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 |
|
12 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
13 |
# define snprintf _snprintf // C compilers have this in stdio.h |
14 |
#endif |
15 |
|
16 |
// some old perls do not have this, try to make it work, no |
17 |
// guarantees, though. if it breaks, you get to keep the pieces. |
18 |
#ifndef UTF8_MAXBYTES |
19 |
# define UTF8_MAXBYTES 13 |
20 |
#endif |
21 |
|
22 |
// three extra for rounding, sign, and end of string |
23 |
#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3) |
24 |
|
25 |
#define F_ASCII 0x00000001UL |
26 |
#define F_LATIN1 0x00000002UL |
27 |
#define F_UTF8 0x00000004UL |
28 |
#define F_INDENT 0x00000008UL |
29 |
#define F_CANONICAL 0x00000010UL |
30 |
#define F_SPACE_BEFORE 0x00000020UL |
31 |
#define F_SPACE_AFTER 0x00000040UL |
32 |
#define F_ALLOW_NONREF 0x00000100UL |
33 |
#define F_SHRINK 0x00000200UL |
34 |
#define F_ALLOW_BLESSED 0x00000400UL |
35 |
#define F_CONV_BLESSED 0x00000800UL |
36 |
#define F_RELAXED 0x00001000UL |
37 |
#define F_ALLOW_UNKNOWN 0x00002000UL |
38 |
#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
39 |
|
40 |
#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
41 |
|
42 |
#define INIT_SIZE 32 // initial scalar size to be allocated |
43 |
#define INDENT_STEP 3 // spaces per indentation level |
44 |
|
45 |
#define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
46 |
|
47 |
#define SB do { |
48 |
#define SE } while (0) |
49 |
|
50 |
#if __GNUC__ >= 3 |
51 |
# define expect(expr,value) __builtin_expect ((expr), (value)) |
52 |
# define INLINE static inline |
53 |
#else |
54 |
# define expect(expr,value) (expr) |
55 |
# define INLINE static |
56 |
#endif |
57 |
|
58 |
#define expect_false(expr) expect ((expr) != 0, 0) |
59 |
#define expect_true(expr) expect ((expr) != 0, 1) |
60 |
|
61 |
#define IN_RANGE_INC(type,val,beg,end) \ |
62 |
((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
63 |
<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
64 |
|
65 |
#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
66 |
|
67 |
#ifdef USE_ITHREADS |
68 |
# define JSON_SLOW 1 |
69 |
# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
70 |
#else |
71 |
# define JSON_SLOW 0 |
72 |
# define JSON_STASH json_stash |
73 |
#endif |
74 |
|
75 |
static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
76 |
static SV *json_true, *json_false; |
77 |
|
78 |
enum { |
79 |
INCR_M_WS = 0, // initial whitespace skipping, must be 0 |
80 |
INCR_M_STR, // inside string |
81 |
INCR_M_BS, // inside backslash |
82 |
INCR_M_C0, // inside comment in initial whitespace sequence |
83 |
INCR_M_C1, // inside comment in other places |
84 |
INCR_M_JSON // outside anything, count nesting |
85 |
}; |
86 |
|
87 |
#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON) |
88 |
|
89 |
typedef struct { |
90 |
U32 flags; |
91 |
U32 max_depth; |
92 |
STRLEN max_size; |
93 |
|
94 |
SV *cb_object; |
95 |
HV *cb_sk_object; |
96 |
|
97 |
// for the incremental parser |
98 |
SV *incr_text; // the source text so far |
99 |
STRLEN incr_pos; // the current offset into the text |
100 |
int incr_nest; // {[]}-nesting level |
101 |
unsigned char incr_mode; |
102 |
} JSON; |
103 |
|
104 |
INLINE void |
105 |
json_init (JSON *json) |
106 |
{ |
107 |
Zero (json, 1, JSON); |
108 |
json->max_depth = 512; |
109 |
} |
110 |
|
111 |
///////////////////////////////////////////////////////////////////////////// |
112 |
// utility functions |
113 |
|
114 |
INLINE SV * |
115 |
get_bool (const char *name) |
116 |
{ |
117 |
SV *sv = get_sv (name, 1); |
118 |
|
119 |
SvREADONLY_on (sv); |
120 |
SvREADONLY_on (SvRV (sv)); |
121 |
|
122 |
return sv; |
123 |
} |
124 |
|
125 |
INLINE void |
126 |
shrink (SV *sv) |
127 |
{ |
128 |
sv_utf8_downgrade (sv, 1); |
129 |
|
130 |
if (SvLEN (sv) > SvCUR (sv) + 1) |
131 |
{ |
132 |
#ifdef SvPV_shrink_to_cur |
133 |
SvPV_shrink_to_cur (sv); |
134 |
#elif defined (SvPV_renew) |
135 |
SvPV_renew (sv, SvCUR (sv) + 1); |
136 |
#endif |
137 |
} |
138 |
} |
139 |
|
140 |
// decode an utf-8 character and return it, or (UV)-1 in |
141 |
// case of an error. |
142 |
// we special-case "safe" characters from U+80 .. U+7FF, |
143 |
// but use the very good perl function to parse anything else. |
144 |
// note that we never call this function for a ascii codepoints |
145 |
INLINE UV |
146 |
decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
147 |
{ |
148 |
if (expect_true (len >= 2 |
149 |
&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf) |
150 |
&& IN_RANGE_INC (char, s[1], 0x80, 0xbf))) |
151 |
{ |
152 |
*clen = 2; |
153 |
return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
154 |
} |
155 |
else |
156 |
return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
157 |
} |
158 |
|
159 |
// likewise for encoding, also never called for ascii codepoints |
160 |
// this function takes advantage of this fact, although current gccs |
161 |
// seem to optimise the check for >= 0x80 away anyways |
162 |
INLINE unsigned char * |
163 |
encode_utf8 (unsigned char *s, UV ch) |
164 |
{ |
165 |
if (expect_false (ch < 0x000080)) |
166 |
*s++ = ch; |
167 |
else if (expect_true (ch < 0x000800)) |
168 |
*s++ = 0xc0 | ( ch >> 6), |
169 |
*s++ = 0x80 | ( ch & 0x3f); |
170 |
else if ( ch < 0x010000) |
171 |
*s++ = 0xe0 | ( ch >> 12), |
172 |
*s++ = 0x80 | ((ch >> 6) & 0x3f), |
173 |
*s++ = 0x80 | ( ch & 0x3f); |
174 |
else if ( ch < 0x110000) |
175 |
*s++ = 0xf0 | ( ch >> 18), |
176 |
*s++ = 0x80 | ((ch >> 12) & 0x3f), |
177 |
*s++ = 0x80 | ((ch >> 6) & 0x3f), |
178 |
*s++ = 0x80 | ( ch & 0x3f); |
179 |
|
180 |
return s; |
181 |
} |
182 |
|
183 |
// convert offset pointer to character index, sv must be string |
184 |
static STRLEN |
185 |
ptr_to_index (SV *sv, char *offset) |
186 |
{ |
187 |
return SvUTF8 (sv) |
188 |
? utf8_distance (offset, SvPVX (sv)) |
189 |
: offset - SvPVX (sv); |
190 |
} |
191 |
|
192 |
///////////////////////////////////////////////////////////////////////////// |
193 |
// fp hell |
194 |
|
195 |
// scan a group of digits, and a trailing exponent |
196 |
static void |
197 |
json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp) |
198 |
{ |
199 |
UV uaccum = 0; |
200 |
int eaccum = 0; |
201 |
|
202 |
for (;;) |
203 |
{ |
204 |
U8 dig = (U8)*s - '0'; |
205 |
|
206 |
if (expect_false (dig >= 10)) |
207 |
{ |
208 |
if (dig == (U8)((U8)'.' - (U8)'0')) |
209 |
{ |
210 |
++s; |
211 |
json_atof_scan1 (s, accum, expo, 1); |
212 |
} |
213 |
else if ((dig | ' ') == 'e' - '0') |
214 |
{ |
215 |
int exp2 = 0; |
216 |
int neg = 0; |
217 |
|
218 |
++s; |
219 |
|
220 |
if (*s == '-') |
221 |
{ |
222 |
++s; |
223 |
neg = 1; |
224 |
} |
225 |
else if (*s == '+') |
226 |
++s; |
227 |
|
228 |
while ((dig = (U8)*s - '0') < 10) |
229 |
exp2 = exp2 * 10 + *s++ - '0'; |
230 |
|
231 |
*expo += neg ? -exp2 : exp2; |
232 |
} |
233 |
|
234 |
break; |
235 |
} |
236 |
|
237 |
++s; |
238 |
|
239 |
uaccum = uaccum * 10 + dig; |
240 |
++eaccum; |
241 |
|
242 |
// if we have too many digits, then recurse for more |
243 |
// we actually do this for rather few digits |
244 |
if (uaccum >= (UV_MAX - 9) / 10) |
245 |
{ |
246 |
if (postdp) *expo -= eaccum; |
247 |
json_atof_scan1 (s, accum, expo, postdp); |
248 |
if (postdp) *expo += eaccum; |
249 |
|
250 |
break; |
251 |
} |
252 |
} |
253 |
|
254 |
// this relies greatly on the quality of the pow () |
255 |
// implementation of the platform, but a good |
256 |
// implementation is hard to beat. |
257 |
if (postdp) *expo -= eaccum; |
258 |
*accum += uaccum * Perl_pow (10., *expo); |
259 |
*expo += eaccum; |
260 |
} |
261 |
|
262 |
static NV |
263 |
json_atof (const char *s) |
264 |
{ |
265 |
NV accum = 0.; |
266 |
int expo = 0; |
267 |
int neg = 0; |
268 |
|
269 |
if (*s == '-') |
270 |
{ |
271 |
++s; |
272 |
neg = 1; |
273 |
} |
274 |
|
275 |
json_atof_scan1 (s, &accum, &expo, 0); |
276 |
|
277 |
return neg ? -accum : accum; |
278 |
} |
279 |
///////////////////////////////////////////////////////////////////////////// |
280 |
// encoder |
281 |
|
282 |
// structure used for encoding JSON |
283 |
typedef struct |
284 |
{ |
285 |
char *cur; // SvPVX (sv) + current output position |
286 |
char *end; // SvEND (sv) |
287 |
SV *sv; // result scalar |
288 |
JSON json; |
289 |
U32 indent; // indentation level |
290 |
UV limit; // escape character values >= this value when encoding |
291 |
} enc_t; |
292 |
|
293 |
INLINE void |
294 |
need (enc_t *enc, STRLEN len) |
295 |
{ |
296 |
if (expect_false (enc->cur + len >= enc->end)) |
297 |
{ |
298 |
STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
299 |
SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
300 |
enc->cur = SvPVX (enc->sv) + cur; |
301 |
enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
302 |
} |
303 |
} |
304 |
|
305 |
INLINE void |
306 |
encode_ch (enc_t *enc, char ch) |
307 |
{ |
308 |
need (enc, 1); |
309 |
*enc->cur++ = ch; |
310 |
} |
311 |
|
312 |
static void |
313 |
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
314 |
{ |
315 |
char *end = str + len; |
316 |
|
317 |
need (enc, len); |
318 |
|
319 |
while (str < end) |
320 |
{ |
321 |
unsigned char ch = *(unsigned char *)str; |
322 |
|
323 |
if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
324 |
{ |
325 |
if (expect_false (ch == '"')) // but with slow exceptions |
326 |
{ |
327 |
need (enc, len += 1); |
328 |
*enc->cur++ = '\\'; |
329 |
*enc->cur++ = '"'; |
330 |
} |
331 |
else if (expect_false (ch == '\\')) |
332 |
{ |
333 |
need (enc, len += 1); |
334 |
*enc->cur++ = '\\'; |
335 |
*enc->cur++ = '\\'; |
336 |
} |
337 |
else |
338 |
*enc->cur++ = ch; |
339 |
|
340 |
++str; |
341 |
} |
342 |
else |
343 |
{ |
344 |
switch (ch) |
345 |
{ |
346 |
case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
347 |
case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
348 |
case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
349 |
case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
350 |
case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
351 |
|
352 |
default: |
353 |
{ |
354 |
STRLEN clen; |
355 |
UV uch; |
356 |
|
357 |
if (is_utf8) |
358 |
{ |
359 |
uch = decode_utf8 (str, end - str, &clen); |
360 |
if (clen == (STRLEN)-1) |
361 |
croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
362 |
} |
363 |
else |
364 |
{ |
365 |
uch = ch; |
366 |
clen = 1; |
367 |
} |
368 |
|
369 |
if (uch < 0x80/*0x20*/ || uch >= enc->limit) |
370 |
{ |
371 |
if (uch >= 0x10000UL) |
372 |
{ |
373 |
if (uch >= 0x110000UL) |
374 |
croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
375 |
|
376 |
need (enc, len += 11); |
377 |
sprintf (enc->cur, "\\u%04x\\u%04x", |
378 |
(int)((uch - 0x10000) / 0x400 + 0xD800), |
379 |
(int)((uch - 0x10000) % 0x400 + 0xDC00)); |
380 |
enc->cur += 12; |
381 |
} |
382 |
else |
383 |
{ |
384 |
need (enc, len += 5); |
385 |
*enc->cur++ = '\\'; |
386 |
*enc->cur++ = 'u'; |
387 |
*enc->cur++ = PL_hexdigit [ uch >> 12 ]; |
388 |
*enc->cur++ = PL_hexdigit [(uch >> 8) & 15]; |
389 |
*enc->cur++ = PL_hexdigit [(uch >> 4) & 15]; |
390 |
*enc->cur++ = PL_hexdigit [(uch >> 0) & 15]; |
391 |
} |
392 |
|
393 |
str += clen; |
394 |
} |
395 |
else if (enc->json.flags & F_LATIN1) |
396 |
{ |
397 |
*enc->cur++ = uch; |
398 |
str += clen; |
399 |
} |
400 |
else if (is_utf8) |
401 |
{ |
402 |
need (enc, len += clen); |
403 |
do |
404 |
{ |
405 |
*enc->cur++ = *str++; |
406 |
} |
407 |
while (--clen); |
408 |
} |
409 |
else |
410 |
{ |
411 |
need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
412 |
enc->cur = encode_utf8 (enc->cur, uch); |
413 |
++str; |
414 |
} |
415 |
} |
416 |
} |
417 |
} |
418 |
|
419 |
--len; |
420 |
} |
421 |
} |
422 |
|
423 |
INLINE void |
424 |
encode_indent (enc_t *enc) |
425 |
{ |
426 |
if (enc->json.flags & F_INDENT) |
427 |
{ |
428 |
int spaces = enc->indent * INDENT_STEP; |
429 |
|
430 |
need (enc, spaces); |
431 |
memset (enc->cur, ' ', spaces); |
432 |
enc->cur += spaces; |
433 |
} |
434 |
} |
435 |
|
436 |
INLINE void |
437 |
encode_space (enc_t *enc) |
438 |
{ |
439 |
need (enc, 1); |
440 |
encode_ch (enc, ' '); |
441 |
} |
442 |
|
443 |
INLINE void |
444 |
encode_nl (enc_t *enc) |
445 |
{ |
446 |
if (enc->json.flags & F_INDENT) |
447 |
{ |
448 |
need (enc, 1); |
449 |
encode_ch (enc, '\n'); |
450 |
} |
451 |
} |
452 |
|
453 |
INLINE void |
454 |
encode_comma (enc_t *enc) |
455 |
{ |
456 |
encode_ch (enc, ','); |
457 |
|
458 |
if (enc->json.flags & F_INDENT) |
459 |
encode_nl (enc); |
460 |
else if (enc->json.flags & F_SPACE_AFTER) |
461 |
encode_space (enc); |
462 |
} |
463 |
|
464 |
static void encode_sv (enc_t *enc, SV *sv); |
465 |
|
466 |
static void |
467 |
encode_av (enc_t *enc, AV *av) |
468 |
{ |
469 |
int i, len = av_len (av); |
470 |
|
471 |
if (enc->indent >= enc->json.max_depth) |
472 |
croak (ERR_NESTING_EXCEEDED); |
473 |
|
474 |
encode_ch (enc, '['); |
475 |
|
476 |
if (len >= 0) |
477 |
{ |
478 |
encode_nl (enc); ++enc->indent; |
479 |
|
480 |
for (i = 0; i <= len; ++i) |
481 |
{ |
482 |
SV **svp = av_fetch (av, i, 0); |
483 |
|
484 |
encode_indent (enc); |
485 |
|
486 |
if (svp) |
487 |
encode_sv (enc, *svp); |
488 |
else |
489 |
encode_str (enc, "null", 4, 0); |
490 |
|
491 |
if (i < len) |
492 |
encode_comma (enc); |
493 |
} |
494 |
|
495 |
encode_nl (enc); --enc->indent; encode_indent (enc); |
496 |
} |
497 |
|
498 |
encode_ch (enc, ']'); |
499 |
} |
500 |
|
501 |
static void |
502 |
encode_hk (enc_t *enc, HE *he) |
503 |
{ |
504 |
encode_ch (enc, '"'); |
505 |
|
506 |
if (HeKLEN (he) == HEf_SVKEY) |
507 |
{ |
508 |
SV *sv = HeSVKEY (he); |
509 |
STRLEN len; |
510 |
char *str; |
511 |
|
512 |
SvGETMAGIC (sv); |
513 |
str = SvPV (sv, len); |
514 |
|
515 |
encode_str (enc, str, len, SvUTF8 (sv)); |
