl function to parse anything else. |
192 |
// note that we never call this function for a ascii codepoints |
193 |
INLINE UV |
194 |
decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
195 |
{ |
196 |
if (expect_true (len >= 2 |
197 |
&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
198 |
&& IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
199 |
{ |
200 |
*clen = 2; |
201 |
return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
202 |
} |
203 |
else |
204 |
return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
205 |
} |
206 |
|
207 |
// likewise for encoding, also never called for ascii codepoints |
208 |
// this function takes advantage of this fact, although current gccs |
209 |
// seem to optimise the check for >= 0x80 away anyways |
210 |
INLINE unsigned char * |
211 |
encode_utf8 (unsigned char *s, UV ch) |
212 |
{ |
213 |
if (expect_false (ch < 0x000080)) |
214 |
*s++ = ch; |
215 |
else if (expect_true (ch < 0x000800)) |
216 |
*s++ = 0xc0 | ( ch >> 6), |
217 |
*s++ = 0x80 | ( ch & 0x3f); |
218 |
else if ( ch < 0x010000) |
219 |
*s++ = 0xe0 | ( ch >> 12), |
220 |
*s++ = 0x80 | ((ch >> 6) & 0x3f), |
221 |
*s++ = 0x80 | ( ch & 0x3f); |
222 |
else if ( ch < 0x110000) |
223 |
*s++ = 0xf0 | ( ch >> 18), |
224 |
*s++ = 0x80 | ((ch >> 12) & 0x3f), |
225 |
*s++ = 0x80 | ((ch >> 6) & 0x3f), |
226 |
*s++ = 0x80 | ( ch & 0x3f); |
227 |
|
228 |
return s; |
229 |
} |
230 |
|
231 |
// convert offset pointer to character index, sv must be string |
232 |
static STRLEN |
233 |
ptr_to_index (SV *sv, char *offset) |
234 |
{ |
235 |
return SvUTF8 (sv) |
236 |
? utf8_distance (offset, SvPVX (sv)) |
237 |
: offset - SvPVX (sv); |
238 |
} |
239 |
|
240 |
///////////////////////////////////////////////////////////////////////////// |
241 |
// fp hell |
242 |
|
243 |
// scan a group of digits, and a trailing exponent |
244 |
static void |
245 |
json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth) |
246 |
{ |
247 |
UV uaccum = 0; |
248 |
int eaccum = 0; |
249 |
|
250 |
// if we recurse too deep, skip all remaining digits |
251 |
// to avoid a stack overflow attack |
252 |
if (expect_false (--maxdepth <= 0)) |
253 |
while (((U8)*s - '0') < 10) |
254 |
++s; |
255 |
|
256 |
for (;;) |
257 |
{ |
258 |
U8 dig = (U8)*s - '0'; |
259 |
|
260 |
if (expect_false (dig >= 10)) |
261 |
{ |
262 |
if (dig == (U8)((U8)'.' - (U8)'0')) |
263 |
{ |
264 |
++s; |
265 |
json_atof_scan1 (s, accum, expo, 1, maxdepth); |
266 |
} |
267 |
else if ((dig | ' ') == 'e' - '0') |
268 |
{ |
269 |
int exp2 = 0; |
270 |
int neg = 0; |
271 |
|
272 |
++s; |
273 |
|
274 |
if (*s == '-') |
275 |
{ |
276 |
++s; |
277 |
neg = 1; |
278 |
} |
279 |
else if (*s == '+') |
280 |
++s; |
281 |
|
282 |
while ((dig = (U8)*s - '0') < 10) |
283 |
exp2 = exp2 * 10 + *s++ - '0'; |
284 |
|
285 |
*expo += neg ? -exp2 : exp2; |
286 |
} |
287 |
|
288 |
break; |
289 |
} |
290 |
|
291 |
++s; |
292 |
|
293 |
uaccum = uaccum * 10 + dig; |
294 |
++eaccum; |
295 |
|
296 |
// if we have too many digits, then recurse for more |
297 |
// we actually do this for rather few digits |
298 |
if (uaccum >= (UV_MAX - 9) / 10) |
299 |
{ |
300 |
if (postdp) *expo -= eaccum; |
301 |
json_atof_scan1 (s, accum, expo, postdp, maxdepth); |
302 |
if (postdp) *expo += eaccum; |
303 |
|
304 |
break; |
305 |
} |
306 |
} |
307 |
|
308 |
// this relies greatly on the quality of the pow () |
309 |
// implementation of the platform, but a good |
310 |
// implementation is hard to beat. |
311 |
// (IEEE 754 conformant ones are required to be exact) |
312 |
if (postdp) *expo -= eaccum; |
313 |
*accum += uaccum * Perl_pow (10., *expo); |
314 |
*expo += eaccum; |
315 |
} |
316 |
|
317 |
static NV |
318 |
json_atof (const char *s) |
319 |
{ |
320 |
NV accum = 0.; |
321 |
int expo = 0; |
322 |
int neg = 0; |
323 |
|
324 |
if (*s == '-') |
325 |
{ |
326 |
++s; |
327 |
neg = 1; |
328 |
} |
329 |
|
330 |
// a recursion depth of ten gives us >>500 bits |
331 |
json_atof_scan1 (s, &accum, &expo, 0, 10); |
332 |
|
333 |
return neg ? -accum : accum; |
334 |
} |
335 |
|
336 |
// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true |
337 |
static int |
338 |
ref_bool_type (SV *sv) |
339 |
{ |
340 |
svtype svt = SvTYPE (sv); |
341 |
|
342 |
if (svt < SVt_PVAV) |
343 |
{ |
344 |
STRLEN len = 0; |
345 |
char *pv = svt ? SvPV (sv, len) : 0; |
346 |
|
347 |
if (len == 1) |
348 |
if (*pv == '1') |
349 |
return 1; |
350 |
else if (*pv == '0') |
351 |
return 0; |
352 |
} |
353 |
|
354 |
return -1; |
355 |
} |
356 |
|
357 |
// returns whether scalar is not a reference in the sense of allow_nonref |
358 |
static int |
359 |
json_nonref (SV *scalar) |
360 |
{ |
361 |
if (!SvROK (scalar)) |
362 |
return 1; |
363 |
|
364 |
scalar = SvRV (scalar); |
365 |
|
366 |
if (SvTYPE (scalar) >= SVt_PVMG) |
367 |
{ |
368 |
if (SvSTASH (scalar) == bool_stash) |
369 |
return 1; |
370 |
|
371 |
if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0) |
372 |
return 1; |
373 |
} |
374 |
|
375 |
return 0; |
376 |
} |
377 |
|
378 |
///////////////////////////////////////////////////////////////////////////// |
379 |
// encoder |
380 |
|
381 |
// structure used for encoding JSON |
382 |
typedef struct |
383 |
{ |
384 |
char *cur; // SvPVX (sv) + current output position |
385 |
char *end; // SvEND (sv) |
386 |
SV *sv; // result scalar |
387 |
JSON json; |
388 |
U32 indent; // indentation level |
389 |
UV limit; // escape character values >= this value when encoding |
390 |
} enc_t; |
391 |
|
392 |
INLINE void |
393 |
need (enc_t *enc, STRLEN len) |
394 |
{ |
395 |
if (expect_false ((uintptr_t)(enc->end - enc->cur) < len)) |
396 |
{ |
397 |
STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
398 |
char *buf = json_sv_grow (enc->sv, cur, len); |
399 |
enc->cur = buf + cur; |
400 |
enc->end = buf + SvLEN (enc->sv) - 1; |
401 |
} |
402 |
} |
403 |
|
404 |
INLINE void |
405 |
encode_ch (enc_t *enc, char ch) |
406 |
{ |
407 |
need (enc, 1); |
408 |
*enc->cur++ = ch; |
409 |
} |
410 |
|
411 |
static void |
412 |
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
413 |
{ |
414 |
char *end = str + len; |
415 |
|
416 |
need (enc, len); |
417 |
|
418 |
while (str < end) |
419 |
{ |
420 |
unsigned char ch = *(unsigned char *)str; |
421 |
|
422 |
if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
423 |
{ |
424 |
if (expect_false (ch == '"')) // but with slow exceptions |
425 |
{ |
426 |
need (enc, len + 1); |
427 |
*enc->cur++ = '\\'; |
428 |
*enc->cur++ = '"'; |
429 |
} |
430 |
else if (expect_false (ch == '\\')) |
431 |
{ |
432 |
need (enc, len + 1); |
433 |
*enc->cur++ = '\\'; |
434 |
*enc->cur++ = '\\'; |
435 |
} |
436 |
else |
437 |
*enc->cur++ = ch; |
438 |
|
439 |
++str; |
440 |
} |
441 |
else |
442 |
{ |
443 |
switch (ch) |
444 |
{ |
445 |
case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
446 |
case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
447 |
case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
448 |
case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
449 |
case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
450 |
|
451 |
default: |
452 |
{ |
453 |
STRLEN clen; |
454 |
UV uch; |
