/* CJSON - JSON support for Lua * * Copyright (c) 2010-2011 Mark Pulford * * Permission is hereby granted, free of charge, to any person obtaining * a copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sublicense, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ /* Caveats: * - JSON "null" values are represented as lightuserdata since Lua * tables cannot contain "nil". Compare with cjson.null. * - Invalid UTF-8 characters are not detected and will be passed * untouched. If required, UTF-8 error checking should be done * outside this library. * - Javascript comments are not part of the JSON spec, and are not * currently supported. * * Note: Decoding is slower than encoding. Lua spends significant * time (30%) managing tables when parsing JSON since it is * difficult to know object/array sizes ahead of time. */ #include "lua_cjson.h" #include #include #include #include "lua.h" #include "lauxlib.h" #include "strbuf.h" /* Support to reset locale to POSIX for strtod() / sprintf(). * Some locales use comma as a decimal separator. This breaks JSON. */ #ifdef USE_POSIX_USELOCALE #ifdef USE_POSIX_SETLOCALE #error Must not define USE_POSIX_USELOCALE and USE_POSIX_SETLOCALE simultaneously. #endif /* unistd.h defines _POSIX_VERSION */ #include #if _POSIX_VERSION >= 200809L /* POSIX.1-2008 adds threadsafe locale support */ #include #elif defined(_POSIX_VERSION) /* Some pre-POSIX.1-2008 operating systems offer xlocale.h instead */ #include #else #error Missing _POSIX_VERSION define #endif /* _POSIX_VERSION */ #define LOCALE_SET_POSIX(x) (x)->saved_locale = uselocale((x)->posix_locale) #define LOCALE_RESTORE(x) uselocale((x)->saved_locale) #elif defined(USE_POSIX_SETLOCALE) /* ANSI C / ISO C90 implementation. Not thread-safe, affects entire process. */ #include #define LOCALE_SET_POSIX(x) \ do { \ (x)->saved_locale = setlocale(LC_NUMERIC, NULL); \ setlocale(LC_NUMERIC, "C"); \ } while(0) #define LOCALE_RESTORE(x) setlocale(LC_NUMERIC, (x)->saved_locale) #else /* Do not work around locale support in strtod() / sprintf() */ #define LOCALE_SET_POSIX(x) do { } while(0) #define LOCALE_RESTORE(x) do { } while(0) #endif /* Workaround for Solaris platforms missing isinf() */ #if !defined(isinf) && (defined(USE_INTERNAL_ISINF) || defined(MISSING_ISINF)) #define isinf(x) (!isnan(x) && isnan((x) - (x))) #endif #define DEFAULT_SPARSE_CONVERT 0 #define DEFAULT_SPARSE_RATIO 2 #define DEFAULT_SPARSE_SAFE 10 #define DEFAULT_MAX_DEPTH 20 #define DEFAULT_ENCODE_REFUSE_BADNUM 1 #define DEFAULT_DECODE_REFUSE_BADNUM 0 #define DEFAULT_ENCODE_KEEP_BUFFER 1 typedef enum { T_OBJ_BEGIN, T_OBJ_END, T_ARR_BEGIN, T_ARR_END, T_STRING, T_NUMBER, T_BOOLEAN, T_NULL, T_COLON, T_COMMA, T_END, T_WHITESPACE, T_ERROR, T_UNKNOWN } json_token_type_t; static const char *json_token_type_name[] = { "T_OBJ_BEGIN", "T_OBJ_END", "T_ARR_BEGIN", "T_ARR_END", "T_STRING", "T_NUMBER", "T_BOOLEAN", "T_NULL", "T_COLON", "T_COMMA", "T_END", "T_WHITESPACE", "T_ERROR", "T_UNKNOWN", NULL }; typedef struct { json_token_type_t ch2token[256]; char escape2char[256]; /* Decoding */ #if 0 char escapes[35][8]; /* Pre-generated escape string buffer */ char *char2escape[256]; /* Encoding */ #endif strbuf_t encode_buf; #if defined(USE_POSIX_USELOCALE) locale_t saved_locale; locale_t posix_locale; #elif defined(USE_POSIX_SETLOCALE) const char *saved_locale; #endif char number_fmt[8]; /* "%.XXg\0" */ int current_depth; int encode_sparse_convert; int encode_sparse_ratio; int encode_sparse_safe; int encode_max_depth; int encode_refuse_badnum; int decode_refuse_badnum; int encode_keep_buffer; int encode_number_precision; } json_config_t; typedef struct { const char *data; int index; strbuf_t *tmp; /* Temporary storage for strings */ json_config_t *cfg; } json_parse_t; typedef struct { json_token_type_t type; int index; union { const char *string; double number; int