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 1.63 by root, Mon Aug 27 01:49:01 2007 UTC vs.
Revision 1.83 by root, Wed Mar 26 22:54:38 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
… …
50 53
51#define SB do { 54#define SB do {
52#define SE } while (0) 55#define SE } while (0)
53 56
54#if __GNUC__ >= 3 57#if __GNUC__ >= 3
55# define expect(expr,value) __builtin_expect ((expr),(value)) 58# define expect(expr,value) __builtin_expect ((expr), (value))
56# define inline inline 59# define INLINE static inline
57#else 60#else
58# define expect(expr,value) (expr) 61# define expect(expr,value) (expr)
59# define inline static 62# define INLINE static
60#endif 63#endif
61 64
62#define expect_false(expr) expect ((expr) != 0, 0) 65#define expect_false(expr) expect ((expr) != 0, 0)
63#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)))
64 71
65#ifdef USE_ITHREADS 72#ifdef USE_ITHREADS
66# define JSON_SLOW 1 73# define JSON_SLOW 1
67# 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))
68#else 75#else
… …
71#endif 78#endif
72 79
73static HV *json_stash, *json_boolean_stash; // JSON::XS:: 80static HV *json_stash, *json_boolean_stash; // JSON::XS::
74static SV *json_true, *json_false; 81static SV *json_true, *json_false;
75 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
76typedef struct { 92typedef struct {
77 U32 flags; 93 U32 flags;
78 SV *cb_object; 94 SV *cb_object;
79 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;
80} JSON; 102} JSON;
81 103
82///////////////////////////////////////////////////////////////////////////// 104/////////////////////////////////////////////////////////////////////////////
83// utility functions 105// utility functions
84 106
85inline void 107INLINE SV *
108get_bool (const char *name)
109{
110 SV *sv = get_sv (name, 1);
111
112 SvREADONLY_on (sv);
113 SvREADONLY_on (SvRV (sv));
114
115 return sv;
116}
117
118INLINE void
86shrink (SV *sv) 119shrink (SV *sv)
87{ 120{
88 sv_utf8_downgrade (sv, 1); 121 sv_utf8_downgrade (sv, 1);
89 if (SvLEN (sv) > SvCUR (sv) + 1) 122 if (SvLEN (sv) > SvCUR (sv) + 1)
90 { 123 {
… …
99// decode an utf-8 character and return it, or (UV)-1 in 132// decode an utf-8 character and return it, or (UV)-1 in
100// case of an error. 133// case of an error.
101// we special-case "safe" characters from U+80 .. U+7FF, 134// we special-case "safe" characters from U+80 .. U+7FF,
102// but use the very good perl function to parse anything else. 135// but use the very good perl function to parse anything else.
103// note that we never call this function for a ascii codepoints 136// note that we never call this function for a ascii codepoints
104inline UV 137INLINE UV
105decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 138decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
106{ 139{
107 if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) 140 if (expect_true (len >= 2
108 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 141 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 142 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110 { 143 {
111 *clen = 2; 144 *clen = 2;
112 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 145 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
113 } 146 }
114 else 147 else
115 { 148 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116 *clen = (STRLEN)-1; 149}
117 return (UV)-1; 150
118 } 151// likewise for encoding, also never called for ascii codepoints
152// this function takes advantage of this fact, although current gccs
153// seem to optimise the check for >= 0x80 away anyways
154INLINE unsigned char *
155encode_utf8 (unsigned char *s, UV ch)
156{
157 if (expect_false (ch < 0x000080))
158 *s++ = ch;
159 else if (expect_true (ch < 0x000800))
160 *s++ = 0xc0 | ( ch >> 6),
161 *s++ = 0x80 | ( ch & 0x3f);
162 else if ( ch < 0x010000)
