Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
5.5.1, 5.6.1, 6.1.1
-
None
Description
The problem described below is also with narrow and wide DECIMAL.
The method TreeNode::getIntVal() performs truncation instead of rounding when converting from DECIMAL to signed integer:
case CalpontSystemCatalog::DECIMAL: |
case CalpontSystemCatalog::UDECIMAL: |
{
|
if (fResultType.colWidth == datatypes::MAXDECIMALWIDTH) |
{
|
return static_cast<int64_t>(fResult.decimalVal.getIntegralPart()); |
}
|
else |
{
|
return (int64_t)(fResult.decimalVal.value / pow((double)10, fResult.decimalVal.scale)); |
}
|
}
|
This makes ColumnStore return MariaDB incompatible values:
RAND()
DROP TABLE IF EXISTS t1; |
CREATE TABLE t1 (a DECIMAL(10,1)) ENGINE=ColumnStore; |
INSERT INTO t1 VALUES (1.4),(1.5); |
SELECT a, RAND(a), RAND(CAST(a AS DECIMAL(10,1))) FROM t1; |
+------+---------------------+--------------------------------+
|
| a | RAND(a) | RAND(CAST(a AS DECIMAL(10,1))) |
|
+------+---------------------+--------------------------------+
|
| 1.4 | 0.40540353712197724 | 0.40540353712197724 |
|
| 1.5 | 0.40540353712197724 | 0.40540353712197724 |
|
+------+---------------------+--------------------------------+
|
Looks wrong. RAND() did not round the seed value from 1.5 to 2.
The expected result is:
DROP TABLE IF EXISTS t1; |
CREATE TABLE t1 (a DECIMAL(10,1)) ENGINE=InnoDB; |
INSERT INTO t1 VALUES (1.4),(1.5); |
SELECT a, RAND(a), RAND(CAST(a AS DECIMAL(10,1))) FROM t1; |
+------+---------------------+--------------------------------+
|
| a | RAND(a) | RAND(CAST(a AS DECIMAL(10,1))) |
|
+------+---------------------+--------------------------------+
|
| 1.4 | 0.40540353712197724 | 0.40540353712197724 |
|
| 1.5 | 0.6555866465490187 | 0.6555866465490187 |
|
+------+---------------------+--------------------------------+
|
TRUNCATE()
DROP TABLE IF EXISTS t1; |
CREATE TABLE t1 (a DECIMAL(10,1)) ENGINE=ColumnStore; |
INSERT INTO t1 VALUES (1.4),(1.5); |
SELECT TRUNCATE(1.2345, a) FROM t1; |
+---------------------+
|
| TRUNCATE(1.2345, a) |
|
+---------------------+
|
| 1.2000 |
|
| 1.2000 |
|
+---------------------+
|
The expected result is:
DROP TABLE IF EXISTS t1; |
CREATE TABLE t1 (a DECIMAL(10,1)) ENGINE=InnoDB; |
INSERT INTO t1 VALUES (1.4),(1.5); |
SELECT TRUNCATE(1.2345, a) FROM t1; |
+---------------------+
|
| TRUNCATE(1.2345, a) |
|
+---------------------+
|
| 1.2000 |
|
| 1.2300 |
|
+---------------------+
|
CONV()
DROP TABLE IF EXISTS t1; |
CREATE TABLE t1 (a DECIMAL(10,1)) ENGINE=ColumnStore; |
INSERT INTO t1 VALUES (15.4),(15.5),(16.0); |
SELECT a, CONV(16,a,10) FROM t1; |
+------+---------------+
|
| a | CONV(16,a,10) |
|
+------+---------------+
|
| 15.4 | 21 |
|
| 15.5 | 21 | <--- this is wrong
|
| 16.0 | 22 |
|
+------+---------------+
|
DROP TABLE IF EXISTS t1; |
CREATE TABLE t1 (a DECIMAL(10,1)) ENGINE=InnoDB; |
INSERT INTO t1 VALUES (15.4),(15.5),(16.0); |
SELECT a, CONV(16,a,10) FROM t1; |
+------+---------------+
|
| a | CONV(16,a,10) |
|
+------+---------------+
|
| 15.4 | 21 |
|
| 15.5 | 22 |
|
| 16.0 | 22 |
|
+------+---------------+
|
