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 1.42 by root, Sat Jun 23 22:56:08 2007 UTC vs.
Revision 1.140 by root, Fri Apr 19 17:09:51 2019 UTC

1#include "EXTERN.h" 1#include "EXTERN.h"
2#include "perl.h" 2#include "perl.h"
3#include "XSUB.h" 3#include "XSUB.h"
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>
10#include <float.h>
11#include <inttypes.h>
9 12
10#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
11# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
12#endif 15#endif
13 16
14// some old perls do not have this, try to make it work, no 17// some old perls do not have this, try to make it work, no
15// guarentees, though. if it breaks, you get to keep the pieces. 18// guarantees, though. if it breaks, you get to keep the pieces.
16#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
17# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
18#endif 21#endif
19 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
36
20#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
21#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
22#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
23#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
24#define F_CANONICAL 0x00000010UL 41#define F_CANONICAL 0x00000010UL
25#define F_SPACE_BEFORE 0x00000020UL 42#define F_SPACE_BEFORE 0x00000020UL
26#define F_SPACE_AFTER 0x00000040UL 43#define F_SPACE_AFTER 0x00000040UL
27#define F_ALLOW_NONREF 0x00000100UL 44#define F_ALLOW_NONREF 0x00000100UL
28#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
46#define F_ALLOW_BLESSED 0x00000400UL
47#define F_CONV_BLESSED 0x00000800UL
29#define F_MAXDEPTH 0xf8000000UL 48#define F_RELAXED 0x00001000UL
30#define S_MAXDEPTH 27 49#define F_ALLOW_UNKNOWN 0x00002000UL
31 50#define F_ALLOW_TAGS 0x00004000UL
32#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
33
34// F_SELFCONVERT? <=> to_json/toJson
35// F_BLESSED? <=> { $__class__$ => }
36 52
37#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
38#define F_DEFAULT (9UL << S_MAXDEPTH)
39 54
40#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
41#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
42 57
43#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
59
60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
44 61
45#define SB do { 62#define SB do {
46#define SE } while (0) 63#define SE } while (0)
47 64
48#if __GNUC__ >= 3 65#if __GNUC__ >= 3
49# define expect(expr,value) __builtin_expect ((expr),(value)) 66# define expect(expr,value) __builtin_expect ((expr), (value))
50# define inline inline 67# define INLINE static inline
51#else 68#else
52# define expect(expr,value) (expr) 69# define expect(expr,value) (expr)
53# define inline static 70# define INLINE static
54#endif 71#endif
55 72
56#define expect_false(expr) expect ((expr) != 0, 0) 73#define expect_false(expr) expect ((expr) != 0, 0)
57#define expect_true(expr) expect ((expr) != 0, 1) 74#define expect_true(expr) expect ((expr) != 0, 1)
58 75
59static HV *json_stash; // JSON::XS:: 76#define IN_RANGE_INC(type,val,beg,end) \
77 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
79
80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
81
82#ifdef USE_ITHREADS
83# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
84# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
85# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
86#else
87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
90#endif
91
92// the amount of HEs to allocate on the stack, when sorting keys
93#define STACK_HES 64
94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_false, *bool_true;
97static SV *sv_json;
98
99enum {
100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
101 INCR_M_TFN, // inside true/false/null
102 INCR_M_NUM, // inside number
103 INCR_M_STR, // inside string
104 INCR_M_BS, // inside backslash
105 INCR_M_C0, // inside comment in initial whitespace sequence
106 INCR_M_C1, // inside comment in other places
107 INCR_M_JSON // outside anything, count nesting
108};
109
110#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
111
112typedef struct {
113 U32 flags;
114 U32 max_depth;
115 STRLEN max_size;
116
117 SV *cb_object;
118 HV *cb_sk_object;
119
120 // for the incremental parser
121 SV *incr_text; // the source text so far
122 STRLEN incr_pos; // the current offset into the text
123 int incr_nest; // {[]}-nesting level
124 unsigned char incr_mode;
125
126 SV *v_false, *v_true;
127} JSON;
128
129INLINE void
130json_init (JSON *json)
131{
132 static const JSON init = { F_ALLOW_NONREF, 512 };
133
134 *json = init;
135}
60 136
61///////////////////////////////////////////////////////////////////////////// 137/////////////////////////////////////////////////////////////////////////////
62// utility functions 138// utility functions
63 139
64static UV * 140INLINE SV *
65SvJSON (SV *sv) 141get_bool (const char *name)
66{ 142{
67 if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) 143 SV *sv = get_sv (name, 1);
68 croak ("object is not of type JSON::XS");
69 144
70 return &SvUVX (SvRV (sv)); 145 SvREADONLY_on (sv);
71} 146 SvREADONLY_on (SvRV (sv));
72 147
73static void 148 return sv;
149}
150
151INLINE void
74shrink (SV *sv) 152shrink (SV *sv)
75{ 153{
76 sv_utf8_downgrade (sv, 1); 154 sv_utf8_downgrade (sv, 1);
155
77 if (SvLEN (sv) > SvCUR (sv) + 1) 156 if (SvLEN (sv) > SvCUR (sv) + 1)
78 { 157 {
79#ifdef SvPV_shrink_to_cur 158#ifdef SvPV_shrink_to_cur
80 SvPV_shrink_to_cur (sv); 159 SvPV_shrink_to_cur (sv);
81#elif defined (SvPV_renew) 160#elif defined (SvPV_renew)
82 SvPV_renew (sv, SvCUR (sv) + 1); 161 SvPV_renew (sv, SvCUR (sv) + 1);
83#endif 162#endif
84 } 163 }
85} 164}
86 165
166/* adds two STRLENs together, slow, and with paranoia */
167static STRLEN
168strlen_sum (STRLEN l1, STRLEN l2)
169{
170 size_t sum = l1 + l2;
171
172 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
173 croak ("JSON::XS: string size overflow");
174
175 return sum;
176}
177
178/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
179static char *
180json_sv_grow (SV *sv, size_t len1, size_t len2)
181{
182 len1 = strlen_sum (len1, len2);
183 len1 = strlen_sum (len1, len1 >> 1);
184
185 if (len1 > 4096 - 24)
186 len1 = (len1 | 4095) - 24;
187
188 return SvGROW (sv, len1);
189}
190
87// decode an utf-8 character and return it, or (UV)-1 in 191// decode a utf-8 character and return it, or (UV)-1 in
88// case of an error. 192// case of an error.
89// we special-case "safe" characters from U+80 .. U+7FF, 193// we special-case "safe" characters from U+80 .. U+7FF,
90// but use the very good perl function to parse anything else. 194// but use the very good perl function to parse anything else.
91// note that we never call this function for a ascii codepoints 195// note that we never call this function for a ascii codepoints
92inline UV 196INLINE UV
93decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) 197decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
94{ 198{
95 if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) 199 if (expect_true (len >= 2
96 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); 200 && IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
97 else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) 201 && IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
98 { 202 {
99 *clen = 2; 203 *clen = 2;
100 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); 204 return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
101 } 205 }
102 else 206 else
207 return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
208}
209
210// likewise for encoding, also never called for ascii codepoints
211// this function takes advantage of this fact, although current gccs
212// seem to optimise the check for >= 0x80 away anyways
213INLINE unsigned char *
214encode_utf8 (unsigned char *s, UV ch)
215{
216 if (expect_false (ch < 0x000080))
217 *s++ = ch;
218 else if (expect_true (ch < 0x000800))
219 *s++ = 0xc0 | ( ch >> 6),
220 *s++ = 0x80 | ( ch & 0x3f);
221 else if ( ch < 0x010000)
222 *s++ = 0xe0 | ( ch >> 12),
223 *s++ = 0x80 | ((ch >> 6) & 0x3f),
224 *s++ = 0x80 | ( ch & 0x3f);
225 else if ( ch < 0x110000)
226 *s++ = 0xf0 | ( ch >> 18),
227 *s++ = 0x80 | ((ch >> 12) & 0x3f),
228 *s++ = 0x80 | ((ch >> 6) & 0x3f),
229 *s++ = 0x80 | ( ch & 0x3f);
230
231 return s;
232}
233
234// convert offset pointer to character index, sv must be string
235static STRLEN
236ptr_to_index (SV *sv, char *offset)
237{
238 return SvUTF8 (sv)
239 ? utf8_distance (offset, SvPVX (sv))
240 : offset - SvPVX (sv);
241}
242
243/////////////////////////////////////////////////////////////////////////////
244// fp hell
245
246// scan a group of digits, and a trailing exponent
247static void
248json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
249{
250 UV uaccum = 0;
251 int eaccum = 0;
252
253 // if we recurse too deep, skip all remaining digits
254 // to avoid a stack overflow attack
255 if (expect_false (--maxdepth <= 0))
256 while (((U8)*s - '0') < 10)
257 ++s;
258
259 for (;;)
260 {
261 U8 dig = (U8)*s - '0';
262
263 if (expect_false (dig >= 10))
264 {
265 if (dig == (U8)((U8)'.' - (U8)'0'))
266 {
267 ++s;
268 json_atof_scan1 (s, accum, expo, 1, maxdepth);
269 }
270 else if ((dig | ' ') == 'e' - '0')
271 {
272 int exp2 = 0;
273 int neg = 0;
274
275 ++s;
276
277 if (*s == '-')
278 {
279 ++s;
280 neg = 1;
281 }
282 else if (*s == '+')
283 ++s;
284
285 while ((dig = (U8)*s - '0') < 10)
286 exp2 = exp2 * 10 + *s++ - '0';
287
288 *expo += neg ? -exp2 : exp2;
289 }
290
291 break;
292 }
293
294 ++s;
295
296 uaccum = uaccum * 10 + dig;
297 ++eaccum;
298
299 // if we have too many digits, then recurse for more
300 // we actually do this for rather few digits
301 if (uaccum >= (UV_MAX - 9) / 10)
302 {
303 if (postdp) *expo -= eaccum;
304 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
305 if (postdp) *expo += eaccum;
306
307 break;
308 }
103 { 309 }
104 *clen = (STRLEN)-1; 310
311 // this relies greatly on the quality of the pow ()
312 // implementation of the platform, but a good
313 // implementation is hard to beat.
314 // (IEEE 754 conformant ones are required to be exact)
315 if (postdp) *expo -= eaccum;
316 *accum += uaccum * Perl_pow (10., *expo);
317 *expo += eaccum;
318}
319
320static NV
321json_atof (const char *s)
322{
323 NV accum = 0.;
324 int expo = 0;
325 int neg = 0;
326
327 if (*s == '-')
328 {
329 ++s;
330 neg = 1;
331 }
332
333 // a recursion depth of ten gives us >>500 bits
334 json_atof_scan1 (s, &accum, &expo, 0, 10);
335
336 return neg ? -accum : accum;
337}
338
339// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
340static int
341ref_bool_type (SV *sv)
342{
343 svtype svt = SvTYPE (sv);
344
345 if (svt < SVt_PVAV)
346 {
347 STRLEN len = 0;
348 char *pv = svt ? SvPV (sv, len) : 0;
349
350 if (len == 1)
351 if (*pv == '1')
352 return 1;
353 else if (*pv == '0')
354 return 0;
355 }
356
357 return -1;
358}
359
360// returns whether scalar is not a reference in the sense of allow_nonref
361static int
362json_nonref (SV *scalar)
363{
364 if (!SvROK (scalar))
365 return 1;
366
367 scalar = SvRV (scalar);
368
369 if (SvTYPE (scalar) >= SVt_PVMG)
370 {
371 if (SvSTASH (scalar) == bool_stash)
105 return (UV)-1; 372 return 1;
373
374 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
375 return 1;
106 } 376 }
377
378 return 0;
107} 379}
108 380
109///////////////////////////////////////////////////////////////////////////// 381/////////////////////////////////////////////////////////////////////////////
110// encoder 382// encoder
111 383
… …
113typedef struct 385typedef struct
114{ 386{
115 char *cur; // SvPVX (sv) + current output position 387 char *cur; // SvPVX (sv) + current output position
116 char *end; // SvEND (sv) 388 char *end; // SvEND (sv)
117 SV *sv; // result scalar 389 SV *sv; // result scalar
118 U32 flags; // F_* 390 JSON json;
119 U32 indent; // indentation level 391 U32 indent; // indentation level
120 U32 maxdepth; // max. indentation/recursion level 392 UV limit; // escape character values >= this value when encoding
121} enc_t; 393} enc_t;
122 394
123inline void 395INLINE void
124need (enc_t *enc, STRLEN len) 396need (enc_t *enc, STRLEN len)
125{ 397{
126 if (expect_false (enc->cur + len >= enc->end)) 398 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
127 { 399 {
128 STRLEN cur = enc->cur - SvPVX (enc->sv); 400 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
129 SvGROW (enc->sv, cur + len + 1); 401 char *buf = json_sv_grow (enc->sv, cur, len);
130 enc->cur = SvPVX (enc->sv) + cur; 402 enc->cur = buf + cur;
131 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 403 enc->end = buf + SvLEN (enc->sv) - 1;
132 } 404 }
133} 405}
134 406
135inline void 407INLINE void
136encode_ch (enc_t *enc, char ch) 408encode_ch (enc_t *enc, char ch)
137{ 409{
138 need (enc, 1); 410 need (enc, 1);
139 *enc->cur++ = ch; 411 *enc->cur++ = ch;
140} 412}
… …
152 424
153 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 425 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
154 { 426 {
155 if (expect_false (ch == '"')) // but with slow exceptions 427 if (expect_false (ch == '"')) // but with slow exceptions
156 { 428 {
157 need (enc, len += 1); 429 need (enc, len + 1);
158 *enc->cur++ = '\\'; 430 *enc->cur++ = '\\';
159 *enc->cur++ = '"'; 431 *enc->cur++ = '"';
160 } 432 }
161 else if (expect_false (ch == '\\')) 433 else if (expect_false (ch == '\\'))
162 { 434 {
163 need (enc, len += 1); 435 need (enc, len + 1);
164 *enc->cur++ = '\\'; 436 *enc->cur++ = '\\';
165 *enc->cur++ = '\\'; 437 *enc->cur++ = '\\';
166 } 438 }
167 else 439 else
168 *enc->cur++ = ch; 440 *enc->cur++ = ch;
… …
171 } 443 }
172 else 444 else
173 { 445 {
174 switch (ch) 446 switch (ch)
175 { 447 {
176 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 448 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
177 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 449 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
178 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 450 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
179 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 451 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
180 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 452 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
181 453
182 default: 454 default:
183 { 455 {
184 STRLEN clen; 456 STRLEN clen;
185 UV uch; 457 UV uch;
… …
194 { 466 {
195 uch = ch; 467 uch = ch;
196 clen = 1; 468 clen = 1;
197 } 469 }
198 470
199 if (uch > 0x10FFFFUL) 471 if (uch < 0x80/*0x20*/ || uch >= enc->limit)
200 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
201
202 if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
203 { 472 {
204 if (uch > 0xFFFFUL) 473 if (uch >= 0x10000UL)
205 { 474 {
475 if (uch >= 0x110000UL)
476 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
477
206 need (enc, len += 11); 478 need (enc, len + 11);
207 sprintf (enc->cur, "\\u%04x\\u%04x", 479 sprintf (enc->cur, "\\u%04x\\u%04x",
208 (int)((uch - 0x10000) / 0x400 + 0xD800), 480 (int)((uch - 0x10000) / 0x400 + 0xD800),
209 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 481 (int)((uch - 0x10000) % 0x400 + 0xDC00));
210 enc->cur += 12; 482 enc->cur += 12;
211 } 483 }
212 else 484 else
213 { 485 {
214 static char hexdigit [16] = "0123456789abcdef";
215 need (enc, len += 5); 486 need (enc, len + 5);
216 *enc->cur++ = '\\'; 487 *enc->cur++ = '\\';
217 *enc->cur++ = 'u'; 488 *enc->cur++ = 'u';
218 *enc->cur++ = hexdigit [ uch >> 12 ]; 489 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
219 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 490 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
220 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 491 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
221 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 492 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
222 } 493 }
223 494
224 str += clen; 495 str += clen;
225 } 496 }
226 else if (enc->flags & F_LATIN1) 497 else if (enc->json.flags & F_LATIN1)
227 { 498 {
228 *enc->cur++ = uch; 499 *enc->cur++ = uch;
229 str += clen; 500 str += clen;
230 } 501 }
231 else if (is_utf8) 502 else if (is_utf8)
232 { 503 {
233 need (enc, len += clen); 504 need (enc, len + clen);
234 do 505 do
235 { 506 {
236 *enc->cur++ = *str++; 507 *enc->cur++ = *str++;
237 } 508 }
238 while (--clen); 509 while (--clen);
239 } 510 }
240 else 511 else
241 { 512 {
242 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 513 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
243 enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); 514 enc->cur = encode_utf8 (enc->cur, uch);
244 ++str; 515 ++str;
245 } 516 }
246 } 517 }
247 } 518 }
248 } 519 }
249 520
250 --len; 521 --len;
251 } 522 }
252} 523}
253 524
254inline void 525INLINE void
255encode_indent (enc_t *enc) 526encode_indent (enc_t *enc)
256{ 527{
257 if (enc->flags & F_INDENT) 528 if (enc->json.flags & F_INDENT)
258 { 529 {
259 int spaces = enc->indent * INDENT_STEP; 530 int spaces = enc->indent * INDENT_STEP;
260 531
261 need (enc, spaces); 532 need (enc, spaces);
262 memset (enc->cur, ' ', spaces); 533 memset (enc->cur, ' ', spaces);
263 enc->cur += spaces; 534 enc->cur += spaces;
264 } 535 }
265} 536}
266 537
267inline void 538INLINE void
268encode_space (enc_t *enc) 539encode_space (enc_t *enc)
269{ 540{
270 need (enc, 1); 541 need (enc, 1);
271 encode_ch (enc, ' '); 542 encode_ch (enc, ' ');
272} 543}
273 544
274inline void 545INLINE void
275encode_nl (enc_t *enc) 546encode_nl (enc_t *enc)
276{ 547{
277 if (enc->flags & F_INDENT) 548 if (enc->json.flags & F_INDENT)