163 *s++ = 0xe0 | ( ch >> 12),
164 *s++ = 0x80 | ((ch >> 6) & 0x3f),
165 *s++ = 0x80 | ( ch & 0x3f);
166 else if ( ch < 0x110000)
167 *s++ = 0xf0 | ( ch >> 18),
168 *s++ = 0x80 | ((ch >> 12) & 0x3f),
169 *s++ = 0x80 | ((ch >> 6) & 0x3f),
170 *s++ = 0x80 | ( ch & 0x3f);
171
172 return s;
119} 173}
120 174
121///////////////////////////////////////////////////////////////////////////// 175/////////////////////////////////////////////////////////////////////////////
122// encoder 176// encoder
123 177
… …
128 char *end; // SvEND (sv) 182 char *end; // SvEND (sv)
129 SV *sv; // result scalar 183 SV *sv; // result scalar
130 JSON json; 184 JSON json;
131 U32 indent; // indentation level 185 U32 indent; // indentation level
132 U32 maxdepth; // max. indentation/recursion level 186 U32 maxdepth; // max. indentation/recursion level
187 UV limit; // escape character values >= this value when encoding
133} enc_t; 188} enc_t;
134 189
135inline void 190INLINE void
136need (enc_t *enc, STRLEN len) 191need (enc_t *enc, STRLEN len)
137{ 192{
138 if (expect_false (enc->cur + len >= enc->end)) 193 if (expect_false (enc->cur + len >= enc->end))
139 { 194 {
140 STRLEN cur = enc->cur - SvPVX (enc->sv); 195 STRLEN cur = enc->cur - SvPVX (enc->sv);
… …
142 enc->cur = SvPVX (enc->sv) + cur; 197 enc->cur = SvPVX (enc->sv) + cur;
143 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 198 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144 } 199 }
145} 200}
146 201
147inline void 202INLINE void
148encode_ch (enc_t *enc, char ch) 203encode_ch (enc_t *enc, char ch)
149{ 204{
150 need (enc, 1); 205 need (enc, 1);
151 *enc->cur++ = ch; 206 *enc->cur++ = ch;
152} 207}
… …
206 { 261 {
207 uch = ch; 262 uch = ch;
208 clen = 1; 263 clen = 1;
209 } 264 }
210 265
211 if (uch > 0x10FFFFUL) 266 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
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 { 267 {
216 if (uch > 0xFFFFUL) 268 if (uch >= 0x10000UL)
217 { 269 {
270 if (uch >= 0x110000UL)
271 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
272
218 need (enc, len += 11); 273 need (enc, len += 11);
219 sprintf (enc->cur, "\\u%04x\\u%04x", 274 sprintf (enc->cur, "\\u%04x\\u%04x",
220 (int)((uch - 0x10000) / 0x400 + 0xD800), 275 (int)((uch - 0x10000) / 0x400 + 0xD800),
221 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 276 (int)((uch - 0x10000) % 0x400 + 0xDC00));
222 enc->cur += 12; 277 enc->cur += 12;
… …
250 while (--clen); 305 while (--clen);
251 } 306 }
252 else 307 else
253 { 308 {
254 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 309 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 310 enc->cur = encode_utf8 (enc->cur, uch);
256 ++str; 311 ++str;
257 } 312 }
258 } 313 }
259 } 314 }
260 } 315 }
261 316
262 --len; 317 --len;
263 } 318 }
264} 319}
265 320
266inline void 321INLINE void
267encode_indent (enc_t *enc) 322encode_indent (enc_t *enc)
268{ 323{
269 if (enc->json.flags & F_INDENT) 324 if (enc->json.flags & F_INDENT)
270 { 325 {
271 int spaces = enc->indent * INDENT_STEP; 326 int spaces = enc->indent * INDENT_STEP;
… …
274 memset (enc->cur, ' ', spaces); 329 memset (enc->cur, ' ', spaces);
275 enc->cur += spaces; 330 enc->cur += spaces;
276 } 331 }
277} 332}
278 333
279inline void 334INLINE void
280encode_space (enc_t *enc) 335encode_space (enc_t *enc)
281{ 336{
282 need (enc, 1); 337 need (enc, 1);
283 encode_ch (enc, ' '); 338 encode_ch (enc, ' ');
284} 339}
285 340
286inline void 341INLINE void
287encode_nl (enc_t *enc) 342encode_nl (enc_t *enc)
288{ 343{
289 if (enc->json.flags & F_INDENT) 344 if (enc->json.flags & F_INDENT)
290 { 345 {
291 need (enc, 1); 346 need (enc, 1);
292 encode_ch (enc, '\n'); 347 encode_ch (enc, '\n');
293 } 348 }
294} 349}
295 350
296inline void 351INLINE void
297encode_comma (enc_t *enc) 352encode_comma (enc_t *enc)
298{ 353{
299 encode_ch (enc, ','); 354 encode_ch (enc, ',');
300 355
301 if (enc->json.flags & F_INDENT) 356 if (enc->json.flags & F_INDENT)
… …
312 int i, len = av_len (av); 367 int i, len = av_len (av);
313 368
314 if (enc->indent >= enc->maxdepth) 369 if (enc->indent >= enc->maxdepth)
315 croak ("data structure too deep (hit recursion limit)"); 370 croak ("data structure too deep (hit recursion limit)");
316 371
317 encode_ch (enc, '['); encode_nl (enc); 372 encode_ch (enc, '[');
318 ++enc->indent; 373
374 if (len >= 0)
375 {
376 encode_nl (enc); ++enc->indent;
319 377
320 for (i = 0; i <= len; ++i) 378 for (i = 0; i <= len; ++i)
321 { 379 {
322 SV **svp = av_fetch (av, i, 0); 380 SV **svp = av_fetch (av, i, 0);
323 381
324 encode_indent (enc); 382 encode_indent (enc);
325 383
326 if (svp) 384 if (svp)
327 encode_sv (enc, *svp); 385 encode_sv (enc, *svp);
328 else 386 else
329 encode_str (enc, "null", 4, 0); 387 encode_str (enc, "null", 4, 0);
330 388
331 if (i < len) 389 if (i < len)
332 encode_comma (enc); 390 encode_comma (enc);
333 } 391 }
334 392
393 encode_nl (enc); --enc->indent; encode_indent (enc);
394 }
395
335 encode_nl (enc); 396 encode_ch (enc, ']');
336
337 --enc->indent;
338 encode_indent (enc); encode_ch (enc, ']');
339} 397}
340 398
341static void 399static void
342encode_hk (enc_t *enc, HE *he) 400encode_hk (enc_t *enc, HE *he)
343{ 401{
… …
391 449
392static void 450static void
393encode_hv (enc_t *enc, HV *hv) 451encode_hv (enc_t *enc, HV *hv)
394{ 452{
395 HE *he; 453 HE *he;
396 int count;
397 454
398 if (enc->indent >= enc->maxdepth) 455 if (enc->indent >= enc->maxdepth)
399 croak ("data structure too deep (hit recursion limit)"); 456 croak ("data structure too deep (hit recursion limit)");
400 457
401 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; 458 encode_ch (enc, '{');
402 459
403 // for canonical output we have to sort by keys first 460 // for canonical output we have to sort by keys first
404 // actually, this is mostly due to the stupid so-called 461 // actually, this is mostly due to the stupid so-called
405 // security workaround added somewhere in 5.8.x. 462 // security workaround added somewhere in 5.8.x.
406 // that randomises hash orderings 463 // that randomises hash orderings
… …
459 516
460 FREETMPS; 517 FREETMPS;
461 LEAVE; 518 LEAVE;
462 } 519 }
463 520
521 encode_nl (enc); ++enc->indent;
522
464 while (count--) 523 while (count--)
465 { 524 {
466 encode_indent (enc); 525 encode_indent (enc);
467 he = hes [count]; 526 he = hes [count];
468 encode_hk (enc, he); 527 encode_hk (enc, he);
469 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); 528 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470 529
471 if (count) 530 if (count)
472 encode_comma (enc); 531 encode_comma (enc);
473 } 532 }
533
534 encode_nl (enc); --enc->indent; encode_indent (enc);
474 } 535 }
475 } 536 }
476 else 537 else
477 { 538 {
478 if (hv_iterinit (hv) || SvMAGICAL (hv)) 539 if (hv_iterinit (hv) || SvMAGICAL (hv))
479 if ((he = hv_iternext (hv))) 540 if ((he = hv_iternext (hv)))
541 {
542 encode_nl (enc); ++enc->indent;
543
480 for (;;) 544 for (;;)
481 { 545 {
482 encode_indent (enc); 546 encode_indent (enc);
483 encode_hk (enc, he); 547 encode_hk (enc, he);
484 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); 548 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485 549
486 if (!(he = hv_iternext (hv))) 550 if (!(he = hv_iternext (hv)))
487 break; 551 break;
488 552
489 encode_comma (enc); 553 encode_comma (enc);
490 } 554 }
491 }
492 555
556 encode_nl (enc); --enc->indent; encode_indent (enc);
557 }
558 }
559
493 encode_nl (enc); 560 encode_ch (enc, '}');
494
495 --enc->indent; encode_indent (enc); encode_ch (enc, '}');
496} 561}
497 562
498// encode objects, arrays and special \0=false and \1=true values. 563// encode objects, arrays and special \0=false and \1=true values.
