Skip to contents

Getting started with protondrive

Source: vignettes/protondrive.Rmd
protondrive.Rmd

This is a third-party package that is not officially supported by Proton.

protondrive manages files on Proton Drive from R. It works by running the official Proton Drive CLI, which is built on the Proton Drive SDK and does all the encryption and API work. The code in this vignette is not run when the package is built, because it needs a signed-in Proton account.

Setup

Install the CLI from https://proton.me/download/drive/cli/index.html, then check that protondrive can find it:

Sign in. A browser window opens, and once you finish signing in the session is stored in your operating system’s secret store. You only need to sign in once per machine:

pd_auth()
pd_has_auth()
#> [1] TRUE

Paths

Proton Drive paths are virtual and always use /. The first part of a path is a section:

path contents
/my-files your own files and folders
/devices computers backed up with the desktop app
/shared-with-me items others have shared with you
/shared-by-me items you have shared
/trash trashed items
/photos, /albums Proton Photos

A path that doesn’t start with / is relative to /my-files, and ~ means /my-files, so these are equivalent:

pd_ls("reports/2026")
pd_ls("~/reports/2026")
pd_ls("/my-files/reports/2026")

If a name itself contains a /, write it as \/. In an R string that is "a\\/b".

The dribble

As in googledrive, a dribble (“drive tibble”) is a tibble with one row per file or folder:

x <- pd_ls("reports")
x
#> # A dribble: 3 × 4
#>   name     path                       id                       drive_resource
#>   <chr>    <chr>                      <pd_id>                  <list>
#> 1 2025     /my-files/reports/2025     xKq3…H9w~Rt7a…kL0=       <named list>
#> 2 2026     /my-files/reports/2026     xKq3…H9w~p2Vd…Qe4=       <named list>
#> 3 q3.csv   /my-files/reports/q3.csv   xKq3…H9w~9sLm…Zb1=       <named list>

drive_resource holds everything the SDK knows about each item. pd_reveal() pulls fields into columns:

x |> pd_reveal(c("type", "size", "modified_time", "shared"))

Every function that takes a file or path argument accepts a path, an ID wrapped in as_id(), or a dribble. Dribbles work well with dplyr:

library(dplyr)

pd_ls("reports", recursive = TRUE) |>
  pd_reveal("modified_time") |>
  filter(modified_time < as.POSIXct("2025-01-01")) |>
  pd_trash()

Uploading and downloading 

write.csv(mtcars, "mtcars.csv", row.names = FALSE)

pd_mkdir("data")
f <- pd_upload("mtcars.csv", path = "data")

Proton Drive never allows two items with the same name in one folder. By default pd_upload() fails if the name is taken. pd_put() (or overwrite = TRUE) uploads a new revision of the existing file instead, which keeps its ID, sharing settings, and version history. The CLI skips the upload entirely if the content is unchanged.

pd_put("mtcars.csv", path = "data")
pd_update(f, "mtcars.csv")

Download to disk, or read straight into R:

pd_download(f, path = "mtcars-copy.csv")
read.csv(text = pd_read_string(f))

Downloads are decrypted and verified by the CLI. Proton Docs and Sheets cannot be downloaded this way.

Organising 

pd_cp(f)                                 # "Copy of mtcars.csv", same folder
pd_cp(f, path = "archive", name = "mtcars-2026.csv")
pd_rename(f, "cars.csv")
pd_mv(f, path = "archive")
pd_size("archive")

Trash 

trashed <- pd_trash(f)
pd_ls("/trash")
pd_untrash(trashed)

pd_rm(f)            # permanent: trashes, then deletes
pd_empty_trash()    # permanent: everything in /trash

The CLI identifies trashed items by name. If two trashed items share a name, pd_untrash() and pd_rm() refuse to guess.

Sharing 

pd_share("reports", c("ana@proton.me", "bo@example.com"), role = "editor")
pd_sharing("reports")
pd_unshare("reports", "bo@example.com")

url <- pd_share_link("reports/q3.csv", password = "s3cret", expiration = Sys.Date() + 7)
pd_link("reports/q3.csv")
pd_unshare_link("reports/q3.csv")

Invitations from others:

inv <- pd_invitations()
pd_accept_invitation(inv$id[[1]])
pd_ls("/shared-with-me")

Using protondrive in scripts and tests

Sign in interactively once. After that, non-interactive R sessions on the same machine reuse the session. To silence status messages, use local_pd_quiet() or with_pd_quiet(). In package tests, skip_if_no_pd() skips when the CLI is missing or not signed in.