]>
Commit | Line | Data |
---|---|---|
15108778 | 1 | #define VERSION "1.0.3" |
2 | ||
3 | /* CJSON - JSON support for Lua | |
4 | * | |
5 | * Copyright (c) 2010-2011 Mark Pulford <mark@kyne.com.au> | |
6 | * | |
7 | * Permission is hereby granted, free of charge, to any person obtaining | |
8 | * a copy of this software and associated documentation files (the | |
9 | * "Software"), to deal in the Software without restriction, including | |
10 | * without limitation the rights to use, copy, modify, merge, publish, | |
11 | * distribute, sublicense, and/or sell copies of the Software, and to | |
12 | * permit persons to whom the Software is furnished to do so, subject to | |
13 | * the following conditions: | |
14 | * | |
15 | * The above copyright notice and this permission notice shall be | |
16 | * included in all copies or substantial portions of the Software. | |
17 | * | |
18 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, | |
19 | * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF | |
20 | * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. | |
21 | * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY | |
22 | * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, | |
23 | * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE | |
24 | * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. | |
25 | */ | |
26 | ||
27 | /* Caveats: | |
28 | * - JSON "null" values are represented as lightuserdata since Lua | |
29 | * tables cannot contain "nil". Compare with cjson.null. | |
30 | * - Invalid UTF-8 characters are not detected and will be passed | |
31 | * untouched. If required, UTF-8 error checking should be done | |
32 | * outside this library. | |
33 | * - Javascript comments are not part of the JSON spec, and are not | |
34 | * currently supported. | |
35 | * | |
36 | * Note: Decoding is slower than encoding. Lua spends significant | |
37 | * time (30%) managing tables when parsing JSON since it is | |
38 | * difficult to know object/array sizes ahead of time. | |
39 | */ | |
40 | ||
41 | #include <assert.h> | |
42 | #include <string.h> | |
43 | #include <math.h> | |
44 | #include <lua.h> | |
45 | #include <lauxlib.h> | |
46 | ||
47 | #include "strbuf.h" | |
48 | ||
49 | #ifdef MISSING_ISINF | |
50 | #define isinf(x) (!isnan(x) && isnan((x) - (x))) | |
51 | #endif | |
52 | ||
53 | #define DEFAULT_SPARSE_CONVERT 0 | |
54 | #define DEFAULT_SPARSE_RATIO 2 | |
55 | #define DEFAULT_SPARSE_SAFE 10 | |
56 | #define DEFAULT_MAX_DEPTH 20 | |
57 | #define DEFAULT_ENCODE_REFUSE_BADNUM 1 | |
58 | #define DEFAULT_DECODE_REFUSE_BADNUM 0 | |
59 | #define DEFAULT_ENCODE_KEEP_BUFFER 1 | |
60 | ||
61 | typedef enum { | |
62 | T_OBJ_BEGIN, | |
63 | T_OBJ_END, | |
64 | T_ARR_BEGIN, | |
65 | T_ARR_END, | |
66 | T_STRING, | |
67 | T_NUMBER, | |
68 | T_BOOLEAN, | |
69 | T_NULL, | |
70 | T_COLON, | |
71 | T_COMMA, | |
72 | T_END, | |
73 | T_WHITESPACE, | |
74 | T_ERROR, | |
75 | T_UNKNOWN | |
76 | } json_token_type_t; | |
77 | ||
78 | static const char *json_token_type_name[] = { | |
79 | "T_OBJ_BEGIN", | |
80 | "T_OBJ_END", | |
81 | "T_ARR_BEGIN", | |
82 | "T_ARR_END", | |
83 | "T_STRING", | |
84 | "T_NUMBER", | |
85 | "T_BOOLEAN", | |
86 | "T_NULL", | |
87 | "T_COLON", | |
88 | "T_COMMA", | |
89 | "T_END", | |
90 | "T_WHITESPACE", | |
91 | "T_ERROR", | |
92 | "T_UNKNOWN", | |
93 | NULL | |
94 | }; | |
95 | ||
96 | typedef struct { | |
97 | json_token_type_t ch2token[256]; | |
98 | char escape2char[256]; /* Decoding */ | |
99 | #if 0 | |
100 | char escapes[35][8]; /* Pre-generated escape string buffer */ | |
101 | char *char2escape[256]; /* Encoding */ | |
102 | #endif | |
103 | strbuf_t encode_buf; | |
104 | char number_fmt[8]; /* "%.XXg\0" */ | |
105 | int current_depth; | |
106 | ||
107 | int encode_sparse_convert; | |
108 | int encode_sparse_ratio; | |
109 | int encode_sparse_safe; | |
110 | int encode_max_depth; | |
111 | int encode_refuse_badnum; | |
112 | int decode_refuse_badnum; | |
113 | int encode_keep_buffer; | |
114 | int encode_number_precision; | |
115 | } json_config_t; | |
116 | ||
117 | typedef struct { | |
118 | const char *data; | |
119 | int index; | |
120 | strbuf_t *tmp; /* Temporary storage for strings */ | |
121 | json_config_t *cfg; | |
122 | } json_parse_t; | |
123 | ||
124 | typedef struct { | |
125 | json_token_type_t type; | |
126 | int index; | |
127 | union { | |
128 | const char *string; | |
129 | double number; | |
130 | int boolean; | |
131 | } value; | |
132 | int string_len; | |
133 | } json_token_t; | |
134 | ||
135 | static const char *char2escape[256] = { | |
136 | "\\u0000", "\\u0001", "\\u0002", "\\u0003", | |
137 | "\\u0004", "\\u0005", "\\u0006", "\\u0007", | |
138 | "\\b", "\\t", "\\n", "\\u000b", | |
139 | "\\f", "\\r", "\\u000e", "\\u000f", | |
140 | "\\u0010", "\\u0011", "\\u0012", "\\u0013", | |
141 | "\\u0014", "\\u0015", "\\u0016", "\\u0017", | |
142 | "\\u0018", "\\u0019", "\\u001a", "\\u001b", | |
143 | "\\u001c", "\\u001d", "\\u001e", "\\u001f", | |
144 | NULL, NULL, "\\\"", NULL, NULL, NULL, NULL, NULL, | |
145 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, "\\/", | |
146 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
147 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
148 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
149 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
150 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
151 | NULL, NULL, NULL, NULL, "\\\\", NULL, NULL, NULL, | |
152 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
153 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
154 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
155 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, "\\u007f", | |
156 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
157 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
158 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
159 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
160 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
161 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
162 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
163 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
164 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
165 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
166 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
167 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
168 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
169 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
170 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
171 | NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, | |
172 | }; | |
173 | ||
174 | static int json_config_key; | |
175 | ||
176 | /* ===== CONFIGURATION ===== */ | |
177 | ||
178 | static json_config_t *json_fetch_config(lua_State *l) | |
179 | { | |
180 | json_config_t *cfg; | |
181 | ||
182 | lua_pushlightuserdata(l, &json_config_key); | |
183 | lua_gettable(l, LUA_REGISTRYINDEX); | |
184 | cfg = lua_touserdata(l, -1); | |
185 | if (!cfg) | |
186 | luaL_error(l, "BUG: Unable to fetch CJSON configuration"); | |
187 | ||
188 | lua_pop(l, 1); | |
189 | ||
190 | return cfg; | |
191 | } | |
192 | ||
193 | static void json_verify_arg_count(lua_State *l, int args) | |
194 | { | |
195 | luaL_argcheck(l, lua_gettop(l) <= args, args + 1, | |
196 | "found too many arguments"); | |
197 | } | |
198 | ||
199 | /* Configures handling of extremely sparse arrays: | |
200 | * convert: Convert extremely sparse arrays into objects? Otherwise error. | |
201 | * ratio: 0: always allow sparse; 1: never allow sparse; >1: use ratio | |
202 | * safe: Always use an array when the max index <= safe */ | |
203 | static int json_cfg_encode_sparse_array(lua_State *l) | |
204 | { | |
205 | json_config_t *cfg; | |
206 | int val; | |
207 | ||
208 | json_verify_arg_count(l, 3); | |
209 | cfg = json_fetch_config(l); | |
210 | ||
211 | switch (lua_gettop(l)) { | |
212 | case 3: | |
213 | val = luaL_checkinteger(l, 3); | |
214 | luaL_argcheck(l, val >= 0, 3, "expected integer >= 0"); | |
215 | cfg->encode_sparse_safe = val; | |
216 | case 2: | |
217 | val = luaL_checkinteger(l, 2); | |
218 | luaL_argcheck(l, val >= 0, 2, "expected integer >= 0"); | |
219 | cfg->encode_sparse_ratio = val; | |
220 | case 1: | |
221 | luaL_argcheck(l, lua_isboolean(l, 1), 1, "expected boolean"); | |
222 | cfg->encode_sparse_convert = lua_toboolean(l, 1); | |
223 | } | |
224 | ||
225 | lua_pushboolean(l, cfg->encode_sparse_convert); | |
226 | lua_pushinteger(l, cfg->encode_sparse_ratio); | |
227 | lua_pushinteger(l, cfg->encode_sparse_safe); | |
228 | ||
229 | return 3; | |
230 | } | |
231 | ||
232 | /* Configures the maximum number of nested arrays/objects allowed when | |
233 | * encoding */ | |
234 | static int json_cfg_encode_max_depth(lua_State *l) | |
235 | { | |
236 | json_config_t *cfg; | |
237 | int depth; | |
238 | ||
239 | json_verify_arg_count(l, 1); | |
240 | cfg = json_fetch_config(l); | |
241 | ||
242 | if (lua_gettop(l)) { | |
243 | depth = luaL_checkinteger(l, 1); | |
244 | luaL_argcheck(l, depth > 0, 1, "expected positive integer"); | |
245 | cfg->encode_max_depth = depth; | |
246 | } | |
247 | ||
248 | lua_pushinteger(l, cfg->encode_max_depth); | |
249 | ||
250 | return 1; | |
251 | } | |
252 | ||
253 | static void json_set_number_precision(json_config_t *cfg, int prec) | |
254 | { | |
255 | cfg->encode_number_precision = prec; | |
256 | sprintf(cfg->number_fmt, "%%.%dg", prec); | |
257 | } | |
258 | ||
259 | /* Configures number precision when converting doubles to text */ | |
260 | static int json_cfg_encode_number_precision(lua_State *l) | |
261 | { | |
262 | json_config_t *cfg; | |
263 | int precision; | |
264 | ||
265 | json_verify_arg_count(l, 1); | |
266 | cfg = json_fetch_config(l); | |
267 | ||
268 | if (lua_gettop(l)) { | |
269 | precision = luaL_checkinteger(l, 1); | |
270 | luaL_argcheck(l, 1 <= precision && precision <= 14, 1, | |
271 | "expected integer between 1 and 14"); | |
272 | json_set_number_precision(cfg, precision); | |
273 | } | |
274 | ||
275 | lua_pushinteger(l, cfg->encode_number_precision); | |
276 | ||
277 | return 1; | |
278 | } | |
279 | ||
280 | /* Configures JSON encoding buffer persistence */ | |
281 | static int json_cfg_encode_keep_buffer(lua_State *l) | |
282 | { | |
283 | json_config_t *cfg; | |
284 | ||
285 | json_verify_arg_count(l, 1); | |
286 | cfg = json_fetch_config(l); | |
287 | ||
288 | if (lua_gettop(l)) { | |
289 | luaL_checktype(l, 1, LUA_TBOOLEAN); | |
290 | cfg->encode_keep_buffer = lua_toboolean(l, 1); | |
291 | } | |
292 | ||
293 | lua_pushboolean(l, cfg->encode_keep_buffer); | |
294 | ||
295 | return 1; | |
296 | } | |
297 | ||
298 | /* On argument: decode enum and set config variables | |
299 | * **options must point to a NULL terminated array of 4 enums | |
300 | * Returns: current enum value */ | |
301 | static void json_enum_option(lua_State *l, const char **options, | |
302 | int *opt1, int *opt2) | |
303 | { | |
304 | int setting; | |
305 | ||
306 | if (lua_gettop(l)) { | |
307 | if (lua_isboolean(l, 1)) | |
308 | setting = lua_toboolean(l, 1) * 3; | |
