Open a Parquet file online

See the schema, the statistics and the rows of an Apache Parquet file (.parquet) in your browser, page by page, even when it holds millions of rows. The file is read on your device and never uploaded.

Drop your .parquet file here

Your file stays on this device: it is read here, in your browser, and never uploaded.

Apache Parquet files of any size: only the parts you look at are read.

How it works

Choose the .parquet file
Drop it on the box above or choose it. Only its footer is read first (the schema and the statistics), then just the rows you look at, so even files of many gigabytes open in seconds. Nothing is sent anywhere.
Look at the data
Page through the rows, 100 or 1,000 at a time, jump to any row, pick the columns to show and search the rows on the page. The schema tab lists every column with its type and statistics.
Export it if you need to
Viewing is free. Every row as CSV, Excel or JSON Lines, with the columns you choose, is a paid export: the options and their prices are shown next to the export button, before you pay anything.

Parquet files, explained

What a Parquet file is

Apache Parquet is a file format for tables of data. Instead of storing one row after another like a CSV file, it stores each column together, compressed, in blocks of rows called row groups, with a footer that describes the columns and their types. That makes it small and fast for data tools, and it is why data engineering and analytics software uses it everywhere: Apache Spark, pandas and Polars, DuckDB, Snowflake, BigQuery, Databricks and the big cloud platforms.

You usually receive one because someone exported data for you: a cloud billing or usage report, a database or data-warehouse export, a dataset from an open-data portal or a machine-learning collection, or a file a data team shared instead of a spreadsheet.

Why it won’t open

A Parquet file is binary and compressed, so double-clicking it does nothing useful and a text editor shows unreadable characters. Excel and Google Sheets cannot open a .parquet file as it is. Reading one normally takes code (Python with pandas, DuckDB, Spark) or a database tool.

What this viewer does

Reads the file in pieces: the footer first, then only the row groups that the page on screen needs. The file is never loaded whole.
Shows the schema: every column with its type (text, integers, decimals with their precision, dates, timestamps with their unit and time zone, lists, structs, maps), whether it can be empty, its compression and its size.
Shows the statistics the file stores for each column (lowest and highest value, number of empty values) without reading any rows.
Lists the file’s details: number of rows and row groups, compression (Snappy, Gzip, Zstandard, Brotli, LZ4), the program that wrote it, the format version and the key-value metadata.
Shows the rows 100 or 1,000 at a time, on any page of the file, with the columns you choose, and searches the rows on the page.
Exports every row, or the columns you choose, as CSV, Excel or JSON Lines, optionally split into parts (paid).

How the values are shown

The viewer keeps every digit the file holds:

Decimals are shown and exported exactly as stored, never rounded through floating-point numbers.
Timestamps are written in ISO 8601 with all the digits of their unit (milliseconds, microseconds or nanoseconds), with “Z” when the column says the values are in UTC. Old INT96 timestamps, as written by Spark and Hive, are read to the nanosecond and shown without a time zone, as the file stores them.
Dates are written as year-month-day.
Lists, structs and maps are written as JSON in one cell. Binary values are shown as base64, unless they are plain text. GeoParquet geometries are shown as WKT.
In the exports: CSV files are UTF-8 with a byte-order mark, so Excel reads accents correctly; a line break inside a value becomes a space, and a value that starts with =, +, - or @ gets an apostrophe in front, so a spreadsheet cannot run it as a formula. Excel sheets hold 1,048,575 rows each, so a longer export continues on a new sheet, and an Excel cell holds 32,767 characters. JSON Lines keeps lists and structs as JSON values.

What it does not do

It does not change the file or write Parquet files.
It does not run SQL queries or join several files.
It does not read encrypted Parquet files or the old LZO compression.
A single column of one row group larger than about 1.5 GB (uncompressed) is too large for a browser tab.
An export stops at about 4 GB (1 GB in Safari and on iPhone and iPad) and keeps the rows written up to there. Exporting fewer columns makes the file smaller.

Other ways to open a Parquet file

DuckDB, free: SELECT * FROM 'file.parquet' in its command-line tool or any DuckDB client.
Python with pandas: pandas.read_parquet("file.parquet"), then to_csv or to_excel.
Ask the sender for a CSV or Excel export if you only need to read a small table.

Your data stays on your device

Data exports are often confidential: customer records, sales, logs, costs. Here the file is read by code running in your browser; no code on the page sends the file, its name or anything in it anywhere, not even to our own domain. The exports are made in the page too.

Apache Parquet, Apache Spark and Apache Arrow are trademarks of the Apache Software Foundation. Microsoft and Excel are trademarks of Microsoft Corporation. Other names belong to their owners and are used only to describe what the file works with. CantOpen is not affiliated with or endorsed by any of them.

Frequently asked questions

How do I open a Parquet file without writing code?

Drop it on this page: the schema, the statistics and the rows appear in your browser, with nothing to install and no account. It works on Windows, Mac, Chromebook and tablets.

Is it free?

Viewing is free: the schema, the statistics, the file details and every row, page by page. Exporting the rows to CSV, Excel or JSON Lines is paid; the options and their prices are shown next to the export button.

Can I convert a Parquet file to CSV or Excel?

Yes. Choose CSV, Excel or JSON Lines and the columns you want. A CSV or JSON Lines export can be split into parts of 100,000 or 1,000,000 rows, in one ZIP; an Excel export continues on a new sheet every 1,048,575 rows.

Is my file uploaded?

No. It is read by code running in your browser and never leaves your device. No code on the page sends the file, its name or its contents anywhere, not even to our own domain.

Which Parquet files does it read?

Files written by Spark, pandas, Polars, DuckDB, Arrow and other tools, with Snappy, Gzip, Zstandard, Brotli, LZ4 or no compression, nested columns, dictionary and delta encodings, and old INT96 timestamps. Encrypted files and LZO compression are not read.

Can it open very large files?

Yes. Only the footer and the row groups you look at are read, so files of many gigabytes and millions of rows open quickly. Exports read the file one row group at a time.

Why do some timestamps end in “Z”?

The “Z” means the file stores the time in UTC. Timestamps without it are stored without a time zone, as local times, and are shown exactly as written.