sdv — open R objects in StatDataViewer

中文说明

An RStudio addin that sends variables from your R session to the StatDataViewer desktop app. You can send them with a keyboard shortcut, from the Addins menu, or through View().

Installation

install.packages("sdv", repos = c("https://sashelper.r-universe.dev",
                                  "https://cloud.r-project.org"))

You can also install the development version from GitHub:

# install.packages("pak")
pak::pak("sashelper/sdv")

You also need StatDataViewer itself, on macOS or Windows.

Usage

How	Setup
Keyboard shortcut ⌘⇧V (Cmd+Shift+V; Ctrl+Shift+V on Windows)	Run `sdv::sdv_install_shortcut()` once, then restart RStudio. Pick another key with `sdv_install_shortcut("Cmd+Shift+K")`.
Addins menu / Command Palette (⇧⌘P)	Search for "StatDataViewer".
`View(x)` and the grid icon in the Environment pane	Turn on with `sdv::sdv_use_as_viewer(TRUE)`; `FALSE` restores RStudio's viewer.
Console	`sdv::sdv_view(x)`, or `library(sdv)` and then `sdv_view(x)`.

Every exported function starts with sdv_, so library(sdv) does not mask functions from other packages.

The Open in StatDataViewer addin decides what to send in this order:

Selected code, in the editor or the console, is evaluated in the global environment and the result is sent. This works for head(mtcars, 3) or a whole pipeline.
With nothing selected, it sends the object name under the cursor. That can be df, df$col, obj@slot, pkg::obj or `a name with spaces`. With the cursor on df in df$col it sends df; on col it sends df$col.
If neither works, or evaluation fails, a prompt asks you for an object name or expression.

RStudio does not let addins add items to its right-click menus. Taking over View() is the closest substitute: once it is on, clicking the grid icon next to an object in the Environment pane opens it in StatDataViewer. To turn it on in every session, add the snippet that sdv_use_as_viewer() prints to your ~/.Rprofile.

Supported objects

R object	Shown in StatDataViewer as
data.frame, tibble, data.table, sf	A table. Row names that aren't the default become a `row_names` column.
matrix, 2-D table / xtabs	A wide table
arrays and tables with 3+ dimensions	A long table
ts / mts, zoo / xts	A table with `time` and `cycle` columns (or `index`) plus the values
atomic vectors, factors, Dates	A `value` column, plus a `name` column for named vectors
lists	A table when all elements have the same length; otherwise a tree with path / class / type / size / preview columns
environments	A list of the objects they contain
functions, formulas, calls	`.R` text
lm / glm	Text of `summary()`
S4, R6, htest and anything else	Text of `print()` and `str()`

StatDataViewer's R reader displays only a few column types, so each column is converted first:

factor → its labels.
POSIXct / POSIXlt → the wall-clock time in the original time zone. The viewer always shows times in UTC, so the values are shifted, and the time zone is recorded in the column label.
Date → stored as double. StatDataViewer versions before the fix mis-read integer-stored dates, such as those from seq() in recent R versions.
list columns, complex, raw → short text summaries.
difftime → a number, with the unit in the column label.
hms → HH:MM:SS text.
integer64 → a number, or text when the value is beyond 2^53.
haven_labelled → the codes, plus an extra *_lbl column holding the labels.
data frame and matrix columns → several columns.

Options

Option	Purpose
`sdv.app`	Path to the `StatDataViewer.app` to use (macOS). Useful if several copies are installed.
`sdv.exe`	Path to `StatDataViewer.exe` (Windows)
`sdv.dir`	Folder for the temporary files; default `tempdir()/sdv`
`sdv.view_types`	`"tabular"` (default) or `"all"`: which objects `View()` sends
`sdv.max_text_lines`	Maximum number of lines in text output
`sdv.list_max_rows`	Maximum number of rows when a list is shown as a tree

How it works

sdv_prepare() converts the object into a plain data frame or into text.
The result is written to a new temporary folder with saveRDS(version = 2), uncompressed. A new folder each time means the viewer never asks "Reload Dataset?". The file is named after the object, so that name is also the tab title.
macOS: open -b com.statdataviewer <file> hands the file to the running app, or starts it first.
Windows: runs StatDataViewer.exe <file>. The app passes the file on to the instance that is already running.

Extending

Add an S3 method for your own class. It can return a data frame (shown as a table), a character vector (shown as text) or NULL:

sdv_prepare.myclass <- function(x, name = "object", ...) as.data.frame(x$data)

Development

Rscript -e 'testthat::test_local()'

test-librdata-compat.R is optional. When the StatDataViewer source tree is present (SDV_SOURCE_DIR), it compiles a small probe with the viewer's R reader under AddressSanitizer. The probe checks, cell by cell, what the viewer would display for every kind of output. The test is skipped everywhere else, including CI.

See RELEASING.md for the release steps.

License

MIT