309 | else | |
310 | setting = luaL_checkoption(l, 1, NULL, options); | |
311 | ||
312 | *opt1 = setting & 1 ? 1 : 0; | |
313 | *opt2 = setting & 2 ? 1 : 0; | |
314 | } else { | |
315 | setting = *opt1 | (*opt2 << 1); | |
316 | } | |
317 | ||
318 | if (setting) | |
319 | lua_pushstring(l, options[setting]); | |
320 | else | |
321 | lua_pushboolean(l, 0); | |
322 | } | |
323 | ||
324 | ||
325 | /* When enabled, rejects: NaN, Infinity, hexidecimal numbers */ | |
326 | static int json_cfg_refuse_invalid_numbers(lua_State *l) | |
327 | { | |
328 | static const char *options_enc_dec[] = { "none", "encode", "decode", | |
329 | "both", NULL }; | |
330 | json_config_t *cfg; | |
331 | ||
332 | json_verify_arg_count(l, 1); | |
333 | cfg = json_fetch_config(l); | |
334 | ||
335 | json_enum_option(l, options_enc_dec, | |
336 | &cfg->encode_refuse_badnum, | |
337 | &cfg->decode_refuse_badnum); | |
338 | ||
339 | return 1; | |
340 | } | |
341 | ||
342 | static int json_destroy_config(lua_State *l) | |
343 | { | |
344 | json_config_t *cfg; | |
345 | ||
346 | cfg = lua_touserdata(l, 1); | |
347 | if (cfg) | |
348 | strbuf_free(&cfg->encode_buf); | |
349 | cfg = NULL; | |
350 | ||
351 | return 0; | |
352 | } | |
353 | ||
354 | static void json_create_config(lua_State *l) | |
355 | { | |
356 | json_config_t *cfg; | |
357 | int i; | |
358 | ||
359 | cfg = lua_newuserdata(l, sizeof(*cfg)); | |
360 | ||
361 | /* Create GC method to clean up strbuf */ | |
362 | lua_newtable(l); | |
363 | lua_pushcfunction(l, json_destroy_config); | |
364 | lua_setfield(l, -2, "__gc"); | |
365 | lua_setmetatable(l, -2); | |
366 | ||
367 | strbuf_init(&cfg->encode_buf, 0); | |
368 | ||
369 | cfg->encode_sparse_convert = DEFAULT_SPARSE_CONVERT; | |
370 | cfg->encode_sparse_ratio = DEFAULT_SPARSE_RATIO; | |
371 | cfg->encode_sparse_safe = DEFAULT_SPARSE_SAFE; | |
372 | cfg->encode_max_depth = DEFAULT_MAX_DEPTH; | |
373 | cfg->encode_refuse_badnum = DEFAULT_ENCODE_REFUSE_BADNUM; | |
374 | cfg->decode_refuse_badnum = DEFAULT_DECODE_REFUSE_BADNUM; | |
375 | cfg->encode_keep_buffer = DEFAULT_ENCODE_KEEP_BUFFER; | |
376 | json_set_number_precision(cfg, 14); | |
377 | ||
378 | /* Decoding init */ | |
379 | ||
380 | /* Tag all characters as an error */ | |
381 | for (i = 0; i < 256; i++) | |
382 | cfg->ch2token[i] = T_ERROR; | |
383 | ||
384 | /* Set tokens that require no further processing */ | |
385 | cfg->ch2token['{'] = T_OBJ_BEGIN; | |
386 | cfg->ch2token['}'] = T_OBJ_END; | |
387 | cfg->ch2token['['] = T_ARR_BEGIN; | |
388 | cfg->ch2token[']'] = T_ARR_END; | |
389 | cfg->ch2token[','] = T_COMMA; | |
390 | cfg->ch2token[':'] = T_COLON; | |
391 | cfg->ch2token['\0'] = T_END; | |
392 | cfg->ch2token[' '] = T_WHITESPACE; | |
393 | cfg->ch2token['\t'] = T_WHITESPACE; | |
394 | cfg->ch2token['\n'] = T_WHITESPACE; | |
395 | cfg->ch2token['\r'] = T_WHITESPACE; | |
396 | ||
397 | /* Update characters that require further processing */ | |
398 | cfg->ch2token['f'] = T_UNKNOWN; /* false? */ | |
399 | cfg->ch2token['i'] = T_UNKNOWN; /* inf, ininity? */ | |
400 | cfg->ch2token['I'] = T_UNKNOWN; | |
401 | cfg->ch2token['n'] = T_UNKNOWN; /* null, nan? */ | |
402 | cfg->ch2token['N'] = T_UNKNOWN; | |
403 | cfg->ch2token['t'] = T_UNKNOWN; /* true? */ | |
404 | cfg->ch2token['"'] = T_UNKNOWN; /* string? */ | |
405 | cfg->ch2token['+'] = T_UNKNOWN; /* number? */ | |
406 | cfg->ch2token['-'] = T_UNKNOWN; | |
407 | for (i = 0; i < 10; i++) | |
408 | cfg->ch2token['0' + i] = T_UNKNOWN; | |
409 | ||
410 | /* Lookup table for parsing escape characters */ | |
411 | for (i = 0; i < 256; i++) | |
412 | cfg->escape2char[i] = 0; /* String error */ | |
413 | cfg->escape2char['"'] = '"'; | |
414 | cfg->escape2char['\\'] = '\\'; | |
415 | cfg->escape2char['/'] = '/'; | |
416 | cfg->escape2char['b'] = '\b'; | |
417 | cfg->escape2char['t'] = '\t'; | |
418 | cfg->escape2char['n'] = '\n'; | |
419 | cfg->escape2char['f'] = '\f'; | |
420 | cfg->escape2char['r'] = '\r'; | |
421 | cfg->escape2char['u'] = 'u'; /* Unicode parsing required */ | |
422 | ||
423 | ||
424 | #if 0 | |
425 | /* Initialise separate storage for pre-generated escape codes. | |
426 | * Escapes 0-31 map directly, 34, 92, 127 follow afterwards to | |
427 | * save memory. */ | |
428 | for (i = 0 ; i < 32; i++) | |
429 | sprintf(cfg->escapes[i], "\\u%04x", i); | |
430 | strcpy(cfg->escapes[8], "\b"); /* Override simpler escapes */ | |
431 | strcpy(cfg->escapes[9], "\t"); | |
432 | strcpy(cfg->escapes[10], "\n"); | |
433 | strcpy(cfg->escapes[12], "\f"); | |
434 | strcpy(cfg->escapes[13], "\r"); | |
435 | strcpy(cfg->escapes[32], "\\\""); /* chr(34) */ | |
436 | strcpy(cfg->escapes[33], "\\\\"); /* chr(92) */ | |
437 | sprintf(cfg->escapes[34], "\\u%04x", 127); /* char(127) */ | |
438 | ||
439 | /* Initialise encoding escape lookup table */ | |
440 | for (i = 0; i < 32; i++) | |
441 | cfg->char2escape[i] = cfg->escapes[i]; | |
442 | for (i = 32; i < 256; i++) | |
443 | cfg->char2escape[i] = NULL; | |
444 | cfg->char2escape[34] = cfg->escapes[32]; | |
445 | cfg->char2escape[92] = cfg->escapes[33]; | |
446 | cfg->char2escape[127] = cfg->escapes[34]; | |
447 | #endif | |
448 | } | |
449 | ||
450 | /* ===== ENCODING ===== */ | |
451 | ||
452 | static void json_encode_exception(lua_State *l, json_config_t *cfg, int lindex, | |
453 | const char *reason) | |
454 | { | |
455 | if (!cfg->encode_keep_buffer) | |
456 | strbuf_free(&cfg->encode_buf); | |
457 | luaL_error(l, "Cannot serialise %s: %s", | |
458 | lua_typename(l, lua_type(l, lindex)), reason); | |
459 | } | |
460 | ||
461 | /* json_append_string args: | |
462 | * - lua_State | |
463 | * - JSON strbuf | |
464 | * - String (Lua stack index) | |
465 | * | |
466 | * Returns nothing. Doesn't remove string from Lua stack */ | |
467 | static void json_append_string(lua_State *l, strbuf_t *json, int lindex) | |
468 | { | |
469 | const char *escstr; | |
470 | int i; | |
471 | const char *str; | |
472 | size_t len; | |
473 | ||
474 | str = lua_tolstring(l, lindex, &len); | |
475 | ||
476 | /* Worst case is len * 6 (all unicode escapes). | |
477 | * This buffer is reused constantly for small strings | |
478 | * If there are any excess pages, they won't be hit anyway. | |
479 | * This gains ~5% speedup. */ | |
480 | strbuf_ensure_empty_length(json, len * 6 + 2); | |
481 | ||
482 | strbuf_append_char_unsafe(json, '\"'); | |
483 | for (i = 0; i < len; i++) { | |
484 | escstr = char2escape[(unsigned char)str[i]]; | |
485 | if (escstr) | |
486 | strbuf_append_string(json, escstr); | |
487 | else | |
488 | strbuf_append_char_unsafe(json, str[i]); | |
489 | } | |
490 | strbuf_append_char_unsafe(json, '\"'); | |
491 | } | |
492 | ||
493 | /* Find the size of the array on the top of the Lua stack | |
494 | * -1 object (not a pure array) | |
495 | * >=0 elements in array | |
496 | */ | |
497 | static int lua_array_length(lua_State *l, json_config_t *cfg) | |
498 | { | |
499 | double k; | |
500 | int max; | |
501 | int items; | |
502 | ||
503 | max = 0; | |
504 | items = 0; | |
505 | ||
506 | lua_pushnil(l); | |
507 | /* table, startkey */ | |
508 | while (lua_next(l, -2) != 0) { | |
509 | /* table, key, value */ | |
510 | if (lua_type(l, -2) == LUA_TNUMBER && | |
511 | (k = lua_tonumber(l, -2))) { | |
512 | /* Integer >= 1 ? */ | |
513 | if (floor(k) == k && k >= 1) { | |
514 | if (k > max) | |
515 | max = k; | |
516 | items++; | |
517 | lua_pop(l, 1); | |
518 | continue; | |
519 | } | |
520 | } | |
521 | ||
522 | /* Must not be an array (non integer key) */ | |
523 | lua_pop(l, 2); | |
524 | return -1; | |
525 | } | |
526 | ||
527 | /* Encode excessively sparse arrays as objects (if enabled) */ | |
528 | if (cfg->encode_sparse_ratio > 0 && | |
529 | max > items * cfg->encode_sparse_ratio && | |
530 | max > cfg->encode_sparse_safe) { | |
531 | if (!cfg->encode_sparse_convert) | |
532 | json_encode_exception(l, cfg, -1, "excessively sparse array"); | |
533 | ||
534 | return -1; | |
535 | } | |
536 | ||
537 | return max; | |
538 | } | |
539 | ||
540 | static void json_encode_descend(lua_State *l, json_config_t *cfg) | |
541 | { | |
542 | cfg->current_depth++; | |
543 | ||
544 | if (cfg->current_depth > cfg->encode_max_depth) { | |
545 | if (!cfg->encode_keep_buffer) | |
546 | strbuf_free(&cfg->encode_buf); | |
547 | luaL_error(l, "Cannot serialise, excessive nesting (%d)", | |
548 | cfg->current_depth); | |
549 | } | |
550 | } | |
551 | ||
552 | static void json_append_data(lua_State *l, json_config_t *cfg, strbuf_t *json); | |
553 | ||
554 | /* json_append_array args: | |
555 | * - lua_State | |
556 | * - JSON strbuf | |
557 | * - Size of passwd Lua array (top of stack) */ | |
558 | static void json_append_array(lua_State *l, json_config_t *cfg, strbuf_t *json, | |
559 | int array_length) | |
560 | { | |
561 | int comma, i; | |
562 | ||
563 | json_encode_descend(l, cfg); | |
564 | ||
565 | strbuf_append_char(json, '['); | |
566 | ||
567 | comma = 0; | |
568 | for (i = 1; i <= array_length; i++) { | |
569 | if (comma) | |
570 | strbuf_append_char(json, ','); | |
571 | else | |
572 | comma = 1; | |
573 | ||
574 | lua_rawgeti(l, -1, i); | |
575 | json_append_data(l, cfg, json); | |
576 | lua_pop(l, 1); | |
577 | } | |
578 | ||
579 | strbuf_append_char(json, ']'); | |
580 | ||
581 | cfg->current_depth--; | |
582 | } | |
583 | ||
584 | static void json_append_number(lua_State *l, strbuf_t *json, int index, | |
585 | json_config_t *cfg) | |
586 | { | |
587 | double num = lua_tonumber(l, index); | |
588 | ||
589 | if (cfg->encode_refuse_badnum && (isinf(num) || isnan(num))) | |
590 | json_encode_exception(l, cfg, index, "must not be NaN or Inf"); | |
591 | ||
592 | /* Lowest double printed with %.14g is 21 characters long: | |
593 | * -1.7976931348623e+308 | |
594 | * | |
595 | * Use 32 to include the \0, and a few extra just in case.. | |
596 | */ | |
597 | strbuf_append_fmt(json, 32, cfg->number_fmt, num); | |
598 | } | |
599 | ||
600 | static void json_append_object(lua_State *l, json_config_t *cfg, | |
601 | strbuf_t *json) | |
602 | { | |
603 | int comma, keytype; | |
604 | ||
605 | json_encode_descend(l, cfg); | |
606 | ||
607 | /* Object */ | |
608 | strbuf_append_char(json, '{'); | |
609 | ||
610 | lua_pushnil(l); | |
611 | /* table, startkey */ | |
612 | comma = 0; | |
613 | while (lua_next(l, -2) != 0) { | |
614 | if (comma) | |
615 | strbuf_append_char(json, ','); | |
616 | else | |
617 | comma = 1; | |
618 | ||
619 | /* table, key, value */ | |
620 | keytype = lua_type(l, -2); | |
621 | if (keytype == LUA_TNUMBER) { | |
622 | strbuf_append_char(json, '"'); | |
623 | json_append_number(l, json, -2, cfg); | |
624 | strbuf_append_mem(json, "\":", 2); | |
625 | } else if (keytype == LUA_TSTRING) { | |
626 | json_append_string(l, json, -2); | |
627 | strbuf_append_char(json, ':'); | |
628 | } else { | |
629 | json_encode_exception(l, cfg, -2, | |
630 | "table key must be a number or string"); | |
631 | /* never returns */ | |
632 | } | |
633 | ||
634 | /* table, key, value */ | |
635 | json_append_data(l, cfg, json); | |
636 | lua_pop(l, 1); | |
637 | /* table, key */ | |
638 | } | |
639 | ||
