Data in R Packages

Overview

R packages can include data in several forms: exported datasets for users, internal data for package functions, raw data files, and dynamic package state. This skill covers all data types, documentation requirements, and CRAN restrictions.

Four Types of Package Data

1. Exported Data (data/)

User-accessible datasets loaded with data() or direct reference.

Location: data/ directory Format: .rda or .RData files Access: data(dataset_name) or direct reference (if LazyData: true) Documentation: Required in R/data.R

2. Internal Data (R/sysdata.rda)

Data used by package functions, not accessible to users.

Location: R/sysdata.rda Format: Single .rda file containing multiple objects Access: Direct reference in package code Documentation: Not required (internal only)

3. Raw Data Files (inst/extdata/)

Non-R data files (CSV, JSON, images, etc.) for users to access.

Location: inst/extdata/ Format: Any file format Access: system.file("extdata", "file.ext", package = "pkg") Documentation: Optional, usually in vignettes/examples

4. Dynamic Package State (environments)

Runtime state stored in package environments.

Location: Environment created in R code Format: In-memory R objects Access: Getter/setter functions Documentation: Document the getter/setter functions

Exported Data (data/)

Creating Exported Data

# Prepare your data:
my_dataset <- data.frame(
  id = 1:100,
  value = rnorm(100),
  category = sample(LETTERS[1:3], 100, replace = TRUE)
)

# Save to data/:
usethis::use_data(my_dataset, overwrite = TRUE)

This creates data/my_dataset.rda.

Multiple Datasets

# Save multiple datasets:
dataset1 <- mtcars[1:10, ]
dataset2 <- iris[1:50, ]

usethis::use_data(dataset1, dataset2, overwrite = TRUE)
# Creates data/dataset1.rda and data/dataset2.rda

Compression Options

# Default compression (gzip):
usethis::use_data(my_dataset)

# Maximum compression (xz - slowest, smallest):
usethis::use_data(my_dataset, compress = "xz")

# Faster compression (bzip2):
usethis::use_data(my_dataset, compress = "bzip2")

# No compression (largest):
usethis::use_data(my_dataset, compress = FALSE)

CRAN recommendation: Use compress = "xz" for data >1MB.

LazyData

Add to DESCRIPTION to make data available without data() call:

LazyData: true

# Without LazyData:
library(mypackage)
data(my_dataset)  # Required
head(my_dataset)

# With LazyData:
library(mypackage)
head(my_dataset)  # Direct access

Note: LazyData loads datasets into namespace but keeps them on disk until accessed (lazy loading).

Internal Data (R/sysdata.rda)

Creating Internal Data

Internal data is for package functions only, not exported to users.

# Create internal lookup tables, constants, etc.:
internal_lookup <- list(
  codes = c(A = 1, B = 2, C = 3),
  thresholds = c(low = 0.05, high = 0.95)
)

internal_constants <- list(
  api_version = "v2",
  default_timeout = 30
)

# Save to R/sysdata.rda:
usethis::use_data(
  internal_lookup,
  internal_constants,
  internal = TRUE,
  overwrite = TRUE
)

All objects saved with internal = TRUE go into a single file: R/sysdata.rda

Using Internal Data

# In package functions, reference directly:
my_function <- function(code) {
  value <- internal_lookup$codes[code]
  # ... use value ...
}

# No need for pkg:::sysdata syntax
# Objects are available in package namespace

When to Use Internal Data

Good uses:

• Lookup tables
• Large constants
• Pre-computed values (avoid recomputation)
• Default configurations

Avoid:

• Data that changes (use environments instead)
• User-facing data (use data/ instead)
• Very large objects (consider lazy loading strategies)

Raw Data Files (inst/extdata/)

Adding Raw Data Files

# Create inst/extdata/ directory:
dir.create("inst/extdata", recursive = TRUE)

# Add files manually or:
usethis::use_directory("inst/extdata")

# Then copy files into inst/extdata/

Common file types:

• CSV, TSV, Excel files
• JSON, XML, YAML
• Images (PNG, JPEG)
• Shapefiles, GeoJSON
• Text files, logs
• Binary formats

Accessing Raw Data Files

# In package code:
get_example_file <- function(filename) {
  system.file("extdata", filename, package = "mypackage")
}

# Usage:
csv_path <- system.file("extdata", "example.csv", package = "mypackage")
data <- read.csv(csv_path)

# Or in exported function:
#' @examples
#' file <- system.file("extdata", "example.csv", package = "mypackage")
#' data <- read_my_data(file)
read_my_data <- function(file) {
  # ...
}

inst/ vs data/

inst/extdata/          # Raw files, any format
├── example.csv        # Access with system.file()
├── sample.json
└── image.png

data/                  # R objects only
├── dataset1.rda       # Access with data() or direct reference
└── dataset2.rda