278 { 549 {
279 need (enc, 1); 550 need (enc, 1);
280 encode_ch (enc, '\n'); 551 encode_ch (enc, '\n');
281 } 552 }
282} 553}
283 554
284inline void 555INLINE void
285encode_comma (enc_t *enc) 556encode_comma (enc_t *enc)
286{ 557{
287 encode_ch (enc, ','); 558 encode_ch (enc, ',');
288 559
289 if (enc->flags & F_INDENT) 560 if (enc->json.flags & F_INDENT)
290 encode_nl (enc); 561 encode_nl (enc);
291 else if (enc->flags & F_SPACE_AFTER) 562 else if (enc->json.flags & F_SPACE_AFTER)
292 encode_space (enc); 563 encode_space (enc);
293} 564}
294 565
295static void encode_sv (enc_t *enc, SV *sv); 566static void encode_sv (enc_t *enc, SV *sv);
296 567
297static void 568static void
298encode_av (enc_t *enc, AV *av) 569encode_av (enc_t *enc, AV *av)
299{ 570{
300 int i, len = av_len (av); 571 int i, len = av_len (av);
301 572
302 if (enc->indent >= enc->maxdepth) 573 if (enc->indent >= enc->json.max_depth)
303 croak ("data structure too deep (hit recursion limit)"); 574 croak (ERR_NESTING_EXCEEDED);
304 575
305 encode_ch (enc, '['); encode_nl (enc); 576 encode_ch (enc, '[');
306 ++enc->indent;
307 577
578 if (len >= 0)
579 {
580 encode_nl (enc); ++enc->indent;
581
308 for (i = 0; i <= len; ++i) 582 for (i = 0; i <= len; ++i)
309 { 583 {
584 SV **svp = av_fetch (av, i, 0);
585
310 encode_indent (enc); 586 encode_indent (enc);
311 encode_sv (enc, *av_fetch (av, i, 0));
312 587
588 if (svp)
589 encode_sv (enc, *svp);
590 else
591 encode_str (enc, "null", 4, 0);
592
313 if (i < len) 593 if (i < len)
314 encode_comma (enc); 594 encode_comma (enc);
315 } 595 }
316 596
597 encode_nl (enc); --enc->indent; encode_indent (enc);
598 }
599
317 encode_nl (enc); 600 encode_ch (enc, ']');
318
319 --enc->indent;
320 encode_indent (enc); encode_ch (enc, ']');
321} 601}
322 602
323static void 603static void
324encode_he (enc_t *enc, HE *he) 604encode_hk (enc_t *enc, HE *he)
325{ 605{
326 encode_ch (enc, '"'); 606 encode_ch (enc, '"');
327 607
328 if (HeKLEN (he) == HEf_SVKEY) 608 if (HeKLEN (he) == HEf_SVKEY)
329 { 609 {
330 SV *sv = HeSVKEY (he); 610 SV *sv = HeSVKEY (he);
331 STRLEN len; 611 STRLEN len;
332 char *str; 612 char *str;
333 613
334 SvGETMAGIC (sv); 614 SvGETMAGIC (sv);
335 str = SvPV (sv, len); 615 str = SvPV (sv, len);
336 616
337 encode_str (enc, str, len, SvUTF8 (sv)); 617 encode_str (enc, str, len, SvUTF8 (sv));
338 } 618 }
339 else 619 else
340 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); 620 encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
341 621
342 encode_ch (enc, '"'); 622 encode_ch (enc, '"');
343 623
344 if (enc->flags & F_SPACE_BEFORE) encode_space (enc); 624 if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
345 encode_ch (enc, ':'); 625 encode_ch (enc, ':');
346 if (enc->flags & F_SPACE_AFTER ) encode_space (enc); 626 if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
347 encode_sv (enc, HeVAL (he));
348} 627}
349 628
350// compare hash entries, used when all keys are bytestrings 629// compare hash entries, used when all keys are bytestrings
351static int 630static int
352he_cmp_fast (const void *a_, const void *b_) 631he_cmp_fast (const void *a_, const void *b_)
… …
357 HE *b = *(HE **)b_; 636 HE *b = *(HE **)b_;
358 637
359 STRLEN la = HeKLEN (a); 638 STRLEN la = HeKLEN (a);
360 STRLEN lb = HeKLEN (b); 639 STRLEN lb = HeKLEN (b);
361 640
362 if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) 641 if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
363 cmp = la - lb; 642 cmp = lb - la;
364 643
365 return cmp; 644 return cmp;
366} 645}
367 646
368// compare hash entries, used when some keys are sv's or utf-x 647// compare hash entries, used when some keys are sv's or utf-x
369static int 648static int
370he_cmp_slow (const void *a, const void *b) 649he_cmp_slow (const void *a, const void *b)
371{ 650{
372 return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); 651 return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
373} 652}
374 653
375static void 654static void
376encode_hv (enc_t *enc, HV *hv) 655encode_hv (enc_t *enc, HV *hv)
377{ 656{
378 int count, i; 657 HE *he;
379 658
380 if (enc->indent >= enc->maxdepth) 659 if (enc->indent >= enc->json.max_depth)
381 croak ("data structure too deep (hit recursion limit)"); 660 croak (ERR_NESTING_EXCEEDED);
382 661
383 encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; 662 encode_ch (enc, '{');
384 663
385 if ((count = hv_iterinit (hv)))
386 {
387 // for canonical output we have to sort by keys first 664 // for canonical output we have to sort by keys first
388 // actually, this is mostly due to the stupid so-called 665 // actually, this is mostly due to the stupid so-called
389 // security workaround added somewhere in 5.8.x. 666 // security workaround added somewhere in 5.8.x
390 // that randomises hash orderings 667 // that randomises hash orderings
391 if (enc->flags & F_CANONICAL) 668 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
669 {
670 int count = hv_iterinit (hv);
671
672 if (SvMAGICAL (hv))
392 { 673 {
674 // need to count by iterating. could improve by dynamically building the vector below
675 // but I don't care for the speed of this special case.
676 // note also that we will run into undefined behaviour when the two iterations
677 // do not result in the same count, something I might care for in some later release.
678
679 count = 0;
680 while (hv_iternext (hv))
681 ++count;
682
683 hv_iterinit (hv);
684 }
685
686 if (count)
687 {
393 int fast = 1; 688 int i, fast = 1;
394 HE *he; 689 HE *hes_stack [STACK_HES];
395#if defined(__BORLANDC__) || defined(_MSC_VER) 690 HE **hes = hes_stack;
396 HE **hes = _alloca (count * sizeof (HE)); 691
397#else 692 // allocate larger arrays on the heap
398 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 693 if (count > STACK_HES)
399#endif 694 {
695 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
696 hes = (HE **)SvPVX (sv);
697 }
400 698
401 i = 0; 699 i = 0;
402 while ((he = hv_iternext (hv))) 700 while ((he = hv_iternext (hv)))
403 { 701 {
404 hes [i++] = he; 702 hes [i++] = he;
… …
426 724
427 FREETMPS; 725 FREETMPS;
428 LEAVE; 726 LEAVE;
429 } 727 }
430 728
431 for (i = 0; i < count; ++i) 729 encode_nl (enc); ++enc->indent;
730
731 while (count--)
432 { 732 {
433 encode_indent (enc); 733 encode_indent (enc);
734 he = hes [count];
434 encode_he (enc, hes [i]); 735 encode_hk (enc, he);
736 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
435 737
436 if (i < count - 1) 738 if (count)
437 encode_comma (enc); 739 encode_comma (enc);
438 } 740 }
439 741
440 encode_nl (enc); 742 encode_nl (enc); --enc->indent; encode_indent (enc);
441 } 743 }
744 }
442 else 745 else
746 {
747 if (hv_iterinit (hv) || SvMAGICAL (hv))
748 if ((he = hv_iternext (hv)))
443 { 749 {
444 HE *he = hv_iternext (hv); 750 encode_nl (enc); ++enc->indent;
445 751
446 for (;;) 752 for (;;)
447 { 753 {
448 encode_indent (enc); 754 encode_indent (enc);
449 encode_he (enc, he); 755 encode_hk (enc, he);
756 encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
450 757
451 if (!(he = hv_iternext (hv))) 758 if (!(he = hv_iternext (hv)))
452 break; 759 break;
453 760
454 encode_comma (enc); 761 encode_comma (enc);
455 } 762 }
456 763
457 encode_nl (enc); 764 encode_nl (enc); --enc->indent; encode_indent (enc);
458 } 765 }
459 } 766 }
460 767
461 --enc->indent; encode_indent (enc); encode_ch (enc, '}'); 768 encode_ch (enc, '}');
462} 769}
463 770
464// encode objects, arrays and special \0=false and \1=true values. 771// encode objects, arrays and special \0=false and \1=true values.
