I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

• Add a stable bucket(count, value) partition transform using SipHash.
• Add WRITE_DISTRIBUTION_MODE = 'hash' for partitioned Fuse tables.
• Normalize distributed write input with TableWritePrepare, evaluate the partition expressions, and route equal partition values through GlobalShuffle before distributed INSERT SELECT writers.
• Add a common node-local Fuse write layout before table.append_data(): evaluate partition expressions, full-sort by partition values, and rebuild blocks using the table's max_rows_per_block.
• Reuse the existing Fuse partition splitter, block and segment serialization, statistics, compaction, and commit paths without introducing a new buffering or spill processor.

The goal is to reduce fragmented blocks when unordered input repeatedly crosses physical partition boundaries, while preserving the existing guarantee that a block or segment never spans physical partitions.

Current scope

• Distributed INSERT SELECT uses GlobalShuffle to route equal evaluated partition values to the same writer node.
• INSERT VALUES, streaming insert, COPY INTO TABLE, local INSERT SELECT, and other common append paths reuse the node-local full-sort write layout.
• UPDATE, REPLACE, MERGE, and multi-table insert do not currently receive the new cross-node hash routing; they retain the existing partition-boundary guarantees.
• Non-Fuse tables and Fuse tables without WRITE_DISTRIBUTION_MODE = 'hash' keep their existing write behavior.
• bucket(count, value) controls persistent physical partition cardinality. It does not configure execution nodes or pipeline lanes.
• Hash distribution reduces partition tails but does not guarantee exactly one block or segment per partition.

Tests

• Unit Test
• Logic Test
• Benchmark Test
• No Test

Type of change

• Bug Fix (non-breaking change which fixes an issue)
• New Feature (non-breaking change which adds functionality)
• Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• Documentation Update
• Refactoring
• Performance Improvement
• Other (please describe):

AI assistance

• AI usage: An AI coding agent assisted with implementation refactoring, CI diagnosis, and logic tests; the final changes were reviewed by the responsible human.
• Responsible human: @KKould
• The responsible human has read every line of this diff and can explain each change

This change is