boolean; } value; int string_len; } json_token_t; static const char *char2escape[256] = { "\\u0000", "\\u0001", "\\u0002", "\\u0003", "\\u0004", "\\u0005", "\\u0006", "\\u0007", "\\b", "\\t", "\\n", "\\u000b", "\\f", "\\r", "\\u000e", "\\u000f", "\\u0010", "\\u0011", "\\u0012", "\\u0013", "\\u0014", "\\u0015", "\\u0016", "\\u0017", "\\u0018", "\\u0019", "\\u001a", "\\u001b", "\\u001c", "\\u001d", "\\u001e", "\\u001f", NULL, NULL, "\\\"", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, "\\/", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, "\\\\", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, "\\u007f", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, }; static int json_config_key; /* ===== CONFIGURATION ===== */ static json_config_t *json_fetch_config(lua_State *l) { json_config_t *cfg; lua_pushlightuserdata(l, &json_config_key); lua_gettable(l, LUA_REGISTRYINDEX); cfg = lua_touserdata(l, -1); if (!cfg) luaL_error(l, "BUG: Unable to fetch CJSON configuration"); lua_pop(l, 1); return cfg; } static void json_verify_arg_count(lua_State *l, int args) { luaL_argcheck(l, lua_gettop(l) <= args, args + 1, "found too many arguments"); } /* Configures handling of extremely sparse arrays: * convert: Convert extremely sparse arrays into objects? Otherwise error. * ratio: 0: always allow sparse; 1: never allow sparse; >1: use ratio * safe: Always use an array when the max index <= safe */ static int json_cfg_encode_sparse_array(lua_State *l) { json_config_t *cfg; int val; json_verify_arg_count(l, 3); cfg = json_fetch_config(l); switch (lua_gettop(l)) { case 3: val = luaL_checkinteger(l, 3); luaL_argcheck(l, val >= 0, 3, "expected integer >= 0"); cfg->encode_sparse_safe = val; case 2: val = luaL_checkinteger(l, 2); luaL_argcheck(l, val >= 0, 2, "expected integer >= 0"); cfg->encode_sparse_ratio = val; case 1: luaL_argcheck(l, lua_isboolean(l, 1), 1, "expected boolean"); cfg->encode_sparse_convert = lua_toboolean(l, 1); } lua_pushboolean(l, cfg->encode_sparse_convert); lua_pushinteger(l, cfg->encode_sparse_ratio); lua_pushinteger(l, cfg->encode_sparse_safe); return 3; } /* Configures the maximum number of nested arrays/objects allowed when * encoding */ static int json_cfg_encode_max_depth(lua_State *l) { json_config_t *cfg; int depth; json_verify_arg_count(l, 1); cfg = json_fetch_config(l); if (lua_gettop(l)) { depth = luaL_checkinteger(l, 1); luaL_argcheck(l, depth > 0, 1, "expected positive integer"); cfg->encode_max_depth = depth; } lua_pushinteger(l, cfg->encode_max_depth); return 1; } static void json_set_number_precision(json_config_t *cfg, int prec) { cfg->encode_number_precision = prec; sprintf(cfg->number_fmt, "%%.%dg", prec); } /* Configures number precision when converting doubles to text */ static int json_cfg_encode_number_precision(lua_State *l) { json_config_t *cfg; int precision; json_verify_arg_count(l, 1); cfg = json_fetch_config(l); if (lua_gettop(l)) { precision = luaL_checkinteger(l, 1); luaL_argcheck(l, 1 <= precision && precision <= 14, 1, "expected integer between 1 and 14"); json_set_number_precision(cfg, precision); } lua_pushinteger(l, cfg->encode_number_precision); return 1; } /* Configures JSON encoding buffer persistence */ static int json_cfg_encode_keep_buffer(lua_State *l) { json_config_t *cfg; json_verify_arg_count(l, 1); cfg = json_fetch_config(l); if (lua_gettop(l)) { luaL_checktype(l, 1, LUA_TBOOLEAN); cfg->encode_keep_buffer = lua_toboolean(l, 1); } lua_pushboolean(l, cfg->encode_keep_buffer); return 1; } /* On argument: decode enum and set config variables * **options must point to a NULL terminated array of 4 enums * Returns: current enum value */ static void json_enum_option(lua_State *l, const char **options, int *opt1, int *opt2) { int setting; if (lua_gettop(l)) { if (lua_isboolean(l, 1)) setting = lua_toboolean(l, 1) * 3; else setting = luaL_checkoption(l, 1, NULL, options); *opt1 = setting & 1 ? 