516 |
} |
517 |
else |
518 |
encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
519 |
|
520 |
encode_ch (enc, '"'); |
521 |
|
522 |
if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
523 |
encode_ch (enc, ':'); |
524 |
if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
525 |
} |
526 |
|
527 |
// compare hash entries, used when all keys are bytestrings |
528 |
static int |
529 |
he_cmp_fast (const void *a_, const void *b_) |
530 |
{ |
531 |
int cmp; |
532 |
|
533 |
HE *a = *(HE **)a_; |
534 |
HE *b = *(HE **)b_; |
535 |
|
536 |
STRLEN la = HeKLEN (a); |
537 |
STRLEN lb = HeKLEN (b); |
538 |
|
539 |
if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
540 |
cmp = lb - la; |
541 |
|
542 |
return cmp; |
543 |
} |
544 |
|
545 |
// compare hash entries, used when some keys are sv's or utf-x |
546 |
static int |
547 |
he_cmp_slow (const void *a, const void *b) |
548 |
{ |
549 |
return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
550 |
} |
551 |
|
552 |
static void |
553 |
encode_hv (enc_t *enc, HV *hv) |
554 |
{ |
555 |
HE *he; |
556 |
|
557 |
if (enc->indent >= enc->json.max_depth) |
558 |
croak (ERR_NESTING_EXCEEDED); |
559 |
|
560 |
encode_ch (enc, '{'); |
561 |
|
562 |
// for canonical output we have to sort by keys first |
563 |
// actually, this is mostly due to the stupid so-called |
564 |
// security workaround added somewhere in 5.8.x |
565 |
// that randomises hash orderings |
566 |
if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv)) |
567 |
{ |
568 |
int count = hv_iterinit (hv); |
569 |
|
570 |
if (SvMAGICAL (hv)) |
571 |
{ |
572 |
// need to count by iterating. could improve by dynamically building the vector below |
573 |
// but I don't care for the speed of this special case. |
574 |
// note also that we will run into undefined behaviour when the two iterations |
575 |
// do not result in the same count, something I might care for in some later release. |
576 |
|
577 |
count = 0; |
578 |
while (hv_iternext (hv)) |
579 |
++count; |
580 |
|
581 |
hv_iterinit (hv); |
582 |
} |
583 |
|
584 |
if (count) |
585 |
{ |
586 |
int i, fast = 1; |
587 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
588 |
HE **hes = _alloca (count * sizeof (HE)); |
589 |
#else |
590 |
HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
591 |
#endif |
592 |
|
593 |
i = 0; |
594 |
while ((he = hv_iternext (hv))) |
595 |
{ |
596 |
hes [i++] = he; |
597 |
if (HeKLEN (he) < 0 || HeKUTF8 (he)) |
598 |
fast = 0; |
599 |
} |
600 |
|
601 |
assert (i == count); |
602 |
|
603 |
if (fast) |
604 |
qsort (hes, count, sizeof (HE *), he_cmp_fast); |
605 |
else |
606 |
{ |
607 |
// hack to forcefully disable "use bytes" |
608 |
COP cop = *PL_curcop; |
609 |
cop.op_private = 0; |
610 |
|
611 |
ENTER; |
612 |
SAVETMPS; |
613 |
|
614 |
SAVEVPTR (PL_curcop); |
615 |
PL_curcop = &cop; |
616 |
|
617 |
qsort (hes, count, sizeof (HE *), he_cmp_slow); |
618 |
|
619 |
FREETMPS; |
620 |
LEAVE; |
621 |
} |
622 |
|
623 |
encode_nl (enc); ++enc->indent; |
624 |
|
625 |
while (count--) |
626 |
{ |
627 |
encode_indent (enc); |
628 |
he = hes [count]; |
629 |
encode_hk (enc, he); |
630 |
encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
631 |
|
632 |
if (count) |
633 |
encode_comma (enc); |
634 |
} |
635 |
|
636 |
encode_nl (enc); --enc->indent; encode_indent (enc); |
637 |
} |
638 |
} |
639 |
else |
640 |
{ |
641 |
if (hv_iterinit (hv) || SvMAGICAL (hv)) |
642 |
if ((he = hv_iternext (hv))) |
643 |
{ |
644 |
encode_nl (enc); ++enc->indent; |
645 |
|
646 |
for (;;) |
647 |
{ |
648 |
encode_indent (enc); |
649 |
encode_hk (enc, he); |
650 |
encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
651 |
|
652 |
if (!(he = hv_iternext (hv))) |
653 |
break; |
654 |
|
655 |
encode_comma (enc); |
656 |
} |
657 |
|
658 |
encode_nl (enc); --enc->indent; encode_indent (enc); |
659 |
} |
660 |
} |
661 |
|
662 |
encode_ch (enc, '}'); |
663 |
} |
664 |
|
665 |
// encode objects, arrays and special \0=false and \1=true values. |
666 |
static void |
667 |
encode_rv (enc_t *enc, SV *sv) |
668 |
{ |
669 |
svtype svt; |
670 |
|
671 |
SvGETMAGIC (sv); |
672 |
svt = SvTYPE (sv); |
673 |
|
674 |
if (expect_false (SvOBJECT (sv))) |
675 |
{ |
676 |
HV *stash = !JSON_SLOW || json_boolean_stash |
677 |
? json_boolean_stash |
678 |
: gv_stashpv ("JSON::XS::Boolean", 1); |
679 |
|
680 |
if (SvSTASH (sv) == stash) |
681 |
{ |
682 |
if (SvIV (sv)) |
683 |
encode_str (enc, "true", 4, 0); |
684 |
else |
685 |
encode_str (enc, "false", 5, 0); |
686 |
} |
687 |
else |
688 |
{ |
689 |
#if 0 |
690 |
if (0 && sv_derived_from (rv, "JSON::Literal")) |
691 |
{ |
692 |
// not yet |
693 |
} |
694 |
#endif |
695 |
if (enc->json.flags & F_CONV_BLESSED) |
696 |
{ |
697 |
// we re-bless the reference to get overload and other niceties right |
698 |
GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
699 |
|
700 |
if (to_json) |
701 |
{ |
702 |
dSP; |
703 |
|
704 |
ENTER; SAVETMPS; PUSHMARK (SP); |
705 |
XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
706 |
|
707 |
// calling with G_SCALAR ensures that we always get a 1 return value |
708 |
PUTBACK; |
709 |
call_sv ((SV *)GvCV (to_json), G_SCALAR); |
710 |
SPAGAIN; |
711 |
|
712 |
// catch this surprisingly common error |
713 |
if (SvROK (TOPs) && SvRV (TOPs) == sv) |
714 |
croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
715 |
|
716 |
sv = POPs; |
717 |
PUTBACK; |
718 |
|
719 |
encode_sv (enc, sv); |
720 |
|
721 |
FREETMPS; LEAVE; |
722 |
} |
723 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
724 |
encode_str (enc, "null", 4, 0); |
725 |
else |
726 |
croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
727 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
728 |
} |
729 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
730 |
encode_str (enc, "null", 4, 0); |
731 |
else |
732 |
croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
733 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
734 |
} |
735 |
} |
736 |
else if (svt == SVt_PVHV) |
737 |
encode_hv (enc, (HV *)sv); |
738 |
else if (svt == SVt_PVAV) |
739 |
encode_av (enc, (AV *)sv); |
740 |
else if (svt < SVt_PVAV) |
741 |
{ |
742 |
STRLEN len = 0; |
743 |
char *pv = svt ? SvPV (sv, len) : 0; |
744 |
|
745 |
if (len == 1 && *pv == '1') |
746 |
encode_str (enc, "true", 4, 0); |
747 |
else if (len == 1 && *pv == '0') |
748 |
encode_str (enc, "false", 5, 0); |
749 |
else if (enc->json.flags & F_ALLOW_UNKNOWN) |
750 |
encode_str (enc, "null", 4, 0); |
751 |
else |
752 |
croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
753 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
754 |
} |
755 |
else if (enc->json.flags & F_ALLOW_UNKNOWN) |
756 |
encode_str (enc, "null", 4, 0); |
757 |
else |
758 |
croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
759 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
760 |
} |
