roccodev / bdat-rs

A library to read and write BDAT tables from all Xenoblade games

bdat-rs

A library to read and write MONOLITHSOFT's proprietary BDAT format, used for data tables in all Xenoblade games.

Usage

Refer to the documentation for detailed info on how to use the library.

You may also find other projects that use the library useful:

• This repository includes a command-line tool to convert BDAT tables to various formats.
• Recordkeeper is a save editor for Xenoblade 3 and Future Redeemed that reads game data from BDAT tables.

If you have a project that uses the library, feel free to add it to the list by submitting a PR.

License

The bdat-rs library is dual-licensed under both Apache-2.0 and MIT.
The bdat-toolset executable is licensed under the GPLv3.

Credits

• vaxherd for xb3tool and research on the Xenoblade 3 BDAT format

bdat-rs's Issues

--hashes <HASHES>

What would be the file format for hashes? I tried a CSV with hashes, and it didn't work. I also didn't see any examples in the code.

XC3 Hashes

declarative parsing and writing

I've been experimenting with making the reading and writing code for modern bdat files more self documenting using binrw. I've linked a gist for code with binrw 0.13.3 that can read and rebuild all bdat files in Xenoblade 3 1:1 with the originals. This code doesn't include converting to and from the ModernTable type since it looks like you're reworking that on a separate branch. It should also be doable to store enum values instead of bytes for each row and still rebuild the data correctly.
https://gist.github.com/ScanMountGoat/a406b36a8d103eed035663ec42c3b9f5

The implementation is a lot shorter, but the main advantage in my opinion is making the code easier to reason about. The basic process for reading would be to parse the bdat data into an owned Bdat. Writing this back to the file without any changes should produce identical bytes. Converting the Bdat to a friendlier representation like ModernTable can be done with from_ and to_ conversion functions. Splitting the read/write logic and conversion logic also means the conversions can be tested by comparing the Bdat structs instead of diffing binary files.

If we want the types to be an implementation detail, we could just have functions that take a reader or bytes and return Vec<ModernTable, Vec<LegacyTable, or an enum with both. This would avoid all the current complexity on the user side of dealing with opaque readers or slices that will probably get converted immediately to tables anyway. The bdat files are tiny, so the performance cost of parsing and converting all tables is low.

XCX (big endian) BDATs

Borrowed tables

Table structures that borrow strings from their buffer, instead of cloning them.

failed to build row hash table

I dont know why, but I keep getting this error and as a result everything gets extracted except for system.bdat

"thread '' panicked at src\table\modern.rs:328:35:
failed to build row hash table: duplicate key Hash(2344178646)"

Screenshot attached.

automated release builds with github actions

This will make it easier for people that don't want to compile the program from source. It looks like it should just be a matter of copying some of ci.yml to a new file with a trigger for releases.

toolset not in workspace and not runnable from repo root

I would like to be able to run the CLI tool with cargo run -p bdat-toolset and check all code with cargo check. I can always cd into the toolset directory, but that creates multiple target directories.

XC1/DE/2 BDATs

Currently, the library only supports BDATs from Xenoblade 3.

Toolset: Inline schema

The ability to pack tables on-the-fly without a schema file, with a syntax like:

bdat-toolset pack --table Table1=Table1.json --table Table2=Table2.json -f json -o file.bdat

CSV -> BDAT

XC1 JP BDATs use Shift-JIS encoding

• Verify and implement