1 : 0; *opt2 = setting & 2 ? 1 : 0; } else { setting = *opt1 | (*opt2 << 1); } if (setting) lua_pushstring(l, options[setting]); else lua_pushboolean(l, 0); } /* When enabled, rejects: NaN, Infinity, hexidecimal numbers */ static int json_cfg_refuse_invalid_numbers(lua_State *l) { static const char *options_enc_dec[] = { "none", "encode", "decode", "both", NULL }; json_config_t *cfg; json_verify_arg_count(l, 1); cfg = json_fetch_config(l); json_enum_option(l, options_enc_dec, &cfg->encode_refuse_badnum, &cfg->decode_refuse_badnum); return 1; } static int json_destroy_config(lua_State *l) { json_config_t *cfg; cfg = lua_touserdata(l, 1); #ifdef USE_POSIX_USELOCALE if (cfg->posix_locale) freelocale(cfg->posix_locale); #endif if (cfg) strbuf_free(&cfg->encode_buf); cfg = NULL; return 0; } static void json_create_config(lua_State *l) { json_config_t *cfg; int i; cfg = lua_newuserdata(l, sizeof(*cfg)); /* Create GC method to clean up strbuf */ lua_newtable(l); lua_pushcfunction(l, json_destroy_config); lua_setfield(l, -2, "__gc"); lua_setmetatable(l, -2); strbuf_init(&cfg->encode_buf, 0); #ifdef USE_POSIX_USELOCALE cfg->saved_locale = NULL; /* Must not lua_error() before cfg->posix_locale has been initialised */ cfg->posix_locale = newlocale(LC_ALL_MASK, "C", NULL); if (!cfg->posix_locale) luaL_error(l, "Failed to create POSIX locale for JSON"); #endif cfg->encode_sparse_convert = DEFAULT_SPARSE_CONVERT; cfg->encode_sparse_ratio = DEFAULT_SPARSE_RATIO; cfg->encode_sparse_safe = DEFAULT_SPARSE_SAFE; cfg->encode_max_depth = DEFAULT_MAX_DEPTH; cfg->encode_refuse_badnum = DEFAULT_ENCODE_REFUSE_BADNUM; cfg->decode_refuse_badnum = DEFAULT_DECODE_REFUSE_BADNUM; cfg->encode_keep_buffer = DEFAULT_ENCODE_KEEP_BUFFER; json_set_number_precision(cfg, 14); /* Decoding init */ /* Tag all characters as an error */ for (i = 0; i < 256; i++) cfg->ch2token[i] = T_ERROR; /* Set tokens that require no further processing */ cfg->ch2token['{'] = T_OBJ_BEGIN; cfg->ch2token['}'] = T_OBJ_END; cfg->ch2token['['] = T_ARR_BEGIN; cfg->ch2token[']'] = T_ARR_END; cfg->ch2token[','] = T_COMMA; cfg->ch2token[':'] = T_COLON; cfg->ch2token['\0'] = T_END; cfg->ch2token[' '] = T_WHITESPACE; cfg->ch2token['\t'] = T_WHITESPACE; cfg->ch2token['\n'] = T_WHITESPACE; cfg->ch2token['\r'] = T_WHITESPACE; /* Update characters that require further processing */ cfg->ch2token['f'] = T_UNKNOWN; /* false? */ cfg->ch2token['i'] = T_UNKNOWN; /* inf, ininity? */ cfg->ch2token['I'] = T_UNKNOWN; cfg->ch2token['n'] = T_UNKNOWN; /* null, nan? */ cfg->ch2token['N'] = T_UNKNOWN; cfg->ch2token['t'] = T_UNKNOWN; /* true? */ cfg->ch2token['"'] = T_UNKNOWN; /* string? */ cfg->ch2token['+'] = T_UNKNOWN; /* number? */ cfg->ch2token['-'] = T_UNKNOWN; for (i = 0; i < 10; i++) cfg->ch2token['0' + i] = T_UNKNOWN; /* Lookup table for parsing escape characters */ for (i = 0; i < 256; i++) cfg->escape2char[i] = 0; /* String error */ cfg->escape2char['"'] = '"'; cfg->escape2char['\\'] = '\\'; cfg->escape2char['/'] = '/'; cfg->escape2char['b'] = '\b'; cfg->escape2char['t'] = '\t'; cfg->escape2char['n'] = '\n'; cfg->escape2char['f'] = '\f'; cfg->escape2char['r'] = '\r'; cfg->escape2char['u'] = 'u'; /* Unicode parsing required */ } /* ===== ENCODING ===== */ static void json_encode_exception(lua_State *l, json_config_t *cfg, int lindex, const char *reason) { if (!cfg->encode_keep_buffer) strbuf_free(&cfg->encode_buf); LOCALE_RESTORE(cfg); luaL_error(l, "Cannot serialise %s: %s", lua_typename(l, lua_type(l, lindex)), reason); } /* json_append_string args: * - lua_State * - JSON strbuf * - String (Lua stack index) * * Returns nothing. Doesn't remove string from Lua stack */ static void json_append_string(lua_State *l, strbuf_t *json, int lindex) { const char *escstr; int i; const char *str; size_t len; str = lua_tolstring(l, lindex, &len); /* Worst case is len * 6 (all unicode escapes). * This buffer is reused constantly for small strings * If there are any excess pages, they won't be hit anyway. * This gains ~5% speedup. */ strbuf_ensure_empty_length(json, len * 6 + 2); strbuf_append_char_unsafe(json, '\"'); for (i = 0; i < len; i++) { escstr = char2escape[(unsigned char)str[i]]; if (escstr) strbuf_append_string(json, escstr); else strbuf_append_char_unsafe(json, str[i]); } strbuf_append_char_unsafe(json, '\"'); } /* Find the size of the array on the top of the Lua stack * -1 object (not a pure array) * >=0 elements in array */ static int lua_array_length(lua_State *l, json_config_t *cfg) { double k; int max; int items; max = 0; items = 0; lua_pushnil(l); /* table, startkey */ while (lua_next(l, -2) != 0) { /* table, key, value */ if (lua_type(l, -2) == LUA_TNUMBER && (k = lua_tonumber(l, -2))) { /* Integer >= 1 ? */ if (floor(k) == k && k >= 1) { if (k > max) max = k; items++; lua_pop(l, 1); continue; } } /* Must not be an array (non integer key) */ lua_pop(l, 2); return -1; } /* Encode excessively sparse arrays as objects (if enabled) */ if (cfg->encode_sparse_ratio > 0 && max > items * cfg->encode_sparse_ratio && max > cfg->encode_sparse_safe) { if (!cfg->encode_sparse_convert) json_encode_exception(l, cfg, -1, "excessively sparse array"); return -1; } return max; } static void json_encode_descend(lua_State *l, json_config_t *cfg) { cfg->current_depth++; if (cfg->current_depth > cfg->encode_max_depth) { if (!cfg->encode_keep_buffer) strbuf_free(&cfg->encode_buf); LOCALE_RESTORE(cfg); luaL_error(l, "Cannot serialise, excessive nesting (%d)", cfg->current_depth); } } static void json_append_data(lua_State *l, json_config_t *cfg, strbuf_t *json); /* json_append_array args: * - lua_State * - JSON strbuf * - Size of passwd Lua array (top of stack) */ static void json_append_array(lua_State *l, json_config_t *cfg, strbuf_t *json, int array_length) { int comma, i; json_encode_descend(l, cfg); strbuf_append_char(json, '['); comma = 0; for (i = 1; i <= array_length; i++) { if (comma) strbuf_append_char(json, ','); else comma = 1; lua_rawgeti(l, -1, i); json_append_data(l, cfg, json); lua_pop(l, 1); } strbuf_append_char(json, ']'); cfg->current_depth--; } static void json_append_number(lua_State *l, strbuf_t *json, int index, json_config_t *cfg) { double num = lua_tonumber(l, index); if (cfg->encode_refuse_badnum && (isinf(num) || isnan(num))) json_encode_exception(l, cfg, index, "must not be NaN or Inf"); if (isnan(num)) { /* Some platforms may print -nan, just hard code it */ strbuf_append_mem(json, "nan", 3); } else { /* Lowest double printed with %.14g is 21 characters long: * -1.7976931348623e+308 * * Use 32 to include the \0, and a few extra just in case.. */ strbuf_append_fmt(json, 32, cfg->number_fmt, num); } } static void json_append_object(lua_State *l, json_config_t *cfg, strbuf_t *json) { int comma, keytype; json_encode_descend(l, cfg); /* Object */ strbuf_append_char(json, '{'); lua_pushnil(l); /* table, startkey */ comma = 0; while (lua_next(l, -2) != 0) { if (comma) strbuf_append_char(json, ','); else comma = 1; /* table, key, value */ keytype = lua_type(l, -2); if (keytype == LUA_TNUMBER) { strbuf_append_char(json, '"'); json_append_number(l, json, -2, cfg); strbuf_append_mem(json, "\":", 2); } else if (keytype == LUA_TSTRING) { json_append_string(l, json, -2); strbuf_append_char(json, ':'); } else { json_encode_exception(l, cfg, -2, "table key must be a number or string"); /* never returns */ } /* table, key, value */ json_append_data(l, cfg, json); lua_pop(l, 1); /* table, key */ } strbuf_append_char(json, '}'); cfg->current_depth--; } /* Serialise Lua data into JSON string. */ static void json_append_data(lua_State *l, json_config_t *cfg, strbuf_t *json) { int len; switch (lua_type(l, -1)) { case