From 049f8ad2f98182ffc48447358aaeed55aec443a5 Mon Sep 17 00:00:00 2001
From: AlexYue <yj976240184@gmail.com>
Date: Fri, 13 Jan 2023 18:33:40 +0800
Subject: [PATCH] [Bug](sort)fix merge sorter might div zero when block bytes
 less than block rows (#15859)

If block bytes are bigger than the corresponding block's rows, then the avg_size_per_row would be zero. Which would end up diving zero in the following logic.
---
 be/src/vec/common/sort/sorter.cpp             |  2 +-
 .../data/demo_p0/streamLoad_action.out        |  8 ++++++
 .../suites/demo_p0/streamLoad_action.groovy   | 26 +++++++++++++++++++
 3 files changed, 35 insertions(+), 1 deletion(-)
 create mode 100644 regression-test/data/demo_p0/streamLoad_action.out

diff --git a/be/src/vec/common/sort/sorter.cpp b/be/src/vec/common/sort/sorter.cpp
index 778ad27fe2..9bc0a1034c 100644
--- a/be/src/vec/common/sort/sorter.cpp
+++ b/be/src/vec/common/sort/sorter.cpp
@@ -43,7 +43,7 @@ Status MergeSorterState::add_sorted_block(Block& block) {
         return Status::OK();
     }
     if (0 == avg_row_bytes_) {
-        avg_row_bytes_ = block.bytes() / rows;
+        avg_row_bytes_ = std::max((std::size_t)1, block.bytes() / rows);
         spill_block_batch_size_ = (BLOCK_SPILL_BATCH_BYTES + avg_row_bytes_ - 1) / avg_row_bytes_;
     }
 
diff --git a/regression-test/data/demo_p0/streamLoad_action.out b/regression-test/data/demo_p0/streamLoad_action.out
new file mode 100644
index 0000000000..445a11c6dc
--- /dev/null
+++ b/regression-test/data/demo_p0/streamLoad_action.out
@@ -0,0 +1,8 @@
+-- This file is automatically generated. You should know what you did if you want to edit this
+-- !sql --
+0
+2
+2
+3
+3
+
diff --git a/regression-test/suites/demo_p0/streamLoad_action.groovy b/regression-test/suites/demo_p0/streamLoad_action.groovy
index 49df370d11..87affd32c4 100644
--- a/regression-test/suites/demo_p0/streamLoad_action.groovy
+++ b/regression-test/suites/demo_p0/streamLoad_action.groovy
@@ -78,4 +78,30 @@ suite("streamLoad_action") {
             assertTrue(json.NumberLoadedRows > 0 && json.LoadBytes > 0)
         }
     }
+
+    // to test merge sort
+    sql """ DROP TABLE IF EXISTS B """
+    sql """
+        CREATE TABLE IF NOT EXISTS B
+        (
+            b_id int
+        )
+        DISTRIBUTED BY HASH(b_id) BUCKETS 1
+        PROPERTIES("replication_num" = "1");
+    """
+
+    sql " INSERT INTO B values (1);"
+
+    qt_sql """
+        SELECT subq_0.`c1` AS c1
+        FROM
+        (SELECT version() AS c0,
+                  ref_0.`id` AS c1
+        FROM test_streamload_action1 AS ref_0
+        LEFT JOIN B AS ref_9 ON (ref_0.`id` = ref_9.`b_id`)
+        WHERE ref_9.`b_id` IS NULL) AS subq_0
+        WHERE subq_0.`c0` IS NOT NULL
+        ORDER BY subq_0.`c1`, subq_0.`c0` DESC
+        LIMIT 5;
+    """
 }