dataeng.tools

Parquet metadata viewer

Your files never leave your browser

Drop a .parquet file to inspect its footer: schema and types, row groups, compression codec and encodings, per-column statistics (min/max, nulls) and embedded key–value metadata. It reads only the metadata — not your data — entirely in your browser (DuckDB-WASM). Nothing is uploaded.

Drop a file or click to browse

Parquet only — read locally in your browser, never uploaded

Frequently asked questions

What does the Parquet inspector show?
The file's writer (created_by), row and row-group counts, format version, per-row-group sizes, and per-column type, compression codec, encodings, null counts, min/max statistics, and any embedded key–value metadata (e.g. ARROW:schema or pandas).
Is my file uploaded anywhere?
No. Everything runs in your browser using DuckDB-WASM. The inspector reads only the Parquet footer, so it never even scans your data — and nothing leaves your device.
How is this different from parquet-tools?
It gives you the same information as parquet-tools meta / schema, but with nothing to install — no Python, Java, or command line. Drop a file and read its metadata in the browser.
Does it work on large files?
Yes. Because it reads only the footer (not the row data), inspecting a multi-gigabyte Parquet file is effectively instant.
What are row groups and why do they matter?
A Parquet file is split into row groups — independent horizontal chunks, each holding column chunks with their own statistics. Query engines skip whole row groups whose min/max stats can't match a filter, so row-group sizing and column ordering directly affect scan performance.