LUA_TSTRING: json_append_string(l, json, -1); break; case LUA_TNUMBER: json_append_number(l, json, -1, cfg); break; case LUA_TBOOLEAN: if (lua_toboolean(l, -1)) strbuf_append_mem(json, "true", 4); else strbuf_append_mem(json, "false", 5); break; case LUA_TTABLE: len = lua_array_length(l, cfg); if (len > 0) json_append_array(l, cfg, json, len); else json_append_object(l, cfg, json); break; case LUA_TNIL: strbuf_append_mem(json, "null", 4); break; case LUA_TLIGHTUSERDATA: if (lua_touserdata(l, -1) == NULL) { strbuf_append_mem(json, "null", 4); break; } default: /* Remaining types (LUA_TFUNCTION, LUA_TUSERDATA, LUA_TTHREAD, * and LUA_TLIGHTUSERDATA) cannot be serialised */ json_encode_exception(l, cfg, -1, "type not supported"); /* never returns */ } } static int json_encode(lua_State *l) { json_config_t *cfg; char *json; int len; /* Can't use json_verify_arg_count() since we need to ensure * there is only 1 argument */ luaL_argcheck(l, lua_gettop(l) == 1, 1, "expected 1 argument"); cfg = json_fetch_config(l); cfg->current_depth = 0; /* Reset the persistent buffer if it exists. * Otherwise allocate a new buffer. */ if (strbuf_allocated(&cfg->encode_buf)) strbuf_reset(&cfg->encode_buf); else strbuf_init(&cfg->encode_buf, 0); LOCALE_SET_POSIX(cfg); json_append_data(l, cfg, &cfg->encode_buf); json = strbuf_string(&cfg->encode_buf, &len); LOCALE_RESTORE(cfg); lua_pushlstring(l, json, len); if (!cfg->encode_keep_buffer) strbuf_free(&cfg->encode_buf); return 1; } /* ===== DECODING ===== */ static void json_process_value(lua_State *l, json_parse_t *json, json_token_t *token); static int hexdigit2int(char hex) { if ('0' <= hex && hex <= '9') return hex - '0'; /* Force lowercase */ hex |= 0x20; if ('a' <= hex && hex <= 'f') return 10 + hex - 'a'; return -1; } static int decode_hex4(const char *hex) { int digit[4]; int i; /* Convert ASCII hex digit to numeric digit * Note: this returns an error for invalid hex digits, including * NULL */ for (i = 0; i < 4; i++) { digit[i] = hexdigit2int(hex[i]); if (digit[i] < 0) { return -1; } } return (digit[0] << 12) + (digit[1] << 8) + (digit[2] << 4) + digit[3]; } /* Converts a Unicode codepoint to UTF-8. * Returns UTF-8 string length, and up to 4 bytes in *utf8 */ static int codepoint_to_utf8(char *utf8, int codepoint) { /* 0xxxxxxx */ if (codepoint <= 0x7F) { utf8[0] = codepoint; return 1; } /* 110xxxxx 10xxxxxx */ if (codepoint <= 0x7FF) { utf8[0] = (codepoint >> 6) | 0xC0; utf8[1] = (codepoint & 0x3F) | 0x80; return 2; } /* 1110xxxx 10xxxxxx 10xxxxxx */ if (codepoint <= 0xFFFF) { utf8[0] = (codepoint >> 12) | 0xE0; utf8[1] = ((codepoint >> 6) & 0x3F) | 0x80; utf8[2] = (codepoint & 0x3F) | 0x80; return 3; } /* 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx */ if (codepoint <= 0x1FFFFF) { utf8[0] = (codepoint >> 18) | 0xF0; utf8[1] = ((codepoint >> 12) & 0x3F) | 0x80; utf8[2] = ((codepoint >> 6) & 0x3F) | 0x80; utf8[3] = (codepoint & 0x3F) | 0x80; return 4; } return 0; } /* Called when index pointing to beginning of UTF-16 code escape: \uXXXX * \u is guaranteed to exist, but the remaining hex characters may be * missing. * Translate to UTF-8 and append to temporary token string. * Must advance index to the next character to be processed. * Returns: 0 success * -1 error */ static int json_append_unicode_escape(json_parse_t *json) { char utf8[4]; /* Surrogate pairs require 4 UTF-8 bytes */ int codepoint; int surrogate_low; int len; int escape_len = 6; /* Fetch UTF-16 code unit */ codepoint = decode_hex4(&json->data[json->index + 2]); if (codepoint < 0) return -1; /* UTF-16 surrogate pairs take the following 2 byte form: * 11011 x yyyyyyyyyy * When x = 0: y is the high 10 bits of the codepoint * x = 1: y is the low 10 bits of the codepoint * * Check for a surrogate pair (high or low) */ if ((codepoint & 0xF800) == 0xD800) { /* Error if the 1st surrogate is not