640 | strbuf_append_char(json, '}'); | |
641 | ||
642 | cfg->current_depth--; | |
643 | } | |
644 | ||
645 | /* Serialise Lua data into JSON string. */ | |
646 | static void json_append_data(lua_State *l, json_config_t *cfg, strbuf_t *json) | |
647 | { | |
648 | int len; | |
649 | ||
650 | switch (lua_type(l, -1)) { | |
651 | case LUA_TSTRING: | |
652 | json_append_string(l, json, -1); | |
653 | break; | |
654 | case LUA_TNUMBER: | |
655 | json_append_number(l, json, -1, cfg); | |
656 | break; | |
657 | case LUA_TBOOLEAN: | |
658 | if (lua_toboolean(l, -1)) | |
659 | strbuf_append_mem(json, "true", 4); | |
660 | else | |
661 | strbuf_append_mem(json, "false", 5); | |
662 | break; | |
663 | case LUA_TTABLE: | |
664 | len = lua_array_length(l, cfg); | |
665 | if (len > 0) | |
666 | json_append_array(l, cfg, json, len); | |
667 | else | |
668 | json_append_object(l, cfg, json); | |
669 | break; | |
670 | case LUA_TNIL: | |
671 | strbuf_append_mem(json, "null", 4); | |
672 | break; | |
673 | case LUA_TLIGHTUSERDATA: | |
674 | if (lua_touserdata(l, -1) == NULL) { | |
675 | strbuf_append_mem(json, "null", 4); | |
676 | break; | |
677 | } | |
678 | default: | |
679 | /* Remaining types (LUA_TFUNCTION, LUA_TUSERDATA, LUA_TTHREAD, | |
680 | * and LUA_TLIGHTUSERDATA) cannot be serialised */ | |
681 | json_encode_exception(l, cfg, -1, "type not supported"); | |
682 | /* never returns */ | |
683 | } | |
684 | } | |
685 | ||
686 | static int json_encode(lua_State *l) | |
687 | { | |
688 | json_config_t *cfg; | |
689 | char *json; | |
690 | int len; | |
691 | ||
692 | /* Can't use json_verify_arg_count() since we need to ensure | |
693 | * there is only 1 argument */ | |
694 | luaL_argcheck(l, lua_gettop(l) == 1, 1, "expected 1 argument"); | |
695 | ||
696 | cfg = json_fetch_config(l); | |
697 | cfg->current_depth = 0; | |
698 | ||
699 | /* Reset the persistent buffer if it exists. | |
700 | * Otherwise allocate a new buffer. */ | |
701 | if (strbuf_allocated(&cfg->encode_buf)) | |
702 | strbuf_reset(&cfg->encode_buf); | |
703 | else | |
704 | strbuf_init(&cfg->encode_buf, 0); | |
705 | ||
706 | json_append_data(l, cfg, &cfg->encode_buf); | |
707 | json = strbuf_string(&cfg->encode_buf, &len); | |
708 | ||
709 | lua_pushlstring(l, json, len); | |
710 | ||
711 | if (!cfg->encode_keep_buffer) | |
712 | strbuf_free(&cfg->encode_buf); | |
713 | ||
714 | return 1; | |
715 | } | |
716 | ||
717 | /* ===== DECODING ===== */ | |
718 | ||
719 | static void json_process_value(lua_State *l, json_parse_t *json, | |
720 | json_token_t *token); | |
721 | ||
722 | static int hexdigit2int(char hex) | |
723 | { | |
724 | if ('0' <= hex && hex <= '9') | |
725 | return hex - '0'; | |
726 | ||
727 | /* Force lowercase */ | |
728 | hex |= 0x20; | |
729 | if ('a' <= hex && hex <= 'f') | |
730 | return 10 + hex - 'a'; | |
731 | ||
732 | return -1; | |
733 | } | |
734 | ||
735 | static int decode_hex4(const char *hex) | |
736 | { | |
737 | int digit[4]; | |
738 | int i; | |
739 | ||
740 | /* Convert ASCII hex digit to numeric digit | |
741 | * Note: this returns an error for invalid hex digits, including | |
742 | * NULL */ | |
743 | for (i = 0; i < 4; i++) { | |
744 | digit[i] = hexdigit2int(hex[i]); | |
745 | if (digit[i] < 0) { | |
746 | return -1; | |
747 | } | |
748 | } | |
749 | ||
750 | return (digit[0] << 12) + | |
751 | (digit[1] << 8) + | |
752 | (digit[2] << 4) + | |
753 | digit[3]; | |
754 | } | |
755 | ||
756 | /* Converts a Unicode codepoint to UTF-8. | |
757 | * Returns UTF-8 string length, and up to 4 bytes in *utf8 */ | |
758 | static int codepoint_to_utf8(char *utf8, int codepoint) | |
759 | { | |
760 | /* 0xxxxxxx */ | |
761 | if (codepoint <= 0x7F) { | |
762 | utf8[0] = codepoint; | |
763 | return 1; | |
764 | } | |
765 | ||
766 | /* 110xxxxx 10xxxxxx */ | |
767 | if (codepoint <= 0x7FF) { | |
768 | utf8[0] = (codepoint >> 6) | 0xC0; | |
769 | utf8[1] = (codepoint & 0x3F) | 0x80; | |
770 | return 2; | |
771 | } | |
772 | ||
773 | /* 1110xxxx 10xxxxxx 10xxxxxx */ | |
774 | if (codepoint <= 0xFFFF) { | |
775 | utf8[0] = (codepoint >> 12) | 0xE0; | |
776 | utf8[1] = ((codepoint >> 6) & 0x3F) | 0x80; | |
777 | utf8[2] = (codepoint & 0x3F) | 0x80; | |
778 | return 3; | |
779 | } | |
780 | ||
781 | /* 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */ | |
782 | if (codepoint <= 0x1FFFFF) { | |
783 | utf8[0] = (codepoint >> 18) | 0xF0; | |
784 | utf8[1] = ((codepoint >> 12) & 0x3F) | 0x80; | |
785 | utf8[2] = ((codepoint >> 6) & 0x3F) | 0x80; | |
786 | utf8[3] = (codepoint & 0x3F) | 0x80; | |
787 | return 4; | |
788 | } | |
789 | ||
790 | return 0; | |
791 | } | |
792 | ||
793 | ||
794 | /* Called when index pointing to beginning of UTF-16 code escape: \uXXXX | |
795 | * \u is guaranteed to exist, but the remaining hex characters may be | |
796 | * missing. | |
797 | * Translate to UTF-8 and append to temporary token string. | |
798 | * Must advance index to the next character to be processed. | |
799 | * Returns: 0 success | |
800 | * -1 error | |
801 | */ | |
802 | static int json_append_unicode_escape(json_parse_t *json) | |
803 | { | |
804 | char utf8[4]; /* Surrogate pairs require 4 UTF-8 bytes */ | |
805 | int codepoint; | |
806 | int surrogate_low; | |
807 | int len; | |
808 | int escape_len = 6; | |
809 | ||
810 | /* Fetch UTF-16 code unit */ | |
811 | codepoint = decode_hex4(&json->data[json->index + 2]); | |
812 | if (codepoint < 0) | |
813 | return -1; | |
814 | ||
815 | /* UTF-16 surrogate pairs take the following 2 byte form: | |
816 | * 11011 x yyyyyyyyyy | |
