tbl_parquet: Open a Parquet dataset as a lazy dplyr tbl

In quak: Query 'Azure Data Lake Storage Gen2' with 'DuckDB'

Open a Parquet dataset as a lazy dplyr tbl

Description

Validates the URL, loads the azure extension, then returns a lazy dplyr::tbl() over the dataset. Use az_conn() first if the connection needs Azure extensions, settings, or secrets.

Usage

tbl_parquet(conn, url, name = NULL, hive_partitioning = FALSE, replace = TRUE)

Arguments

conn	A DuckDB connection.
url	Character scalar. Azure Blob URL. Supports glob patterns for multi-file datasets (e.g. "abfss://[e-mail]/data/*.parquet").
name	Optional character scalar. Name to register the view under in DuckDB. When NULL (default) the dataset is scanned directly.
hive_partitioning	Logical. Enable Hive partition inference from the directory structure. Default FALSE.
replace	Logical. Replace an existing view of the same name. Default TRUE. Ignored when name = NULL.

Details

When name is NULL the dataset is queried directly via read_parquet() with no persistent object registered on the connection. When name is supplied the dataset is first registered as a VIEW via load_parquet(), then referenced by name. Glob patterns (e.g. "*.parquet") are supported in url for multi-file datasets.

Value

A dplyr::tbl() backed by the Parquet dataset.

Examples

## Not run: 
# Requires a live Azure account, credentials, and network access.
conn <- az_conn()
tbl_parquet(conn, "abfss://container@account/data/*.parquet") |>
  dplyr::collect()

## End(Not run)

quak documentation built on June 9, 2026, 5:09 p.m.