foreverzer0 / sharpnbt Goto Github PK

Why does TagBuilder have to have a name?

In some cases, I don't need a name

For example this data

string testNbtStr1 = "{Count:1b,id:\"minecraft:netherite_sword\",tag:{Damage:0,Enchantments:[{id:\"minecraft:sharpness\",lvl:1s}]}}";

When I tried to install the package into the project (framework 4.7.2), the installation failed with an error.

Error Could not install package 'SharpNBT 1.2.0'. You are trying to install this package into a project that targets '.NETFramework,Version=v4.7.2', but the package does not contain any assembly references or content files that are compatible with that framework. For more information, contact the package author.

Leverage modern runtime features with the new(-ish) Utf8JsonWriter class. The current implementation uses the ISerializable interface, which is not near as flexible, nor convenient to implement in a way that uses the same structure as the NBT.

Title...

Attempting to parse an integer array fails with a SyntaxError if the array contains a negative number.
This effectively means that any SNBT containing a UUID fails as those are length 4 Integer Arrays and UUID format guarantees a negative number here.

A simple check to confirm:

StringNbt.Parse("{uuid:[I; 0, 0, 0, -1]}");

throws

System.Data.SyntaxErrorException: 'Syntax error at index 19: Invalid character '-' in integer array.'

There is a few types in he library that seem to be abstractions for the sake of abstraction, and do not provide anything beneficial. While they do provide a common "parent" to classes like Compound/List, the actual need for such an abstraction is dubious in practical application, and an existing interface such as IEnumerable would suffice if such a need existed.

Being tied to a common concrete class makes both types less flexible.

SharpNBT uses byte, but the wiki mentions that it is a signed byte https://wiki.vg/NBT.

Hello,

I've just created a little console application to see whether SharpNBT suits my needs, I need a small utility to extract nbt tags from a bedrock level.dat file (and be able to edit and write back). I'm still not sure whether SharpNBT is able to read the information from that file format...

But in short, I've found the following problem:
An unhandled exception of type 'System.InvalidCastException' occurred in System.Private.CoreLib.dll: 'Unable to cast object of type 'SharpNBT.StringTag' to type 'SharpNBT.CompoundTag'.'

The code which caused this unhandled exception is pretty simple:

using SharpNBT;

namespace NbtTeset 
{
    class Program
    {
        static void Main(string[] args)
        {
            string file = "/home/david/Downloads/20220828/level.dat";
            var tag = NbtFile.Read(file, FormatOptions.BedrockFile, CompressionType.AutoDetect);
            Console.WriteLine(tag.PrettyPrinted());
        }
    }
}

The exception is thrown at the line where tag is declared and initialized with the result of NbtFile.Read(). I don't know if I am doing something wrong in here, if SharpNBT is unable to handle bedrock worlds or if there is a bug in SharpNBT - that's why I created this issue and hope you may be able to tell me what I'm doing wrong or whether it's really an issue :)

One additional info regarding the file I want to read: it's a world from Bedrock Dedicated Server..

The current CompoundTag implementation behaves akin to a dictionary, but through layers of abstraction, it is actually backed by a list (see #15 ). This works in practice, but breaks the constant-time lookup speed of the CompoundTag.

This proposal is complete #15 and provide and use an actual dictionary for storing child tags.

There is flawed logic in the Stringify method of the CompoundTag class, and likely the ListTag as well. When appending children to the StringBuilder, it fails to call Stringify on them, therefore appending the ToString representation.

This is the data I receive

如果我直接序列化json,NBT 格式和原来的就不对了

If I serialize json directly, the NBT format is wrong with the original one

JsonConvert.SerializeObject(tag);

SharpNBT.SNBT.StringNbt.Parse 之后如何转回原来的？

How do I switch back to the original after SharpNBT.SNBT.StringNbt.Parse?

Add support for container-like tags to support trailing commas without error.

"list": ["Hello", "world",]

This can be enabled via a configuration for the scanner, being disabled by default.

Hi. I am trying to read from the Minecraft level.dat file for a project i am working on. This is my code:

CompoundTag WorldData = NbtFile.Read(WorldFolder + @".\level.dat", FormatOptions.Java, CompressionType.AutoDetect);
var seed = WorldData["Data"]["WorldGenSettings"]["Seed"];

It is giving me the error stating that I cant use [] on the Array.

Rework the "PrettyPrint" logic to output to a Stream/TextWriter, which defaults to the standard output when not specified. This would provide more flexibility than StringBuilder object.

In fact, I did not know how to modify *dat files. I wanted to change the name of the world in Minecraft Java, and I did not know how to modify

why ?

coode:

string testNbtStr3 = "{Count:1b,id:\"minecraft:netherite_sword\",tag:{Damage:0,Enchantments:[{id:\"minecraft:looting\",lvl:3s},{id:\"minecraft:smite\",lvl:5s},{id:\"minecraft:sweeping\",lvl:3s}],RepairCost:7,display:{Name:'{\"extra\":[{\"text\":\"我是修改的名字\"}],\"text\":\"\"}'}}}";
CompoundTag? tags = StringNbt.Parse(testNbtStr3);

The SNBT parser appears to rely heavily on Regex but the regexes aren't compiled.

Specifying the RegexOptions.Compiled in the Regex constructor will yield a massive speedup.

