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 1.67 by root, Wed Nov 28 14:01:01 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
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 void
86shrink (SV *sv) 108shrink (SV *sv)
87{ 109{
88 sv_utf8_downgrade (sv, 1); 110 sv_utf8_downgrade (sv, 1);
89 if (SvLEN (sv) > SvCUR (sv) + 1) 111 if (SvLEN (sv) > SvCUR (sv) + 1)
90 { 112 {
99// 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
100// case of an error. 122// case of an error.
101// we special-case "safe" characters from U+80 .. U+7FF, 123// we special-case "safe" characters from U+80 .. U+7FF,
102// but use the very good perl function to parse anything else. 124// but use the very good perl function to parse anything else.
103// note that we never call this function for a ascii codepoints 125// note that we never call this function for a ascii codepoints
104inline UV 126INLINE UV
105decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 127decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
106{ 128{
107 if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) 129 if (expect_true (len >= 2
108 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 130 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 131 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110 { 132 {
111 *clen = 2; 133 *clen = 2;
112 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 134 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
113 } 135 }
114 else 136 else
115 { 137 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116 *clen = (STRLEN)-1; 138}
117 return (UV)-1; 139
118 } 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;
119} 162}
120 163
121///////////////////////////////////////////////////////////////////////////// 164/////////////////////////////////////////////////////////////////////////////
122// encoder 165// encoder
123 166
128 char *end; // SvEND (sv) 171 char *end; // SvEND (sv)
129 SV *sv; // result scalar 172 SV *sv; // result scalar
130 JSON json; 173 JSON json;
131 U32 indent; // indentation level 174 U32 indent; // indentation level
132 U32 maxdepth; // max. indentation/recursion level 175 U32 maxdepth; // max. indentation/recursion level
176 UV limit; // escape character values >= this value when encoding
133} enc_t; 177} enc_t;
134 178
135inline void 179INLINE void
136need (enc_t *enc, STRLEN len) 180need (enc_t *enc, STRLEN len)
137{ 181{
138 if (expect_false (enc->cur + len >= enc->end)) 182 if (expect_false (enc->cur + len >= enc->end))
139 { 183 {
140 STRLEN cur = enc->cur - SvPVX (enc->sv); 184 STRLEN cur = enc->cur - SvPVX (enc->sv);
142 enc->cur = SvPVX (enc->sv) + cur; 186 enc->cur = SvPVX (enc->sv) + cur;
143 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 187 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144 } 188 }
145} 189}
146 190
147inline void 191INLINE void
148encode_ch (enc_t *enc, char ch) 192encode_ch (enc_t *enc, char ch)
149{ 193{
150 need (enc, 1); 194 need (enc, 1);
151 *enc->cur++ = ch; 195 *enc->cur++ = ch;
152} 196}
206 { 250 {
207 uch = ch; 251 uch = ch;
208 clen = 1; 252 clen = 1;
209 } 253 }
210 254
211 if (uch > 0x10FFFFUL) 255 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 { 256 {
216 if (uch > 0xFFFFUL) 257 if (uch >= 0x10000UL)
217 { 258 {
259 if (uch >= 0x110000UL)
260 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
261
218 need (enc, len += 11); 262 need (enc, len += 11);
219 sprintf (enc->cur, "\\u%04x\\u%04x", 263 sprintf (enc->cur, "\\u%04x\\u%04x",
220 (int)((uch - 0x10000) / 0x400 + 0xD800), 264 (int)((uch - 0x10000) / 0x400 + 0xD800),
221 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 265 (int)((uch - 0x10000) % 0x400 + 0xDC00));
222 enc->cur += 12; 266 enc->cur += 12;
250 while (--clen); 294 while (--clen);
251 } 295 }
252 else 296 else
253 { 297 {
254 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
255 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 299 enc->cur = encode_utf8 (enc->cur, uch);
256 ++str; 300 ++str;
257 } 301 }
258 } 302 }
259 } 303 }
260 } 304 }
261 305
262 --len; 306 --len;
263 } 307 }
264} 308}
265 309
266inline void 310INLINE void
267encode_indent (enc_t *enc) 311encode_indent (enc_t *enc)
