ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/JSON-XS/XS.xs
(Generate patch)

Comparing JSON-XS/XS.xs (file contents):
Revision 1.62 by root, Sun Aug 26 22:27:32 2007 UTC vs.
Revision 1.77 by root, Tue Mar 25 06:37:38 2008 UTC

… …
4 4
5#include <assert.h> 5#include <assert.h>
6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h>
9#include <float.h> 10#include <float.h>
10 11
11#if defined(__BORLANDC__) || defined(_MSC_VER) 12#if defined(__BORLANDC__) || defined(_MSC_VER)
12# define snprintf _snprintf // C compilers have this in stdio.h 13# define snprintf _snprintf // C compilers have this in stdio.h
13#endif 14#endif
… …
15// some old perls do not have this, try to make it work, no 16// 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// guarentees, though. if it breaks, you get to keep the pieces.
17#ifndef UTF8_MAXBYTES 18#ifndef UTF8_MAXBYTES
18# define UTF8_MAXBYTES 13 19# define UTF8_MAXBYTES 13
19#endif 20#endif
21
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
20 23
21#define F_ASCII 0x00000001UL 24#define F_ASCII 0x00000001UL
22#define F_LATIN1 0x00000002UL 25#define F_LATIN1 0x00000002UL
23#define F_UTF8 0x00000004UL 26#define F_UTF8 0x00000004UL
24#define F_INDENT 0x00000008UL 27#define F_INDENT 0x00000008UL
… …
27#define F_SPACE_AFTER 0x00000040UL 30#define F_SPACE_AFTER 0x00000040UL
28#define F_ALLOW_NONREF 0x00000100UL 31#define F_ALLOW_NONREF 0x00000100UL
29#define F_SHRINK 0x00000200UL 32#define F_SHRINK 0x00000200UL
30#define F_ALLOW_BLESSED 0x00000400UL 33#define F_ALLOW_BLESSED 0x00000400UL
31#define F_CONV_BLESSED 0x00000800UL 34#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL
36
32#define F_MAXDEPTH 0xf8000000UL 37#define F_MAXDEPTH 0xf8000000UL
33#define S_MAXDEPTH 27 38#define S_MAXDEPTH 27
34#define F_MAXSIZE 0x01f00000UL 39#define F_MAXSIZE 0x01f00000UL
35#define S_MAXSIZE 20 40#define S_MAXSIZE 20
36#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 41#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
… …
48 53
49#define SB do { 54#define SB do {
50#define SE } while (0) 55#define SE } while (0)
51 56
52#if __GNUC__ >= 3 57#if __GNUC__ >= 3
53# define expect(expr,value) __builtin_expect ((expr),(value)) 58# define expect(expr,value) __builtin_expect ((expr), (value))
54# define inline inline 59# define INLINE static inline
55#else 60#else
56# define expect(expr,value) (expr) 61# define expect(expr,value) (expr)
57# define inline static 62# define INLINE static
58#endif 63#endif
59 64
60#define expect_false(expr) expect ((expr) != 0, 0) 65#define expect_false(expr) expect ((expr) != 0, 0)
61#define expect_true(expr) expect ((expr) != 0, 1) 66#define expect_true(expr) expect ((expr) != 0, 1)
67
68#define IN_RANGE_INC(type,val,beg,end) \
69 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
70 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
62 71
63#ifdef USE_ITHREADS 72#ifdef USE_ITHREADS
64# define JSON_SLOW 1 73# define JSON_SLOW 1
65# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 74# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
66#else 75#else
… …
69#endif 78#endif
70 79
71static HV *json_stash, *json_boolean_stash; // JSON::XS:: 80static HV *json_stash, *json_boolean_stash; // JSON::XS::
72static SV *json_true, *json_false; 81static SV *json_true, *json_false;
73 82
83enum {
84 INCR_M_WS = 0, // initial whitespace skipping, must be 0
85 INCR_M_STR, // inside string
86 INCR_M_BS, // inside backslash
87 INCR_M_JSON // outside anything, count nesting
88};
89
90#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
91
74typedef struct { 92typedef struct {
75 U32 flags; 93 U32 flags;
76 SV *cb_object; 94 SV *cb_object;
77 HV *cb_sk_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 int incr_mode;
78} JSON; 102} JSON;
79 103
80///////////////////////////////////////////////////////////////////////////// 104/////////////////////////////////////////////////////////////////////////////
81// utility functions 105// utility functions
82 106
83inline void 107INLINE void
84shrink (SV *sv) 108shrink (SV *sv)
85{ 109{
86 sv_utf8_downgrade (sv, 1); 110 sv_utf8_downgrade (sv, 1);
87 if (SvLEN (sv) > SvCUR (sv) + 1) 111 if (SvLEN (sv) > SvCUR (sv) + 1)
88 { 112 {
… …
97// decode an utf-8 character and return it, or (UV)-1 in 121// decode an utf-8 character and return it, or (UV)-1 in
98// case of an error. 122// case of an error.
