'SAS Studio'-Style Interactive Dataset Viewer

An interactive dataset viewer that renders a fast, scrollable grid with a column-selection panel, per-column property metadata, and a names-versus-labels header toggle, modelled on the 'SAS Studio' table viewer. Runs from one codebase in interactive 'Shiny' apps and in static HTML documents, with a free-text row filter, header sort, and CSV export, and handles large datasets without row sampling by querying them in the browser with 'DuckDB-WASM'.

datasetviewer

An interactive, SAS Studio-style dataset viewer for R. datasetviewer renders a dataset in a fast, scrollable grid modelled on the SAS Studio table viewer, from one htmlwidget that runs in interactive Shiny apps and in static HTML documents.

• Fast on large data, no sampling. The data is sent to the browser as Parquet and queried in place with DuckDB-WASM, so filtering, sorting, and scrolling stay responsive on datasets that overwhelm DOM-bound grids or server-paged tables.
• SAS Studio layout. A column-selection panel with char/num type icons and a collapse toggle, a property pane (Label, Name, Length, Type, Format), a names-versus-labels header toggle, click-a-header to sort (Shift-click for multi-column sort), and a free-text “Filter Table Rows” expression.
• Clinical metadata. With the sibling artoo package installed, column labels, formats, and storage lengths are read from a labelled or CDISC-conformed frame. Plain data frames work with zero artoo dependency.
• Runs offline / behind a firewall. The grid, filters, code view, and CSV export are bundled in the package. The DuckDB-WASM engine loads from a CDN by default but is fetched into the package at install time, so a Shiny app can serve it to browsers with no internet at runtime.

Installation

Install the released version from CRAN:

install.packages("datasetviewer")

Or the development version from GitHub:

# install.packages("pak")
pak::pak("vthanik/datasetviewer")
# or
remotes::install_github("vthanik/datasetviewer")

Usage

library(datasetviewer)

# a plain data frame
dataset_viewer(mtcars)

# an artoo-conformed frame, headers shown as labels
adsl <- artoo::read_dataset("adsl.parquet")
dataset_viewer(adsl, view = "labels")

Click a column header to select it, then click again to cycle its sort — ascending, descending, then back to unsorted. Shift-click additional headers to build a multi-column sort, each shown with its direction and priority (AGE ↑1, SEX ↓2). In the right-click menu, Sort Ascending/Descending add that column to the sort and Clear Sorting removes just that column; you can also add a filter or copy the column or its header. Click the funnel in the toolbar to edit the filter expression directly (for example AGE > 50 and SEX = "M").

In Shiny

library(shiny)
ui <- fluidPage(datasetviewerOutput("viewer", height = "500px"))
server <- function(input, output, session) {
  output$viewer <- renderDatasetViewer(dataset_viewer(adsl))
  # input$viewer_filter, input$viewer_sort, input$viewer_columns,
  # input$viewer_view track the user's current view
}
shinyApp(ui, server)

Offline and corporate use

Everything except the DuckDB-WASM query engine (~35 MB, too large to ship in a package) is bundled and works offline. The engine loads from a CDN by default, but — like the arrow package’s C++ library — it is fetched into the package at install time, so a Shiny app can serve it to browsers with no internet at runtime. Steer the install-time fetch for air-gapped or mirrored environments:

The install never fails if the engine cannot be fetched; the widget falls back to the CDN at runtime. For self-contained HTML documents, set options(datasetviewer.use_local_engine = FALSE) to load the engine from the CDN instead of embedding it. See vignette("datasetviewer") for details.