[
https://issues.apache.org/jira/browse/CALCITE-5877?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mihai Budiu updated CALCITE-5877:
---------------------------------
Description:
Adding the following test to RelToSqlConverterTest produces an AssertionFailure:
{code:java}
@Test void testNumericScaleMod() {
final String sql = "SELECT MOD(CAST(2 AS DECIMAL(39, 20)), 2)";
final String expected = "";
sql(sql).withPostgresqlModifiedDecimalTypeSystem()
.ok(expected);
}
{code}
The exception is here:
{code:java}
java.lang.AssertionError
at
org.apache.calcite.rel.type.RelDataTypeSystem.deriveDecimalModType(RelDataTypeSystem.java:383)
at
org.apache.calcite.sql.type.ReturnTypes.lambda$static$11(ReturnTypes.java:849)
at
org.apache.calcite.sql.type.SqlTypeTransformCascade.inferReturnType(SqlTypeTransformCascade.java:58)
at
org.apache.calcite.sql.type.SqlReturnTypeInferenceChain.inferReturnType(SqlReturnTypeInferenceChain.java:55)
{code}
In this line:
{code:java}
int scale = Math.max(s1, s2);
assert scale <= getMaxNumericScale();
{code}
I suspect that doing a Math.min with the getMaxNumericScale (instead of an
Assert) is sufficient to fix this particular problem. The question is whether
the compiler should really enforce the invariant that this assertion checks.
was:
Adding the following test to RelToSqlConverterTest produces an AssertionFailure:
{code:java}
@Test void testNumericScaleMod() {
final String sql = "SELECT MOD(CAST(2 AS DECIMAL(39, 20)), 2)";
final String expected = "";
sql(sql).withPostgresqlModifiedDecimalTypeSystem()
.ok(expected);
}
{code}
The exception is here:
{code:java}
java.lang.AssertionError
at
org.apache.calcite.rel.type.RelDataTypeSystem.deriveDecimalModType(RelDataTypeSystem.java:383)
at
org.apache.calcite.sql.type.ReturnTypes.lambda$static$11(ReturnTypes.java:849)
at
org.apache.calcite.sql.type.SqlTypeTransformCascade.inferReturnType(SqlTypeTransformCascade.java:58)
at
org.apache.calcite.sql.type.SqlReturnTypeInferenceChain.inferReturnType(SqlReturnTypeInferenceChain.java:55)
{code}
In this line:
{code:java}
int scale = Math.max(s1, s2);
assert scale <= getMaxNumericScale();
{code}
I suspect that doing a Math.max with the getMaxNumericScale (instead of an
Assert) is sufficient to fix this particular problem. The question is whether
the compiler should really enforce the invariant that this assertion checks.
> Calcite crashes with AssertionError when compiling MOD with arguments with
> large scales
> ---------------------------------------------------------------------------------------
>
> Key: CALCITE-5877
> URL: https://issues.apache.org/jira/browse/CALCITE-5877
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.35.0
> Reporter: Mihai Budiu
> Priority: Minor
> Labels: pull-request-available
> Fix For: 1.36.0
>
>
> Adding the following test to RelToSqlConverterTest produces an
> AssertionFailure:
> {code:java}
> @Test void testNumericScaleMod() {
> final String sql = "SELECT MOD(CAST(2 AS DECIMAL(39, 20)), 2)";
> final String expected = "";
> sql(sql).withPostgresqlModifiedDecimalTypeSystem()
> .ok(expected);
> }
> {code}
> The exception is here:
> {code:java}
> java.lang.AssertionError
> at
> org.apache.calcite.rel.type.RelDataTypeSystem.deriveDecimalModType(RelDataTypeSystem.java:383)
> at
> org.apache.calcite.sql.type.ReturnTypes.lambda$static$11(ReturnTypes.java:849)
> at
> org.apache.calcite.sql.type.SqlTypeTransformCascade.inferReturnType(SqlTypeTransformCascade.java:58)
> at
> org.apache.calcite.sql.type.SqlReturnTypeInferenceChain.inferReturnType(SqlReturnTypeInferenceChain.java:55)
> {code}
> In this line:
> {code:java}
> int scale = Math.max(s1, s2);
> assert scale <= getMaxNumericScale();
> {code}
> I suspect that doing a Math.min with the getMaxNumericScale (instead of an
> Assert) is sufficient to fix this particular problem. The question is whether
> the compiler should really enforce the invariant that this assertion checks.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