Use inst/extdata/ when:

• Non-R formats (CSV, JSON, etc.)
• Files users need paths to
• Multiple related files
• Files for examples/vignettes

Use data/ when:

• R objects for analysis
• Data ready to use in R
• Common datasets for package functions

Documenting Data

Documenting Exported Data

Create R/data.R to document all datasets:

# R/data.R

#' World Health Organization TB data
#'
#' A subset of data from the World Health Organization Global Tuberculosis
#' Report, containing TB cases by country, year, age, and sex.
#'
#' @format A data frame with 7,240 rows and 60 columns:
#' \describe{
#'   \item{country}{Character. Country name}
#'   \item{iso2}{Character. 2-letter ISO country code}
#'   \item{iso3}{Character. 3-letter ISO country code}
#'   \item{year}{Integer. Year of observation (1995-2013)}
#'   \item{new_sp_m014}{Integer. New smear-positive cases in males aged 0-14}
#'   \item{new_sp_m1524}{Integer. New smear-positive cases in males aged 15-24}
#' }
#'
#' @source World Health Organization Global Tuberculosis Report
#'   \url{https://www.who.int/teams/global-tuberculosis-programme/data}
#'
#' @examples
#' head(who)
#' summary(who$year)
#' table(who$country)
"who"

Required tags:

• @format - describe structure and columns
• Title and description (always)

Recommended tags:

• @source - where data came from
• @examples - how to use the data

Data Documentation Templates

Data Frame

#' Customer transaction data
#'
#' Sample transaction data for 1,000 customers over one year,
#' including purchase amounts, dates, and categories.
#'
#' @format A data frame with 1,000 rows and 5 columns:
#' \describe{
#'   \item{customer_id}{Character. Unique customer identifier}
#'   \item{transaction_date}{Date. Date of transaction}
#'   \item{amount}{Numeric. Transaction amount in USD}
#'   \item{category}{Factor. Product category (Electronics, Clothing, Food)}
#'   \item{region}{Character. Customer region (North, South, East, West)}
#' }
#'
#' @details
#' Data was generated synthetically to represent typical e-commerce
#' transaction patterns. Amounts range from $5 to $500.
#'
#' @source Generated using simulation based on real e-commerce patterns
#'
#' @examples
#' head(transactions)
#'
#' # Summary by category
#' aggregate(amount ~ category, data = transactions, FUN = mean)
#'
#' # Transactions over time
#' plot(transactions$transaction_date, transactions$amount)
"transactions"

List

#' Configuration defaults
#'
#' Default configuration settings for the package.
#'
#' @format A list with components:
#' \describe{
#'   \item{api}{List. API configuration:}
#'     \itemize{
#'       \item \code{endpoint}: Character. Base API URL
#'       \item \code{timeout}: Numeric. Request timeout in seconds
#'       \item \code{retries}: Integer. Number of retry attempts
#'     }
#'   \item{cache}{List. Cache settings:}
#'     \itemize{
#'       \item \code{enabled}: Logical. Whether caching is enabled
#'       \item \code{max_size}: Numeric. Maximum cache size in MB
#'     }
#' }
#'
#' @examples
#' config_defaults$api$endpoint
#' config_defaults$cache$enabled
"config_defaults"

Vector

#' Built-in color palette
#'
#' A vector of hex color codes for data visualization.
#'
#' @format A named character vector of length 12:
#' \describe{
#'   \item{Names}{Color names (red, blue, green, etc.)}
#'   \item{Values}{Hex color codes}
#' }
#'
#' @examples
#' palette_colors
#' palette_colors["blue"]
#'
#' # Use in plot
#' plot(1:12, col = palette_colors, pch = 16, cex = 2)
"palette_colors"

Matrix

#' Correlation matrix example
#'
#' Sample correlation matrix for demonstration purposes.
#'
#' @format A 10x10 numeric matrix with row and column names
#'   representing variables v1 through v10. Values are correlations
#'   ranging from -1 to 1.
#'
#' @examples
#' correlation_matrix
#' diag(correlation_matrix)  # All 1s (self-correlation)
"correlation_matrix"

data-raw/ Workflow

Keep data preparation scripts separate from package code.