817 | * When x = 0: y is the high 10 bits of the codepoint | |
818 | * x = 1: y is the low 10 bits of the codepoint | |
819 | * | |
820 | * Check for a surrogate pair (high or low) */ | |
821 | if ((codepoint & 0xF800) == 0xD800) { | |
822 | /* Error if the 1st surrogate is not high */ | |
823 | if (codepoint & 0x400) | |
824 | return -1; | |
825 | ||
826 | /* Ensure the next code is a unicode escape */ | |
827 | if (json->data[json->index + escape_len] != '\\' || | |
828 | json->data[json->index + escape_len + 1] != 'u') { | |
829 | return -1; | |
830 | } | |
831 | ||
832 | /* Fetch the next codepoint */ | |
833 | surrogate_low = decode_hex4(&json->data[json->index + 2 + escape_len]); | |
834 | if (surrogate_low < 0) | |
835 | return -1; | |
836 | ||
837 | /* Error if the 2nd code is not a low surrogate */ | |
838 | if ((surrogate_low & 0xFC00) != 0xDC00) | |
839 | return -1; | |
840 | ||
841 | /* Calculate Unicode codepoint */ | |
842 | codepoint = (codepoint & 0x3FF) << 10; | |
843 | surrogate_low &= 0x3FF; | |
844 | codepoint = (codepoint | surrogate_low) + 0x10000; | |
845 | escape_len = 12; | |
846 | } | |
847 | ||
848 | /* Convert codepoint to UTF-8 */ | |
849 | len = codepoint_to_utf8(utf8, codepoint); | |
850 | if (!len) | |
851 | return -1; | |
852 | ||
853 | /* Append bytes and advance parse index */ | |
854 | strbuf_append_mem_unsafe(json->tmp, utf8, len); | |
855 | json->index += escape_len; | |
856 | ||
857 | return 0; | |
858 | } | |
859 | ||
860 | static void json_set_token_error(json_token_t *token, json_parse_t *json, | |
861 | const char *errtype) | |
862 | { | |
863 | token->type = T_ERROR; | |
864 | token->index = json->index; | |
865 | token->value.string = errtype; | |
866 | } | |
867 | ||
868 | static void json_next_string_token(json_parse_t *json, json_token_t *token) | |
869 | { | |
870 | char *escape2char = json->cfg->escape2char; | |
871 | char ch; | |
872 | ||
873 | /* Caller must ensure a string is next */ | |
874 | assert(json->data[json->index] == '"'); | |
875 | ||
876 | /* Skip " */ | |
877 | json->index++; | |
878 | ||
879 | /* json->tmp is the temporary strbuf used to accumulate the | |
880 | * decoded string value. */ | |
881 | strbuf_reset(json->tmp); | |
882 | while ((ch = json->data[json->index]) != '"') { | |
883 | if (!ch) { | |
884 | /* Premature end of the string */ | |
885 | json_set_token_error(token, json, "unexpected end of string"); | |
886 | return; | |
887 | } | |
888 | ||
889 | /* Handle escapes */ | |
890 | if (ch == '\\') { | |
891 | /* Fetch escape character */ | |
892 | ch = json->data[json->index + 1]; | |
893 | ||
894 | /* Translate escape code and append to tmp string */ | |
895 | ch = escape2char[(unsigned char)ch]; | |
896 | if (ch == 'u') { | |
897 | if (json_append_unicode_escape(json) == 0) | |
898 | continue; | |
899 | ||
900 | json_set_token_error(token, json, | |
901 | "invalid unicode escape code"); | |
902 | return; | |
903 | } | |
904 | if (!ch) { | |
905 | json_set_token_error(token, json, "invalid escape code"); | |
906 | return; | |
907 | } | |
908 | ||
909 | /* Skip '\' */ | |
910 | json->index++; | |
911 | } | |
912 | /* Append normal character or translated single character | |
913 | * Unicode escapes are handled above */ | |
914 | strbuf_append_char_unsafe(json->tmp, ch); | |
915 | json->index++; | |
916 | } | |
917 | json->index++; /* Eat final quote (") */ | |
918 | ||
919 | strbuf_ensure_null(json->tmp); | |
920 | ||
921 | token->type = T_STRING; | |
922 | token->value.string = strbuf_string(json->tmp, &token->string_len); | |
923 | } | |
924 | ||
925 | /* JSON numbers should take the following form: | |
926 | * -?(0|[1-9]|[1-9][0-9]+)(.[0-9]+)?([eE][-+]?[0-9]+)? | |
927 | * | |
928 | * json_next_number_token() uses strtod() which allows other forms: | |
929 | * - numbers starting with '+' | |
930 | * - NaN, -NaN, infinity, -infinity | |
931 | * - hexidecimal numbers | |
932 | * - numbers with leading zeros | |
933 | * | |
934 | * json_is_invalid_number() detects "numbers" which may pass strtod()'s | |
935 | * error checking, but should not be allowed with strict JSON. | |
936 | * | |
937 | * json_is_invalid_number() may pass numbers which cause strtod() | |
938 | * to generate an error. | |
939 | */ | |
940 | static int json_is_invalid_number(json_parse_t *json) | |
941 | { | |
942 | int i = json->index; | |
943 | ||
944 | /* Reject numbers starting with + */ | |
945 | if (json->data[i] == '+') | |
946 | return 1; | |
947 | ||
948 | /* Skip minus sign if it exists */ | |
949 | if (json->data[i] == '-') | |
950 | i++; | |
951 | ||
952 | /* Reject numbers starting with 0x, or leading zeros */ | |
953 | if (json->data[i] == '0') { | |
954 | int ch2 = json->data[i + 1]; | |
955 | ||
956 | if ((ch2 | 0x20) == 'x' || /* Hex */ | |
957 | ('0' <= ch2 && ch2 <= '9')) /* Leading zero */ | |
958 | return 1; | |
959 | ||
960 | return 0; | |
961 | } else if (json->data[i] <= '9') { | |
962 | return 0; /* Ordinary number */ | |
963 | } | |
964 | ||
965 | ||
966 | /* Reject inf/nan */ | |
967 | if (!strncasecmp(&json->data[i], "inf", 3)) | |
968 | return 1; | |
969 | if (!strncasecmp(&json->data[i], "nan", 3)) | |
970 | return 1; | |
971 | ||
972 | /* Pass all other numbers which may still be invalid, but | |
973 | * strtod() will catch them. */ | |
974 | return 0; | |
975 | } | |
976 | ||
977 | static void json_next_number_token(json_parse_t *json, json_token_t *token) | |
978 | { | |
979 | const char *startptr; | |
980 | char *endptr; | |
981 | ||
982 | token->type = T_NUMBER; | |
983 | startptr = &json->data[json->index]; | |
