Difference between revisions of "NBT"

From wiki.vg
Jump to navigation Jump to search
m (<pre> instead of <code>)
(→‎Libraries: add craftflow-nbt implementation)
 
(72 intermediate revisions by 42 users not shown)
Line 1: Line 1:
The Named Binary Tag (NBT) file format is an extremely simple, albeit annoying (did we really need yet ''another'' format?)<sup>[See Discussion]</sup> structured binary format used by the [http://www.minecraft.net Minecraft] game for a variety of things. Due to this, several third-party utilities now also utilize the format. You may find example files at: http://wiki.vg/NBT
+
The Named Binary Tag (NBT) file format is an extremely simple and efficient structured binary format used by Minecraft for a variety of things. Due to this, several third-party utilities now also utilize the format. You may find example files at the bottom of this article.
 +
 
 +
Mojang has released a reference implementation along with their Anvil conversion tool, available from [https://web.archive.org/web/20190710093131/https://mojang.com/2012/02/new-minecraft-map-format-anvil/ this archived page]
  
 
== Current Uses ==
 
== Current Uses ==
Line 10: Line 12:
 
** Chunk data (see [[Region Files]])
 
** Chunk data (see [[Region Files]])
  
Unfortunately, the NBT files you can encounter as a developer will be stored in three different ways, just to make things interesting.
+
NBT files you can encounter as a developer will be stored in three different ways, mainly the second variation as per Notch's original specification.
 
* Uncompressed,
 
* Uncompressed,
 
* [[wikipedia:Gzip|gzip'd]],
 
* [[wikipedia:Gzip|gzip'd]],
Line 19: Line 21:
 
There are many, many libraries for manipulating NBT, written in several languages, and often several per language. For example,
 
There are many, many libraries for manipulating NBT, written in several languages, and often several per language. For example,
  
* [https://github.com/FliPPeh/cNBT C],
+
{| class="wikitable sortable" style="width: auto; text-align: center;"
* [https://github.com/fragmer/fNbt C#],
+
|-style="background:#eee"
* [https://github.com/Dav1dde/nbd D],
+
!Name
* [https://github.com/toqueteos/minero/tree/master/proto/nbt Go],
+
!Description
* [https://github.com/flow/nbt Java],
+
!Language
* [https://github.com/sjmulder/nbt-js Javascript],  
+
|-
* [https://github.com/TheFrozenFire/PHP-NBT-Decoder-Encoder PHP],
+
! [https://github.com/nickelpro/cNBT cNBT]
* [https://github.com/twoolie/NBT Python],
+
| As simple (and, consequently, as fast) as possible NBT file parser.
* [https://github.com/mental/nbtfile Ruby],
+
| {{C}}
* [https://github.com/PistonDevelopers/hematite_nbt Rust],
+
|-
* [https://github.com/drXor/ScalaNBT Scala],
+
! [https://sr.ht/~azbantium/libnbt/ libnbt]
* You get the idea…
+
| A lightweight library to work with NBT, written in C.
 +
| {{C}}
 +
|-
 +
! [https://github.com/PrismLauncher/libnbtplusplus libnbt++]
 +
| A free C++ library for Minecraft's file format Named Binary Tag (NBT).
 +
| {{C++}}
 +
|-
 +
! [https://github.com/SpockBotMC/cpp-nbt cpp-nbt]
 +
| A C++23 header-only library for reading/writing Minecraft NBT data.
 +
| {{C++}}
 +
|-
 +
! [https://github.com/Bentechy66/venom Venom]
 +
| An NBT decoder written in pure Elixir.
 +
| [https://elixir-lang.org/ Elixir]
 +
|-
 +
! [https://github.com/Tnze/go-mc/tree/master/nbt go-nbt]
 +
| This package implements the Named Binary Tag format of Minecraft.
 +
| [https://go.dev/ Go]
 +
|-
 +
! [https://github.com/BitBuf/nbt BitBuf/nbt]
 +
| Flexible and intuitive library for reading and writing Minecraft's NBT format.
 +
| {{Java}}
 +
|-
 +
! [https://github.com/jglrxavpok/hephaistos hephaistos]
 +
| This library is both a NBT library and a Minecraft Anvil format library.
 +
| {{Java}} & {{Kotlin}}
 +
|-
 +
! [https://github.com/TheNullicorn/Nedit Nedit]
 +
| A simple, lightweight NBT parsing library
 +
| {{Java}}
 +
|-
 +
! [https://github.com/sjmulder/nbt-js NBT.js]
 +
| A JavaScript parser and serializer for NBT archives.
 +
| {{JavaScript}}
 +
|-
 +
! [https://github.com/Offroaders123/NBTify NBTify]
 +
| Parser & writer NBT library intended for usage in a web browser.
 +
| {{JavaScript}}, {{TypeScript}}
 +
|-
 +
! [https://github.com/BenWoodworth/knbt knbt]
 +
| Kotlin NBT library for kotlinx.serialization.
 +
| {{Kotlin}}
 +
|-
 +
! [https://github.com/luizrcs/KotlinNBT KotlinNBT]
 +
| With a builder DSL and type-safety
 +
| {{Kotlin}}
 +
|-
 +
! [https://gist.github.com/camdenorrb/bec73c5608267f0232bd8f5c42e0784d simpleNBT]
 +
| Streams, ByteBuffer, NIO, Endianness, Zlib, Gzip, Any Input/Output, Examples in Comments.
 +
| {{Kotlin}}
 +
|-
 +
! [https://github.com/Nimberite-Development/TagForge-Nim TagForge]
 +
| A library made for the serialisation and deserialisation of MC NBT!
 +
| [https://nim-lang.org/ Nim]
 +
|-
 +
! [https://github.com/aternosorg/php-nbt php-nbt]
 +
| A full PHP implementation of Minecraft's Named Binary Tag (NBT) format.
 +
| {{PHP}}
 +
|-
 +
! [https://github.com/twoolie/NBT NBT]
 +
| This is mainly a Named Binary Tag parser & writer library.
 +
| {{Python}}
 +
|-
 +
! [https://github.com/CorentinPtrl/hematite_nbt HematiteNBT]
 +
| A full-featured Rust crate for Minecraft's NBT file format, including Serde support.
 +
| {{Rust}}
 +
|-
 +
! [https://github.com/Rusty-Quartz/quartz_nbt Quartz NBT]
 +
| Provides support for encoding and decoding Minecraft's NBT format.
 +
| {{Rust}}
 +
|-
 +
! [https://github.com/azalea-rs/simdnbt SimdNBT]
 +
| A very fast NBT serializer and deserializer.
 +
| {{Rust}}
 +
|-
 +
! [https://github.com/owengage/fastnbt fastnbt]
 +
| Fast serde serializer and deserializer for Minecraft's NBT and Anvil formats
 +
| {{Rust}}
 +
|-
 +
! [https://github.com/shenjackyuanjie/nbt-rust shen-nbt5]
 +
| A Fast NBT parser/writer (warning: lot's of unsafe (from author))
 +
| {{Rust}}
 +
|-
 +
! [https://github.com/CrabCraftDev/CrabNBT CrabNBT]
 +
| Up-to-date Rust crate for easy and intuitive working with NBT data.
 +
| {{Rust}}
 +
|-
 +
! [https://crates.io/crates/craftflow-nbt CraftFlow-NBT]
 +
| Serde based NBT binary format implementation.
 +
| {{Rust}}
 +
|-
 +
! [https://github.com/drXor/ScalaNBT ScalaNBT]
 +
| Scala library for NBT io with some Mojangson support.
 +
| [https://www.scala-lang.org/ Scala]
 +
|-
 +
|}
  
