Digest every Parquet table in a release directory

One row per table, where a "table" is either a {name}.parquet file or a {name}/ directory of partitions. This is the per-release manifest the sync compares — the Parquet counterpart of the score-COG registry.

Usage

parquet_manifest(dir, con = NULL)

Arguments

dir

directory holding the release's Parquet

con

optional DuckDB connection to reuse

Value

a data frame, one row per table, sorted by name