Fix!: use sql() instead of gen() for time data type data hash computation

Author: georgesittas

sqlmesh/core/model/kind.py

@@ -711,7 +711,7 @@ def data_hash_values(self) -> t.List[t.Optional[str]]:
             gen(self.valid_from_name),
             gen(self.valid_to_name),
             str(self.invalidate_hard_deletes),
-            gen(self.time_data_type),
+            self.time_data_type.sql(self.dialect),
             gen(self.batch_size) if self.batch_size is not None else None,
         ]
 

sqlmesh/migrations/v0083_use_sql_for_scd_time_data_type_data_hash.py

@@ -0,0 +1,5 @@
+"""Use sql(...) instead of gen when computing the data hash of the time data type."""
+
+
+def migrate(state_sync, **kwargs):  # type: ignore
+    pass

tests/core/test_model.py

@@ -9587,3 +9587,43 @@ def test_semicolon_is_not_included_in_model_state(tmp_path, assert_exp_eq):
     plan = ctx.plan(no_prompts=True, auto_apply=True)
 
     assert not plan.context_diff.modified_snapshots
+
+
+def test_scd_time_data_type_does_not_cause_diff_after_deserialization() -> None:
+    for dialect in (
+        "athena",
+        "bigquery",
+        "clickhouse",
+        "databricks",
+        "duckdb",
+        "dune",
+        "hive",
+        "mysql",
+        "postgres",
+        "presto",
+        "redshift",
+        "snowflake",
+        "spark",
+        "trino",
+        "tsql",
+    ):
+        sql = f"""
+        MODEL (
+          name test_schema.test_model,
+          kind SCD_TYPE_2_BY_COLUMN (
+            unique_key ARRAY(col),
+            columns ARRAY(col),
+            invalidate_hard_deletes TRUE,
+            on_destructive_change error
+          ),
+          dialect {dialect}
+        );
+
+        SELECT
+          1 AS col
+        """
+
+        model = load_sql_based_model(d.parse(sql))
+        deserialized_model = SqlModel.parse_raw(model.json())
+
+        assert model.data_hash == deserialized_model.data_hash