455 |
|
456 |
if (is_utf8) |
457 |
{ |
458 |
uch = decode_utf8 (str, end - str, &clen); |
459 |
if (clen == (STRLEN)-1) |
460 |
croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
461 |
} |
462 |
else |
463 |
{ |
464 |
uch = ch; |
465 |
clen = 1; |
466 |
} |
467 |
|
468 |
if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
469 |
{ |
470 |
if (uch >= 0x10000UL) |
471 |
{ |
472 |
if (uch >= 0x110000UL) |
473 |
croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
474 |
|
475 |
need (enc, len + 11); |
476 |
sprintf (enc->cur, "\\u%04x\\u%04x", |
477 |
(int)((uch - 0x10000) / 0x400 + 0xD800), |
478 |
(int)((uch - 0x10000) % 0x400 + 0xDC00)); |
479 |
enc->cur += 12; |
480 |
} |
481 |
else |
482 |
{ |
483 |
need (enc, len + 5); |
484 |
*enc->cur++ = '\\'; |
485 |
*enc->cur++ = 'u'; |
486 |
*enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
487 |
*enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
488 |
*enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
489 |
*enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
490 |
} |
491 |
|
492 |
str += clen; |
493 |
} |
494 |
else if (enc->json.flags & F_LATIN1) |
495 |
{ |
496 |
*enc->cur++ = uch; |
497 |
str += clen; |
498 |
} |
499 |
else if (is_utf8) |
500 |
{ |
501 |
need (enc, len + clen); |
502 |
do |
503 |
{ |
504 |
*enc->cur++ = *str++; |
505 |
} |
506 |
while (--clen); |
507 |
} |
508 |
else |
509 |
{ |
510 |
need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
511 |
enc->cur = encode_utf8 (enc->cur, uch); |
512 |
++str; |
513 |
} |
514 |
} |
515 |
} |
516 |
} |
517 |
|
518 |
--len; |
519 |
} |
520 |
} |
521 |
|
522 |
INLINE void |
523 |
encode_indent (enc_t *enc) |
524 |
{ |
525 |
if (enc->json.flags & F_INDENT) |
526 |
{ |
527 |
int spaces = enc->indent * INDENT_STEP; |
528 |
|
529 |
need (enc, spaces); |
530 |
memset (enc->cur, ' ', spaces); |
531 |
enc->cur += spaces; |
532 |
} |
533 |
} |
534 |
|
535 |
INLINE void |
536 |
encode_space (enc_t *enc) |
537 |
{ |
538 |
need (enc, 1); |
539 |
encode_ch (enc, ' '); |
540 |
} |
541 |
|
542 |
INLINE void |
543 |
encode_nl (enc_t *enc) |
544 |
{ |
545 |
if (enc->json.flags & F_INDENT) |
546 |
{ |
547 |
need (enc, 1); |
548 |
encode_ch (enc, '\n'); |
549 |
} |
550 |
} |
551 |
|
552 |
INLINE void |
553 |
encode_comma (enc_t *enc) |
554 |
{ |
555 |
encode_ch (enc, ','); |
556 |
|
557 |
if (enc->json.flags & F_INDENT) |
558 |
encode_nl (enc); |
559 |
else if (enc->json.flags & F_SPACE_AFTER) |
560 |
encode_space (enc); |
561 |
} |
562 |
|
563 |
static void encode_sv (enc_t *enc, SV *sv); |
564 |
|
565 |
static void |
566 |
encode_av (enc_t *enc, AV *av) |
567 |
{ |
568 |
int i, len = av_len (av); |
569 |
|
570 |
if (enc->indent >= enc->json.max_depth) |
571 |
croak (ERR_NESTING_EXCEEDED); |
572 |
|
573 |
encode_ch (enc, '['); |
574 |
|
575 |
if (len >= 0) |
576 |
{ |
577 |
encode_nl (enc); ++enc->indent; |
578 |
|
579 |
for (i = 0; i <= len; ++i) |
580 |
{ |
581 |
SV **svp = av_fetch (av, i, 0); |
582 |
|
583 |
encode_indent (enc); |
584 |
|
585 |
if (svp) |
586 |
encode_sv (enc, *svp); |
587 |
else |
588 |
encode_str (enc, "null", 4, 0); |
589 |
|
590 |
if (i < len) |
591 |
encode_comma (enc); |
592 |
} |
593 |
|
594 |
encode_nl (enc); --enc->indent; encode_indent (enc); |
595 |
} |
596 |
|
597 |
encode_ch (enc, ']'); |
598 |
} |
599 |
|
600 |
static void |
601 |
encode_hk (enc_t *enc, HE *he) |
602 |
{ |
603 |
encode_ch (enc, '"'); |
604 |
|
605 |
if (HeKLEN (he) == HEf_SVKEY) |
606 |
{ |
607 |
SV *sv = HeSVKEY (he); |
608 |
STRLEN len; |
609 |
char *str; |
610 |
|
611 |
SvGETMAGIC (sv); |
612 |
str = SvPV (sv, len); |
613 |
|
614 |
encode_str (enc, str, len, SvUTF8 (sv)); |
615 |
} |
616 |
else |
617 |
encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
618 |
|
619 |
encode_ch (enc, '"'); |
620 |
|
621 |
if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
622 |
encode_ch (enc, ':'); |
623 |
if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
624 |
} |
625 |
|
626 |
// compare hash entries, used when all keys are bytestrings |
627 |
static int |
628 |
he_cmp_fast (const void *a_, const void *b_) |
629 |
{ |
630 |
int cmp; |
631 |
|
632 |
HE *a = *(HE **)a_; |
633 |
HE *b = *(HE **)b_; |
634 |
|
635 |
STRLEN la = HeKLEN (a); |
636 |
STRLEN lb = HeKLEN (b); |
637 |
|
638 |
if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
639 |
cmp = lb - la; |
640 |
|
641 |
return cmp; |
642 |
} |
643 |
|
644 |
// compare hash entries, used when some keys are sv's or utf-x |
645 |
static int |
646 |
he_cmp_slow (const void *a, const void *b) |
647 |
{ |
648 |
return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
649 |
} |
650 |
|
651 |
static void |
652 |
encode_hv (enc_t *enc, HV *hv) |
653 |
{ |
654 |
HE *he; |
655 |
|
656 |
if (enc->indent >= enc->json.max_depth) |
657 |
croak (ERR_NESTING_EXCEEDED); |
658 |
|
659 |
encode_ch (enc, '{'); |
660 |
|
661 |
// for canonical output we have to sort by keys first |
662 |
// actually, this is mostly due to the stupid so-called |
663 |
// security workaround added somewhere in 5.8.x |
664 |
// that randomises hash orderings |
665 |
if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
666 |
{ |
667 |
int count = hv_iterinit (hv); |
668 |
|
669 |
if (SvMAGICAL (hv)) |
670 |
{ |
671 |
// need to count by iterating. could improve by dynamically building the vector below |
672 |
// but I don't care for the speed of this special case. |
673 |
// note also that we will run into undefined behaviour when the two iterations |
674 |
// do not result in the same count, something I might care for in some later release. |
675 |
|
676 |
count = 0; |
677 |
while (hv_iternext (hv)) |
678 |
++count; |
679 |
|
680 |
hv_iterinit (hv); |
681 |
} |
682 |
|
683 |
if (count) |
684 |
{ |
685 |
int i, fast = 1; |
686 |
HE *hes_stack [STACK_HES]; |
687 |
HE **hes = hes_stack; |
688 |
|
689 |
// allocate larger arrays on the heap |
690 |
if (count > STACK_HES) |
691 |
{ |
692 |
SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes))); |
693 |
hes = (HE **)SvPVX (sv); |
694 |
} |
695 |
|
696 |
i = 0; |
697 |
while ((he = hv_iternext (hv))) |
698 |
{ |
699 |
hes [i++] = he; |
700 |
if (HeKLEN (he) < 0 || HeKUTF8 (he)) |
701 |
fast = 0; |
702 |
} |
703 |
|
704 |
assert (i == count); |
705 |
|
706 |
if (fast) |
707 |
qsort (hes, count, sizeof (HE *), he_cmp_fast); |
708 |
else |
709 |
{ |
710 |
// hack to forcefully disable "use bytes" |
711 |
COP cop = *PL_curcop; |
712 |
cop.op_private = 0; |
713 |
|
714 |
ENTER; |
715 |
SAVETMPS; |
716 |
|
717 |
SAVEVPTR (PL_curcop); |
718 |
PL_curcop = &cop; |
719 |
|
720 |
qsort (hes, count, sizeof (HE *), he_cmp_slow); |
721 |
|
722 |
FREETMPS; |
723 |
LEAVE; |
724 |
} |
725 |
|
726 |
encode_nl (enc); ++enc->indent; |
727 |
|
728 |
while (count--) |
729 |
{ |
730 |
encode_indent (enc); |
731 |
he = hes [count]; |
732 |
encode_hk (enc, he); |
733 |
encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
734 |
|
735 |
if (count) |
736 |
encode_comma (enc); |
737 |
} |
738 |
|
739 |
encode_nl (enc); --enc->indent; encode_indent (enc); |
740 |
} |
741 |
} |
742 |
else |
743 |
{ |
744 |
if (hv_iterinit (hv) || SvMAGICAL (hv)) |
745 |
if ((he = hv_iternext (hv))) |
746 |
{ |
747 |
encode_nl (enc); ++enc->indent; |
748 |
|
749 |
for (;;) |
750 |
{ |
751 |
encode_indent (enc); |
752 |
encode_hk (enc, he); |
753 |
encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
754 |
|
755 |
if (!(he = hv_iternext (hv))) |
756 |
break; |
757 |
|
758 |
encode_comma (enc); |
759 |
} |
760 |
|
761 |
encode_nl (enc); --enc->indent; encode_indent (enc); |
762 |
} |
763 |
} |
764 |
|
765 |
encode_ch (enc, '}'); |
766 |
} |
767 |
|
768 |
// encode objects, arrays and special \0=false and \1=true values. |
769 |
static void |
770 |
encode_rv (enc_t *enc, SV *sv) |
771 |
{ |
772 |
svtype svt; |
773 |
GV *method; |
774 |
|
775 |
SvGETMAGIC (sv); |
776 |
svt = SvTYPE (sv); |
777 |
|
778 |
if (expect_false (SvOBJECT (sv))) |
779 |
{ |
780 |
HV *stash = SvSTASH (sv); |
781 |
|
782 |
if (stash == bool_stash) |
783 |
{ |
784 |
if (SvIV (sv)) encode_str (enc, "true" , 4, 0); |
785 |
else encode_str (enc, "false", 5, 0); |
786 |
} |
787 |
else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0))) |
788 |
{ |
789 |
int count; |
790 |
dSP; |
791 |
|
792 |
ENTER; SAVETMPS; |
793 |
SAVESTACK_POS (); |
794 |
PUSHMARK (SP); |
795 |
EXTEND (SP, 2); |
796 |
// we re-bless the reference to get overload and other niceties right |
797 |
PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
798 |
PUSHs (sv_json); |
799 |
|
800 |
PUTBACK; |
801 |
count = call_sv ((SV *)GvCV (method), G_ARRAY); |
802 |
SPAGAIN; |
803 |
|
804 |
// catch this surprisingly common error |
805 |
if (SvROK (TOPs) && SvRV (TOPs) == sv) |
806 |
croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
807 |
|
808 |
encode_ch (enc, '('); |
809 |
encode_ch (enc, '"'); |
810 |
encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash)); |
811 |
encode_ch (enc, '"'); |
812 |
encode_ch (enc, ')'); |
813 |
encode_ch (enc, '['); |
814 |
|
815 |
while (count) |
816 |
{ |
817 |
encode_sv (enc, SP[1 - count--]); |
818 |
|
819 |
if (count) |
820 |
encode_ch (enc, ','); |
821 |
} |
822 |
|
823 |
encode_ch (enc, ']'); |
824 |
|
825 |
FREETMPS; LEAVE; |
826 |
} |
827 |
else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0))) |
828 |
{ |
829 |
dSP; |
830 |
|
831 |
ENTER; SAVETMPS; |
832 |
PUSHMARK (SP); |
833 |
// we re-bless the reference to get overload and other niceties right |
834 |
XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash)); |
835 |
|
836 |
// calling with G_SCALAR ensures that we always get a 1 return value |
837 |
PUTBACK; |
838 |
call_sv ((SV *)GvCV (method), G_SCALAR); |
839 |
SPAGAIN; |
840 |
|
841 |
// catch this surprisingly common error |
842 |
if (SvROK (TOPs) && SvRV (TOPs) == sv) |
843 |
croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
844 |
|
845 |
sv = POPs; |
846 |
PUTBACK; |
847 |
|
848 |
encode_sv (enc, sv); |
849 |
|
850 |
FREETMPS; LEAVE; |
851 |
} |
852 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
853 |
encode_str (enc, "null", 4, 0); |
854 |
else |
855 |
croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)", |
856 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
857 |
} |
858 |
else if (svt == SVt_PVHV) |
859 |
encode_hv (enc, (HV *)sv); |
860 |
else if (svt == SVt_PVAV) |
861 |
encode_av (enc, (AV *)sv); |
862 |
else if (svt < SVt_PVAV) |
863 |
{ |
864 |
int bool_type = ref_bool_type (sv); |
865 |
|
866 |
if (bool_type == 1) |
867 |
encode_str (enc, "true", 4, 0); |
868 |
else if (bool_type == 0) |
869 |
encode_str (enc, "false", 5, 0); |
870 |
else if (enc->json.flags & F_ALLOW_UNKNOWN) |
871 |
encode_str (enc, "null", 4, 0); |
872 |
else |
873 |
croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
874 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
875 |
} |
876 |
else if (enc->json.flags & F_ALLOW_UNKNOWN) |
877 |
encode_str (enc, "null", 4, 0); |
878 |
else |
879 |
croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
880 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
881 |
} |
882 |
|
883 |
static void |
884 |
encode_sv (enc_t *enc, SV *sv) |
885 |
{ |
886 |
SvGETMAGIC (sv); |
887 |
|
888 |
if (SvPOKp (sv)) |
889 |
{ |
890 |
STRLEN len; |
891 |
char *str = SvPV (sv, len); |
892 |
encode_ch (enc, '"'); |
893 |
encode_str (enc, str, len, SvUTF8 (sv)); |
894 |
encode_ch (enc, '"'); |
895 |
} |
896 |
else if (SvNOKp (sv)) |
897 |
{ |
898 |
// trust that perl will do the right thing w.r.t. JSON syntax. |
899 |
need (enc, NV_DIG + 32); |
900 |
Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
901 |
enc->cur += strlen (enc->cur); |
902 |
} |
903 |
else if (SvIOKp (sv)) |
904 |
{ |
905 |
// we assume we can always read an IV as a UV and vice versa |
906 |
// we assume two's complement |
907 |
// we assume no aliasing issues in the union |
908 |
if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
909 |
: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
910 |
{ |
911 |
// optimise the "small number case" |
912 |
// code will likely be branchless and use only a single multiplication |
913 |
// works for numbers up to 59074 |
914 |
I32 i = SvIVX (sv); |
915 |
U32 u; |
916 |
char digit, nz = 0; |
917 |
|
918 |
need (enc, 6); |
919 |
|
920 |
*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
921 |
u = i < 0 ? -i : i; |
922 |
|
923 |
// convert to 4.28 fixed-point representation |
924 |
u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
925 |
|
926 |
// now output digit by digit, each time masking out the integer part |
927 |
// and multiplying by 5 while moving the decimal point one to the right, |
928 |
// resulting in a net multiplication by 10. |
929 |
// we always write the digit to memory but conditionally increment |
930 |
// the pointer, to enable the use of conditional move instructions. |
931 |
digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
932 |
digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
933 |
digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
934 |
digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
935 |
digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
936 |
} |
937 |
else |
938 |
{ |
939 |
// large integer, use the (rather slow) snprintf way. |
940 |
need (enc, IVUV_MAXCHARS); |
941 |
enc->cur += |
942 |
SvIsUV(sv) |
943 |
? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
944 |
: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
945 |
} |
946 |
} |
947 |
else if (SvROK (sv)) |
948 |
encode_rv (enc, SvRV (sv)); |
949 |
else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
950 |
encode_str (enc, "null", 4, 0); |
951 |
else |
952 |
croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data", |
953 |
SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
954 |
} |
955 |
|
956 |
static SV * |
957 |
encode_json (SV *scalar, JSON *json) |
958 |
{ |
959 |
enc_t enc; |
960 |
|
961 |
if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar)) |
962 |
croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
963 |
|
964 |
enc.json = *json; |
965 |
enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
966 |
enc.cur = SvPVX (enc.sv); |
967 |
enc.end = SvEND (enc.sv); |
968 |
enc.indent = 0; |
969 |
enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
