| 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 <float.h> |
| 10 |
|
| 11 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
| 12 |
# define snprintf _snprintf // C compilers have this in stdio.h |
| 13 |
#endif |
| 14 |
|
| 15 |
// some old perls do not have this, try to make it work, no |
| 16 |
// guarentees, though. if it breaks, you get to keep the pieces. |
| 17 |
#ifndef UTF8_MAXBYTES |
| 18 |
# define UTF8_MAXBYTES 13 |
| 19 |
#endif |
| 20 |
|
| 21 |
#define F_ASCII 0x00000001UL |
| 22 |
#define F_LATIN1 0x00000002UL |
| 23 |
#define F_UTF8 0x00000004UL |
| 24 |
#define F_INDENT 0x00000008UL |
| 25 |
#define F_CANONICAL 0x00000010UL |
| 26 |
#define F_SPACE_BEFORE 0x00000020UL |
| 27 |
#define F_SPACE_AFTER 0x00000040UL |
| 28 |
#define F_ALLOW_NONREF 0x00000100UL |
| 29 |
#define F_SHRINK 0x00000200UL |
| 30 |
#define F_ALLOW_BLESSED 0x00000400UL |
| 31 |
#define F_CONV_BLESSED 0x00000800UL |
| 32 |
#define F_RELAXED 0x00001000UL |
| 33 |
|
| 34 |
#define F_MAXDEPTH 0xf8000000UL |
| 35 |
#define S_MAXDEPTH 27 |
| 36 |
#define F_MAXSIZE 0x01f00000UL |
| 37 |
#define S_MAXSIZE 20 |
| 38 |
#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
| 39 |
|
| 40 |
#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
| 41 |
#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
| 42 |
|
| 43 |
#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
| 44 |
#define F_DEFAULT (9UL << S_MAXDEPTH) |
| 45 |
|
| 46 |
#define INIT_SIZE 32 // initial scalar size to be allocated |
| 47 |
#define INDENT_STEP 3 // spaces per indentation level |
| 48 |
|
| 49 |
#define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
| 50 |
|
| 51 |
#define SB do { |
| 52 |
#define SE } while (0) |
| 53 |
|
| 54 |
#if __GNUC__ >= 3 |
| 55 |
# define expect(expr,value) __builtin_expect ((expr),(value)) |
| 56 |
# define inline inline |
| 57 |
#else |
| 58 |
# define expect(expr,value) (expr) |
| 59 |
# define inline static |
| 60 |
#endif |
| 61 |
|
| 62 |
#define expect_false(expr) expect ((expr) != 0, 0) |
| 63 |
#define expect_true(expr) expect ((expr) != 0, 1) |
| 64 |
|
| 65 |
#ifdef USE_ITHREADS |
| 66 |
# define JSON_SLOW 1 |
| 67 |
# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
| 68 |
#else |
| 69 |
# define JSON_SLOW 0 |
| 70 |
# define JSON_STASH json_stash |
| 71 |
#endif |
| 72 |
|
| 73 |
static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
| 74 |
static SV *json_true, *json_false; |
| 75 |
|
| 76 |
typedef struct { |
| 77 |
U32 flags; |
| 78 |
SV *cb_object; |
| 79 |
HV *cb_sk_object; |
| 80 |
} JSON; |
| 81 |
|
| 82 |
///////////////////////////////////////////////////////////////////////////// |
| 83 |
// utility functions |
| 84 |
|
| 85 |
inline void |
| 86 |
shrink (SV *sv) |
| 87 |
{ |
| 88 |
sv_utf8_downgrade (sv, 1); |
| 89 |
if (SvLEN (sv) > SvCUR (sv) + 1) |
| 90 |
{ |
| 91 |
#ifdef SvPV_shrink_to_cur |
| 92 |
SvPV_shrink_to_cur (sv); |
| 93 |
#elif defined (SvPV_renew) |
| 94 |
SvPV_renew (sv, SvCUR (sv) + 1); |
| 95 |
#endif |
| 96 |
} |
| 97 |
} |
| 98 |
|
| 99 |
// decode an utf-8 character and return it, or (UV)-1 in |
| 100 |
// case of an error. |
| 101 |
// we special-case "safe" characters from U+80 .. U+7FF, |
| 102 |
// but use the very good perl function to parse anything else. |
| 103 |
// note that we never call this function for a ascii codepoints |
| 104 |
inline UV |
| 105 |
decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
| 106 |
{ |
| 107 |
if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
| 108 |
return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
| 109 |
else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
| 110 |
{ |
| 111 |
*clen = 2; |
| 112 |
return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
| 113 |
} |
| 114 |
else |
| 115 |
{ |
| 116 |
*clen = (STRLEN)-1; |
| 117 |
return (UV)-1; |
| 118 |
} |
| 119 |
} |
| 120 |
|
| 121 |
///////////////////////////////////////////////////////////////////////////// |
| 122 |
// encoder |
| 123 |
|
| 124 |
// structure used for encoding JSON |
| 125 |
typedef struct |
| 126 |
{ |
| 127 |
char *cur; // SvPVX (sv) + current output position |
| 128 |
char *end; // SvEND (sv) |
| 129 |
SV *sv; // result scalar |
| 130 |
JSON json; |
| 131 |
U32 indent; // indentation level |
| 132 |
U32 maxdepth; // max. indentation/recursion level |
| 133 |
} enc_t; |
| 134 |
|
| 135 |
inline void |
| 136 |
need (enc_t *enc, STRLEN len) |
| 137 |
{ |
| 138 |
if (expect_false (enc->cur + len >= enc->end)) |
| 139 |
{ |
| 140 |
STRLEN cur = enc->cur - SvPVX (enc->sv); |
| 141 |
SvGROW (enc->sv, cur + len + 1); |
| 142 |
enc->cur = SvPVX (enc->sv) + cur; |
| 143 |
enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
| 144 |
} |
| 145 |
} |
| 146 |
|
| 147 |
inline void |
| 148 |
encode_ch (enc_t *enc, char ch) |
| 149 |
{ |
| 150 |
need (enc, 1); |
| 151 |
*enc->cur++ = ch; |
| 152 |
} |
| 153 |
|
| 154 |
static void |
| 155 |
encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8) |
| 156 |
{ |
| 157 |
char *end = str + len; |
| 158 |
|
| 159 |
need (enc, len); |
| 160 |
|
| 161 |
while (str < end) |
| 162 |
{ |
| 163 |
unsigned char ch = *(unsigned char *)str; |
| 164 |
|
| 165 |
if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
| 166 |
{ |
| 167 |
if (expect_false (ch == '"')) // but with slow exceptions |
| 168 |
{ |
| 169 |
need (enc, len += 1); |
| 170 |
*enc->cur++ = '\\'; |
| 171 |
*enc->cur++ = '"'; |
| 172 |
} |
| 173 |
else if (expect_false (ch == '\\')) |
| 174 |
{ |
| 175 |
need (enc, len += 1); |
| 176 |
*enc->cur++ = '\\'; |
| 177 |
*enc->cur++ = '\\'; |
| 178 |
} |
| 179 |
else |
