Iceberg Metadata Explainer

Write rows to a partitioned table — date is bucketed by month, state by hash into 10 buckets — and watch metadata evolve. Runs entirely in your browser.

in-memory catalog in-memory S3 icebird (JS)

Table Data

date mo state b10 value del

Snapshot Timeline

Each write creates a new Iceberg snapshot. Scrub the slider to time-travel.

Metadata Structure


        


        

          
Engine scan at the selected snapshot — position deletes applied (merge-on-read).

          

        


        

          

        


        

          

          
The manifest list is an Avro file. Each row points at a manifest file.

          

        


        

          
Manifest files list data files and their partition values.

          

        


        

          
Actual Parquet data files — row contents and footer statistics.