Apache Iceberg Go 0.5.0 Release

The Apache Iceberg community is pleased to announce version 0.5.0 of iceberg-go.

This release covers approximately five months of development since the 0.4.0 release in October 2025 and is the result of merging over 110 PRs from 31 contributors, including 18 first-time contributors. See the full changelog for the complete list of changes.

iceberg-go is a native Go implementation of the Apache Iceberg table format, providing libraries for reading, writing, and managing Iceberg tables in Go applications.

Release Highlights

Iceberg V3 Table Spec Support

This release significantly advances iceberg-go's implementation of the Iceberg V3 table specification:

• Table Metadata V3: Full V3 metadata format support is now implemented
• Unknown Type: Support for the Unknown type introduced in the V3 spec
• TimestampNano and TimestampNanoTz: New timestamp nanosecond types required by V3 are now supported
• Row Lineage Tracking: Row lineage metadata is now tracked and the snapshot producer wires V3 row-lineage state correctly, with a fix for first-row-id assignment in V3 manifest lists
• Source IDs: Support for source-ids metadata was added

View Support

iceberg-go now has substantial support for Iceberg views:

• View Spec: Initial implementation of the View specification including serialization, deserialization, and JSON validation
• View Metadata Builder: Full view metadata builder and UpdateView method in the REST catalog
• Load View: Load existing views from catalogs
• Hive Catalog View Operations: View create, load, drop, and list support in the Hive catalog
• RCK Compatibility: Field ID preservation and error handling for REST Catalog Kit compatibility

Schema Evolution

• Schema evolution: Full schema evolution support was added, enabling AddColumn, RenameColumn, UpdateColumnType, and related operations
• HighestFieldID fixes: Corrected HighestFieldID to include the List element ID and fixed a related bug

Delete File Support

iceberg-go now supports both strategies for deleting rows from Iceberg tables:

• Copy-on-Write deletes: Implemented copy-on-write delete mode for rewriting data files on delete
• Merge-on-Read deletes: Implemented merge-on-read delete mode for writing position and equality delete files
• Table overwrite: General table overwrite functionality with a clean options-based API
• Add/replace data files: Functions to directly add or replace DataFiles in a table

Hive Catalog

• Initial Hive Catalog support: Full Hive metastore catalog was added, supporting table creation, listing, loading, and dropping
• View operations in Hive: View support was added alongside the table operations

Puffin File Format

A Puffin reader and writer was implemented, enabling iceberg-go to read and write Puffin statistics files used for table statistics and NDV sketches.

Catalog Improvements

• AuthManager support: Pluggable authentication manager for the REST catalog, enabling custom credential flows
• WithHeaders: A WithHeaders catalog option allows injecting custom HTTP headers into REST catalog requests
• Staged table creation: Support for staged table creation in the REST catalog
• ListNamespaces pagination: Pagination support was added to ListNamespaces
• RegisterTable fix: Corrected the RegisterTable REST endpoint
• renameTable 204 handling: Properly handle 204 No Content responses on renameTable

Write Performance

• Partitioned write throughput: Optimized partitioned write throughput by reducing allocations in the hot path
• Gzip metadata: Support for both reading and writing gzip-compressed metadata JSON files

IO Improvements

• IO registry pattern: Refactored IO scheme registration to use a registry pattern, making it easier to add custom IO backends
• S3 signing: Allow s3.signer.uri property and only reject explicit s3.remote-signing-enabled=true; also set x-amz-content-sha256 before signing
• Azure SAS tokens: Updated SAS token behaviour to match the Java implementation

Bug Fixes

Notable bug fixes in this release include:

• Fix truncate transform for negative integers
• Fix data race in yieldDataFiles
• Remove partition type cache causing concurrent map write panic
• Preserve snapshot log history
• Handle missing parent snapshots in ExpireSnapshots
• Add AssertRefSnapshotID requirements to ExpireSnapshots
• Support INT32/INT64 physical types for decimal columns
• Fix decimal literal MarshalBinary
• Fix index out of range in buildManifestEvaluator
• Fix response body leaks in the REST catalog
• Close writers on error for every exit path
• Fix findMinSeq from manifest list

Breaking Changes

Some of these changes are breaking changes that need to be called out:

• You now need to add import _ "github.com/apache/iceberg-go/io/gocloud" to register the storage implementations for S3, GCS, Azure and in-memory blobs. They are no longer registered by default, only the local filesystem implementation exists by default.
• io.ParseAWSConfig is now in the gocloud package, though it wasn't expected that this would necessarily be used by many consumers anyways.
• The table.Metadata interface has two new methods PartitionSpecByID and NextRowID, any packages that implemented their own Metadata object to meet the interface would need to add these methods. Though it is not expected that consumers would do this.
• The rest.Catalog.CreateView function signature has changed (#629)
• The fields of catalog.CreateTableCfg have been moved into an embedded commonCreateCfg struct (#754). Field access should still work via the inherited properties, but you should use NewCreateTableCfg() and the With* methods rather than attempting to construct this directly.

New Contributors

Welcome to all 18 first-time contributors: @hamilton-earthscope, @itaigilo, @ankur-anand, @0xMALVEE, @el10savio, @rambleraptor, @vjanelle, @hdnpth, @dhananjaykrutika, @hackintoshrao, @csumisha, @ErenDursun, @dontirun, @alexandre-normand, @agaddis02, @mkuznets, @lovromazgon, @zhjwpku

Getting Involved

The iceberg-go project welcomes contributions. We use GitHub issues for tracking work and the Apache Iceberg Community Slack for discussions.

The easiest way to get started is to:

1. Try iceberg-go with your workloads and report any issues you encounter
2. Review the contributor guide
3. Look for good first issues

For more information, visit the iceberg-go repository.