970 |
: enc.json.flags & F_LATIN1 ? 0x000100UL |
971 |
: 0x110000UL; |
972 |
|
973 |
SvPOK_only (enc.sv); |
974 |
encode_sv (&enc, scalar); |
975 |
encode_nl (&enc); |
976 |
|
977 |
SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
978 |
*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
979 |
|
980 |
if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
981 |
SvUTF8_on (enc.sv); |
982 |
|
983 |
if (enc.json.flags & F_SHRINK) |
984 |
shrink (enc.sv); |
985 |
|
986 |
return enc.sv; |
987 |
} |
988 |
|
989 |
///////////////////////////////////////////////////////////////////////////// |
990 |
// decoder |
991 |
|
992 |
// structure used for decoding JSON |
993 |
typedef struct |
994 |
{ |
995 |
char *cur; // current parser pointer |
996 |
char *end; // end of input string |
997 |
const char *err; // parse error, if != 0 |
998 |
JSON json; |
999 |
U32 depth; // recursion depth |
1000 |
U32 maxdepth; // recursion depth limit |
1001 |
} dec_t; |
1002 |
|
1003 |
INLINE void |
1004 |
decode_comment (dec_t *dec) |
1005 |
{ |
1006 |
// only '#'-style comments allowed a.t.m. |
1007 |
|
1008 |
while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
1009 |
++dec->cur; |
1010 |
} |
1011 |
|
1012 |
INLINE void |
1013 |
decode_ws (dec_t *dec) |
1014 |
{ |
1015 |
for (;;) |
1016 |
{ |
1017 |
char ch = *dec->cur; |
1018 |
|
1019 |
if (ch > 0x20) |
1020 |
{ |
1021 |
if (expect_false (ch == '#')) |
1022 |
{ |
1023 |
if (dec->json.flags & F_RELAXED) |
1024 |
decode_comment (dec); |
1025 |
else |
1026 |
break; |
1027 |
} |
1028 |
else |
1029 |
break; |
1030 |
} |
1031 |
else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
1032 |
break; // parse error, but let higher level handle it, gives better error messages |
1033 |
|
1034 |
++dec->cur; |
1035 |
} |
1036 |
} |
1037 |
|
1038 |
#define ERR(reason) SB dec->err = reason; goto fail; SE |
1039 |
|
1040 |
#define EXPECT_CH(ch) SB \ |
1041 |
if (*dec->cur != ch) \ |
1042 |
ERR (# ch " expected"); \ |
1043 |
++dec->cur; \ |
1044 |
SE |
1045 |
|
1046 |
#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
1047 |
#define DEC_DEC_DEPTH --dec->depth |
1048 |
|
1049 |
static SV *decode_sv (dec_t *dec); |
1050 |
|
1051 |
static signed char decode_hexdigit[256]; |
1052 |
|
1053 |
static UV |
1054 |
decode_4hex (dec_t *dec) |
1055 |
{ |
1056 |
signed char d1, d2, d3, d4; |
1057 |
unsigned char *cur = (unsigned char *)dec->cur; |
1058 |
|
1059 |
d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
1060 |
d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
1061 |
d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
1062 |
d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
1063 |
|
1064 |
dec->cur += 4; |
1065 |
|
1066 |
return ((UV)d1) << 12 |
1067 |
| ((UV)d2) << 8 |
1068 |
| ((UV)d3) << 4 |
1069 |
| ((UV)d4); |
1070 |
|
1071 |
fail: |
1072 |
return (UV)-1; |
1073 |
} |
1074 |
|
1075 |
static SV * |
1076 |
decode_str (dec_t *dec) |
1077 |
{ |
1078 |
SV *sv = 0; |
1079 |
int utf8 = 0; |
1080 |
char *dec_cur = dec->cur; |
1081 |
|
1082 |
do |
1083 |
{ |
1084 |
char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
1085 |
char *cur = buf; |
1086 |
|
1087 |
do |
1088 |
{ |
1089 |
unsigned char ch = *(unsigned char *)dec_cur++; |
1090 |
|
1091 |
if (expect_false (ch == '"')) |
1092 |
{ |
1093 |
--dec_cur; |
1094 |
break; |
1095 |
} |
1096 |
else if (expect_false (ch == '\\')) |
1097 |
{ |
1098 |
switch (*dec_cur) |
1099 |
{ |
1100 |
case '\\': |
1101 |
case '/': |
1102 |
case '"': *cur++ = *dec_cur++; break; |
1103 |
|
1104 |
case 'b': ++dec_cur; *cur++ = '\010'; break; |
1105 |
case 't': ++dec_cur; *cur++ = '\011'; break; |
1106 |
case 'n': ++dec_cur; *cur++ = '\012'; break; |
1107 |
case 'f': ++dec_cur; *cur++ = '\014'; break; |
1108 |
case 'r': ++dec_cur; *cur++ = '\015'; break; |
1109 |
|
1110 |
case 'u': |
1111 |
{ |
1112 |
UV lo, hi; |
1113 |
++dec_cur; |
1114 |
|
1115 |
dec->cur = dec_cur; |
1116 |
hi = decode_4hex (dec); |
1117 |
dec_cur = dec->cur; |
1118 |
if (hi == (UV)-1) |
1119 |
goto fail; |
1120 |
|
1121 |
// possibly a surrogate pair |
1122 |
if (hi >= 0xd800) |
1123 |
if (hi < 0xdc00) |
1124 |
{ |
1125 |
if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
1126 |
ERR ("missing low surrogate character in surrogate pair"); |
1127 |
|
1128 |
dec_cur += 2; |
1129 |
|
1130 |
dec->cur = dec_cur; |
1131 |
lo = decode_4hex (dec); |
1132 |
dec_cur = dec->cur; |
1133 |
if (lo == (UV)-1) |
1134 |
goto fail; |
1135 |
|
1136 |
if (lo < 0xdc00 || lo >= 0xe000) |
1137 |
ERR ("surrogate pair expected"); |
1138 |
|
1139 |
hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
1140 |
} |
1141 |
else if (hi < 0xe000) |
1142 |
ERR ("missing high surrogate character in surrogate pair"); |
1143 |
|
1144 |
if (hi >= 0x80) |
1145 |
{ |
1146 |
utf8 = 1; |
1147 |
|
1148 |
cur = encode_utf8 (cur, hi); |
1149 |
} |
1150 |
else |
1151 |
*cur++ = hi; |
1152 |
} |
1153 |
break; |
1154 |
|
1155 |
default: |
1156 |
--dec_cur; |
1157 |
ERR ("illegal backslash escape sequence in string"); |
1158 |
} |
1159 |
} |
1160 |
else if (expect_true (ch >= 0x20 && ch < 0x80)) |
1161 |
*cur++ = ch; |
1162 |
else if (ch >= 0x80) |
1163 |
{ |
1164 |
STRLEN clen; |
1165 |
|
1166 |
--dec_cur; |
1167 |
|
1168 |
decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1169 |
if (clen == (STRLEN)-1) |
1170 |
ERR ("malformed UTF-8 character in JSON string"); |
1171 |
|
1172 |
do |
1173 |
*cur++ = *dec_cur++; |
1174 |
while (--clen); |
1175 |
|
1176 |
utf8 = 1; |
1177 |
} |
1178 |
else if (ch == '\t' && dec->json.flags & F_RELAXED) |
1179 |
*cur++ = ch; |
1180 |
else |
1181 |
{ |
1182 |
--dec_cur; |
1183 |
|
1184 |
if (!ch) |
1185 |
ERR ("unexpected end of string while parsing JSON string"); |
1186 |
else |
1187 |
ERR ("invalid character encountered while parsing JSON string"); |
1188 |
} |
1189 |
} |
1190 |
while (cur < buf + SHORT_STRING_LEN); |
1191 |
|
1192 |
{ |
1193 |
STRLEN len = cur - buf; |
1194 |
|
1195 |
if (sv) |
1196 |
{ |
1197 |
STRLEN cur = SvCUR (sv); |
1198 |
|
1199 |
if (SvLEN (sv) - cur <= len) |
1200 |
json_sv_grow (sv, cur, len); |
1201 |
|
1202 |
memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1203 |
SvCUR_set (sv, SvCUR (sv) + len); |
1204 |
} |
1205 |
else |
1206 |
sv = newSVpvn (buf, len); |
1207 |
} |
1208 |
} |
1209 |
while (*dec_cur != '"'); |
1210 |
|
1211 |
++dec_cur; |
1212 |
|
1213 |
if (sv) |
1214 |
{ |
1215 |
SvPOK_only (sv); |
1216 |
*SvEND (sv) = 0; |
1217 |
|
1218 |
if (utf8) |
1219 |
SvUTF8_on (sv); |
1220 |
} |
1221 |
else |
1222 |
sv = newSVpvn ("", 0); |
1223 |
|
1224 |
dec->cur = dec_cur; |
1225 |
return sv; |
1226 |
|
1227 |
fail: |
1228 |
dec->cur = dec_cur; |
1229 |
return 0; |
1230 |
} |
1231 |
|
1232 |
static SV * |
1233 |
decode_num (dec_t *dec) |
1234 |
{ |
1235 |
int is_nv = 0; |
1236 |
char *start = dec->cur; |
1237 |
|
1238 |
// [minus] |
1239 |
if (*dec->cur == '-') |
1240 |
++dec->cur; |
1241 |
|
1242 |
if (*dec->cur == '0') |
1243 |
{ |
1244 |
++dec->cur; |
1245 |
if (*dec->cur >= '0' && *dec->cur <= '9') |
1246 |
ERR ("malformed number (leading zero must not be followed by another digit)"); |
1247 |
} |
1248 |
else if (*dec->cur < '0' || *dec->cur > '9') |