99// we special-case "safe" characters from U+80 .. U+7FF, 123// we special-case "safe" characters from U+80 .. U+7FF,
100// but use the very good perl function to parse anything else. 124// but use the very good perl function to parse anything else.
101// note that we never call this function for a ascii codepoints 125// note that we never call this function for a ascii codepoints
102inline UV 126INLINE UV
103decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 127decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
104{ 128{
105 if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) 129 if (expect_true (len >= 2
106 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 130 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
107 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 131 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
108 { 132 {
109 *clen = 2; 133 *clen = 2;
110 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 134 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
111 } 135 }
112 else 136 else
113 { 137 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
114 *clen = (STRLEN)-1; 138}
115 return (UV)-1; 139
116 } 140// likewise for encoding, also never called for ascii codepoints
141// this function takes advantage of this fact, although current gccs
142// seem to optimise the check for >= 0x80 away anyways
143INLINE unsigned char *
144encode_utf8 (unsigned char *s, UV ch)
145{
146 if (expect_false (ch < 0x000080))
147 *s++ = ch;
148 else if (expect_true (ch < 0x000800))
149 *s++ = 0xc0 | ( ch >> 6),
150 *s++ = 0x80 | ( ch & 0x3f);
151 else if ( ch < 0x010000)
152 *s++ = 0xe0 | ( ch >> 12),
153 *s++ = 0x80 | ((ch >> 6) & 0x3f),
154 *s++ = 0x80 | ( ch & 0x3f);
155 else if ( ch < 0x110000)
156 *s++ = 0xf0 | ( ch >> 18),
157 *s++ = 0x80 | ((ch >> 12) & 0x3f),
158 *s++ = 0x80 | ((ch >> 6) & 0x3f),
159 *s++ = 0x80 | ( ch & 0x3f);
160
161 return s;
117} 162}
118 163
119///////////////////////////////////////////////////////////////////////////// 164/////////////////////////////////////////////////////////////////////////////
120// encoder 165// encoder
121 166
… …
126 char *end; // SvEND (sv) 171 char *end; // SvEND (sv)
127 SV *sv; // result scalar 172 SV *sv; // result scalar
128 JSON json; 173 JSON json;
129 U32 indent; // indentation level 174 U32 indent; // indentation level
130 U32 maxdepth; // max. indentation/recursion level 175 U32 maxdepth; // max. indentation/recursion level
176 UV limit; // escape character values >= this value when encoding
131} enc_t; 177} enc_t;
132 178
133inline void 179INLINE void
134need (enc_t *enc, STRLEN len) 180need (enc_t *enc, STRLEN len)
135{ 181{
136 if (expect_false (enc->cur + len >= enc->end)) 182 if (expect_false (enc->cur + len >= enc->end))
137 { 183 {
138 STRLEN cur = enc->cur - SvPVX (enc->sv); 184 STRLEN cur = enc->cur - SvPVX (enc->sv);
… …
140 enc->cur = SvPVX (enc->sv) + cur; 186 enc->cur = SvPVX (enc->sv) + cur;
141 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 187 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
142 } 188 }
143} 189}
144 190
145inline void 191INLINE void
146encode_ch (enc_t *enc, char ch) 192encode_ch (enc_t *enc, char ch)
147{ 193{
148 need (enc, 1); 194 need (enc, 1);
149 *enc->cur++ = ch; 195 *enc->cur++ = ch;
150} 196}
… …
204 { 250 {
205 uch = ch; 251 uch = ch;
206 clen = 1; 252 clen = 1;
207 } 253 }
208 254
209 if (uch > 0x10FFFFUL) 255 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
210 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
211
212 if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
213 { 256 {
214 if (uch > 0xFFFFUL) 257 if (uch >= 0x10000UL)
215 { 258 {
259 if (uch >= 0x110000UL)
260 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
261
216 need (enc, len += 11); 262 need (enc, len += 11);
217 sprintf (enc->cur, "\\u%04x\\u%04x", 263 sprintf (enc->cur, "\\u%04x\\u%04x",
218 (int)((uch - 0x10000) / 0x400 + 0xD800), 264 (int)((uch - 0x10000) / 0x400 + 0xD800),
219 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 265 (int)((uch - 0x10000) % 0x400 + 0xDC00));
220 enc->cur += 12; 266 enc->cur += 12;
… …
248 while (--clen); 294 while (--clen);
249 } 295 }
250 else 296 else
251 { 297 {
252 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 298 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
253 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 299 enc->cur = encode_utf8 (enc->cur, uch);
254 ++str; 300 ++str;
255 } 301 }
256 } 302 }
257 } 303 }
258 } 304 }
259 305
260 --len; 306 --len;
261 } 307 }
262} 308}
263 309
264inline void 310INLINE void
265encode_indent (enc_t *enc) 311encode_indent (enc_t *enc)
266{ 312{