high */ if (codepoint & 0x400) return -1; /* Ensure the next code is a unicode escape */ if (json->data[json->index + escape_len] != '\\' || json->data[json->index + escape_len + 1] != 'u') { return -1; } /* Fetch the next codepoint */ surrogate_low = decode_hex4(&json->data[json->index + 2 + escape_len]); if (surrogate_low < 0) return -1; /* Error if the 2nd code is not a low surrogate */ if ((surrogate_low & 0xFC00) != 0xDC00) return -1; /* Calculate Unicode codepoint */ codepoint = (codepoint & 0x3FF) << 10; surrogate_low &= 0x3FF; codepoint = (codepoint | surrogate_low) + 0x10000; escape_len = 12; } /* Convert codepoint to UTF-8 */ len = codepoint_to_utf8(utf8, codepoint); if (!len) return -1; /* Append bytes and advance parse index */ strbuf_append_mem_unsafe(json->tmp, utf8, len); json->index += escape_len; return 0; } static void json_set_token_error(json_token_t *token, json_parse_t *json, const char *errtype) { token->type = T_ERROR; token->index = json->index; token->value.string = errtype; } static void json_next_string_token(json_parse_t *json, json_token_t *token) { char *escape2char = json->cfg->escape2char; char ch; /* Caller must ensure a string is next */ assert(json->data[json->index] == '"'); /* Skip " */ json->index++; /* json->tmp is the temporary strbuf used to accumulate the * decoded string value. */ strbuf_reset(json->tmp); while ((ch = json->data[json->index]) != '"') { if (!ch) { /* Premature end of the string */ json_set_token_error(token, json, "unexpected end of string"); return; } /* Handle escapes */ if (ch == '\\') { /* Fetch escape character */ ch = json->data[json->index + 1]; /* Translate escape code and append to tmp string */ ch = escape2char[(unsigned char)ch]; if (ch == 'u') { if (json_append_unicode_escape(json) == 0) continue; json_set_token_error(token, json, "invalid unicode escape code"); return; } if (!ch) { json_set_token_error(token, json, "invalid escape code"); return; } /* Skip '\' */ json->index++; } /* Append normal character or translated single character * Unicode escapes are handled above */ strbuf_append_char_unsafe(json->tmp, ch); json->index++; } json->index++; /* Eat final quote (") */ strbuf_ensure_null(json->tmp); token->type = T_STRING; token->value.string = strbuf_string(json->tmp, &token->string_len); } /* JSON numbers should take the following form: * -?(0|[1-9]|[1-9][0-9]+)(.[0-9]+)?([eE][-+]?[0-9]+)? * * json_next_number_token() uses strtod() which allows other forms: * - numbers starting with '+' * - NaN, -NaN, infinity, -infinity * - hexidecimal numbers * - numbers with leading zeros * * json_is_invalid_number() detects "numbers" which may pass strtod()'s * error checking, but should not be allowed with strict JSON. * * json_is_invalid_number() may pass numbers which cause strtod() * to generate an error. */ static int json_is_invalid_number(json_parse_t *json) { int i = json->index; /* Reject numbers starting with + */ if (json->data[i] == '+') return 1; /* Skip minus sign if it exists */ if (json->data[i] == '-') i++; /* Reject numbers starting with 0x, or leading zeros */ if (json->data[i] == '0') { int ch2 = json->data[i + 1]; if ((ch2 | 0x20) == 'x' || /* Hex */ ('0' <= ch2 && ch2 <= '9')) /* Leading zero */ return 1; return 0; } else if (json->data[i] <= '9') { return 0; /* Ordinary number */ } /* Reject inf/nan */ if (!strncasecmp(&json->data[i], "inf", 3)) return 1; if (!strncasecmp(&json->data[i], "nan", 3)) return 1; /* Pass all other numbers which may still be invalid, but * strtod() will catch them. */ return 0; } static void json_next_number_token(json_parse_t *json, json_token_t *token) { const char *startptr; char *endptr; token->type = T_NUMBER; startptr = &json->data[json->index]; token->value.number = strtod(&json->data[json->index], &endptr); if (startptr == endptr) json_set_token_error(token, json, "invalid