Setup

# Create data-raw/ directory and template script:
usethis::use_data_raw("dataset_name")

This creates:

• data-raw/ directory
• data-raw/dataset_name.R script
• Adds ^data-raw$ to .Rbuildignore

Data Preparation Script

# data-raw/customer_data.R

## Code to prepare `customer_data` dataset

library(dplyr)
library(lubridate)

# Read raw data:
raw_data <- read.csv("~/Downloads/raw_customer_data.csv")

# Clean and process:
customer_data <- raw_data %>%
  # Clean column names:
  janitor::clean_names() %>%
  # Parse dates:
  mutate(
    transaction_date = ymd(transaction_date),
    signup_date = ymd(signup_date)
  ) %>%
  # Filter to relevant period:
  filter(
    transaction_date >= "2020-01-01",
    transaction_date <= "2023-12-31"
  ) %>%
  # Select and rename:
  select(
    customer_id = id,
    transaction_date,
    amount = transaction_amount,
    category = product_category,
    region = customer_region
  ) %>%
  # Remove duplicates:
  distinct() %>%
  # Sort:
  arrange(transaction_date)

# Save to package:
usethis::use_data(customer_data, overwrite = TRUE, compress = "xz")

Benefits of data-raw/

• Reproducibility: Anyone can recreate the data
• Documentation: Scripts document data transformations
• Version control: Track changes to data preparation
• Separation: Keep raw data separate from package
• Updates: Easy to update data with new source files

Multiple Datasets

# data-raw/prepare_all_data.R

# Prepare dataset 1:
source("data-raw/dataset1.R")

# Prepare dataset 2:
source("data-raw/dataset2.R")

# Or in single file:
dataset1 <- prepare_dataset1()
usethis::use_data(dataset1, overwrite = TRUE)

dataset2 <- prepare_dataset2()
usethis::use_data(dataset2, overwrite = TRUE)

CRAN Size Limits

Size Restrictions

CRAN limits:

• Total package size: 5 MB (compressed)
• Per subdirectory: 1 MB (recommended)
• Larger packages require justification

Check size:

# Build package:
pkgbuild::build()

# Check .tar.gz size:
file.size("../mypackage_1.0.0.tar.gz") / 1024^2  # MB

Reducing Data Size

1. Compression

# Use maximum compression:
usethis::use_data(dataset, compress = "xz")

# Compare sizes:
save(dataset, file = "test_gzip.rda", compress = "gzip")
save(dataset, file = "test_bzip2.rda", compress = "bzip2")
save(dataset, file = "test_xz.rda", compress = "xz")

file.size("test_gzip.rda") / 1024   # KB
file.size("test_bzip2.rda") / 1024
file.size("test_xz.rda") / 1024

2. Subsetting

# Instead of full dataset:
# full_data <- huge_dataset  # 10 MB

# Use representative sample:
sample_data <- huge_dataset %>%
  sample_n(1000) %>%
  select(key_columns)  # Select only essential columns

usethis::use_data(sample_data, compress = "xz")

3. Move to inst/extdata/

# Instead of:
# usethis::use_data(large_dataset)  # data/large_dataset.rda

# Use raw format in inst/extdata/:
write.csv(large_dataset, "inst/extdata/large_dataset.csv.gz")

# Users access with:
#' @examples
#' data_path <- system.file("extdata", "large_dataset.csv.gz", package = "pkg")
#' data <- read.csv(data_path)

4. External Data Packages

For very large data, create separate data package:

mypackage/          # Main package
mypackage.data/     # Data-only package (optional dependency)

# In main package:
# Add to Suggests:
Suggests: mypackage.data

# In functions:
get_full_data <- function() {
  if (!requireNamespace("mypackage.data", quietly = TRUE)) {
    stop("Install mypackage.data: install.packages('mypackage.data')")
  }
  mypackage.data::full_dataset
}

5. Download on Demand

# Store data externally, download as needed:
get_external_data <- function(cache = TRUE) {
  cache_file <- rappdirs::user_cache_dir("mypackage")
  data_file <- file.path(cache_file, "data.rds")

  if (cache && file.exists(data_file)) {
    return(readRDS(data_file))
  }

  # Download:
  url <- "https://example.com/data.rds"
  data <- readRDS(url(url))

  if (cache) {
    dir.create(cache_file, recursive = TRUE, showWarnings = FALSE)
    saveRDS(data, data_file)
  }

  data
}

Dynamic Package State

Use environments for runtime state, not exported data.

Package Environment Pattern

# R/state.R

# Create package environment:
pkg_env <- new.env(parent = emptyenv())

# Initialize in .onLoad():
.onLoad <- function(libname, pkgname) {
  pkg_env$cache <- new.env(parent = emptyenv())
  pkg_env$config <- list(
    api_key = NULL,
    verbose = FALSE
  )
}

# Getters and setters:
#' Get configuration value
#' @param key Configuration key
#' @export
get_config <- function(key) {
  pkg_env$config[[key]]
}

#' Set configuration value
#' @param key Configuration key
#' @param value New value
#' @export
set_config <- function(key, value) {
  pkg_env$config[[key]] <- value
  invisible(value)
}