499static void 564static void
500encode_rv (enc_t *enc, SV *sv) 565encode_rv (enc_t *enc, SV *sv)
… …
608 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 673 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609 enc->cur += strlen (enc->cur); 674 enc->cur += strlen (enc->cur);
610 } 675 }
611 else if (SvIOKp (sv)) 676 else if (SvIOKp (sv))
612 { 677 {
613 // we assume we can always read an IV as a UV 678 // we assume we can always read an IV as a UV and vice versa
614 if (SvUV (sv) & ~(UV)0x7fff) 679 // we assume two's complement
615 { 680 // we assume no aliasing issues in the union
616 // large integer, use the (rather slow) snprintf way. 681 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617 need (enc, sizeof (UV) * 3); 682 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
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 { 683 {
625 // optimise the "small number case" 684 // optimise the "small number case"
626 // code will likely be branchless and use only a single multiplication 685 // code will likely be branchless and use only a single multiplication
686 // works for numbers up to 59074
627 I32 i = SvIV (sv); 687 I32 i = SvIVX (sv);
628 U32 u; 688 U32 u;
629 char digit, nz = 0; 689 char digit, nz = 0;
630 690
631 need (enc, 6); 691 need (enc, 6);
632 692
… …
638 698
639 // now output digit by digit, each time masking out the integer part 699 // 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, 700 // and multiplying by 5 while moving the decimal point one to the right,
641 // resulting in a net multiplication by 10. 701 // resulting in a net multiplication by 10.
642 // we always write the digit to memory but conditionally increment 702 // we always write the digit to memory but conditionally increment
643 // the pointer, to ease the usage of conditional move instructions. 703 // the pointer, to enable the use of conditional move instructions.
644 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; 704 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
645 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; 705 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
646 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; 706 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
647 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; 707 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
648 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' 708 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
709 }
710 else
711 {
712 // large integer, use the (rather slow) snprintf way.
713 need (enc, IVUV_MAXCHARS);
714 enc->cur +=
715 SvIsUV(sv)
716 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
717 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
649 } 718 }
650 } 719 }
651 else if (SvROK (sv)) 720 else if (SvROK (sv))
652 encode_rv (enc, SvRV (sv)); 721 encode_rv (enc, SvRV (sv));
653 else if (!SvOK (sv)) 722 else if (!SvOK (sv))
… …
669 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 738 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670 enc.cur = SvPVX (enc.sv); 739 enc.cur = SvPVX (enc.sv);
671 enc.end = SvEND (enc.sv); 740 enc.end = SvEND (enc.sv);
672 enc.indent = 0; 741 enc.indent = 0;
673 enc.maxdepth = DEC_DEPTH (enc.json.flags); 742 enc.maxdepth = DEC_DEPTH (enc.json.flags);
743 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
744 : enc.json.flags & F_LATIN1 ? 0x000100UL
745 : 0x110000UL;
674 746
675 SvPOK_only (enc.sv); 747 SvPOK_only (enc.sv);
676 encode_sv (&enc, scalar); 748 encode_sv (&enc, scalar);
677 749
678 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 750 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
… …
699 JSON json; 771 JSON json;
700 U32 depth; // recursion depth 772 U32 depth; // recursion depth
701 U32 maxdepth; // recursion depth limit 773 U32 maxdepth; // recursion depth limit
702} dec_t; 774} dec_t;
703 775
704inline void 776INLINE void
777decode_comment (dec_t *dec)
778{
779 // only '#'-style comments allowed a.t.m.
