data.library

!!! IN DEVELOPMENT !!!

Simple R library to input and output datasets.

• Data (data.table, tibble, data.frame) is outputted (output()) as individual RDS objects.
• Data is inputted (input()) as a combined (data.table, tibble, data.frame) object.
• Ability to specify a subset of variables to input.

Usage

# Install package.
remotes::install_github("KyleHaynes/data.library")
# or
devtools::install_github("KyleHaynes/data.library")

# Load required packages
library(data.library)

# Output mtcars.
output(mtcars, "./mtcars")

# Input mtcars.
data <- input("./mtcars")

# Input a subset of variables.
data <- input("./mtcars", vars_regex = "c|w")
# ℹ Importing the following variables: "carb", "cyl", "qsec", and "wt".
# ! Not importing the following variables: "am", "disp", "drat", "gear", "hp", "mpg", and "vs".
# ℹ Imported `carb`` in 0 minutes.
# ℹ Imported `cyl`` in 0 minutes.
# ℹ Imported `qsec`` in 0 minutes.
# ℹ Imported `wt`` in 0 minutes.
# ℹ Folder last modified: 2024-12-01 05:43:10.138605.
# ℹ Overall time taken: 0 minutes.

# You can also define global options.
options("data.library.path" = "./iris")

# Output iris.
output(iris)

# Input iris.
i <- input()

# Check equality.
all.equal(iris, i)
# [1] TRUE

Why not feather?

• Network benchmarks (~100MB/s) indicate reading RDS is faster.
• feather can't output complex data structures (e.g. lists).

(basic, basic, basic) benchmarks (GNAF-Core, 15.4m rows, 27 vars):

system.time({
    x <- read_feather("\\\\Server\\9tb\\New folder\\gnaf.feather", columns = c("ADDRESS_DETAIL_PID", "DATE_CREATED", "ADDRESS_LABEL"))
}, gcFirst = F)
# 32 seconds

system.time({
    d <- input("\\\\Server\\9tb\\New folder\\gnap", vars_vec = c("ADDRESS_DETAIL_PID", "DATE_CREATED", "ADDRESS_LABEL"))
}, gcFirst = F)
# 23 seconds

system.time({
    x <- read_feather("\\\\Server\\9tb\\New folder\\gnaf.feather")
}, gcFirst = F)
# 97 seconds

system.time({
    d <- input("\\\\Server\\9tb\\New folder\\gnap")
}, gcFirst = F)
# 90 seconds

To Do

• Lots
• Streamlined creation of metadata
• Better doco
• Vignettes