268{ 312{
269 if (enc->json.flags & F_INDENT) 313 if (enc->json.flags & F_INDENT)
270 { 314 {
271 int spaces = enc->indent * INDENT_STEP; 315 int spaces = enc->indent * INDENT_STEP;
274 memset (enc->cur, ' ', spaces); 318 memset (enc->cur, ' ', spaces);
275 enc->cur += spaces; 319 enc->cur += spaces;
276 } 320 }
277} 321}
278 322
279inline void 323INLINE void
280encode_space (enc_t *enc) 324encode_space (enc_t *enc)
281{ 325{
282 need (enc, 1); 326 need (enc, 1);
283 encode_ch (enc, ' '); 327 encode_ch (enc, ' ');
284} 328}
285 329
286inline void 330INLINE void
287encode_nl (enc_t *enc) 331encode_nl (enc_t *enc)
288{ 332{
289 if (enc->json.flags & F_INDENT) 333 if (enc->json.flags & F_INDENT)
290 { 334 {
291 need (enc, 1); 335 need (enc, 1);
292 encode_ch (enc, '\n'); 336 encode_ch (enc, '\n');
293 } 337 }
294} 338}
295 339
296inline void 340INLINE void
297encode_comma (enc_t *enc) 341encode_comma (enc_t *enc)
298{ 342{
299 encode_ch (enc, ','); 343 encode_ch (enc, ',');
300 344
301 if (enc->json.flags & F_INDENT) 345 if (enc->json.flags & F_INDENT)
619 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 663 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
620 enc->cur += strlen (enc->cur); 664 enc->cur += strlen (enc->cur);
621 } 665 }
622 else if (SvIOKp (sv)) 666 else if (SvIOKp (sv))
623 { 667 {
624 // 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
625 if (SvUV (sv) & ~(UV)0x7fff) 669 // we assume two's complement
626 { 670 // we assume no aliasing issues in the union
627 // large integer, use the (rather slow) snprintf way. 671 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
628 need (enc, sizeof (UV) * 3); 672 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
629 enc->cur +=
630 SvIsUV(sv)
631 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
632 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
633 }
634 else
635 { 673 {
636 // optimise the "small number case" 674 // optimise the "small number case"
637 // 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
638 I32 i = SvIV (sv); 677 I32 i = SvIVX (sv);
639 U32 u; 678 U32 u;
640 char digit, nz = 0; 679 char digit, nz = 0;
641 680
642 need (enc, 6); 681 need (enc, 6);
643 682
649 688
650 // 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
651 // 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,
652 // resulting in a net multiplication by 10. 691 // resulting in a net multiplication by 10.
653 // we always write the digit to memory but conditionally increment 692 // we always write the digit to memory but conditionally increment
654 // the pointer, to ease the usage of conditional move instructions. 693 // the pointer, to enable the use of conditional move instructions.
655 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;
656 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;
657 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;
658 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;
659 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));
660 } 708 }
661 } 709 }
662 else if (SvROK (sv)) 710 else if (SvROK (sv))
663 encode_rv (enc, SvRV (sv)); 711 encode_rv (enc, SvRV (sv));
664 else if (!SvOK (sv)) 712 else if (!SvOK (sv))
680 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 728 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
681 enc.cur = SvPVX (enc.sv); 729 enc.cur = SvPVX (enc.sv);
682 enc.end = SvEND (enc.sv); 730 enc.end = SvEND (enc.sv);
683 enc.indent = 0; 731 enc.indent = 0;
684 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;
685 736
686 SvPOK_only (enc.sv); 737 SvPOK_only (enc.sv);
687 encode_sv (&enc, scalar); 738 encode_sv (&enc, scalar);
688 739
689 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 740 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
710 JSON json; 761 JSON json;
711 U32 depth; // recursion depth 762 U32 depth; // recursion depth
712 U32 maxdepth; // recursion depth limit 763 U32 maxdepth; // recursion depth limit
713} dec_t; 764} dec_t;
714 765
715inline void 766INLINE void
716decode_comment (dec_t *dec) 767decode_comment (dec_t *dec)
717{ 768{
718 // only '#'-style comments allowed a.t.m. 769 // only '#'-style comments allowed a.t.m.
719 770
720 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d) 771 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
721 ++dec->cur; 772 ++dec->cur;
722} 773}
723 774
724inline void 775INLINE void
725decode_ws (dec_t *dec) 776decode_ws (dec_t *dec)
726{ 777{
727 for (;;) 778 for (;;)
728 { 779 {
729 char ch = *dec->cur; 780 char ch = *dec->cur;
855 906
856 if (hi >= 0x80) 907 if (hi >= 0x80)
857 { 908 {
858 utf8 = 1; 909 utf8 = 1;
859 910
860 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); 911 cur = encode_utf8 (cur, hi);
861 } 912 }
862 else 913 else
863 *cur++ = hi; 914 *cur++ = hi;
864 } 915 }
865 break; 916 break;
867 default: 918 default:
868 --dec_cur; 919 --dec_cur;
869 ERR ("illegal backslash escape sequence in string"); 920 ERR ("illegal backslash escape sequence in string");
870 } 921 }
871 } 922 }
872 else if (expect_true (ch >= 0x20 && ch <= 0x7f)) 923 else if (expect_true (ch >= 0x20 && ch < 0x80))
873 *cur++ = ch; 924 *cur++ = ch;
874 else if (ch >= 0x80) 925 else if (ch >= 0x80)
875 { 926 {
876 STRLEN clen; 927 STRLEN clen;
877 UV uch; 928 UV uch;
1000 1051
1001 if (!is_nv) 1052 if (!is_nv)
1002 { 1053 {
1003 int len = dec->cur - start; 1054 int len = dec->cur - start;
1004 1055
1005 // 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
1006 if (*start == '-') 1057 if (*start == '-')
1007 switch (len) 1058 switch (len)
1008 { 1059 {
1009 case 2: return newSViv (-( start [1] - '0' * 1)); 1060 case 2: return newSViv (-( start [1] - '0' * 1));
1010 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1061 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
1011 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));
1012 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));
1013 } 1065 }
1014 else 1066 else
1015 switch (len) 1067 switch (len)
1016 { 1068 {
1017 case 1: return newSViv ( start [0] - '0' * 1); 1069 case 1: return newSViv ( start [0] - '0' * 1);
1018 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1070 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1019 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);
1020 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);
1021 } 1074 }
1022 1075
1023 { 1076 {
1024 UV uv; 1077 UV uv;
1025 int numtype = grok_number (start, len, &uv); 1078 int numtype = grok_number (start, len, &uv);
1133 char *p = dec->cur; 1186 char *p = dec->cur;
1134 char *e = p + 24; // only try up to 24 bytes 1187 char *e = p + 24; // only try up to 24 bytes
1135 1188
1136 for (;;) 1189 for (;;)
1137 { 1190 {
1138 // the >= 0x80 is true on most architectures 1191 // the >= 0x80 is false on most architectures
1139 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1192 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1140 { 1193 {
1141 // slow path, back up and use decode_str 1194 // slow path, back up and use decode_str
1142 SV *key = decode_str (dec); 1195 SV *key = decode_str (dec);
1143 if (!key) 1196 if (!key)
1274 1327
1275static SV * 1328static SV *
1276decode_sv (dec_t *dec) 1329decode_sv (dec_t *dec)
1277{ 1330{
1278 // the beauty of JSON: you need exactly one character lookahead 1331 // the beauty of JSON: you need exactly one character lookahead
1279 // to parse anything. 1332 // to parse everything.
1280 switch (*dec->cur) 1333 switch (*dec->cur)
1281 { 1334 {
1282 case '"': ++dec->cur; return decode_str (dec); 1335 case '"': ++dec->cur; return decode_str (dec);
1283 case '[': ++dec->cur; return decode_av (dec); 1336 case '[': ++dec->cur; return decode_av (dec);
1284 case '{': ++dec->cur; return decode_hv (dec); 1337 case '{': ++dec->cur; return decode_hv (dec);
1285 1338
1286 case '-': 1339 case '-':
1287 case '0': case '1': case '2': case '3': case '4': 1340 case '0': case '1': case '2': case '3': case '4':
1288 case '5': case '6': case '7': case '8': case '9': 1341 case '5': case '6': case '7': case '8': case '9':
1289 return decode_num (dec); 1342 return decode_num (dec);
1335fail: 1388fail:
1336 return 0; 1389 return 0;
1337} 1390}
1338 1391
1339static SV * 1392static SV *
1340decode_json (SV *string, JSON *json, UV *offset_return) 1393decode_json (SV *string, JSON *json, STRLEN *offset_return)
1341{ 1394{
1342 dec_t dec; 1395 dec_t dec;
1343 UV offset; 1396 STRLEN offset;
1344 SV *sv; 1397 SV *sv;
1345 1398
1346 SvGETMAGIC (string); 1399 SvGETMAGIC (string);
1347 SvUPGRADE (string, SVt_PV); 1400 SvUPGRADE (string, SVt_PV);
1348 1401
1421 1474
1422 return sv; 1475 return sv;
1423} 1476}
1424 1477
1425///////////////////////////////////////////////////////////////////////////// 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/////////////////////////////////////////////////////////////////////////////
1426// XS interface functions 1595// XS interface functions
1427 1596
1428MODULE = JSON::XS PACKAGE = JSON::XS 1597MODULE = JSON::XS PACKAGE = JSON::XS
1429 1598
1430BOOT: 1599BOOT:
1457{ 1626{
1458 SV *pv = NEWSV (0, sizeof (JSON)); 1627 SV *pv = NEWSV (0, sizeof (JSON));
1459 SvPOK_only (pv); 1628 SvPOK_only (pv);
1460 Zero (SvPVX (pv), 1, JSON); 1629 Zero (SvPVX (pv), 1, JSON);
1461 ((JSON *)SvPVX (pv))->flags = F_DEFAULT; 1630 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1462 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 )));
1463} 1635}
1464 1636
1465void ascii (JSON *self, int enable = 1) 1637void ascii (JSON *self, int enable = 1)
1466 ALIAS: 1638 ALIAS:
1467 ascii = F_ASCII 1639 ascii = F_ASCII
1494 get_utf8 = F_UTF8 1666 get_utf8 = F_UTF8
1495 get_indent = F_INDENT 1667 get_indent = F_INDENT
1496 get_canonical = F_CANONICAL 1668 get_canonical = F_CANONICAL
1497 get_space_before = F_SPACE_BEFORE 1669 get_space_before = F_SPACE_BEFORE
1498 get_space_after = F_SPACE_AFTER 1670 get_space_after = F_SPACE_AFTER
1499 get_pretty = F_PRETTY
1500 get_allow_nonref = F_ALLOW_NONREF 1671 get_allow_nonref = F_ALLOW_NONREF
1501 get_shrink = F_SHRINK 1672 get_shrink = F_SHRINK
1502 get_allow_blessed = F_ALLOW_BLESSED 1673 get_allow_blessed = F_ALLOW_BLESSED
1503 get_convert_blessed = F_CONV_BLESSED 1674 get_convert_blessed = F_CONV_BLESSED
1504 get_relaxed = F_RELAXED 1675 get_relaxed = F_RELAXED
1588 XPUSHs (decode_json (jsonstr, self, 0)); 1759 XPUSHs (decode_json (jsonstr, self, 0));
1589 1760
1590void decode_prefix (JSON *self, SV *jsonstr) 1761void decode_prefix (JSON *self, SV *jsonstr)
1591 PPCODE: 1762 PPCODE:
1592{ 1763{
1593 UV offset; 1764 STRLEN offset;
1594 EXTEND (SP, 2); 1765 EXTEND (SP, 2);
1595 PUSHs (decode_json (jsonstr, self, &offset)); 1766 PUSHs (decode_json (jsonstr, self, &offset));
1596 PUSHs (sv_2mortal (newSVuv (offset))); 1767 PUSHs (sv_2mortal (newSVuv (offset)));
1597} 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
1598 1830
1599void DESTROY (JSON *self) 1831void DESTROY (JSON *self)
1600 CODE: 1832 CODE:
1601 SvREFCNT_dec (self->cb_sk_object); 1833 SvREFCNT_dec (self->cb_sk_object);
1602 SvREFCNT_dec (self->cb_object); 1834 SvREFCNT_dec (self->cb_object);
1835 SvREFCNT_dec (self->incr_text);
1603 1836
1604PROTOTYPES: ENABLE 1837PROTOTYPES: ENABLE
1605 1838
1606void to_json (SV *scalar) 1839void encode_json (SV *scalar)
1840 ALIAS:
1841 to_json_ = 0
1842 encode_json = F_UTF8
1607 PPCODE: 1843 PPCODE:
1608{ 1844{
1609 JSON json = { F_DEFAULT | F_UTF8 }; 1845 JSON json = { F_DEFAULT | ix };
1610 XPUSHs (encode_json (scalar, &json)); 1846 XPUSHs (encode_json (scalar, &json));
1611} 1847}
1612 1848
1613void from_json (SV *jsonstr) 1849void decode_json (SV *jsonstr)
1850 ALIAS:
1851 from_json_ = 0
1852 decode_json = F_UTF8
1614 PPCODE: 1853 PPCODE:
1615{ 1854{
1616 JSON json = { F_DEFAULT | F_UTF8 }; 1855 JSON json = { F_DEFAULT | ix };
1617 XPUSHs (decode_json (jsonstr, &json, 0)); 1856 XPUSHs (decode_json (jsonstr, &json, 0));
1618} 1857}
1619 1858
1859

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines