fix(plugin): MCOL-6088 This ensures that SELECT col AS alias ... GROUP BY alias works correctly even when col is wrapped in an implicit aggregate wrapper.

mariadb-LeonidFedorov · mariadb-AlekseiBukhalov · commit 5ae2e2133cf1 · 2025-11-25T16:18:23.000+01:00
diff --git a/dbcon/mysql/ha_mcs_execplan.cpp b/dbcon/mysql/ha_mcs_execplan.cpp
@@ -5848,8 +5848,11 @@ int processGroupBy(SELECT_LEX& select_lex, gp_walk_info& gwi, const bool withRol
 
       if (sc)
       {
+        // buildSimpleColumn succeeded - we have a valid table column reference
+        // This table column takes precedence over any SELECT aliases (SQL standard behavior)
+        // Just check if it's in the SELECT list to set orderPos (optional optimization)
         bool found = false;
-        for (uint32_t j = 0; j < gwi.returnedCols.size(); j++)
+        for (uint32_t j = 0; !found && j < gwi.returnedCols.size(); j++)
         {
           if (sc->sameColumn(gwi.returnedCols[j].get()))
           {
@@ -5858,25 +5861,95 @@ int processGroupBy(SELECT_LEX& select_lex, gp_walk_info& gwi, const bool withRol
             break;
           }
         }
-        for (uint32_t j = 0; !found && j < gwi.returnedCols.size(); j++)
+        // Note: We do NOT fall back to alias matching here.
+        // The SimpleColumn from buildSimpleColumn is the correct table column to use.
+      }
+      else if (rc)
+      {
+        // rc is not a SimpleColumn (e.g., could be an aggregate or other type)
+        // Try to match by alias name
+        for (uint32_t j = 0; j < gwi.returnedCols.size(); j++)
         {
-          if (strcasecmp(sc->alias().c_str(), gwi.returnedCols[j]->alias().c_str()) == 0)
+          if (ifp->name.length && strcasecmp(ifp->name.str, gwi.returnedCols[j].get()->alias().c_str()) == 0)
           {
+            ReturnedColumn* matched_col = gwi.returnedCols[j].get();
+
+            // If it's an aggregate column, try to unwrap it to get the underlying column
+            AggregateColumn* agg_check = dynamic_cast<AggregateColumn*>(matched_col);
+            if (agg_check)
+            {
+              // Check if the aggregate has parameters and use the first parameter if it's not an aggregate
+              // itself
+              if (!agg_check->aggParms().empty())
+              {
+                ReturnedColumn* inner_col = agg_check->aggParms()[0].get();
+                AggregateColumn* inner_agg = dynamic_cast<AggregateColumn*>(inner_col);
+
+                if (!inner_agg)
+                {
+                  // Use the inner non-aggregate column
+                  matched_col = inner_col;
+                }
+                else
+                {
+                  // It's a nested aggregate, skip it
+                  continue;
+                }
+              }
+              else
+              {
+                // Aggregate with no parameters (like COUNT(*)), skip it
+                continue;
+              }
+            }
+
             delete rc;
-            rc = gwi.returnedCols[j].get()->clone();
+            rc = matched_col->clone();
             rc->orderPos(j);
             break;
           }
         }
       }
       else
       {
+        // buildSimpleColumn() returned NULL - this might be due to ambiguity
+        // or column not found. Try to find a matching alias in the SELECT list.
         for (uint32_t j = 0; j < gwi.returnedCols.size(); j++)
         {
-          if (ifp->name.length && string(ifp->name.str) == gwi.returnedCols[j].get()->alias())
+          if (ifp->name.length && strcasecmp(ifp->name.str, gwi.returnedCols[j].get()->alias().c_str()) == 0)
           {
-            delete rc;
-            rc = gwi.returnedCols[j].get()->clone();
+            ReturnedColumn* matched_col = gwi.returnedCols[j].get();
+
+            // If it's an aggregate column, try to unwrap it to get the underlying column
+            AggregateColumn* agg_check = dynamic_cast<AggregateColumn*>(matched_col);
+            if (agg_check)
+            {
+              // Check if the aggregate has parameters and use the first parameter if it's not an aggregate
+              // itself
+              if (!agg_check->aggParms().empty())
+              {
+                ReturnedColumn* inner_col = agg_check->aggParms()[0].get();
+                AggregateColumn* inner_agg = dynamic_cast<AggregateColumn*>(inner_col);
+
+                if (!inner_agg)
+                {
+                  // Use the inner non-aggregate column
+                  matched_col = inner_col;
+                }
+                else
+                {
+                  // It's a nested aggregate, skip it
+                  continue;
+                }
+              }
+              else
+              {
+                // Aggregate with no parameters (like COUNT(*)), skip it
+                continue;
+              }
+            }
+
+            rc = matched_col->clone();
             rc->orderPos(j);
             break;
           }
diff --git a/mysql-test/columnstore/basic/r/group_by_ambiguous.result b/mysql-test/columnstore/basic/r/group_by_ambiguous.result
@@ -0,0 +1,50 @@
+DROP DATABASE IF EXISTS group_by_ambiguous_test;
+CREATE DATABASE group_by_ambiguous_test;
+USE group_by_ambiguous_test;
+CREATE TABLE `test2` (
+`id` INT UNSIGNED NOT NULL,
+`GBP` DOUBLE UNSIGNED NOT NULL,
+`cat_id` INT UNSIGNED NOT NULL,
+`cat_id_level_0` INT UNSIGNED NOT NULL
+) ENGINE=Columnstore;
+INSERT INTO test2 VALUES (1, 100.5, 10, 20);
+INSERT INTO test2 VALUES (2, 200.5, 10, 30);
+INSERT INTO test2 VALUES (3, 300.5, 20, 20);
+#
+# Test case: GROUP BY with ambiguous column reference
+# This should not throw a ColumnStore error, but resolve to the table column (SQL standard)
+#
+SELECT cat_id_level_0 AS entity_id, SUM(GBP) AS spend, cat_id_level_0 AS cat_id
+FROM test2
+GROUP BY cat_id;
+entity_id	spend	cat_id
+20	300.5	20
+30	301	30
+Warnings:
+Warning	1052	Column 'cat_id' in GROUP BY is ambiguous
+DROP TABLE test2;
+#
+# Test case: GROUP BY with alias that matches a column but points to different column
+#
+CREATE TABLE t1 (col1 int, col2 varchar(5), col_t1 int) ENGINE=Columnstore;
+INSERT INTO t1 VALUES(10,'hello',10);
+INSERT INTO t1 VALUES(20,'hello',20);
+INSERT INTO t1 VALUES(30,'hello',30);
+INSERT INTO t1 VALUES(10,'bye',10);
+INSERT INTO t1 VALUES(10,'sam',10);
+INSERT INTO t1 VALUES(10,'bob',10);
+# GROUP BY col2 is ambiguous: matches both table column and SELECT alias
+# SQL standard: prefers table column col2 (varchar) over alias
+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;
+col2
+10
+10
+10
+60
+Warnings:
+Warning	1052	Column 'col2' in GROUP BY is ambiguous
+#
+# Cleanup
+#
+DROP TABLE t1;
+DROP DATABASE group_by_ambiguous_test;
diff --git a/mysql-test/columnstore/basic/r/mcol-4525.result b/mysql-test/columnstore/basic/r/mcol-4525.result
@@ -77,18 +77,22 @@ INSERT INTO t1 VALUES(10,'bye',10);
 INSERT INTO t1 VALUES(10,'sam',10);
 INSERT INTO t1 VALUES(10,'bob',10);
 SET columnstore_select_handler=ON;
-SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2;
-ERROR 42000: The storage engine for the table doesn't support MCS-2016: Non supported item 'col2' on the GROUP BY list.
+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;
+col2
+10
+10
+10
+60
 SELECT col1 c FROM t1 ORDER BY AVG(col1);
 c
 10
 SET columnstore_select_handler=AUTO;
-SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2;
+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;
 col2
 10
 10
-60
 10
+60
 SELECT col1 c FROM t1 ORDER BY AVG(col1);
 c
 10
diff --git a/mysql-test/columnstore/basic/t/group_by_ambiguous.test b/mysql-test/columnstore/basic/t/group_by_ambiguous.test
@@ -0,0 +1,51 @@
+--source ../include/have_columnstore.inc
+
+--disable_warnings
+DROP DATABASE IF EXISTS group_by_ambiguous_test;
+--enable_warnings
+
+CREATE DATABASE group_by_ambiguous_test;
+USE group_by_ambiguous_test;
+
+CREATE TABLE `test2` (
+  `id` INT UNSIGNED NOT NULL,
+  `GBP` DOUBLE UNSIGNED NOT NULL,
+  `cat_id` INT UNSIGNED NOT NULL,
+  `cat_id_level_0` INT UNSIGNED NOT NULL
+) ENGINE=Columnstore;
+
+INSERT INTO test2 VALUES (1, 100.5, 10, 20);
+INSERT INTO test2 VALUES (2, 200.5, 10, 30);
+INSERT INTO test2 VALUES (3, 300.5, 20, 20);
+
+--echo #
+--echo # Test case: GROUP BY with ambiguous column reference
+--echo # This should not throw a ColumnStore error, but resolve to the table column (SQL standard)
+--echo #
+
+SELECT cat_id_level_0 AS entity_id, SUM(GBP) AS spend, cat_id_level_0 AS cat_id
+FROM test2
+GROUP BY cat_id;
+DROP TABLE test2;
+
+--echo #
+--echo # Test case: GROUP BY with alias that matches a column but points to different column
+--echo #
+
+CREATE TABLE t1 (col1 int, col2 varchar(5), col_t1 int) ENGINE=Columnstore;
+INSERT INTO t1 VALUES(10,'hello',10);
+INSERT INTO t1 VALUES(20,'hello',20);
+INSERT INTO t1 VALUES(30,'hello',30);
+INSERT INTO t1 VALUES(10,'bye',10);
+INSERT INTO t1 VALUES(10,'sam',10);
+INSERT INTO t1 VALUES(10,'bob',10);
+
+--echo # GROUP BY col2 is ambiguous: matches both table column and SELECT alias
+--echo # SQL standard: prefers table column col2 (varchar) over alias
+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;
+
+--echo #
+--echo # Cleanup
+--echo #
+DROP TABLE t1;
+DROP DATABASE group_by_ambiguous_test;
diff --git a/mysql-test/columnstore/basic/t/mcol-4525.test b/mysql-test/columnstore/basic/t/mcol-4525.test
@@ -83,11 +83,10 @@ INSERT INTO t1 VALUES(10,'bye',10);
 INSERT INTO t1 VALUES(10,'sam',10);
 INSERT INTO t1 VALUES(10,'bob',10);
 SET columnstore_select_handler=ON;
---error 1178
-SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2;
+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;
 SELECT col1 c FROM t1 ORDER BY AVG(col1);
 SET columnstore_select_handler=AUTO;
-SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2;
+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;
 SELECT col1 c FROM t1 ORDER BY AVG(col1);
 DROP TABLE t1;
 

Original file line number	Diff line number	Diff line change
`@@ -77,18 +77,22 @@ INSERT INTO t1 VALUES(10,'bye',10);`
`77`	`77`	`INSERT INTO t1 VALUES(10,'sam',10);`
`78`	`78`	`INSERT INTO t1 VALUES(10,'bob',10);`
`79`	`79`	`SET columnstore_select_handler=ON;`
`80`		`-SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2;`
`81`		`-ERROR 42000: The storage engine for the table doesn't support MCS-2016: Non supported item 'col2' on the GROUP BY list.`
	`80`	`+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;`
	`81`	`+col2`
	`82`	`+10`
	`83`	`+10`
	`84`	`+10`
	`85`	`+60`
`82`	`86`	`SELECT col1 c FROM t1 ORDER BY AVG(col1);`
`83`	`87`	`c`
`84`	`88`	`10`
`85`	`89`	`SET columnstore_select_handler=AUTO;`
`86`		`-SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2;`
	`90`	`+SELECT SUM(col1) AS col2 FROM t1 GROUP BY col2 ORDER BY col2;`
`87`	`91`	`col2`
`88`	`92`	`10`
`89`	`93`	`10`
`90`		`-60`
`91`	`94`	`10`
	`95`	`+60`
`92`	`96`	`SELECT col1 c FROM t1 ORDER BY AVG(col1);`
`93`	`97`	`c`
`94`	`98`	`10`