/*
 * [2012-06-09 First Version]
 * - provides strongly typed node classes and lists / dictionaries
 * - provides easy access to class members / array items / data values
 * - the parser now properly identifies types. So generating JSON with this framework should work.
 * - only double quotes (") are used for quoting strings.
 * - provides "casting" properties to easily convert to / from those types:
 *   int / float / double / bool
 * - provides a common interface for each node so no explicit casting is required.
 * - the parser tries to avoid errors, but if malformed JSON is parsed the result is more or less undefined
 * - It can serialize/deserialize a node tree into/from an experimental compact binary format. It might
 *   be handy if you want to store things in a file and don't want it to be easily modifiable
 * 
 * [2012-12-17 Update]
 * - Added internal JSONLazyCreator class which simplifies the construction of a JSON tree
 *   Now you can simple reference any item that doesn't exist yet and it will return a JSONLazyCreator
 *   The class determines the required type by it's further use, creates the type and removes itself.
 * - Added binary serialization / deserialization.
 * - Added support for BZip2 zipped binary format. Requires the SharpZipLib ( http://www.icsharpcode.net/opensource/sharpziplib/ )
 *   The usage of the SharpZipLib library can be disabled by removing or commenting out the USE_SharpZipLib define at the top
 * - The serializer uses different types when it comes to store the values. Since my data values
 *   are all of type string, the serializer will "try" which format fits best. The order is: int, float, double, bool, string.
 *   It's not the most efficient way but for a moderate amount of data it should work on all platforms.
 * 
 * [2017-03-08 Update]
 * - Optimised parsing by using a StringBuilder for token. This prevents performance issues when large
 *   string data fields are contained in the json data.
 * - Finally refactored the badly named JSONClass into JSONObject.
 * - Replaced the old JSONData class by distict typed classes ( JSONString, JSONNumber, JSONBool, JSONNull ) this
 *   allows to propertly convert the node tree back to json without type information loss. The actual value
 *   parsing now happens at parsing time and not when you actually access one of the casting properties.
 * 
 * [2017-04-11 Update]
 * - Fixed parsing bug where empty string values have been ignored.
 * - Optimised "ToString" by using a StringBuilder internally. This should heavily improve performance for large files
 * - Changed the overload of "ToString(string aIndent)" to "ToString(int aIndent)"
 * 
 * [2017-11-29 Update]
 * - Removed the IEnumerator implementations on JSONArray & JSONObject and replaced it with a common
 *   struct Enumerator in JSONNode that should avoid garbage generation. The enumerator always works
 *   on KeyValuePair<string, JSONNode>, even for JSONArray.
 * - Added two wrapper Enumerators that allows for easy key or value enumeration. A JSONNode now has
 *   a "Keys" and a "Values" enumerable property. Those are also struct enumerators / enumerables
 * - A KeyValuePair<string, JSONNode> can now be implicitly converted into a JSONNode. This allows
 *   a foreach loop over a JSONNode to directly access the values only. Since KeyValuePair as well as
 *   all the Enumerators are structs, no garbage is allocated.
 * - To add Linq support another "LinqEnumerator" is available through the "Linq" property. This
 *   enumerator does implement the generic IEnumerable interface so most Linq extensions can be used
 *   on this enumerable object. This one does allocate memory as it's a wrapper class.
 * - The Escape method now escapes all control characters (# < 32) in strings as uncode characters
 *   (\uXXXX) and if the static bool JSONNode.forceASCII is set to true it will also escape all
 *   characters # > 127. This might be useful if you require an ASCII output. Though keep in mind
 *   when your strings contain many non-ascii characters the strings become much longer (x6) and are
 *   no longer human readable.
 * - The node types JSONObject and JSONArray now have an "Inline" boolean switch which will default to
 *   false. It can be used to serialize this element inline even you serialize with an indented format
 *   This is useful for arrays containing numbers so it doesn't place every number on a new line
 * - Extracted the binary serialization code into a seperate extension file. All classes are now declared
 *   as "partial" so an extension file can even add a new virtual or abstract method / interface to
 *   JSONNode and override it in the concrete type classes. It's of course a hacky approach which is
 *   generally not recommended, but i wanted to keep everything tightly packed.
 * - Added a static CreateOrGet method to the JSONNull class. Since this class is immutable it could
 *   be reused without major problems. If you have a lot null fields in your data it will help reduce
 *   the memory / garbage overhead. I also added a static setting (reuseSameInstance) to JSONNull
 *   (default is true) which will change the behaviour of "CreateOrGet". If you set this to false
 *   CreateOrGet will not reuse the cached instance but instead create a new JSONNull instance each time.
 *   I made the JSONNull constructor private so if you need to create an instance manually use
 *   JSONNull.CreateOrGet()
 * 
 * [2018-01-09 Update]
 * - Changed all double.TryParse and double.ToString uses to use the invariant culture to avoid problems
 *   on systems with a culture that uses a comma as decimal point.
 * 
 * [2018-01-26 Update]
 * - Added AsLong. Note that a JSONNumber is stored as double and can't represent all long values. However
 *   storing it as string would work.
 * - Added static setting "JSONNode.longAsString" which controls the default type that is used by the
 *   LazyCreator when using AsLong
 * 
 * [2018-04-25 Update]
 *  - Added support for parsing single values (JSONBool, JSONString, JSONNumber, JSONNull) as top level value.
 * 
 * [2019-02-18 Update]
 *  - Added HasKey(key) and GetValueOrDefault(key, default) to the JSONNode class to provide way to read
 *    values conditionally without creating a LazyCreator
 * 
 * [2019-03-25 Update]
 *  - Added static setting "allowLineComments" to the JSONNode class which is true by default. This allows
 *    "//" line comments when parsing json text as long as it's not within quoted text. All text after // up
 *    to the end of the line is completely ignored / skipped. This makes it easier to create human readable
 *    and editable files. Note that stripped comments are not read, processed or preserved in any way. So
 *    this feature is only relevant for human created files.
 *  - Explicitly strip BOM (Byte Order Mark) when parsing to avoid getting it leaked into a single primitive
 *    value. That's a rare case but better safe than sorry.
 *  - Allowing adding the empty string as key
 *  
 * [2019-12-10 Update]
 *  - Added Clone() method to JSONNode to allow cloning of a whole node tree.
 * 
 * [2020-09-19 Update]
 *  - Added Clear() method to JSONNode.
 *  - The parser will now automatically mark arrays or objects as inline when it doesn't contain any
 *    new line characters. This should more or less preserve the layout.
 *  - Added new extension file "SimpleJSONDotNetTypes.cs" to provide support for some basic .NET types
 *    like decimal, char, byte, sbyte, short, ushort, uint, DateTime, TimeSpan and Guid as well as some
 *    nullable types.
 *  - Fixed an error in the Unity extension file. The Color component order was wrong (it was argb, now it's rgba)
 *  - There are now two static float variables (ColorDefaultAlpha and Color32DefaultAlpha) to specify the default
 *    alpha values when reading UnityEngine.Color / Color32 values where the alpha value is absent. The default
 *    values are 1.0f and 255 respectively.
 */