Dev.to
7/25/2026

The original headline is: "ClickHouse 26.3 Removes Experimental Hypothesis Skip Index: Migration Guide"
Original: Day 99 - Efficient Random Sampling with system.numbers_mt: Parallel Number Generation in ClickHouse® 26.3
Short summary
ClickHouse 26.3 removes the experimental Hypothesis Skip Index (TYPE hypothesis), which precomputed whether boolean expressions could be true within data granules. Tables using this index must be updated before upgrading or schema restoration will fail. The article explains how skip indexes work, why the feature was removed, and provides migration steps including ALTER TABLE DROP INDEX.
- •ClickHouse 26.3 removes experimental Hypothesis Skip Index — schemas must be updated before upgrading
- •The index stored three-state metadata per granule but never matured beyond experimental status
- •Migration: use SHOW CREATE TABLE to find affected tables, then DROP INDEX before upgrade
Generated with AI, which can make mistakes.
Is this a good recommendation for you?