 
Unless you have specific goals or licence requirements, it is ''extremely recommended'' to go with one of the existing libraries.
 
Unless you have specific goals or licence requirements, it is ''extremely recommended'' to go with one of the existing libraries.
Line 36: Line 133:
 
=== Utilities ===
 
=== Utilities ===
 
Almost every 3rd-party Minecraft application uses NBT on some level. There also exist several dedicated NBT editors, which will likely be useful to you if you are developing an NBT library of your own. These include:
 
Almost every 3rd-party Minecraft application uses NBT on some level. There also exist several dedicated NBT editors, which will likely be useful to you if you are developing an NBT library of your own. These include:
* [http://www.minecraftforum.net/topic/6661-nbtedit/ NBTEdit] (C#, Mono-capable), one of the very first NBT editors.
+
* [https://www.minecraftforum.net/forums/mapping-and-modding-java-edition/minecraft-tools/1262665-nbtexplorer-nbt-editor-for-windows-and-mac NBTExplorer] (C#) NBT Directory-tree interface that fully supports the Minecraft .mcr/.mca region files.
* [http://gerritg.de/?p=152 NEINedit] (Obj-C), an OS X specific editor.
+
* [https://web.archive.org/web/20180428090808/http://gerritg.de/wp/archives/152 NEINedit] (Obj-C), an OS X specific editor.
 
* [https://bitbucket.org/zzzeek/nbt2yaml nbt2yaml] (Python), provides command-line editing of NBT via the YAML format, as well as a fast and minimalist NBT parsing/rendering API.
 
* [https://bitbucket.org/zzzeek/nbt2yaml nbt2yaml] (Python), provides command-line editing of NBT via the YAML format, as well as a fast and minimalist NBT parsing/rendering API.
 +
* [https://github.com/C4K3/nbted nbted] (Rust; CC0), provides command-line editing of NBT files via your $EDITOR
 +
* [https://git.sleeping.town/unascribed/unbted unbted] (Rust; GPL-v3) Command-line interactive NBT editor
 +
* [https://github.com/midnightfreddie/nbt2json nbt2json] (Golang; MIT) Command-line utility for NBT to JSON/YAML conversion and back. MCPE-NBT support. Can be used as library.
 +
* [https://github.com/tryashtar/nbt-studio NBT Studio] (C#) A visual editor similar to NBT-Explorer (A claimed, spiritual successor). Supports Minecraft Bedrock Edition files and SNBT.
 +
* [https://sourceforge.net/projects/nbtfsutils/ NBTFS] (C; MPL-2) Editing NBT as using a file system; supports Minecraft region files.
 +
* [https://github.com/Lenni0451/ImNbt ImNbt] (Java) ImGui based NBT Directory-tree interface that supports Minecraft Java and Bedrock Edition NBT. Features diffing, searching and exporting in different NBT formats.
 +
* [https://offroaders123.github.io/Dovetail/ Dovetail] (JavaScript) web-based NBT editor for Java and Bedrock Edition.
  
 
== Specification ==
 
== Specification ==
The NBT file format is extremely simple, and writing a library capable of reading/writing it is a simple affair. There are 11 datatypes supported by this format, and one type used to close compound tags. It is strongly advised to read this entire section or you may run into issues.
+
The NBT file format is extremely simple, and writing a library capable of reading/writing it is a simple affair. There are 13 datatypes supported by this format, one of which is used to close compound tags. It is strongly advised to read this entire section or you may run into issues.
 +
 
 +
 
 +
{{Warning|Heads up! Since 1.20.2 NBT sent over the network has a subtle but critical specification change, refer to the Network NBT section below for more information}}
  
  
Line 50: Line 157:
 
! Payload Size (Bytes)
 
! Payload Size (Bytes)
 
! Description
 
! Description
|-
+
|- id=Specification:end_tag
 
| 0  
 
| 0  
| TAG_End
+
| [[#Specification:end_tag|TAG_End]]
 
| 0   
 
| 0   
| Signifies the end of a TAG_Compound. It is only ever used inside a TAG_Compound, and is not named despite being in a TAG_Compound
+
| Signifies the end of a TAG_Compound. It is only ever used inside a TAG_Compound, a TAG_List that has it's type id set to TAG_Compound or as the type for a TAG_List if the length is 0 or negative, and is not named even when in a TAG_Compound
|-
+
|- id=Specification:byte_tag
 
| 1
 
| 1
| TAG_Byte
+
| [[#Specification:byte_tag|TAG_Byte]]
 
| 1
 
| 1
 
| A single signed byte
 
| A single signed byte
|-
+
|- id=Specification:short_tag
 
| 2
 
| 2
| TAG_Short
+
| [[#Specification:short_tag|TAG_Short]]
 
| 2
 
| 2
 
| A single signed, big endian 16 bit integer
 
| A single signed, big endian 16 bit integer
|-
+
|- id=Specification:int_tag
 
| 3
 
| 3
| TAG_Int
+
| [[#Specification:int_tag|TAG_Int]]
 
| 4
 
| 4
 
| A single signed, big endian 32 bit integer
 
| A single signed, big endian 32 bit integer
|-
+
|- id=Specification:long_tag
 
| 4
 
| 4
| TAG_Long
+
| [[#Specification:long_tag|TAG_Long]]
 
| 8
 
| 8
 
| A single signed, big endian 64 bit integer
 
| A single signed, big endian 64 bit integer
|-
+
|- id=Specification:float_tag
 
| 5
 
| 5
| TAG_Float
+
| [[#Specification:float_tag|TAG_Float]]
 
| 4
 
| 4
| A single, big endian [[wikipedia:IEEE 754-2008|IEEE-754]] single-precision floating point number
+
| A single, big endian [[wikipedia:IEEE 754-2008|IEEE-754]] single-precision floating point number ([[wikipedia:NaN|NaN]] possible)
|-
+
|- id=Specification:double_tag
 
| 6
 
| 6
| TAG_Double
+
| [[#Specification:double_tag|TAG_Double]]
 
| 8
 
| 8
| A single, big endian [[wikipedia:IEEE 754-2008|IEEE-754]] double-precision floating point number
+
| A single, big endian [[wikipedia:IEEE 754-2008|IEEE-754]] double-precision floating point number ([[wikipedia:NaN|NaN]] possible)
|-
+
|- id=Specification:byte_array_tag
 
| 7
 
| 7
| TAG_Byte_Array
+
| [[#Specification:byte_array_tag|TAG_Byte_Array]]
 
| ...
 
| ...
 
| A length-prefixed array of '''signed''' bytes. The prefix is a '''signed''' integer (thus 4 bytes)
 
| A length-prefixed array of '''signed''' bytes. The prefix is a '''signed''' integer (thus 4 bytes)
|-
+
|- id=Specification:string_tag
 
| 8
 
| 8
| TAG_String
+
| [[#Specification:string_tag|TAG_String]]
 
| ...
 
| ...
| A length-prefixed [[wikipedia:UTF-8|UTF-8]] string. The prefix is an '''unsigned''' short (thus 2 bytes) signifying the length of the string in bytes
+
| A length-prefixed [https://docs.oracle.com/javase/8/docs/api/java/io/DataInput.html#modified-utf-8 modified UTF-8] string. The prefix is an '''unsigned''' short (thus 2 bytes) signifying the length of the string in bytes
|-
+
|- id=Specification:list_tag
 
| 9
 
| 9
| TAG_List
+
| [[#Specification:list_tag|TAG_List]]
 
| ...
 
| ...
| A list of '''nameless''' tags, all of the same type. The list is prefixed with the <code>Type ID</code> of the items it contains (thus 1 byte), and the length of the list as a '''signed''' integer (a further 4 bytes)
+
| A list of '''nameless''' tags, all of the same type. The list is prefixed with the <code>Type ID</code> of the items it contains (thus 1 byte), and the length of the list as a '''signed''' integer (a further 4 bytes).  If the length of the list is 0 or negative, the type may be 0 (TAG_End) but otherwise it must be any other type.  (The notchian implementation uses TAG_End in that situation, but another reference implementation by Mojang uses 1 instead; parsers should accept any type if the length is <= 0).
|-
+
|- id=Specification:compound_tag
 
| 10
 
| 10
| TAG_Compound
+
| [[#Specification:compound_tag|TAG_Compound]]
 
| ...
 
| ...
| Effectively a list of a '''named''' tags. Order is not guaranteed.
+
| Effectively a list of '''named''' tags. Order is not guaranteed.
|-
+
|- id=Specification:int_array_tag
 
| 11
 
| 11
| TAG_Int_Array
+
| [[#Specification:int_array_tag|TAG_Int_Array]]
 
| ...
 
| ...
 
| A length-prefixed array of '''signed''' integers. The prefix is a '''signed''' integer (thus 4 bytes) and indicates the number of 4 byte integers.
 
| A length-prefixed array of '''signed''' integers. The prefix is a '''signed''' integer (thus 4 bytes) and indicates the number of 4 byte integers.
 +
|- id=Specification:long_array_tag
 +
| 12
 +
| [[#Specification:long_array_tag|TAG_Long_Array]]
 +
| ...
 +
| A length-prefixed array of '''signed''' longs. The prefix is a '''signed''' integer (thus 4 bytes) and indicates the number of 8 byte longs.
 
|}
 
|}
  
 
There are a couple of simple things to remember:
 
There are a couple of simple things to remember:
* The datatypes representing numbers are in big-endian in the PC version, but Pocket Version is in little-endian. Unless you're using Java, you will most likely have to swap it to little-endian. See [[wikipedia:Endianness|the Wikipedia article on Endianness]].
+
* The datatypes representing numbers are in big-endian in Java edition, but Bedrock edition changes things up a bit. See the below section on Bedrock edition
* Every NBT file will '''always''' implicitly be inside a tag compound, and also begin with a TAG_Compound
+
* Every NBT file will '''always''' implicitly be inside a tag compound, and also begin with a TAG_Compound (except in Bedrock edition, see below)
* The structure of a NBT file is defined by the TAG_List and TAG_Compound types, as such a tag itself will only contain the payload, but depending on what the tag is contained within may contain additional headers. I.e. if it's inside a Compound, then each tag will begin with the TAG_id, and then a string (the tag's name), and finally the payload. While in a list it will be only the payload, as there is no name and the tag type is given in the beginning of the list.
+
* The structure of a NBT file is defined by the TAG_List and TAG_Compound types, as such a tag itself will only contain the payload, but depending on what the tag is contained within may contain additional headers. I.e. if it's inside a Compound, then each tag will begin with the TAG_id, and then a [https://docs.oracle.com/javase/8/docs/api/java/io/DataInput.html#modified-utf-8 modified UTF-8] string (the tag's name), and finally the payload. While in a list it will be only the payload, as there is no name and the tag type is given in the beginning of the list.
  
 
For example, here's the example layout of a <code>TAG_Short</code> on disk:
 
For example, here's the example layout of a <code>TAG_Short</code> on disk:
Line 141: Line 253:
  
 
If this <code>TAG_Short</code> had been in a <code>TAG_List</code>, it would have been nothing more than the payload, since the type is implied and tags within the first level of a list are nameless.
 
If this <code>TAG_Short</code> had been in a <code>TAG_List</code>, it would have been nothing more than the payload, since the type is implied and tags within the first level of a list are nameless.
 +
 +
=== Network NBT (Java Edition) ===
 +
Since 1.20.2 (Protocol 764) NBT sent over the network has been updated to exclude the name from the root <code>TAG_COMPOUND</code>, this essentially boils down to the following.
 +
{| class="wikitable"
 +
|-
 +
|
 +
! Type ID
 +
! Length of Name
 +
! Name
 +
! Payload
 +
|-
 +
! < 1.20.2 (Protocol 764)
 +
| <code>0x0a</code>
 +
| <code>0x00 0x00</code>
 +
| (Empty name)
 +
| <code>0x02 0x09</code>
 +
|-
 +
! >= 1.20.2 (Protocol 764)
 +
| <code>0x0a</code>
 +
| N/A
 +
| N/A
 +
| <code>0x02 0x09</code>
 +
|}
 +
Remember - This only applies to network NBT. Player data, world data, etc... will not be affected.
 +
 +
=== Bedrock edition ===
 +
Bedrock edition makes a couple of significant changes to the NBT format. First of all, first tag in an NBT file can sometimes be a TAG_List instead of a TAG_Compound. Additionally, NBT data is encoded in one of two different formats, a little-endian version intended for writing to disk, and a VarInt version intended for transport over the network.
 +
 +
==== Little-endian ====
 +
Identical to the big-endian format used by Java edition, but all numbers are encoded in little-endian. This includes the 16-bit length prefix before tag names and TAG_String values, as well as TAG_Float and TAG_Double values.
 +
 +
==== VarInt ====
 +
This format is a bit more complex than the others. The differences from Java edition's big-endian format are as follows:
 +
* TAG_Short, TAG_Float and TAG_Double values are encoded as their little-endian counterparts
 +
* TAG_Int values and the length prefixes for TAG_List, TAG_Byte_Array, TAG_Int_Array and TAG_Long_Array are encoded as [https://developers.google.com/protocol-buffers/docs/encoding#varints VarInts with ZigZag encoding]
 +
* TAG_Long values are encoded as [https://developers.google.com/protocol-buffers/docs/encoding#varints VarLongs with ZigZag encoding]
 +
* All strings (Tag names and TAG_String values) are length-prefixed with a normal [https://wiki.vg/Protocol#VarInt_and_VarLong VarInt]
  