1249 |
ERR ("malformed number (no digits after initial minus)"); |
1250 |
else |
1251 |
do |
1252 |
{ |
1253 |
++dec->cur; |
1254 |
} |
1255 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
1256 |
|
1257 |
// [frac] |
1258 |
if (*dec->cur == '.') |
1259 |
{ |
1260 |
++dec->cur; |
1261 |
|
1262 |
if (*dec->cur < '0' || *dec->cur > '9') |
1263 |
ERR ("malformed number (no digits after decimal point)"); |
1264 |
|
1265 |
do |
1266 |
{ |
1267 |
++dec->cur; |
1268 |
} |
1269 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
1270 |
|
1271 |
is_nv = 1; |
1272 |
} |
1273 |
|
1274 |
// [exp] |
1275 |
if (*dec->cur == 'e' || *dec->cur == 'E') |
1276 |
{ |
1277 |
++dec->cur; |
1278 |
|
1279 |
if (*dec->cur == '-' || *dec->cur == '+') |
1280 |
++dec->cur; |
1281 |
|
1282 |
if (*dec->cur < '0' || *dec->cur > '9') |
1283 |
ERR ("malformed number (no digits after exp sign)"); |
1284 |
|
1285 |
do |
1286 |
{ |
1287 |
++dec->cur; |
1288 |
} |
1289 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
1290 |
|
1291 |
is_nv = 1; |
1292 |
} |
1293 |
|
1294 |
if (!is_nv) |
1295 |
{ |
1296 |
int len = dec->cur - start; |
1297 |
|
1298 |
// special case the rather common 1..5-digit-int case |
1299 |
if (*start == '-') |
1300 |
switch (len) |
1301 |
{ |
1302 |
case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
1303 |
case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
1304 |
case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1305 |
case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1306 |
case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1307 |
} |
1308 |
else |
1309 |
switch (len) |
1310 |
{ |
1311 |
case 1: return newSViv ( start [0] - '0' * 1); |
1312 |
case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1313 |
case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1314 |
case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1315 |
case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1316 |
} |
1317 |
|
1318 |
{ |
1319 |
UV uv; |
1320 |
int numtype = grok_number (start, len, &uv); |
1321 |
if (numtype & IS_NUMBER_IN_UV) |
1322 |
if (numtype & IS_NUMBER_NEG) |
1323 |
{ |
1324 |
if (uv < (UV)IV_MIN) |
1325 |
return newSViv (-(IV)uv); |
1326 |
} |
1327 |
else |
1328 |
return newSVuv (uv); |
1329 |
} |
1330 |
|
1331 |
len -= *start == '-' ? 1 : 0; |
1332 |
|
1333 |
// does not fit into IV or UV, try NV |
1334 |
if (len <= NV_DIG) |
1335 |
// fits into NV without loss of precision |
1336 |
return newSVnv (json_atof (start)); |
1337 |
|
1338 |
// everything else fails, convert it to a string |
1339 |
return newSVpvn (start, dec->cur - start); |
1340 |
} |
1341 |
|
1342 |
// loss of precision here |
1343 |
return newSVnv (json_atof (start)); |
1344 |
|
1345 |
fail: |
1346 |
return 0; |
1347 |
} |
1348 |
|
1349 |
static SV * |
1350 |
decode_av (dec_t *dec) |
1351 |
{ |
1352 |
AV *av = newAV (); |
1353 |
|
1354 |
DEC_INC_DEPTH; |
1355 |
decode_ws (dec); |
1356 |
|
1357 |
if (*dec->cur == ']') |
1358 |
++dec->cur; |
1359 |
else |
1360 |
for (;;) |
1361 |
{ |
1362 |
SV *value; |
1363 |
|
1364 |
value = decode_sv (dec); |
1365 |
if (!value) |
1366 |
goto fail; |
1367 |
|
1368 |
av_push (av, value); |
1369 |
|
1370 |
decode_ws (dec); |
1371 |
|
1372 |
if (*dec->cur == ']') |
1373 |
{ |
1374 |
++dec->cur; |
1375 |
break; |
1376 |
} |
1377 |
|
1378 |
if (*dec->cur != ',') |
1379 |
ERR (", or ] expected while parsing array"); |
1380 |
|
1381 |
++dec->cur; |
1382 |
|
1383 |
decode_ws (dec); |
1384 |
|
1385 |
if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
1386 |
{ |
1387 |
++dec->cur; |
1388 |
break; |
1389 |
} |
1390 |
} |
1391 |
|
1392 |
DEC_DEC_DEPTH; |
1393 |
return newRV_noinc ((SV *)av); |
1394 |
|
1395 |
fail: |
1396 |
SvREFCNT_dec (av); |
1397 |
DEC_DEC_DEPTH; |
1398 |
return 0; |
1399 |
} |
1400 |
|
1401 |
static SV * |
1402 |
decode_hv (dec_t *dec) |
1403 |
{ |
1404 |
SV *sv; |
1405 |
HV *hv = newHV (); |
1406 |
|
1407 |
DEC_INC_DEPTH; |
1408 |
decode_ws (dec); |
1409 |
|
1410 |
if (*dec->cur == '}') |
1411 |
++dec->cur; |
1412 |
else |
1413 |
for (;;) |
1414 |
{ |
1415 |
EXPECT_CH ('"'); |
1416 |
|
1417 |
// heuristic: assume that |
1418 |
// a) decode_str + hv_store_ent are abysmally slow. |
1419 |
// b) most hash keys are short, simple ascii text. |
1420 |
// => try to "fast-match" such strings to avoid |
1421 |
// the overhead of decode_str + hv_store_ent. |
1422 |
{ |
1423 |
SV *value; |
1424 |
char *p = dec->cur; |
1425 |
char *e = p + 24; // only try up to 24 bytes |
1426 |
|
1427 |
for (;;) |
1428 |
{ |
1429 |
// the >= 0x80 is false on most architectures |
1430 |
if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1431 |
{ |
1432 |
// slow path, back up and use decode_str |
1433 |
SV *key = decode_str (dec); |
1434 |
if (!key) |
1435 |
goto fail; |
1436 |
|
1437 |
decode_ws (dec); EXPECT_CH (':'); |
1438 |
|
1439 |
decode_ws (dec); |
1440 |
value = decode_sv (dec); |
1441 |
if (!value) |
1442 |
{ |
1443 |
SvREFCNT_dec (key); |
1444 |
goto fail; |
1445 |
} |
1446 |
|
1447 |
hv_store_ent (hv, key, value, 0); |
1448 |
SvREFCNT_dec (key); |
1449 |
|
1450 |
break; |
1451 |
} |
1452 |
else if (*p == '"') |
1453 |
{ |
1454 |
// fast path, got a simple key |
1455 |
char *key = dec->cur; |
1456 |
int len = p - key; |
1457 |
dec->cur = p + 1; |
1458 |
|
1459 |
decode_ws (dec); EXPECT_CH (':'); |
1460 |
|
1461 |
decode_ws (dec); |
1462 |
value = decode_sv (dec); |
1463 |
if (!value) |
1464 |
goto fail; |
1465 |
|
1466 |
hv_store (hv, key, len, value, 0); |
1467 |
|
1468 |
break; |
1469 |
} |
1470 |
|
1471 |
++p; |
1472 |
} |
1473 |
} |
1474 |
|
1475 |
decode_ws (dec); |
1476 |
|
1477 |
if (*dec->cur == '}') |
1478 |
{ |
1479 |
++dec->cur; |
1480 |
break; |
1481 |
} |
1482 |
|
1483 |
if (*dec->cur != ',') |
1484 |
ERR (", or } expected while parsing object/hash"); |
1485 |
|
1486 |
++dec->cur; |
1487 |
|
1488 |
decode_ws (dec); |
1489 |
|
1490 |
if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
1491 |
{ |
1492 |
++dec->cur; |
1493 |
break; |
1494 |
} |
1495 |
} |
1496 |
|
1497 |
DEC_DEC_DEPTH; |
1498 |
sv = newRV_noinc ((SV *)hv); |
1499 |
|
1500 |
// check filter callbacks |
1501 |
if (dec->json.flags & F_HOOK) |
1502 |
{ |
1503 |
if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
1504 |
{ |
1505 |
HE *cb, *he; |
1506 |
|
1507 |
hv_iterinit (hv); |
1508 |
he = hv_iternext (hv); |
1509 |
hv_iterinit (hv); |
1510 |
|
1511 |
// the next line creates a mortal sv each time it's called. |
1512 |
// might want to optimise this for common cases. |
1513 |
cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1514 |
|
1515 |
if (cb) |
1516 |
{ |
1517 |
dSP; |
1518 |
int count; |
1519 |
|
1520 |
ENTER; SAVETMPS; |
1521 |
SAVESTACK_POS (); |
1522 |
PUSHMARK (SP); |
1523 |
XPUSHs (HeVAL (he)); |
1524 |
sv_2mortal (sv); |
1525 |
|
1526 |
PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1527 |
|
1528 |
if (count == 1) |
1529 |
{ |
1530 |
sv = newSVsv (POPs); |
1531 |
FREETMPS; LEAVE; |
1532 |
return sv; |
1533 |
} |
1534 |
|
1535 |
SvREFCNT_inc (sv); |
1536 |
FREETMPS; LEAVE; |
1537 |
} |
1538 |
} |
1539 |
|
1540 |
if (dec->json.cb_object) |
1541 |
{ |
