Normalize is_iceberg Field for Consistency in Snowflake with QUOTED_I…

…DENTIFIERS_IGNORE_CASE (#1229)

* Reproduce. Test fail. Fix. Test green.

* Add changelog.

* Make a test that tests the actual behavior without hardcoding the quoted flag.

* I can't believe how elegant a solution this is.

Use a session-only configuration.

* Fix test.

* Simplify logic for this by moving this away from a quoting situation.

Columns that unquoted are capitalized regardless of the quoting ignore flag. So we let it be uppercase and then normalize it down to lowercase in Python.

* edit comments.

---------

Co-authored-by: Mike Alfare <[email protected]>

.changes/unreleased/Fixes-20241104-172349.yaml

@@ -0,0 +1,6 @@
+kind: Fixes
+body: Iceberg quoting ignore fix.
+time: 2024-11-04T17:23:49.706297-08:00
+custom:
+    Author: versusfacit
+    Issue: "1227"

dbt/adapters/snowflake/impl.py

@@ -263,6 +263,11 @@ def list_relations_without_caching(
         # this can be collapsed once Snowflake adds is_iceberg to show objects
         columns = ["database_name", "schema_name", "name", "kind", "is_dynamic"]
         if self.behavior.enable_iceberg_materializations.no_warn:
+            # The QUOTED_IDENTIFIERS_IGNORE_CASE setting impacts column names like
+            # is_iceberg which is created by dbt, but it does not affect the case
+            # of column values in Snowflake's SHOW OBJECTS query! This
+            # normalization step ensures metadata queries are handled consistently.
+            schema_objects = schema_objects.rename(column_names={"IS_ICEBERG": "is_iceberg"})
             columns.append("is_iceberg")
 
         return [self._parse_list_relations_result(obj) for obj in schema_objects.select(columns)]

dbt/include/snowflake/macros/adapters.sql

@@ -135,7 +135,6 @@
 {% endmacro %}
 
 {% macro snowflake__list_relations_without_caching(schema_relation, max_iter=10, max_results_per_iter=10000) %}
-
   {%- set max_total_results = max_results_per_iter * max_iter -%}
   {%- set sql -%}
     {% if schema_relation is string %}
@@ -147,7 +146,7 @@
     {# -- Gated for performance reason. If you don't want Iceberg, you shouldn't pay the
        -- latency penalty. #}
     {% if adapter.behavior.enable_iceberg_materializations.no_warn %}
-      select all_objects.*, is_iceberg as "is_iceberg"
+      select all_objects.*, is_iceberg
       from table(result_scan(last_query_id(-1))) all_objects
       left join INFORMATION_SCHEMA.tables as all_tables
         on all_tables.table_name = all_objects."name"

tests/functional/adapter/list_relations_tests/test_show_objects.py

@@ -3,6 +3,8 @@
 
 import pytest
 
+from pathlib import Path
+
 from dbt.adapters.factory import get_adapter_by_type
 from dbt.adapters.snowflake import SnowflakeRelation
 
@@ -41,8 +43,32 @@
 """
 )
 
+_MODEL_ICEBERG = """
+{{
+  config(
+    materialized = "table",
+    table_format="iceberg",
+    external_volume="s3_iceberg_snow",
+  )
+}}
+
+select 1
+"""
+
+
+class ShowObjectsBase:
+    @staticmethod
+    def list_relations_without_caching(project) -> List[SnowflakeRelation]:
+        my_adapter = get_adapter_by_type("snowflake")
+        schema = my_adapter.Relation.create(
+            database=project.database, schema=project.test_schema, identifier=""
+        )
+        with get_connection(my_adapter):
+            relations = my_adapter.list_relations_without_caching(schema)
+        return relations
+
 
-class TestShowObjects:
+class TestShowObjects(ShowObjectsBase):
     views: int = 10
     tables: int = 10
     dynamic_tables: int = 10
@@ -66,16 +92,6 @@ def setup(self, project):
         run_dbt(["seed"])
         run_dbt(["run"])
 
-    @staticmethod
-    def list_relations_without_caching(project) -> List[SnowflakeRelation]:
-        my_adapter = get_adapter_by_type("snowflake")
-        schema = my_adapter.Relation.create(
-            database=project.database, schema=project.test_schema, identifier=""
-        )
-        with get_connection(my_adapter):
-            relations = my_adapter.list_relations_without_caching(schema)
-        return relations
-
     def test_list_relations_without_caching(self, project):
         relations = self.list_relations_without_caching(project)
         assert len([relation for relation in relations if relation.is_view]) == self.views
@@ -87,3 +103,25 @@ def test_list_relations_without_caching(self, project):
             len([relation for relation in relations if relation.is_dynamic_table])
             == self.dynamic_tables
         )
+
+
+class TestShowIcebergObjects(ShowObjectsBase):
+    @pytest.fixture(scope="class")
+    def project_config_update(self):
+        return {"flags": {"enable_iceberg_materializations": True}}
+
+    @pytest.fixture(scope="class")
+    def models(self):
+        return {"my_model.sql": _MODEL_ICEBERG}
+
+    def test_quoting_ignore_flag_doesnt_break_iceberg_metadata(self, project):
+        """https://github.com/dbt-labs/dbt-snowflake/issues/1227
+
+        The list relations function involves a metadata sub-query. Regardless of
+        QUOTED_IDENTIFIERS_IGNORE_CASE, this function will fail without proper
+        normalization within the encapsulating python function after the macro invocation
+        returns. This test verifies that normalization is working.
+        """
+        run_dbt(["run"])
+
+        self.list_relations_without_caching(project)