vigicaen (version 0.15.5)

dt_parquet: Read parquet and convert to data.table

Description

Load data IN- our OUT- of memory. File extension can be omitted.

Usage

dt_parquet(path_base, name = NULL, ext = ".parquet", in_memory = TRUE)

Value

A data.table if in_memory is set to TRUE, a parquet Table if in_memory is set to FALSE.

Arguments

path_base

A character string, providing the path to read from.

name

Optional. A character string. The file name (if absent from path_base).

ext

Optional. A character string. The file extension.

in_memory

Logical, should data be loaded in memory?

Details

Output is a data.table. For meddra and whodrug tables, it is still a good option to load data in-memory. This function is wrapping arrow::read_parquet(), dplyr::collect() and data.table::as.data.table() altogether. If you want to load OUT of memory, set arg in_memory to FALSE. Be careful that doing so will change the function output format. For this latter case, the output is not a data.table, so there is no practical benefit as compared to using arrow::read_parquet() directly, with as_data_frame = FALSE.

See Also

tb_vigibase(), tb_who(), tb_meddra()

Examples

# Say you have a data.frame stored in a parquet format, such as this one
demo <-
  data.table::data.table(
    UMCReportId = c(1, 2, 3, 4),
    AgeGroup = c(1, 7, 7, 8)
  ) |>
  arrow::as_arrow_table()

tmp_folder <- paste0(tempdir(), "/dtparquetex")
dir.create(tmp_folder)
path_data <- paste0(tmp_folder, "/")

arrow::write_parquet(demo,
                     sink = paste0(path_data, "demo.parquet")
)

# Now you have a new session without demo
rm(demo)

# You may import the file directly to data.table format with dt_parquet
demo <-
  dt_parquet(path_data, "demo")

# Clean up (required for CRAN checks)
unlink(tmp_folder, recursive = TRUE)