| 180 |
*enc->cur++ = ch; |
| 181 |
|
| 182 |
++str; |
| 183 |
} |
| 184 |
else |
| 185 |
{ |
| 186 |
switch (ch) |
| 187 |
{ |
| 188 |
case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
| 189 |
case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
| 190 |
case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
| 191 |
case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
| 192 |
case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
| 193 |
|
| 194 |
default: |
| 195 |
{ |
| 196 |
STRLEN clen; |
| 197 |
UV uch; |
| 198 |
|
| 199 |
if (is_utf8) |
| 200 |
{ |
| 201 |
uch = decode_utf8 (str, end - str, &clen); |
| 202 |
if (clen == (STRLEN)-1) |
| 203 |
croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
| 204 |
} |
| 205 |
else |
| 206 |
{ |
| 207 |
uch = ch; |
| 208 |
clen = 1; |
| 209 |
} |
| 210 |
|
| 211 |
if (uch > 0x10FFFFUL) |
| 212 |
croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
| 213 |
|
| 214 |
if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
| 215 |
{ |
| 216 |
if (uch > 0xFFFFUL) |
| 217 |
{ |
| 218 |
need (enc, len += 11); |
| 219 |
sprintf (enc->cur, "\\u%04x\\u%04x", |
| 220 |
(int)((uch - 0x10000) / 0x400 + 0xD800), |
| 221 |
(int)((uch - 0x10000) % 0x400 + 0xDC00)); |
| 222 |
enc->cur += 12; |
| 223 |
} |
| 224 |
else |
| 225 |
{ |
| 226 |
static char hexdigit [16] = "0123456789abcdef"; |
| 227 |
need (enc, len += 5); |
| 228 |
*enc->cur++ = '\\'; |
| 229 |
*enc->cur++ = 'u'; |
| 230 |
*enc->cur++ = hexdigit [ uch >> 12 ]; |
| 231 |
*enc->cur++ = hexdigit [(uch >> 8) & 15]; |
| 232 |
*enc->cur++ = hexdigit [(uch >> 4) & 15]; |
| 233 |
*enc->cur++ = hexdigit [(uch >> 0) & 15]; |
| 234 |
} |
| 235 |
|
| 236 |
str += clen; |
| 237 |
} |
| 238 |
else if (enc->json.flags & F_LATIN1) |
| 239 |
{ |
| 240 |
*enc->cur++ = uch; |
| 241 |
str += clen; |
| 242 |
} |
| 243 |
else if (is_utf8) |
| 244 |
{ |
| 245 |
need (enc, len += clen); |
| 246 |
do |
| 247 |
{ |
| 248 |
*enc->cur++ = *str++; |
| 249 |
} |
| 250 |
while (--clen); |
| 251 |
} |
| 252 |
else |
| 253 |
{ |
| 254 |
need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
| 255 |
enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
| 256 |
++str; |
| 257 |
} |
| 258 |
} |
| 259 |
} |
| 260 |
} |
| 261 |
|
| 262 |
--len; |
| 263 |
} |
| 264 |
} |
| 265 |
|
| 266 |
inline void |
| 267 |
encode_indent (enc_t *enc) |
| 268 |
{ |
| 269 |
if (enc->json.flags & F_INDENT) |
| 270 |
{ |
| 271 |
int spaces = enc->indent * INDENT_STEP; |
| 272 |
|
| 273 |
need (enc, spaces); |
| 274 |
memset (enc->cur, ' ', spaces); |
| 275 |
enc->cur += spaces; |
| 276 |
} |
| 277 |
} |
| 278 |
|
| 279 |
inline void |
| 280 |
encode_space (enc_t *enc) |
| 281 |
{ |
| 282 |
need (enc, 1); |
| 283 |
encode_ch (enc, ' '); |
| 284 |
} |
| 285 |
|
| 286 |
inline void |
| 287 |
encode_nl (enc_t *enc) |
| 288 |
{ |
| 289 |
if (enc->json.flags & F_INDENT) |
| 290 |
{ |
| 291 |
need (enc, 1); |
| 292 |
encode_ch (enc, '\n'); |
| 293 |
} |
| 294 |
} |
| 295 |
|
| 296 |
inline void |
| 297 |
encode_comma (enc_t *enc) |
| 298 |
{ |
| 299 |
encode_ch (enc, ','); |
| 300 |
|
| 301 |
if (enc->json.flags & F_INDENT) |
| 302 |
encode_nl (enc); |
| 303 |
else if (enc->json.flags & F_SPACE_AFTER) |
| 304 |
encode_space (enc); |
| 305 |
} |
| 306 |
|
| 307 |
static void encode_sv (enc_t *enc, SV *sv); |
| 308 |
|
| 309 |
static void |
| 310 |
encode_av (enc_t *enc, AV *av) |
| 311 |
{ |
| 312 |
int i, len = av_len (av); |
| 313 |
|
| 314 |
if (enc->indent >= enc->maxdepth) |
| 315 |
croak ("data structure too deep (hit recursion limit)"); |
| 316 |
|
| 317 |
encode_ch (enc, '['); encode_nl (enc); |
| 318 |
++enc->indent; |
| 319 |
|
| 320 |
for (i = 0; i <= len; ++i) |
| 321 |
{ |
| 322 |
SV **svp = av_fetch (av, i, 0); |
| 323 |
|
| 324 |
encode_indent (enc); |
| 325 |
|
| 326 |
if (svp) |
| 327 |
encode_sv (enc, *svp); |
| 328 |
else |
| 329 |
encode_str (enc, "null", 4, 0); |
| 330 |
|
| 331 |
if (i < len) |
| 332 |
encode_comma (enc); |
| 333 |
} |
| 334 |
|
| 335 |
encode_nl (enc); |
| 336 |
|
| 337 |
--enc->indent; |
| 338 |
encode_indent (enc); encode_ch (enc, ']'); |
| 339 |
} |
| 340 |
|
| 341 |
static void |
| 342 |
encode_hk (enc_t *enc, HE *he) |
| 343 |
{ |
| 344 |
encode_ch (enc, '"'); |
| 345 |
|
| 346 |
if (HeKLEN (he) == HEf_SVKEY) |
| 347 |
{ |
| 348 |
SV *sv = HeSVKEY (he); |
| 349 |
STRLEN len; |
| 350 |
char *str; |
| 351 |
|
| 352 |
SvGETMAGIC (sv); |
| 353 |
str = SvPV (sv, len); |
| 354 |
|
| 355 |
encode_str (enc, str, len, SvUTF8 (sv)); |
| 356 |
} |
| 357 |
else |
| 358 |
encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
| 359 |
|
| 360 |
encode_ch (enc, '"'); |
| 361 |
|
| 362 |
if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
| 363 |
encode_ch (enc, ':'); |
| 364 |
if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
| 365 |
} |
| 366 |
|
| 367 |
// compare hash entries, used when all keys are bytestrings |
| 368 |
static int |
| 369 |
he_cmp_fast (const void *a_, const void *b_) |
| 370 |
{ |
| 371 |
int cmp; |
| 372 |
|
| 373 |
HE *a = *(HE **)a_; |
| 374 |
HE *b = *(HE **)b_; |
| 375 |
|
| 376 |
STRLEN la = HeKLEN (a); |
| 377 |
STRLEN lb = HeKLEN (b); |
| 378 |
|
| 379 |
if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
| 380 |
cmp = lb - la; |
| 381 |
|
| 382 |
return cmp; |
| 383 |
} |
| 384 |
|
| 385 |
// compare hash entries, used when some keys are sv's or utf-x |
| 386 |
static int |
| 387 |
he_cmp_slow (const void *a, const void *b) |
| 388 |
{ |
| 389 |
return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
| 390 |
} |
| 391 |
|
| 392 |
static void |
| 393 |
