|
The current plugin code sets the ColType in the projection RowGroup for UDAF using UDAF's first parameter. This might produce errors with a UDAF that has multiple arguments. Consider the example:
MariaDB [tpch1]> select * from (select CCHAR9, regr_count(CDECIMAL9_2, length(CCHAR9)) f from datatypetestm1 group by 1) a;
|
+-----------+------+
|
| CCHAR9 | f |
|
+-----------+------+
|
| rrrrrrrr | 0 |
|
| rrrrrrr | 0 |
|
| iii | 0 |
|
| i | 0 |
|
| ii | 0 |
|
| iiii | 0 |
|
| zzzzzzzzz | 0 |
|
| aaaaaaaaa | 0 |
|
| rrrrrrrrr | 0 |
|
| NULL | 0 |
|
+-----------+------+
|
10 rows in set (0.021 sec)
|
|
MariaDB [tpch1]> select * from (select CCHAR9, regr_count(CDECIMAL9_2, length(CCHAR9)) f from datatypetestm1 where CCHAR9 = 'ii' group by 1) a;
|
+--------+------+
|
| CCHAR9 | f |
|
+--------+------+
|
| ii | 0 |
|
+--------+------+
|
1 row in set (0.033 sec)
|
The f column contains 0 b/c f's ColType is a DECIMAL(9,2) so that Decimal TypeHandler first prints f value as '0.2' and stores it in the MDB's Field. MDB rounds the result to the nearest INT value that is 0 storing the string representation of the DECIMAL(9,2).
Here is what MCS must return.
MariaDB [test]> select * from (select CCHAR9, regr_count(CDECIMAL9_2, length(CCHAR9)) f from datatypetestm1 group by 1) a;
|
+-----------+------+
|
| CCHAR9 | f |
|
+-----------+------+
|
| rrrrrrrr | 2 |
|
| rrrrrrr | 2 |
|
| iii | 2 |
|
| i | 2 |
|
| ii | 2 |
|
| iiii | 2 |
|
| zzzzzzzzz | 2 |
|
| aaaaaaaaa | 2 |
|
| rrrrrrrrr | 4 |
|
| NULL | 0 |
|
+-----------+------+
|
10 rows in set (0.117 sec)
|
|
MariaDB [test]> select * from (select CCHAR9, regr_count(CDECIMAL9_2, length(CCHAR9)) f from datatypetestm1 where CCHAR9 = 'ii' group by 1) a;
|
+--------+------+
|
| CCHAR9 | f |
|
+--------+------+
|
| ii | 2 |
|
+--------+------+
|
1 row in set (0.037 sec)
|
|
MariaDB [test]>
|
|