DuckDB Website and Documentation

The anofox_statistics extension provides a comprehensive suite of statistical regression and inference tools directly within SQL. It supports a wide range of models including:

• Linear Models: OLS, Ridge, Elastic Net, LARS/LassoLars, WLS, and recursive least squares.
• Robust Estimators: Huber, RANSAC, and Theil-Sen.
• Generalized Linear Models (GLMs): Poisson, Negative Binomial, Binomial, Tweedie, Gamma, and Logistic (including offset support).
• Advanced Modeling: Mixed-effects/hierarchical GLMs (with random slopes and crossed/nested factors), AFT survival regression, and time-series regression.
• Inference & Priors: Explicit priors with Laplace intervals and empirical-Bayes shrinkage.

All models include full diagnostics and inference capabilities.

The DuckDB Swift API provides a native interface for Swift platforms (Apple platforms and Linux) with the following capabilities:

• Enhanced SQL Dialect: Supports window functions, correlated subqueries, collations, and complex types (Swift arrays and structs).
• File Support: Native import/export for JSON, CSV, and Parquet.
• Strong Typing: Result sets can be cast directly to native Swift types.
• Concurrency: Core types conform to Sendable, making them safe for use in asynchronous and parallel processing workflows.

The datasketches extension utilizes the Apache DataSketches library to provide efficient approximate data analysis within DuckDB. It is designed for:

• Computing approximate distinct item counts.
• Estimating quantiles.
• Serializing sketches for storage or transfer.

The DuckDB C++ API is built around three primary abstractions: DuckDB (the database instance), Connection (a session to the instance), and QueryResult (the result of a query).

Warning: The C++ API is considered internal and is not guaranteed to be stable. For building production applications, it is recommended to use the C API instead.

To use the API, include the duckdb.hpp header file.

#include "duckdb.hpp"

The opendal extension integrates Apache OpenDAL into DuckDB as a virtual filesystem. It enables transparent read and write access to multiple storage services through a unified SQL interface.

Supported Storage Services

• fs:// / file:// — Local filesystem
• s3:// — Amazon S3 and S3-compatible stores (MinIO, R2, etc.)
• memory:// — In-memory storage (useful for testing)

Key Features

• Read & write Parquet, CSV, and JSON files through any supported service.
• Glob patterns work transparently (e.g., read_parquet('s3://bucket/**/*.parquet')).
• Scoped secrets with per-service configuration via DuckDB's secret manager.
• Configurable layers for retry, timeout, I/O concurrency, and caching.
• Coexistence with DuckDB's native httpfs extension via the opendal_override_native_filesystems setting.

DuckDB provides an Arrow Database Connectivity (ADBC) driver. ADBC is a C-style API that enables code portability between different database systems using Apache Arrow for data transfer. This allows for efficient, zero-copy data movement between DuckDB and your application.

Note: The DuckDB ADBC driver currently supports ADBC version 0.7.

The lerobot extension provides capabilities to:

• Read LeRobot v3 metadata and frame data.
• Create datasets using the COPY command.
• Decode RGB video frames.
• Select temporal observation windows using SQL.

Integration Note: TorchCodec integration is available through a separate optional Python package.

Licensing Note: The extension source is Apache-2.0. Native binaries statically link LGPL FFmpeg, libaom, dav1d, and zlib. GPL codecs are excluded from the default build.

DuckDB supports a wide variety of data formats and storage locations. You can import data from local files (CSV, Parquet, JSON, Excel) or directly query remote cloud storage (S3, GCS, HTTP(S), Cloudflare R2). You can also query external database systems like MySQL, PostgreSQL, and SQLite directly.

Key capabilities include:

• File Formats: CSV, Parquet, JSON, and Excel (via Spatial extension).
• Cloud Storage: Direct access to S3, S3 Express One, GCS, and Cloudflare R2.
• Database Integration: Direct querying of MySQL, PostgreSQL, and SQLite.
• Binary/Text: Direct reading of raw binary or text files.

The Delta Extension is currently considered experimental because both the extension and its dependency, delta-kernel-rs, are in early stages.

Supported Features

• All data types
• Filter and projection pushdown
• File skipping based on filter pushdown
• Deletion vectors
• Partitioned tables
• Fully parallel scanning

Supported Platforms

• linux_amd64
• linux_arm64
• osx_amd64
• osx_arm64

Planned Features

Future updates aim to add support for:

• Write support
• Column mapping
• Time travel
• Variant, RowIds
• Wasm support