encode_hv (enc_t *enc, HV *hv) |
| 394 |
{ |
| 395 |
HE *he; |
| 396 |
int count; |
| 397 |
|
| 398 |
if (enc->indent >= enc->maxdepth) |
| 399 |
croak ("data structure too deep (hit recursion limit)"); |
| 400 |
|
| 401 |
encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
| 402 |
|
| 403 |
// for canonical output we have to sort by keys first |
| 404 |
// actually, this is mostly due to the stupid so-called |
| 405 |
// security workaround added somewhere in 5.8.x. |
| 406 |
// that randomises hash orderings |
| 407 |
if (enc->json.flags & F_CANONICAL) |
| 408 |
{ |
| 409 |
int count = hv_iterinit (hv); |
| 410 |
|
| 411 |
if (SvMAGICAL (hv)) |
| 412 |
{ |
| 413 |
// need to count by iterating. could improve by dynamically building the vector below |
| 414 |
// but I don't care for the speed of this special case. |
| 415 |
// note also that we will run into undefined behaviour when the two iterations |
| 416 |
// do not result in the same count, something I might care for in some later release. |
| 417 |
|
| 418 |
count = 0; |
| 419 |
while (hv_iternext (hv)) |
| 420 |
++count; |
| 421 |
|
| 422 |
hv_iterinit (hv); |
| 423 |
} |
| 424 |
|
| 425 |
if (count) |
| 426 |
{ |
| 427 |
int i, fast = 1; |
| 428 |
#if defined(__BORLANDC__) || defined(_MSC_VER) |
| 429 |
HE **hes = _alloca (count * sizeof (HE)); |
| 430 |
#else |
| 431 |
HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
| 432 |
#endif |
| 433 |
|
| 434 |
i = 0; |
| 435 |
while ((he = hv_iternext (hv))) |
| 436 |
{ |
| 437 |
hes [i++] = he; |
| 438 |
if (HeKLEN (he) < 0 || HeKUTF8 (he)) |
| 439 |
fast = 0; |
| 440 |
} |
| 441 |
|
| 442 |
assert (i == count); |
| 443 |
|
| 444 |
if (fast) |
| 445 |
qsort (hes, count, sizeof (HE *), he_cmp_fast); |
| 446 |
else |
| 447 |
{ |
| 448 |
// hack to forcefully disable "use bytes" |
| 449 |
COP cop = *PL_curcop; |
| 450 |
cop.op_private = 0; |
| 451 |
|
| 452 |
ENTER; |
| 453 |
SAVETMPS; |
| 454 |
|
| 455 |
SAVEVPTR (PL_curcop); |
| 456 |
PL_curcop = &cop; |
| 457 |
|
| 458 |
qsort (hes, count, sizeof (HE *), he_cmp_slow); |
| 459 |
|
| 460 |
FREETMPS; |
| 461 |
LEAVE; |
| 462 |
} |
| 463 |
|
| 464 |
while (count--) |
| 465 |
{ |
| 466 |
encode_indent (enc); |
| 467 |
he = hes [count]; |
| 468 |
encode_hk (enc, he); |
| 469 |
encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
| 470 |
|
| 471 |
if (count) |
| 472 |
encode_comma (enc); |
| 473 |
} |
| 474 |
} |
| 475 |
} |
| 476 |
else |
| 477 |
{ |
| 478 |
if (hv_iterinit (hv) || SvMAGICAL (hv)) |
| 479 |
if ((he = hv_iternext (hv))) |
| 480 |
for (;;) |
| 481 |
{ |
| 482 |
encode_indent (enc); |
| 483 |
encode_hk (enc, he); |
| 484 |
encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
| 485 |
|
| 486 |
if (!(he = hv_iternext (hv))) |
| 487 |
break; |
| 488 |
|
| 489 |
encode_comma (enc); |
| 490 |
} |
| 491 |
} |
| 492 |
|
| 493 |
encode_nl (enc); |
| 494 |
|
| 495 |
--enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
| 496 |
} |
| 497 |
|
| 498 |
// encode objects, arrays and special \0=false and \1=true values. |
| 499 |
static void |
| 500 |
encode_rv (enc_t *enc, SV *sv) |
| 501 |
{ |
| 502 |
svtype svt; |
| 503 |
|
| 504 |
SvGETMAGIC (sv); |
| 505 |
svt = SvTYPE (sv); |
| 506 |
|
| 507 |
if (expect_false (SvOBJECT (sv))) |
| 508 |
{ |
| 509 |
HV *stash = !JSON_SLOW || json_boolean_stash |
| 510 |
? json_boolean_stash |
| 511 |
: gv_stashpv ("JSON::XS::Boolean", 1); |
| 512 |
|
| 513 |
if (SvSTASH (sv) == stash) |
| 514 |
{ |
| 515 |
if (SvIV (sv)) |
| 516 |
encode_str (enc, "true", 4, 0); |
| 517 |
else |
| 518 |
encode_str (enc, "false", 5, 0); |
| 519 |
} |
| 520 |
else |
| 521 |
{ |
| 522 |
#if 0 |
| 523 |
if (0 && sv_derived_from (rv, "JSON::Literal")) |
| 524 |
{ |
| 525 |
// not yet |
| 526 |
} |
| 527 |
#endif |
| 528 |
if (enc->json.flags & F_CONV_BLESSED) |
| 529 |
{ |
| 530 |
// we re-bless the reference to get overload and other niceties right |
| 531 |
GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
| 532 |
|
| 533 |
if (to_json) |
| 534 |
{ |
| 535 |
dSP; |
| 536 |
|
| 537 |
ENTER; SAVETMPS; PUSHMARK (SP); |
| 538 |
XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
| 539 |
|
| 540 |
// calling with G_SCALAR ensures that we always get a 1 return value |
| 541 |
PUTBACK; |
| 542 |
call_sv ((SV *)GvCV (to_json), G_SCALAR); |
| 543 |
SPAGAIN; |
| 544 |
|
| 545 |
// catch this surprisingly common error |
| 546 |
if (SvROK (TOPs) && SvRV (TOPs) == sv) |
| 547 |
croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
| 548 |
|
| 549 |
sv = POPs; |
| 550 |
PUTBACK; |
| 551 |
|
| 552 |
encode_sv (enc, sv); |
| 553 |
|
| 554 |
FREETMPS; LEAVE; |
| 555 |
} |
| 556 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
| 557 |
encode_str (enc, "null", 4, 0); |
| 558 |
else |
| 559 |
croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
| 560 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
| 561 |
} |
| 562 |
else if (enc->json.flags & F_ALLOW_BLESSED) |
| 563 |
encode_str (enc, "null", 4, 0); |
| 564 |
else |
| 565 |
croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
| 566 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
| 567 |
} |
| 568 |
} |
| 569 |
else if (svt == SVt_PVHV) |
| 570 |
encode_hv (enc, (HV *)sv); |
| 571 |
else if (svt == SVt_PVAV) |
| 572 |
encode_av (enc, (AV *)sv); |
| 573 |
else if (svt < SVt_PVAV) |
| 574 |
{ |
| 575 |
STRLEN len = 0; |
| 576 |
char *pv = svt ? SvPV (sv, len) : 0; |
| 577 |
|
| 578 |
if (len == 1 && *pv == '1') |
| 579 |
encode_str (enc, "true", 4, 0); |
| 580 |
else if (len == 1 && *pv == '0') |
| 581 |
encode_str (enc, "false", 5, 0); |
| 582 |
else |
| 583 |
croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
| 584 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
| 585 |
} |
| 586 |
else |
| 587 |
croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
| 588 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
| 589 |
} |
| 590 |
|
| 591 |
static void |
| 592 |
encode_sv (enc_t *enc, SV *sv) |
| 593 |
{ |
| 594 |
SvGETMAGIC (sv); |
| 595 |
|
| 596 |
if (SvPOKp (sv)) |
| 597 |
{ |
| 598 |
STRLEN len; |
| 599 |
char *str = SvPV (sv, len); |
| 600 |
encode_ch (enc, '"'); |
| 601 |
encode_str (enc, str, len, SvUTF8 (sv)); |
| 602 |
encode_ch (enc, '"'); |
| 603 |
} |
| 604 |
else if (SvNOKp (sv)) |
| 605 |
{ |
| 606 |
// trust that perl will do the right thing w.r.t. JSON syntax. |
| 607 |
need (enc, NV_DIG + 32); |
| 608 |
Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
| 609 |
enc->cur += strlen (enc->cur); |
| 610 |
} |
| 611 |
else if (SvIOKp (sv)) |
| 612 |
{ |
| 613 |
// we assume we can always read an IV as a UV |
| 614 |
if (SvUV (sv) & ~(UV)0x7fff) |
| 615 |
{ |
| 616 |
// large integer, use the (rather slow) snprintf way. |
| 617 |
need (enc, sizeof (UV) * 3); |
| 618 |
enc->cur += |
| 619 |
SvIsUV(sv) |
| 620 |
? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
| 621 |
: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
| 622 |
} |
| 623 |
else |
| 624 |
{ |
| 625 |
// optimise the "small number case" |
| 626 |
// code will likely be branchless and use only a single multiplication |
| 627 |
I32 i = SvIV (sv); |
| 628 |
U32 u; |
| 629 |
char digit, nz = 0; |
| 630 |
|
| 631 |
need (enc, 6); |
| 632 |
|
| 633 |
*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
| 634 |
u = i < 0 ? -i : i; |
| 635 |
|
| 636 |
// convert to 4.28 fixed-point representation |
| 637 |
u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
| 638 |
|
| 639 |
// now output digit by digit, each time masking out the integer part |
| 640 |
// and multiplying by 5 while moving the decimal point one to the right, |
| 641 |
// resulting in a net multiplication by 10. |
| 642 |
// we always write the digit to memory but conditionally increment |
| 643 |
// the pointer, to ease the usage of conditional move instructions. |
| 644 |
digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
| 645 |
digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
| 646 |
digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
| 647 |
digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
| 648 |
digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
| 649 |
} |
| 650 |
} |
| 651 |
else if (SvROK (sv)) |
| 652 |
encode_rv (enc, SvRV (sv)); |
| 653 |
else if (!SvOK (sv)) |
| 654 |
encode_str (enc, "null", 4, 0); |
| 655 |
else |
| 656 |
croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
| 657 |
SvPV_nolen (sv), SvFLAGS (sv)); |
| 658 |
} |
| 659 |
|
| 660 |
static SV * |
| 661 |
encode_json (SV *scalar, JSON *json) |
| 662 |
{ |
| 663 |
enc_t enc; |
| 664 |
|
| 665 |
if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
| 666 |
croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
| 667 |
|
| 668 |
enc.json = *json; |
| 669 |
enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
| 670 |
enc.cur = SvPVX (enc.sv); |
| 671 |
enc.end = SvEND (enc.sv); |
| 672 |
enc.indent = 0; |
| 673 |
enc.maxdepth = DEC_DEPTH (enc.json.flags); |
| 674 |
|
| 675 |
SvPOK_only (enc.sv); |
| 676 |
encode_sv (&enc, scalar); |
| 677 |
|
| 678 |
SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
| 679 |
*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
| 680 |
|
| 681 |
if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
| 682 |
SvUTF8_on (enc.sv); |
| 683 |
|
| 684 |
if (enc.json.flags & F_SHRINK) |
| 685 |
shrink (enc.sv); |
| 686 |
|
| 687 |
return enc.sv; |
| 688 |
} |
| 689 |
|
| 690 |
///////////////////////////////////////////////////////////////////////////// |
| 691 |
// decoder |
| 692 |
|
| 693 |
// structure used for decoding JSON |
| 694 |
typedef struct |
| 695 |
{ |
| 696 |
char *cur; // current parser pointer |
| 697 |
char *end; // end of input string |
| 698 |
const char *err; // parse error, if != 0 |
| 699 |
JSON json; |
| 700 |
U32 depth; // recursion depth |
| 701 |
U32 maxdepth; // recursion depth limit |
| 702 |
} dec_t; |
| 703 |
|
| 704 |
inline void |
| 705 |
decode_comment (dec_t *dec) |
| 706 |
{ |
| 707 |
// only '#'-style comments allowed a.t.m. |
| 708 |
|
| 709 |
while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) |
| 710 |
++dec->cur; |
| 711 |
} |
| 712 |
|
| 713 |
inline void |
| 714 |
decode_ws (dec_t *dec) |
| 715 |
{ |
| 716 |
for (;;) |
| 717 |
{ |
| 718 |
char ch = *dec->cur; |
| 719 |
|
| 720 |
if (ch > 0x20) |
| 721 |
{ |
| 722 |
if (expect_false (ch == '#')) |
| 723 |
{ |
| 724 |
if (dec->json.flags & F_RELAXED) |
| 725 |
decode_comment (dec); |
| 726 |
else |
| 727 |
break; |
| 728 |
} |
| 729 |
else |
| 730 |
break; |
| 731 |
} |
| 732 |
else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09) |
| 733 |
break; // parse error, but let higher level handle it, gives better error messages |
| 734 |
|
| 735 |
++dec->cur; |
| 736 |
} |
| 737 |
} |
| 738 |
|
| 739 |
#define ERR(reason) SB dec->err = reason; goto fail; SE |
| 740 |
|
| 741 |
#define EXPECT_CH(ch) SB \ |
| 742 |
if (*dec->cur != ch) \ |
| 743 |
ERR (# ch " expected"); \ |
| 744 |
++dec->cur; \ |
| 745 |
SE |
| 746 |
|
| 747 |
#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
| 748 |
#define DEC_DEC_DEPTH --dec->depth |
| 749 |
|
| 750 |
static SV *decode_sv (dec_t *dec); |
| 751 |
|
| 752 |
static signed char decode_hexdigit[256]; |
| 753 |
|
| 754 |
