Compatibility

Backward Compatibility

Backward compatibility refers to the ability of a newer DuckDB version to read storage files created by an older DuckDB version. Version 0.10 is the first release of DuckDB that supports backward compatibility in the storage format. DuckDB v0.10 can read and operate on files created by the previous DuckDB version – DuckDB v0.9.

For future DuckDB versions, our goal is to ensure that any DuckDB version released after can read files created by previous versions, starting from this release. We want to ensure that the file format is fully backward compatible. This allows you to keep data stored in DuckDB files around and guarantees that you will be able to read the files without having to worry about which version the file was written with or having to convert files between versions.

Forward Compatibility

Forward compatibility refers to the ability of an older DuckDB version to read storage files produced by a newer DuckDB version. DuckDB v0.9 is partially forward compatible with DuckDB v0.10. Certain files created by DuckDB v0.10 can be read by DuckDB v0.9.

Forward compatibility is provided on a best effort basis. While stability of the storage format is important – there are still many improvements and innovations that we want to make to the storage format in the future. As such, forward compatibility may be (partially) broken on occasion.

How to Move Between Storage Formats

When you update DuckDB and open an old database file, you might encounter an error message about incompatible storage formats, pointing to this page. To move your database(s) to newer format you only need the older and the newer DuckDB executable.

Open your database file with the older DuckDB and run the SQL statement EXPORT DATABASE 'tmp'. This allows you to save the whole state of the current database in use inside folder tmp. The content of the tmp folder will be overridden, so choose an empty/non yet existing location. Then, start the newer DuckDB and execute IMPORT DATABASE 'tmp' (pointing to the previously populated folder) to load the database, which can be then saved to the file you pointed DuckDB to.

A Bash script to achieve this (to be adapted with the file names and executable locations) is the following

/older/duckdb mydata.old.db -c "EXPORT DATABASE 'tmp'"
/newer/duckdb mydata.new.db -c "IMPORT DATABASE 'tmp'"

After this, mydata.db will remain in the old format, mydata.new.db will contain the same data but in a format accessible by the more recent DuckDB version, and the folder tmp will hold the same data in a universal format as different files.

Check EXPORT documentation for more details on the syntax.

Explicit Storage Versions

DuckDB v1.2.0 introduced the STORAGE_VERSION option, which allows explicitly specifying the storage version. Using this, you can opt-in to newer forwards-incompatible features:

ATTACH 'file.db' (STORAGE_VERSION 'v1.2.0');

This setting specifies the minimum DuckDB version that should be able to read the database file. When database files are written with this option, the resulting files cannot be opened by older DuckDB released versions than the specified version. They can be read by the specified version and all newer versions of DuckDB.

If you attach to DuckDB databases, you can query the storage versions using the following command:

SELECT database_name, tags FROM duckdb_databases();

This shows the storage versions:

┌───────────────┬───────────────────────────────────┐
│ database_name │               tags                │
│    varchar    │       map(varchar, varchar)       │
├───────────────┼───────────────────────────────────┤
│ file1         │ {storage_version=v1.2.0}          │
│ file2         │ {storage_version=v1.0.0 - v1.1.3} │
│ ...           │ ...                               │
└───────────────┴───────────────────────────────────┘

This means that file2 can be opened by past DuckDB versions while file1 is compatible only with v1.2.0 (or future versions).

Converting Between Storage Versions

To convert from the new format to the old format for compatibility, use the following sequence in DuckDB v1.2.0+:

ATTACH 'file1.db';
ATTACH 'converted_file.db' (STORAGE_VERSION 'v1.0.0');
COPY FROM DATABASE file1 TO converted_file;

Storage Header

DuckDB files start with a uint64_t which contains a checksum for the main header, followed by four magic bytes (DUCK), followed by the storage version number in a uint64_t.

hexdump -n 20 -C mydata.db

00000000  01 d0 e2 63 9c 13 39 3e  44 55 43 4b 2b 00 00 00  |...c..9>DUCK+...|
00000010  00 00 00 00                                       |....|
00000014

A simple example of reading the storage version using Python is below.

import struct

pattern = struct.Struct('<8x4sQ')

with open('test/sql/storage_version/storage_version.db', 'rb') as fh:
    print(pattern.unpack(fh.read(pattern.size)))

Storage Version Table

For changes in each given release, check out the change log on GitHub. To see the commits that changed each storage version, see the commit log.

Compression

DuckDB uses lightweight compression. Note that compression is only applied to persistent databases and is not applied to in-memory instances.

Compression Algorithms

The compression algorithms supported by DuckDB include the following:

• Constant Encoding
• Run-Length Encoding (RLE)
• Bit Packing
• Frame of Reference (FOR)
• Dictionary Encoding
• Fast Static Symbol Table (FSST) – VLDB 2020 paper
• Adaptive Lossless Floating-Point Compression (ALP) – SIGMOD 2024 paper
• Chimp – VLDB 2022 paper
• Patas

Disk Usage

The disk usage of DuckDB's format depends on a number of factors, including the data type and the data distribution, the compression methods used, etc. As a rough approximation, loading 100 GB of uncompressed CSV files into a DuckDB database file will require 25 GB of disk space, while loading 100 GB of Parquet files will require 120 GB of disk space.

Row Groups

DuckDB's storage format stores the data in row groups, i.e., horizontal partitions of the data. This concept is equivalent to Parquet's row groups. Several features in DuckDB, including parallelism and compression are based on row groups.

Troubleshooting

Error Message When Opening an Incompatible Database File

When opening a database file that has been written by a different DuckDB version from the one you are using, the following error message may occur:

Error: unable to open database "...": Serialization Error: Failed to deserialize: ...

The message implies that the database file was created with a newer DuckDB version and uses features that are backward incompatible with the DuckDB version used to read the file.

There are two potential workarounds:

1. Update your DuckDB version to the latest stable version.
2. Open the database with the latest version of DuckDB, export it to a standard format (e.g., Parquet), then import it using to any version of DuckDB. See the EXPORT/IMPORT DATABASE statements for details.