pins for R

The pins package allows you to publish data, models, and other R objects to shared locations called boards.

A board is an abstraction representing a storage backend (e.g., a local folder, Posit Connect, Amazon S3, Google Cloud Storage, etc.). Once a board is initialized, you use it to write (pin) or read objects.

Key workflow:

1. Create a board: Initialize a board object using functions like board_folder() or board_connect().
2. Pin an object: Use pin_write() to save an object to the board with a specific name.
3. Read an object: Use pin_read() to retrieve the object from the board using its name.

library(pins)

# 1. Create a board
board <- board_temp()

# 2. Pin an object
board |> pin_write(head(mtcars), "mtcars")

# 3. Read the object
board |> pin_read("mtcars")

You can install the stable version of pins from CRAN or the development version from GitHub.

To install from CRAN:

install.packages("pins")

To install the development version using pak:

# install.packages("pak")
pak::pak("rstudio/pins-r")

Use pin_write() to save an object to a board. It requires the board object, the object itself, and a name for the pin.

By default, pins may choose a file format like parquet. You can explicitly specify the storage format using the type argument (e.g., "rds", "arrow", "csv", or "json").

board |> pin_write(head(mtcars), "mtcars")

# Specifying a type
board |> pin_write(tidy_sales_data, "sales-summary", type = "rds")

Use pin_read() to retrieve an object from a board by its name.

board |> pin_read("mtcars")

Pins supports various storage backends through different board initialization functions:

• Local/Temporary:
  • board_temp(): Creates a temporary board that is automatically deleted when the R session ends. Useful for testing.
  • board_folder(): Stores pins in a specified directory (e.g., a networked drive or Dropbox).
• Cloud & Enterprise:
  • board_connect(): Connects to Posit Connect.
  • board_databricks(): Uses Databricks Volumes.
  • board_s3(): Uses Amazon S3.
  • board_azure(): Uses Azure blob storage.
  • board_gcs(): Uses Google Cloud Storage.

Refer to vignette("pins") for detailed documentation on cloud boards.

board <- board_temp()
board <- board_folder("path/to/folder")
board <- board_connect()
board <- board_s3(bucket = "my-bucket")