Skip to Content
SourcesDuckLake

DuckLake

DuckLake

Scan a DuckLake — its catalog in Postgres, MySQL, SQLite or DuckDB, its data in object storage.

Category
Warehouse & Lakehouse
Source type
DUCKLAKE
Produces
table

DuckLake keeps a lakehouse’s metadata in an ordinary SQL database and its data as Parquet files in a bucket or folder. This source attaches the lake read-only and reads every table in it as it is now.

What you need to connect

The metadata catalog: a connection string such as postgres:dbname=lake host=db.example.com user=reader, a mysql: string, a sqlite: path, or a DuckDB file path. Its password goes in the masked password field rather than in the string.

The data files are found where the catalog recorded them. Add storage keys for S3, Azure or GCS (HMAC) when the scanning environment can’t already read that location, and a data path when the files have moved.

What Classifyre reads

One asset per table, across every schema in the lake or only the ones you list. Rows come from the lake’s current snapshot: deleted rows are absent and a renamed column reads as one column.

Shared behaviour · SQL databases

One asset per table or view, never one per row. The asset carries the table's structure — database, schema, table name, object type, its columns and their types, and a row-count estimate — and its content is a sample of real rows, formatted so a detector reads actual values rather than a schema dump.

How many rows, and which ones, is entirely up to the sampling strategy. Large tables are paged through by key rather than by OFFSET, so a scan that stops halfway can resume from where it left off instead of re-reading from the top.

Read-only throughout. The connector issues catalog queries and bounded SELECTs. Nothing is written back, and a read-only account is the right account to give it.

Metadata on every asset

Asset kind · table

FieldTypeAlways presentWhat it is
databasestringYesDatabase or catalog name
table_namestringYesTable name
table_typestringYesObject type (TABLE/VIEW)
schemastringNoSchema name
columnsobject[]NoColumns as {name, type} objects
row_countintegerNoEstimated number of rows
table_formatstringNoStorage format of the table (iceberg, delta, hudi, paimon, lance, native, external, view)
format_versionintegerNoTable format version (Delta reader version, Iceberg format version, Hudi table version)
table_versionstringNoCurrent version the table was read at (Delta version, Iceberg metadata file, Paimon snapshot, Hudi instant, Lance version)
snapshot_idstringNoCurrent snapshot id (Iceberg, Paimon)
num_filesintegerNoNumber of live data files
partition_columnsstring[]NoPartition column names
history_lengthintegerNoNumber of commits in the table history (Delta)
hudi_table_typestringNoHudi table type (COPY_ON_WRITE or MERGE_ON_READ)

Worth knowing

  • Read-only. The lake is attached read-only; nothing is written to the catalog or the data path.
  • No row estimate is shown: DuckLake’s own statistics still count rows that have since been deleted.

Configuration

Beyond the fields below, every source also has the settings shared by all of them: the sampling strategy, the augmentation notebook that enriches assets after extraction, the detectors to run, the scan schedule, and the compute limits for its scan jobs.

Required

Without these, the source will not save.

FieldTypeRequiredWhat it doesDefault
requiredobjectYes—no extra properties—
catalogstringYesDuckLake metadata catalog: postgres:dbname=lake host=db.example.com user=reader, mysql:..., sqlite:/path/meta.sqlite, or a DuckDB file path—

Secrets

Stored encrypted and never shown again after you save them. See Configuration & Fields.

FieldTypeRequiredWhat it doesDefault
maskedobjectYes—no extra properties—
azure_connection_stringstringNoAzure connection string for the data files—
catalog_passwordstringNoPassword for a Postgres or MySQL metadata catalog—
gcs_hmac_key_idstringNoGCS HMAC key for the data files—
gcs_hmac_secretstringNoGCS HMAC secret—
s3_access_key_idstringNoS3 access key for the data files—
s3_secret_access_keystringNoS3 secret key for the data files—
s3_session_tokenstringNoS3 session token—

Optional

Everything you can tune. Sensible defaults apply when you leave them alone.

FieldTypeRequiredWhat it doesDefault
optionalobjectNo—no extra properties—
connectionobjectNo—no extra properties—
connection.data_pathstringNoWhere the data files are, when it differs from what the catalog recorded—
connection.metadata_schemastringNoSchema of the metadata tables in a Postgres or MySQL catalog—
connection.s3_endpoint_urlstringNoS3 endpoint when it is not AWS (MinIO, R2, ...)—
connection.s3_regionstringNoS3 region—
scopeobjectNo—no extra properties—
scope.exclude_schemasarrayNoSkip these schemas—
scope.exclude_schemas[]stringNo——
scope.include_objectsarrayNoOnly these tables (schema.table)—
scope.include_objects[]stringNo——
scope.include_schemasarrayNoOnly these schemas—
scope.include_schemas[]stringNo——
scope.table_limitintegerNoMaximum number of tables to scan per runmin 1—
Last updated on