diff --git a/internal/formatter.go b/internal/formatter.go
index cb6f323..25a0df8 100644
--- a/internal/formatter.go
+++ b/internal/formatter.go
@@ -800,12 +800,6 @@ func (n *FilterScanNode) FormatSQL(ctx context.Context) (string, error) {
 	if err != nil {
 		return "", err
 	}
-	nodeMap := nodeMapFromContext(ctx)
-	for _, node := range nodeMap.FindNodeFromResolvedNode(n.node) {
-		if _, ok := node.(*parsed_ast.HavingNode); ok {
-			return fmt.Sprintf("%s HAVING %s", input, filter), nil
-		}
-	}
 	currentQuery := string(removeExpressions.ReplaceAllString(input, ""))
 
 	// Qualify the statement if the input is not wrapped in parens
@@ -980,15 +974,31 @@ func (n *SetOperationScanNode) FormatSQL(ctx context.Context) (string, error) {
 	case ast.SetOperationTypeExceptDistinct:
 		opType = "EXCEPT"
 	default:
-		opType = "UNKONWN"
+		opType = "UNKNOWN"
 	}
 	var queries []string
 	for _, item := range n.node.InputItemList() {
+		var outputColumns []string
+		for _, outputColumn := range item.OutputColumnList() {
+			outputColumns = append(outputColumns, fmt.Sprintf("`%s`", uniqueColumnName(ctx, outputColumn)))
+		}
 		query, err := newNode(item).FormatSQL(ctx)
 		if err != nil {
 			return "", err
 		}
-		queries = append(queries, query)
+
+		formattedInput, err := formatInput(query)
+		if err != nil {
+			return "", err
+		}
+
+		queries = append(
+			queries,
+			fmt.Sprintf("SELECT %s %s",
+				strings.Join(outputColumns, ", "),
+				formattedInput,
+			),
+		)
 	}
 	columnMaps := []string{}
 	if len(n.node.InputItemList()) != 0 {
@@ -1354,11 +1364,35 @@ func (n *ExplainStmtNode) FormatSQL(ctx context.Context) (string, error) {
 	return "", nil
 }
 
+// FormatSQL Formats the outermost query statement that runs and produces rows of output, like a SELECT
+// The node's `OutputColumnList()` gives user-visible column names that should be returned. There may be duplicate names,
+// and multiple output columns may reference the same column from `Query()`
+// https://github.com/google/zetasql/blob/master/docs/resolved_ast.md#ResolvedQueryStmt
 func (n *QueryStmtNode) FormatSQL(ctx context.Context) (string, error) {
 	if n.node == nil {
 		return "", nil
 	}
-	return newNode(n.node.Query()).FormatSQL(ctx)
+	input, err := newNode(n.node.Query()).FormatSQL(ctx)
+	if err != nil {
+		return "", err
+	}
+
+	var columns []string
+	for _, outputColumnNode := range n.node.OutputColumnList() {
+		columns = append(
+			columns,
+			fmt.Sprintf("`%s` AS `%s`",
+				uniqueColumnName(ctx, outputColumnNode.Column()),
+				outputColumnNode.Name(),
+			),
+		)
+	}
+
+	return fmt.Sprintf(
+		"SELECT %s FROM (%s)",
+		strings.Join(columns, ", "),
+		input,
+	), nil
 }
 
 func (n *CreateDatabaseStmtNode) FormatSQL(ctx context.Context) (string, error) {
diff --git a/query_test.go b/query_test.go
index 4905f7c..929bb9d 100644
--- a/query_test.go
+++ b/query_test.go
@@ -39,6 +39,27 @@ func TestQuery(t *testing.T) {
 		expectedRows [][]interface{}
 		expectedErr  string
 	}{
+		// Regression test for https://github.com/goccy/go-zetasqlite/issues/191
+		{
+			name: "distinct union",
+			query: `WITH toks AS (SELECT true AS x, 1 AS y)
+					SELECT DISTINCT x, x as y FROM toks`,
+			expectedRows: [][]interface{}{{true, true}},
+		},
+		{
+			name: "with scan union all",
+			query: `(WITH toks AS (SELECT 1 AS x) SELECT x FROM toks)
+UNION ALL
+(WITH toks2 AS (SELECT 2 AS x) SELECT x FROM toks2)`,
+			expectedRows: [][]interface{}{{int64(1)}, {int64(2)}},
+		},
+		{
+			name: "having with union all",
+			query: `(WITH toks AS (SELECT 1 AS x) SELECT COUNT(x) AS total_rows FROM toks WHERE x > 0 HAVING total_rows >= 0)
+UNION ALL
+(WITH toks2 AS (SELECT 2 AS x) SELECT COUNT(x) AS total_rows FROM toks2 WHERE x > 0 HAVING total_rows >= 0)`,
+			expectedRows: [][]interface{}{{int64(1)}, {int64(1)}},
+		},
 		// priority 2 operator
 		{
 			name:         "unary plus operator",