465static void 772static void
466encode_rv (enc_t *enc, SV *sv) 773encode_rv (enc_t *enc, SV *sv)
467{ 774{
468 svtype svt; 775 svtype svt;
776 GV *method;
469 777
470 SvGETMAGIC (sv); 778 SvGETMAGIC (sv);
471 svt = SvTYPE (sv); 779 svt = SvTYPE (sv);
472 780
781 if (expect_false (SvOBJECT (sv)))
782 {
783 HV *stash = SvSTASH (sv);
784
785 if (stash == bool_stash)
786 {
787 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
788 else encode_str (enc, "false", 5, 0);
789 }
790 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
791 {
792 int count;
793 dSP;
794
795 ENTER; SAVETMPS;
796 PUSHMARK (SP);
797 EXTEND (SP, 2);
798 // we re-bless the reference to get overload and other niceties right
799 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
800 PUSHs (sv_json);
801
802 PUTBACK;
803 count = call_sv ((SV *)GvCV (method), G_ARRAY);
804 SPAGAIN;
805
806 // catch this surprisingly common error
807 if (SvROK (TOPs) && SvRV (TOPs) == sv)
808 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
809
810 encode_ch (enc, '(');
811 encode_ch (enc, '"');
812 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
813 encode_ch (enc, '"');
814 encode_ch (enc, ')');
815 encode_ch (enc, '[');
816
817 if (count)
818 {
819 int i;
820
821 for (i = 0; i < count - 1; ++i)
822 {
823 encode_sv (enc, SP[i + 1 - count]);
824 encode_ch (enc, ',');
825 }
826
827 encode_sv (enc, TOPs);
828 SP -= count;
829 }
830
831 PUTBACK;
832
833 encode_ch (enc, ']');
834
835 FREETMPS; LEAVE;
836 }
837 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
838 {
839 dSP;
840
841 ENTER; SAVETMPS;
842 PUSHMARK (SP);
843 // we re-bless the reference to get overload and other niceties right
844 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
845
846 // calling with G_SCALAR ensures that we always get a 1 return value
847 PUTBACK;
848 call_sv ((SV *)GvCV (method), G_SCALAR);
849 SPAGAIN;
850
851 // catch this surprisingly common error
852 if (SvROK (TOPs) && SvRV (TOPs) == sv)
853 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
854
855 sv = POPs;
856 PUTBACK;
857
858 encode_sv (enc, sv);
859
860 FREETMPS; LEAVE;
861 }
862 else if (enc->json.flags & F_ALLOW_BLESSED)
863 encode_str (enc, "null", 4, 0);
864 else
865 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
866 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
867 }
473 if (svt == SVt_PVHV) 868 else if (svt == SVt_PVHV)
474 encode_hv (enc, (HV *)sv); 869 encode_hv (enc, (HV *)sv);
475 else if (svt == SVt_PVAV) 870 else if (svt == SVt_PVAV)
476 encode_av (enc, (AV *)sv); 871 encode_av (enc, (AV *)sv);
477 else if (svt < SVt_PVAV) 872 else if (svt < SVt_PVAV)
478 { 873 {
479 if (SvNIOK (sv) && SvIV (sv) == 0) 874 int bool_type = ref_bool_type (sv);
875
876 if (bool_type == 1)
877 encode_str (enc, "true", 4, 0);
878 else if (bool_type == 0)
480 encode_str (enc, "false", 5, 0); 879 encode_str (enc, "false", 5, 0);
481 else if (SvNIOK (sv) && SvIV (sv) == 1) 880 else if (enc->json.flags & F_ALLOW_UNKNOWN)
482 encode_str (enc, "true", 4, 0); 881 encode_str (enc, "null", 4, 0);
483 else 882 else
484 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 883 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
485 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 884 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
486 } 885 }
886 else if (enc->json.flags & F_ALLOW_UNKNOWN)
887 encode_str (enc, "null", 4, 0);
487 else 888 else
488 croak ("encountered %s, but JSON can only represent references to arrays or hashes", 889 croak ("encountered %s, but JSON can only represent references to arrays or hashes",
489 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 890 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
490} 891}
491 892
… …
509 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); 910 Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
510 enc->cur += strlen (enc->cur); 911 enc->cur += strlen (enc->cur);
511 } 912 }
512 else if (SvIOKp (sv)) 913 else if (SvIOKp (sv))
513 { 914 {
514 // we assume we can always read an IV as a UV 915 // we assume we can always read an IV as a UV and vice versa
515 if (SvUV (sv) & ~(UV)0x7fff) 916 // we assume two's complement
516 { 917 // we assume no aliasing issues in the union
517 // large integer, use the (rather slow) snprintf way. 918 if (SvIsUV (sv) ? SvUVX (sv) <= 59000
518 need (enc, sizeof (UV) * 3); 919 : SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
519 enc->cur +=
520 SvIsUV(sv)
521 ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
522 : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
523 }
524 else
525 { 920 {
526 // optimise the "small number case" 921 // optimise the "small number case"
527 // code will likely be branchless and use only a single multiplication 922 // code will likely be branchless and use only a single multiplication
923 // works for numbers up to 59074
528 I32 i = SvIV (sv); 924 I32 i = SvIVX (sv);
529 U32 u; 925 U32 u;
530 char digit, nz = 0; 926 char digit, nz = 0;
531 927
532 need (enc, 6); 928 need (enc, 6);
533 929
… …
539 935
540 // now output digit by digit, each time masking out the integer part 936 // now output digit by digit, each time masking out the integer part
541 // and multiplying by 5 while moving the decimal point one to the right, 937 // and multiplying by 5 while moving the decimal point one to the right,
542 // resulting in a net multiplication by 10. 938 // resulting in a net multiplication by 10.
543 // we always write the digit to memory but conditionally increment 939 // we always write the digit to memory but conditionally increment
544 // the pointer, to ease the usage of conditional move instructions. 940 // the pointer, to enable the use of conditional move instructions.
545 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; 941 digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
546 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; 942 digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
547 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; 943 digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
548 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; 944 digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
549 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' 945 digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
550 } 946 }
947 else
948 {
949 // large integer, use the (rather slow) snprintf way.
950 need (enc, IVUV_MAXCHARS);
951 enc->cur +=
952 SvIsUV(sv)
953 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
954 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
955 }
551 } 956 }
552 else if (SvROK (sv)) 957 else if (SvROK (sv))
553 encode_rv (enc, SvRV (sv)); 958 encode_rv (enc, SvRV (sv));
554 else if (!SvOK (sv)) 959 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
555 encode_str (enc, "null", 4, 0); 960 encode_str (enc, "null", 4, 0);
556 else 961 else
557 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 962 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
558 SvPV_nolen (sv), SvFLAGS (sv)); 963 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
559} 964}
560 965
561static SV * 966static SV *
562encode_json (SV *scalar, U32 flags) 967encode_json (SV *scalar, JSON *json)
563{ 968{
564 enc_t enc; 969 enc_t enc;
565 970
566 if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) 971 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
567 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 972 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
568 973
569 enc.flags = flags; 974 enc.json = *json;
570 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 975 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
571 enc.cur = SvPVX (enc.sv); 976 enc.cur = SvPVX (enc.sv);
572 enc.end = SvEND (enc.sv); 977 enc.end = SvEND (enc.sv);
573 enc.indent = 0; 978 enc.indent = 0;
574 enc.maxdepth = DEC_DEPTH (flags); 979 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
980 : enc.json.flags & F_LATIN1 ? 0x000100UL
981 : 0x110000UL;
575 982
576 SvPOK_only (enc.sv); 983 SvPOK_only (enc.sv);
577 encode_sv (&enc, scalar); 984 encode_sv (&enc, scalar);
985 encode_nl (&enc);
578 986
579 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 987 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
580 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 988 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
581 989
582 if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8))) 990 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
583 SvUTF8_on (enc.sv); 991 SvUTF8_on (enc.sv);
584 992
585 if (enc.flags & F_SHRINK) 993 if (enc.json.flags & F_SHRINK)
586 shrink (enc.sv); 994 shrink (enc.sv);
587 995
588 return enc.sv; 996 return enc.sv;
589} 997}
590 998
… …
595typedef struct 1003typedef struct
596{ 1004{
597 char *cur; // current parser pointer 1005 char *cur; // current parser pointer
598 char *end; // end of input string 1006 char *end; // end of input string
599 const char *err; // parse error, if != 0 1007 const char *err; // parse error, if != 0
600 U32 flags; // F_* 1008 JSON json;
601 U32 depth; // recursion depth 1009 U32 depth; // recursion depth
602 U32 maxdepth; // recursion depth limit 1010 U32 maxdepth; // recursion depth limit
603} dec_t; 1011} dec_t;
604 1012
605inline void 1013INLINE void
1014decode_comment (dec_t *dec)
1015{
1016 // only '#'-style comments allowed a.t.m.
