Overall

The world format has become a little more complicated over the years. It is now broken into chunks.

Type Formats

The types used in the world file are below, all are stored little-endian.

Framework

The world file has an overall section structure. The very first Int32 in the file is the map version. This will be used to determine which fields and sections to load and which to skip.

If the version is at least 135, a 7-byte "relogic" string follows the version number. This helps identify the world file. Following that is a 1-byte filetype. World files have type 2. Finally there are 12 bytes of revision and favorite data that you can ignore.

If the version is at least 88, the next thing the file is the number of sections as an Int16. This is followed by a list of Int32s (one for each section) which are offsets from the beginning of the file to each section. Finally, there is an Int16 which represents the number of tiles in the tile mask, and that many bits (rounded up to the nearest byte) following which represent whether or not that tile has UV coordinates or not.

Header

The first section in the file is the Header. The header contains general information about the world, including what state the world is in, what bosses have been killed, etc.

Some fields are only present in certain versions of the map. I've included the minimum map version for each field. If the current version is less than that, you should act as if that field doesn't exist.

Tile Data

The second section is the Tile Data. The tile data is a little more complicated. Some fields are only present if previous booleans are true.

The tiles are stored column-first. So the first tile is at 0,0, the second tile is at 1,0. Also, in later versions, there is RLE compression, which runs vertically for the current column only.

Finally, some tile types are marked "Important", and contain texture coordinates, other tiles have their texture coordinates calculated at runtime. This is determined by the bit mask read after the section offsets.

If RLE is present and greater than 0, you should copy the current tile and all of its information RLE times down. RLE will never be greater than the number of tiles you have left in the current column.

Chests

Section 3 contains all the chests in the world along with their contents.

Signs

Section 4 is a list of all the signs in the world.

NPCs

Section 5 is a list of all the NPCs in the world. The way NPCs are stored is each NPC is prefixed with a boolean, keep reading NPCs until that boolean is false.

If version is at least 140, there's a second list, stored in the same fashion, of homeless NPCs.

Entities

Section 6 contains a list of Entities if the version is at least 122.

Each entity has a different data layout:

For Entity Type 0: Training Dummy

For Entity Type 1: Item Frame

For Entity Type 2: Logic Sensor

Pressure Plates

Section 7 (if version is at least 170) contains the list of weighted pressure plates in the world.

Town Manager

Section 8 (if version is at least 189) contains the town manager. This contains the homes of each NPC, even if the NPC is currently dead.

Bestiary

Section 9 (if version is at least 210) contains the bestiary.

Creative Powers

Section 10 (if version at least 220), contains the list of creative powers active. This is the stuff found in the journey mode menu. Is time stopped, is biome spreading disabled, etc.

I haven't bothered figuring this format out.