 
=== Examples ===
 
=== Examples ===
Line 186: Line 335:
 
| <code>08</code>
 
| <code>08</code>
 
| <code>00 04</code>
 
| <code>00 04</code>
| <code>6e 6a 6d 65</code>
+
| <code>6e 61 6d 65</code>
 
| <code>00 09</code>
 
| <code>00 09</code>
 
| <code>42 61 6e 61 6e 72 61 6d 61</code>
 
| <code>42 61 6e 61 6e 72 61 6d 61</code>
Line 212: Line 361:
 
     TAG_Int('intTest'): 2147483647
 
     TAG_Int('intTest'): 2147483647
 
     TAG_Byte('byteTest'): 127
 
     TAG_Byte('byteTest'): 127
     TAG_String('stringTest'): 'HELLO WORLD THIS IS A TEST STRING \xc3\x85\xc3\x84\xc3\x96!'
+
     TAG_String('stringTest'): 'HELLO WORLD THIS IS A TEST STRING \xc5\xc4\xd6!'
 
     TAG_List('listTest (long)'): 5 entries
 
     TAG_List('listTest (long)'): 5 entries
 
     {
 
     {
Line 251: Line 400:
 
       TAG_Compound(None): 3 entries
 
       TAG_Compound(None): 3 entries
 
       {
 
       {
         TAG_Byte('hideAddress'): 0 (Don't hide the IP address)
+
         TAG_Byte('acceptTextures'): 1 (Automatically accept resourcepacks from this server)
 +
        TAG_String('ip'): '199.167.132.229:25620'
 
         TAG_String('name'): 'Dainz1 - Creative'
 
         TAG_String('name'): 'Dainz1 - Creative'
        TAG_String('ip'): '199.167.132.229:25620'
 
 
          
 
          
 
       }
 
       }
 
       TAG_Compound(None): 3 entries
 
       TAG_Compound(None): 3 entries
 
       {
 
       {
         TAG_Byte('hideAddress'): 1
+
         TAG_String('icon'): 'iVBORw0KGgoAAAANUhEUgAAAEAAAABACA...' (The base64-encoded server icon. Trimmed here for the example's sake)
 +
        TAG_String('ip'): '76.127.122.65:25565'
 
         TAG_String('name'): 'minstarmin4'
 
         TAG_String('name'): 'minstarmin4'
        TAG_String('ip'): '76.127.122.65:25565'
 
 
          
 
          
 
       }
 
       }
Line 437: Line 586:
 
* [https://raw.github.com/Dav1dde/nbd/master/test/hello_world.nbt test.nbt/hello_world.nbt] (uncompressed),
 
* [https://raw.github.com/Dav1dde/nbd/master/test/hello_world.nbt test.nbt/hello_world.nbt] (uncompressed),
 
* [https://raw.github.com/Dav1dde/nbd/master/test/bigtest.nbt bigtest.nbt] (gzip compressed)
 
* [https://raw.github.com/Dav1dde/nbd/master/test/bigtest.nbt bigtest.nbt] (gzip compressed)
 +
* [https://github.com/VADemon/nbd/raw/5de7a3f37569e1ffee11afbc017ae08e2c24523e/test/Player-nan-value.dat NaN-value-double.dat] (compressed, origin version unknown)
 +
* [https://web.archive.org/web/20110723210920/http://www.minecraft.net/docs/NBT.txt NBT.txt] (original NBT specification)
 +
 +
[[Category:Protocol_Details]]
 +
[[Category:Minecraft_Modern]]
 +
[[Category:File_Formats]]

Latest revision as of 09:49, 26 October 2024

The Named Binary Tag (NBT) file format is an extremely simple and efficient structured binary format used by Minecraft for a variety of things. Due to this, several third-party utilities now also utilize the format. You may find example files at the bottom of this article.

Mojang has released a reference implementation along with their Anvil conversion tool, available from this archived page

Current Uses

The NBT format is currently used in several places, chiefly:

  • In the Protocol as part of Slot Data
  • Multiplayer saved server list (servers.dat).
  • Player data (both single player and multiplayer, one file per player). This includes such things as inventory and location.
  • Saved worlds (both single player and multiplayer).
    • World index file (level.dat) that contains general information (spawn point, time of day, etc...)
    • Chunk data (see Region Files)

NBT files you can encounter as a developer will be stored in three different ways, mainly the second variation as per Notch's original specification.

  • Uncompressed,
  • gzip'd,
  • zlib'd (aka DEFLATE with a few bytes extra)

Libraries

There are many, many libraries for manipulating NBT, written in several languages, and often several per language. For example,

Name Description Language
cNBT As simple (and, consequently, as fast) as possible NBT file parser. C
libnbt A lightweight library to work with NBT, written in C. C
libnbt++ A free C++ library for Minecraft's file format Named Binary Tag (NBT). C++
cpp-nbt A C++23 header-only library for reading/writing Minecraft NBT data. C++
Venom An NBT decoder written in pure Elixir. Elixir
go-nbt This package implements the Named Binary Tag format of Minecraft. Go
BitBuf/nbt Flexible and intuitive library for reading and writing Minecraft's NBT format. Java
hephaistos This library is both a NBT library and a Minecraft Anvil format library. Java & Kotlin
Nedit A simple, lightweight NBT parsing library Java
NBT.js A JavaScript parser and serializer for NBT archives. JavaScript
NBTify Parser & writer NBT library intended for usage in a web browser. JavaScript, TypeScript
knbt Kotlin NBT library for kotlinx.serialization. Kotlin
KotlinNBT With a builder DSL and type-safety Kotlin
simpleNBT Streams, ByteBuffer, NIO, Endianness, Zlib, Gzip, Any Input/Output, Examples in Comments. Kotlin
TagForge A library made for the serialisation and deserialisation of MC NBT! Nim
php-nbt A full PHP implementation of Minecraft's Named Binary Tag (NBT) format. PHP
NBT This is mainly a Named Binary Tag parser & writer library. Python
HematiteNBT A full-featured Rust crate for Minecraft's NBT file format, including Serde support. Rust
Quartz NBT Provides support for encoding and decoding Minecraft's NBT format. Rust
SimdNBT A very fast NBT serializer and deserializer. Rust
fastnbt Fast serde serializer and deserializer for Minecraft's NBT and Anvil formats Rust
shen-nbt5 A Fast NBT parser/writer (warning: lot's of unsafe (from author)) Rust
CrabNBT Up-to-date Rust crate for easy and intuitive working with NBT data. Rust
CraftFlow-NBT Serde based NBT binary format implementation. Rust
ScalaNBT Scala library for NBT io with some Mojangson support. Scala

Unless you have specific goals or licence requirements, it is extremely recommended to go with one of the existing libraries.

Utilities

Almost every 3rd-party Minecraft application uses NBT on some level. There also exist several dedicated NBT editors, which will likely be useful to you if you are developing an NBT library of your own. These include:

  • NBTExplorer (C#) NBT Directory-tree interface that fully supports the Minecraft .mcr/.mca region files.
  • NEINedit (Obj-C), an OS X specific editor.
  • nbt2yaml (Python), provides command-line editing of NBT via the YAML format, as well as a fast and minimalist NBT parsing/rendering API.
  • nbted (Rust; CC0), provides command-line editing of NBT files via your $EDITOR
  • unbted (Rust; GPL-v3) Command-line interactive NBT editor
  • nbt2json (Golang; MIT) Command-line utility for NBT to JSON/YAML conversion and back. MCPE-NBT support. Can be used as library.
  • NBT Studio (C#) A visual editor similar to NBT-Explorer (A claimed, spiritual successor). Supports Minecraft Bedrock Edition files and SNBT.
  • NBTFS (C; MPL-2) Editing NBT as using a file system; supports Minecraft region files.
  • ImNbt (Java) ImGui based NBT Directory-tree interface that supports Minecraft Java and Bedrock Edition NBT. Features diffing, searching and exporting in different NBT formats.
  • Dovetail (JavaScript) web-based NBT editor for Java and Bedrock Edition.

Specification

The NBT file format is extremely simple, and writing a library capable of reading/writing it is a simple affair. There are 13 datatypes supported by this format, one of which is used to close compound tags. It is strongly advised to read this entire section or you may run into issues.


Warning.png Heads up! Since 1.20.2 NBT sent over the network has a subtle but critical specification change, refer to the Network NBT section below for more information


Type ID Type Name Payload Size (Bytes) Description
0 TAG_End 0 Signifies the end of a TAG_Compound. It is only ever used inside a TAG_Compound, a TAG_List that has it's type id set to TAG_Compound or as the type for a TAG_List if the length is 0 or negative, and is not named even when in a TAG_Compound
1 TAG_Byte 1 A single signed byte
2 TAG_Short 2 A single signed, big endian 16 bit integer
3 TAG_Int 4 A single signed, big endian 32 bit integer
4 TAG_Long 8 A single signed, big endian 64 bit integer
5 TAG_Float 4 A single, big endian IEEE-754 single-precision floating point number (NaN possible)
6 TAG_Double 8 A single, big endian IEEE-754 double-precision floating point number (NaN possible)
7 TAG_Byte_Array ... A length-prefixed array of signed bytes. The prefix is a signed integer (thus 4 bytes)
8 TAG_String ... A length-prefixed modified UTF-8 string. The prefix is an unsigned short (thus 2 bytes) signifying the length of the string in bytes
9 TAG_List ... A list of nameless tags, all of the same type. The list is prefixed with the Type ID of the items it contains (thus 1 byte), and the length of the list as a signed integer (a further 4 bytes). If the length of the list is 0 or negative, the type may be 0 (TAG_End) but otherwise it must be any other type. (The notchian implementation uses TAG_End in that situation, but another reference implementation by Mojang uses 1 instead; parsers should accept any type if the length is <= 0).
10 TAG_Compound ... Effectively a list of named tags. Order is not guaranteed.
11 TAG_Int_Array ... A length-prefixed array of signed integers. The prefix is a signed integer (thus 4 bytes) and indicates the number of 4 byte integers.
12 TAG_Long_Array ... A length-prefixed array of signed longs. The prefix is a signed integer (thus 4 bytes) and indicates the number of 8 byte longs.

