Parquet Reader: View .parquet Files Online

Open Apache Parquet files and browse their rows as a paginated table.

About Parquet Reader

Drop a .parquet file and the page parses it locally with the hyparquet decoder, then renders the contents as a paginated grid. Row count, file size, and column count appear in the header. An Export as JSON button writes every row to a downloadable .json file so you can move the data into another tool.

Apache Parquet is a columnar storage format used across data lakes, Spark jobs, and pandas pipelines. Each file packs row groups, column chunks, and footer metadata into a single binary blob optimized for analytical scans. Inspecting one without a script normally means firing up Python or DuckDB, which is overkill when you just want to see what's inside.

This reader uses hyparquet, a pure-JavaScript Parquet decoder. The file is parsed in the browser tab. Each row is decoded into a plain object and the result is paginated 50 rows at a time so the UI stays responsive even for large files. Columns inferred from the first row drive the table header, so nested types render as JSON strings rather than being dropped.

The Export as JSON button serializes the full result set. BigInt fields are converted to decimal strings and byte arrays show as length tags, both of which JSON can round-trip safely. Because everything runs locally, the same workflow handles confidential extracts without any upload step.

How to use the Parquet Reader
  1. 1

    Open the file

    Pick a .parquet file. The decoder parses footer metadata, row groups, and column pages directly in your browser.

  2. 2

    Browse rows

    Rows render as a table with 50 entries per page. Prev and Next move through the set; row counts and column counts stay visible.

  3. 3

    Export if needed

    Click Export as JSON to download every row as a JSON file with BigInts converted to strings for portability.

Common use cases

Peek inside a data lake extract

Verify what columns and row counts are in a Parquet file pulled from S3 or HDFS before loading it into a notebook.

Inspect a Spark or pandas output

A pipeline wrote a .parquet partition. Open one of the part files to confirm the rows look right before retraining a model.

Hand-share data with a non-engineer

Send a Parquet file plus this URL so a colleague can browse the data without installing pandas or DuckDB.

Convert for another tool

Export the table as JSON to feed it into a tool that doesn't speak Parquet natively, such as a no-code dashboard.

Frequently asked questions
Is my Parquet file uploaded anywhere?

No. hyparquet runs as a JavaScript module in your tab. The file bytes never leave the browser and there is no server-side component.

What compression codecs are supported?

Snappy, GZIP, ZSTD, and uncompressed pages all decode in the browser. Brotli is supported as well. Files written by Spark, pandas, DuckDB, and Arrow open cleanly.

How large a file can I open?

Browsers cap heap memory around 2 to 4 GB depending on the build. Files up to a few hundred megabytes open quickly. Beyond that you may want to filter the columns or row groups before reading.

Why are some columns shown as JSON text?

Lists, structs, and maps are nested. The viewer serializes them inline so the cell content remains readable. The exported JSON keeps the original nested shape.

Does it support encrypted Parquet?

Parquet Modular Encryption is not supported. The reader handles the standard open format only.

developerconverterformatter