Schemity Tests Round-Trip Fidelity on Snowflake Schemas, Fixes Lost Clustering Keys

The author, who builds the desktop ERD tool Schemity, tested whether a schema could survive a round trip through the tool on Snowflake. Two schemas were used: TPC-DS as shipped in Snowflake's sample data, and a six-table feature schema using Snowflake-specific functionality. Before recent fixes, the rebuilt feature schema failed partway through because an expression-based clustering key could not be read, and clustering keys, transient tables, identity settings, and autoincrement columns were lost. After the fixes, the generated SQL runs as written, TPC-DS shows no catalog differences, and the feature schema differs only in two comments. Both schemas also matched their originals across 14 behavioural tests. This post is the final installment in a series testing round-trip fidelity across the databases Schemity supports.