There are a couple of simple things to remember:

  • The datatypes representing numbers are in big-endian in Java edition, but Bedrock edition changes things up a bit. See the below section on Bedrock edition
  • Every NBT file will always implicitly be inside a tag compound, and also begin with a TAG_Compound (except in Bedrock edition, see below)
  • The structure of a NBT file is defined by the TAG_List and TAG_Compound types, as such a tag itself will only contain the payload, but depending on what the tag is contained within may contain additional headers. I.e. if it's inside a Compound, then each tag will begin with the TAG_id, and then a modified UTF-8 string (the tag's name), and finally the payload. While in a list it will be only the payload, as there is no name and the tag type is given in the beginning of the list.

For example, here's the example layout of a TAG_Short on disk:

Type ID Length of Name Name Payload
Decoded 2 9 shortTest 32767
On Disk (in hex) 02 00 09 73 68 6F 72 74 54 65 73 74 7F FF

If this TAG_Short had been in a TAG_List, it would have been nothing more than the payload, since the type is implied and tags within the first level of a list are nameless.

Network NBT (Java Edition)

Since 1.20.2 (Protocol 764) NBT sent over the network has been updated to exclude the name from the root TAG_COMPOUND, this essentially boils down to the following.

Type ID Length of Name Name Payload
< 1.20.2 (Protocol 764) 0x0a 0x00 0x00 (Empty name) 0x02 0x09
>= 1.20.2 (Protocol 764) 0x0a N/A N/A 0x02 0x09

Remember - This only applies to network NBT. Player data, world data, etc... will not be affected.

Bedrock edition

Bedrock edition makes a couple of significant changes to the NBT format. First of all, first tag in an NBT file can sometimes be a TAG_List instead of a TAG_Compound. Additionally, NBT data is encoded in one of two different formats, a little-endian version intended for writing to disk, and a VarInt version intended for transport over the network.

Little-endian

Identical to the big-endian format used by Java edition, but all numbers are encoded in little-endian. This includes the 16-bit length prefix before tag names and TAG_String values, as well as TAG_Float and TAG_Double values.

VarInt

This format is a bit more complex than the others. The differences from Java edition's big-endian format are as follows:

  • TAG_Short, TAG_Float and TAG_Double values are encoded as their little-endian counterparts
  • TAG_Int values and the length prefixes for TAG_List, TAG_Byte_Array, TAG_Int_Array and TAG_Long_Array are encoded as VarInts with ZigZag encoding
  • TAG_Long values are encoded as VarLongs with ZigZag encoding
  • All strings (Tag names and TAG_String values) are length-prefixed with a normal VarInt

Examples

There are two defacto example files used for testing your implementation (test.nbt & bigtest.nbt), originally provided by Markus. The example output provided below was generated using PyNBT's debug-nbt tool.

test.nbt

This first example is an uncompressed "Hello World" NBT example. Should you parse it correctly, you will get a structure similar to the following:

  TAG_Compound('hello world'): 1 entry
  {
    TAG_String('name'): 'Bananrama'
  }

Here is the example explained:

(The entire thing is implicitly inside a compound) Type ID (first element in the implicit compound) Length of name of the root compound Name of the root compound Type ID of first element in root compound Length of name of first element in root Name of first element Length of string String Tag end (of root compound)
Decoded Compound 11 hello world String 4 name 9 Bananrama
On Disk (in hex) 0a 00 0b 68 65 6c 6c 6f 20 77 6f 72 6c 64 08 00 04 6e 61 6d 65 00 09 42 61 6e 61 6e 72 61 6d 61 00

bigtest.nbt

This second example is a gzip compressed test of every available tag. If your program can successfully parse this file, then you've done well. Note that the tags under TAG_List do not have a name, as mentioned above.

  TAG_Compound('Level'): 11 entries
  {
    TAG_Compound('nested compound test'): 2 entries
    {
      TAG_Compound('egg'): 2 entries
      {
        TAG_String('name'): 'Eggbert'
        TAG_Float('value'): 0.5
      }
      TAG_Compound('ham'): 2 entries
      {
        TAG_String('name'): 'Hampus'
        TAG_Float('value'): 0.75
      }
    }
    TAG_Int('intTest'): 2147483647
    TAG_Byte('byteTest'): 127
    TAG_String('stringTest'): 'HELLO WORLD THIS IS A TEST STRING \xc5\xc4\xd6!'
    TAG_List('listTest (long)'): 5 entries
    {
      TAG_Long(None): 11
      TAG_Long(None): 12
      TAG_Long(None): 13
      TAG_Long(None): 14
      TAG_Long(None): 15
    }
    TAG_Double('doubleTest'): 0.49312871321823148
    TAG_Float('floatTest'): 0.49823147058486938
    TAG_Long('longTest'): 9223372036854775807L
    TAG_List('listTest (compound)'): 2 entries
    {
      TAG_Compound(None): 2 entries
      {
        TAG_Long('created-on'): 1264099775885L
        TAG_String('name'): 'Compound tag #0'
      }
      TAG_Compound(None): 2 entries
      {
        TAG_Long('created-on'): 1264099775885L
        TAG_String('name'): 'Compound tag #1'
      }
    }
    TAG_Byte_Array('byteArrayTest (the first 1000 values of (n*n*255+n*7)%100, starting with n=0 (0, 62, 34, 16, 8, ...))'): [1000 bytes]
    TAG_Short('shortTest'): 32767
  }

servers.dat

The servers.dat file contains a list of multiplayer servers you've added to the game. To mix things up a bit, this file will always be uncompressed. Below is an example of the structure seen in servers.dat.

