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