dr_source_iceberg()
Uses a caller-owned DuckDB connection with the Iceberg extension already
loaded. Configure REST catalogs and credentials outside product definitions.
Sources use iceberg_scan() on a metadata file or table location. Pin a
snapshot ID to reproduce a read; retention remains the catalog's policy.
Usage
dr_source_iceberg(connection, path, snapshot = NULL)
dr_target_iceberg(connection, table, mode = c("create", "append"))Arguments
- connection
- Caller-owned DuckDB connection.
- path
- Iceberg metadata file or table location.
- snapshot
- Optional snapshot ID as a decimal string, never a double.
- table
- Fully qualified DBI::Id in an attached Iceberg catalog.
- mode
- Create a new table or append to an existing one.
Value
A source or target specification.
Details
Targets require DBI::Id(catalog=, schema=, table=) in an attached Iceberg
REST catalog. Create fails if the table exists; append requires an existing
table. The stored candidate is checked lazily inside a transaction
before commit. Catalogs must support staged writes and transactions; uncertain
failures are propagated, with no nontransactional fallback. No replace, branch/merge,
snapshot-retention or cross-table atomicity guarantee is made.
Each write stores a
This experimental adapter delegates to DuckDB's Iceberg extension. Writes are not DataRaft lake releases: no registry release, lineage edge or managed rollback is created. Use a managed lake target when release governance is required. Lazy validation avoids collecting the entire table by default; custom quality rules may still collect data.
dataraft.write-id Iceberg table property. If DuckDB
reports an error after commit, the adapter checks this marker and the row
count before accepting the already published write; uncertain outcomes fail.This experimental adapter delegates to DuckDB's Iceberg extension. Writes are not DataRaft lake releases: no registry release, lineage edge or managed rollback is created. Use a managed lake target when release governance is required. Lazy validation avoids collecting the entire table by default; custom quality rules may still collect data.
Documentation source snapshot · View source at 2f9432d ↗
Code examples are documented source examples; they are not executed in this website build.
Code examples are documented source examples; they are not executed in this website build.