1542 |
dSP; |
1543 |
int count; |
1544 |
|
1545 |
ENTER; SAVETMPS; |
1546 |
SAVESTACK_POS (); |
1547 |
PUSHMARK (SP); |
1548 |
XPUSHs (sv_2mortal (sv)); |
1549 |
|
1550 |
PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1551 |
|
1552 |
if (count == 1) |
1553 |
{ |
1554 |
sv = newSVsv (POPs); |
1555 |
FREETMPS; LEAVE; |
1556 |
return sv; |
1557 |
} |
1558 |
|
1559 |
SvREFCNT_inc (sv); |
1560 |
FREETMPS; LEAVE; |
1561 |
} |
1562 |
} |
1563 |
|
1564 |
return sv; |
1565 |
|
1566 |
fail: |
1567 |
SvREFCNT_dec (hv); |
1568 |
DEC_DEC_DEPTH; |
1569 |
return 0; |
1570 |
} |
1571 |
|
1572 |
static SV * |
1573 |
decode_tag (dec_t *dec) |
1574 |
{ |
1575 |
SV *tag = 0; |
1576 |
SV *val = 0; |
1577 |
|
1578 |
if (!(dec->json.flags & F_ALLOW_TAGS)) |
1579 |
ERR ("malformed JSON string, neither array, object, number, string or atom"); |
1580 |
|
1581 |
++dec->cur; |
1582 |
|
1583 |
decode_ws (dec); |
1584 |
|
1585 |
tag = decode_sv (dec); |
1586 |
if (!tag) |
1587 |
goto fail; |
1588 |
|
1589 |
if (!SvPOK (tag)) |
1590 |
ERR ("malformed JSON string, (tag) must be a string"); |
1591 |
|
1592 |
decode_ws (dec); |
1593 |
|
1594 |
if (*dec->cur != ')') |
1595 |
ERR (") expected after tag"); |
1596 |
|
1597 |
++dec->cur; |
1598 |
|
1599 |
decode_ws (dec); |
1600 |
|
1601 |
val = decode_sv (dec); |
1602 |
if (!val) |
1603 |
goto fail; |
1604 |
|
1605 |
if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV) |
1606 |
ERR ("malformed JSON string, tag value must be an array"); |
1607 |
|
1608 |
{ |
1609 |
AV *av = (AV *)SvRV (val); |
1610 |
int i, len = av_len (av) + 1; |
1611 |
HV *stash = gv_stashsv (tag, 0); |
1612 |
SV *sv; |
1613 |
|
1614 |
if (!stash) |
1615 |
ERR ("cannot decode perl-object (package does not exist)"); |
1616 |
|
1617 |
GV *method = gv_fetchmethod_autoload (stash, "THAW", 0); |
1618 |
|
1619 |
if (!method) |
1620 |
ERR ("cannot decode perl-object (package does not have a THAW method)"); |
1621 |
|
1622 |
dSP; |
1623 |
|
1624 |
ENTER; SAVETMPS; |
1625 |
PUSHMARK (SP); |
1626 |
EXTEND (SP, len + 2); |
1627 |
// we re-bless the reference to get overload and other niceties right |
1628 |
PUSHs (tag); |
1629 |
PUSHs (sv_json); |
1630 |
|
1631 |
for (i = 0; i < len; ++i) |
1632 |
PUSHs (*av_fetch (av, i, 1)); |
1633 |
|
1634 |
PUTBACK; |
1635 |
call_sv ((SV *)GvCV (method), G_SCALAR); |
1636 |
SPAGAIN; |
1637 |
|
1638 |
SvREFCNT_dec (tag); |
1639 |
SvREFCNT_dec (val); |
1640 |
sv = SvREFCNT_inc (POPs); |
1641 |
|
1642 |
PUTBACK; |
1643 |
|
1644 |
FREETMPS; LEAVE; |
1645 |
|
1646 |
return sv; |
1647 |
} |
1648 |
|
1649 |
fail: |
1650 |
SvREFCNT_dec (tag); |
1651 |
SvREFCNT_dec (val); |
1652 |
return 0; |
1653 |
} |
1654 |
|
1655 |
static SV * |
1656 |
decode_sv (dec_t *dec) |
1657 |
{ |
1658 |
// the beauty of JSON: you need exactly one character lookahead |
1659 |
// to parse everything. |
1660 |
switch (*dec->cur) |
1661 |
{ |
1662 |
case '"': ++dec->cur; return decode_str (dec); |
1663 |
case '[': ++dec->cur; return decode_av (dec); |
1664 |
case '{': ++dec->cur; return decode_hv (dec); |
1665 |
case '(': return decode_tag (dec); |
1666 |
|
1667 |
case '-': |
1668 |
case '0': case '1': case '2': case '3': case '4': |
1669 |
case '5': case '6': case '7': case '8': case '9': |
1670 |
return decode_num (dec); |
1671 |
|
1672 |
case 'f': |
1673 |
if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1674 |
{ |
1675 |
dec->cur += 5; |
1676 |
|
1677 |
if (expect_false (!dec->json.v_false)) |
1678 |
dec->json.v_false = GET_BOOL (false); |
1679 |
|
1680 |
return newSVsv (dec->json.v_false); |
1681 |
} |
1682 |
else |
1683 |
ERR ("'false' expected"); |
1684 |
|
1685 |
break; |
1686 |
|
1687 |
case 't': |
1688 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1689 |
{ |
1690 |
dec->cur += 4; |
1691 |
|
1692 |
if (expect_false (!dec->json.v_true)) |
1693 |
dec->json.v_true = GET_BOOL (true); |
1694 |
|
1695 |
return newSVsv (dec->json.v_true); |
1696 |
} |
1697 |
else |
1698 |
ERR ("'true' expected"); |
1699 |
|
1700 |
break; |
1701 |
|
1702 |
case 'n': |
1703 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
1704 |
{ |
1705 |
dec->cur += 4; |
1706 |
return newSVsv (&PL_sv_undef); |
1707 |
} |
1708 |
else |
1709 |
ERR ("'null' expected"); |
1710 |
|
1711 |
break; |
1712 |
|
1713 |
default: |
1714 |
ERR ("malformed JSON string, neither tag, array, object, number, string or atom"); |
1715 |
break; |
1716 |
} |
1717 |
|
1718 |
fail: |
1719 |
return 0; |
1720 |
} |
1721 |
|
1722 |
static SV * |
1723 |
decode_json (SV *string, JSON *json, STRLEN *offset_return) |
1724 |
{ |
1725 |
dec_t dec; |
1726 |
SV *sv; |
1727 |
|
1728 |
/* work around bugs in 5.10 where manipulating magic values |
1729 |
* makes perl ignore the magic in subsequent accesses. |
1730 |
* also make a copy of non-PV values, to get them into a clean |
1731 |
* state (SvPV should do that, but it's buggy, see below). |
1732 |
* |
1733 |
* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16), |
1734 |
* as reported by Reini Urban. |
1735 |
*/ |
1736 |
/*SvGETMAGIC (string);*/ |
1737 |
if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string)) |
1738 |
string = sv_2mortal (newSVsv (string)); |
1739 |
|
1740 |
SvUPGRADE (string, SVt_PV); |
1741 |
|
1742 |
/* work around a bug in perl 5.10, which causes SvCUR to fail an |
1743 |
* assertion with -DDEBUGGING, although SvCUR is documented to |
1744 |
* return the xpv_cur field which certainly exists after upgrading. |
1745 |
* according to nicholas clark, calling SvPOK fixes this. |
1746 |
* But it doesn't fix it, so try another workaround, call SvPV_nolen |
1747 |
* and hope for the best. |
1748 |
* Damnit, SvPV_nolen still trips over yet another assertion. This |
1749 |
* assertion business is seriously broken, try yet another workaround |
1750 |
* for the broken -DDEBUGGING. |
1751 |
*/ |
1752 |
{ |
1753 |
#ifdef DEBUGGING |
1754 |
STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
1755 |
#else |
1756 |
STRLEN offset = SvCUR (string); |
1757 |
#endif |
1758 |
|
1759 |
if (offset > json->max_size && json->max_size) |
1760 |
croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1761 |
(unsigned long)SvCUR (string), (unsigned long)json->max_size); |
1762 |
} |
1763 |
|
1764 |
if (DECODE_WANTS_OCTETS (json)) |
1765 |
sv_utf8_downgrade (string, 0); |
1766 |
else |
1767 |
sv_utf8_upgrade (string); |
1768 |
|
1769 |
SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1770 |
|
1771 |
dec.json = *json; |
1772 |
dec.cur = SvPVX (string); |
1773 |
dec.end = SvEND (string); |
1774 |
dec.err = 0; |
1775 |
dec.depth = 0; |
1776 |
|
1777 |
if (dec.json.cb_object || dec.json.cb_sk_object) |
1778 |
dec.json.flags |= F_HOOK; |
1779 |
|
1780 |
*dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1781 |
|
1782 |
decode_ws (&dec); |
1783 |
sv = decode_sv (&dec); |
1784 |
|
1785 |
if (offset_return) |
1786 |
*offset_return = dec.cur - SvPVX (string); |
1787 |
else if (sv) |
1788 |
{ |
1789 |
// check for trailing garbage |
1790 |
decode_ws (&dec); |
1791 |
|
1792 |
if (*dec.cur) |
1793 |
{ |
1794 |
dec.err = "garbage after JSON object"; |
1795 |
SvREFCNT_dec (sv); |
1796 |
sv = 0; |
1797 |
} |
1798 |
} |
