Seed data for exercising diff functionality across connectors. All files create a customers table (key column id) with the same base dataset, adapted per engine's SQL dialect and type system.

• postgres_customers.sql — the baseline / "source A". Has rows 1-8, 10, 11.
• All other files (databricks_customers.sql, mysql_customers.sql, snowflake_customers.sql, bigquery_customers.sql, cockroachdb_customers.sql, yugabyte_customers.sql, tidb_customers.sql) are "source B" variants. Each has rows 1-7, 9, 10, 12.

Diffing Postgres against any one of them (with id as the key) produces the same result:

Expected summary: 7 matched, 2 added, 2 deleted, 1 modified (mismatched columns on row 10: amount, status).

Diffing any two non-Postgres sides against each other (e.g. MySQL vs. Snowflake) should report a full match (0 differences) since they all share the identical "source B" dataset — useful as a sanity check that the connector/algorithm pairing itself introduces no false positives.

1. Apply postgres_customers.sql to a Postgres instance.
2. Apply the matching fixture to whichever other engine you want to test.
3. Run dimer-diff, select the two sources, table customers on both sides, key column id.

Cockroach DB and YugabyteDB are PostgreSQL wire-compatible and TiDB is MySQL wire-compatible, so their fixtures reuse the same DDL/data as Postgres/MySQL — only the CLI/tooling used to apply them and the connection env vars differ.