Fix optimize_aggregation_in_order in case of empty result set

Note, that this is not complete fix, see the next two patches.

Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>

src/Processors/Transforms/MergingAggregatedTransform.cpp

@@ -23,7 +23,11 @@ void MergingAggregatedTransform::consume(Chunk chunk)
         LOG_TRACE(log, "Reading blocks of partially aggregated data.");
     }
 
-    total_input_rows += chunk.getNumRows();
+    size_t input_rows = chunk.getNumRows();
+    if (!input_rows)
+        return;
+
+    total_input_rows += input_rows;
     ++total_input_blocks;
 
     const auto & info = chunk.getChunkInfo();

tests/queries/0_stateless/02176_optimize_aggregation_in_order_empty.reference

@@ -0,0 +1,8 @@
+-- { echoOn }
+
+-- regression for optimize_aggregation_in_order with empty result set
+-- that cause at first
+--   "Chunk should have AggregatedChunkInfo in GroupingAggregatedTransform"
+-- at first and after
+--   "Chunk should have AggregatedChunkInfo in GroupingAggregatedTransform"
+select count() from remote('127.{1,2}', currentDatabase(), data_02176) where key = 0 group by key settings optimize_aggregation_in_order=1;

tests/queries/0_stateless/02176_optimize_aggregation_in_order_empty.sql

@@ -0,0 +1,14 @@
+drop table if exists data_02176;
+create table data_02176 (key Int) Engine=MergeTree() order by key;
+
+-- { echoOn }
+
+-- regression for optimize_aggregation_in_order with empty result set
+-- that cause at first
+--   "Chunk should have AggregatedChunkInfo in GroupingAggregatedTransform"
+-- at first and after
+--   "Chunk should have AggregatedChunkInfo in GroupingAggregatedTransform"
+select count() from remote('127.{1,2}', currentDatabase(), data_02176) where key = 0 group by key settings optimize_aggregation_in_order=1;
+
+-- { echoOff }
+drop table data_02176;