Query Apache Hudi COW and MOR tables directly in DuckDB — full scans, time-travel (as_of), incremental windows, and table metadata — powered by the Rust hudi-rs kernel.
Installing and Loading
INSTALL hudi FROM community;
LOAD hudi;
Example
-- Read a Hudi COW table
LOAD hudi;
SELECT * FROM hudi_scan('/path/to/hudi_table')
WHERE dt = '2026-01-01';
-- Time-travel to a specific instant
SELECT * FROM hudi_scan('/path/to/hudi_table', as_of = '20260101000000000');
-- Incremental read of new records since the last checkpoint
SELECT * FROM hudi_scan('/path/to/hudi_table', incremental = true);
About hudi
duckdb-hudi
DuckDB extension that reads Apache Hudi tables (COW and MOR) using the hudi-rs Rust kernel. Source: https://github.com/alitrack/duckdb-hudi
Install
INSTALL hudi FROM community;
LOAD hudi;
Features
- Full scan of COW and MOR Hudi tables
- Time-travel via
as_of(query a table at a specific instant) - Incremental reads (new/changed records since a checkpoint)
- Table schema + metadata queries
- Parallel scanning
Build note
The extension is built from the DuckDB C-API Rust template and its hudi
dependency pulls in the Lance crate, which runs prost-build at compile
time. Building therefore requires protoc to be available (this extension
declares requires_toolchains: rust; the protobuf compiler is the one
toolchain not yet in the default community build environment — see the PR
description for the resolution).
Added Functions
| function_name | function_type | description | comment | examples |
|---|---|---|---|---|
| hudi_files | scalar | NULL | NULL | |
| hudi_info | scalar | NULL | NULL | |
| read_hudi | table | NULL | NULL |
Overloaded Functions
This extension does not add any function overloads.
Added Types
This extension does not add any types.
Added Settings
This extension does not add any settings.