number"); else json->index += endptr - startptr; /* Skip the processed number */ return; } /* Fills in the token struct. * T_STRING will return a pointer to the json_parse_t temporary string * T_ERROR will leave the json->index pointer at the error. */ static void json_next_token(json_parse_t *json, json_token_t *token) { json_token_type_t *ch2token = json->cfg->ch2token; int ch; /* Eat whitespace. FIXME: UGLY */ token->type = ch2token[(unsigned char)json->data[json->index]]; while (token->type == T_WHITESPACE) token->type = ch2token[(unsigned char)json->data[++json->index]]; token->index = json->index; /* Don't advance the pointer for an error or the end */ if (token->type == T_ERROR) { json_set_token_error(token, json, "invalid token"); return; } if (token->type == T_END) { return; } /* Found a known single character token, advance index and return */ if (token->type != T_UNKNOWN) { json->index++; return; } /* Process characters which triggered T_UNKNOWN */ ch = json->data[json->index]; /* Must use strncmp() to match the front of the JSON string. * JSON identifier must be lowercase. * When strict_numbers if disabled, either case is allowed for * Infinity/NaN (since we are no longer following the spec..) */ if (ch == '"') { json_next_string_token(json, token); return; } else if (ch == '-' || ('0' <= ch && ch <= '9')) { if (json->cfg->decode_refuse_badnum && json_is_invalid_number(json)) { json_set_token_error(token, json, "invalid number"); return; } json_next_number_token(json, token); return; } else if (!strncmp(&json->data[json->index], "true", 4)) { token->type = T_BOOLEAN; token->value.boolean = 1; json->index += 4; return; } else if (!strncmp(&json->data[json->index], "false", 5)) { token->type = T_BOOLEAN; token->value.boolean = 0; json->index += 5; return; } else if (!strncmp(&json->data[json->index], "null", 4)) { token->type = T_NULL; json->index += 4; return; } else if (!json->cfg->decode_refuse_badnum && json_is_invalid_number(json)) { /* When refuse_badnum is disabled, only attempt to process * numbers we know are invalid JSON (Inf, NaN, hex) * This is required to generate an appropriate token error, * otherwise all bad tokens will register as "invalid number" */ json_next_number_token(json, token); return; } /* Token starts with t/f/n but isn't recognised above. */ json_set_token_error(token, json, "invalid token"); } /* This function does not return. * DO NOT CALL WITH DYNAMIC MEMORY ALLOCATED. * The only supported exception is the temporary parser string * json->tmp struct. * json and token should exist on the stack somewhere. * luaL_error() will long_jmp and release the stack */ static void json_throw_parse_error(lua_State *l, json_parse_t *json, const char *exp, json_token_t *token) { const char *found; strbuf_free(json->tmp); if (token->type == T_ERROR) found = token->value.string; else found = json_token_type_name[token->type]; LOCALE_RESTORE(json->cfg); /* Note: token->index is 0 based, display starting from 1 */ luaL_error(l, "Expected %s but found %s at character %d", exp, found, token->index + 1); } static void json_decode_checkstack(lua_State *l, json_parse_t *json, int n) { if (lua_checkstack(l, n)) return; strbuf_free(json->tmp); LOCALE_RESTORE(json->cfg); luaL_error(l, "Too many nested data structures"); } static void json_parse_object_context(lua_State *l, json_parse_t *json) { json_token_t token; /* 3 slots required: * .., table, key, value */ json_decode_checkstack(l, json, 3); lua_newtable(l); json_next_token(json, &token); /* Handle empty objects */ if (token.type == T_OBJ_END) { return; } while (1) { if (token.type != T_STRING) json_throw_parse_error(l, json, "object key string", &token); /* Push key */ lua_pushlstring(l, token.value.string, token.string_len); json_next_token(json, &token); if (token.type != T_COLON) json_throw_parse_error(l, json, "colon", &token); /* Fetch