780
781 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
782 ++dec->cur;
783}
784
785INLINE void
705decode_ws (dec_t *dec) 786decode_ws (dec_t *dec)
706{ 787{
707 for (;;) 788 for (;;)
708 { 789 {
709 char ch = *dec->cur; 790 char ch = *dec->cur;
710 791
711 if (ch > 0x20 792 if (ch > 0x20)
793 {
794 if (expect_false (ch == '#'))
795 {
796 if (dec->json.flags & F_RELAXED)
797 decode_comment (dec);
798 else
799 break;
800 }
801 else
802 break;
803 }
712 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) 804 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
713 break; 805 break; // parse error, but let higher level handle it, gives better error messages
714
715 if (ch == '#' && dec->json.flags & F_RELAXED)
716 ++dec->cur;
717 806
718 ++dec->cur; 807 ++dec->cur;
719 } 808 }
720} 809}
721 810
… …
827 916
828 if (hi >= 0x80) 917 if (hi >= 0x80)
829 { 918 {
830 utf8 = 1; 919 utf8 = 1;
831 920
832 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); 921 cur = encode_utf8 (cur, hi);
833 } 922 }
834 else 923 else
835 *cur++ = hi; 924 *cur++ = hi;
836 } 925 }
837 break; 926 break;
… …
839 default: 928 default:
840 --dec_cur; 929 --dec_cur;
841 ERR ("illegal backslash escape sequence in string"); 930 ERR ("illegal backslash escape sequence in string");
842 } 931 }
843 } 932 }
844 else if (expect_true (ch >= 0x20 && ch <= 0x7f)) 933 else if (expect_true (ch >= 0x20 && ch < 0x80))
845 *cur++ = ch; 934 *cur++ = ch;
846 else if (ch >= 0x80) 935 else if (ch >= 0x80)
847 { 936 {
848 STRLEN clen; 937 STRLEN clen;
849 UV uch; 938 UV uch;
… …
972 1061
973 if (!is_nv) 1062 if (!is_nv)
974 { 1063 {
975 int len = dec->cur - start; 1064 int len = dec->cur - start;
976 1065
977 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so 1066 // special case the rather common 1..5-digit-int case
978 if (*start == '-') 1067 if (*start == '-')
979 switch (len) 1068 switch (len)
980 { 1069 {
981 case 2: return newSViv (-( start [1] - '0' * 1)); 1070 case 2: return newSViv (-( start [1] - '0' * 1));
982 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1071 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
983 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1072 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
984 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1073 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1074 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
985 } 1075 }
986 else 1076 else
987 switch (len) 1077 switch (len)
988 { 1078 {
989 case 1: return newSViv ( start [0] - '0' * 1); 1079 case 1: return newSViv ( start [0] - '0' * 1);
990 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1080 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
991 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1081 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
992 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1082 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1083 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
993 } 1084 }
994 1085
995 { 1086 {
996 UV uv; 1087 UV uv;
997 int numtype = grok_number (start, len, &uv); 1088 int numtype = grok_number (start, len, &uv);
… …
1105 char *p = dec->cur; 1196 char *p = dec->cur;
1106 char *e = p + 24; // only try up to 24 bytes 1197 char *e = p + 24; // only try up to 24 bytes
1107 1198
1108 for (;;) 1199 for (;;)
1109 { 1200 {
1110 // the >= 0x80 is true on most architectures 1201 // the >= 0x80 is false on most architectures
1111 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1202 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1112 { 1203 {
1113 // slow path, back up and use decode_str 1204 // slow path, back up and use decode_str
1114 SV *key = decode_str (dec); 1205 SV *key = decode_str (dec);
1115 if (!key) 1206 if (!key)
… …
1246 1337
1247static SV * 1338static SV *
1248decode_sv (dec_t *dec) 1339decode_sv (dec_t *dec)
1249{ 1340{
1250 // the beauty of JSON: you need exactly one character lookahead 1341 // the beauty of JSON: you need exactly one character lookahead
1251 // to parse anything. 1342 // to parse everything.
