Free · Private · Browser-based

Large Parquet File Viewer

Open a .parquet file, preview its rows and inspect its schema. Filter and convert the full file to CSV without uploading it.

Open files up to 5 GiB in the local viewer. Files opened in this local viewer stay in your browser unless you choose an upload tool.

  1. Choose one Parquet file
  2. Inspect the preview and filter rows
  3. Open Export to download CSV
View and query large files with SQL

Upload up to 5 GiB for SQL filters, sorting and summaries across the full file. Local viewing above keeps file contents in your browser.

Analyze with SQL

Checking analysis availability…

Preview the data

Inspect up to 100,000 rows while conversion reads the full file.

Inspect the schema

See column names, types, and compression details.

Convert the full file

Filters apply to every row, including rows beyond the preview.

Preview a large file without loading every row

This viewer reads slices of your local Parquet file and displays up to 100,000 rows. The file-size limit is 5 GiB, but a file with very large row groups or decoded pages can still exceed browser memory limits. Save such a file with smaller row groups before trying again.

A preview is only part of the table. Filters run across all source rows when you export CSV; sorting is available only when every row fits in the preview. Use Analyze with SQL above when you need server-side ordering or summaries across the complete file. That separate action sends your file for processing.

The Parquet file viewer also accepts files up to 5 GiB, with paged viewing and editing for supported layouts. Schema and decoded-page limits still apply. To change values and save a Parquet copy, check the Parquet editor and its supported schemas. Try the million-row Flights file in the sample Parquet datasets before opening your own larger file.

Other ways to work with your Parquet file

Prefer a script or command line? These local examples complete the same task with installed tools. Review their format and memory limitations before using them on your data.

View Parquet with Python

Run this Python script locally with Apache Arrow. The script reads your local file rather than sending it to this website.

Before you start

  • Install Python 3 and create an isolated environment using the commands below.
  • Work in a folder containing input.parquet. Change the filenames to your own paths and use a new output name to avoid overwriting a file.
  1. Save the Python code as task.py. On macOS/Linux run .venv/bin/python task.py; on Windows PowerShell run .venv\Scripts\python.exe task.py.
Set up Python on macOS/Linux
python3 -m venv .venv
.venv/bin/python -m pip install pyarrow

Set up Python on Windows PowerShell
py -m venv .venv
.venv\Scripts\python.exe -m pip install pyarrow

task.py
import pyarrow.parquet as pq

source = pq.ParquetFile("input.parquet")
print(source.schema_arrow)
print("Rows:", source.metadata.num_rows)
batch = next(source.iter_batches(batch_size=10), None)
print([] if batch is None else batch.to_pylist())

Limits and output

  • This prints the schema and at most ten rows. It does not modify the file or export the complete table.
  • A row preview is not a complete data-quality check. Compressed pages, wide values and nested columns can still require substantial memory.

Documentation: Apache Arrow Parquet documentation

View Parquet with DuckDB

Use the DuckDB command-line client to work directly with a local Parquet file.

Before you start

  • Install the DuckDB CLI using the official installation instructions, then confirm duckdb --version.
  • Work in a folder containing input.parquet. Change the filenames to your own paths and use a new output name to avoid overwriting a file.
Run in a terminal
duckdb -c "DESCRIBE SELECT * FROM read_parquet('input.parquet'); SELECT * FROM read_parquet('input.parquet') LIMIT 10;"

Limits and output

  • This prints the schema and at most ten rows. It does not modify the file or export the complete table.
  • A row preview is not a complete data-quality check. Compressed pages, wide values and nested columns can still require substantial memory.
  • SQL results have no guaranteed row order without ORDER BY. Choose a stable key when order matters.

Documentation: Install the DuckDB CLI; DuckDB Parquet documentation

View Parquet with ClickHouse

Run ClickHouse local on your computer; no ClickHouse server or cloud account is required for this local-file example.

Before you start

  • Install the ClickHouse binary for your operating system from the official instructions and confirm clickhouse local --version.
  • Work in a folder containing input.parquet. Change the filenames to your own paths and use a new output name to avoid overwriting a file.
Run in a terminal
clickhouse local --query "DESCRIBE TABLE file('input.parquet', Parquet); SELECT * FROM file('input.parquet', Parquet) LIMIT 10 FORMAT PrettyCompact;"

Limits and output

  • This prints the schema and at most ten rows. It does not modify the file or export the complete table.
  • A row preview is not a complete data-quality check. Compressed pages, wide values and nested columns can still require substantial memory.
  • The file path is local to clickhouse local. A remote ClickHouse server would read its own filesystem, not yours. SQL row order requires an explicit ORDER BY.

Documentation: Install ClickHouse; ClickHouse local documentation

Plain answers

Parquet viewer questions

How do I open a Parquet file?

Choose or drop a .parquet file up to 5 GiB. The viewer reads it locally and shows its schema and a preview of up to 100,000 rows. No installation or account is needed.

Is this Parquet reader private?

Local viewing and conversion do not upload your selected file. The separate Analyze with SQL workflow uploads it directly to private server storage; filenames, SQL, columns, and cell values are not recorded in operational telemetry.

What is the file-size limit?

Local input files can be up to 5 GiB. For server-side viewing, upload up to 5 GiB in the separate Analyze with SQL section and query the full table. The preview is limited to 100,000 rows. Very large column chunks, pages or nested row groups may require you to save the source with smaller row groups. CSV downloads are limited to 1 GB; supported browsers also offer Save CSV to disk for larger output.

Can I filter and sort Parquet rows?

Filters apply to the preview and to every row during CSV conversion. Sorting is available when the preview contains the whole file. It is disabled for partial previews so the download never claims to sort unseen rows.

Can I convert Parquet to CSV?

Yes. Open the file, optionally filter it, and select Export active view. Conversion reads the full file locally, including rows beyond the preview. When all rows fit the preview, sorting also carries into the download.

Which compression formats can the viewer read?

The browser parser includes uncompressed, Snappy, Gzip, Brotli, LZ4, LZ4 Raw, and Zstandard decompression support. Some unusual nested schemas may still display differently from flat tabular files.