Data Browser

Sight includes a data browser for viewing Stata datasets in VS Code. Open .dta files directly in the editor, or call vview from Stata to send the current dataset to VS Code — either way, data renders in a webview panel with a virtualized grid that handles large datasets efficiently.

The vview Command

vview is a Stata ado-file that exports the current dataset and opens it in the Sight Data Browser. It supports the same syntax as browse:

sysuse auto, clear
vview                          // View all data
vview price mpg weight         // View specific variables
vview if foreign == 1          // View with if condition
vview in 1/100                 // View first 100 observations
vview price mpg if foreign, replace  // Refresh existing panel

Options

Option	Description
varlist	Variables to display (default: all)
if / in	Subset observations using Stata's standard qualifiers
replace	Refresh the existing panel instead of opening a new one
name(string)	Custom tab name (default: current filename)
rows(integer)	Cap the number of displayed observations. Applied after if/in filtering — useful when you want to limit output without knowing the exact observation count (e.g., vview, rows(500) on a large dataset)

browse in console Stata

In console Stata (the CLI), the built-in browse command does not exist — it is a GUI-only feature. Sight ships a small browse.ado alongside vview.ado so that, in the CLI only, browse becomes an alias for vview:

browse                 // same as vview
browse price mpg if foreign

This does not affect the Stata GUI. There, the built-in browse is resolved before any ado-file on your path, so the native Data Editor opens exactly as before; Sight's browse.ado never runs. The alias also guards on c(console) as a safeguard.

The standard abbreviations of browse — brows, brow, bro, and br — are aliased too. Stata does not auto-abbreviate ado-file command names, so Sight ships a small forwarder ado for each (br.ado, bro.ado, etc.). In the GUI, the built-in browse command and its abbreviations are resolved first, so all of these still open the native Data Editor; in the CLI they forward to vview. Because the alias forwards to vview, only vview's options apply; native-browse-only options such as nolabel are not supported in the CLI.

Installation

Sight automatically installs its Stata commands (vview.ado, browse.ado, and the browse abbreviation forwarders brows/brow/bro/br) the first time the extension activates. You'll see a one-time prompt asking for permission.

• macOS: ~/ado/
• Windows: %USERPROFILE%\ado\personal\
• Linux: ~/ado/personal/

On macOS, Sight defaults to OLDPLACE (~/ado/) rather than PERSONAL to avoid sandbox-related writes into ~/Documents/Stata/....

If you already have your own copy of any of the generic command names (browse, brows, brow, bro, br) on the ado-path, Sight leaves that file untouched and installs only the ones it owns (vview.ado is always installed, since vview is Sight's own command name).

You can override the install location with the sight.personalAdoDir setting, or manually install with the Sight: Install Stata Commands command from the Command Palette.

To re-trigger the install prompt (e.g., after declining), run Sight: Reset Stata Commands Install Permission from the Command Palette. To remove the installed files (Sight-owned copies only), run Sight: Uninstall Stata Commands.

How It Works

When you run vview in Stata:

1. Stata saves the (optionally subsetted) dataset to a temporary .dta file in ~/.sight/browse/
2. A JSON sidecar file is written alongside it with metadata (variable count, observation count, source file, Stata's working directory, etc.)
3. A signal file is created that triggers the VS Code extension
4. The extension claims the signal (atomically, so only one VS Code window processes it), reads the metadata, and opens the data browser panel
5. The temporary files are cleaned up automatically

Multi-window support: When multiple VS Code windows are open, Sight prioritizes the window whose workspace folder matches Stata's current working directory. This means the data browser is more likely to open in the "right" window when you have multiple projects open.

Opening .dta Files Directly

You can also open .dta files directly in the data browser without using vview:

• Double-click a .dta file in the VS Code Explorer
• Right-click a .dta file and select Open in Sight Data Browser
• Run Sight: Open Data Browser from the Command Palette

This supports Stata formats 113–119 (Stata 8 through Stata 18).

Features

• Virtualized grid: Only visible rows are loaded into memory, so large datasets scroll smoothly
• Column resizing: Drag column borders to resize; widths are persisted across sessions
• Column visibility: Hide/show columns from the column header context menu
• Value labels: Toggle between raw values, formatted values, and value labels
• Missing values: Extended missing values (.a through .z) are highlighted; style configurable via sight.dataBrowser.missingValueStyle
• Theme-aware: The grid automatically adapts to your VS Code color theme
• Sorting: Sort rows by one or more columns (see below)
• Filtering: Filter rows by per-column predicates (see below)

Sorting Rows

Right-click a column header to sort:

• Sort ascending / Sort descending replaces any active sort with this column.
• Add ascending/descending to sort appends the column as the next sort key (also available by holding Shift while clicking Sort ascending/descending). This builds a multi-column sort: rows are ordered by the first key, ties broken by the second, and so on.
• Clear sort on this column / Clear all sorts remove keys.

Sorted headers show a small arrow (▲ ascending, ▼ descending). With more than one key, each header also shows its 1-based priority number. The toolbar grows a chip strip listing the active keys; click a chip to flip its direction, remove it, or move it to first. The status bar appends a sorted by … summary.