1252 switch (*dec->cur) 1343 switch (*dec->cur)
1253 { 1344 {
1254 case '"': ++dec->cur; return decode_str (dec); 1345 case '"': ++dec->cur; return decode_str (dec);
1255 case '[': ++dec->cur; return decode_av (dec); 1346 case '[': ++dec->cur; return decode_av (dec);
1256 case '{': ++dec->cur; return decode_hv (dec); 1347 case '{': ++dec->cur; return decode_hv (dec);
1257 1348
1258 case '-': 1349 case '-':
1259 case '0': case '1': case '2': case '3': case '4': 1350 case '0': case '1': case '2': case '3': case '4':
1260 case '5': case '6': case '7': case '8': case '9': 1351 case '5': case '6': case '7': case '8': case '9':
1261 return decode_num (dec); 1352 return decode_num (dec);
… …
1263 case 't': 1354 case 't':
1264 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1355 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1265 { 1356 {
1266 dec->cur += 4; 1357 dec->cur += 4;
1267#if JSON_SLOW 1358#if JSON_SLOW
1268 json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); 1359 json_true = get_bool ("JSON::XS::true");
1269#endif 1360#endif
1270 return SvREFCNT_inc (json_true); 1361 return newSVsv (json_true);
1271 } 1362 }
1272 else 1363 else
1273 ERR ("'true' expected"); 1364 ERR ("'true' expected");
1274 1365
1275 break; 1366 break;
… …
1277 case 'f': 1368 case 'f':
1278 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) 1369 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1279 { 1370 {
1280 dec->cur += 5; 1371 dec->cur += 5;
1281#if JSON_SLOW 1372#if JSON_SLOW
1282 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1373 json_false = get_bool ("JSON::XS::false");
1283#endif 1374#endif
1284 return SvREFCNT_inc (json_false); 1375 return newSVsv (json_false);
1285 } 1376 }
1286 else 1377 else
1287 ERR ("'false' expected"); 1378 ERR ("'false' expected");
1288 1379
1289 break; 1380 break;
… …
1307fail: 1398fail:
1308 return 0; 1399 return 0;
1309} 1400}
1310 1401
1311static SV * 1402static SV *
1312decode_json (SV *string, JSON *json, UV *offset_return) 1403decode_json (SV *string, JSON *json, STRLEN *offset_return)
1313{ 1404{
1314 dec_t dec; 1405 dec_t dec;
1315 UV offset; 1406 STRLEN offset;
1316 SV *sv; 1407 SV *sv;
1317 1408
1318 SvGETMAGIC (string); 1409 SvGETMAGIC (string);
1319 SvUPGRADE (string, SVt_PV); 1410 SvUPGRADE (string, SVt_PV);
1320 1411
… …
1393 1484
1394 return sv; 1485 return sv;
1395} 1486}
1396 1487
1397///////////////////////////////////////////////////////////////////////////// 1488/////////////////////////////////////////////////////////////////////////////
1489// incremental parser
1490
1491static void
1492incr_parse (JSON *self)
1493{
1494 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1495
1496 for (;;)
1497 {
1498 //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
1499 switch (self->incr_mode)
1500 {
1501 // only used for intiial whitespace skipping
1502 case INCR_M_WS:
1503 for (;;)
1504 {
1505 if (*p > 0x20)
1506 {
1507 self->incr_mode = INCR_M_JSON;
1508 goto incr_m_json;
1509 }
1510 else if (!*p)
1511 goto interrupt;
1512
1513 ++p;
1514 }
1515
1516 // skip a single char inside a string (for \\-processing)
1517 case INCR_M_BS:
1518 if (!*p)
1519 goto interrupt;
1520
1521 ++p;
1522 self->incr_mode = INCR_M_STR;
1523 goto incr_m_str;
1524
1525 // inside a string
1526 case INCR_M_STR:
1527 incr_m_str:
1528 for (;;)
1529 {
1530 if (*p == '"')
1531 {
1532 ++p;
1533 self->incr_mode = INCR_M_JSON;
1534
1535 if (!self->incr_nest)
1536 goto interrupt;
1537
1538 goto incr_m_json;
1539 }
1540 else if (*p == '\\')
1541 {
1542 ++p; // "virtually" consumes character after \
1543
1544 if (!*p) // if at end of string we have to switch modes
1545 {
1546 self->incr_mode = INCR_M_BS;
1547 goto interrupt;
1548 }
1549 }
1550 else if (!*p)
1551 goto interrupt;
1552
1553 ++p;
1554 }
1555
1556 // after initial ws, outside string
1557 case INCR_M_JSON:
1558 incr_m_json:
1559 for (;;)
1560 {
1561 switch (*p++)
1562 {
1563 case 0:
1564 --p;
1565 goto interrupt;
1566
1567 case 0x09:
1568 case 0x0a:
1569 case 0x0d:
1570 case 0x20:
1571 if (!self->incr_nest)
1572 {
1573 --p; // do not eat the whitespace, let the next round do it
1574 goto interrupt;
1575 }
1576 break;
1577
1578 case '"':
1579 self->incr_mode = INCR_M_STR;
1580 goto incr_m_str;
1581
1582 case '[':
1583 case '{':
1584 ++self->incr_nest;
1585 break;
1586
1587 case ']':
1588 case '}':
1589 if (!--self->incr_nest)
1590 goto interrupt;
1591 }
1592 }
1593 }
1594
1595 modechange:
1596 ;
1597 }
1598
1599interrupt:
1600 self->incr_pos = p - SvPVX (self->incr_text);
1601 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1602}
1603
1604/////////////////////////////////////////////////////////////////////////////
1398// XS interface functions 1605// XS interface functions
1399 1606
1400MODULE = JSON::XS PACKAGE = JSON::XS 1607MODULE = JSON::XS PACKAGE = JSON::XS
1401 1608
1402BOOT: 1609BOOT:
… …
1411 : -1; 1618 : -1;
1412 1619
1413 json_stash = gv_stashpv ("JSON::XS" , 1); 1620 json_stash = gv_stashpv ("JSON::XS" , 1);
1414 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 1621 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1415 1622
1416 json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); 1623 json_true = get_bool ("JSON::XS::true");
1417 json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); 1624 json_false = get_bool ("JSON::XS::false");
1418} 1625}
1419 1626
1420PROTOTYPES: DISABLE 1627PROTOTYPES: DISABLE
1421 1628
1422void CLONE (...) 1629void CLONE (...)