1017
1018 while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
1019 ++dec->cur;
1020}
1021
1022INLINE void
606decode_ws (dec_t *dec) 1023decode_ws (dec_t *dec)
607{ 1024{
608 for (;;) 1025 for (;;)
609 { 1026 {
610 char ch = *dec->cur; 1027 char ch = *dec->cur;
611 1028
612 if (ch > 0x20 1029 if (ch > 0x20)
1030 {
1031 if (expect_false (ch == '#'))
1032 {
1033 if (dec->json.flags & F_RELAXED)
1034 decode_comment (dec);
1035 else
1036 break;
1037 }
1038 else
1039 break;
1040 }
613 || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) 1041 else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
614 break; 1042 break; // parse error, but let higher level handle it, gives better error messages
615 1043
616 ++dec->cur; 1044 ++dec->cur;
617 } 1045 }
618} 1046}
619 1047
… …
623 if (*dec->cur != ch) \ 1051 if (*dec->cur != ch) \
624 ERR (# ch " expected"); \ 1052 ERR (# ch " expected"); \
625 ++dec->cur; \ 1053 ++dec->cur; \
626 SE 1054 SE
627 1055
628#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1056#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
629#define DEC_DEC_DEPTH --dec->depth 1057#define DEC_DEC_DEPTH --dec->depth
630 1058
631static SV *decode_sv (dec_t *dec); 1059static SV *decode_sv (dec_t *dec);
632 1060
633static signed char decode_hexdigit[256]; 1061static signed char decode_hexdigit[256];
… …
725 1153
726 if (hi >= 0x80) 1154 if (hi >= 0x80)
727 { 1155 {
728 utf8 = 1; 1156 utf8 = 1;
729 1157
730 cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); 1158 cur = encode_utf8 (cur, hi);
731 } 1159 }
732 else 1160 else
733 *cur++ = hi; 1161 *cur++ = hi;
734 } 1162 }
735 break; 1163 break;
… …
737 default: 1165 default:
738 --dec_cur; 1166 --dec_cur;
739 ERR ("illegal backslash escape sequence in string"); 1167 ERR ("illegal backslash escape sequence in string");
740 } 1168 }
741 } 1169 }
742 else if (expect_true (ch >= 0x20 && ch <= 0x7f)) 1170 else if (expect_true (ch >= 0x20 && ch < 0x80))
743 *cur++ = ch; 1171 *cur++ = ch;
744 else if (ch >= 0x80) 1172 else if (ch >= 0x80)
745 { 1173 {
746 STRLEN clen; 1174 STRLEN clen;
747 UV uch;
748 1175
749 --dec_cur; 1176 --dec_cur;
750 1177
751 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1178 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
752 if (clen == (STRLEN)-1) 1179 if (clen == (STRLEN)-1)
753 ERR ("malformed UTF-8 character in JSON string"); 1180 ERR ("malformed UTF-8 character in JSON string");
754 1181
755 do 1182 do
756 *cur++ = *dec_cur++; 1183 *cur++ = *dec_cur++;
757 while (--clen); 1184 while (--clen);
758 1185
759 utf8 = 1; 1186 utf8 = 1;
760 } 1187 }
1188 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1189 *cur++ = ch;
761 else 1190 else
762 { 1191 {
763 --dec_cur; 1192 --dec_cur;
764 1193
765 if (!ch) 1194 if (!ch)
… …
773 { 1202 {
774 STRLEN len = cur - buf; 1203 STRLEN len = cur - buf;
775 1204
776 if (sv) 1205 if (sv)
777 { 1206 {
778 SvGROW (sv, SvCUR (sv) + len + 1); 1207 STRLEN cur = SvCUR (sv);
1208
1209 if (SvLEN (sv) - cur <= len)
1210 json_sv_grow (sv, cur, len);
1211
779 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1212 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
780 SvCUR_set (sv, SvCUR (sv) + len); 1213 SvCUR_set (sv, SvCUR (sv) + len);
781 } 1214 }
782 else 1215 else
783 sv = newSVpvn (buf, len); 1216 sv = newSVpvn (buf, len);
… …
868 is_nv = 1; 1301 is_nv = 1;
869 } 1302 }
870 1303
871 if (!is_nv) 1304 if (!is_nv)
872 { 1305 {
1306 int len = dec->cur - start;
1307
873 // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so 1308 // special case the rather common 1..5-digit-int case
874 if (*start == '-') 1309 if (*start == '-')
875 switch (dec->cur - start) 1310 switch (len)
876 { 1311 {
877 case 2: return newSViv (-( start [1] - '0' * 1)); 1312 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
878 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1313 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
879 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1314 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
880 case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1315 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1316 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
881 } 1317 }
882 else 1318 else
883 switch (dec->cur - start) 1319 switch (len)
884 { 1320 {
885 case 1: return newSViv ( start [0] - '0' * 1); 1321 case 1: return newSViv ( start [0] - '0' * 1);
886 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1322 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
887 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1323 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
888 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1324 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1325 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
889 } 1326 }
890 1327
891 { 1328 {
892 UV uv; 1329 UV uv;
893 int numtype = grok_number (start, dec->cur - start, &uv); 1330 int numtype = grok_number (start, len, &uv);
894 if (numtype & IS_NUMBER_IN_UV) 1331 if (numtype & IS_NUMBER_IN_UV)
895 if (numtype & IS_NUMBER_NEG) 1332 if (numtype & IS_NUMBER_NEG)
896 { 1333 {
897 if (uv < (UV)IV_MIN) 1334 if (uv < (UV)IV_MIN)
898 return newSViv (-(IV)uv); 1335 return newSViv (-(IV)uv);
899 } 1336 }
900 else 1337 else
901 return newSVuv (uv); 1338 return newSVuv (uv);
902
903 // here would likely be the place for bigint support
904 } 1339 }
905 }
906 1340
907 // if we ever support bigint or bigfloat, this is the place for bigfloat 1341 len -= *start == '-' ? 1 : 0;
1342
1343 // does not fit into IV or UV, try NV
1344 if (len <= NV_DIG)
1345 // fits into NV without loss of precision
1346 return newSVnv (json_atof (start));
1347
1348 // everything else fails, convert it to a string
1349 return newSVpvn (start, dec->cur - start);
1350 }
1351
1352 // loss of precision here
908 return newSVnv (Atof (start)); 1353 return newSVnv (json_atof (start));
909 1354
910fail: 1355fail:
911 return 0; 1356 return 0;
912} 1357}
913 1358
… …
937 if (*dec->cur == ']') 1382 if (*dec->cur == ']')
938 { 1383 {
939 ++dec->cur; 1384 ++dec->cur;
940 break; 1385 break;
941 } 1386 }
942 1387
943 if (*dec->cur != ',') 1388 if (*dec->cur != ',')
944 ERR (", or ] expected while parsing array"); 1389 ERR (", or ] expected while parsing array");
945 1390
946 ++dec->cur; 1391 ++dec->cur;
1392
1393 decode_ws (dec);
1394
1395 if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
1396 {
1397 ++dec->cur;
1398 break;
1399 }
947 } 1400 }
948 1401
949 DEC_DEC_DEPTH; 1402 DEC_DEC_DEPTH;
950 return newRV_noinc ((SV *)av); 1403 return newRV_noinc ((SV *)av);
951 1404
… …
956} 1409}
957 1410
958static SV * 1411static SV *
959decode_hv (dec_t *dec) 1412decode_hv (dec_t *dec)
960{ 1413{
1414 SV *sv;
961 HV *hv = newHV (); 1415 HV *hv = newHV ();
962 1416
963 DEC_INC_DEPTH; 1417 DEC_INC_DEPTH;
964 decode_ws (dec); 1418 decode_ws (dec);
965 1419
966 if (*dec->cur == '}') 1420 if (*dec->cur == '}')
967 ++dec->cur; 1421 ++dec->cur;
968 else 1422 else
969 for (;;) 1423 for (;;)
970 { 1424 {
1425 EXPECT_CH ('"');
1426
1427 // heuristic: assume that
1428 // a) decode_str + hv_store_ent are abysmally slow.
1429 // b) most hash keys are short, simple ascii text.
1430 // => try to "fast-match" such strings to avoid
1431 // the overhead of decode_str + hv_store_ent.
1432 {
971 SV *key, *value; 1433 SV *value;
1434 char *p = dec->cur;
1435 char *e = p + 24; // only try up to 24 bytes
972 1436
973 decode_ws (dec); EXPECT_CH ('"'); 1437 for (;;)
974
975 key = decode_str (dec);
976 if (!key)
977 goto fail;
978
979 decode_ws (dec); EXPECT_CH (':');
980
981 value = decode_sv (dec);
982 if (!value)
983 { 1438 {
1439 // the >= 0x80 is false on most architectures
1440 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1441 {
1442 // slow path, back up and use decode_str
1443 SV *key = decode_str (dec);
1444 if (!key)
1445 goto fail;
1446
1447 decode_ws (dec); EXPECT_CH (':');
1448
1449 decode_ws (dec);
1450 value = decode_sv (dec);
1451 if (!value)
1452 {
1453 SvREFCNT_dec (key);
1454 goto fail;
1455 }
1456
1457 hv_store_ent (hv, key, value, 0);
984 SvREFCNT_dec (key); 1458 SvREFCNT_dec (key);
1459
1460 break;
1461 }
1462 else if (*p == '"')
1463 {
1464 // fast path, got a simple key
1465 char *key = dec->cur;
1466 int len = p - key;
1467 dec->cur = p + 1;
1468
1469 decode_ws (dec); EXPECT_CH (':');
1470
1471 decode_ws (dec);
1472 value = decode_sv (dec);
1473 if (!value)
985 goto fail; 1474 goto fail;
1475
1476 hv_store (hv, key, len, value, 0);
1477
1478 break;
1479 }
1480
1481 ++p;
986 } 1482 }
987 1483 }
988 hv_store_ent (hv, key, value, 0);
989 SvREFCNT_dec (key);
990 1484
991 decode_ws (dec); 1485 decode_ws (dec);
992 1486
993 if (*dec->cur == '}') 1487 if (*dec->cur == '}')
994 { 1488 {
… …
998 1492
999 if (*dec->cur != ',') 1493 if (*dec->cur != ',')
1000 ERR (", or } expected while parsing object/hash"); 1494 ERR (", or } expected while parsing object/hash");
1001 1495
1002 ++dec->cur; 1496 ++dec->cur;
1497
1498 decode_ws (dec);
1499
1500 if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
1501 {
1502 ++dec->cur;
1503 break;
1504 }
1003 } 1505 }
1004 1506
1005 DEC_DEC_DEPTH; 1507 DEC_DEC_DEPTH;
1006 return newRV_noinc ((SV *)hv); 1508 sv = newRV_noinc ((SV *)hv);
1509
1510 // check filter callbacks
1511 if (expect_false (dec->json.flags & F_HOOK))
1512 {
1513 if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1514 {
1515 HE *cb, *he;
1516
1517 hv_iterinit (hv);
1518 he = hv_iternext (hv);
1519 hv_iterinit (hv);
1520
1521 // the next line creates a mortal sv each time it's called.
1522 // might want to optimise this for common cases.
