Merge branch 'main' into chore/ci-dbt-1.11

Author: cmgoffena13

sqlmesh/core/model/meta.py

@@ -321,17 +321,19 @@ def _column_descriptions_validator(
         if isinstance(vs, (exp.Tuple, exp.Array)):
             vs = vs.expressions
 
-        raw_col_descriptions = (
-            vs
+        # Normalize each part while it is still an identifier, so that a quoted
+        # column keeps its case on dialects where quoting makes it significant.
+        col_descriptions = (
+            {normalize_identifiers(k, dialect=dialect).name: v for k, v in vs.items()}
             if isinstance(vs, dict)
-            else {".".join([part.this for part in v.this.parts]): v.expression.name for v in vs}
+            else {
+                ".".join(
+                    normalize_identifiers(part, dialect=dialect).name for part in v.this.parts
+                ): v.expression.name
+                for v in vs
+            }
         )
 
-        col_descriptions = {
-            normalize_identifiers(k, dialect=dialect).name: v
-            for k, v in raw_col_descriptions.items()
-        }
-
         columns_to_types = data.get("columns_to_types_")
         if columns_to_types:
             from sqlmesh.core.console import get_console

tests/core/test_model.py

@@ -1000,6 +1000,75 @@ def test_column_descriptions(sushi_context, assert_exp_eq):
     assert model.column_descriptions == {"id": "primary key", "foo": "bar"}
 
 
+def test_column_descriptions_quoted_identifier():
+    expressions = d.parse(
+        """
+        MODEL (
+            name db.table,
+            kind FULL,
+            dialect snowflake,
+            column_descriptions (
+                "myColumn" = 'a case-sensitive column',
+                other_column = 'an unquoted column'
+            )
+        );
+
+        SELECT 1 AS "myColumn", 2 AS other_column
+    """
+    )
+    model = load_sql_based_model(expressions, dialect="snowflake")
+
+    # A quoted key keeps its case, an unquoted one is still normalized.
+    assert model.column_descriptions == {
+        "myColumn": "a case-sensitive column",
+        "OTHER_COLUMN": "an unquoted column",
+    }
+    assert set(model.column_descriptions) <= set(model.columns_to_types)
+
+
+def test_column_descriptions_dotted_identifier():
+    # A nested field is looked up by its dotted path, so every part normalizes on its own.
+    expressions = d.parse(
+        """
+        MODEL (
+            name db.table,
+            kind FULL,
+            dialect bigquery,
+            column_descriptions (
+                record.`myField` = 'a nested field'
+            )
+        );
+
+        SELECT STRUCT(1 AS `myField`) AS record
+    """
+    )
+    model = load_sql_based_model(expressions, dialect="bigquery")
+
+    assert model.column_descriptions == {"record.myfield": "a nested field"}
+
+    expressions = d.parse(
+        """
+        MODEL (
+            name db.table,
+            kind FULL,
+            dialect snowflake,
+            column_descriptions (
+                nested.field = 'an unquoted path',
+                "MyStruct"."myField" = 'a quoted path'
+            )
+        );
+
+        SELECT 1 AS c
+    """
+    )
+    model = load_sql_based_model(expressions, dialect="snowflake")
+
+    assert model.column_descriptions == {
+        "NESTED.FIELD": "an unquoted path",
+        "MyStruct.myField": "a quoted path",
+    }
+
+
 def test_model_jinja_macro_reference_extraction():
     @macro()
     def test_macro(**kwargs) -> None: