1. Zero-Setup Schema Inspection
Verify column names, physical types (INT64, DOUBLE, BYTE_ARRAY), logical types, and row-group counts in seconds without spinning up Python, Spark, or DuckDB.
THE TOOLBOX / DATA & ANALYTICS
Drop any Apache Parquet (.parquet) dataset to inspect its Thrift schema, row groups, and tabular data right in your browser—then filter and export to CSV or JSON with zero uploads.
| # |
|---|
PARQUET VS. CSV & JSON
| Format | Layout | Built-In Schema | Typical Size vs. CSV |
|---|---|---|---|
| Apache Parquet (.parquet) | Columnar Binary | Yes (strict typed footer) | 10% – 25% |
| CSV (.csv) | Row-Oriented Text | No (all strings) | 100% (baseline) |
| JSON / NDJSON | Row-Oriented Text | Implicit per record | 140% – 220% |
Verify column names, physical types (INT64, DOUBLE, BYTE_ARRAY), logical types, and row-group counts in seconds without spinning up Python, Spark, or DuckDB.
Filter rows across all columns and click Export CSV to hand clean slices directly to Excel or Google Sheets.
Unpack compressed tarballs in our TAR.GZ Extractor, validate JSON schemas in the JSON Schema Validator, or convert YAML/TOML in our YAML to JSON Converter.
Drop any .parquet file into the Parquet viewer above. It parses the Thrift footer metadata, column schema, row groups, and tabular data directly inside your browser so you can filter rows and inspect types instantly.
Yes. Load your .parquet file (and optionally filter rows using the search box), then click "Export CSV" or "Export JSON" to download a standard spreadsheet or JSON array.
Never. All Parquet footer parsing, decompression, filtering, and CSV/JSON conversion execute 100% locally inside your browser tab.
Parquet stores data by column rather than by row, allowing dictionary encoding, run-length encoding, and Snappy/GZIP compression per column—typically shrinking file size by 70%–90% compared to plain CSV while preserving strict data types.
To keep browser rendering fast and responsive, the viewer reads up to the first 500 rows for instant inspection and export while still displaying the exact full row count and schema from the Parquet footer.
Standard physical and logical types including INT32, INT64, FLOAT, DOUBLE, BOOLEAN, BYTE_ARRAY (UTF-8 strings), timestamps, and dates are decoded and displayed cleanly.