1799 |
|
1800 |
if (!sv) |
1801 |
{ |
1802 |
SV *uni = sv_newmortal (); |
1803 |
|
1804 |
// horrible hack to silence warning inside pv_uni_display |
1805 |
COP cop = *PL_curcop; |
1806 |
cop.cop_warnings = pWARN_NONE; |
1807 |
ENTER; |
1808 |
SAVEVPTR (PL_curcop); |
1809 |
PL_curcop = &cop; |
1810 |
pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1811 |
LEAVE; |
1812 |
|
1813 |
croak ("%s, at character offset %d (before \"%s\")", |
1814 |
dec.err, |
1815 |
(int)ptr_to_index (string, dec.cur), |
1816 |
dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1817 |
} |
1818 |
|
1819 |
sv = sv_2mortal (sv); |
1820 |
|
1821 |
if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv)) |
1822 |
croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1823 |
|
1824 |
return sv; |
1825 |
} |
1826 |
|
1827 |
///////////////////////////////////////////////////////////////////////////// |
1828 |
// incremental parser |
1829 |
|
1830 |
static void |
1831 |
incr_parse (JSON *self) |
1832 |
{ |
1833 |
const char *p = SvPVX (self->incr_text) + self->incr_pos; |
1834 |
|
1835 |
// the state machine here is a bit convoluted and could be simplified a lot |
1836 |
// but this would make it slower, so... |
1837 |
|
1838 |
for (;;) |
1839 |
{ |
1840 |
//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D |
1841 |
switch (self->incr_mode) |
1842 |
{ |
1843 |
// only used for initial whitespace skipping |
1844 |
case INCR_M_WS: |
1845 |
for (;;) |
1846 |
{ |
1847 |
if (*p > 0x20) |
1848 |
{ |
1849 |
if (*p == '#') |
1850 |
{ |
1851 |
self->incr_mode = INCR_M_C0; |
1852 |
goto incr_m_c; |
1853 |
} |
1854 |
else |
1855 |
{ |
1856 |
self->incr_mode = INCR_M_JSON; |
1857 |
goto incr_m_json; |
1858 |
} |
1859 |
} |
1860 |
else if (!*p) |
1861 |
goto interrupt; |
1862 |
|
1863 |
++p; |
1864 |
} |
1865 |
|
1866 |
// skip a single char inside a string (for \\-processing) |
1867 |
case INCR_M_BS: |
1868 |
if (!*p) |
1869 |
goto interrupt; |
1870 |
|
1871 |
++p; |
1872 |
self->incr_mode = INCR_M_STR; |
1873 |
goto incr_m_str; |
1874 |
|
1875 |
// inside #-style comments |
1876 |
case INCR_M_C0: |
1877 |
case INCR_M_C1: |
1878 |
incr_m_c: |
1879 |
for (;;) |
1880 |
{ |
1881 |
if (*p == '\n') |
1882 |
{ |
1883 |
self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
1884 |
break; |
1885 |
} |
1886 |
else if (!*p) |
1887 |
goto interrupt; |
1888 |
|
1889 |
++p; |
1890 |
} |
1891 |
|
1892 |
break; |
1893 |
|
1894 |
// inside a string |
1895 |
case INCR_M_STR: |
1896 |
incr_m_str: |
1897 |
for (;;) |
1898 |
{ |
1899 |
if (*p == '"') |
1900 |
{ |
1901 |
++p; |
1902 |
self->incr_mode = INCR_M_JSON; |
1903 |
|
1904 |
if (!self->incr_nest) |
1905 |
goto interrupt; |
1906 |
|
1907 |
goto incr_m_json; |
1908 |
} |
1909 |
else if (*p == '\\') |
1910 |
{ |
1911 |
++p; // "virtually" consumes character after \ |
1912 |
|
1913 |
if (!*p) // if at end of string we have to switch modes |
1914 |
{ |
1915 |
self->incr_mode = INCR_M_BS; |
1916 |
goto interrupt; |
1917 |
} |
1918 |
} |
1919 |
else if (!*p) |
1920 |
goto interrupt; |
1921 |
|
1922 |
++p; |
1923 |
} |
1924 |
|
1925 |
// after initial ws, outside string |
1926 |
case INCR_M_JSON: |
1927 |
incr_m_json: |
1928 |
for (;;) |
1929 |
{ |
1930 |
switch (*p++) |
1931 |
{ |
1932 |
case 0: |
1933 |
--p; |
1934 |
goto interrupt; |
1935 |
|
1936 |
case 0x09: |
1937 |
case 0x0a: |
1938 |
case 0x0d: |
1939 |
case 0x20: |
1940 |
if (!self->incr_nest) |
1941 |
{ |
1942 |
--p; // do not eat the whitespace, let the next round do it |
1943 |
goto interrupt; |
1944 |
} |
1945 |
break; |
1946 |
|
1947 |
case '"': |
1948 |
self->incr_mode = INCR_M_STR; |
1949 |
goto incr_m_str; |
1950 |
|
1951 |
case '[': |
1952 |
case '{': |
1953 |
case '(': |
1954 |
if (++self->incr_nest > self->max_depth) |
1955 |
croak (ERR_NESTING_EXCEEDED); |
1956 |
break; |
1957 |
|
1958 |
case ']': |
1959 |
case '}': |
1960 |
if (--self->incr_nest <= 0) |
1961 |
goto interrupt; |
1962 |
break; |
1963 |
|
1964 |
case ')': |
1965 |
--self->incr_nest; |
1966 |
break; |
1967 |
|
1968 |
case '#': |
1969 |
self->incr_mode = INCR_M_C1; |
1970 |
goto incr_m_c; |
1971 |
} |
1972 |
} |
1973 |
} |
1974 |
|
1975 |
modechange: |
1976 |
; |
1977 |
} |
1978 |
|
1979 |
interrupt: |
1980 |
self->incr_pos = p - SvPVX (self->incr_text); |
1981 |
//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
1982 |
//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1983 |
} |
1984 |
|
1985 |
///////////////////////////////////////////////////////////////////////////// |
1986 |
// XS interface functions |
1987 |
|
1988 |
MODULE = JSON::XS PACKAGE = JSON::XS |
1989 |
|
1990 |
BOOT: |
1991 |
{ |
1992 |
int i; |
1993 |
|
1994 |
for (i = 0; i < 256; ++i) |
1995 |
decode_hexdigit [i] = |
1996 |
i >= '0' && i <= '9' ? i - '0' |
1997 |
: i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1998 |
: i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1999 |
: -1; |
2000 |
|
2001 |
json_stash = gv_stashpv ("JSON::XS" , 1); |
2002 |
bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1); |
2003 |
bool_false = get_bool ("Types::Serialiser::false"); |
2004 |
bool_true = get_bool ("Types::Serialiser::true"); |
2005 |
|
2006 |
sv_json = newSVpv ("JSON", 0); |
2007 |
SvREADONLY_on (sv_json); |
2008 |
|
2009 |
CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
2010 |
} |
2011 |
|
2012 |
PROTOTYPES: DISABLE |
2013 |
|
2014 |
void CLONE (...) |
2015 |
CODE: |
2016 |
// as long as these writes are atomic, the race should not matter |
2017 |
// as existing threads either already use 0, or use the old value, |
2018 |
// which is sitll correct for the initial thread. |
2019 |
json_stash = 0; |
2020 |
bool_stash = 0; |
2021 |
bool_false = 0; |
2022 |
bool_true = 0; |
2023 |
|
2024 |
void new (char *klass) |
2025 |
PPCODE: |
2026 |
{ |
2027 |
SV *pv = NEWSV (0, sizeof (JSON)); |
2028 |
SvPOK_only (pv); |
2029 |
json_init ((JSON *)SvPVX (pv)); |
2030 |
XPUSHs (sv_2mortal (sv_bless ( |
2031 |
newRV_noinc (pv), |
2032 |
strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
2033 |
))); |
2034 |
} |
2035 |
|
2036 |
void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0) |
2037 |
PPCODE: |
2038 |
self->v_false = newSVsv (v_false); |
2039 |
self->v_true = newSVsv (v_true); |
2040 |
XPUSHs (ST (0)); |
2041 |
|
2042 |
void get_boolean_values (JSON *self) |
2043 |
PPCODE: |
2044 |
if (self->v_false && self->v_true) |
2045 |
{ |
2046 |
EXTEND (SP, 2); |
2047 |
PUSHs (self->v_false); |
2048 |
PUSHs (self->v_true); |
2049 |
} |
2050 |
|
2051 |
void ascii (JSON *self, int enable = 1) |
2052 |
ALIAS: |
2053 |
ascii = F_ASCII |
2054 |
latin1 = F_LATIN1 |
2055 |
utf8 = F_UTF8 |
2056 |
indent = F_INDENT |
2057 |
canonical = F_CANONICAL |
2058 |
space_before = F_SPACE_BEFORE |
2059 |
space_after = F_SPACE_AFTER |
2060 |
pretty = F_PRETTY |
2061 |
allow_nonref = F_ALLOW_NONREF |
2062 |
shrink = F_SHRINK |
2063 |
allow_blessed = F_ALLOW_BLESSED |
2064 |
convert_blessed = F_CONV_BLESSED |
2065 |
relaxed = F_RELAXED |
2066 |
allow_unknown = F_ALLOW_UNKNOWN |
2067 |
allow_tags = F_ALLOW_TAGS |
2068 |
PPCODE: |
2069 |
{ |
2070 |
if (enable) |
2071 |
self->flags |= ix; |
2072 |
else |
2073 |
