Iceberg Doctor
Iceberg Doctor checks IDs and references across schemas, partition specs, sort orders, snapshots, refs, and the current manifest-list result. Findings include stable codes and JSON path/value evidence.
Iceberg inspection for macOS
Debug Iceberg tables.
Check metadata integrity, compare retained snapshots, and explain how a DataFusion SQL query affects Iceberg candidate files without executing it.
Core workflowsMetadata-first
- 01Iceberg DoctorCheck IDs and references
- 02Snapshot comparisonReview retained snapshots
- 03Query planAnalyze, then run SQL
# Investigate an Iceberg table.
Start with metadata and manifest statistics. Read data-file payloads only when the investigation requires them.
Snapshot comparison
Choose snapshot A and B. Compare reported summary metrics, top-level schema changes by stable field ID, and active manifest inventories. Missing counts stay unknown instead of becoming zero.
Query planning and execution
Write DataFusion SQL once. Visualize the physical plan, inspect Iceberg candidate files and deletes, review suggestions, then run the query when the plan looks right.
Layout and deletes
Review partition inventory and skew, evolved specs, file counts and bytes, exact or sampled size percentiles, small files, and active delete artifacts.
Metadata, files, and SQL
Browse metadata JSON, history, manifests, and manifest entries. Open Parquet row groups, codecs, encodings, bounds, and null counts, or run SQL with a configured query engine.
# Supported sources.
Open catalog tables or a selected local metadata.json file. Browse individual files in Amazon S3 or local directories.
- AWS Glue
- Catalog databases and Iceberg tables
- Amazon S3 Tables
- Buckets, namespaces, and managed Iceberg tables
- Iceberg REST
- Namespaces and Iceberg tables
- Cloudflare R2
- Iceberg tables through R2 Data Catalog
- Amazon S3
- Objects, metadata files, and data files
- Local
- Directories, data files, and individual metadata.json tables
Table format
Apache Iceberg
File inspection
Parquet · Avro · CSV · JSON · JSON Lines
# What Column Kit reads.
Read scope depends on the tool and stays visible in the workspace.
01 Iceberg Doctor checks the metadata document and the current manifest-list result already loaded by Column Kit. It does not verify object existence or read every manifest entry.
02 Snapshot comparison reads two manifest lists when available. Query-plan and layout analysis can read current manifest entries. These tools do not scan data-file payloads.
03 Query-plan evaluation uses Iceberg's metadata planner. Returned candidate tasks are capped and metadata planning has a time limit; the task cap does not bound manifest entries Iceberg may inspect. Partial totals remain lower bounds, and the result does not include per-stage manifest I/O counters.
04 Opening a file, previewing rows, or running SQL reads data only when requested. Table-property edits are separate, reviewed, and confirmed actions; Column Kit does not automatically rewrite or optimize tables.
# Availability.
Column Kit is in beta for macOS. It is free during beta. The installer is not available yet.
macOS installer coming soon- Version
- 0.1.0 beta
- Platform
- macOS
- Price
- Free during beta