kermalis / endianbinaryio Goto Github PK

For example, if using BytesToInt16s<ushort>(), and a negative short value is read, the crash will occur because a ushort will not hold the value. I thought it'd cast to a ushort.

ReadObject should call .Read() on IBinarySerializable, instead it creates a new object and reads each member within it.

field value is never set

Every property is checked for attributes even if the types will not use the attributes. This is very slow.

String null termination checking should be better; currently it will default to null termination unless there's a string length attribute. Attributes should be enforced.

The char counts used are for C# strings, which are UTF16 by default. Therefore, a surrogate pair in a UTF16 C# string is 2 chars, but 1 char in UTF32. Writing will write 2 chars technically, and you'd want to read 1 UTF32 char. However, it will only be the first half of the surrogate pair when converted to a string, since the char array will be shrunk down to 1

The custom implementation may flip the bytes too many times or too little.

Reading should throw an error on EndianBinaryIO's side because the underlying type would be unknown.
Writing should not throw; it instead should use value.GetType() instead of typeof(TEnum) to fetch the correct type rather than Enum itself in this case

Array, primitive, interface, enum, and pointer types should be blocked.

Not sure how I didn't notice, but you get an ArgumentException every time you try to read one

Should String and char[] be flipped with endianness (note BigEndianUnicode)?
Will the current implementation always work? Encoding sizes for each char might be different depending on the char, so I think PeekChar and ReadChar will be broken under some circumstances.

UTF requires variable length chars for some (except for UTF-32)
I knew this when I started the library, but I was never sure how to tackle this properly with streams.

Crashes when setting the position of the stream back, since the length of the buffer is longer than the amount of bytes read by the stream

Dispose only disposes of the underlying stream, which the user should be doing.

What license of this library under? It would be good to add a license file to the project - either publish it under a permissive license, or, alternately, under LGPL license (the library is being used by https://github.com/Kermalis/VGMusicStudio project, whose source code is under LGPL license).