Enable in-process queued scalar-UDF execution — rducks_enable

Switches a Rducks-enabled DuckDB connection to an inproc_concurrent execution plan for subsequent scalar-UDF registrations and updates the native runtime backend. This backend preserves R's thread discipline: DuckDB worker-side scalar-UDF callbacks submit chunk requests to an extension-owned queue, and the recorded main R thread drains the queue and performs all R API work. This is a same-process scheduling mode, not a performance promise; R function calls are still serialized on the main R thread.

Usage

rducks_enable_inproc(con, threads = NULL, external_threads = NULL)

Arguments

con: A duckdb_connection already enabled with rducks_enable().
threads: Optional positive integer to set with PRAGMA threads before enabling the in-process backend. Use NULL to leave unchanged.
external_threads: Optional positive integer to set with SET external_threads before enabling the in-process backend. Use NULL to leave unchanged. For actual DuckDB worker concurrency, keep this smaller than threads (for example threads = 4, external_threads = 1).

Value

con, invisibly.

Details

This is a compatibility helper for the arrow_r/arrow_c in-process queue. New code can call rducks_set_execution_plan() directly with rducks_execution_plan("arrow_r", "inproc_concurrent") or rducks_execution_plan("arrow_c", "inproc_concurrent"). Select the plan before registering scalar UDFs whose reported execution plan should be the queued in-process path.

Examples

# \donttest{
db <- duckdb::dbConnect(duckdb::duckdb(config = list(allow_unsigned_extensions = "true")))
rducks_enable(db)
rducks_enable_inproc(db)
rducks_release(db)
DBI::dbDisconnect(db)
# }