Merge pull request #995 from Altinity/backport/24.8/83844

Backport ClickHouse#83844 to 24.8: Support references to unknown tables in remote table functions

Author: zvonand

src/Analyzer/Resolve/QueryAnalyzer.cpp

@@ -4629,7 +4629,26 @@ void QueryAnalyzer::resolveTableFunction(QueryTreeNodePtr & table_function_node,
                 table_function_node_to_resolve_typed->getArgumentsNode() = table_function_argument_function->getArgumentsNode();
 
                 QueryTreeNodePtr table_function_node_to_resolve = std::move(table_function_node_to_resolve_typed);
-                resolveTableFunction(table_function_node_to_resolve, scope, expressions_visitor, true /*nested_table_function*/);
+                if (table_function_argument_function_name == "view")
+                {
+                    /// Subquery in view() table function can reference tables that don't exist on the initiator.
+                    /// In the following example `users` table may be not available on the initiator:
+                    /// SELECT *
+                    /// FROM remoteSecure(<address>, view(
+                    ///     SELECT
+                    ///         t1.age,
+                    ///         t1.name,
+                    ///         t2.name
+                    ///     FROM users AS t1
+                    ///     INNER JOIN users AS t2 ON t1.uid = t2.uid
+                    /// ), <user>, <password>)
+                    /// SETTINGS prefer_localhost_replica = 0
+                    skip_analysis_arguments_indexes.push_back(table_function_argument_index);
+                }
+                else
+                {
+                    resolveTableFunction(table_function_node_to_resolve, scope, expressions_visitor, true /*nested_table_function*/);
+                }
 
                 result_table_function_arguments.push_back(std::move(table_function_node_to_resolve));
                 continue;

src/Storages/StorageDistributed.cpp

@@ -759,9 +759,28 @@ QueryTreeNodePtr buildQueryTreeDistributed(SelectQueryInfo & query_info,
         if (table_expression_modifiers)
             table_function_node->setTableExpressionModifiers(*table_expression_modifiers);
 
-        QueryAnalysisPass query_analysis_pass;
-        QueryTreeNodePtr node = table_function_node;
-        query_analysis_pass.run(node, query_context);
+        /// Subquery in table function `view` may reference tables that don't exist on the initiator.
+        if (table_function_node->getTableFunctionName() == "view")
+        {
+            auto get_column_options = GetColumnsOptions(GetColumnsOptions::All).withExtendedObjects().withVirtuals();
+            auto column_names_and_types = distributed_storage_snapshot->getColumns(get_column_options);
+
+            StorageID fake_storage_id = StorageID::createEmpty();
+            if (auto * table_node = query_info.table_expression->as<TableNode>())
+                fake_storage_id = table_node->getStorage()->getStorageID();
+            else if (auto * original_table_function_node = query_info.table_expression->as<TableFunctionNode>())
+                fake_storage_id = original_table_function_node->getStorage()->getStorageID();
+
+            auto storage = std::make_shared<StorageDummy>(fake_storage_id, ColumnsDescription{column_names_and_types});
+
+            table_function_node->resolve({}, std::move(storage), query_context, /*unresolved_arguments_indexes_=*/{ 0 });
+        }
+        else
+        {
+            QueryAnalysisPass query_analysis_pass;
+            QueryTreeNodePtr node = table_function_node;
+            query_analysis_pass.run(node, query_context);
+        }
 
         replacement_table_expression = std::move(table_function_node);
     }
@@ -809,9 +828,7 @@ void StorageDistributed::read(
 
     if (settings.allow_experimental_analyzer)
     {
-        StorageID remote_storage_id = StorageID::createEmpty();
-        if (!remote_table_function_ptr)
-            remote_storage_id = StorageID{remote_database, remote_table};
+        StorageID remote_storage_id = StorageID{remote_database, remote_table};
 
         auto query_tree_distributed = buildQueryTreeDistributed(modified_query_info,
             query_info.merge_storage_snapshot ? query_info.merge_storage_snapshot : storage_snapshot,

tests/integration/test_remote_function_view/__init__.py

@@ -0,0 +1,16 @@
+<clickhouse>
+    <remote_servers>
+        <cluster>
+            <shard>
+                <replica>
+                    <host>node1</host>
+                    <port>9000</port>
+                </replica>
+                <replica>
+                    <host>node2</host>
+                    <port>9000</port>
+                </replica>
+            </shard>
+        </cluster>
+    </remote_servers>
+</clickhouse>

tests/integration/test_remote_function_view/configs/clusters.xml

@@ -0,0 +1,48 @@
+import pytest
+
+from helpers.cluster import ClickHouseCluster
+
+cluster = ClickHouseCluster(__file__)
+node1 = cluster.add_instance("node1", main_configs=["configs/clusters.xml"])
+node2 = cluster.add_instance("node2", main_configs=["configs/clusters.xml"])
+
+
+@pytest.fixture(scope="module")
+def start_cluster():
+    try:
+        cluster.start()
+
+        node2.query(
+            """
+                CREATE TABLE test_table(
+                    APIKey UInt32,
+                    CustomAttributeId UInt64,
+                    ProfileIDHash UInt64,
+                    DeviceIDHash UInt64,
+                    Data String)
+                ENGINE = SummingMergeTree()
+                ORDER BY (APIKey, CustomAttributeId, ProfileIDHash, DeviceIDHash, intHash32(DeviceIDHash))
+            """
+        )
+        yield cluster
+
+    finally:
+        cluster.shutdown()
+
+
+def test_remote(start_cluster):
+    assert (
+        node1.query(
+            "SELECT 1 FROM remote('node2', view(SELECT * FROM default.test_table)) WHERE (APIKey = 137715) AND (CustomAttributeId IN (45, 66)) AND (ProfileIDHash != 0) LIMIT 1"
+        )
+        == ""
+    )
+
+
+def test_remote_fail(start_cluster):
+    assert (
+        "Unknown table expression identifier 'default.table_not_exists'"
+        in node1.query_and_get_error(
+            "SELECT 1 FROM remote('node2', view(SELECT * FROM default.table_not_exists)) WHERE (APIKey = 137715) AND (CustomAttributeId IN (45, 66)) AND (ProfileIDHash != 0) LIMIT 1"
+        )
+    )