  TAG_Compound(''): 1 entry
  {
    TAG_List('servers'): 2 entries
    {
      TAG_Compound(None): 3 entries
      {
        TAG_Byte('acceptTextures'): 1 (Automatically accept resourcepacks from this server)
        TAG_String('ip'): '199.167.132.229:25620'
        TAG_String('name'): 'Dainz1 - Creative'
        
      }
      TAG_Compound(None): 3 entries
      {
        TAG_String('icon'): 'iVBORw0KGgoAAAANUhEUgAAAEAAAABACA...' (The base64-encoded server icon. Trimmed here for the example's sake)
        TAG_String('ip'): '76.127.122.65:25565'
        TAG_String('name'): 'minstarmin4'
        
      }
    }
  }

level.dat

This final example is of a single player level.dat, which is compressed using gzip. Notice the player's inventory and general world details such as spawn position, world name, and the game seed.

  TAG_Compound(''): 1 entry
  {
    TAG_Compound('Data'): 17 entries
    {
      TAG_Byte('raining'): 0
      TAG_Long('RandomSeed'): 3142388825013346304L
      TAG_Int('SpawnX'): 0
      TAG_Int('SpawnZ'): 0
      TAG_Long('LastPlayed'): 1323133681772L
      TAG_Int('GameType'): 1
      TAG_Int('SpawnY'): 63
      TAG_Byte('MapFeatures'): 1
      TAG_Compound('Player'): 24 entries
      {
        TAG_Int('XpTotal'): 0
        TAG_Compound('abilities'): 4 entries
        {
          TAG_Byte('instabuild'): 1
          TAG_Byte('flying'): 1
          TAG_Byte('mayfly'): 1
          TAG_Byte('invulnerable'): 1
        }
        TAG_Int('XpLevel'): 0
        TAG_Int('Score'): 0
        TAG_Short('Health'): 20
        TAG_List('Inventory'): 13 entries
        {
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 1
            TAG_Byte('Slot'): 0
            TAG_Short('id'): 24
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 1
            TAG_Byte('Slot'): 1
            TAG_Short('id'): 25
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 1
            TAG_Byte('Slot'): 2
            TAG_Short('id'): 326
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 1
            TAG_Byte('Slot'): 3
            TAG_Short('id'): 29
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 10
            TAG_Byte('Slot'): 4
            TAG_Short('id'): 69
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 3
            TAG_Byte('Slot'): 5
            TAG_Short('id'): 33
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 43
            TAG_Byte('Slot'): 6
            TAG_Short('id'): 356
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 64
            TAG_Byte('Slot'): 7
            TAG_Short('id'): 331
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 20
            TAG_Byte('Slot'): 8
            TAG_Short('id'): 76
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 64
            TAG_Byte('Slot'): 9
            TAG_Short('id'): 331
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 1
            TAG_Byte('Slot'): 10
            TAG_Short('id'): 323
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 16
            TAG_Byte('Slot'): 11
            TAG_Short('id'): 331
            TAG_Short('Damage'): 0
          }
          TAG_Compound(None): 4 entries
          {
            TAG_Byte('Count'): 1
            TAG_Byte('Slot'): 12
            TAG_Short('id'): 110
            TAG_Short('Damage'): 0
          }
        }
        TAG_Short('HurtTime'): 0
        TAG_Short('Fire'): -20
        TAG_Float('foodExhaustionLevel'): 0.0
        TAG_Float('foodSaturationLevel'): 5.0
        TAG_Int('foodTickTimer'): 0
        TAG_Short('SleepTimer'): 0
        TAG_Short('DeathTime'): 0
        TAG_List('Rotation'): 2 entries
        {
          TAG_Float(None): 1151.9342041015625
          TAG_Float(None): 32.249679565429688
        }
        TAG_Float('XpP'): 0.0
        TAG_Float('FallDistance'): 0.0
        TAG_Short('Air'): 300
        TAG_List('Motion'): 3 entries
        {
          TAG_Double(None): -2.9778325794951344e-11
          TAG_Double(None): -0.078400001525878907
          TAG_Double(None): 1.1763942772801152e-11
        }
        TAG_Int('Dimension'): 0
        TAG_Byte('OnGround'): 1
        TAG_List('Pos'): 3 entries
        {
          TAG_Double(None): 256.87499499518492
          TAG_Double(None): 112.62000000476837
          TAG_Double(None): -34.578128612797634
        }
        TAG_Byte('Sleeping'): 0
        TAG_Short('AttackTime'): 0
        TAG_Int('foodLevel'): 20
      }
      TAG_Int('thunderTime'): 2724
      TAG_Int('version'): 19132
      TAG_Int('rainTime'): 5476
      TAG_Long('Time'): 128763
      TAG_Byte('thundering'): 1
      TAG_Byte('hardcore'): 0
      TAG_Long('SizeOnDisk'): 0
      TAG_String('LevelName'): 'Sandstone Test World'
    }
  }

Download