CHANGELOG.md

Cyotek.Data.Nbt Change Log

[3.0.0-alpha] - 2016-12-27

This release concentrates on cleaning up the code by removing unused or rarely used features, solving boxing issues, removing some hacks, generally reworking the API and increasing the testing coverage. Some minor code changes may be required due to the level of breaking change this release introduces.

A fair chunk of both the library code and test code are now generated via T4 templates.

Added

• With the base Tag class no longer offering a Value property, GetValue and SetValue methods are available. However, note that these methods will box value types
• The TagWriter class can now be used to directly write NBT data to a stream without having to first create concrete Tag classes (e.g. similar to using XmlWriter over XmlDocument)
• The base Tag class now implements IEquatable<Tag> (boxing)
• Each concrete tag class now implements IEquatable<> (non-boxing)
• Added indexers to TagCompound
• Added TagCompound.Count
• Added Benchmarks project testing the different serialization methods. Unsurprisingly, XML is many times slower than binary, and writing NBT documents without constructing Tag objects is faster than creating and then saving a NbtDocument
• Added new TagDictionary.AddRange overloads
• All library code is now covered by tests

Fixed

• All tags created internally by the library use TagFactory and avoid all of the boxing issues present in previous versions
• XmlTagReader crashed if empty byte or int array values were present
• Calling XmlTagReader.IsNbtDocument would return true if a type attribute was found, regardless of if the value was TagCompound or not
• TagDictionary.Add(name, object) didn't support TagDictionary or TagCollection values
• TagCollection.Add(object) now correctly supports TagDictionary and TagCollection values
• TagCollection.Add will now correctly throw exceptions if named tags are added
• Resolved an infinite loop reading empty TagCompound or TagList data stored in XML files

Changed

• Tag names should now be empty when not set rather than null
• TagReader.ReadCollection and ReadDictionary renamed to ReadList and ReadCompound to match their NBT types.
• TagFactory.Create methods have had their parameters shuffled so that name comes first, mirroring TagDictionary and other TagFactory methods
• TagCollection contents will automatically set the ListType based on the first value added when no explicit type is defined
• TagDictionary.Add methods now return the appropriate concrete Tag instance for the value's data type
• TagCompound.GetBoolValue renamed to GetBooleanValue
• TagCompound.Query should no longer throw exceptions when a match cannot be found, but instead returns null

Removed

• Removed the WriteTagOptions and ReadTagOptions enumerations, plus removed any overloaded method supplying these options. Each reader and writer now maintains its own state to know when it should or should not be doing things without having to be told
• Removed the ITag interface as it was a pointless level of abstraction and there is already an abstract Tag class
• Similarly, ITagReader and ITagWriter have also been removed
• Removed the default constructors from tag readers and writers
• Removed the default Tag.Value implementation. Each Tag implementation has a strongly typed Value property without boxing.
• Removed all events from the Tag object as they added overhead without being used in most use cases
• Removed the TagException class as it was unused
• Removed the TagEventArgs class as it was unused
• Removed Tag.ToString(string) overloads
• Removed Tag.CanRemove and Tag.Remove(), as they are too situational
• Tag properties are no longer virtual
• Removed the CompressionOptions enum and related support from NbtDocument. When saving to a file, XML will be uncompressed, binary will be gzipped. When saving to a Stream, you can pass in your own GZipStream, DeflateStream or equivalent
• All TagCollection.Add method overloads that accepted a name argument have been removed
• Removed secondary data type helpers from TagCollection
• Removed TagCompound.GetEnumValue

[2.1.0] - 2016-02-24

Added

• Added new constructor to XmlTagWriter allowing you to specify a XmlWriter. Useful for when calling Write* methods directly, without first using WriteDocument
• Added new constructor to XmlTagReader allowing you to specify a XmlReader. Useful for when calling Read* methods directly, without first using ReadDocument

Fixed

• XML documents that either didn't include any whitespace between elements or that used self closing tags weren't loaded correctly

[2.0.0] - 2016-02-16

This major update includes breaking changes, mostly around the ITagReader and ITagWriter interfaces and implementing classes. The NbtDocument class has been updated so that methods such as Load and Save now can accept a Stream rather than always being forced to use a file.

I've also changed the license to the GNU Lesser General Public License, version 2.1. Originally I said it was licensed under Creative Commons Share Alike to match the original code this library was derived from, but it was actually LGPL all along.

