Parquet Viewer
Reads a Parquet file directly in your browser using hyparquet, a pure-JavaScript parser (no WebAssembly). Row count and full column schema appear immediately from the file's own metadata, with types translated from Parquet's internal type codes into plain labels (String, Integer, Number, Boolean, Timestamp, ...). Browse the data 50 rows at a time rather than loading an entire file into the page at once. Every standard compression codec is supported — Snappy natively, plus GZip, Brotli, ZSTD, LZ4, and LZO via hyparquet-compressors. Runs entirely client-side — nothing is uploaded.
About this tool
Reads a Parquet file — the columnar storage format used throughout data engineering, data science, and ML pipelines — directly in your browser, using hyparquet, a pure-JavaScript parser with no WebAssembly and no server involved. Choose a file and the row count and full column schema appear immediately, read straight from the file's own metadata footer without needing to scan the actual data. Browsing the data pages through 50 rows at a time rather than loading an entire (potentially huge) file into the page at once. Compression is fully covered, not just the common case: hyparquet handles Snappy natively, and this tool adds hyparquet-compressors on top for GZip, Brotli, ZSTD, LZ4, and LZO — the full set of codecs a real-world Parquet file might use, so "this file uses a codec we don't support" isn't a failure mode here. Column types are translated from Parquet's internal type system (BYTE_ARRAY with a UTF8 annotation, INT64 with a TIMESTAMP_MILLIS annotation, and so on) into plain labels like String, Integer, and Timestamp.
What you can do
- Reads real Parquet files via hyparquet — a pure-JS parser, no WASM, no upload
- Every compression codec Parquet uses — Snappy, GZip, Brotli, ZSTD, LZ4, LZO — not just the common one
- Schema shown as readable column types (String, Integer, Timestamp, ...), not raw Parquet type codes
- Pages through large files 50 rows at a time instead of loading everything into the page at once
How to use
Step 1
Choose a .parquet file
Row count and schema appear immediately from the file's own metadata.
Step 2
Browse the data
The first 50 rows load automatically — page forward to see more.
Step 3
Check the schema
Every column's real type, translated from Parquet's internal type codes.
Example
A Parquet file's schema shows up as a simple table:
Schema
id Integer name String score Number active Boolean
These are hyparquet's own type detections, translated to plain labels — a BYTE_ARRAY column annotated UTF8 shows as "String", not the raw Parquet type code.
How your data is handled
This tool is labeled 🟢 Runs in your browser: hyparquet parses the file entirely client-side — nothing is uploaded to a server, logged, or retained once you leave the page. The whole file is read into your browser's memory once you choose it, and discarded when you navigate away or refresh.
Frequently asked questions
Which compression codecs are supported?+
All of them — Snappy comes from hyparquet itself, and GZip, Brotli, ZSTD, LZ4, and LZO come from the hyparquet-compressors package this tool also includes. A Parquet file using any standard codec should open without issue.
Why does it only show 50 rows at a time?+
Parquet files can be very large — some far bigger than a browser tab could comfortably hold as fully-rendered HTML rows. Paging 50 at a time keeps browsing fast and memory-light regardless of the file's total size, while the row count and schema (read from the file's metadata, not the data itself) still show the true totals immediately.
Does this support nested/struct columns?+
This version shows top-level columns as a flat table, matching how a CSV-style view naturally works. Deeply nested struct, list, or map columns aren't expanded — a reasonable next step if you need to inspect one specifically.
Is my file uploaded anywhere?+
No — hyparquet parses the file entirely in your browser. Nothing about its contents is ever sent to a server.
What if my file uses GeoParquet geometry columns?+
hyparquet detects GeoParquet geometry columns and decodes them to GeoJSON automatically; this tool currently displays that as its raw decoded value rather than rendering it on a map — for a map view of geospatial data, try the GeoJSON / KML / GPX / CSV Viewer once you've exported the geometry.