TRACER RECART · IN-BROWSER DATA TOOL

Inspect, query and edit Parquet in your browser.

Open a local file, an HTTPS URL, or a public/presigned S3-compatible object. Browse large datasets with virtual loading, edit cells, run SQL, inspect internals and export CSV, JSON or Parquet without sending the file through our servers.

✓ In-browser processing✓ No upload proxy✓ No account✓ No artificial row limit

⇩

Open a Parquet source

Choose a local file, enter an HTTPS URL, or use an S3/R2 presigned URL.

Drop a .parquet file here or choose one from this device.

● Local files stay on this device. Remote files are fetched directly from their source by your browser and never pass through TRACER RECART or a backend proxy.

Reading Parquet footer and metadata…

Compressed storage by column

Aggregated across all row groups using column-chunk metadata.

File metadata

#	Name	Physical type	Logical type	Repetition	Field ID

Row group	Rows	Uncompressed	Compressed	Ratio	Column chunks

Row group	Column path	Physical type	Codec	Encodings	Values	Compressed	Uncompressed	Min	Max	Nulls

First 50 rows

Preview decoding happens in this browser and may take longer than metadata inspection on very large or heavily-compressed files.

Raw metadata

How to inspect a Parquet source

Choose a local .parquet file, enter a public HTTPS URL, or paste an S3-compatible presigned URL.
For remote sources, the browser first verifies CORS and efficient byte-range support. Unsupported servers are rejected instead of triggering a hidden full-file download.
Browse rows, inspect the schema and row groups, run SQL, edit supported cells, and export CSV, JSON or Parquet in this tab.

Guides: Viewing large Parquet files · Understanding row groups and column statistics

Why inspect Parquet internals?

A Parquet file is more than a table. Its physical layout affects scan speed, compression, predicate pushdown and storage cost. The data workspace adds virtual browsing, in-browser editing and SQL without requiring Python, DuckDB Desktop or an upload service.

Row groups

Horizontal partitions of rows. Their size influences parallelism, pruning granularity and I/O behavior.

Column chunks

Each row group stores every column separately. Chunk metadata reveals codec, encoding and byte footprint.

Statistics

Min, max and null counts can let engines skip data. Missing or weak statistics can reduce pruning efficiency.

Privacy by architecture

Local files are accessed through the browser File API. Remote URL and S3-compatible sources are requested directly by your browser with byte ranges. Parsing, queries, editing and exports run in JavaScript/WebAssembly in this tab. TRACER RECART has no file upload or remote-file proxy endpoint, and presigned URLs are not stored.