self->flags &= ~ix; |
2074 |
|
2075 |
XPUSHs (ST (0)); |
2076 |
} |
2077 |
|
2078 |
void get_ascii (JSON *self) |
2079 |
ALIAS: |
2080 |
get_ascii = F_ASCII |
2081 |
get_latin1 = F_LATIN1 |
2082 |
get_utf8 = F_UTF8 |
2083 |
get_indent = F_INDENT |
2084 |
get_canonical = F_CANONICAL |
2085 |
get_space_before = F_SPACE_BEFORE |
2086 |
get_space_after = F_SPACE_AFTER |
2087 |
get_allow_nonref = F_ALLOW_NONREF |
2088 |
get_shrink = F_SHRINK |
2089 |
get_allow_blessed = F_ALLOW_BLESSED |
2090 |
get_convert_blessed = F_CONV_BLESSED |
2091 |
get_relaxed = F_RELAXED |
2092 |
get_allow_unknown = F_ALLOW_UNKNOWN |
2093 |
get_allow_tags = F_ALLOW_TAGS |
2094 |
PPCODE: |
2095 |
XPUSHs (boolSV (self->flags & ix)); |
2096 |
|
2097 |
void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
2098 |
PPCODE: |
2099 |
self->max_depth = max_depth; |
2100 |
XPUSHs (ST (0)); |
2101 |
|
2102 |
U32 get_max_depth (JSON *self) |
2103 |
CODE: |
2104 |
RETVAL = self->max_depth; |
2105 |
OUTPUT: |
2106 |
RETVAL |
2107 |
|
2108 |
void max_size (JSON *self, U32 max_size = 0) |
2109 |
PPCODE: |
2110 |
self->max_size = max_size; |
2111 |
XPUSHs (ST (0)); |
2112 |
|
2113 |
int get_max_size (JSON *self) |
2114 |
CODE: |
2115 |
RETVAL = self->max_size; |
2116 |
OUTPUT: |
2117 |
RETVAL |
2118 |
|
2119 |
void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
2120 |
PPCODE: |
2121 |
{ |
2122 |
SvREFCNT_dec (self->cb_object); |
2123 |
self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
2124 |
|
2125 |
XPUSHs (ST (0)); |
2126 |
} |
2127 |
|
2128 |
void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
2129 |
PPCODE: |
2130 |
{ |
2131 |
if (!self->cb_sk_object) |
2132 |
self->cb_sk_object = newHV (); |
2133 |
|
2134 |
if (SvOK (cb)) |
2135 |
hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
2136 |
else |
2137 |
{ |
2138 |
hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
2139 |
|
2140 |
if (!HvKEYS (self->cb_sk_object)) |
2141 |
{ |
2142 |
SvREFCNT_dec (self->cb_sk_object); |
2143 |
self->cb_sk_object = 0; |
2144 |
} |
2145 |
} |
2146 |
|
2147 |
XPUSHs (ST (0)); |
2148 |
} |
2149 |
|
2150 |
void encode (JSON *self, SV *scalar) |
2151 |
PPCODE: |
2152 |
PUTBACK; scalar = encode_json (scalar, self); SPAGAIN; |
2153 |
XPUSHs (scalar); |
2154 |
|
2155 |
void decode (JSON *self, SV *jsonstr) |
2156 |
PPCODE: |
2157 |
PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN; |
2158 |
XPUSHs (jsonstr); |
2159 |
|
2160 |
void decode_prefix (JSON *self, SV *jsonstr) |
2161 |
PPCODE: |
2162 |
{ |
2163 |
SV *sv; |
2164 |
STRLEN offset; |
2165 |
PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN; |
2166 |
EXTEND (SP, 2); |
2167 |
PUSHs (sv); |
2168 |
PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset)))); |
2169 |
} |
2170 |
|
2171 |
void incr_parse (JSON *self, SV *jsonstr = 0) |
2172 |
PPCODE: |
2173 |
{ |
2174 |
if (!self->incr_text) |
2175 |
self->incr_text = newSVpvn ("", 0); |
2176 |
|
2177 |
/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */ |
2178 |
if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text)) |
2179 |
if (DECODE_WANTS_OCTETS (self)) |
2180 |
{ |
2181 |
if (self->incr_pos) |
2182 |
self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text), |
2183 |
(U8 *)SvPVX (self->incr_text) + self->incr_pos); |
2184 |
|
2185 |
sv_utf8_downgrade (self->incr_text, 0); |
2186 |
} |
2187 |
else |
2188 |
{ |
2189 |
sv_utf8_upgrade (self->incr_text); |
2190 |
|
2191 |
if (self->incr_pos) |
2192 |
self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
2193 |
- (U8 *)SvPVX (self->incr_text); |
2194 |
} |
2195 |
|
2196 |
// append data, if any |
2197 |
if (jsonstr) |
2198 |
{ |
2199 |
/* make sure both strings have same encoding */ |
2200 |
if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text)) |
2201 |
if (SvUTF8 (jsonstr)) |
2202 |
sv_utf8_downgrade (jsonstr, 0); |
2203 |
else |
2204 |
sv_utf8_upgrade (jsonstr); |
2205 |
|
2206 |
/* and then just blindly append */ |
2207 |
{ |
2208 |
STRLEN len; |
2209 |
const char *str = SvPV (jsonstr, len); |
2210 |
STRLEN cur = SvCUR (self->incr_text); |
2211 |
|
2212 |
if (SvLEN (self->incr_text) - cur <= len) |
2213 |
json_sv_grow (self->incr_text, cur, len); |
2214 |
|
2215 |
Move (str, SvEND (self->incr_text), len, char); |
2216 |
SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
2217 |
*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
2218 |
} |
2219 |
} |
2220 |
|
2221 |
if (GIMME_V != G_VOID) |
2222 |
do |
2223 |
{ |
2224 |
SV *sv; |
2225 |
STRLEN offset; |
2226 |
|
2227 |
if (!INCR_DONE (self)) |
2228 |
{ |
2229 |
incr_parse (self); |
2230 |
|
2231 |
if (self->incr_pos > self->max_size && self->max_size) |
2232 |
croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
2233 |
(unsigned long)self->incr_pos, (unsigned long)self->max_size); |
2234 |
|
2235 |
if (!INCR_DONE (self)) |
2236 |
{ |
2237 |
// as an optimisation, do not accumulate white space in the incr buffer |
2238 |
if (self->incr_mode == INCR_M_WS && self->incr_pos) |
2239 |
{ |
2240 |
self->incr_pos = 0; |
2241 |
SvCUR_set (self->incr_text, 0); |
2242 |
} |
2243 |
|
2244 |
break; |
2245 |
} |
2246 |
} |
2247 |
|
2248 |
PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN; |
2249 |
XPUSHs (sv); |
2250 |
|
2251 |
self->incr_pos -= offset; |
2252 |
self->incr_nest = 0; |
2253 |
self->incr_mode = 0; |
2254 |
|
2255 |
sv_chop (self->incr_text, SvPVX (self->incr_text) + offset); |
2256 |
} |
2257 |
while (GIMME_V == G_ARRAY); |
2258 |
} |
2259 |
|
2260 |
SV *incr_text (JSON *self) |
2261 |
ATTRS: lvalue |
2262 |
CODE: |
2263 |
{ |
2264 |
if (self->incr_pos) |
2265 |
croak ("incr_text can not be called when the incremental parser already started parsing"); |
2266 |
|
2267 |
RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
2268 |
} |
2269 |
OUTPUT: |
2270 |
RETVAL |
2271 |
|
2272 |
void incr_skip (JSON *self) |
2273 |
CODE: |
2274 |
{ |
2275 |
if (self->incr_pos) |
2276 |
{ |
2277 |
sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
2278 |
self->incr_pos = 0; |
2279 |
self->incr_nest = 0; |
2280 |
self->incr_mode = 0; |
2281 |
} |
2282 |
} |
2283 |
|
2284 |
void incr_reset (JSON *self) |
2285 |
CODE: |
2286 |
{ |
2287 |
SvREFCNT_dec (self->incr_text); |
2288 |
self->incr_text = 0; |
2289 |
self->incr_pos = 0; |
2290 |
self->incr_nest = 0; |
2291 |
self->incr_mode = 0; |
2292 |
} |
2293 |
|
2294 |
void DESTROY (JSON *self) |
2295 |
CODE: |
2296 |
SvREFCNT_dec (self->v_false); |
2297 |
SvREFCNT_dec (self->v_true); |
2298 |
SvREFCNT_dec (self->cb_sk_object); |
2299 |
SvREFCNT_dec (self->cb_object); |
2300 |
SvREFCNT_dec (self->incr_text); |
2301 |
|
2302 |
PROTOTYPES: ENABLE |
2303 |
|
2304 |
void encode_json (SV *scalar) |
2305 |
PPCODE: |
2306 |
{ |
2307 |
JSON json; |
2308 |
json_init (&json); |
2309 |
json.flags |= F_UTF8; |
2310 |
PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN; |
2311 |
XPUSHs (scalar); |
2312 |
} |
2313 |
|
2314 |
void decode_json (SV *jsonstr) |
2315 |
PPCODE: |
2316 |
{ |
2317 |
JSON json; |
2318 |
json_init (&json); |
2319 |
json.flags |= F_UTF8; |
2320 |
PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN; |
2321 |
XPUSHs (jsonstr); |
2322 |
} |
2323 |
|