static UV |
| 755 |
decode_4hex (dec_t *dec) |
| 756 |
{ |
| 757 |
signed char d1, d2, d3, d4; |
| 758 |
unsigned char *cur = (unsigned char *)dec->cur; |
| 759 |
|
| 760 |
d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
| 761 |
d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
| 762 |
d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
| 763 |
d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
| 764 |
|
| 765 |
dec->cur += 4; |
| 766 |
|
| 767 |
return ((UV)d1) << 12 |
| 768 |
| ((UV)d2) << 8 |
| 769 |
| ((UV)d3) << 4 |
| 770 |
| ((UV)d4); |
| 771 |
|
| 772 |
fail: |
| 773 |
return (UV)-1; |
| 774 |
} |
| 775 |
|
| 776 |
static SV * |
| 777 |
decode_str (dec_t *dec) |
| 778 |
{ |
| 779 |
SV *sv = 0; |
| 780 |
int utf8 = 0; |
| 781 |
char *dec_cur = dec->cur; |
| 782 |
|
| 783 |
do |
| 784 |
{ |
| 785 |
char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
| 786 |
char *cur = buf; |
| 787 |
|
| 788 |
do |
| 789 |
{ |
| 790 |
unsigned char ch = *(unsigned char *)dec_cur++; |
| 791 |
|
| 792 |
if (expect_false (ch == '"')) |
| 793 |
{ |
| 794 |
--dec_cur; |
| 795 |
break; |
| 796 |
} |
| 797 |
else if (expect_false (ch == '\\')) |
| 798 |
{ |
| 799 |
switch (*dec_cur) |
| 800 |
{ |
| 801 |
case '\\': |
| 802 |
case '/': |
| 803 |
case '"': *cur++ = *dec_cur++; break; |
| 804 |
|
| 805 |
case 'b': ++dec_cur; *cur++ = '\010'; break; |
| 806 |
case 't': ++dec_cur; *cur++ = '\011'; break; |
| 807 |
case 'n': ++dec_cur; *cur++ = '\012'; break; |
| 808 |
case 'f': ++dec_cur; *cur++ = '\014'; break; |
| 809 |
case 'r': ++dec_cur; *cur++ = '\015'; break; |
| 810 |
|
| 811 |
case 'u': |
| 812 |
{ |
| 813 |
UV lo, hi; |
| 814 |
++dec_cur; |
| 815 |
|
| 816 |
dec->cur = dec_cur; |
| 817 |
hi = decode_4hex (dec); |
| 818 |
dec_cur = dec->cur; |
| 819 |
if (hi == (UV)-1) |
| 820 |
goto fail; |
| 821 |
|
| 822 |
// possibly a surrogate pair |
| 823 |
if (hi >= 0xd800) |
| 824 |
if (hi < 0xdc00) |
| 825 |
{ |
| 826 |
if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
| 827 |
ERR ("missing low surrogate character in surrogate pair"); |
| 828 |
|
| 829 |
dec_cur += 2; |
| 830 |
|
| 831 |
dec->cur = dec_cur; |
| 832 |
lo = decode_4hex (dec); |
| 833 |
dec_cur = dec->cur; |
| 834 |
if (lo == (UV)-1) |
| 835 |
goto fail; |
| 836 |
|
| 837 |
if (lo < 0xdc00 || lo >= 0xe000) |
| 838 |
ERR ("surrogate pair expected"); |
| 839 |
|
| 840 |
hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
| 841 |
} |
| 842 |
else if (hi < 0xe000) |
| 843 |
ERR ("missing high surrogate character in surrogate pair"); |
| 844 |
|
| 845 |
if (hi >= 0x80) |
| 846 |
{ |
| 847 |
utf8 = 1; |
| 848 |
|
| 849 |
cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
| 850 |
} |
| 851 |
else |
| 852 |
*cur++ = hi; |
| 853 |
} |
| 854 |
break; |
| 855 |
|
| 856 |
default: |
| 857 |
--dec_cur; |
| 858 |
ERR ("illegal backslash escape sequence in string"); |
| 859 |
} |
| 860 |
} |
| 861 |
else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
| 862 |
*cur++ = ch; |
| 863 |
else if (ch >= 0x80) |
| 864 |
{ |
| 865 |
STRLEN clen; |
| 866 |
UV uch; |
| 867 |
|
| 868 |
--dec_cur; |
| 869 |
|
| 870 |
uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
| 871 |
if (clen == (STRLEN)-1) |
| 872 |
ERR ("malformed UTF-8 character in JSON string"); |
| 873 |
|
| 874 |
do |
| 875 |
*cur++ = *dec_cur++; |
| 876 |
while (--clen); |
| 877 |
|
| 878 |
utf8 = 1; |
| 879 |
} |
| 880 |
else |
| 881 |
{ |
| 882 |
--dec_cur; |
| 883 |
|
| 884 |
if (!ch) |
| 885 |
ERR ("unexpected end of string while parsing JSON string"); |
| 886 |
else |
| 887 |
ERR ("invalid character encountered while parsing JSON string"); |
| 888 |
} |
| 889 |
} |
| 890 |
while (cur < buf + SHORT_STRING_LEN); |
| 891 |
|
| 892 |
{ |
| 893 |
STRLEN len = cur - buf; |
| 894 |
|
| 895 |
if (sv) |
| 896 |
{ |
| 897 |
SvGROW (sv, SvCUR (sv) + len + 1); |
| 898 |
memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
| 899 |
SvCUR_set (sv, SvCUR (sv) + len); |
| 900 |
} |
| 901 |
else |
| 902 |
sv = newSVpvn (buf, len); |
| 903 |
} |
| 904 |
} |
| 905 |
while (*dec_cur != '"'); |
| 906 |
|
| 907 |
++dec_cur; |
| 908 |
|
| 909 |
if (sv) |
| 910 |
{ |
| 911 |
SvPOK_only (sv); |
| 912 |
*SvEND (sv) = 0; |
| 913 |
|
| 914 |
if (utf8) |
| 915 |
SvUTF8_on (sv); |
| 916 |
} |
| 917 |
else |
| 918 |
sv = newSVpvn ("", 0); |
| 919 |
|
| 920 |
dec->cur = dec_cur; |
| 921 |
return sv; |
| 922 |
|
| 923 |
fail: |
| 924 |
dec->cur = dec_cur; |
| 925 |
return 0; |
| 926 |
} |
| 927 |
|
| 928 |
static SV * |
| 929 |
decode_num (dec_t *dec) |
| 930 |
{ |
| 931 |
int is_nv = 0; |
| 932 |
char *start = dec->cur; |
| 933 |
|
| 934 |
// [minus] |
| 935 |
if (*dec->cur == '-') |
| 936 |
++dec->cur; |
| 937 |
|
| 938 |
if (*dec->cur == '0') |
| 939 |
{ |
| 940 |
++dec->cur; |
| 941 |
if (*dec->cur >= '0' && *dec->cur <= '9') |
| 942 |
ERR ("malformed number (leading zero must not be followed by another digit)"); |
| 943 |
} |
| 944 |
else if (*dec->cur < '0' || *dec->cur > '9') |
| 945 |
ERR ("malformed number (no digits after initial minus)"); |
| 946 |
else |
| 947 |
do |
| 948 |
{ |
| 949 |
++dec->cur; |
| 950 |
} |
| 951 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
| 952 |
|
| 953 |
// [frac] |
| 954 |
if (*dec->cur == '.') |
| 955 |
{ |
| 956 |
++dec->cur; |
| 957 |
|
| 958 |
if (*dec->cur < '0' || *dec->cur > '9') |
| 959 |
ERR ("malformed number (no digits after decimal point)"); |
| 960 |
|
| 961 |
do |
| 962 |
{ |
| 963 |
++dec->cur; |
| 964 |
} |
| 965 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
| 966 |
|
| 967 |
is_nv = 1; |
| 968 |
} |
| 969 |
|
| 970 |
// [exp] |