1523 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1524
1525 if (cb)
1526 {
1527 dSP;
1528 int count;
1529
1530 ENTER; SAVETMPS;
1531 PUSHMARK (SP);
1532 XPUSHs (HeVAL (he));
1533 sv_2mortal (sv);
1534
1535 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1536
1537 if (count == 1)
1538 {
1539 sv = newSVsv (POPs);
1540 PUTBACK;
1541 FREETMPS; LEAVE;
1542 return sv;
1543 }
1544 else if (count)
1545 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1546
1547 PUTBACK;
1548
1549 SvREFCNT_inc (sv);
1550
1551 FREETMPS; LEAVE;
1552 }
1553 }
1554
1555 if (dec->json.cb_object)
1556 {
1557 dSP;
1558 int count;
1559
1560 ENTER; SAVETMPS;
1561 PUSHMARK (SP);
1562 XPUSHs (sv_2mortal (sv));
1563
1564 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1565
1566 if (count == 1)
1567 sv = newSVsv (POPs);
1568 else if (count == 0)
1569 SvREFCNT_inc (sv);
1570 else
1571 croak ("filter_json_object callbacks must not return more than one scalar");
1572
1573 PUTBACK;
1574
1575 FREETMPS; LEAVE;
1576 }
1577 }
1578
1579 return sv;
1007 1580
1008fail: 1581fail:
1009 SvREFCNT_dec (hv); 1582 SvREFCNT_dec (hv);
1010 DEC_DEC_DEPTH; 1583 DEC_DEC_DEPTH;
1011 return 0; 1584 return 0;
1012} 1585}
1013 1586
1014static SV * 1587static SV *
1588decode_tag (dec_t *dec)
1589{
1590 SV *tag = 0;
1591 SV *val = 0;
1592
1593 if (!(dec->json.flags & F_ALLOW_TAGS))
1594 ERR ("malformed JSON string, neither array, object, number, string or atom");
1595
1596 ++dec->cur;
1597
1598 decode_ws (dec);
1599
1600 tag = decode_sv (dec);
1601 if (!tag)
1602 goto fail;
1603
1604 if (!SvPOK (tag))
1605 ERR ("malformed JSON string, (tag) must be a string");
1606
1607 decode_ws (dec);
1608
1609 if (*dec->cur != ')')
1610 ERR (") expected after tag");
1611
1612 ++dec->cur;
1613
1614 decode_ws (dec);
1615
1616 val = decode_sv (dec);
1617 if (!val)
1618 goto fail;
1619
1620 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1621 ERR ("malformed JSON string, tag value must be an array");
1622
1623 {
1624 AV *av = (AV *)SvRV (val);
1625 int i, len = av_len (av) + 1;
1626 HV *stash = gv_stashsv (tag, 0);
1627 SV *sv;
1628
1629 if (!stash)
1630 ERR ("cannot decode perl-object (package does not exist)");
1631
1632 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1633
1634 if (!method)
1635 ERR ("cannot decode perl-object (package does not have a THAW method)");
1636
1637 dSP;
1638
1639 ENTER; SAVETMPS;
1640 PUSHMARK (SP);
1641 EXTEND (SP, len + 2);
1642 // we re-bless the reference to get overload and other niceties right
1643 PUSHs (tag);
1644 PUSHs (sv_json);
1645
1646 for (i = 0; i < len; ++i)
1647 PUSHs (*av_fetch (av, i, 1));
1648
1649 PUTBACK;
1650 call_sv ((SV *)GvCV (method), G_SCALAR);
1651 SPAGAIN;
1652
1653 SvREFCNT_dec (tag);
1654 SvREFCNT_dec (val);
1655 sv = SvREFCNT_inc (POPs);
1656
1657 PUTBACK;
1658
1659 FREETMPS; LEAVE;
1660
1661 return sv;
1662 }
1663
1664fail:
1665 SvREFCNT_dec (tag);
1666 SvREFCNT_dec (val);
1667 return 0;
1668}
1669
1670static SV *
1015decode_sv (dec_t *dec) 1671decode_sv (dec_t *dec)
1016{ 1672{
1017 decode_ws (dec);
1018
1019 // the beauty of JSON: you need exactly one character lookahead 1673 // the beauty of JSON: you need exactly one character lookahead
1020 // to parse anything. 1674 // to parse everything.
1021 switch (*dec->cur) 1675 switch (*dec->cur)
1022 { 1676 {
1023 case '"': ++dec->cur; return decode_str (dec); 1677 case '"': ++dec->cur; return decode_str (dec);
1024 case '[': ++dec->cur; return decode_av (dec); 1678 case '[': ++dec->cur; return decode_av (dec);
1025 case '{': ++dec->cur; return decode_hv (dec); 1679 case '{': ++dec->cur; return decode_hv (dec);
1680 case '(': return decode_tag (dec);
1026 1681
1027 case '-': 1682 case '-':
1028 case '0': case '1': case '2': case '3': case '4': 1683 case '0': case '1': case '2': case '3': case '4':
1029 case '5': case '6': case '7': case '8': case '9': 1684 case '5': case '6': case '7': case '8': case '9':
1030 return decode_num (dec); 1685 return decode_num (dec);
1031 1686
1687 case 'f':
1688 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1689 {
1690 dec->cur += 5;
1691
1692 if (expect_false (!dec->json.v_false))
1693 dec->json.v_false = GET_BOOL (false);
1694
1695 return newSVsv (dec->json.v_false);
1696 }
1697 else
1698 ERR ("'false' expected");
1699
1700 break;
1701
1032 case 't': 1702 case 't':
1033 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1703 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1034 { 1704 {
1035 dec->cur += 4; 1705 dec->cur += 4;
1706
1707 if (expect_false (!dec->json.v_true))
1708 dec->json.v_true = GET_BOOL (true);
1709
1036 return newSViv (1); 1710 return newSVsv (dec->json.v_true);
1037 } 1711 }
1038 else 1712 else
1039 ERR ("'true' expected"); 1713 ERR ("'true' expected");
1040
1041 break;
1042
1043 case 'f':
1044 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1045 {
1046 dec->cur += 5;
1047 return newSViv (0);
1048 }
1049 else
1050 ERR ("'false' expected");
1051 1714
1052 break; 1715 break;
1053 1716
1054 case 'n': 1717 case 'n':
1055 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1718 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
… …
1061 ERR ("'null' expected"); 1724 ERR ("'null' expected");
1062 1725
1063 break; 1726 break;
1064 1727
1065 default: 1728 default:
1066 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1729 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1067 break; 1730 break;
1068 } 1731 }
1069 1732
1070fail: 1733fail:
1071 return 0; 1734 return 0;
1072} 1735}
1073 1736
1074static SV * 1737static SV *
1075decode_json (SV *string, U32 flags, UV *offset_return) 1738decode_json (SV *string, JSON *json, STRLEN *offset_return)
1076{ 1739{
1077 dec_t dec; 1740 dec_t dec;
1078 UV offset;
1079 SV *sv; 1741 SV *sv;
1080 1742
1743 /* work around bugs in 5.10 where manipulating magic values
1744 * makes perl ignore the magic in subsequent accesses.
1745 * also make a copy of non-PV values, to get them into a clean
1746 * state (SvPV should do that, but it's buggy, see below).
1747 *
1748 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1749 * as reported by Reini Urban.
1750 */
1081 SvGETMAGIC (string); 1751 /*SvGETMAGIC (string);*/
1752 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1753 string = sv_2mortal (newSVsv (string));
1754
1082 SvUPGRADE (string, SVt_PV); 1755 SvUPGRADE (string, SVt_PV);
1083 1756
1084 if (flags & F_UTF8) 1757 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1758 * assertion with -DDEBUGGING, although SvCUR is documented to
1759 * return the xpv_cur field which certainly exists after upgrading.
1760 * according to nicholas clark, calling SvPOK fixes this.
1761 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1762 * and hope for the best.
1763 * Damnit, SvPV_nolen still trips over yet another assertion. This
1764 * assertion business is seriously broken, try yet another workaround
1765 * for the broken -DDEBUGGING.
1766 */
1767 {
1768#ifdef DEBUGGING
1769 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1770#else
1771 STRLEN offset = SvCUR (string);
1772#endif
1773
1774 if (offset > json->max_size && json->max_size)
1775 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1776 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1777 }
1778
1779 if (DECODE_WANTS_OCTETS (json))
1085 sv_utf8_downgrade (string, 0); 1780 sv_utf8_downgrade (string, 0);
1086 else 1781 else
1087 sv_utf8_upgrade (string); 1782 sv_utf8_upgrade (string);
1088 1783
1089 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1784 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1090 1785
1091 dec.flags = flags; 1786 dec.json = *json;
1092 dec.cur = SvPVX (string); 1787 dec.cur = SvPVX (string);
1093 dec.end = SvEND (string); 1788 dec.end = SvEND (string);
1094 dec.err = 0; 1789 dec.err = 0;
1095 dec.depth = 0; 1790 dec.depth = 0;
1096 dec.maxdepth = DEC_DEPTH (dec.flags); 1791
1792 if (dec.json.cb_object || dec.json.cb_sk_object)
1793 dec.json.flags |= F_HOOK;
1097 1794
1098 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1795 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1796
1797 decode_ws (&dec);
1099 sv = decode_sv (&dec); 1798 sv = decode_sv (&dec);
1100 1799
1101 if (!(offset_return || !sv)) 1800 if (offset_return)
1801 *offset_return = dec.cur - SvPVX (string);
1802 else if (sv)
1102 { 1803 {
1103 // check for trailing garbage 1804 // check for trailing garbage
1104 decode_ws (&dec); 1805 decode_ws (&dec);
1105 1806
1106 if (*dec.cur) 1807 if (dec.cur != dec.end)
1107 { 1808 {
1108 dec.err = "garbage after JSON object"; 1809 dec.err = "garbage after JSON object";
1109 SvREFCNT_dec (sv); 1810 SvREFCNT_dec (sv);
1110 sv = 0; 1811 sv = 0;
1111 } 1812 }
1112 }
1113
1114 if (offset_return || !sv)
1115 {
1116 offset = dec.flags & F_UTF8
1117 ? dec.cur - SvPVX (string)
1118 : utf8_distance (dec.cur, SvPVX (string));
1119
1120 if (offset_return)
1121 *offset_return = offset;
1122 } 1813 }
1123 1814
1124 if (!sv) 1815 if (!sv)
1125 { 1816 {
1126 SV *uni = sv_newmortal (); 1817 SV *uni = sv_newmortal ();
… …
1132 SAVEVPTR (PL_curcop); 1823 SAVEVPTR (PL_curcop);
1133 PL_curcop = &cop; 1824 PL_curcop = &cop;
1134 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1825 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1135 LEAVE; 1826 LEAVE;
1136 1827
1137 croak ("%s, at character offset %d [\"%s\"]", 1828 croak ("%s, at character offset %d (before \"%s\")",
1138 dec.err, 1829 dec.err,
1139 (int)offset, 1830 (int)ptr_to_index (string, dec.cur),
1140 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1831 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1141 } 1832 }
1142 1833
1143 sv = sv_2mortal (sv); 1834 sv = sv_2mortal (sv);
1144 1835
1145 if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1836 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1146 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1837 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1147 1838
1148 return sv; 1839 return sv;
1840}
1841
1842/////////////////////////////////////////////////////////////////////////////
1843// incremental parser
1844
1845static void
1846incr_parse (JSON *self)
1847{
1848 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1849
1850 // the state machine here is a bit convoluted and could be simplified a lot
1851 // but this would make it slower, so...
1852
1853 for (;;)
1854 {
1855 switch (self->incr_mode)
1856 {
1857 // reached end of a scalar, see if we are inside a nested structure or not
1858 end_of_scalar:
1859 self->incr_mode = INCR_M_JSON;
1860
1861 if (self->incr_nest) // end of a scalar inside array, object or tag
1862 goto incr_m_json;
1863 else // end of scalar outside structure, json text ends here
1864 goto interrupt;
1865
1866 // only used for initial whitespace skipping
1867 case INCR_M_WS:
1868 for (;;)
1869 {
1870 if (*p > 0x20)
1871 {
1872 if (*p == '#')
1873 {
1874 self->incr_mode = INCR_M_C0;
1875 goto incr_m_c;
1876 }
1877 else
1878 {
1879 self->incr_mode = INCR_M_JSON;
1880 goto incr_m_json;
1881 }
1882 }
1883 else if (!*p)
1884 goto interrupt;
1885
1886 ++p;
1887 }
1888
1889 // skip a single char inside a string (for \\-processing)
1890 case INCR_M_BS:
1891 if (!*p)
1892 goto interrupt;
1893
1894 ++p;
1895 self->incr_mode = INCR_M_STR;
1896 goto incr_m_str;
1897
1898 // inside #-style comments
1899 case INCR_M_C0:
1900 case INCR_M_C1:
1901 incr_m_c:
1902 for (;;)
1903 {
1904 if (*p == '\n')
1905 {
1906 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1907 break;
1908 }
1909 else if (!*p)
1910 goto interrupt;
1911
1912 ++p;
1913 }
1914
1915 break;
1916
1917 // inside true/false/null
1918 case INCR_M_TFN:
1919 incr_m_tfn:
1920 for (;;)
1921 switch (*p++)
1922 {
1923 case 'r': case 'u': case 'e': // tRUE, falsE, nUll
1924 case 'a': case 'l': case 's': // fALSe, nuLL
1925 // allowed
1926 break;
1927
1928 default:
1929 --p;
1930 goto end_of_scalar;
1931 }
1932
1933 // inside a number
1934 case INCR_M_NUM:
1935 incr_m_num:
1936 for (;;)
1937 switch (*p++)
1938 {
1939 case 'e': case 'E': case '.': case '+':
1940 case '-':
1941 case '0': case '1': case '2': case '3': case '4':
1942 case '5': case '6': case '7': case '8': case '9':
1943 // allowed
1944 break;
1945
1946 default:
1947 --p;
1948 goto end_of_scalar;
1949 }
1950
1951 // inside a string
1952 case INCR_M_STR:
1953 incr_m_str:
1954 for (;;)
1955 {
1956 if (*p == '"')
1957 {
1958 ++p;
1959 goto end_of_scalar;
1960 }
1961 else if (*p == '\\')
1962 {
1963 ++p; // "virtually" consumes character after \
1964
1965 if (!*p) // if at end of string we have to switch modes
1966 {
1967 self->incr_mode = INCR_M_BS;
1968 goto interrupt;
1969 }
1970 }
1971 else if (!*p)
1972 goto interrupt;
1973
1974 ++p;
1975 }
1976
1977 // after initial ws, outside string
1978 case INCR_M_JSON:
1979 incr_m_json:
1980 for (;;)
1981 {
1982 switch (*p++)
1983 {
1984 case 0:
1985 --p;
1986 goto interrupt;
1987
1988 case 0x09:
1989 case 0x0a:
1990 case 0x0d:
1991 case 0x20:
1992 if (!self->incr_nest)
1993 {
1994 --p; // do not eat the whitespace, let the next round do it
1995 goto interrupt;
1996 }
1997 break;
1998
1999 // the following three blocks handle scalars. this makes the parser
2000 // more strict than required inside arrays or objects, and could
2001 // be moved to a special case on the toplevel (except strings)
2002 case 't':
2003 case 'f':
2004 case 'n':
2005 self->incr_mode = INCR_M_TFN;
2006 goto incr_m_tfn;
2007
2008 case '-':
2009 case '0': case '1': case '2': case '3': case '4':
2010 case '5': case '6': case '7': case '8': case '9':
2011 self->incr_mode = INCR_M_NUM;
2012 goto incr_m_num;
2013
2014 case '"':
2015 self->incr_mode = INCR_M_STR;
2016 goto incr_m_str;
2017
2018 case '[':
2019 case '{':
2020 case '(':
2021 if (++self->incr_nest > self->max_depth)
2022 croak (ERR_NESTING_EXCEEDED);
2023 break;
2024
2025 case ']':
2026 case '}':
2027 if (--self->incr_nest <= 0)
2028 goto interrupt;
2029 break;
2030
2031 case ')':
2032 --self->incr_nest;
2033 break;
2034
2035 case '#':
2036 self->incr_mode = INCR_M_C1;
2037 goto incr_m_c;
2038 }
2039 }
2040 }
2041
2042 modechange:
2043 ;
2044 }
2045
2046interrupt:
2047 self->incr_pos = p - SvPVX (self->incr_text);
2048 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
2049 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1149} 2050}
1150 2051
1151///////////////////////////////////////////////////////////////////////////// 2052/////////////////////////////////////////////////////////////////////////////
1152// XS interface functions 2053// XS interface functions
1153 2054
… …
1162 i >= '0' && i <= '9' ? i - '0' 2063 i >= '0' && i <= '9' ? i - '0'
1163 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2064 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1164 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2065 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1165 : -1; 2066 : -1;
1166 2067
1167 json_stash = gv_stashpv ("JSON::XS", 1); 2068 json_stash = gv_stashpv ("JSON::XS" , 1);
2069 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
2070 bool_false = get_bool ("Types::Serialiser::false");
2071 bool_true = get_bool ("Types::Serialiser::true");
2072
2073 sv_json = newSVpv ("JSON", 0);
2074 SvREADONLY_on (sv_json);
2075
2076 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1168} 2077}
1169 2078
1170PROTOTYPES: DISABLE 2079PROTOTYPES: DISABLE
1171 2080
1172SV *new (char *dummy) 2081void CLONE (...)
1173 CODE: 2082 CODE:
1174 RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); 2083 // as long as these writes are atomic, the race should not matter
2084 // as existing threads either already use 0, or use the old value,
2085 // which is sitll correct for the initial thread.
2086 json_stash = 0;
2087 bool_stash = 0;
2088 bool_false = 0;
2089 bool_true = 0;
2090
2091void new (char *klass)
2092 PPCODE:
2093{
2094 SV *pv = NEWSV (0, sizeof (JSON));
2095 SvPOK_only (pv);
2096 json_init ((JSON *)SvPVX (pv));
2097 XPUSHs (sv_2mortal (sv_bless (
2098 newRV_noinc (pv),
2099 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
2100 )));
2101}
2102
2103void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2104 PPCODE:
2105 self->v_false = newSVsv (v_false);
2106 self->v_true = newSVsv (v_true);
2107 XPUSHs (ST (0));
2108
2109void get_boolean_values (JSON *self)
2110 PPCODE:
2111 if (self->v_false && self->v_true)
2112 {
2113 EXTEND (SP, 2);
2114 PUSHs (self->v_false);
2115 PUSHs (self->v_true);
2116 }
2117
2118void ascii (JSON *self, int enable = 1)
2119 ALIAS:
2120 ascii = F_ASCII
2121 latin1 = F_LATIN1
2122 utf8 = F_UTF8
2123 indent = F_INDENT
2124 canonical = F_CANONICAL
2125 space_before = F_SPACE_BEFORE
2126 space_after = F_SPACE_AFTER
2127 pretty = F_PRETTY
2128 allow_nonref = F_ALLOW_NONREF
2129 shrink = F_SHRINK
2130 allow_blessed = F_ALLOW_BLESSED
2131 convert_blessed = F_CONV_BLESSED
2132 relaxed = F_RELAXED
2133 allow_unknown = F_ALLOW_UNKNOWN
2134 allow_tags = F_ALLOW_TAGS
2135 PPCODE:
2136{
2137 if (enable)
2138 self->flags |= ix;
2139 else
2140 self->flags &= ~ix;
2141
2142 XPUSHs (ST (0));
2143}
2144
2145void get_ascii (JSON *self)
2146 ALIAS:
2147 get_ascii = F_ASCII
2148 get_latin1 = F_LATIN1
2149 get_utf8 = F_UTF8
2150 get_indent = F_INDENT
2151 get_canonical = F_CANONICAL
2152 get_space_before = F_SPACE_BEFORE
2153 get_space_after = F_SPACE_AFTER
2154 get_allow_nonref = F_ALLOW_NONREF
2155 get_shrink = F_SHRINK
2156 get_allow_blessed = F_ALLOW_BLESSED
2157 get_convert_blessed = F_CONV_BLESSED
2158 get_relaxed = F_RELAXED
2159 get_allow_unknown = F_ALLOW_UNKNOWN
2160 get_allow_tags = F_ALLOW_TAGS
2161 PPCODE:
2162 XPUSHs (boolSV (self->flags & ix));
2163
2164void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
2165 PPCODE:
2166 self->max_depth = max_depth;
2167 XPUSHs (ST (0));
2168
2169U32 get_max_depth (JSON *self)
2170 CODE:
2171 RETVAL = self->max_depth;
1175 OUTPUT: 2172 OUTPUT:
1176 RETVAL 2173 RETVAL
1177 2174
1178SV *ascii (SV *self, int enable = 1) 2175void max_size (JSON *self, U32 max_size = 0)
1179 ALIAS: 2176 PPCODE:
1180 ascii = F_ASCII 2177 self->max_size = max_size;
1181 latin1 = F_LATIN1 2178 XPUSHs (ST (0));
1182 utf8 = F_UTF8 2179
1183 indent = F_INDENT 2180int get_max_size (JSON *self)
1184 canonical = F_CANONICAL
1185 space_before = F_SPACE_BEFORE
1186 space_after = F_SPACE_AFTER
1187 pretty = F_PRETTY
1188 allow_nonref = F_ALLOW_NONREF
1189 shrink = F_SHRINK
1190 CODE: 2181 CODE:
1191{ 2182 RETVAL = self->max_size;
1192 UV *uv = SvJSON (self);
1193 if (enable)
1194 *uv |= ix;
1195 else
1196 *uv &= ~ix;
1197
1198 RETVAL = newSVsv (self);
1199}
1200 OUTPUT: 2183 OUTPUT:
1201 RETVAL 2184 RETVAL
1202 2185
1203SV *max_depth (SV *self, UV max_depth = 0x80000000UL) 2186void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
2187 PPCODE:
2188{
2189 SvREFCNT_dec (self->cb_object);
2190 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
2191
2192 XPUSHs (ST (0));
2193}
2194
2195void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
2196 PPCODE:
2197{
2198 if (!self->cb_sk_object)
2199 self->cb_sk_object = newHV ();
2200
2201 if (SvOK (cb))
2202 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
2203 else
2204 {
2205 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
2206
2207 if (!HvKEYS (self->cb_sk_object))
2208 {
2209 SvREFCNT_dec (self->cb_sk_object);
2210 self->cb_sk_object = 0;
2211 }
2212 }
2213
2214 XPUSHs (ST (0));
2215}
2216
2217void encode (JSON *self, SV *scalar)
2218 PPCODE:
2219 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2220 XPUSHs (scalar);
2221
2222void decode (JSON *self, SV *jsonstr)
2223 PPCODE:
2224 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2225 XPUSHs (jsonstr);
2226
2227void decode_prefix (JSON *self, SV *jsonstr)
2228 PPCODE:
2229{
2230 SV *sv;
2231 STRLEN offset;
2232 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
2233 EXTEND (SP, 2);
2234 PUSHs (sv);
2235 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
2236}
2237
2238void incr_parse (JSON *self, SV *jsonstr = 0)
2239 PPCODE:
2240{
2241 if (!self->incr_text)
2242 self->incr_text = newSVpvn ("", 0);
2243
2244 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2245 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2246 if (DECODE_WANTS_OCTETS (self))
2247 {
2248 if (self->incr_pos)
2249 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2250 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2251
2252 sv_utf8_downgrade (self->incr_text, 0);
2253 }
2254 else
2255 {
2256 sv_utf8_upgrade (self->incr_text);
2257
2258 if (self->incr_pos)
2259 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2260 - (U8 *)SvPVX (self->incr_text);
2261 }
2262
2263 // append data, if any
2264 if (jsonstr)
2265 {
2266 /* make sure both strings have same encoding */
2267 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2268 if (SvUTF8 (jsonstr))
2269 sv_utf8_downgrade (jsonstr, 0);
2270 else
2271 sv_utf8_upgrade (jsonstr);
2272
2273 /* and then just blindly append */
2274 {
2275 STRLEN len;
2276 const char *str = SvPV (jsonstr, len);
2277 STRLEN cur = SvCUR (self->incr_text);
2278
2279 if (SvLEN (self->incr_text) - cur <= len)
2280 json_sv_grow (self->incr_text, cur, len);
2281
2282 Move (str, SvEND (self->incr_text), len, char);
2283 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
2284 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
2285 }
2286 }
2287
2288 if (GIMME_V != G_VOID)
2289 do
2290 {
2291 SV *sv;
2292 STRLEN offset;
2293
2294 if (!INCR_DONE (self))
2295 {
2296 incr_parse (self);
2297
2298 if (self->incr_pos > self->max_size && self->max_size)
2299 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2300 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2301
2302 if (!INCR_DONE (self))
2303 {
2304 // as an optimisation, do not accumulate white space in the incr buffer
2305 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2306 {
2307 self->incr_pos = 0;
2308 SvCUR_set (self->incr_text, 0);
2309 }
2310
2311 break;
2312 }
2313 }
2314
2315 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2316 XPUSHs (sv);
2317
2318 self->incr_pos -= offset;
2319 self->incr_nest = 0;
2320 self->incr_mode = 0;
2321
2322 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
2323 }
2324 while (GIMME_V == G_ARRAY);
2325}
2326
2327SV *incr_text (JSON *self)
2328 ATTRS: lvalue
1204 CODE: 2329 CODE:
1205{ 2330{
1206 UV *uv = SvJSON (self); 2331 if (self->incr_pos)
1207 UV log2 = 0; 2332 croak ("incr_text can not be called when the incremental parser already started parsing");
1208 2333
1209 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; 2334 RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1210
1211 while ((1UL << log2) < max_depth)
1212 ++log2;
1213
1214 *uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1215
1216 RETVAL = newSVsv (self);
1217} 2335}
1218 OUTPUT: 2336 OUTPUT:
1219 RETVAL 2337 RETVAL
1220 2338
2339void incr_skip (JSON *self)
2340 CODE:
2341{
2342 if (self->incr_pos)
2343 {
2344 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
2345 self->incr_pos = 0;
2346 self->incr_nest = 0;
2347 self->incr_mode = 0;
2348 }
2349}
2350
2351void incr_reset (JSON *self)
2352 CODE:
2353{
2354 SvREFCNT_dec (self->incr_text);
2355 self->incr_text = 0;
2356 self->incr_pos = 0;
2357 self->incr_nest = 0;
2358 self->incr_mode = 0;
2359}
2360
2361void DESTROY (JSON *self)
2362 CODE:
2363 SvREFCNT_dec (self->v_false);
2364 SvREFCNT_dec (self->v_true);
2365 SvREFCNT_dec (self->cb_sk_object);
2366 SvREFCNT_dec (self->cb_object);
2367 SvREFCNT_dec (self->incr_text);
2368
2369PROTOTYPES: ENABLE
2370
1221void encode (SV *self, SV *scalar) 2371void encode_json (SV *scalar)
1222 PPCODE: 2372 PPCODE:
1223 XPUSHs (encode_json (scalar, *SvJSON (self))); 2373{
2374 JSON json;
2375 json_init (&json);
2376 json.flags |= F_UTF8;
2377 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2378 XPUSHs (scalar);
2379}
1224 2380
1225void decode (SV *self, SV *jsonstr) 2381void decode_json (SV *jsonstr)
1226 PPCODE: 2382 PPCODE:
1227 XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
1228
1229void decode_prefix (SV *self, SV *jsonstr)
1230 PPCODE:
1231{ 2383{
1232 UV offset; 2384 JSON json;
1233 EXTEND (SP, 2); 2385 json_init (&json);
1234 PUSHs (decode_json (jsonstr, *SvJSON (self), &offset)); 2386 json.flags |= F_UTF8;
1235 PUSHs (sv_2mortal (newSVuv (offset))); 2387 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2388 XPUSHs (jsonstr);
1236} 2389}
1237 2390
1238PROTOTYPES: ENABLE
1239
1240void to_json (SV *scalar)
1241 ALIAS:
1242 objToJson = 0
1243 PPCODE:
1244 XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1245
1246void from_json (SV *jsonstr)
1247 ALIAS:
1248 jsonToObj = 0
1249 PPCODE:
1250 XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1251

Diff Legend

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