[BugFix] make max_by/min_by's output result column always nullable (S…

…tarRocks#55116)

Signed-off-by: before-Sunrise <[email protected]>

be/src/exec/aggregator.cpp

@@ -42,8 +42,8 @@
 
 namespace starrocks {
 
-static const std::unordered_set<std::string> ALWAYS_NULLABLE_RESULT_AGG_FUNCS = {"variance_samp", "var_samp",
-                                                                                 "stddev_samp", "covar_samp", "corr"};
+static const std::unordered_set<std::string> ALWAYS_NULLABLE_RESULT_AGG_FUNCS = {
+        "variance_samp", "var_samp", "stddev_samp", "covar_samp", "corr", "max_by_v2", "min_by_v2"};
 
 static const std::string AGG_STATE_UNION_SUFFIX = "_union";
 static const std::string AGG_STATE_MERGE_SUFFIX = "_merge";

test/sql/test_max_min_by_not_filter_nulls_with_nulls/R/test_max_min_by_with_empty_table

@@ -0,0 +1,36 @@
+-- name: test_max_min_by_with_empty_table
+CREATE TABLE `primary_table_with_null_partition` (
+                `k1` date not null,
+                `k2` datetime not null,
+                `k3` varchar(20) not null,
+                `k4` varchar(20) not null,
+                `k5` boolean not null,
+                `v1` tinyint,
+                `v2` smallint,
+                `v3` int,
+                `v4` bigint,
+                `v5` largeint,
+                `v6` float,
+                `v7` double,
+                `v8` decimal(27,9)
+            ) 
+            PRIMARY KEY(`k1`, `k2`,`k3`)
+            COMMENT "OLAP"
+            PARTITION BY RANGE(`k1`)
+            (
+                PARTITION `p202006` VALUES LESS THAN ("2020-07-01"),
+                PARTITION `p202007` VALUES LESS THAN ("2020-08-01"),
+                PARTITION `p202008` VALUES LESS THAN ("2020-09-01")
+            )
+            DISTRIBUTED BY HASH(`k1`, `k2`) BUCKETS 3
+            PROPERTIES (
+                "replication_num" = "1",
+                "enable_persistent_index" = "true",
+                "storage_format" = "v2" 
+            );
+-- result:
+-- !result
+select max_by(k1,k1) from primary_table_with_null_partition;
+-- result:
+None
+-- !result

test/sql/test_max_min_by_not_filter_nulls_with_nulls/T/test_max_min_by_with_empty_table

@@ -0,0 +1,32 @@
+-- name: test_max_min_by_with_empty_table
+CREATE TABLE `primary_table_with_null_partition` (
+                `k1` date not null,
+                `k2` datetime not null,
+                `k3` varchar(20) not null,
+                `k4` varchar(20) not null,
+                `k5` boolean not null,
+                `v1` tinyint,
+                `v2` smallint,
+                `v3` int,
+                `v4` bigint,
+                `v5` largeint,
+                `v6` float,
+                `v7` double,
+                `v8` decimal(27,9)
+            ) 
+            PRIMARY KEY(`k1`, `k2`,`k3`)
+            COMMENT "OLAP"
+            PARTITION BY RANGE(`k1`)
+            (
+                PARTITION `p202006` VALUES LESS THAN ("2020-07-01"),
+                PARTITION `p202007` VALUES LESS THAN ("2020-08-01"),
+                PARTITION `p202008` VALUES LESS THAN ("2020-09-01")
+            )
+            DISTRIBUTED BY HASH(`k1`, `k2`) BUCKETS 3
+            PROPERTIES (
+                "replication_num" = "1",
+                "enable_persistent_index" = "true",
+                "storage_format" = "v2" 
+            );
+
+select max_by(k1,k1) from primary_table_with_null_partition;