267 if (enc->json.flags & F_INDENT) 313 if (enc->json.flags & F_INDENT)
268 { 314 {
269 int spaces = enc->indent * INDENT_STEP; 315 int spaces = enc->indent * INDENT_STEP;
… …
272 memset (enc->cur, ' ', spaces); 318 memset (enc->cur, ' ', spaces);
273 enc->cur += spaces; 319 enc->cur += spaces;
274 } 320 }
275} 321}
276 322
277inline void 323INLINE void
278encode_space (enc_t *enc) 324encode_space (enc_t *enc)
279{ 325{
280 need (enc, 1); 326 need (enc, 1);
281 encode_ch (enc, ' '); 327 encode_ch (enc, ' ');
282} 328}
283 329
284inline void 330INLINE void
285encode_nl (enc_t *enc) 331encode_nl (enc_t *enc)
286{ 332{
287 if (enc->json.flags & F_INDENT) 333 if (enc->json.flags & F_INDENT)
288 { 334 {
289 need (enc, 1); 335 need (enc, 1);
290 encode_ch (enc, '\n'); 336 encode_ch (enc, '\n');
291 } 337 }
292} 338}
293 339
294inline void 340INLINE void
295encode_comma (enc_t *enc) 341encode_comma (enc_t *enc)
296{ 342{
297 encode_ch (enc, ','); 343 encode_ch (enc, ',');
298 344
299 if (enc->json.flags & F_INDENT) 345 if (enc->json.flags & F_INDENT)
… …
310 int i, len = av_len (av); 356 int i, len = av_len (av);
311 357
312 if (enc->indent >= enc->maxdepth) 358 if (enc->indent >= enc->maxdepth)
313 croak ("data structure too deep (hit recursion limit)"); 359 croak ("data structure too deep (hit recursion limit)");
314 360
315 encode_ch (enc, '['); encode_nl (enc); 361 encode_ch (enc, '[');
316 ++enc->indent; 362
363 if (len >= 0)
364 {
365 encode_nl (enc); ++enc->indent;
317 366
318 for (i = 0; i <= len; ++i) 367 for (i = 0; i <= len; ++i)
319 { 368 {
320 SV **svp = av_fetch (av, i, 0); 369 SV **svp = av_fetch (av, i, 0);
321 370
322 encode_indent (enc); 371 encode_indent (enc);
323 372
324 if (svp) 373 if (svp)
325 encode_sv (enc, *svp); 374 encode_sv (enc, *svp);
326 else 375 else
327 encode_str (enc, "null", 4, 0); 376 encode_str (enc, "null", 4, 0);
328 377
329 if (i < len) 378 if (i < len)
330 encode_comma (enc); 379 encode_comma (enc);
331 } 380 }
332 381
382 encode_nl (enc); --enc->indent; encode_indent (enc);
383 }
384
333 encode_nl (enc); 385 encode_ch (enc, ']');
334
335 --enc->indent;
336 encode_indent (enc); encode_ch (enc, ']');
337} 386}
338 387
339static void 388static void
340encode_hk (enc_t *enc, HE *he) 389encode_hk (enc_t *enc, HE *he)
341{ 390{
… …
394 int count; 443 int count;
395 444
396 if (enc->indent >= enc->maxdepth) 445 if (enc->indent >= enc->maxdepth)
397 croak ("data structure too deep (hit recursion limit)"); 446 croak ("data structure too deep (hit recursion limit)");
398 447
399 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; 448 encode_ch (enc, '{');
400 449
401 // for canonical output we have to sort by keys first 450 // for canonical output we have to sort by keys first
402 // actually, this is mostly due to the stupid so-called 451 // actually, this is mostly due to the stupid so-called
403 // security workaround added somewhere in 5.8.x. 452 // security workaround added somewhere in 5.8.x.
404 // that randomises hash orderings 453 // that randomises hash orderings
… …
457 506
458 FREETMPS; 507 FREETMPS;
459 LEAVE; 508 LEAVE;
460 } 509 }
461 510
511 encode_nl (enc); ++enc->indent;
512
462 while (count--) 513 while (count--)
463 { 514 {
464 encode_indent (enc); 515 encode_indent (enc);
465 he = hes [count]; 516 he = hes [count];
466 encode_hk (enc, he); 517 encode_hk (enc, he);
467 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); 518 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
468 519
469 if (count) 520 if (count)
470 encode_comma (enc); 521 encode_comma (enc);
471 } 522 }
523
524 encode_nl (enc); --enc->indent; encode_indent (enc);
472 } 525 }
473 } 526 }
474 else 527 else
475 { 528 {
476 if (hv_iterinit (hv) || SvMAGICAL (hv)) 529 if (hv_iterinit (hv) || SvMAGICAL (hv))
477 if ((he = hv_iternext (hv))) 530 if ((he = hv_iternext (hv)))
531 {
532 encode_nl (enc); ++enc->indent;
533
478 for (;;) 534 for (;;)
479 { 535 {
480 encode_indent (enc); 536 encode_indent (enc);
481 encode_hk (enc, he); 537 encode_hk (enc, he);
482 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); 538 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
483 539
484 if (!(he = hv_iternext (hv))) 540 if (!(he = hv_iternext (hv)))
485 break; 541 break;
486 542
487 encode_comma (enc); 543 encode_comma (enc);
488 } 544 }
489 }
490 545
546 encode_nl (enc); --enc->indent; encode_indent (enc);
547 }
548 }
549
491 encode_nl (enc); 550 encode_ch (enc, '}');
492
493 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
494} 551}
495 552
496// encode objects, arrays and special \0=false and \1=true values. 553// encode objects, arrays and special \0=false and \1=true values.