| 971 |
if (*dec->cur == 'e' || *dec->cur == 'E') |
| 972 |
{ |
| 973 |
++dec->cur; |
| 974 |
|
| 975 |
if (*dec->cur == '-' || *dec->cur == '+') |
| 976 |
++dec->cur; |
| 977 |
|
| 978 |
if (*dec->cur < '0' || *dec->cur > '9') |
| 979 |
ERR ("malformed number (no digits after exp sign)"); |
| 980 |
|
| 981 |
do |
| 982 |
{ |
| 983 |
++dec->cur; |
| 984 |
} |
| 985 |
while (*dec->cur >= '0' && *dec->cur <= '9'); |
| 986 |
|
| 987 |
is_nv = 1; |
| 988 |
} |
| 989 |
|
| 990 |
if (!is_nv) |
| 991 |
{ |
| 992 |
int len = dec->cur - start; |
| 993 |
|
| 994 |
// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
| 995 |
if (*start == '-') |
| 996 |
switch (len) |
| 997 |
{ |
| 998 |
case 2: return newSViv (-( start [1] - '0' * 1)); |
| 999 |
case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
| 1000 |
case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
| 1001 |
case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
| 1002 |
} |
| 1003 |
else |
| 1004 |
switch (len) |
| 1005 |
{ |
| 1006 |
case 1: return newSViv ( start [0] - '0' * 1); |
| 1007 |
case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
| 1008 |
case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
| 1009 |
case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
| 1010 |
} |
| 1011 |
|
| 1012 |
{ |
| 1013 |
UV uv; |
| 1014 |
int numtype = grok_number (start, len, &uv); |
| 1015 |
if (numtype & IS_NUMBER_IN_UV) |
| 1016 |
if (numtype & IS_NUMBER_NEG) |
| 1017 |
{ |
| 1018 |
if (uv < (UV)IV_MIN) |
| 1019 |
return newSViv (-(IV)uv); |
| 1020 |
} |
| 1021 |
else |
| 1022 |
return newSVuv (uv); |
| 1023 |
} |
| 1024 |
|
| 1025 |
len -= *start == '-' ? 1 : 0; |
| 1026 |
|
| 1027 |
// does not fit into IV or UV, try NV |
| 1028 |
if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
| 1029 |
#if defined (LDBL_DIG) |
| 1030 |
|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
| 1031 |
#endif |
| 1032 |
) |
| 1033 |
// fits into NV without loss of precision |
| 1034 |
return newSVnv (Atof (start)); |
| 1035 |
|
| 1036 |
// everything else fails, convert it to a string |
| 1037 |
return newSVpvn (start, dec->cur - start); |
| 1038 |
} |
| 1039 |
|
| 1040 |
// loss of precision here |
| 1041 |
return newSVnv (Atof (start)); |
| 1042 |
|
| 1043 |
fail: |
| 1044 |
return 0; |
| 1045 |
} |
| 1046 |
|
| 1047 |
static SV * |
| 1048 |
decode_av (dec_t *dec) |
| 1049 |
{ |
| 1050 |
AV *av = newAV (); |
| 1051 |
|
| 1052 |
DEC_INC_DEPTH; |
| 1053 |
decode_ws (dec); |
| 1054 |
|
| 1055 |
if (*dec->cur == ']') |
| 1056 |
++dec->cur; |
| 1057 |
else |
| 1058 |
for (;;) |
| 1059 |
{ |
| 1060 |
SV *value; |
| 1061 |
|
| 1062 |
value = decode_sv (dec); |
| 1063 |
if (!value) |
| 1064 |
goto fail; |
| 1065 |
|
| 1066 |
av_push (av, value); |
| 1067 |
|
| 1068 |
decode_ws (dec); |
| 1069 |
|
| 1070 |
if (*dec->cur == ']') |
| 1071 |
{ |
| 1072 |
++dec->cur; |
| 1073 |
break; |
| 1074 |
} |
| 1075 |
|
| 1076 |
if (*dec->cur != ',') |
| 1077 |
ERR (", or ] expected while parsing array"); |
| 1078 |
|
| 1079 |
++dec->cur; |
| 1080 |
|
| 1081 |
decode_ws (dec); |
| 1082 |
|
| 1083 |
if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
| 1084 |
{ |
| 1085 |
++dec->cur; |
| 1086 |
break; |
| 1087 |
} |
| 1088 |
} |
| 1089 |
|
| 1090 |
DEC_DEC_DEPTH; |
| 1091 |
return newRV_noinc ((SV *)av); |
| 1092 |
|
| 1093 |
fail: |
| 1094 |
SvREFCNT_dec (av); |
| 1095 |
DEC_DEC_DEPTH; |
| 1096 |
return 0; |
| 1097 |
} |
| 1098 |
|
| 1099 |
static SV * |
| 1100 |
decode_hv (dec_t *dec) |
| 1101 |
{ |
| 1102 |
SV *sv; |
| 1103 |
HV *hv = newHV (); |
| 1104 |
|
| 1105 |
DEC_INC_DEPTH; |
| 1106 |
decode_ws (dec); |
| 1107 |
|
| 1108 |
if (*dec->cur == '}') |
| 1109 |
++dec->cur; |
| 1110 |
else |
| 1111 |
for (;;) |
| 1112 |
{ |
| 1113 |
EXPECT_CH ('"'); |
| 1114 |
|
| 1115 |
// heuristic: assume that |
| 1116 |
// a) decode_str + hv_store_ent are abysmally slow. |
| 1117 |
// b) most hash keys are short, simple ascii text. |
| 1118 |
// => try to "fast-match" such strings to avoid |
| 1119 |
// the overhead of decode_str + hv_store_ent. |
| 1120 |
{ |
| 1121 |
SV *value; |
| 1122 |
char *p = dec->cur; |
| 1123 |
char *e = p + 24; // only try up to 24 bytes |
| 1124 |
|
| 1125 |
for (;;) |
| 1126 |
{ |
| 1127 |
// the >= 0x80 is true on most architectures |
| 1128 |
if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
| 1129 |
{ |
| 1130 |
// slow path, back up and use decode_str |
| 1131 |
SV *key = decode_str (dec); |
| 1132 |
if (!key) |
| 1133 |
goto fail; |
| 1134 |
|
| 1135 |
decode_ws (dec); EXPECT_CH (':'); |
| 1136 |
|
| 1137 |
decode_ws (dec); |
| 1138 |
value = decode_sv (dec); |
| 1139 |
if (!value) |
| 1140 |
{ |
| 1141 |
SvREFCNT_dec (key); |
| 1142 |
goto fail; |
| 1143 |
} |
| 1144 |
|
| 1145 |
hv_store_ent (hv, key, value, 0); |
| 1146 |
SvREFCNT_dec (key); |
| 1147 |
|
| 1148 |
break; |
| 1149 |
} |
| 1150 |
else if (*p == '"') |
| 1151 |
{ |
| 1152 |
// fast path, got a simple key |
| 1153 |
char *key = dec->cur; |
| 1154 |
int len = p - key; |
| 1155 |
dec->cur = p + 1; |
| 1156 |
|
| 1157 |
decode_ws (dec); EXPECT_CH (':'); |
| 1158 |
|
| 1159 |
decode_ws (dec); |
| 1160 |
value = decode_sv (dec); |
| 1161 |
if (!value) |
| 1162 |
goto fail; |
| 1163 |
|
| 1164 |
hv_store (hv, key, len, value, 0); |
| 1165 |
|
| 1166 |
break; |
| 1167 |
} |
| 1168 |
|
| 1169 |
++p; |
| 1170 |
} |
| 1171 |
} |
| 1172 |
|
| 1173 |
decode_ws (dec); |
| 1174 |
|
| 1175 |
if (*dec->cur == '}') |
| 1176 |
{ |
| 1177 |
++dec->cur; |
| 1178 |
break; |
| 1179 |
} |
| 1180 |
|
| 1181 |
if (*dec->cur != ',') |
| 1182 |
ERR (", or } expected while parsing object/hash"); |
