Feature tour

This is the whole extension in the order you would meet it. Every screenshot is the real thing; nothing here is a mock-up. Each section links to the guide that covers it properly.

The grid

The grid is a custom editor sitting on top of the same text document VS Code already has open. That single design decision is what makes everything else behave:

• Save, undo, redo and the dirty dot are VS Code’s own. An edit in a cell is a text edit on the document. Ctrl+Z walks back through it, Git sees a normal diff, and an external change to the file on disk reloads the grid.
• Virtualised rendering keeps scrolling smooth well past 100,000 rows.
• Delimiter detection covers comma, tab, semicolon and pipe, with a per-file override in the toolbar.
• Column types are inferred as integer, number, boolean, date or text. The type drives right-aligned numbers, numeric rather than lexical sorting, and which filter operators you are offered.
• Range selection works like a spreadsheet, copies to and pastes from Excel, and puts count, sum, average, minimum and maximum in the status bar.

The grid editorEditing, saving, large files, and everything the toolbar does.

Excel-style AutoFilter

Click the caret in any column header, or press Alt+Down. You get the dropdown you already know:

• A searchable list of distinct values with row counts, tri-state (Select All) and an explicit (Blanks) entry.
• Condition filters appropriate to the column’s type — contains, begins with, greater than, between, before, is empty — two at a time, joined by And or Or, with * and ? wildcards.
• Cascading value lists. The values on offer reflect the filters already applied to other columns, so you can never pick a combination that matches zero rows.
• Sorting, and a one-click clear per column.

Filtering, search and sortEvery operator, how cascading works, and multi-level sort.

Find, replace and sort

Ctrl+F opens find and replace over cells rather than over text: match case, whole cell and regular expression modes, live highlighting of every match, next and previous, and a replace-all that lands as one undo step.

The Sort dialog does up to three levels and asks the question that matters: sort the view, or write the new row order to the file.

Row and column surgery

Cell menu

Column menu

Insert, duplicate, move, delete and hide — for rows and columns, on a single item or a whole selection. Beyond that, the operations you would otherwise open a spreadsheet for:

Reshape

Split one column into several on a delimiter or a width. Merge several into one with a separator. Transpose the entire table.

Clean

Remove empty rows, remove duplicate rows, trim whitespace, change case, and normalize rows whose field count does not match the header.

Fill

Fill down with Ctrl+D and right with Ctrl+R, or fill a series — numbers, dates, or a repeating pattern.

Statistics

Per column: count, distinct, blanks, minimum, maximum, shortest and longest value, and the most frequent values with their counts. For numeric columns it adds sum, mean, median, standard deviation and quartiles.

This is usually the fastest way to answer what is actually in this file — before you write a single query.

Charts

A histogram for numeric columns, a top-values bar chart for everything else. One click from the statistics panel, drawn with Chart.js inside the editor.

AnalysisStatistics, charts and SQL in detail.

SQL

The SQL panel loads the file into an in-memory SQLite database — real SQLite, compiled to WebAssembly — as a table named csv. That means joins onto itself, window functions, CTEs, GROUP BY, the lot:

SELECT category,
       COUNT(*)        AS products,
       ROUND(AVG(price), 2) AS avg_price,
       SUM(qty)        AS units
FROM csv
WHERE in_stock = 'true'
GROUP BY category
ORDER BY units DESC;

Results copy to the clipboard or open as a new CSV in their own grid.

Note

Nothing leaves your machine. The database is created in memory from the file you have open and discarded when you close the panel.

Pipelines

The same cleanup usually happens more than once. Describe it once as a *.csvpipe.json file, then run it on demand, when the file opens, or when it is saved.

Tier	Steps
No-code	trim, change case, fill empty, find and replace, rename, select, drop, sort, dedupe, limit, split column, merge columns, transpose, remove empty rows, random sample
Low-code	filter and compute — one-line expressions where columns are variables and 38 helper functions are in scope
Code	inline JavaScript, a workspace script file, a SQLite query, or any external command over stdin and stdout

{
  "name": "Clean sales export",
  "applyTo": ["data/sales-*.csv"],
  "trigger": "onSave",
  "output": { "target": "file", "format": "json", "path": "out/${name}-summary.json" },
  "steps": [
    { "kind": "trim" },
    { "kind": "filter", "expression": "in_stock && units_sold > 0" },
    { "kind": "compute", "column": "margin", "expression": "round(revenue * 0.18, 2)" },
    { "kind": "sql", "query": "SELECT category, SUM(revenue) AS revenue FROM csv GROUP BY category" }
  ]
}

Because a command step simply pipes the table through your shell, anything already installed works: pandas, R, jq, awk, csvkit, Miller, qsv.

Trust matters here

Steps that evaluate code — expression, js, script, sql, command — never run in an untrusted workspace, and can be switched off entirely with csv.pipelines.allowScripts.

Pipelines guideWriting, running and triggering pipelines, with the expression helper reference.

Pipeline schemaEvery field and every step kind, generated from the schema the extension validates against.

Export

The whole file, or just the selection, to:

• JSON — an array of objects keyed by header
• Markdown — a pipe table you can paste into a README
• HTML — a standalone table
• SQL — CREATE TABLE plus INSERT statements
• Another delimiter — CSV to TSV, TSV to pipe-delimited, and so on

Selections can go straight to the clipboard in a form Excel and Google Sheets both understand.

The sidebar

The CSV icon in the activity bar opens three views.

Settings

Every option, in place

Grouped, showing its current value, editable with a toggle for booleans, a picker for choices and a validated prompt for numbers. It also surfaces the per-file header and delimiter overrides, which are otherwise invisible once set from the toolbar. A control in the view title chooses whether edits go to User or Workspace settings.

Pipelines

Everything runnable, ranked

Workspace pipelines with the ones matching the current file first. Run or edit either one in a single click.

Files

Every delimited file

Every .csv, .tsv, .tab and .psv in the workspace, opening straight into the grid or into the text editor.

Text mode

Not everything wants a grid. Open as text puts you back in the plain editor, where the extension still helps:

• Each column gets its own colour, via semantic highlighting that respects your theme.
• Hovering a cell names its column and position.
• The status bar tracks the column under the cursor.
• Rows whose field count differs from the header are reported in the Problems panel.

Try all of it in your browserA live grid running the extension's own parsing, type inference and statistics code — no install required.