Smarter deployments in Tinybird: faster schema changes for the most common operations

Tinybird has shipped ALTER TABLE support in its deployment engine, replacing the previous brute-force approach of full table rebuilds for every schema change. Previously, even simple operations like adding a column triggered a create-backfill-swap cycle that could take minutes on large tables. Now, the deployment engine classifies each change into four categories (NoChange, MetadataOnly, AlterOperation, Schema) and applies in-place ALTER TABLE statements for supported operations: TTL changes, column additions/drops, nullable modifications, default value changes, and index changes. A deploy that previously took minutes on a 200GB table now completes in under 5 seconds. The system includes safety checks for pending mutations, column droppability, and downstream impact, plus rollback query generation. Materialized Views also gain a DEPLOYMENT_METHOD 'alter' directive to apply query changes via ALTER TABLE MODIFY QUERY instead of full recreation.