| 1183 |
|
| 1184 |
++dec->cur; |
| 1185 |
|
| 1186 |
decode_ws (dec); |
| 1187 |
|
| 1188 |
if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
| 1189 |
{ |
| 1190 |
++dec->cur; |
| 1191 |
break; |
| 1192 |
} |
| 1193 |
} |
| 1194 |
|
| 1195 |
DEC_DEC_DEPTH; |
| 1196 |
sv = newRV_noinc ((SV *)hv); |
| 1197 |
|
| 1198 |
// check filter callbacks |
| 1199 |
if (dec->json.flags & F_HOOK) |
| 1200 |
{ |
| 1201 |
if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
| 1202 |
{ |
| 1203 |
HE *cb, *he; |
| 1204 |
|
| 1205 |
hv_iterinit (hv); |
| 1206 |
he = hv_iternext (hv); |
| 1207 |
hv_iterinit (hv); |
| 1208 |
|
| 1209 |
// the next line creates a mortal sv each time its called. |
| 1210 |
// might want to optimise this for common cases. |
| 1211 |
cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
| 1212 |
|
| 1213 |
if (cb) |
| 1214 |
{ |
| 1215 |
dSP; |
| 1216 |
int count; |
| 1217 |
|
| 1218 |
ENTER; SAVETMPS; PUSHMARK (SP); |
| 1219 |
XPUSHs (HeVAL (he)); |
| 1220 |
|
| 1221 |
PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
| 1222 |
|
| 1223 |
if (count == 1) |
| 1224 |
{ |
| 1225 |
sv = newSVsv (POPs); |
| 1226 |
FREETMPS; LEAVE; |
| 1227 |
return sv; |
| 1228 |
} |
| 1229 |
|
| 1230 |
FREETMPS; LEAVE; |
| 1231 |
} |
| 1232 |
} |
| 1233 |
|
| 1234 |
if (dec->json.cb_object) |
| 1235 |
{ |
| 1236 |
dSP; |
| 1237 |
int count; |
| 1238 |
|
| 1239 |
ENTER; SAVETMPS; PUSHMARK (SP); |
| 1240 |
XPUSHs (sv_2mortal (sv)); |
| 1241 |
|
| 1242 |
PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
| 1243 |
|
| 1244 |
if (count == 1) |
| 1245 |
{ |
| 1246 |
sv = newSVsv (POPs); |
| 1247 |
FREETMPS; LEAVE; |
| 1248 |
return sv; |
| 1249 |
} |
| 1250 |
|
| 1251 |
SvREFCNT_inc (sv); |
| 1252 |
FREETMPS; LEAVE; |
| 1253 |
} |
| 1254 |
} |
| 1255 |
|
| 1256 |
return sv; |
| 1257 |
|
| 1258 |
fail: |
| 1259 |
SvREFCNT_dec (hv); |
| 1260 |
DEC_DEC_DEPTH; |
| 1261 |
return 0; |
| 1262 |
} |
| 1263 |
|
| 1264 |
static SV * |
| 1265 |
decode_sv (dec_t *dec) |
| 1266 |
{ |
| 1267 |
// the beauty of JSON: you need exactly one character lookahead |
| 1268 |
// to parse anything. |
| 1269 |
switch (*dec->cur) |
| 1270 |
{ |
| 1271 |
case '"': ++dec->cur; return decode_str (dec); |
| 1272 |
case '[': ++dec->cur; return decode_av (dec); |
| 1273 |
case '{': ++dec->cur; return decode_hv (dec); |
| 1274 |
|
| 1275 |
case '-': |
| 1276 |
case '0': case '1': case '2': case '3': case '4': |
| 1277 |
case '5': case '6': case '7': case '8': case '9': |
| 1278 |
return decode_num (dec); |
| 1279 |
|
| 1280 |
case 't': |
| 1281 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
| 1282 |
{ |
| 1283 |
dec->cur += 4; |
| 1284 |
#if JSON_SLOW |
| 1285 |
json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
| 1286 |
#endif |
| 1287 |
return SvREFCNT_inc (json_true); |
| 1288 |
} |
| 1289 |
else |
| 1290 |
ERR ("'true' expected"); |
| 1291 |
|
| 1292 |
break; |
| 1293 |
|
| 1294 |
case 'f': |
| 1295 |
if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
| 1296 |
{ |
| 1297 |
dec->cur += 5; |
| 1298 |
#if JSON_SLOW |
| 1299 |
json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
| 1300 |
#endif |
| 1301 |
return SvREFCNT_inc (json_false); |
| 1302 |
} |
| 1303 |
else |
| 1304 |
ERR ("'false' expected"); |
| 1305 |
|
| 1306 |
break; |
| 1307 |
|
| 1308 |
case 'n': |
| 1309 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
| 1310 |
{ |
| 1311 |
dec->cur += 4; |
| 1312 |
return newSVsv (&PL_sv_undef); |
| 1313 |
} |
| 1314 |
else |
| 1315 |
ERR ("'null' expected"); |
| 1316 |
|
| 1317 |
break; |
| 1318 |
|
| 1319 |
default: |
| 1320 |
ERR ("malformed JSON string, neither array, object, number, string or atom"); |
| 1321 |
break; |
| 1322 |
} |
| 1323 |
|
| 1324 |
fail: |
| 1325 |
return 0; |
| 1326 |
} |
| 1327 |
|
| 1328 |
static SV * |
| 1329 |
decode_json (SV *string, JSON *json, UV *offset_return) |
| 1330 |
{ |
| 1331 |
dec_t dec; |
| 1332 |
UV offset; |
| 1333 |
SV *sv; |
| 1334 |
|
| 1335 |
SvGETMAGIC (string); |
| 1336 |
SvUPGRADE (string, SVt_PV); |
| 1337 |
|
| 1338 |
if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
| 1339 |
croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
| 1340 |
(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
| 1341 |
|
| 1342 |
if (json->flags & F_UTF8) |
| 1343 |
sv_utf8_downgrade (string, 0); |
| 1344 |
else |
| 1345 |
sv_utf8_upgrade (string); |
| 1346 |
|
| 1347 |
SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
| 1348 |
|
| 1349 |
dec.json = *json; |
| 1350 |
dec.cur = SvPVX (string); |
| 1351 |
dec.end = SvEND (string); |
| 1352 |
dec.err = 0; |
| 1353 |
dec.depth = 0; |
| 1354 |
dec.maxdepth = DEC_DEPTH (dec.json.flags); |
| 1355 |
|
| 1356 |
if (dec.json.cb_object || dec.json.cb_sk_object) |
| 1357 |
dec.json.flags |= F_HOOK; |
| 1358 |
|
| 1359 |
*dec.end = 0; // this should basically be a nop, too, but make sure it's there |
| 1360 |
|
| 1361 |
decode_ws (&dec); |
| 1362 |
sv = decode_sv (&dec); |
| 1363 |
|
| 1364 |
if (!(offset_return || !sv)) |
| 1365 |
{ |
| 1366 |
// check for trailing garbage |
| 1367 |
decode_ws (&dec); |
| 1368 |
|
| 1369 |
if (*dec.cur) |
| 1370 |
{ |
| 1371 |
dec.err = "garbage after JSON object"; |
| 1372 |
SvREFCNT_dec (sv); |
| 1373 |
sv = 0; |
| 1374 |
} |
| 1375 |
} |
| 1376 |
|
| 1377 |
if (offset_return || !sv) |
| 1378 |
{ |
| 1379 |
offset = dec.json.flags & F_UTF8 |
| 1380 |
? dec.cur - SvPVX (string) |
| 1381 |
: utf8_distance (dec.cur, SvPVX (string)); |
| 1382 |
|
| 1383 |
if (offset_return) |
| 1384 |
*offset_return = offset; |
| 1385 |
} |
| 1386 |
|
| 1387 |
if (!sv) |
| 1388 |
{ |
| 1389 |