Keyboard shortcuts (when the grid is focused and a column is selected):

Shortcut	Action
Shift+Alt+A	Sort the focused column ascending (replace)
Shift+Alt+D	Sort the focused column descending (replace)
Shift+Alt+0	Clear all sorts

Sort details

• Missing values sort last in both ascending and descending order. (This is a data-browser convention, not Stata's "missing is larger than any number" ordering — it keeps real values together at the top however you sort.)
• Labels are WYSIWYG: a value-labelled column sorts by the displayed label when Labels is on, and by the underlying numeric code when it is off. Toggling Labels re-sorts such a column. Display Formats never affect order.
• String columns use natural numeric-aware collation, so file_2 sorts before file_10.
• Sorting runs in the extension host against the on-disk .dta file; Copy column follows the displayed (sorted) order.
• The active sort is remembered per dataset (and dataset shape) and restored the next time you open it, unless sight.dataBrowser.persistSort is disabled.

Filtering Rows

Right-click a column header and choose Filter… (or Edit filter… if the column already has one) to open the filter editor. Pick a condition, enter the value(s), and click Apply (or press Enter). One filter per column; applying a new filter replaces the column's existing one. Multiple columns' filters combine with AND — a row is shown only if it passes every enabled filter.

The available conditions depend on the column's type:

Column kind	Conditions
Numeric	Compare (=, ≠, <, ≤, >, ≥), Between, Not between, Is empty / Is not empty
Value-labelled numeric	Is one of / Is not one of (a checklist of labels), plus all numeric conditions
String	Contains, Does not contain, Starts with, Ends with, Equals, Not equals, Matches regex, Is empty / Is not empty
Date	Compare, Between, Not between, Is empty / Is not empty

Each applied filter becomes a chip in the toolbar's second row. A chip shows a ✓/✗ enabled glyph and a short predicate summary; click its body to edit, or use the ⋯ menu to Enable/Disable or Remove it. A trailing ✕ clears all filters. The status bar appends filtered to N of M (P%).

Keyboard shortcuts (when the grid is focused and a column is selected):

Shortcut	Action
Shift+Alt+F	Open the filter editor for the focused column
Shift+Alt+X	Clear the focused column's filter
Shift+Alt+9	Clear all filters

Filter details

• Missing values (., .a–.z) fail every predicate by default. Tick Include missing in the editor to keep missing rows as well. The Is empty condition matches only missing rows (and ignores the checkbox).
• Numeric "Between" shows a histogram brush: drag the two thumbs (or nudge with arrow keys; hold Shift for a 10× step) to set the low/high bounds, which stay in sync with the typed inputs.
• Value-labelled numerics filter by the underlying numeric code, not the displayed label. The label checklist is for convenience only, so a filter survives toggling Labels on or off. Display Formats never affect which rows match.
• Dates are entered as calendar dates (or date-times for %tc/%tC clock formats) and converted to Stata's internal day/millisecond domain. Only daily (%td, %d) and clock (%tc, %tC) formats are treated as dates; other %t… formats (weekly, monthly, quarterly, …) filter as plain numerics by their stored code.
• Regular expressions are validated live in the editor; an invalid pattern can't be applied. Note that a pathological pattern (catastrophic backtracking) runs synchronously over every row on a large dataset and can briefly stall the view — this is self-inflicted and limited to your own typed pattern.
• Filtering runs in the extension host against the on-disk .dta file; the visible row numbers reflect the filtered sequence, and Copy column follows the displayed (filtered, then sorted) order.
• Active filters are remembered per dataset (and dataset shape) and restored the next time you open it, unless sight.dataBrowser.persistFilters is disabled. Only the filter definitions are stored; the surviving rows are always recomputed against the current data.

Layout Persistence

Column widths and visibility settings are saved automatically in VS Code's global state, keyed by the source dataset path. Layouts persist across sessions and survive dataset refreshes.

Storage Limit

To prevent unbounded storage growth, Sight limits the number of stored layout entries. When the limit is reached, the least recently saved entries are evicted first.

Setting	Default	Description
sight.dataBrowser.maxStoredLayouts	10000	Maximum layout entries stored. Each dataset may use multiple entries for alias keys (source path, temp path, basename).

With the default limit of 10,000 entries and ~3-4 alias keys per dataset, this accommodates roughly 2,500-3,000 unique datasets before eviction begins.

Settings

Setting	Default	Description
sight.dataBrowser.missingValueStyle	"foreground"	How to highlight missing values: "foreground" (colorize text), "background" (tint cell), or "none"
sight.dataBrowser.persistSort	true	Remember and restore the row sort per dataset (matching shape)
sight.dataBrowser.persistFilters	true	Remember and restore the row filters per dataset (matching shape)
sight.dataBrowser.maxStoredLayouts	10000	Maximum stored layout entries (see above)
sight.personalAdoDir	(platform default)	Path where Sight installs its Stata commands