value */ json_next_token(json, &token); json_process_value(l, json, &token); /* Set key = value */ lua_rawset(l, -3); json_next_token(json, &token); if (token.type == T_OBJ_END) return; if (token.type != T_COMMA) json_throw_parse_error(l, json, "comma or object end", &token); json_next_token(json, &token); } } /* Handle the array context */ static void json_parse_array_context(lua_State *l, json_parse_t *json) { json_token_t token; int i; /* 2 slots required: * .., table, value */ json_decode_checkstack(l, json, 2); lua_newtable(l); json_next_token(json, &token); /* Handle empty arrays */ if (token.type == T_ARR_END) return; for (i = 1; ; i++) { json_process_value(l, json, &token); lua_rawseti(l, -2, i); /* arr[i] = value */ json_next_token(json, &token); if (token.type == T_ARR_END) return; if (token.type != T_COMMA) json_throw_parse_error(l, json, "comma or array end", &token); json_next_token(json, &token); } } /* Handle the "value" context */ static void json_process_value(lua_State *l, json_parse_t *json, json_token_t *token) { switch (token->type) { case T_STRING: lua_pushlstring(l, token->value.string, token->string_len); break;; case T_NUMBER: lua_pushnumber(l, token->value.number); break;; case T_BOOLEAN: lua_pushboolean(l, token->value.boolean); break;; case T_OBJ_BEGIN: json_parse_object_context(l, json); break;; case T_ARR_BEGIN: json_parse_array_context(l, json); break;; case T_NULL: /* In Lua, setting "t[k] = nil" will delete k from the table. * Hence a NULL pointer lightuserdata object is used instead */ lua_pushlightuserdata(l, NULL); break;; default: json_throw_parse_error(l, json, "value", token); } } /* json_text must be null terminated string */ static void lua_json_decode(lua_State *l, const char *json_text, int json_len) { json_parse_t json; json_token_t token; json.cfg = json_fetch_config(l); json.data = json_text; json.index = 0; /* Ensure the temporary buffer can hold the entire string. * This means we no longer need to do length checks since the decoded * string must be smaller than the entire json string */ json.tmp = strbuf_new(json_len); LOCALE_SET_POSIX(json.cfg); json_next_token(&json, &token); json_process_value(l, &json, &token); /* Ensure there is no more input left */ json_next_token(&json, &token); if (token.type != T_END) json_throw_parse_error(l, &json, "the end", &token); LOCALE_RESTORE(json.cfg); strbuf_free(json.tmp); } static int json_decode(lua_State *l) { const char *json; size_t len; json_verify_arg_count(l, 1); json = luaL_checklstring(l, 1, &len); /* Detect Unicode other than UTF-8 (see RFC 4627, Sec 3) * * CJSON can support any simple data type, hence only the first * character is guaranteed to be ASCII (at worst: '"'). This is * still enough to detect whether the wrong encoding is in use. */ if (len >= 2 && (!json[0] || !json[1])) luaL_error(l, "JSON parser does not support UTF-16 or UTF-32"); lua_json_decode(l, json, len); return 1; } /* ===== INITIALISATION ===== */ int luaopen_cjson(lua_State *l) { luaL_Reg reg[] = { { "encode", json_encode }, { "decode", json_decode }, { "encode_sparse_array", json_cfg_encode_sparse_array }, { "encode_max_depth", json_cfg_encode_max_depth }, { "encode_number_precision", json_cfg_encode_number_precision }, { "encode_keep_buffer", json_cfg_encode_keep_buffer }, { "refuse_invalid_numbers", json_cfg_refuse_invalid_numbers }, { NULL, NULL } }; /* Use json_config_key as a pointer. * It's faster than using a config string, and more unique */ lua_pushlightuserdata(l, &json_config_key); json_create_config(l); lua_settable(l, LUA_REGISTRYINDEX); luaL_register(l, "cjson", reg); /* Set cjson.null */ lua_pushlightuserdata(l, NULL); lua_setfield(l, -2, "null"); /* Set cjson.version */ lua_pushliteral(l, "1.0.4"); lua_setfield(l, -2, "version"); /* Return cjson table */ return 1; } /* vi:ai et sw=4 ts=4: */