… …
1429{ 1636{
1430 SV *pv = NEWSV (0, sizeof (JSON)); 1637 SV *pv = NEWSV (0, sizeof (JSON));
1431 SvPOK_only (pv); 1638 SvPOK_only (pv);
1432 Zero (SvPVX (pv), 1, JSON); 1639 Zero (SvPVX (pv), 1, JSON);
1433 ((JSON *)SvPVX (pv))->flags = F_DEFAULT; 1640 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1434 XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); 1641 XPUSHs (sv_2mortal (sv_bless (
1642 newRV_noinc (pv),
1643 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1644 )));
1435} 1645}
1436 1646
1437void ascii (JSON *self, int enable = 1) 1647void ascii (JSON *self, int enable = 1)
1438 ALIAS: 1648 ALIAS:
1439 ascii = F_ASCII 1649 ascii = F_ASCII
… …
1457 self->flags &= ~ix; 1667 self->flags &= ~ix;
1458 1668
1459 XPUSHs (ST (0)); 1669 XPUSHs (ST (0));
1460} 1670}
1461 1671
1672void get_ascii (JSON *self)
1673 ALIAS:
1674 get_ascii = F_ASCII
1675 get_latin1 = F_LATIN1
1676 get_utf8 = F_UTF8
1677 get_indent = F_INDENT
1678 get_canonical = F_CANONICAL
1679 get_space_before = F_SPACE_BEFORE
1680 get_space_after = F_SPACE_AFTER
1681 get_allow_nonref = F_ALLOW_NONREF
1682 get_shrink = F_SHRINK
1683 get_allow_blessed = F_ALLOW_BLESSED
1684 get_convert_blessed = F_CONV_BLESSED
1685 get_relaxed = F_RELAXED
1686 PPCODE:
1687 XPUSHs (boolSV (self->flags & ix));
1688
1462void max_depth (JSON *self, UV max_depth = 0x80000000UL) 1689void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1463 PPCODE: 1690 PPCODE:
1464{ 1691{
1465 UV log2 = 0; 1692 UV log2 = 0;
1466 1693
… …
1472 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); 1699 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1473 1700
1474 XPUSHs (ST (0)); 1701 XPUSHs (ST (0));
1475} 1702}
1476 1703
1704U32 get_max_depth (JSON *self)
1705 CODE:
1706 RETVAL = DEC_DEPTH (self->flags);
1707 OUTPUT:
1708 RETVAL
1709
1477void max_size (JSON *self, UV max_size = 0) 1710void max_size (JSON *self, UV max_size = 0)
1478 PPCODE: 1711 PPCODE:
1479{ 1712{
1480 UV log2 = 0; 1713 UV log2 = 0;
1481 1714
… …
1487 1720
1488 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); 1721 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1489 1722
1490 XPUSHs (ST (0)); 1723 XPUSHs (ST (0));
1491} 1724}
1725
1726int get_max_size (JSON *self)
1727 CODE:
1728 RETVAL = DEC_SIZE (self->flags);
1729 OUTPUT:
1730 RETVAL
1492 1731
1493void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 1732void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1494 PPCODE: 1733 PPCODE:
1495{ 1734{
1496 SvREFCNT_dec (self->cb_object); 1735 SvREFCNT_dec (self->cb_object);
… …
1530 XPUSHs (decode_json (jsonstr, self, 0)); 1769 XPUSHs (decode_json (jsonstr, self, 0));
1531 1770
1532void decode_prefix (JSON *self, SV *jsonstr) 1771void decode_prefix (JSON *self, SV *jsonstr)
1533 PPCODE: 1772 PPCODE:
1534{ 1773{
1535 UV offset; 1774 STRLEN offset;
1536 EXTEND (SP, 2); 1775 EXTEND (SP, 2);
1537 PUSHs (decode_json (jsonstr, self, &offset)); 1776 PUSHs (decode_json (jsonstr, self, &offset));
1538 PUSHs (sv_2mortal (newSVuv (offset))); 1777 PUSHs (sv_2mortal (newSVuv (offset)));
1778}
1779
1780void incr_parse (JSON *self, SV *jsonstr = 0)
1781 PPCODE:
1782{
1783 if (!self->incr_text)
1784 self->incr_text = newSVpvn ("", 0);
1785
1786 // append data, if any
1787 if (jsonstr)
1788 {
1789 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
1790 {
1791 /* utf-8-ness differs, need to upgrade */
1792 sv_utf8_upgrade (self->incr_text);
1793
1794 if (self->incr_pos)
1795 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1796 - (U8 *)SvPVX (self->incr_text);
1797 }
1798
1799 {
1800 STRLEN len;
1801 const char *str = SvPV (jsonstr, len);
1802 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
1803 Move (str, SvEND (self->incr_text), len, char);
1804 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1805 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1806 }
1807 }
1808
1809 if (GIMME_V != G_VOID)
1810 do
1811 {
1812 STRLEN offset;
1813
1814 if (!INCR_DONE (self))
1815 {
1816 incr_parse (self);
1817 if (!INCR_DONE (self))
1818 break;
1819 }
1820
1821 XPUSHs (decode_json (self->incr_text, self, &offset));
1822
1823 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1824 self->incr_pos -= offset;
1825 self->incr_nest = 0;
1826 self->incr_mode = 0;
1827 }
1828 while (GIMME_V == G_ARRAY);
1829}
1830
1831SV *incr_text (JSON *self)
1832 ATTRS: lvalue
1833 CODE:
1834{
1835 if (self->incr_pos)
1836 croak ("incr_text can not be called when the incremental parser already started parsing");
1837
1838 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1839}
1840 OUTPUT:
1841 RETVAL
1842
1843void incr_skip (JSON *self)
1844 CODE:
1845{
1846 if (self->incr_pos)
1847 {
1848 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
1849 self->incr_pos = 0;
1850 self->incr_nest = 0;
1851 self->incr_mode = 0;
1852 }
1539} 1853}
1540 1854
1541void DESTROY (JSON *self) 1855void DESTROY (JSON *self)
1542 CODE: 1856 CODE:
1543 SvREFCNT_dec (self->cb_sk_object); 1857 SvREFCNT_dec (self->cb_sk_object);
1544 SvREFCNT_dec (self->cb_object); 1858 SvREFCNT_dec (self->cb_object);
1859 SvREFCNT_dec (self->incr_text);
1545 1860
1546PROTOTYPES: ENABLE 1861PROTOTYPES: ENABLE
1547 1862
1548void to_json (SV *scalar) 1863void encode_json (SV *scalar)
1864 ALIAS:
1865 to_json_ = 0
1866 encode_json = F_UTF8
1549 PPCODE: 1867 PPCODE:
1550{ 1868{
1551 JSON json = { F_DEFAULT | F_UTF8 }; 1869 JSON json = { F_DEFAULT | ix };
1552 XPUSHs (encode_json (scalar, &json)); 1870 XPUSHs (encode_json (scalar, &json));
1553} 1871}
1554 1872
1555void from_json (SV *jsonstr) 1873void decode_json (SV *jsonstr)
1874 ALIAS:
1875 from_json_ = 0
1876 decode_json = F_UTF8
1556 PPCODE: 1877 PPCODE:
1557{ 1878{
1558 JSON json = { F_DEFAULT | F_UTF8 }; 1879 JSON json = { F_DEFAULT | ix };
1559 XPUSHs (decode_json (jsonstr, &json, 0)); 1880 XPUSHs (decode_json (jsonstr, &json, 0));
1560} 1881}
1561 1882
1883

Diff Legend

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