CSV Studio

A Swiss Army knife for CSV in VS Code. You get a typed grid to view and edit data, Excel-style filters, multi-column sorting, SQL queries, column transforms including decryption, column profiles, lint and repair, comparison, split and merge, and export to almost anything.

For a guided tour, read the blog post CSV Studio: Query, Filter and Decrypt CSV Files Without Leaving VS Code.

Open it

• Click the split-view icon in the editor title bar of any .csv / .tsv / .psv file (Open CSV Studio to the Side).
• Or right-click a file in the Explorer and choose Open in CSV Studio, or use Reopen Editor With… → CSV Studio.
• To make it the default, add "workbench.editorAssociations": { "*.csv": "csvStudio.editor" } to your settings.

CSV Studio edits the real file. A cell edit rewrites only that field, so the rest of the file stays exactly as it was: quoting, delimiter and line endings are unchanged. Undo, save and git work normally. Keep the text editor open beside it and the two stay in sync.

Features

Grid and column types

• A virtualized grid that stays smooth with hundreds of thousands of rows. Columns can be resized, dragged to reorder, hidden and frozen, and long text can wrap.
• Column types are detected automatically:
  • Numbers: integer, decimal (including 1.234,5), percent, currency.
  • Dates: dates (ISO, dd/mm/yyyy and mm/dd/yyyy are told apart automatically) and datetimes.
  • Other: boolean, email, URL, UUID, JSON.
• Click a type badge to override it. Overrides are remembered per file.
• Types control sorting, filters and alignment, and how values are stored in SQL. Cells that don't match their column's type are underlined in red.
• Values with leading zeros (007) and numbers with more than 15 digits (card numbers, long IDs) are kept as text, so they never lose data.
• The delimiter, header row, BOM and line endings are detected automatically. The chip in the toolbar lets you read the file differently, for example with no header row or a semicolon delimiter.

Filter, sort and search

• Shift+click column headers to sort by several columns. Empty values always sort last.
• The ⏷ button on each column opens a filter that matches the column's type:
  • Numbers and dates: >, ≥, <, ≤, between.
  • Text: contains, regex, starts with.
  • Any column: empty / not empty, or only cells that don't match the column's type.
• The same panel lists the column's distinct values with counts, with checkboxes, like Excel.
• Right-click any cell for Filter to this value, Exclude this value, and ≥ / ≤ this value.
• Filters and sorts show as chips you can remove. Global search covers every cell.
• Tools → Apply current sort to file makes the sort permanent.

SQL queries

Open the SQL panel, or press ⌘⌥Q / Ctrl+Alt+Q. The file is an in-memory SQLite table called data, also available under its file name:

SELECT region, COUNT(*) AS orders, ROUND(SUM(qty * unit_price), 2) AS revenue
FROM data
WHERE order_date >= '2024-01-01'
GROUP BY region
ORDER BY revenue DESC

• Columns are stored as their detected types. $1,200 becomes the number 1200, and dates become ISO text, so strftime() works.
• + table (or CSV Studio: Add Table from File…) adds other CSV files as tables you can JOIN.
• Extra functions:
  • REGEXP, regexp_replace, regexp_extract.
  • Every transform below that has a SQL name: decrypt(card, 'my-key'), mask(x, 0, 4), hash(x), jwt_payload(token), b64decode(x)…
  • Your own JavaScript transforms, as js_<name>(x).
• Autocomplete, Examples ▾ built from your actual columns, a ƒ Functions reference, history and saved queries.
• From view turns your current filters and sort into the equivalent SQL.
• Run with ⌘/Ctrl+Enter, or select text to run just part of the query.
• Save / Export ▾ writes all result rows to a new CSV, JSON, YAML, Markdown, HTML or SQL file, or a new tab. It can also copy the result as TSV, or replace the file with it.

Save results to another file

Export saves the current filtered and sorted view, or just the rows you selected (click and Shift/Cmd+click row numbers). You can export all columns or only the visible ones in display order, and choose:

• Format: CSV, TSV, JSON, NDJSON, YAML, Markdown, HTML or SQL INSERT statements.
• CSV options: delimiter, quoting (only when needed / all / non-numeric), header on or off, LF or CRLF.
• Encoding: UTF-8, UTF-8 with BOM (Excel-friendly), UTF-16 or Windows-1252.

Column transforms, including decryption

Open Transform from the toolbar, a column header or a cell. Pick a column and a transform, and see a live preview of 50 rows before and after, including any per-row errors. Then write the result:

• Virtual column: shown in the grid and usable in SQL, but never written to the file. Ideal for decrypted data.
• New column next to the source column.
• Replace values in place (⌘Z undoes it).

Keys and passwords: run CSV Studio: Set Secret…. Secrets are stored in your OS keychain through VS Code's SecretStorage. They are never written to settings or files, and transforms refer to them only by name.

Your own JavaScript: run CSV Studio: New JavaScript Transform…. It creates .vscode/csv-studio/transforms/<name>.js, which you can keep in version control and share with your team:

export const label = 'Decrypt legacy card';

export default function (value, row, ctx) {
  if (!value) return value;
  const key = helpers.deriveKey(ctx.secret('legacy-key'), 'hex', 'aes-256-cbc');
  const iv = Buffer.from(row.iv, 'hex');          // IV stored in another column
  return helpers.aesDecrypt(Buffer.from(value, 'base64'), { algorithm: 'aes-256-cbc', key, iv }).toString('utf8');
}

• The function gets value, row (the cells keyed by column name) and ctx (rowIndex, secret(name), state), plus Node's crypto, Buffer and helpers.
• It runs in a separate context with no require, file system, network or eval, and with a time limit for each value.
• It is not a hard security sandbox. Transforms therefore only run in trusted workspaces, and each new or changed file must be approved once before it runs.
• Transforms that run synchronously can also be called from SQL as js_<name>(value).

Profile and data quality

• Profile shows, for the current view:
  • Counts: rows, empty (%), distinct.
  • Type match, with a shortcut to show only the invalid cells.
  • For numbers and dates: min, max, mean, median, P95, sum.
  • Text length range, and the top values with bars. Click a bar to filter to that value.

• Lint: ragged rows, unterminated quotes, blank lines, mixed line endings, duplicate or empty header names, and a trailing empty column. These appear in the Problems panel and as a banner.
• Lint & Repair shows a diff before it changes anything. It pads short rows, names extra columns, drops blank lines and fixes quoting.

More tools (Tools ▾ or the Command Palette)

• Edit: insert, duplicate and delete rows (several at once). Rename, insert, move and delete columns. Clear cells.
• Change delimiter / quoting / line endings: rewrites the whole file.
• Split file: one file per value of a column, or every N rows.
• Merge / append several CSV files, matching columns by name, with an optional source_file column.
• Compare with another CSV by key column(s): lists added, removed and changed rows, with the before and after value of each changed cell.
• Convert: CSV to JSON (optionally turning a.b columns into nested objects), YAML, Markdown, HTML or SQL; and JSON to CSV.
• Text editor: rainbow-colored columns, the column name on hover, and Row 12 · price in the status bar.
• Large files: Open Large CSV (Read-Only) reads files from disk, including those over VS Code's 50 MB extension limit. Filtering, SQL, transforms (as virtual columns) and export all still work.

Settings

Development

npm install
node test/fixtures/generate.mjs     # sales.csv (encrypted card column), customers.csv, broken.csv
npm run build                       # extension (esbuild) + webview (Vite)
npm test                            # unit tests (vitest)
npm run test:integration            # smoke tests inside VS Code
npm run package                     # build a .vsix

The fixtures' card column is encrypted with the demo key 000102…1e1f (see test/fixtures/generate.mjs). Save it as the secret demo-key, then try SELECT decrypt(card, 'demo-key') FROM data.

Layout: src/core holds the pure, unit-tested logic (parser, dialect, types, view, edit, SQL, transforms, export, compare). src/extension is the VS Code host. webview/ is the React UI.