497static void 554static void
498encode_rv (enc_t *enc, SV *sv) 555encode_rv (enc_t *enc, SV *sv)
… …
606 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 663 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
607 enc->cur += strlen (enc->cur); 664 enc->cur += strlen (enc->cur);
608 } 665 }
609 else if (SvIOKp (sv)) 666 else if (SvIOKp (sv))
610 { 667 {
611 // we assume we can always read an IV as a UV 668 // we assume we can always read an IV as a UV and vice versa
612 if (SvUV (sv) & ~(UV)0x7fff) 669 // we assume two's complement
613 { 670 // we assume no aliasing issues in the union
614 // large integer, use the (rather slow) snprintf way. 671 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
615 need (enc, sizeof (UV) * 3); 672 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
616 enc->cur +=
617 SvIsUV(sv)
618 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
619 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
620 }
621 else
622 { 673 {
623 // optimise the "small number case" 674 // optimise the "small number case"
624 // code will likely be branchless and use only a single multiplication 675 // code will likely be branchless and use only a single multiplication
676 // works for numbers up to 59074
625 I32 i = SvIV (sv); 677 I32 i = SvIVX (sv);
626 U32 u; 678 U32 u;
627 char digit, nz = 0; 679 char digit, nz = 0;
628 680
629 need (enc, 6); 681 need (enc, 6);
630 682
… …
636 688
637 // now output digit by digit, each time masking out the integer part 689 // now output digit by digit, each time masking out the integer part
638 // and multiplying by 5 while moving the decimal point one to the right, 690 // and multiplying by 5 while moving the decimal point one to the right,
639 // resulting in a net multiplication by 10. 691 // resulting in a net multiplication by 10.
640 // we always write the digit to memory but conditionally increment 692 // we always write the digit to memory but conditionally increment
641 // the pointer, to ease the usage of conditional move instructions. 693 // the pointer, to enable the use of conditional move instructions.
642 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; 694 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
643 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; 695 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
644 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; 696 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
645 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; 697 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
646 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' 698 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
699 }
700 else
701 {
702 // large integer, use the (rather slow) snprintf way.
703 need (enc, IVUV_MAXCHARS);
704 enc->cur +=
705 SvIsUV(sv)
706 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
707 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
647 } 708 }
648 } 709 }
649 else if (SvROK (sv)) 710 else if (SvROK (sv))
650 encode_rv (enc, SvRV (sv)); 711 encode_rv (enc, SvRV (sv));
651 else if (!SvOK (sv)) 712 else if (!SvOK (sv))
… …
667 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 728 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
668 enc.cur = SvPVX (enc.sv); 729 enc.cur = SvPVX (enc.sv);
669 enc.end = SvEND (enc.sv); 730 enc.end = SvEND (enc.sv);
670 enc.indent = 0; 731 enc.indent = 0;
671 enc.maxdepth = DEC_DEPTH (enc.json.flags); 732 enc.maxdepth = DEC_DEPTH (enc.json.flags);
733 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
734 : enc.json.flags & F_LATIN1 ? 0x000100UL
735 : 0x110000UL;
672 736
673 SvPOK_only (enc.sv); 737 SvPOK_only (enc.sv);
674 encode_sv (&enc, scalar); 738 encode_sv (&enc, scalar);
675 739
676 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 740 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
… …
697 JSON json; 761 JSON json;
698 U32 depth; // recursion depth 762 U32 depth; // recursion depth
699 U32 maxdepth; // recursion depth limit 763 U32 maxdepth; // recursion depth limit
700} dec_t; 764} dec_t;
701 765
702inline void 766INLINE void
767decode_comment (dec_t *dec)
768{
769 // only '#'-style comments allowed a.t.m.
