Parquet & Delta Lake Workbench

Data, Parquet metadata and Delta history in one place

The workbench has three views. Data is the spreadsheet editor. Parquet details shows the schema, row groups, column chunks, codecs, encodings, sizes, statistics and file-level key-value metadata. Delta history shows each transaction version, operation, timestamp, add/remove counts, protocol or metadata changes and the resulting active-file count.

Edit Parquet coherently

Edit cells, add or remove rows and columns, rename columns, change types, or paste a rectangular block from Excel, LibreOffice or Google Sheets. File-level key-value metadata can also be edited. Physical metadata such as row-group offsets and statistics is deliberately not directly editable: it is regenerated from the rewritten Parquet file so the result remains internally consistent.

Delta Lake history stays immutable

A Delta Lake table is a folder containing Parquet files and a _delta_log. Existing log versions are shown read-only. You can inspect their raw actions, load the snapshot from an older version into the data editor, compare that version with the latest snapshot, or create a new RESTORE commit. Editing current data or table properties and downloading Delta appends a new transaction version while retaining the original log and data files in the ZIP.

What a Delta edit writes

When an existing Delta table is edited, the workbench writes a new Parquet file, adds remove actions for the previously active files, emits updated table metadata when needed, and appends a new Delta JSON log version. A new table still starts at version 0. Existing history is never rewritten.

Limits

• The editor loads up to 200,000 active rows at once to avoid exhausting browser memory.
• Delta import currently needs JSON transaction history starting at version 0. Checkpoint-only histories whose earlier JSON logs were removed are not opened yet.
• Restoring an older version requires its referenced Parquet files to still be present in the selected folder or ZIP.
• Edited Delta exports are rewritten as an unpartitioned current snapshot, while the previous transaction history and files remain available for time travel.

Are my files uploaded?

No. Parquet and Delta inspection and editing runs with JavaScript and WebAssembly running in your browser tab. Your files are read from your device and the results are created on your device; nothing is sent to gratistools.be or anyone else, which is also why there are no queues, daily limits or watermarks. See the privacy page.