fix(schema_loader): quote case-sensitive column names in external models

create_external_models wrote column names bare, so a Postgres column
"ID" was read back as id. A name is now quoted only when the loader's
identifier normalization would change it. Names that already round-trip
are written as before.

Before the fix the new test failed with '"ID"' != 'ID' on Postgres and
'"lower_case"' != 'lower_case' on Snowflake.

Fixes #6058

Signed-off-by: ReguiguiMohamed <mohamedreguigui2004@gmail.com>

Author: ReguiguiMohamed

sqlmesh/core/schema_loader.py

@@ -6,6 +6,7 @@
 
 from sqlglot import exp
 from sqlglot.dialects.dialect import DialectType
+from sqlglot.optimizer.normalize_identifiers import normalize_identifiers
 
 from sqlmesh.core.console import get_console
 from sqlmesh.core.engine_adapter import EngineAdapter
@@ -98,7 +99,12 @@ def get_columns(
     """
     try:
         columns = adapter.columns(table, include_pseudo_columns=True)
-        return {c: dtype.sql(dialect=dialect) for c, dtype in columns.items()}
+        result = {}
+        for c, dtype in columns.items():
+            if normalize_identifiers(c, dialect=dialect).name != c:
+                c = exp.to_identifier(c).sql(dialect=dialect, identify=True)
+            result[c] = dtype.sql(dialect=dialect)
+        return result
     except Exception as e:
         msg = f"Unable to get schema for '{table}': '{e}'."
         if strict:

tests/core/test_schema_loader.py

@@ -389,6 +389,52 @@ def test_no_internal_model_conversion(tmp_path: Path, mocker: MockerFixture):
         create_external_model(**row, dialect="bigquery")
 
 
+@pytest.mark.parametrize(
+    "dialect, columns, expected_keys",
+    [
+        (
+            "postgres",
+            ["ID", "MixedCase", "lower_case", "a.b"],
+            ['"ID"', '"MixedCase"', "lower_case", "a.b"],
+        ),
+        ("snowflake", ["ID", "lower_case"], ["ID", '"lower_case"']),
+        ("clickhouse", ["a\\b"], ["a\\b"]),
+    ],
+)
+def test_create_external_models_quotes_case_sensitive_columns(
+    tmp_path: Path,
+    mocker: MockerFixture,
+    dialect: str,
+    columns: t.List[str],
+    expected_keys: t.List[str],
+):
+    engine_adapter_mock = mocker.Mock()
+    engine_adapter_mock.columns.return_value = {
+        name: exp.DataType.build("text") for name in columns
+    }
+
+    state_reader_mock = mocker.Mock()
+    state_reader_mock.nodes_exist.return_value = set()
+
+    model = SqlModel(name="a", query=parse_one("select * FROM tbl"))
+
+    filename = tmp_path / c.EXTERNAL_MODELS_YAML
+    create_external_models_file(
+        filename,
+        {"a": model},  # type: ignore
+        engine_adapter_mock,
+        state_reader_mock,
+        dialect,
+    )
+
+    schema = yaml.load(filename)
+    assert list(schema[0]["columns"]) == expected_keys
+
+    external_model = create_external_model(**schema[0], dialect=dialect)
+    assert external_model.columns_to_types is not None
+    assert list(external_model.columns_to_types) == columns
+
+
 def test_missing_table(tmp_path: Path):
     config = Config(gateways=GatewayConfig(connection=DuckDBConnectionConfig()))
     context = Context(paths=[str(tmp_path.absolute())], config=config)