In ParseList, when determining if an container is a ListTag or array of numbers, it checks if the next in the sequence is a number:

if (char.IsNumber(c) || c == ',')

This should also include a check for a number that has a leading + or -.

Apply the immutability to many NBT types.

The following types should be completely immutable after initialization:

• BoolTag ( #17 )
• ByteTag
• ShortTag
• IntTag
• LongTag
• FloatTag
• Doubletag
• StringTag

The following types should be of a fixed size, but allow individual elements to be changed.

• ByteArraytag
• IntArrayTag
• LongArraytag

The following will behave normally, allowing insertion/deletion

• ListTag
• CompoundTag

The TagReader and TagWriter class still use the naive and legacy method of BitConverter.GetBytes paired with Array.Reverse to get the correct byte-order.

Replace all these method calls with the use of stackalloc spans and employ BinaryPrimitives methods to read/write in the appropriate endian. This will make them free of any heap-allocations in additional to being computationally less expensive.

I found some strange things. When parsing, it traverses a ListTag whose type is SharpNBT.TagType.List. Their child item Type is SharpNBT.TagType.Compound, but when I build ListTag

If I

ListTag listTag = new ListTag(NBTEnchants, TagType.List);

Then

   CompoundTag? dicJarrTag = DicToTag(dicJArryItem);

   listTag.Add(dicJarrTag); //An error will be reported here

Exception:

System.ArrayTypeMismatchException: "Incorrect tag type added to this collection."

why is that? Or should my ListTag TagType be set to CompoundTag? So can it still be the same as the original one?

Problem

There are currently three different ways that NBT can be serialized:

I will reserve disparaging Microsoft for needlessly complicating one of the most simple serialization formats in existence. It is par for the course, and not even unexpected anymore. I would be more surprised if they didn't mess it up.

Regardless, the TagReader and TagWriter class accounts for all scenarios using the format options passed to its constructor. While this works in practice, it means most methods have to do additional checks to see what the appropriate way to read the data is.

Proposal

Make TagReader and TagWriter abstract, with a factory method to create an instance. Internally these different ways of reading/writing can be split into separate classes that do not require any runtime checks.

Something akin to this psuedo-code:

public static TagReader Create(Stream stream, Format format)
{
    return format switch
    {
        Format.Java => new JavaReader(stream),
        Format.BedrockFile => new BedrockFileReader(stream),
        Format.BedrockNetwork => new BedrockNetworkReader(stream),
        _ => throw new ArgumentOutOfRangeException()
    };
}

The actual implementations can be completely transparent and need not even be part of the public API.

The SNBT parser lacks support for the Boolean type with values true or false. Internally this type is mapped to Byte when converting to NBT.

Although I think it is important to somehow preserve the original value came from a Boolean such that when rewriting snbt files, the Boolean fields remain as true or false and don't get rewritten as 1b or 0b. Rewriting to 1b or 0b would cause unnecessary changes to appear when diffing snbts. It also makes manual editing more error prone as one loses any indication that the field has only two valid values.

So perhaps an extra property could be added to ByteTag indicating it is really a Boolean, or a BooleanTag could be introduced that has conversions for ByteTag in order to prevent information loss.

If you're using the recommended Contexts workflow, you end up getting log spam every time a Context is disposed.

I know there's ways to suppress the Console temporarily yourself, but it'd be nice to have the option to disable this within TagBuilder.
Thanks for an otherwise solid library 🩷

When comparing compound tags the bar for equality is set extremely low. Two compound objects are equal if they have the same type (guaranteed) and their names are the same.

This is because the CompoundTag class does not override the Tag.Equals method leading to this bizarre behaviour. A very simple check is this:

using SharpNBT.SNBT;
Console.WriteLine(StringNbt.Parse("{value:385}") == StringNbt.Parse("{data:[1,2,3,4]}"));

This will pass even though the two compounds aren't remotely the same.

This is a great open source!
But I encountered some problems in the process of using, I don't know how to solve this problem, my technical level is relatively low

I can't find Value in "Tag", how can I get this value

There was a breaking change in .NET 6 in DeflateStream and GZipStream. These streams can now read less bytes than requested by the Read methods count parameter. You now have to call the Read method multiple times until your buffer is filled or 0 is returned (end of base stream).

In TagReader there are a few instances where BaseStream.Read is used but the return value is ignored. For example: When reading Minecraft chunk files the NBT data is compressed by GZipStream or ZLibStream (uses DeflateStream). Chunk files can contain really long LongArrayTags that are read in a single Read call. Due to the .NET 6 change it is possible that not all bytes are read in a single call. This will mess up the read data, the stream position and the Crc in ZLibStream that will throw an InvalidDataException.

If required I can upload a region file and an example code to reproduce the problem.

Possible solutions:
Create an extension method e.g. ReadFixedLength with the same signature as Stream.Read but without return value. This will call Stream.Read in a loop until all count bytes are read. This should throw an EndOfStreamException if it couldn't read the required number of bytes from the base steam.

CompoundTag How can to StringNbt ？

The BoolTag is an addition to this library that does not exist in the spec. This proposal is to remove it completely, and merge its little functionality into ByteTag with additional IsBool and BoolValue properties, as well as a constructor that accepts a boolean.

When reading from binary, it is impossible to differentiate from a ByteTag, so its only purpose is really for display when parsed from SNBT. That small functionality can be easily achieved by the ByteTag class.