SV *uni = sv_newmortal (); |
| 1390 |
|
| 1391 |
// horrible hack to silence warning inside pv_uni_display |
| 1392 |
COP cop = *PL_curcop; |
| 1393 |
cop.cop_warnings = pWARN_NONE; |
| 1394 |
ENTER; |
| 1395 |
SAVEVPTR (PL_curcop); |
| 1396 |
PL_curcop = &cop; |
| 1397 |
pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
| 1398 |
LEAVE; |
| 1399 |
|
| 1400 |
croak ("%s, at character offset %d [\"%s\"]", |
| 1401 |
dec.err, |
| 1402 |
(int)offset, |
| 1403 |
dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
| 1404 |
} |
| 1405 |
|
| 1406 |
sv = sv_2mortal (sv); |
| 1407 |
|
| 1408 |
if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
| 1409 |
croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
| 1410 |
|
| 1411 |
return sv; |
| 1412 |
} |
| 1413 |
|
| 1414 |
///////////////////////////////////////////////////////////////////////////// |
| 1415 |
// XS interface functions |
| 1416 |
|
| 1417 |
MODULE = JSON::XS PACKAGE = JSON::XS |
| 1418 |
|
| 1419 |
BOOT: |
| 1420 |
{ |
| 1421 |
int i; |
| 1422 |
|
| 1423 |
for (i = 0; i < 256; ++i) |
| 1424 |
decode_hexdigit [i] = |
| 1425 |
i >= '0' && i <= '9' ? i - '0' |
| 1426 |
: i >= 'a' && i <= 'f' ? i - 'a' + 10 |
| 1427 |
: i >= 'A' && i <= 'F' ? i - 'A' + 10 |
| 1428 |
: -1; |
| 1429 |
|
| 1430 |
json_stash = gv_stashpv ("JSON::XS" , 1); |
| 1431 |
json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
| 1432 |
|
| 1433 |
json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
| 1434 |
json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
| 1435 |
} |
| 1436 |
|
| 1437 |
PROTOTYPES: DISABLE |
| 1438 |
|
| 1439 |
void CLONE (...) |
| 1440 |
CODE: |
| 1441 |
json_stash = 0; |
| 1442 |
json_boolean_stash = 0; |
| 1443 |
|
| 1444 |
void new (char *klass) |
| 1445 |
PPCODE: |
| 1446 |
{ |
| 1447 |
SV *pv = NEWSV (0, sizeof (JSON)); |
| 1448 |
SvPOK_only (pv); |
| 1449 |
Zero (SvPVX (pv), 1, JSON); |
| 1450 |
((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
| 1451 |
XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); |
| 1452 |
} |
| 1453 |
|
| 1454 |
void ascii (JSON *self, int enable = 1) |
| 1455 |
ALIAS: |
| 1456 |
ascii = F_ASCII |
| 1457 |
latin1 = F_LATIN1 |
| 1458 |
utf8 = F_UTF8 |
| 1459 |
indent = F_INDENT |
| 1460 |
canonical = F_CANONICAL |
| 1461 |
space_before = F_SPACE_BEFORE |
| 1462 |
space_after = F_SPACE_AFTER |
| 1463 |
pretty = F_PRETTY |
| 1464 |
allow_nonref = F_ALLOW_NONREF |
| 1465 |
shrink = F_SHRINK |
| 1466 |
allow_blessed = F_ALLOW_BLESSED |
| 1467 |
convert_blessed = F_CONV_BLESSED |
| 1468 |
relaxed = F_RELAXED |
| 1469 |
PPCODE: |
| 1470 |
{ |
| 1471 |
if (enable) |
| 1472 |
self->flags |= ix; |
| 1473 |
else |
| 1474 |
self->flags &= ~ix; |
| 1475 |
|
| 1476 |
XPUSHs (ST (0)); |
| 1477 |
} |
| 1478 |
|
| 1479 |
void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
| 1480 |
PPCODE: |
| 1481 |
{ |
| 1482 |
UV log2 = 0; |
| 1483 |
|
| 1484 |
if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
| 1485 |
|
| 1486 |
while ((1UL << log2) < max_depth) |
| 1487 |
++log2; |
| 1488 |
|
| 1489 |
self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
| 1490 |
|
| 1491 |
XPUSHs (ST (0)); |
| 1492 |
} |
| 1493 |
|
| 1494 |
void max_size (JSON *self, UV max_size = 0) |
| 1495 |
PPCODE: |
| 1496 |
{ |
| 1497 |
UV log2 = 0; |
| 1498 |
|
| 1499 |
if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
| 1500 |
if (max_size == 1) max_size = 2; |
| 1501 |
|
| 1502 |
while ((1UL << log2) < max_size) |
| 1503 |
++log2; |
| 1504 |
|
| 1505 |
self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
| 1506 |
|
| 1507 |
XPUSHs (ST (0)); |
| 1508 |
} |
| 1509 |
|
| 1510 |
void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
| 1511 |
PPCODE: |
| 1512 |
{ |
| 1513 |
SvREFCNT_dec (self->cb_object); |
| 1514 |
self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
| 1515 |
|
| 1516 |
XPUSHs (ST (0)); |
| 1517 |
} |
| 1518 |
|
| 1519 |
void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
| 1520 |
PPCODE: |
| 1521 |
{ |
| 1522 |
if (!self->cb_sk_object) |
| 1523 |
self->cb_sk_object = newHV (); |
| 1524 |
|
| 1525 |
if (SvOK (cb)) |
| 1526 |
hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
| 1527 |
else |
| 1528 |
{ |
| 1529 |
hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
| 1530 |
|
| 1531 |
if (!HvKEYS (self->cb_sk_object)) |
| 1532 |
{ |
| 1533 |
SvREFCNT_dec (self->cb_sk_object); |
| 1534 |
self->cb_sk_object = 0; |
| 1535 |
} |
| 1536 |
} |
| 1537 |
|
| 1538 |
XPUSHs (ST (0)); |
| 1539 |
} |
| 1540 |
|
| 1541 |
void encode (JSON *self, SV *scalar) |
| 1542 |
PPCODE: |
| 1543 |
XPUSHs (encode_json (scalar, self)); |
| 1544 |
|
| 1545 |
void decode (JSON *self, SV *jsonstr) |
| 1546 |
PPCODE: |
| 1547 |
XPUSHs (decode_json (jsonstr, self, 0)); |
| 1548 |
|
| 1549 |
void decode_prefix (JSON *self, SV *jsonstr) |
| 1550 |
PPCODE: |
| 1551 |
{ |
| 1552 |
UV offset; |
| 1553 |
EXTEND (SP, 2); |
| 1554 |
PUSHs (decode_json (jsonstr, self, &offset)); |
| 1555 |
PUSHs (sv_2mortal (newSVuv (offset))); |
| 1556 |
} |
| 1557 |
|
| 1558 |
void DESTROY (JSON *self) |
| 1559 |
CODE: |
| 1560 |
SvREFCNT_dec (self->cb_sk_object); |
| 1561 |
SvREFCNT_dec (self->cb_object); |
| 1562 |
|
| 1563 |
PROTOTYPES: ENABLE |
| 1564 |
|
| 1565 |
void to_json (SV *scalar) |
| 1566 |
PPCODE: |
| 1567 |
{ |
| 1568 |
JSON json = { F_DEFAULT | F_UTF8 }; |
| 1569 |
XPUSHs (encode_json (scalar, &json)); |
| 1570 |
} |
| 1571 |
|
| 1572 |
void from_json (SV *jsonstr) |
| 1573 |
PPCODE: |
| 1574 |
{ |
| 1575 |
JSON json = { F_DEFAULT | F_UTF8 }; |
| 1576 |
XPUSHs (decode_json (jsonstr, &json, 0)); |
| 1577 |
} |
| 1578 |
|
