Open and query Parquet files on your desktop

Parquet is a binary, columnar format, so a text editor or a spreadsheet cannot show it. DuckLocal opens .parquet files as tables you can browse, filter, query with SQL and export. Everything runs on your computer, and the file is never uploaded to an online viewer.

Free and open source for macOS, Windows and Linux. No account. No telemetry.

From a .parquet file to an answer

No Python environment, notebook or cluster needed to look inside a file.

  1. Open the file

    Drop a file onto the window, open a whole folder, or use a pattern such as ./data/2026-0*/*.parquet. On a Mac, double-click the file in Finder. Each file becomes a view named after it.

  2. Look at what is in it

    The schema sidebar lists every column and its type. The Overview tab summarizes each column of a result: nulls, distinct values, range and distribution.

  3. Query, chart and export

    Filter and aggregate with DuckDB SQL, switch the result to a chart, and export the complete result to CSV or Parquet.

Inspect the file itself

A Parquet file carries its own schema, row groups, compression and column statistics in its footer. DuckDB reads them with plain SQL, so you can check how a file was written before you scan a single row.

Data sources · Results, overview and charts

-- columns and types
DESCRIBE SELECT * FROM 'trips.parquet';

-- row groups, compression and statistics
SELECT * FROM parquet_metadata('trips.parquet');

Preview without opening anything

On macOS, select a Parquet file in Finder and press Space. The Quick Look preview shows the columns and their types, the first 50 rows, and from the footer the row count, row groups, compression and the writer.

The preview ships inside DuckLocal.app. It never scans the whole file and downloads nothing. More about Quick Look previews

  • Your file is read where it is and is not uploaded
  • Queries read only the columns they need
  • The source file is never changed
  • No account and no telemetry
  • Open source under Apache-2.0

One file, or thousands

Parquet data rarely comes as a single file. Open it the way it is stored.

  • *.parquetOpen every file that matches a pattern
  • folder/Walk a directory recursively
  • read_parquet()Read many files as one table in SQL
  • https://Query a remote Parquet file over HTTPS
  • s3://Buckets and S3-compatible storage, when you configure it
  • Export → CSVTurn a query result into CSV
  • Export → ParquetWrite CSV, JSON or Excel data as Parquet
  • JOINCombine Parquet with CSV, JSON, Excel and DuckDB tables

Made for files a spreadsheet cannot open

The results grid shows up to 100,000 rows of a result, and says so when it stops. Aggregations, filters and joins still run over the whole file, and an export reruns the query and writes every row.

So the usual workflow is to let SQL reduce millions of rows to the answer you need, rather than scrolling through them.

Parquet viewer questions

How do I open a Parquet file on a Mac?
Install DuckLocal and double-click the .parquet file in Finder, or drop it onto the DuckLocal window. To peek without opening it, select the file and press Space for a Quick Look preview. DuckLocal needs macOS 12 or later on Apple silicon.
How do I open a Parquet file on Windows or Linux?
Start DuckLocal and drop the file onto the window, or choose it from the Open menu. From a terminal, ducklocal ./file.parquet opens it directly.
Can I convert Parquet to CSV?
Yes. Query the file and choose Export → CSV; the export reruns the query and writes the complete result. It works the other way too: open a CSV file and export it as Parquet.
How large a Parquet file can DuckLocal open?
DuckDB reads Parquet column by column and only the parts a query needs, so files with many millions of rows are practical on a laptop. The grid displays up to 100,000 rows of a result, while aggregations run over the full file.
Is my Parquet file uploaded anywhere?
No. DuckLocal is a desktop app with DuckDB embedded. It reads the file from its path on your computer, requires no account and sends no telemetry.

Open your first Parquet file

Download DuckLocal and drop a .parquet file onto the window. The columns appear before you write any SQL.