770
771 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
772 ++dec->cur;
773}
774
775INLINE void
703decode_ws (dec_t *dec) 776decode_ws (dec_t *dec)
704{ 777{
705 for (;;) 778 for (;;)
706 { 779 {
707 char ch = *dec->cur; 780 char ch = *dec->cur;
708 781
709 if (ch > 0x20 782 if (ch > 0x20)
783 {
784 if (expect_false (ch == '#'))
785 {
786 if (dec->json.flags & F_RELAXED)
787 decode_comment (dec);
788 else
789 break;
790 }
791 else
792 break;
793 }
710 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) 794 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
711 break; 795 break; // parse error, but let higher level handle it, gives better error messages
712 796
713 ++dec->cur; 797 ++dec->cur;
714 } 798 }
715} 799}
716 800
… …
822 906
823 if (hi >= 0x80) 907 if (hi >= 0x80)
824 { 908 {
825 utf8 = 1; 909 utf8 = 1;
826 910
827 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); 911 cur = encode_utf8 (cur, hi);
828 } 912 }
829 else 913 else
830 *cur++ = hi; 914 *cur++ = hi;
831 } 915 }
832 break; 916 break;
… …
834 default: 918 default:
835 --dec_cur; 919 --dec_cur;
836 ERR ("illegal backslash escape sequence in string"); 920 ERR ("illegal backslash escape sequence in string");
837 } 921 }
838 } 922 }
839 else if (expect_true (ch >= 0x20 && ch <= 0x7f)) 923 else if (expect_true (ch >= 0x20 && ch < 0x80))
840 *cur++ = ch; 924 *cur++ = ch;
841 else if (ch >= 0x80) 925 else if (ch >= 0x80)
842 { 926 {
843 STRLEN clen; 927 STRLEN clen;
844 UV uch; 928 UV uch;
… …
967 1051
968 if (!is_nv) 1052 if (!is_nv)
969 { 1053 {
970 int len = dec->cur - start; 1054 int len = dec->cur - start;
971 1055
972 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so 1056 // special case the rather common 1..5-digit-int case
973 if (*start == '-') 1057 if (*start == '-')
974 switch (len) 1058 switch (len)
975 { 1059 {
976 case 2: return newSViv (-( start [1] - '0' * 1)); 1060 case 2: return newSViv (-( start [1] - '0' * 1));
977 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1061 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
978 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1062 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
979 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1063 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1064 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
980 } 1065 }
981 else 1066 else
982 switch (len) 1067 switch (len)
983 { 1068 {
984 case 1: return newSViv ( start [0] - '0' * 1); 1069 case 1: return newSViv ( start [0] - '0' * 1);
985 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1070 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
986 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1071 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
987 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1072 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1073 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
988 } 1074 }
989 1075
990 { 1076 {
991 UV uv; 1077 UV uv;
992 int numtype = grok_number (start, len, &uv); 1078 int numtype = grok_number (start, len, &uv);
… …
1053 1139
1054 if (*dec->cur != ',') 1140 if (*dec->cur != ',')
1055 ERR (", or ] expected while parsing array"); 1141 ERR (", or ] expected while parsing array");
1056 1142
1057 ++dec->cur; 1143 ++dec->cur;
1144
1145 decode_ws (dec);
1146
1147 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1148 {
1149 ++dec->cur;
1150 break;
1151 }
1058 } 1152 }
1059 1153
1060 DEC_DEC_DEPTH; 1154 DEC_DEC_DEPTH;
1061 return newRV_noinc ((SV *)av); 1155 return newRV_noinc ((SV *)av);
1062 1156
… …
1078 if (*dec->cur == '}') 1172 if (*dec->cur == '}')
1079 ++dec->cur; 1173 ++dec->cur;
1080 else 1174 else
1081 for (;;) 1175 for (;;)
1082 { 1176 {
1083 decode_ws (dec); EXPECT_CH ('"'); 1177 EXPECT_CH ('"');
1084 1178
1085 // heuristic: assume that 1179 // heuristic: assume that
1086 // a) decode_str + hv_store_ent are abysmally slow. 1180 // a) decode_str + hv_store_ent are abysmally slow.
1087 // b) most hash keys are short, simple ascii text. 1181 // b) most hash keys are short, simple ascii text.
1088 // => try to "fast-match" such strings to avoid 1182 // => try to "fast-match" such strings to avoid
… …
1092 char *p = dec->cur; 1186 char *p = dec->cur;
1093 char *e = p + 24; // only try up to 24 bytes 1187 char *e = p + 24; // only try up to 24 bytes
1094 1188
1095 for (;;) 1189 for (;;)
1096 { 1190 {
1097 // the >= 0x80 is true on most architectures 1191 // the >= 0x80 is false on most architectures
1098 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1192 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1099 { 1193 {
1100 // slow path, back up and use decode_str 1194 // slow path, back up and use decode_str
1101 SV *key = decode_str (dec); 1195 SV *key = decode_str (dec);
1102 if (!key) 1196 if (!key)
1103 goto fail; 1197 goto fail;
1104 1198
1105 decode_ws (dec); EXPECT_CH (':'); 1199 decode_ws (dec); EXPECT_CH (':');
1106 1200
1201 decode_ws (dec);
1107 value = decode_sv (dec); 1202 value = decode_sv (dec);
1108 if (!value) 1203 if (!value)
1109 { 1204 {
1110 SvREFCNT_dec (key); 1205 SvREFCNT_dec (key);
1111 goto fail; 1206 goto fail;
… …
1123 int len = p - key; 1218 int len = p - key;
1124 dec->cur = p + 1; 1219 dec->cur = p + 1;
1125 1220
1126 decode_ws (dec); EXPECT_CH (':'); 1221 decode_ws (dec); EXPECT_CH (':');
1127 1222
1223 decode_ws (dec);
1128 value = decode_sv (dec); 1224 value = decode_sv (dec);
1129 if (!value) 1225 if (!value)
1130 goto fail; 1226 goto fail;
1131 1227
1132 hv_store (hv, key, len, value, 0); 1228 hv_store (hv, key, len, value, 0);
… …
1148 1244
1149 if (*dec->cur != ',') 1245 if (*dec->cur != ',')
1150 ERR (", or } expected while parsing object/hash"); 1246 ERR (", or } expected while parsing object/hash");
1151 1247
1152 ++dec->cur; 1248 ++dec->cur;
1249
1250 decode_ws (dec);
1251
1252 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1253 {
1254 ++dec->cur;
1255 break;
1256 }
1153 } 1257 }
1154 1258
1155 DEC_DEC_DEPTH; 1259 DEC_DEC_DEPTH;
1156 sv = newRV_noinc ((SV *)hv); 1260 sv = newRV_noinc ((SV *)hv);
1157 1261
… …
1222} 1326}
1223 1327
1224static SV * 1328static SV *
1225decode_sv (dec_t *dec) 1329decode_sv (dec_t *dec)
1226{ 1330{
1227 decode_ws (dec);
1228
1229 // the beauty of JSON: you need exactly one character lookahead 1331 // the beauty of JSON: you need exactly one character lookahead
1230 // to parse anything. 1332 // to parse everything.
1231 switch (*dec->cur) 1333 switch (*dec->cur)
1232 { 1334 {
1233 case '"': ++dec->cur; return decode_str (dec); 1335 case '"': ++dec->cur; return decode_str (dec);
1234 case '[': ++dec->cur; return decode_av (dec); 1336 case '[': ++dec->cur; return decode_av (dec);
1235 case '{': ++dec->cur; return decode_hv (dec); 1337 case '{': ++dec->cur; return decode_hv (dec);
1236 1338
1237 case '-': 1339 case '-':
1238 case '0': case '1': case '2': case '3': case '4': 1340 case '0': case '1': case '2': case '3': case '4':
1239 case '5': case '6': case '7': case '8': case '9': 1341 case '5': case '6': case '7': case '8': case '9':
1240 return decode_num (dec); 1342 return decode_num (dec);
… …
1286fail: 1388fail:
1287 return 0; 1389 return 0;
1288} 1390}
1289 1391
1290static SV * 1392static SV *
1291decode_json (SV *string, JSON *json, UV *offset_return) 1393decode_json (SV *string, JSON *json, STRLEN *offset_return)
1292{ 1394{
1293 dec_t dec; 1395 dec_t dec;
1294 UV offset; 1396 STRLEN offset;
1295 SV *sv; 1397 SV *sv;
1296 1398
1297 SvGETMAGIC (string); 1399 SvGETMAGIC (string);
1298 SvUPGRADE (string, SVt_PV); 1400 SvUPGRADE (string, SVt_PV);
1299 1401
… …
1317 1419
1318 if (dec.json.cb_object || dec.json.cb_sk_object) 1420 if (dec.json.cb_object || dec.json.cb_sk_object)
1319 dec.json.flags |= F_HOOK; 1421 dec.json.flags |= F_HOOK;
1320 1422
1321 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1423 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1424
1425 decode_ws (&dec);
1322 sv = decode_sv (&dec); 1426 sv = decode_sv (&dec);
1323 1427
1324 if (!(offset_return || !sv)) 1428 if (!(offset_return || !sv))
1325 { 1429 {
1326 // check for trailing garbage 1430 // check for trailing garbage
… …
1370 1474
1371 return sv; 1475 return sv;
1372} 1476}
1373 1477
1374///////////////////////////////////////////////////////////////////////////// 1478/////////////////////////////////////////////////////////////////////////////
1479// incremental parser
1480
1481static void
1482incr_parse (JSON *self)
1483{
1484 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1485
1486 for (;;)
1487 {
1488 //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
1489 switch (self->incr_mode)
1490 {
1491 // only used for intiial whitespace skipping
1492 case INCR_M_WS:
1493 for (;;)
1494 {
1495 if (*p > 0x20)
1496 {
1497 self->incr_mode = INCR_M_JSON;
1498 goto incr_m_json;
1499 }
1500 else if (!*p)
1501 goto interrupt;
1502
1503 ++p;
1504 }
1505
1506 // skip a single char inside a string (for \\-processing)
1507 case INCR_M_BS:
1508 if (!*p)
1509 goto interrupt;
1510
1511 ++p;
1512 self->incr_mode = INCR_M_STR;
1513 goto incr_m_str;
1514
1515 // inside a string
1516 case INCR_M_STR:
1517 incr_m_str:
1518 for (;;)
1519 {
1520 if (*p == '"')
1521 {
1522 ++p;
1523 self->incr_mode = INCR_M_JSON;
1524
1525 if (!self->incr_nest)
1526 goto interrupt;
1527
1528 goto incr_m_json;
1529 }
1530 else if (*p == '\\')
1531 {
1532 ++p; // "virtually" consumes character after \
1533
1534 if (!*p) // if at end of string we have to switch modes
1535 {
1536 self->incr_mode = INCR_M_BS;
1537 goto interrupt;
1538 }
1539 }
1540 else if (!*p)
1541 goto interrupt;
1542
1543 ++p;
1544 }
1545
1546 // after initial ws, outside string
1547 case INCR_M_JSON:
1548 incr_m_json:
1549 for (;;)
1550 {
1551 switch (*p++)
1552 {
1553 case 0:
1554 --p;
1555 goto interrupt;
1556
1557 case 0x09:
1558 case 0x0a:
1559 case 0x0d:
1560 case 0x20:
1561 if (!self->incr_nest)
1562 {
1563 --p; // do not eat the whitespace, let the next round do it
1564 goto interrupt;
1565 }
1566 break;
1567
1568 case '"':
1569 self->incr_mode = INCR_M_STR;
1570 goto incr_m_str;
1571
1572 case '[':
1573 case '{':
1574 ++self->incr_nest;
1575 break;
1576
1577 case ']':
1578 case '}':
1579 if (!--self->incr_nest)
1580 goto interrupt;
1581 }
1582 }
1583 }
1584
1585 modechange:
1586 ;
1587 }
1588
1589interrupt:
1590 self->incr_pos = p - SvPVX (self->incr_text);
1591 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1592}
1593
1594/////////////////////////////////////////////////////////////////////////////
1375// XS interface functions 1595// XS interface functions
1376 1596
1377MODULE = JSON::XS PACKAGE = JSON::XS 1597MODULE = JSON::XS PACKAGE = JSON::XS
1378 1598
1379BOOT: 1599BOOT:
… …
1406{ 1626{
1407 SV *pv = NEWSV (0, sizeof (JSON)); 1627 SV *pv = NEWSV (0, sizeof (JSON));
1408 SvPOK_only (pv); 1628 SvPOK_only (pv);
1409 Zero (SvPVX (pv), 1, JSON); 1629 Zero (SvPVX (pv), 1, JSON);
1410 ((JSON *)SvPVX (pv))->flags = F_DEFAULT; 1630 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1411 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); 1631 XPUSHs (sv_2mortal (sv_bless (
1632 newRV_noinc (pv),
1633 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1634 )));
1412} 1635}
1413 1636
1414void ascii (JSON *self, int enable = 1) 1637void ascii (JSON *self, int enable = 1)
1415 ALIAS: 1638 ALIAS:
1416 ascii = F_ASCII 1639 ascii = F_ASCII
… …
1423 pretty = F_PRETTY 1646 pretty = F_PRETTY
1424 allow_nonref = F_ALLOW_NONREF 1647 allow_nonref = F_ALLOW_NONREF
1425 shrink = F_SHRINK 1648 shrink = F_SHRINK
1426 allow_blessed = F_ALLOW_BLESSED 1649 allow_blessed = F_ALLOW_BLESSED
1427 convert_blessed = F_CONV_BLESSED 1650 convert_blessed = F_CONV_BLESSED
1651 relaxed = F_RELAXED
1428 PPCODE: 1652 PPCODE:
1429{ 1653{
1430 if (enable) 1654 if (enable)
1431 self->flags |= ix; 1655 self->flags |= ix;
1432 else 1656 else
1433 self->flags &= ~ix; 1657 self->flags &= ~ix;
1434 1658
1435 XPUSHs (ST (0)); 1659 XPUSHs (ST (0));
1436} 1660}
1437 1661
1662void get_ascii (JSON *self)
1663 ALIAS:
1664 get_ascii = F_ASCII
1665 get_latin1 = F_LATIN1
1666 get_utf8 = F_UTF8
1667 get_indent = F_INDENT
1668 get_canonical = F_CANONICAL
1669 get_space_before = F_SPACE_BEFORE
1670 get_space_after = F_SPACE_AFTER
1671 get_allow_nonref = F_ALLOW_NONREF
1672 get_shrink = F_SHRINK
1673 get_allow_blessed = F_ALLOW_BLESSED
1674 get_convert_blessed = F_CONV_BLESSED
1675 get_relaxed = F_RELAXED
1676 PPCODE:
1677 XPUSHs (boolSV (self->flags & ix));
1678
1438void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1679void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1439 PPCODE: 1680 PPCODE:
1440{ 1681{
1441 UV log2 = 0; 1682 UV log2 = 0;
1442 1683
… …
1448 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); 1689 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1449 1690
1450 XPUSHs (ST (0)); 1691 XPUSHs (ST (0));
1451} 1692}
1452 1693
1694U32 get_max_depth (JSON *self)
1695 CODE:
1696 RETVAL = DEC_DEPTH (self->flags);
1697 OUTPUT:
1698 RETVAL
1699
1453void max_size (JSON *self, UV max_size = 0) 1700void max_size (JSON *self, UV max_size = 0)
1454 PPCODE: 1701 PPCODE:
1455{ 1702{
1456 UV log2 = 0; 1703 UV log2 = 0;
1457 1704
… …
1463 1710
1464 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); 1711 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1465 1712
1466 XPUSHs (ST (0)); 1713 XPUSHs (ST (0));
1467} 1714}
1715
1716int get_max_size (JSON *self)
1717 CODE:
1718 RETVAL = DEC_SIZE (self->flags);
1719 OUTPUT:
1720 RETVAL
1468 1721
1469void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 1722void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1470 PPCODE: 1723 PPCODE:
1471{ 1724{
1472 SvREFCNT_dec (self->cb_object); 1725 SvREFCNT_dec (self->cb_object);
… …
1506 XPUSHs (decode_json (jsonstr, self, 0)); 1759 XPUSHs (decode_json (jsonstr, self, 0));
1507 1760
1508void decode_prefix (JSON *self, SV *jsonstr) 1761void decode_prefix (JSON *self, SV *jsonstr)
1509 PPCODE: 1762 PPCODE:
1510{ 1763{
1511 UV offset; 1764 STRLEN offset;
1512 EXTEND (SP, 2); 1765 EXTEND (SP, 2);
1513 PUSHs (decode_json (jsonstr, self, &offset)); 1766 PUSHs (decode_json (jsonstr, self, &offset));
1514 PUSHs (sv_2mortal (newSVuv (offset))); 1767 PUSHs (sv_2mortal (newSVuv (offset)));
1515} 1768}
1769
1770void incr_parse (JSON *self, SV *jsonstr = 0)
1771 PPCODE:
1772{
1773 if (!self->incr_text)
1774 self->incr_text = newSVpvn ("", 0);
1775
1776 // append data, if any
1777 if (jsonstr)
1778 {
1779 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
1780 {
1781 /* utf-8-ness differs, need to upgrade */
1782 sv_utf8_upgrade (self->incr_text);
1783
1784 if (self->incr_pos)
1785 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1786 - (U8 *)SvPVX (self->incr_text);
1787 }
1788
1789 {
1790 STRLEN len;
1791 const char *str = SvPV (jsonstr, len);
1792 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
1793 Move (str, SvEND (self->incr_text), len, char);
1794 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1795 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1796 }
1797 }
1798
1799 if (GIMME_V != G_VOID)
1800 do
1801 {
1802 STRLEN offset;
1803
1804 incr_parse (self);
1805
1806 if (!INCR_DONE (self))
1807 break;
1808
1809 XPUSHs (decode_json (self->incr_text, self, &offset));
1810
1811 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1812 self->incr_pos -= offset;
1813 self->incr_nest = 0;
1814 self->incr_mode = 0;
1815 }
1816 while (GIMME_V == G_ARRAY);
1817}
1818
1819SV *incr_text (JSON *self)
1820 ATTRS: lvalue
1821 CODE:
1822{
1823 if (self->incr_pos)
1824 croak ("incr_text can only be called after a successful incr_parse call in scalar context %d", self->incr_pos);//D
1825
1826 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1827}
1828 OUTPUT:
1829 RETVAL
1516 1830
1517void DESTROY (JSON *self) 1831void DESTROY (JSON *self)
1518 CODE: 1832 CODE:
1519 SvREFCNT_dec (self->cb_sk_object); 1833 SvREFCNT_dec (self->cb_sk_object);
1520 SvREFCNT_dec (self->cb_object); 1834 SvREFCNT_dec (self->cb_object);
1835 SvREFCNT_dec (self->incr_text);
1521 1836
1522PROTOTYPES: ENABLE 1837PROTOTYPES: ENABLE
1523 1838
1524void to_json (SV *scalar) 1839void encode_json (SV *scalar)
1840 ALIAS:
1841 to_json_ = 0
1842 encode_json = F_UTF8
1525 PPCODE: 1843 PPCODE:
1526{ 1844{
1527 JSON json = { F_DEFAULT | F_UTF8 }; 1845 JSON json = { F_DEFAULT | ix };
1528 XPUSHs (encode_json (scalar, &json)); 1846 XPUSHs (encode_json (scalar, &json));
1529} 1847}
1530 1848
1531void from_json (SV *jsonstr) 1849void decode_json (SV *jsonstr)
1850 ALIAS:
1851 from_json_ = 0
1852 decode_json = F_UTF8
1532 PPCODE: 1853 PPCODE:
1533{ 1854{
1534 JSON json = { F_DEFAULT | F_UTF8 }; 1855 JSON json = { F_DEFAULT | ix };
1535 XPUSHs (decode_json (jsonstr, &json, 0)); 1856 XPUSHs (decode_json (jsonstr, &json, 0));
1536} 1857}
1537 1858
1859

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines