Interface DuckDbQueryCompilerConfigs

Mappings of table name in kysely to duckdb table expressions.

Duckdb can read external source(file, url or database) as table like: SELECT * FROM read_json_objects('path/to/file/*.json'). You can use raw duckdb table expression as table name, but it may be too long, preserving too many implementation details.

This mappings is used to replace table name string to duckdb table expression.

Keys can be plain table names or schema-qualified names (e.g., "schema.table").

• Plain table names match only when no schema is specified in the query.
• Schema-qualified keys match when using .withSchema("schema").selectFrom("table").
• When a schema is specified but not found in any mapping key, the table is resolved normally (useful for attached databases like Postgres via ATTACH).

Example

const dialect = new DuckDbDialect({
 database: db,
 tableMappings: {
   // Matches: db.selectFrom("person")
   person: 'read_json_object("s3://my-bucket/person.json")',
   // Matches: db.withSchema("archive").selectFrom("person")
   "archive.person": 'read_parquet("s3://my-bucket/archive/person.parquet")',
 }
});

const db = new Kysely<Database>({ dialect });

// Uses the "person" mapping (reads from JSON)
await db.selectFrom("person").selectAll().execute();

// Uses the "archive.person" mapping (reads from Parquet)
await db.withSchema("archive").selectFrom("person").selectAll().execute();

// No mapping for "neon.person", queries the attached database directly
await db.withSchema("neon").selectFrom("person").selectAll().execute();
Copy