Skip to content
Kun Ren edited this page May 5, 2022 · 9 revisions

R session watcher enables the communication between vscode-R and a live R session to support the following features:

  • Watch any R session
  • Show value of session symbols on hover
  • Provide completion for session symbols
  • View() any objects including data frames and list objects
  • Show plot output on update and plot history
  • Show htmlwidgets, documentation and shiny apps in WebView

Basic usage

To enable this feature, turn on r.sessionWatcher in VSCode settings, reload or restart VSCode, and the session watcher will be activated automatically on R sessions launched by vscode-R via R: Create R Terminal command.

If you previously appended the source(...) line to ~/.Rprofile, you may safely remove it since the configuration for basic usage is automated. It is now only necessary for advanced usage described below.

Advanced usage (for self-managed R sessions)

For advanced users to work with self-managed R sessions (e.g. manually launched R terminal or started in tmux or screen window), some extra configuration is needed. Follow the steps below to make R session watcher work with any external R session:

  1. Turn on r.sessionWatcher in VSCode settings.

  2. Edit .Rprofile in your home directory by running the following code in R:

    file.edit("~/.Rprofile")
  3. Append the following code to the file:

    if (interactive() && Sys.getenv("RSTUDIO") == "") {
      source(file.path(Sys.getenv(if (.Platform$OS.type == "windows") "USERPROFILE" else "HOME"), ".vscode-R", "init.R"))
    }
  4. Restart or Reload Window in VSCode

If the workspace folder you open in VSCode already has a .Rprofile, you need to append the code above in this file too because ~/.Rprofile will not be executed when a local .Rprofile is found.

The script only works with environment variable TERM_PROGRAM=vscode. The script will not take effect with R sessions started in a tmux or screen window that does not have it, unless this environment variable is manually set before sourcing init.R, for example, you may insert a line Sys.setenv(TERM_PROGRAM="vscode") before it.

Available functions and options

When the session watcher is initialized on session startup, a local environment named tools:vscode is attached and the following functions are made available for user to interoperate with VSCode:

# Attach vscode-R with same workspace folder to current session.
.vsc.attach()

# A customizable View() where title and viewer can be specified.
.vsc.view(x, title, viewer)

# Browse an URL in a WebView (used by e.g. shiny apps, R html help).
.vsc.browser(url, title, ..., viewer)

# Show viewer from a HTML file or htmlwidget object (used by most htmlwidgets).
.vsc.viewer(url, title, ..., viewer)

# Show page viewer from a HTML file or htmlwidget object (used by e.g. profvis).
.vsc.page_viewer(url, title, ..., viewer)

All WebView-related functions have a viewer argument which could be one of the values defined in vscode-api#ViewColumn, .e.g. "Active" (current editor), "Two" (editor group 2), or "Besides" (always show besides the current editor).

The following options are available for user to customize the session watcher functionality and behavior:

# Watch global environemnt symbols to provide hover-on and
#   completion-after session symbol.
# Only specify in .Rprofile since it only takes effect on session startup.
options(vsc.globalenv = TRUE | FALSE)

# Which view column to show the plot file on graphics update
# Use FALSE to diable plot watcher so that the default R plot device is used.
# Only specify in .Rprofile since it only takes effect on session startup.
options(vsc.plot = "Two" | "Active" | "Beside" | FALSE)

# The arguments for the png device to replay user graphics to show in VSCode.
# Ignored if options(vsc.plot = FALSE).
options(vsc.dev.args = NULL | list(width = 800, height = 600))

# Which view column to show the WebView triggered by browser (e.g. shiny apps)?
# Use FALSE to open in external web browser.
options(vsc.browser = "Active" | "Beside" | "Two" | FALSE)

# Which view column to show the WebView triggered by viewer (e.g. htmlwidgets)?
# Use FALSE to open in external web browser.
options(vsc.viewer = "Two" | "Active" | "Beside" | FALSE)

# Which view column to show the WebView triggered by page_viewer (e.g. profvis)?
# Use FALSE to open in external web browser.
options(vsc.page_viewer = "Active" | "Beside" | "Two" | FALSE)

# Which view column to show the WebView triggered by View()?
# Use FALSE for R's native View(), which should be specified in .Rprofile
#   since it only takes effect on session startup.
options(vsc.view = "Two" | "Active" | "Beside" | FALSE)

# Which view column to show the WebView triggered by help panel
# (e.g. after sending `?mean` to terminal)?
# Use FALSE to disable help panel and revert to old behaviour.
options(vsc.helpPanel = "Two" | "Active" | "Beside" | FALSE)

# How much of the object to show on hover and autocomplete detail?
# As controlled by max.level arg of str().
# Use 0 (or 1) is the default - literal value or object type and dimensions
# Use 2 to show list contents, data frame columns, and example values.
options(vsc.str.max.level = 0 | 2 )

# Show object size in the workspace viewer tooltip
# The R session could become laggy if there are large character vectors in
#   objects in the global environment when this option is `TRUE`.
options(vsc.show_object_size = FALSE | TRUE)

# Use the httpgd-based plot viewer.
# Requires the R package `httpgd` to be installed.
options(vsc.use_httpgd = FALSE | TRUE)

The first values are the default and all subsequent values after | are available choices. The "Two" | "Active" | "Besides" are popular values from ViewColumn to specify which view column should the corresponding tab appears in VSCode.

How to disable it

For the case of basic usage, turning off r.sessionWatcher in VSCode settings is sufficient to disable R session watcher.

For the case of advanced usage, user should, in addition, comment out or remove the source(...) line appended to ~/.Rprofile.

How it works

  • When vscode-R is activated with session watcher enabled, it deploys the initialization script to ~/.vscode-R/init.R.
  • vscode-R watches ~/.vscode-R/request.log for requests from user R sessions.
  • When a new R session is created, it sources init.R to initialize the session watcher and writes attach request to ~/.vscode-R/request.log.
  • vscode-R reads the attach request and knows the working directory and session temp directory ({tempDir}) of the attaching session.
  • vscode-R watches {tempDir}/vscode-R/globalenv.json for global environment info and {tempDir}/vscode-R/plot.png for plot graphics.
  • In the R session, the global environment info will be updated on each evaluation of top-level expression.
  • When user creates or updates a plot, the {tempDir}/vscode-R/plot.png is updated, and vscode-R will open the plot file.
  • When user calls View() with a data frame, list, environment, or any other object, the request is written to ~/.vscode-R/request.log and vscode-R will open a WebView to show the data or open a text document to show the content of the object.
  • When user calls the viewer (e.g. htmlwidget, provis) or browser (e.g. shiny app, HTML help documentation), the request is written to ~/.vscode-R/request.log and vscode-R will open a WebView to present the viewer content.

R sessions started from the workspace root folder or a subfolder will be automatically attached. The session watcher is designed to work in a wide range of scenarios:

  • Official R terminal or radian console
  • R session started by vscode-R or user
  • R session in a tmux or screen window
  • Multiple R sessions in VSCode terminal
  • Multiple R sessions in tmux windows or panes.
  • Multi-root workspace in VSCode
  • Switch between multiple running R sessions
  • Remote Development via SSH, WSL and Docker

The status bar item shows the process id of the attached R session. Click the status bar item and it will attach to currently active session.

Attached R process

R session watcher

The R terminal used in the screenshot is radian which is cross-platform and supports syntax highlighting, auto-completion and many other features.