test_that("local csv files", { cars <- tempfile() write.csv(mtcars, cars) df <- open_dataset(cars, format = "csv") expect_true(inherits(df, "tbl_duckdb_connection")) unlink(cars) close_connection() }) test_that("duckdb_s3_config", { skip_if_offline() skip_on_cran() status <- duckdb_s3_config( s3_access_key_id = "YOUR_ACCESS_KEY_ID", s3_secret_access_key = "YOUR_SECRET_ACCESS_KEY", s3_endpoint = "YOUR_S3_ENDPOINT", s3_region = "YOUR_S3_REGION", s3_uploader_max_filesize = "800GB", s3_uploader_max_parts_per_file = 1000, s3_url_compatibility_mode = FALSE, s3_url_style = "vhost", s3_use_ssl = TRUE) expect_identical(status, 0) }) test_that("https", { skip_if_offline() skip_on_cran() base <- paste0("https://github.com/duckdb/duckdb/raw/main/", "data/parquet-testing/hive-partitioning/union_by_name/") f1 <- paste0(base, "x=1/f1.parquet") f2 <- paste0(base, "x=1/f2.parquet") f3 <- paste0(base, "x=2/f2.parquet") conn <- cached_connection() ds <- open_dataset( c(f1,f2,f3), conn = conn, unify_schemas = TRUE) expect_s3_class(ds, "tbl") df <- dplyr::collect(ds) expect_s3_class(df, "data.frame") close_connection(conn) }) test_that("close_connection", { skip_on_cran() close_connection() close_connection() expect_true(TRUE) }) test_that("s3", { skip_if_offline() skip_on_cran() close_connection() parquet <- "s3://gbif-open-data-us-east-1/occurrence/2023-06-01/occurrence.parquet" gbif <- open_dataset(parquet, anonymous = TRUE, s3_region="us-east-1") expect_s3_class(gbif, "tbl_dbi") expect_s3_class(gbif, "tbl") })