# Cache functions:
cache_get <- function(key) {
  pkg_env$cache[[key]]
}

cache_set <- function(key, value) {
  pkg_env$cache[[key]] <- value
  invisible(value)
}

cache_clear <- function() {
  rm(list = ls(pkg_env$cache), envir = pkg_env$cache)
  invisible(NULL)
}

When to Use Environments vs data/

Use environments for:

• Runtime configuration
• Caches
• Mutable state
• Session-specific data

Use data/ for:

• Static datasets
• Examples and documentation
• Reference data
• Immutable package constants

Complete Example

Full Data Package Setup

# 1. Create data-raw/ script:
usethis::use_data_raw("customers")

# data-raw/customers.R

## Code to prepare `customers` dataset

library(dplyr)

# Read and clean:
customers <- read.csv("raw/customers.csv") %>%
  filter(active == TRUE) %>%
  select(id, name, region, signup_date) %>%
  arrange(signup_date)

# Save:
usethis::use_data(customers, overwrite = TRUE, compress = "xz")

# 2. Document in R/data.R:

#' Customer data
#'
#' Sample customer records for demonstration and testing.
#'
#' @format A data frame with 500 rows and 4 columns:
#' \describe{
#'   \item{id}{Character. Unique customer ID}
#'   \item{name}{Character. Customer name}
#'   \item{region}{Factor. Geographic region (North, South, East, West)}
#'   \item{signup_date}{Date. Account creation date}
#' }
#'
#' @source Generated from internal CRM system (anonymized)
#'
#' @examples
#' head(customers)
#' table(customers$region)
"customers"

# 3. Add LazyData to DESCRIPTION:
LazyData: true

# 4. Build and check:
devtools::document()
devtools::check()

Common Pitfalls

1. Forgetting LazyData

Without LazyData, users must call data() explicitly:

# Add to DESCRIPTION:
LazyData: true

2. Not Documenting Data

CRAN requires documentation for all exported data:

# Create R/data.R with documentation for each dataset

3. Exceeding Size Limits

# Check package size:
devtools::build()
# Look at .tar.gz file size

# Reduce with:
# - compress = "xz"
# - Subset data
# - Move to inst/extdata/
# - External data package

4. Wrong File Extension

# WRONG:
usethis::use_data(data, file = "data.RData")

# RIGHT:
usethis::use_data(data)  # Automatically creates .rda

5. Including data-raw/ in Package

Should be in .Rbuildignore:

^data-raw$

usethis::use_data_raw() adds this automatically.

6. Using save() Instead of usethis::use_data()

# AVOID:
save(dataset, file = "data/dataset.rda")

# PREFER:
usethis::use_data(dataset)  # Better defaults, automatic compression

7. Not Using system.file() for inst/extdata/

# WRONG:
data <- read.csv("inst/extdata/example.csv")  # Doesn't work after installation

# RIGHT:
path <- system.file("extdata", "example.csv", package = "mypackage")
data <- read.csv(path)

8. Documenting Internal Data

Internal data (R/sysdata.rda) should NOT be documented:

# No documentation needed for internal data
usethis::use_data(internal_obj, internal = TRUE)

9. Missing @format Tag

# INCOMPLETE:
#' My data
"mydata"

# COMPLETE:
#' My data
#' @format A data frame with 100 rows and 5 columns
"mydata"

10. Hardcoding Paths in data-raw/

# FRAGILE:
raw <- read.csv("C:/Users/Me/Desktop/data.csv")

# BETTER:
raw <- read.csv("raw_data/data.csv")  # Relative to project root

# BEST:
raw <- read.csv(here::here("raw_data", "data.csv"))

Quick Reference

Creating Data

# Exported data:
usethis::use_data(dataset, compress = "xz")

# Internal data:
usethis::use_data(internal_obj, internal = TRUE)

# Data preparation:
usethis::use_data_raw("dataset")

# Raw files:
# Manually copy to inst/extdata/

Accessing Data

# Exported data:
data(dataset)          # Explicit load
dataset                # Direct (if LazyData: true)

# Internal data:
internal_obj           # Direct in package code

# Raw files:
system.file("extdata", "file.csv", package = "pkg")

Documentation Template

#' Dataset title
#'
#' Description of what the dataset contains.
#'
#' @format A data frame with X rows and Y columns:
#' \describe{
#'   \item{col1}{Type. Description}
#'   \item{col2}{Type. Description}
#' }
#'
#' @source Where the data came from
#'
#' @examples
#' head(dataset)
"dataset"

Size Management

# Check size:
devtools::build()

# Optimize compression:
usethis::use_data(data, compress = "xz")

# Subset:
data_sample <- data[1:1000, ]

Resources

• R Packages: Data
• Writing R Extensions: Data
• usethis package documentation