Added

• Added overloads accepting Stream arguments to LoadDocument, Load, Save, IsNbtDocument and GetDocumentFormat methods
• When saving a document, you can now specify Auto for compression which essentially lets the serializer choose the compression (on for binary, off for XML)

Changed

• The ITagWriter and ITagReader interfaces have also been intensively refactored to use separate enums for options and compression (see NbtOptions in the Removed section), and to use a source Stream rather than a filename, and to remove some very silly API designs
• The Custom value for the NbtFormat enum has been renamed to Invalid reflecting the fact that custom support is no longer supported natively
• Added (hopefully) correct detection of gzip or deflate streams, removing the awful practice of relying on exceptions to drop between the formats
• All interfaces, classes and enums which strictly deal with serialization or de-serialization have been moved to the Serialization namespace
• Methods which would previously accept null input, or missing files such as IsNbtDocument, GetDocumentName and GetDocument format now throw the appropriate ArgumentNullException or FileNotFoundException exceptions

Removed

• Removed the NbtOptions enum, replacing it with CompressionOption, WriteTagOption and ReadTagOption which appropriate options. The SingleUse flag has been fully removed as it was a hack for XML support
• Removed the ITagEditor interface, the TagEditorAttribute class and all defined instances of TagEditor. The editor support doesn't really belong in the core library, but rather by the separate GUI library.
• Removed the TagReader and TagWriter classes as they were fairly pointless
• Removed the ReaderType and WriterType properties from NbtDocument along with constructors that accepted types, NbtFormat values, or file names.
• Removed the static NbtDocument.DefaultFormat property
• Removed the static IsDeflateDocument, IsGzipDocument, IsXmlDocument and IsRawDocument methods from NbtDocument - they never should have been public in the first place, and have been replaced with proper stream detection
• NbtDocument no longer supports custom serialization options

Fixed

• The ITagWriter interface is now correctly public

1.0.4.5

Changed

• Reformatted using C# 6 syntax, and minor corrections to avoid double casts

Fixed

• Added culture info to date conversions

1.0.4.4

Added

• Added ToString overrides to TagCollection and TagDictionary

Fixed

• Saving an empty byte array didn't store the zero length size correctly, causing a crash when trying to reload the document. Documents affected by this bug are essentially corrupt and can't be loaded.

1.0.4.3

Fixed

• Removed EnumExtensions class and replaced all HasFlag calls with bitwise statements

1.0.4.2

Fixed

• Added missing GetByteValue to TagCompound

1.0.4.1

Fixed

• NbtDocument.GetDocumentName now automatically returns false if the format of the document is not XML or binary.
• Removed exception when lists have a type of 0, as although it makes sense as it's essentially an invalid tag, some Minecraft tags are using zero type lists.

1.0.4.0

Added

• Added tests for Tag.ToValueString
• ICollectionTag now explicitly inherits from ITag so you don't have to cast back to the former if you only have a reference to the latter.
• Added new TagEventArgs helper class

Fixed

• TagIntArray.ToValueString and TagByteArray.ToValueString now return appropriate values instead of the .NET type name for their respective arrays.

1.0.3.6

Added

• Added AddRange method to TagDictionary
• Added AddRange and Add(bool) methods to TagCollection
• Added tests for TagDictionary, TagCollection and TagEnd, and added some other tests for classes without full coverage [still more to go]
• Added new NbtOptions.HeaderOnly flag, allowing the name of a tag to be read without loading the full body

Changed

• Tag.Value is now virtual
• Setting the value of the TagEnd class now has no effect, and retrieving the value always returns null

Fixed

• Calling NbtDocument.GetDocumentName no longer loads the entire NBT document to return a single string value
• TagDictionary.Add(Guid) and TagCollection.Add(Guid) now return ITag instead of void
• Removed AddIfNotDefault methods from TagDictionary
• BinaryTagReader.Load always passed in default options rather than whatever the caller had specified

1.0.3.5

Fixed

• Changed TagEditor attributes on tag classes to remove fixed version information

1.0.3.4

Changed

• Code refactoring according to Resharper suggestions

Fixed

• Fixed corrupt document tree and crash when loading XML document with self closing tags

1.0.3.3

Added

• Added a number of TagCollection.Add overloads without the name parameter to make it somewhat easier to populate lists via code.
• Added a few more tests

Fixed

• Fixed problems trying to load invalid documents with the NbtDocument class and having them misidentified as "custom" documents.
• Fixed an infinite loop when loading an XML file containing a list tag with no children

1.0.3.2 and earlier

• Change history not available in this document