get rid of unneeded getSampleBlock calls

Author: filimonov

src/Storages/StorageDistributed.cpp

@@ -1048,7 +1048,6 @@ void StorageDistributed::read(
 
     SelectQueryInfo modified_query_info = query_info;
 
-    std::vector<Block> all_headers;
     std::vector<SelectQueryInfo> all_query_infos;
 
     const auto & settings = local_context->getSettingsRef();
@@ -1081,8 +1080,9 @@ void StorageDistributed::read(
 
         if (!additional_storages.empty())
         {
-            for (const auto & extra_storage : additional_storages)
+            for (size_t i = 0; i < additional_storages.size(); ++i)
             {
+                const auto & extra_storage = additional_storages[i];
                 // Create a modified query info with the additional predicate
                 SelectQueryInfo additional_query_info = query_info;
 
@@ -1092,17 +1092,12 @@ void StorageDistributed::read(
                     extra_storage.storage_ast,
                     extra_storage.storage_filter_ast);
 
-                // TODO: do we need it? somewhere here the DESCRIBE TABLE is triggered, try to avoid it.
-                auto additional_header = InterpreterSelectQueryAnalyzer::getSampleBlock(additional_query_tree, local_context, SelectQueryOptions(processed_stage).analyze());
-
-                for (auto & column : additional_header)
-                    column.column = column.column->convertToFullColumnIfConst();
+                // We don't need to get sample block for additional storages as it triggers unnecessary analysis
+                // and causes issues with virtual columns that don't exist in remote tables
 
                 additional_query_info.query = queryNodeToDistributedSelectQuery(additional_query_tree);
                 additional_query_info.query_tree = std::move(additional_query_tree);
 
-                // TODO: do we need it?
-                all_headers.push_back(additional_header);
                 all_query_infos.push_back(additional_query_info);
             }
         }
@@ -1122,20 +1117,19 @@ void StorageDistributed::read(
 
         if (!additional_storages.empty())
         {
-            for (const auto & extra_storage : additional_storages)
+            for (size_t i = 0; i < additional_storages.size(); ++i)
             {
+                const auto & extra_storage = additional_storages[i];
                 SelectQueryInfo additional_query_info = query_info;
 
-                // TODO: do we need it? somewhere here the DESCRIBE TABLE is triggered, try to avoid it.
-                auto additional_header = InterpreterSelectQuery(additional_query_info.query, local_context, SelectQueryOptions(processed_stage).analyze()).getSampleBlock();
+                // We don't need to get sample block for additional storages as it triggers unnecessary analysis
+                // and causes issues with virtual columns that don't exist in remote tables
 
                 additional_query_info.query = ClusterProxy::rewriteSelectQuery(
                     local_context, additional_query_info.query,
                     "", "", extra_storage.storage_ast,
                     extra_storage.storage_filter_ast);
 
-                // TODO: do we need it?
-                all_headers.push_back(additional_header);
                 all_query_infos.push_back(additional_query_info);
             }
         }
@@ -1192,7 +1186,6 @@ void StorageDistributed::read(
     {
         auto additional_query_info = all_query_infos[i];
         const auto & storage = additional_storages[i].storage;
-        auto additional_header = all_headers[i];
 
         // Create a new query plan for this additional storage
         QueryPlan additional_plan;