984 | token->value.number = strtod(&json->data[json->index], &endptr); | |
985 | if (startptr == endptr) | |
986 | json_set_token_error(token, json, "invalid number"); | |
987 | else | |
988 | json->index += endptr - startptr; /* Skip the processed number */ | |
989 | ||
990 | return; | |
991 | } | |
992 | ||
993 | /* Fills in the token struct. | |
994 | * T_STRING will return a pointer to the json_parse_t temporary string | |
995 | * T_ERROR will leave the json->index pointer at the error. | |
996 | */ | |
997 | static void json_next_token(json_parse_t *json, json_token_t *token) | |
998 | { | |
999 | json_token_type_t *ch2token = json->cfg->ch2token; | |
1000 | int ch; | |
1001 | ||
1002 | /* Eat whitespace. FIXME: UGLY */ | |
1003 | token->type = ch2token[(unsigned char)json->data[json->index]]; | |
1004 | while (token->type == T_WHITESPACE) | |
1005 | token->type = ch2token[(unsigned char)json->data[++json->index]]; | |
1006 | ||
1007 | token->index = json->index; | |
1008 | ||
1009 | /* Don't advance the pointer for an error or the end */ | |
1010 | if (token->type == T_ERROR) { | |
1011 | json_set_token_error(token, json, "invalid token"); | |
1012 | return; | |
1013 | } | |
1014 | ||
1015 | if (token->type == T_END) { | |
1016 | return; | |
1017 | } | |
1018 | ||
1019 | /* Found a known single character token, advance index and return */ | |
1020 | if (token->type != T_UNKNOWN) { | |
1021 | json->index++; | |
1022 | return; | |
1023 | } | |
1024 | ||
1025 | /* Process characters which triggered T_UNKNOWN */ | |
1026 | ch = json->data[json->index]; | |
1027 | ||
1028 | /* Must use strncmp() to match the front of the JSON string. | |
1029 | * JSON identifier must be lowercase. | |
1030 | * When strict_numbers if disabled, either case is allowed for | |
1031 | * Infinity/NaN (since we are no longer following the spec..) */ | |
1032 | if (ch == '"') { | |
1033 | json_next_string_token(json, token); | |
1034 | return; | |
1035 | } else if (ch == '-' || ('0' <= ch && ch <= '9')) { | |
1036 | if (json->cfg->decode_refuse_badnum && json_is_invalid_number(json)) { | |
1037 | json_set_token_error(token, json, "invalid number"); | |
1038 | return; | |
1039 | } | |
1040 | json_next_number_token(json, token); | |
1041 | return; | |
1042 | } else if (!strncmp(&json->data[json->index], "true", 4)) { | |
1043 | token->type = T_BOOLEAN; | |
1044 | token->value.boolean = 1; | |
1045 | json->index += 4; | |
1046 | return; | |
1047 | } else if (!strncmp(&json->data[json->index], "false", 5)) { | |
1048 | token->type = T_BOOLEAN; | |
1049 | token->value.boolean = 0; | |
1050 | json->index += 5; | |
1051 | return; | |
1052 | } else if (!strncmp(&json->data[json->index], "null", 4)) { | |
1053 | token->type = T_NULL; | |
1054 | json->index += 4; | |
1055 | return; | |
1056 | } else if (!json->cfg->decode_refuse_badnum && | |
1057 | json_is_invalid_number(json)) { | |
1058 | /* When refuse_badnum is disabled, only attempt to process | |
1059 | * numbers we know are invalid JSON (Inf, NaN, hex) | |
1060 | * This is required to generate an appropriate token error, | |
1061 | * otherwise all bad tokens will register as "invalid number" | |
1062 | */ | |
1063 | json_next_number_token(json, token); | |
1064 | return; | |
1065 | } | |
1066 | ||
1067 | /* Token starts with t/f/n but isn't recognised above. */ | |
1068 | json_set_token_error(token, json, "invalid token"); | |
1069 | } | |
1070 | ||
1071 | /* This function does not return. | |
1072 | * DO NOT CALL WITH DYNAMIC MEMORY ALLOCATED. | |
1073 | * The only supported exception is the temporary parser string | |
1074 | * json->tmp struct. | |
1075 | * json and token should exist on the stack somewhere. | |
1076 | * luaL_error() will long_jmp and release the stack */ | |
1077 | static void json_throw_parse_error(lua_State *l, json_parse_t *json, | |
1078 | const char *exp, json_token_t *token) | |
1079 | { | |
1080 | const char *found; | |
1081 | ||
1082 | strbuf_free(json->tmp); | |
1083 | ||
1084 | if (token->type == T_ERROR) | |
1085 | found = token->value.string; | |
1086 | else | |
1087 | found = json_token_type_name[token->type]; | |
1088 | ||
1089 | /* Note: token->index is 0 based, display starting from 1 */ | |
1090 | luaL_error(l, "Expected %s but found %s at character %d", | |
1091 | exp, found, token->index + 1); | |
1092 | } | |
1093 | ||
1094 | static void json_decode_checkstack(lua_State *l, json_parse_t *json, int n) | |
1095 | { | |
1096 | if (lua_checkstack(l, n)) | |
1097 | return; | |
1098 | ||
1099 | strbuf_free(json->tmp); | |
1100 | luaL_error(l, "Too many nested data structures"); | |
1101 | } | |
1102 | ||
1103 | static void json_parse_object_context(lua_State *l, json_parse_t *json) | |
1104 | { | |
1105 | json_token_t token; | |
1106 | ||
1107 | /* 3 slots required: | |
1108 | * .., table, key, value */ | |
1109 | json_decode_checkstack(l, json, 3); | |
1110 | ||
1111 | lua_newtable(l); | |
1112 | ||
1113 | json_next_token(json, &token); | |
1114 | ||
1115 | /* Handle empty objects */ | |
1116 | if (token.type == T_OBJ_END) { | |
1117 | return; | |
1118 | } | |
1119 | ||
1120 | while (1) { | |
1121 | if (token.type != T_STRING) | |
1122 | json_throw_parse_error(l, json, "object key string", &token); | |
1123 | ||
1124 | /* Push key */ | |
1125 | lua_pushlstring(l, token.value.string, token.string_len); | |
1126 | ||
1127 | json_next_token(json, &token); | |
1128 | if (token.type != T_COLON) | |
1129 | json_throw_parse_error(l, json, "colon", &token); | |
1130 | ||
1131 | /* Fetch value */ | |
1132 | json_next_token(json, &token); | |
1133 | json_process_value(l, json, &token); | |
1134 | ||
1135 | /* Set key = value */ | |
1136 | lua_rawset(l, -3); | |
1137 | ||
1138 | json_next_token(json, &token); | |
1139 | ||
1140 | if (token.type == T_OBJ_END) | |
1141 | return; | |
1142 | ||
1143 | if (token.type != T_COMMA) | |
1144 | json_throw_parse_error(l, json, "comma or object end", &token); | |
1145 | ||
1146 | json_next_token(json, &token); | |
1147 | } | |
1148 | } | |
1149 | ||
1150 | /* Handle the array context */ | |
1151 | static void json_parse_array_context(lua_State *l, json_parse_t *json) | |
1152 | { | |
1153 | json_token_t token; | |
1154 | int i; | |
1155 | ||
1156 | /* 2 slots required: | |
1157 | * .., table, value */ | |
1158 | json_decode_checkstack(l, json, 2); | |
1159 | ||
1160 | lua_newtable(l); | |
1161 | ||
1162 | json_next_token(json, &token); | |
1163 | ||
1164 | /* Handle empty arrays */ | |
1165 | if (token.type == T_ARR_END) | |
1166 | return; | |
1167 | ||
1168 | for (i = 1; ; i++) { | |
1169 | json_process_value(l, json, &token); | |
1170 | lua_rawseti(l, -2, i); /* arr[i] = value */ | |
1171 | ||
1172 | json_next_token(json, &token); | |
1173 | ||
1174 | if (token.type == T_ARR_END) | |
1175 | return; | |
1176 | ||
1177 | if (token.type != T_COMMA) | |
1178 | json_throw_parse_error(l, json, "comma or array end", &token); | |
1179 | ||
1180 | json_next_token(json, &token); | |
1181 | } | |
1182 | } | |
1183 | ||
1184 | /* Handle the "value" context */ | |
1185 | static void json_process_value(lua_State *l, json_parse_t *json, | |
1186 | json_token_t *token) | |
1187 | { | |
1188 | switch (token->type) { | |
1189 | case T_STRING: | |
1190 | lua_pushlstring(l, token->value.string, token->string_len); | |
1191 | break;; | |
1192 | case T_NUMBER: | |
1193 | lua_pushnumber(l, token->value.number); | |
1194 | break;; | |
1195 | case T_BOOLEAN: | |
1196 | lua_pushboolean(l, token->value.boolean); | |
1197 | break;; | |
1198 | case T_OBJ_BEGIN: | |
1199 | json_parse_object_context(l, json); | |
1200 | break;; | |
1201 | case T_ARR_BEGIN: | |
1202 | json_parse_array_context(l, json); | |
1203 | break;; | |
1204 | case T_NULL: | |
1205 | /* In Lua, setting "t[k] = nil" will delete k from the table. | |
1206 | * Hence a NULL pointer lightuserdata object is used instead */ | |
1207 | lua_pushlightuserdata(l, NULL); | |
1208 | break;; | |
1209 | default: | |
1210 | json_throw_parse_error(l, json, "value", token); | |
1211 | } | |
1212 | } | |
1213 | ||
1214 | /* json_text must be null terminated string */ | |
1215 | static void lua_json_decode(lua_State *l, const char *json_text, int json_len) | |
1216 | { | |
1217 | json_parse_t json; | |
1218 | json_token_t token; | |
1219 | ||
1220 | json.cfg = json_fetch_config(l); | |
1221 | json.data = json_text; | |
1222 | json.index = 0; | |
1223 | ||
1224 | /* Ensure the temporary buffer can hold the entire string. | |
1225 | * This means we no longer need to do length checks since the decoded | |
1226 | * string must be smaller than the entire json string */ | |
1227 | json.tmp = strbuf_new(json_len); | |
1228 | ||
1229 | json_next_token(&json, &token); | |
1230 | json_process_value(l, &json, &token); | |
1231 | ||
1232 | /* Ensure there is no more input left */ | |
1233 | json_next_token(&json, &token); | |
1234 | ||
1235 | if (token.type != T_END) | |
1236 | json_throw_parse_error(l, &json, "the end", &token); | |
1237 | ||
1238 | strbuf_free(json.tmp); | |
1239 | } | |
1240 | ||
1241 | static int json_decode(lua_State *l) | |
1242 | { | |
1243 | const char *json; | |
1244 | size_t len; | |
1245 | ||
1246 | json_verify_arg_count(l, 1); | |
1247 | ||
1248 | json = luaL_checklstring(l, 1, &len); | |
1249 | ||
1250 | /* Detect Unicode other than UTF-8 (see RFC 4627, Sec 3) | |
1251 | * | |
1252 | * CJSON can support any simple data type, hence only the first | |
1253 | * character is guaranteed to be ASCII (at worst: '"'). This is | |
1254 | * still enough to detect whether the wrong encoding is in use. */ | |
1255 | if (len >= 2 && (!json[0] || !json[1])) | |
1256 | luaL_error(l, "JSON parser does not support UTF-16 or UTF-32"); | |
1257 | ||
1258 | lua_json_decode(l, json, len); | |
1259 | ||
1260 | return 1; | |
1261 | } | |
1262 | ||
1263 | /* ===== INITIALISATION ===== */ | |
1264 | ||
1265 | int luaopen_cjson(lua_State *l) | |
1266 | { | |
1267 | luaL_Reg reg[] = { | |
1268 | { "encode", json_encode }, | |
1269 | { "decode", json_decode }, | |
1270 | { "encode_sparse_array", json_cfg_encode_sparse_array }, | |
1271 | { "encode_max_depth", json_cfg_encode_max_depth }, | |
1272 | { "encode_number_precision", json_cfg_encode_number_precision }, | |
1273 | { "encode_keep_buffer", json_cfg_encode_keep_buffer }, | |
1274 | { "refuse_invalid_numbers", json_cfg_refuse_invalid_numbers }, | |
1275 | { NULL, NULL } | |
1276 | }; | |
1277 | ||
1278 | /* Use json_fetch_config as a pointer. | |
1279 | * It's faster than using a config string, and more unique */ | |
1280 | lua_pushlightuserdata(l, &json_config_key); | |
1281 | json_create_config(l); | |
1282 | lua_settable(l, LUA_REGISTRYINDEX); | |
1283 | ||
1284 | luaL_register(l, "cjson", reg); | |
1285 | ||
1286 | /* Set cjson.null */ | |
1287 | lua_pushlightuserdata(l, NULL); | |
1288 | lua_setfield(l, -2, "null"); | |
1289 | ||
1290 | /* Set cjson.version */ | |
1291 | lua_pushliteral(l, VERSION); | |
1292 | lua_setfield(l, -2, "version"); | |
1293 | ||
1294 | /* Return cjson table */ | |
1295 | return 1; | |
1296 | } | |
1297 | ||
1298 | /* vi:ai et sw=4 ts=4: | |
1299 | */ |