761 |
|
762 |
static void |
763 |
encode_sv (enc_t *enc, SV *sv) |
764 |
{ |
765 |
SvGETMAGIC (sv); |
766 |
|
767 |
if (SvPOKp (sv)) |
768 |
{ |
769 |
STRLEN len; |
770 |
char *str = SvPV (sv, len); |
771 |
encode_ch (enc, '"'); |
772 |
encode_str (enc, str, len, SvUTF8 (sv)); |
773 |
encode_ch (enc, '"'); |
774 |
} |
775 |
else if (SvNOKp (sv)) |
776 |
{ |
777 |
// trust that perl will do the right thing w.r.t. JSON syntax. |
778 |
need (enc, NV_DIG + 32); |
779 |
Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
780 |
enc->cur += strlen (enc->cur); |
781 |
} |
782 |
else if (SvIOKp (sv)) |
783 |
{ |
784 |
// we assume we can always read an IV as a UV and vice versa |
785 |
// we assume two's complement |
786 |
// we assume no aliasing issues in the union |
787 |
if (SvIsUV (sv) ? SvUVX (sv) <= 59000 |
788 |
: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000) |
789 |
{ |
790 |
// optimise the "small number case" |
791 |
// code will likely be branchless and use only a single multiplication |
792 |
// works for numbers up to 59074 |
793 |
I32 i = SvIVX (sv); |
794 |
U32 u; |
795 |
char digit, nz = 0; |
796 |
|
797 |
need (enc, 6); |
798 |
|
799 |
*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
800 |
u = i < 0 ? -i : i; |
801 |
|
802 |
// convert to 4.28 fixed-point representation |
803 |
u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
804 |
|
805 |
// now output digit by digit, each time masking out the integer part |
806 |
// and multiplying by 5 while moving the decimal point one to the right, |
807 |
// resulting in a net multiplication by 10. |
808 |
// we always write the digit to memory but conditionally increment |
809 |
// the pointer, to enable the use of conditional move instructions. |
810 |
digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5; |
811 |
digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5; |
812 |
digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5; |
813 |
digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5; |
814 |
digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
815 |
} |
816 |
else |
817 |
{ |
818 |
// large integer, use the (rather slow) snprintf way. |
819 |
need (enc, IVUV_MAXCHARS); |
820 |
enc->cur += |
821 |
SvIsUV(sv) |
822 |
? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) |
823 |
: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); |
824 |
} |
825 |
} |
826 |
else if (SvROK (sv)) |
827 |
encode_rv (enc, SvRV (sv)); |
828 |
else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) |
829 |
encode_str (enc, "null", 4, 0); |
830 |
else |
831 |
croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
832 |
SvPV_nolen (sv), SvFLAGS (sv)); |
833 |
} |
834 |
|
835 |
static SV * |
836 |
encode_json (SV *scalar, JSON *json) |
837 |
{ |
838 |
enc_t enc; |
839 |
|
840 |
if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
841 |
croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
842 |
|
843 |
enc.json = *json; |
844 |
enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
845 |
enc.cur = SvPVX (enc.sv); |
846 |
enc.end = SvEND (enc.sv); |
847 |
enc.indent = 0; |
848 |
enc.limit = enc.json.flags & F_ASCII ? 0x000080UL |
849 |
: enc.json.flags & F_LATIN1 ? 0x000100UL |
850 |
: 0x110000UL; |
851 |
|
852 |
SvPOK_only (enc.sv); |
853 |
encode_sv (&enc, scalar); |
854 |
encode_nl (&enc); |
855 |
|
856 |
SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
857 |
*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
858 |
|
859 |
if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
860 |
SvUTF8_on (enc.sv); |
861 |
|
862 |
if (enc.json.flags & F_SHRINK) |
863 |
shrink (enc.sv); |
864 |
|
865 |
return enc.sv; |
866 |
} |
867 |
|
868 |
///////////////////////////////////////////////////////////////////////////// |
869 |
// decoder |
870 |
|
871 |
// structure used for decoding JSON |
872 |
typedef struct |
873 |
{ |
874 |
char *cur; // current parser pointer |
875 |
char *end; // end of input string |
876 |
const char *err; // parse error, if != 0 |
877 |
JSON json; |
878 |
U32 depth; // recursion depth |
879 |
U32 maxdepth; // recursion depth limit |
880 |
} dec_t; |
881 |
|
882 |
INLINE void |
883 |
decode_comment (dec_t *dec) |
884 |
{ |
885 |
// only '#'-style comments allowed a.t.m. |
886 |
|
887 |
while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
888 |
++dec->cur; |
889 |
} |
890 |
|
891 |
INLINE void |
892 |
decode_ws (dec_t *dec) |
893 |
{ |
894 |
for (;;) |
895 |
{ |
896 |
char ch = *dec->cur; |
897 |
|
898 |
if (ch > 0x20) |
899 |
{ |
900 |
if (expect_false (ch == '#')) |
901 |
{ |
902 |
if (dec->json.flags & F_RELAXED) |
903 |
decode_comment (dec); |
904 |
else |
905 |
break; |
906 |
} |
907 |
else |
908 |
break; |
909 |
} |
910 |
else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
911 |
break; // parse error, but let higher level handle it, gives better error messages |
912 |
|
913 |
++dec->cur; |
914 |
} |
915 |
} |
916 |
|
917 |
#define ERR(reason) SB dec->err = reason; goto fail; SE |
918 |
|
919 |
#define EXPECT_CH(ch) SB \ |
920 |
if (*dec->cur != ch) \ |
921 |
ERR (# ch " expected"); \ |
922 |
++dec->cur; \ |
923 |
SE |
924 |
|
925 |
#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED) |
926 |
#define DEC_DEC_DEPTH --dec->depth |
927 |
|
928 |
static SV *decode_sv (dec_t *dec); |
929 |
|
930 |
static signed char decode_hexdigit[256]; |
931 |
|
932 |
static UV |
933 |
decode_4hex (dec_t *dec) |
934 |
{ |
935 |
signed char d1, d2, d3, d4; |
936 |
unsigned char *cur = (unsigned char *)dec->cur; |
937 |
|
938 |
d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
939 |
d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
940 |
d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
941 |
d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
942 |
|
943 |
dec->cur += 4; |
944 |
|
945 |
return ((UV)d1) << 12 |
946 |
| ((UV)d2) << 8 |
947 |
| ((UV)d3) << 4 |
948 |
| ((UV)d4); |
949 |
|
950 |
fail: |
951 |
return (UV)-1; |
952 |
} |
953 |
|
954 |
static SV * |
955 |
decode_str (dec_t *dec) |
956 |
{ |
957 |
SV *sv = 0; |
958 |
int utf8 = 0; |
959 |
char *dec_cur = dec->cur; |
960 |
|
961 |
do |
962 |
{ |
963 |
char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
964 |
char *cur = buf; |
965 |
|
966 |
do |
967 |
{ |
968 |
unsigned char ch = *(unsigned char *)dec_cur++; |
969 |
|
970 |
if (expect_false (ch == '"')) |
971 |
{ |
972 |
--dec_cur; |
973 |
break; |
974 |
} |
975 |
else if (expect_false (ch == '\\')) |
976 |
{ |
977 |
switch (*dec_cur) |
978 |
{ |
979 |
case '\\': |
980 |
case '/': |
981 |
case '"': *cur++ = *dec_cur++; break; |
982 |
|
983 |
case 'b': ++dec_cur; *cur++ = '\010'; break; |
984 |
case 't': ++dec_cur; *cur++ = '\011'; break; |
985 |
case 'n': ++dec_cur; *cur++ = '\012'; break; |
986 |
case 'f': ++dec_cur; *cur++ = '\014'; break; |
987 |
case 'r': ++dec_cur; *cur++ = '\015'; break; |
988 |
|
989 |
case 'u': |
990 |
{ |
991 |
UV lo, hi; |
992 |
++dec_cur; |
993 |
|
994 |
dec->cur = dec_cur; |
995 |
hi = decode_4hex (dec); |
996 |
dec_cur = dec->cur; |
997 |
if (hi == (UV)-1) |
998 |
goto fail; |
999 |
|
1000 |
// possibly a surrogate pair |
1001 |
if (hi >= 0xd800) |
1002 |
if (hi < 0xdc00) |
1003 |
{ |
1004 |
if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
1005 |
ERR ("missing low surrogate character in surrogate pair"); |
1006 |
|
1007 |
dec_cur += 2; |
1008 |
|
1009 |
dec->cur = dec_cur; |
1010 |
lo = decode_4hex (dec); |
1011 |
dec_cur = dec->cur; |
1012 |
if (lo == (UV)-1) |
1013 |
goto fail; |
1014 |
|
1015 |
if (lo < 0xdc00 || lo >= 0xe000) |
1016 |
ERR ("surrogate pair expected"); |
1017 |
|
1018 |
hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
1019 |
} |
1020 |
else if (hi < 0xe000) |
1021 |
ERR ("missing high surrogate character in surrogate pair"); |
1022 |
|
1023 |
if (hi >= 0x80) |
1024 |
{ |
1025 |
utf8 = 1; |
1026 |
|
1027 |
cur = encode_utf8 (cur, hi); |
1028 |
} |
1029 |
else |
1030 |
*cur++ = hi; |
1031 |
} |
1032 |
break; |
1033 |
|
1034 |
default: |
1035 |
--dec_cur; |
1036 |
ERR ("illegal backslash escape sequence in string"); |
1037 |
} |
1038 |
} |
1039 |
else if (expect_true (ch >= 0x20 && ch < 0x80)) |
1040 |
*cur++ = ch; |
1041 |
else if (ch >= 0x80) |
1042 |
{ |
1043 |
STRLEN clen; |
1044 |
|
1045 |
--dec_cur; |
1046 |
|
1047 |
decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
1048 |
if (clen == (STRLEN)-1) |
1049 |
ERR ("malformed UTF-8 character in JSON string"); |
1050 |
|
1051 |
do |
1052 |
*cur++ = *dec_cur++; |
1053 |
while (--clen); |
1054 |
|
1055 |
utf8 = 1; |
1056 |
} |
1057 |
else |
1058 |
{ |
1059 |
--dec_cur; |
1060 |
|
1061 |
if (!ch) |
1062 |
ERR ("unexpected end of string while parsing JSON string"); |
1063 |
else |
1064 |
ERR ("invalid character encountered while parsing JSON string"); |
1065 |
} |
1066 |
} |
1067 |
while (cur < buf + SHORT_STRING_LEN); |
1068 |
|
1069 |
{ |
1070 |
STRLEN len = cur - buf; |
1071 |
|
1072 |
if (sv) |
1073 |
{ |
1074 |
STRLEN cur = SvCUR (sv); |
1075 |
|
1076 |
if (SvLEN (sv) <= cur + len) |
1077 |
SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
1078 |
|
1079 |
memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
1080 |
SvCUR_set (sv, SvCUR (sv) + len); |
1081 |
} |
1082 |
else |
1083 |
sv = newSVpvn (buf, len); |
1084 |
} |
1085 |
} |
1086 |
while (*dec_cur != '"'); |
1087 |
|
1088 |
++dec_cur; |
1089 |
|
1090 |
if (sv) |
1091 |
{ |
1092 |
SvPOK_only (sv); |
1093 |
*SvEND (sv) = 0; |
1094 |
|
1095 |
if (utf8) |
1096 |
SvUTF8_on (sv); |
1097 |
} |
1098 |
else |
1099 |
sv = newSVpvn ("", 0); |
1100 |
|
1101 |
dec->cur = dec_cur; |
1102 |
return sv; |
1103 |
|
1104 |
fail: |
1105 |
dec->cur = dec_cur; |
1106 |
return 0; |
1107 |
} |
1108 |
|
1109 |
static SV * |
1110 |
decode_num (dec_t *dec) |
1111 |
{ |
1112 |
int is_nv = 0; |
1113 |
char *start = dec->cur; |
1114 |
|
1115 |
// [minus] |
1116 |
if (*dec->cur == '-') |
1117 |
++dec->cur; |
1118 |
|
1119 |
if (*dec->cur == '0') |
1120 |
{ |
1121 |
++dec->cur; |
1122 |
if (*dec->cur >= '0' && *dec->cur <= '9') |
1123 |
ERR ("malformed number (leading zero must not be followed by another digit)"); |
1124 |
} |
1125 |
else if (*dec->cur < '0' || *dec->cur > '9') |
1126 |
ERR ("malformed number (no digits after initial minus)"); |
1127 |
else |
1128 |
do |
1129 |
{ |
1130 |
++dec->cur; |
1131 |
} |
1132 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
1133 |
|
1134 |
// [frac] |
1135 |
if (*dec->cur == '.') |
1136 |
{ |
1137 |
++dec->cur; |
1138 |
|
1139 |
if (*dec->cur < '0' || *dec->cur > '9') |
1140 |
ERR ("malformed number (no digits after decimal point)"); |
1141 |
|
1142 |
do |
1143 |
{ |
1144 |
++dec->cur; |
1145 |
} |
1146 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
1147 |
|
1148 |
is_nv = 1; |
1149 |
} |
1150 |
|
1151 |
// [exp] |
1152 |
if (*dec->cur == 'e' || *dec->cur == 'E') |
1153 |
{ |
1154 |
++dec->cur; |
1155 |
|
1156 |
if (*dec->cur == '-' || *dec->cur == '+') |
1157 |
++dec->cur; |
1158 |
|
1159 |
if (*dec->cur < '0' || *dec->cur > '9') |
1160 |
ERR ("malformed number (no digits after exp sign)"); |
1161 |
|
1162 |
do |
1163 |
{ |
1164 |
++dec->cur; |
1165 |
} |
1166 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
1167 |
|
1168 |
is_nv = 1; |
1169 |
} |
1170 |
|
1171 |
if (!is_nv) |
1172 |
{ |
1173 |
int len = dec->cur - start; |
1174 |
|
1175 |
// special case the rather common 1..5-digit-int case |
1176 |
if (*start == '-') |
1177 |
switch (len) |
1178 |
{ |
1179 |
case 2: return newSViv (-(IV)( start [1] - '0' * 1)); |
1180 |
case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11)); |
1181 |
case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
1182 |
case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
1183 |
case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); |
1184 |
} |
1185 |
else |
1186 |
switch (len) |
1187 |
{ |
1188 |
case 1: return newSViv ( start [0] - '0' * 1); |
1189 |
case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
1190 |
case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
1191 |
case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
1192 |
case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); |
1193 |
} |
1194 |
|
1195 |
{ |
1196 |
UV uv; |
1197 |
int numtype = grok_number (start, len, &uv); |
1198 |
if (numtype & IS_NUMBER_IN_UV) |
1199 |
if (numtype & IS_NUMBER_NEG) |
1200 |
{ |
1201 |
if (uv < (UV)IV_MIN) |
1202 |
return newSViv (-(IV)uv); |
1203 |
} |
1204 |
else |
1205 |
return newSVuv (uv); |
1206 |
} |
1207 |
|
1208 |
len -= *start == '-' ? 1 : 0; |
1209 |
|
1210 |
// does not fit into IV or UV, try NV |
1211 |
if (len <= NV_DIG) |
1212 |
// fits into NV without loss of precision |
1213 |
return newSVnv (json_atof (start)); |
1214 |
|
1215 |
// everything else fails, convert it to a string |
1216 |
return newSVpvn (start, dec->cur - start); |
1217 |
} |
1218 |
|
1219 |
// loss of precision here |
1220 |
return newSVnv (json_atof (start)); |
1221 |
|
1222 |
fail: |
1223 |
return 0; |
1224 |
} |
1225 |
|
1226 |
static SV * |
1227 |
decode_av (dec_t *dec) |
1228 |
{ |
1229 |
AV *av = newAV (); |
1230 |
|
1231 |
DEC_INC_DEPTH; |
1232 |
decode_ws (dec); |
1233 |
|
1234 |
if (*dec->cur == ']') |
1235 |
++dec->cur; |
1236 |
else |
1237 |
for (;;) |
1238 |
{ |
1239 |
SV *value; |
1240 |
|
1241 |
value = decode_sv (dec); |
1242 |
if (!value) |
1243 |
goto fail; |
1244 |
|
1245 |
av_push (av, value); |
1246 |
|
1247 |
decode_ws (dec); |
1248 |
|
1249 |
if (*dec->cur == ']') |
1250 |
{ |
1251 |
++dec->cur; |
1252 |
break; |
1253 |
} |
1254 |
|
1255 |
if (*dec->cur != ',') |
1256 |
ERR (", or ] expected while parsing array"); |
1257 |
|
1258 |
++dec->cur; |
1259 |
|
1260 |
decode_ws (dec); |
1261 |
|
1262 |
if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
1263 |
{ |
1264 |
++dec->cur; |
1265 |
break; |
1266 |
} |
1267 |
} |
1268 |
|
1269 |
DEC_DEC_DEPTH; |
1270 |
return newRV_noinc ((SV *)av); |
1271 |
|
1272 |
fail: |
1273 |
SvREFCNT_dec (av); |
1274 |
DEC_DEC_DEPTH; |
1275 |
return 0; |
1276 |
} |
1277 |
|
1278 |
static SV * |
1279 |
decode_hv (dec_t *dec) |
1280 |
{ |
1281 |
SV *sv; |
1282 |
HV *hv = newHV (); |
1283 |
|
1284 |
DEC_INC_DEPTH; |
1285 |
decode_ws (dec); |
1286 |
|
1287 |
if (*dec->cur == '}') |
1288 |
++dec->cur; |
1289 |
else |
1290 |
for (;;) |
1291 |
{ |
1292 |
EXPECT_CH ('"'); |
1293 |
|
1294 |
// heuristic: assume that |
1295 |
// a) decode_str + hv_store_ent are abysmally slow. |
1296 |
// b) most hash keys are short, simple ascii text. |
1297 |
// => try to "fast-match" such strings to avoid |
1298 |
// the overhead of decode_str + hv_store_ent. |
1299 |
{ |
1300 |
SV *value; |
1301 |
char *p = dec->cur; |
1302 |
char *e = p + 24; // only try up to 24 bytes |
1303 |
|
1304 |
for (;;) |
1305 |
{ |
1306 |
// the >= 0x80 is false on most architectures |
1307 |
if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
1308 |
{ |
1309 |
// slow path, back up and use decode_str |
1310 |
SV *key = decode_str (dec); |
1311 |
if (!key) |
1312 |
goto fail; |
1313 |
|
1314 |
decode_ws (dec); EXPECT_CH (':'); |
1315 |
|
1316 |
decode_ws (dec); |
1317 |
value = decode_sv (dec); |
1318 |
if (!value) |
1319 |
{ |
1320 |
SvREFCNT_dec (key); |
1321 |
goto fail; |
1322 |
} |
1323 |
|
1324 |
hv_store_ent (hv, key, value, 0); |
1325 |
SvREFCNT_dec (key); |
1326 |
|
1327 |
break; |
1328 |
} |
1329 |
else if (*p == '"') |
1330 |
{ |
1331 |
// fast path, got a simple key |
1332 |
char *key = dec->cur; |
1333 |
int len = p - key; |
1334 |
dec->cur = p + 1; |
1335 |
|
1336 |
decode_ws (dec); EXPECT_CH (':'); |
1337 |
|
1338 |
decode_ws (dec); |
1339 |
value = decode_sv (dec); |
1340 |
if (!value) |
1341 |
goto fail; |
1342 |
|
1343 |
hv_store (hv, key, len, value, 0); |
1344 |
|
1345 |
break; |
1346 |
} |
1347 |
|
1348 |
++p; |
1349 |
} |
1350 |
} |
1351 |
|
1352 |
decode_ws (dec); |
1353 |
|
1354 |
if (*dec->cur == '}') |
1355 |
{ |
1356 |
++dec->cur; |
1357 |
break; |
1358 |
} |
1359 |
|
1360 |
if (*dec->cur != ',') |
1361 |
ERR (", or } expected while parsing object/hash"); |
1362 |
|
1363 |
++dec->cur; |
1364 |
|
1365 |
decode_ws (dec); |
1366 |
|
1367 |
if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
1368 |
{ |
1369 |
++dec->cur; |
1370 |
break; |
1371 |
} |
1372 |
} |
1373 |
|
1374 |
DEC_DEC_DEPTH; |
1375 |
sv = newRV_noinc ((SV *)hv); |
1376 |
|
1377 |
// check filter callbacks |
1378 |
if (dec->json.flags & F_HOOK) |
1379 |
{ |
1380 |
if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
1381 |
{ |
1382 |
HE *cb, *he; |
1383 |
|
1384 |
hv_iterinit (hv); |
1385 |
he = hv_iternext (hv); |
1386 |
hv_iterinit (hv); |
1387 |
|
1388 |
// the next line creates a mortal sv each time its called. |
1389 |
// might want to optimise this for common cases. |
1390 |
cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
1391 |
|
1392 |
if (cb) |
1393 |
{ |
1394 |
dSP; |
1395 |
int count; |
1396 |
|
1397 |
ENTER; SAVETMPS; PUSHMARK (SP); |
1398 |
XPUSHs (HeVAL (he)); |
1399 |
|
1400 |
PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
1401 |
|
1402 |
if (count == 1) |
1403 |
{ |
1404 |
sv = newSVsv (POPs); |
1405 |
FREETMPS; LEAVE; |
1406 |
return sv; |
1407 |
} |
1408 |
|
1409 |
FREETMPS; LEAVE; |
1410 |
} |
1411 |
} |
1412 |
|
1413 |
if (dec->json.cb_object) |
1414 |
{ |
1415 |
dSP; |
1416 |
int count; |
1417 |
|
1418 |
ENTER; SAVETMPS; PUSHMARK (SP); |
1419 |
XPUSHs (sv_2mortal (sv)); |
1420 |
|
1421 |
PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
1422 |
|
1423 |
if (count == 1) |
1424 |
{ |
1425 |
sv = newSVsv (POPs); |
1426 |
FREETMPS; LEAVE; |
1427 |
return sv; |
1428 |
} |
1429 |
|
1430 |
SvREFCNT_inc (sv); |
1431 |
FREETMPS; LEAVE; |
1432 |
} |
1433 |
} |
1434 |
|
1435 |
return sv; |
1436 |
|
1437 |
fail: |
1438 |
SvREFCNT_dec (hv); |
1439 |
DEC_DEC_DEPTH; |
1440 |
return 0; |
1441 |
} |
1442 |
|
1443 |
static SV * |
1444 |
decode_sv (dec_t *dec) |
1445 |
{ |
1446 |
// the beauty of JSON: you need exactly one character lookahead |
1447 |
// to parse everything. |
1448 |
switch (*dec->cur) |
1449 |
{ |
1450 |
case '"': ++dec->cur; return decode_str (dec); |
1451 |
case '[': ++dec->cur; return decode_av (dec); |
1452 |
case '{': ++dec->cur; return decode_hv (dec); |
1453 |
|
1454 |
case '-': |
1455 |
case '0': case '1': case '2': case '3': case '4': |
1456 |
case '5': case '6': case '7': case '8': case '9': |
1457 |
return decode_num (dec); |
1458 |
|
1459 |
case 't': |
1460 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1461 |
{ |
1462 |
dec->cur += 4; |
1463 |
#if JSON_SLOW |
1464 |
json_true = get_bool ("JSON::XS::true"); |
1465 |
#endif |
1466 |
return newSVsv (json_true); |
1467 |
} |
1468 |
else |
1469 |
ERR ("'true' expected"); |
1470 |
|
1471 |
break; |
1472 |
|
1473 |
case 'f': |
1474 |
if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1475 |
{ |
1476 |
dec->cur += 5; |
1477 |
#if JSON_SLOW |
1478 |
json_false = get_bool ("JSON::XS::false"); |
1479 |
#endif |
1480 |
return newSVsv (json_false); |
1481 |
} |
1482 |
else |
1483 |
ERR ("'false' expected"); |
1484 |
|
1485 |
break; |
1486 |
|
1487 |
case 'n': |
1488 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
1489 |
{ |
1490 |
dec->cur += 4; |
1491 |
return newSVsv (&PL_sv_undef); |
1492 |
} |
1493 |
else |
1494 |
ERR ("'null' expected"); |
1495 |
|
1496 |
break; |
1497 |
|
1498 |
default: |
1499 |
ERR ("malformed JSON string, neither array, object, number, string or atom"); |
1500 |
break; |
1501 |
} |
1502 |
|
1503 |
fail: |
1504 |
return 0; |
1505 |
} |
1506 |
|
1507 |
static SV * |
1508 |
decode_json (SV *string, JSON *json, char **offset_return) |
1509 |
{ |
1510 |
dec_t dec; |
1511 |
SV *sv; |
1512 |
|
1513 |
/* work around bugs in 5.10 where manipulating magic values |
1514 |
* will perl ignore the magic in subsequent accesses |
1515 |
*/ |
1516 |
/*SvGETMAGIC (string);*/ |
1517 |
if (SvMAGICAL (string)) |
1518 |
string = sv_2mortal (newSVsv (string)); |
1519 |
|
1520 |
SvUPGRADE (string, SVt_PV); |
1521 |
|
1522 |
/* work around a bug in perl 5.10, which causes SvCUR to fail an |
1523 |
* assertion with -DDEBUGGING, although SvCUR is documented to |
1524 |
* return the xpv_cur field which certainly exists after upgrading. |
1525 |
* according to nicholas clark, calling SvPOK fixes this. |
1526 |
* But it doesn't fix it, so try another workaround, call SvPV_nolen |
1527 |
* and hope for the best. |
1528 |
* Damnit, SvPV_nolen still trips over yet another assertion. This |
1529 |
* assertion business is seriously broken, try yet another workaround |
1530 |
* for the broken -DDEBUGGING. |
1531 |
*/ |
1532 |
{ |
1533 |
#ifdef DEBUGGING |
1534 |
STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
1535 |
#else |
1536 |
STRLEN offset = SvCUR (string); |
1537 |
#endif |
1538 |
|
1539 |
if (offset > json->max_size && json->max_size) |
1540 |
croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1541 |
(unsigned long)SvCUR (string), (unsigned long)json->max_size); |
1542 |
} |
1543 |
|
1544 |
if (json->flags & F_UTF8) |
1545 |
sv_utf8_downgrade (string, 0); |
1546 |
else |
1547 |
sv_utf8_upgrade (string); |
1548 |
|
1549 |
SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1550 |
|
1551 |
dec.json = *json; |
1552 |
dec.cur = SvPVX (string); |
1553 |
dec.end = SvEND (string); |
1554 |
dec.err = 0; |
1555 |
dec.depth = 0; |
1556 |
|
1557 |
if (dec.json.cb_object || dec.json.cb_sk_object) |
1558 |
dec.json.flags |= F_HOOK; |
1559 |
|
1560 |
*dec.end = 0; // this should basically be a nop, too, but make sure it's there |
1561 |
|
1562 |
decode_ws (&dec); |
1563 |
sv = decode_sv (&dec); |
1564 |
|
1565 |
if (offset_return) |
1566 |
*offset_return = dec.cur; |
1567 |
|
1568 |
if (!(offset_return || !sv)) |
1569 |
{ |
1570 |
// check for trailing garbage |
1571 |
decode_ws (&dec); |
1572 |
|
1573 |
if (*dec.cur) |
1574 |
{ |
1575 |
dec.err = "garbage after JSON object"; |
1576 |
SvREFCNT_dec (sv); |
1577 |
sv = 0; |
1578 |
} |
1579 |
} |
1580 |
|
1581 |
if (!sv) |
1582 |
{ |
1583 |
SV *uni = sv_newmortal (); |
1584 |
|
1585 |
// horrible hack to silence warning inside pv_uni_display |
1586 |
COP cop = *PL_curcop; |
1587 |
cop.cop_warnings = pWARN_NONE; |
1588 |
ENTER; |
1589 |
SAVEVPTR (PL_curcop); |
1590 |
PL_curcop = &cop; |
1591 |
pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1592 |
LEAVE; |
1593 |
|
1594 |
croak ("%s, at character offset %d (before \"%s\")", |
1595 |
dec.err, |
1596 |
ptr_to_index (string, dec.cur), |
1597 |
dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1598 |
} |
1599 |
|
1600 |
sv = sv_2mortal (sv); |
1601 |
|
1602 |
if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1603 |
croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
1604 |
|
1605 |
return sv; |
1606 |
} |
1607 |
|
1608 |
///////////////////////////////////////////////////////////////////////////// |
1609 |
// incremental parser |
1610 |
|
1611 |
static void |
1612 |
incr_parse (JSON *self) |
1613 |
{ |
1614 |
const char *p = SvPVX (self->incr_text) + self->incr_pos; |
1615 |
|
1616 |
// the state machine here is a bit convoluted and could be simplified a lot |
1617 |
// but this would make it slower, so... |
1618 |
|
1619 |
for (;;) |
1620 |
{ |
1621 |
//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 |
1622 |
switch (self->incr_mode) |
1623 |
{ |
1624 |
// only used for initial whitespace skipping |
1625 |
case INCR_M_WS: |
1626 |
for (;;) |
1627 |
{ |
1628 |
if (*p > 0x20) |
1629 |
{ |
1630 |
if (*p == '#') |
1631 |
{ |
1632 |
self->incr_mode = INCR_M_C0; |
1633 |
goto incr_m_c; |
1634 |
} |
1635 |
else |
1636 |
{ |
1637 |
self->incr_mode = INCR_M_JSON; |
1638 |
goto incr_m_json; |
1639 |
} |
1640 |
} |
1641 |
else if (!*p) |
1642 |
goto interrupt; |
1643 |
|
1644 |
++p; |
1645 |
} |
1646 |
|
1647 |
// skip a single char inside a string (for \\-processing) |
1648 |
case INCR_M_BS: |
1649 |
if (!*p) |
1650 |
goto interrupt; |
1651 |
|
1652 |
++p; |
1653 |
self->incr_mode = INCR_M_STR; |
1654 |
goto incr_m_str; |
1655 |
|
1656 |
// inside #-style comments |
1657 |
case INCR_M_C0: |
1658 |
case INCR_M_C1: |
1659 |
incr_m_c: |
1660 |
for (;;) |
1661 |
{ |
1662 |
if (*p == '\n') |
1663 |
{ |
1664 |
self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON; |
1665 |
break; |
1666 |
} |
1667 |
else if (!*p) |
1668 |
goto interrupt; |
1669 |
|
1670 |
++p; |
1671 |
} |
1672 |
|
1673 |
break; |
1674 |
|
1675 |
// inside a string |
1676 |
case INCR_M_STR: |
1677 |
incr_m_str: |
1678 |
for (;;) |
1679 |
{ |
1680 |
if (*p == '"') |
1681 |
{ |
1682 |
++p; |
1683 |
self->incr_mode = INCR_M_JSON; |
1684 |
|
1685 |
if (!self->incr_nest) |
1686 |
goto interrupt; |
1687 |
|
1688 |
goto incr_m_json; |
1689 |
} |
1690 |
else if (*p == '\\') |
1691 |
{ |
1692 |
++p; // "virtually" consumes character after \ |
1693 |
|
1694 |
if (!*p) // if at end of string we have to switch modes |
1695 |
{ |
1696 |
self->incr_mode = INCR_M_BS; |
1697 |
goto interrupt; |
1698 |
} |
1699 |
} |
1700 |
else if (!*p) |
1701 |
goto interrupt; |
1702 |
|
1703 |
++p; |
1704 |
} |
1705 |
|
1706 |
// after initial ws, outside string |
1707 |
case INCR_M_JSON: |
1708 |
incr_m_json: |
1709 |
for (;;) |
1710 |
{ |
1711 |
switch (*p++) |
1712 |
{ |
1713 |
case 0: |
1714 |
--p; |
1715 |
goto interrupt; |
1716 |
|
1717 |
case 0x09: |
1718 |
case 0x0a: |
1719 |
case 0x0d: |
1720 |
case 0x20: |
1721 |
if (!self->incr_nest) |
1722 |
{ |
1723 |
--p; // do not eat the whitespace, let the next round do it |
1724 |
goto interrupt; |
1725 |
} |
1726 |
break; |
1727 |
|
1728 |
case '"': |
1729 |
self->incr_mode = INCR_M_STR; |
1730 |
goto incr_m_str; |
1731 |
|
1732 |
case '[': |
1733 |
case '{': |
1734 |
if (++self->incr_nest > self->max_depth) |
1735 |
croak (ERR_NESTING_EXCEEDED); |
1736 |
break; |
1737 |
|
1738 |
case ']': |
1739 |
case '}': |
1740 |
if (--self->incr_nest <= 0) |
1741 |
goto interrupt; |
1742 |
break; |
1743 |
|
1744 |
case '#': |
1745 |
self->incr_mode = INCR_M_C1; |
1746 |
goto incr_m_c; |
1747 |
} |
1748 |
} |
1749 |
} |
1750 |
|
1751 |
modechange: |
1752 |
; |
1753 |
} |
1754 |
|
1755 |
interrupt: |
1756 |
self->incr_pos = p - SvPVX (self->incr_text); |
1757 |
//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D |
1758 |
//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D |
1759 |
} |
1760 |
|
1761 |
///////////////////////////////////////////////////////////////////////////// |
1762 |
// XS interface functions |
1763 |
|
1764 |
MODULE = JSON::XS PACKAGE = JSON::XS |
1765 |
|
1766 |
BOOT: |
1767 |
{ |
1768 |
int i; |
1769 |
|
1770 |
for (i = 0; i < 256; ++i) |
1771 |
decode_hexdigit [i] = |
1772 |
i >= '0' && i <= '9' ? i - '0' |
1773 |
: i >= 'a' && i <= 'f' ? i - 'a' + 10 |
1774 |
: i >= 'A' && i <= 'F' ? i - 'A' + 10 |
1775 |
: -1; |
1776 |
|
1777 |
json_stash = gv_stashpv ("JSON::XS" , 1); |
1778 |
json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
1779 |
|
1780 |
json_true = get_bool ("JSON::XS::true"); |
1781 |
json_false = get_bool ("JSON::XS::false"); |
1782 |
|
1783 |
CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */ |
1784 |
} |
1785 |
|
1786 |
PROTOTYPES: DISABLE |
1787 |
|
1788 |
void CLONE (...) |
1789 |
CODE: |
1790 |
json_stash = 0; |
1791 |
json_boolean_stash = 0; |
1792 |
|
1793 |
void new (char *klass) |
1794 |
PPCODE: |
1795 |
{ |
1796 |
SV *pv = NEWSV (0, sizeof (JSON)); |
1797 |
SvPOK_only (pv); |
1798 |
json_init ((JSON *)SvPVX (pv)); |
1799 |
XPUSHs (sv_2mortal (sv_bless ( |
1800 |
newRV_noinc (pv), |
1801 |
strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) |
1802 |
))); |
1803 |
} |
1804 |
|
1805 |
void ascii (JSON *self, int enable = 1) |
1806 |
ALIAS: |
1807 |
ascii = F_ASCII |
1808 |
latin1 = F_LATIN1 |
1809 |
utf8 = F_UTF8 |
1810 |
indent = F_INDENT |
1811 |
canonical = F_CANONICAL |
1812 |
space_before = F_SPACE_BEFORE |
1813 |
space_after = F_SPACE_AFTER |
1814 |
pretty = F_PRETTY |
1815 |
allow_nonref = F_ALLOW_NONREF |
1816 |
shrink = F_SHRINK |
1817 |
allow_blessed = F_ALLOW_BLESSED |
1818 |
convert_blessed = F_CONV_BLESSED |
1819 |
relaxed = F_RELAXED |
1820 |
allow_unknown = F_ALLOW_UNKNOWN |
1821 |
PPCODE: |
1822 |
{ |
1823 |
if (enable) |
1824 |
self->flags |= ix; |
1825 |
else |
1826 |
self->flags &= ~ix; |
1827 |
|
1828 |
XPUSHs (ST (0)); |
1829 |
} |
1830 |
|
1831 |
void get_ascii (JSON *self) |
1832 |
ALIAS: |
1833 |
get_ascii = F_ASCII |
1834 |
get_latin1 = F_LATIN1 |
1835 |
get_utf8 = F_UTF8 |
1836 |
get_indent = F_INDENT |
1837 |
get_canonical = F_CANONICAL |
1838 |
get_space_before = F_SPACE_BEFORE |
1839 |
get_space_after = F_SPACE_AFTER |
1840 |
get_allow_nonref = F_ALLOW_NONREF |
1841 |
get_shrink = F_SHRINK |
1842 |
get_allow_blessed = F_ALLOW_BLESSED |
1843 |
get_convert_blessed = F_CONV_BLESSED |
1844 |
get_relaxed = F_RELAXED |
1845 |
get_allow_unknown = F_ALLOW_UNKNOWN |
1846 |
PPCODE: |
1847 |
XPUSHs (boolSV (self->flags & ix)); |
1848 |
|
1849 |
void max_depth (JSON *self, U32 max_depth = 0x80000000UL) |
1850 |
PPCODE: |
1851 |
self->max_depth = max_depth; |
1852 |
XPUSHs (ST (0)); |
1853 |
|
1854 |
U32 get_max_depth (JSON *self) |
1855 |
CODE: |
1856 |
RETVAL = self->max_depth; |
1857 |
OUTPUT: |
1858 |
RETVAL |
1859 |
|
1860 |
void max_size (JSON *self, U32 max_size = 0) |
1861 |
PPCODE: |
1862 |
self->max_size = max_size; |
1863 |
XPUSHs (ST (0)); |
1864 |
|
1865 |
int get_max_size (JSON *self) |
1866 |
CODE: |
1867 |
RETVAL = self->max_size; |
1868 |
OUTPUT: |
1869 |
RETVAL |
1870 |
|
1871 |
void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
1872 |
PPCODE: |
1873 |
{ |
1874 |
SvREFCNT_dec (self->cb_object); |
1875 |
self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
1876 |
|
1877 |
XPUSHs (ST (0)); |
1878 |
} |
1879 |
|
1880 |
void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
1881 |
PPCODE: |
1882 |
{ |
1883 |
if (!self->cb_sk_object) |
1884 |
self->cb_sk_object = newHV (); |
1885 |
|
1886 |
if (SvOK (cb)) |
1887 |
hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
1888 |
else |
1889 |
{ |
1890 |
hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
1891 |
|
1892 |
if (!HvKEYS (self->cb_sk_object)) |
1893 |
{ |
1894 |
SvREFCNT_dec (self->cb_sk_object); |
1895 |
self->cb_sk_object = 0; |
1896 |
} |
1897 |
} |
1898 |
|
1899 |
XPUSHs (ST (0)); |
1900 |
} |
1901 |
|
1902 |
void encode (JSON *self, SV *scalar) |
1903 |
PPCODE: |
1904 |
XPUSHs (encode_json (scalar, self)); |
1905 |
|
1906 |
void decode (JSON *self, SV *jsonstr) |
1907 |
PPCODE: |
1908 |
XPUSHs (decode_json (jsonstr, self, 0)); |
1909 |
|
1910 |
void decode_prefix (JSON *self, SV *jsonstr) |
1911 |
PPCODE: |
1912 |
{ |
1913 |
char *offset; |
1914 |
EXTEND (SP, 2); |
1915 |
PUSHs (decode_json (jsonstr, self, &offset)); |
1916 |
PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset)))); |
1917 |
} |
1918 |
|
1919 |
void incr_parse (JSON *self, SV *jsonstr = 0) |
1920 |
PPCODE: |
1921 |
{ |
1922 |
if (!self->incr_text) |
1923 |
self->incr_text = newSVpvn ("", 0); |
1924 |
|
1925 |
// append data, if any |
1926 |
if (jsonstr) |
1927 |
{ |
1928 |
if (SvUTF8 (jsonstr)) |
1929 |
{ |
1930 |
if (!SvUTF8 (self->incr_text)) |
1931 |
{ |
1932 |
/* utf-8-ness differs, need to upgrade */ |
1933 |
sv_utf8_upgrade (self->incr_text); |
1934 |
|
1935 |
if (self->incr_pos) |
1936 |
self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos) |
1937 |
- (U8 *)SvPVX (self->incr_text); |
1938 |
} |
1939 |
} |
1940 |
else if (SvUTF8 (self->incr_text)) |
1941 |
sv_utf8_upgrade (jsonstr); |
1942 |
|
1943 |
{ |
1944 |
STRLEN len; |
1945 |
const char *str = SvPV (jsonstr, len); |
1946 |
STRLEN cur = SvCUR (self->incr_text); |
1947 |
|
1948 |
if (SvLEN (self->incr_text) <= cur + len) |
1949 |
SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
1950 |
|
1951 |
Move (str, SvEND (self->incr_text), len, char); |
1952 |
SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); |
1953 |
*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there |
1954 |
} |
1955 |
} |
1956 |
|
1957 |
if (GIMME_V != G_VOID) |
1958 |
do |
1959 |
{ |
1960 |
char *offset; |
1961 |
|
1962 |
if (!INCR_DONE (self)) |
1963 |
{ |
1964 |
incr_parse (self); |
1965 |
|
1966 |
if (self->incr_pos > self->max_size && self->max_size) |
1967 |
croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
1968 |
(unsigned long)self->incr_pos, (unsigned long)self->max_size); |
1969 |
|
1970 |
if (!INCR_DONE (self)) |
1971 |
break; |
1972 |
} |
1973 |
|
1974 |
XPUSHs (decode_json (self->incr_text, self, &offset)); |
1975 |
|
1976 |
self->incr_pos -= offset - SvPVX (self->incr_text); |
1977 |
self->incr_nest = 0; |
1978 |
self->incr_mode = 0; |
1979 |
|
1980 |
sv_chop (self->incr_text, offset); |
1981 |
} |
1982 |
while (GIMME_V == G_ARRAY); |
1983 |
} |
1984 |
|
1985 |
SV *incr_text (JSON *self) |
1986 |
ATTRS: lvalue |
1987 |
CODE: |
1988 |
{ |
1989 |
if (self->incr_pos) |
1990 |
croak ("incr_text can not be called when the incremental parser already started parsing"); |
1991 |
|
1992 |
RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef; |
1993 |
} |
1994 |
OUTPUT: |
1995 |
RETVAL |
1996 |
|
1997 |
void incr_skip (JSON *self) |
1998 |
CODE: |
1999 |
{ |
2000 |
if (self->incr_pos) |
2001 |
{ |
2002 |
sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos); |
2003 |
self->incr_pos = 0; |
2004 |
self->incr_nest = 0; |
2005 |
self->incr_mode = 0; |
2006 |
} |
2007 |
} |
2008 |
|
2009 |
void incr_reset (JSON *self) |
2010 |
CODE: |
2011 |
{ |
2012 |
SvREFCNT_dec (self->incr_text); |
2013 |
self->incr_text = 0; |
2014 |
self->incr_pos = 0; |
2015 |
self->incr_nest = 0; |
2016 |
self->incr_mode = 0; |
2017 |
} |
2018 |
|
2019 |
void DESTROY (JSON *self) |
2020 |
CODE: |
2021 |
SvREFCNT_dec (self->cb_sk_object); |
2022 |
SvREFCNT_dec (self->cb_object); |
2023 |
SvREFCNT_dec (self->incr_text); |
2024 |
|
2025 |
PROTOTYPES: ENABLE |
2026 |
|
2027 |
void encode_json (SV *scalar) |
2028 |
ALIAS: |
2029 |
to_json_ = 0 |
2030 |
encode_json = F_UTF8 |
2031 |
PPCODE: |
2032 |
{ |
2033 |
JSON json; |
2034 |
json_init (&json); |
2035 |
json.flags |= ix; |
2036 |
XPUSHs (encode_json (scalar, &json)); |
2037 |
} |
2038 |
|
2039 |
void decode_json (SV *jsonstr) |
2040 |
ALIAS: |
2041 |
from_json_ = 0 |
2042 |
decode_json = F_UTF8 |
2043 |
PPCODE: |
2044 |
{ |
2045 |
JSON json; |
2046 |
json_init (&json); |
2047 |
json.flags |= ix; |
2048 |
XPUSHs (decode_json (jsonstr, &json, 0)); |
2049 |
} |
2050 |
|