boxr: A high-level R interface for the API

Brendan Rocks


Getting Set Up

To use boxr, you need to enable API access for your account. The process is slightly annoying. You only need to do it once - it takes around 2 minutes.

1. ‘Create an app’

At, click ‘My Apps’ in the top right-hand corner, log in, and create a new ‘app’ for your account. You can call it anything you like. This won’t do anything remotely like creating an app, but it does allow you to access your account via the API.

2. Set OAuth2 Parameters

On the next screen, you’ll want to set Content API Access Only, and http://localhost as your redirect_uri as in the screenshot below.

Setting OAuth2.0 parameters

Setting OAuth2.0 parameters

3. Connect boxr to your account

This means passing your client_id and client_secret to the box_auth function. These strings are not enough for someone to access your account maliciously. However, it’s still a good idea to keep them safe, and out of any files or code which might be shared with others.



And paste/type the client_id and client_secret when prompted. If these are valid, a browser window should open, for you to formally grant yourself access to your files at

4. And you’re done

If box_auth() worked successfully, you won’t need to do any of this again, and thanks to the magic of httr everything should just work. Your client_id and client_secret will be securely stored in your R environment variables, your hashed OAuth2.0 token will stored at ~/.boxr-oauth, .gitignore’d if necessary, and automatically refreshed when needed.


Basic Operations

Aside from file upload/download, boxr provides functions which mirror base R operations for local files.

Directory wide Operations

Cloud storage services can complement version control systems for code, which aren’t well suited to large binary files (e.g. databases, .RData, or heaps of pdfs). box explicitly versions binary files, keeping old ones, and making it easy fall back to an older copy.

boxr provides git style facilities to upload, download, and synchronize the contents of entire local and remote directories. At the time of writing, the API does not support this directly, and so boxr recursively loops through directory structures.

Synch a whole directory!

Synch a whole directory!

These functions all have overwrite and delete parameters, which are set to FALSE by default.

Disclaimer: is no replacement for a VCS/remote-database, and familiar verbs are no guarantee of expected behavior! Do check the function documentation before jumping in.


boxr’s functions have been designed to be ‘pipable’. Here’s a little example:


# 'nycflights13.json' is the same as nycflights13::flights, if you want to 
# follow along at home


box_search("nycflights13.json") %>%                # Find a remote file
  box_read() %>%                                   # Download it as a data.frame
    group_by(origin, dest, month) %>%              #   Do some, er, cutting edge 
    summarise(mu = mean(arr_delay), n = n()) %>%   #   analysis with dplyr!
  box_write("delay_summary.xlsx") %>%              # Convert to .xlsx, upload
  box_add_description("Check out these averages!") # Add a description to your file!

File/Folder IDs

Are how identifies things. You can find them in an item’s URL:

Finding file and folder ids

Finding file and folder ids



boxr is by default rather verbose, printing status to the console with cat. This is ‘rude’ package behaviour, and may cause unwanted output if used in conjunction with the excellent knitr package.

To supress messages produced using cat, set boxr’s verbose option with:

options(boxr.verbose = FALSE)


boxr aims to expedite data analysis/communication/distribution. Other ways to manipulate a account include:

Managing your client id & secret

If you don’t like the idea of typing credentials into your console, you can put them straight into ~/.Renviron yourself, prior to the R session:


(Note the final blank line).

Reporting Bugs

boxr is a realtively new package. If you find anything that looks like a bug while using it, please report it!

The best way to do this is via a GitHub issue, at: