Skip to content

Open large CSV and Parquet files

Use vastgrid when you want to inspect a large table with spreadsheet controls: scroll through rows, sort a column, filter records, or select cells to see their sum. It is free and MIT licensed.

Try 1,000,000 rows now — no account, install or file upload.

What you need Start here Where the data goes
Inspect a file without installing anything Open in the browser Stays in your browser; no upload
Work with files that exceed browser memory npx vastgrid-cli sales.parquet DuckDB on your computer
Stay in your editor Install the VS Code extension DuckDB in the extension
Add the grid to your own app React quick start Your choice of browser or server engine

The browser also reads TSV, JSON Lines and Excel (.xlsx, .xlsm). DuckDB databases open through the CLI, VS Code and the other DuckDB-backed apps. See the format and save comparison.

  1. Open the sample and wait for the status bar to show 1,000,000 rows.
  2. Open the unit_price column menu (▾) and choose Sort descending.
  3. Drag across a range of prices. The status bar shows the selection’s count, sum and average.
  4. Use Open another file to try your own CSV or Parquet file.

The sample generates a CSV in a Web Worker and reads it with vastgrid-js, the memory engine. It does not download a million-row dataset. The 200,000-row orders demo is another way to explore editing, grouping and formulas.

Files opened here are not uploaded. Edits stay in the page until you export them; closing or reloading the page loses unsaved edits. Use the toolbar to export CSV, TSV, JSON or Excel. The browser does not overwrite your original file.

With Node.js 22.18 or later installed, run:

Terminal window
npx vastgrid-cli sales.parquet

Replace sales.parquet with your file’s path. For example:

Terminal window
npx vastgrid-cli ./data/orders.csv

The command opens the grid in your browser while DuckDB reads the file on your computer. For a filename with spaces, put the path in quotes. See the CLI guide for options, exports and saving CSV/TSV edits.

If you already use VS Code, install vastgrid and open your file in its grid editor.

The two engines have different limits:

  • Browser: memory and operation costs depend on the file, columns, device and browser. Parquet reads row groups on demand; sorting or grouping can still need more memory than browsing. If an operation exceeds the memory limit, use the CLI or VS Code. See browser engine limits.
  • DuckDB: can use temporary disk space for larger workloads. In the published M2 Pro benchmark, a 100-million-row Parquet file (1.38 GB) returned its first sorted rows in 953 ms, with the rest of the view built 3 seconds later. These are engine timings, excluding HTTP and rendering; they are not a promise for every file or device.

The million-row browser sample and the 100-million-row DuckDB benchmark are different workloads. Measurements and reproduction steps document both engines.

If vastgrid is useful for your next data task, star it on GitHub to find it again. If a file does not open, report its format, size and the error. Use a small synthetic example when the original data is private.