1 // Copyright (c) 2006-2008 The Chromium Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style license that can be 3 // found in the LICENSE file. 4 // 5 // A JSON parser. Converts strings of JSON into a Value object (see 6 // base/values.h). 7 // http://www.ietf.org/rfc/rfc4627.txt?number=4627 8 // 9 // Known limitations/deviations from the RFC: 10 // - Only knows how to parse ints within the range of a signed 32 bit int and 11 // decimal numbers within a double. 12 // - Assumes input is encoded as UTF8. The spec says we should allow UTF-16 13 // (BE or LE) and UTF-32 (BE or LE) as well. 14 // - We limit nesting to 100 levels to prevent stack overflow (this is allowed 15 // by the RFC). 16 // - A Unicode FAQ ("http://unicode.org/faq/utf_bom.html") writes a data 17 // stream may start with a Unicode Byte-Order-Mark (U+FEFF), i.e. the input 18 // UTF-8 string for the JSONReader::JsonToValue() function may start with a 19 // UTF-8 BOM (0xEF, 0xBB, 0xBF). 20 // To avoid the function from mis-treating a UTF-8 BOM as an invalid 21 // character, the function skips a Unicode BOM at the beginning of the 22 // Unicode string (converted from the input UTF-8 string) before parsing it. 23 // 24 // TODO(tc): Add a parsing option to to relax object keys being wrapped in 25 // double quotes 26 // TODO(tc): Add an option to disable comment stripping 27 // TODO(aa): Consider making the constructor public and the static Read() method 28 // only a convenience for the common uses with more complex configuration going 29 // on the instance. 30 31 #ifndef BASE_JSON_JSON_READER_H_ 32 #define BASE_JSON_JSON_READER_H_ 33 34 #include <string> 35 36 #include "base/basictypes.h" 37 38 // Chromium and Chromium OS check out gtest to different places, so we're 39 // unable to compile on both if we include gtest_prod.h here. Instead, include 40 // its only contents -- this will need to be updated if the macro ever changes. 41 #define FRIEND_TEST(test_case_name, test_name)\ 42 friend class test_case_name##_##test_name##_Test 43 44 class Value; 45 46 namespace base { 47 48 class JSONReader { 49 public: 50 // A struct to hold a JS token. 51 class Token { 52 public: 53 enum Type { 54 OBJECT_BEGIN, // { 55 OBJECT_END, // } 56 ARRAY_BEGIN, // [ 57 ARRAY_END, // ] 58 STRING, 59 NUMBER, 60 BOOL_TRUE, // true 61 BOOL_FALSE, // false 62 NULL_TOKEN, // null 63 LIST_SEPARATOR, // , 64 OBJECT_PAIR_SEPARATOR, // : 65 END_OF_INPUT, 66 INVALID_TOKEN, 67 }; Token(Type t,const wchar_t * b,int len)68 Token(Type t, const wchar_t* b, int len) 69 : type(t), begin(b), length(len) {} 70 71 Type type; 72 73 // A pointer into JSONReader::json_pos_ that's the beginning of this token. 74 const wchar_t* begin; 75 76 // End should be one char past the end of the token. 77 int length; 78 79 // Get the character that's one past the end of this token. NextChar()80 wchar_t NextChar() { 81 return *(begin + length); 82 } 83 }; 84 85 // Error messages that can be returned. 86 static const char* kBadRootElementType; 87 static const char* kInvalidEscape; 88 static const char* kSyntaxError; 89 static const char* kTrailingComma; 90 static const char* kTooMuchNesting; 91 static const char* kUnexpectedDataAfterRoot; 92 static const char* kUnsupportedEncoding; 93 static const char* kUnquotedDictionaryKey; 94 95 JSONReader(); 96 97 // Reads and parses |json|, returning a Value. The caller owns the returned 98 // instance. If |json| is not a properly formed JSON string, returns NULL. 99 // If |allow_trailing_comma| is true, we will ignore trailing commas in 100 // objects and arrays even though this goes against the RFC. 101 static Value* Read(const std::string& json, bool allow_trailing_comma); 102 103 // Reads and parses |json| like Read(). |error_message_out| is optional. If 104 // specified and NULL is returned, |error_message_out| will be populated with 105 // a string describing the error. Otherwise, |error_message_out| is 106 // unmodified. 107 static Value* ReadAndReturnError(const std::string& json, 108 bool allow_trailing_comma, 109 std::string* error_message_out); 110 111 // Returns the error message if the last call to JsonToValue() failed. If the 112 // last call did not fail, returns a valid empty string. error_message()113 std::string error_message() { return error_message_; } 114 115 // Reads and parses |json|, returning a Value. The caller owns the returned 116 // instance. If |json| is not a properly formed JSON string, returns NULL and 117 // a detailed error can be retrieved from |error_message()|. 118 // If |check_root| is true, we require that the root object be an object or 119 // array. Otherwise, it can be any valid JSON type. 120 // If |allow_trailing_comma| is true, we will ignore trailing commas in 121 // objects and arrays even though this goes against the RFC. 122 Value* JsonToValue(const std::string& json, bool check_root, 123 bool allow_trailing_comma); 124 125 private: 126 static std::string FormatErrorMessage(int line, int column, 127 const char* description); 128 129 DISALLOW_COPY_AND_ASSIGN(JSONReader); 130 131 FRIEND_TEST(JSONReaderTest, Reading); 132 FRIEND_TEST(JSONReaderTest, ErrorMessages); 133 134 // Recursively build Value. Returns NULL if we don't have a valid JSON 135 // string. If |is_root| is true, we verify that the root element is either 136 // an object or an array. 137 Value* BuildValue(bool is_root); 138 139 // Parses a sequence of characters into a Token::NUMBER. If the sequence of 140 // characters is not a valid number, returns a Token::INVALID_TOKEN. Note 141 // that DecodeNumber is used to actually convert from a string to an 142 // int/double. 143 Token ParseNumberToken(); 144 145 // Try and convert the substring that token holds into an int or a double. If 146 // we can (ie., no overflow), return the value, else return NULL. 147 Value* DecodeNumber(const Token& token); 148 149 // Parses a sequence of characters into a Token::STRING. If the sequence of 150 // characters is not a valid string, returns a Token::INVALID_TOKEN. Note 151 // that DecodeString is used to actually decode the escaped string into an 152 // actual wstring. 153 Token ParseStringToken(); 154 155 // Convert the substring into a value string. This should always succeed 156 // (otherwise ParseStringToken would have failed). 157 Value* DecodeString(const Token& token); 158 159 // Grabs the next token in the JSON stream. This does not increment the 160 // stream so it can be used to look ahead at the next token. 161 Token ParseToken(); 162 163 // Increments |json_pos_| past leading whitespace and comments. 164 void EatWhitespaceAndComments(); 165 166 // If |json_pos_| is at the start of a comment, eat it, otherwise, returns 167 // false. 168 bool EatComment(); 169 170 // Checks if |json_pos_| matches str. 171 bool NextStringMatch(const std::wstring& str); 172 173 // Creates the error message that will be returned to the caller. The current 174 // line and column are determined and added into the final message. 175 void SetErrorMessage(const char* description, const wchar_t* error_pos); 176 177 // Pointer to the starting position in the input string. 178 const wchar_t* start_pos_; 179 180 // Pointer to the current position in the input string. 181 const wchar_t* json_pos_; 182 183 // Used to keep track of how many nested lists/dicts there are. 184 int stack_depth_; 185 186 // A parser flag that allows trailing commas in objects and arrays. 187 bool allow_trailing_comma_; 188 189 // Contains the error message for the last call to JsonToValue(), if any. 190 std::string error_message_; 191 }; 192 193 } // namespace base 194 195 #endif // BASE_JSON_JSON_READER_H_ 196