[
https://issues.apache.org/jira/browse/CALCITE-7269?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dmitry Sysolyatin updated CALCITE-7269:
---------------------------------------
Description:
Test case [1]
{code:java}
@Test void testLambdaExpressionWithRowParameter() {
final String sql = "select \"EXISTS\"(array(ROW(true, false)), x ->
x.\"EXPR$1\")";
fixture()
.withFactory(c ->
c.withOperatorTable(t ->
SqlValidatorTest.operatorTableFor(SqlLibrary.SPARK)))
.withSql(sql)
.ok();
}
{code}
Exception:
{code:java}
>From line 1, column 47 to line 1, column 56: Param 'X.EXPR$1' not found in
>lambda expression '`X` -> `X`.`EXPR$1`'
org.apache.calcite.runtime.CalciteContextException: From line 1, column 47 to
line 1, column 56: Param 'X.EXPR$1' not found in lambda expression '`X` ->
`X`.`EXPR$1`'
at
java.base/jdk.internal.reflect.DirectConstructorHandleAccessor.newInstance(DirectConstructorHandleAccessor.java:62)
at
java.base/java.lang.reflect.Constructor.newInstanceWithCaller(Constructor.java:502)
at
java.base/java.lang.reflect.Constructor.newInstance(Constructor.java:486)
at
org.apache.calcite.runtime.Resources$ExInstWithCause.ex(Resources.java:511)
at org.apache.calcite.sql.SqlUtil.newContextException(SqlUtil.java:960)
at org.apache.calcite.sql.SqlUtil.newContextException(SqlUtil.java:945)
{code}
There are three things about lambda implementation:
1. SqlLambdaScope.fullyQualified returns the parameter only if the
SqlIdentifier matches exactly. If we change it to just use getComponent(0),
then…
2. SqlLambdaScope sets lambda parameters to ANY. But ANY does not have fields.
3. LambdaOperandTypeChecker sets the actual types. I think this is not the
right place to set actual types [2]. (or it is ?)
[1]
[https://github.com/apache/calcite/commit/44bf468d867a6623e905c98ec8e3c64daa2711b7]
[2]
[https://github.com/apache/calcite/blob/7ecd358b2e4678e8c74e6689a235912361aa6a8a/core/src/main/java/org/apache/calcite/sql/type/OperandTypes.java#L1815-L1817]
was:
Test case [1]
{code:java}
@Test void testLambdaExpressionWithRowParameter() {
final String sql = "select \"EXISTS\"(array(ROW(true, false)), x ->
x.\"EXPR$1\")";
fixture()
.withFactory(c ->
c.withOperatorTable(t ->
SqlValidatorTest.operatorTableFor(SqlLibrary.SPARK)))
.withSql(sql)
.ok();
}
{code}
Exception:
{code:java}
>From line 1, column 47 to line 1, column 56: Param 'X.EXPR$1' not found in
>lambda expression '`X` -> `X`.`EXPR$1`'
org.apache.calcite.runtime.CalciteContextException: From line 1, column 47 to
line 1, column 56: Param 'X.EXPR$1' not found in lambda expression '`X` ->
`X`.`EXPR$1`'
at
java.base/jdk.internal.reflect.DirectConstructorHandleAccessor.newInstance(DirectConstructorHandleAccessor.java:62)
at
java.base/java.lang.reflect.Constructor.newInstanceWithCaller(Constructor.java:502)
at
java.base/java.lang.reflect.Constructor.newInstance(Constructor.java:486)
at
org.apache.calcite.runtime.Resources$ExInstWithCause.ex(Resources.java:511)
at org.apache.calcite.sql.SqlUtil.newContextException(SqlUtil.java:960)
at org.apache.calcite.sql.SqlUtil.newContextException(SqlUtil.java:945)
{code}
There are three things about lambda implementation:
1. SqlLambdaScope.fullyQualified returns the parameter only if the
SqlIdentifier matches exactly. If we change it to just use getComponent(0),
then…
2. SqlLambdaScope sets lambda parameters to ANY. But ANY does not have fields.
3. The most interesting part: LambdaOperandTypeChecker sets the actual types. I
think this is not the right place to set actual types [2]. (or it is ?)
[1]
[https://github.com/apache/calcite/commit/44bf468d867a6623e905c98ec8e3c64daa2711b7]
[2]
[https://github.com/apache/calcite/blob/7ecd358b2e4678e8c74e6689a235912361aa6a8a/core/src/main/java/org/apache/calcite/sql/type/OperandTypes.java#L1815-L1817]
> SqlValidator throws exception if lambda parameter is struct
> -----------------------------------------------------------
>
> Key: CALCITE-7269
> URL: https://issues.apache.org/jira/browse/CALCITE-7269
> Project: Calcite
> Issue Type: Bug
> Components: core
> Affects Versions: 1.41.0
> Reporter: Dmitry Sysolyatin
> Priority: Major
>
> Test case [1]
> {code:java}
> @Test void testLambdaExpressionWithRowParameter() {
> final String sql = "select \"EXISTS\"(array(ROW(true, false)), x ->
> x.\"EXPR$1\")";
> fixture()
> .withFactory(c ->
> c.withOperatorTable(t ->
> SqlValidatorTest.operatorTableFor(SqlLibrary.SPARK)))
> .withSql(sql)
> .ok();
> }
> {code}
> Exception:
> {code:java}
> From line 1, column 47 to line 1, column 56: Param 'X.EXPR$1' not found in
> lambda expression '`X` -> `X`.`EXPR$1`'
> org.apache.calcite.runtime.CalciteContextException: From line 1, column 47 to
> line 1, column 56: Param 'X.EXPR$1' not found in lambda expression '`X` ->
> `X`.`EXPR$1`'
> at
> java.base/jdk.internal.reflect.DirectConstructorHandleAccessor.newInstance(DirectConstructorHandleAccessor.java:62)
> at
> java.base/java.lang.reflect.Constructor.newInstanceWithCaller(Constructor.java:502)
> at
> java.base/java.lang.reflect.Constructor.newInstance(Constructor.java:486)
> at
> org.apache.calcite.runtime.Resources$ExInstWithCause.ex(Resources.java:511)
> at org.apache.calcite.sql.SqlUtil.newContextException(SqlUtil.java:960)
> at org.apache.calcite.sql.SqlUtil.newContextException(SqlUtil.java:945)
> {code}
> There are three things about lambda implementation:
> 1. SqlLambdaScope.fullyQualified returns the parameter only if the
> SqlIdentifier matches exactly. If we change it to just use getComponent(0),
> then…
> 2. SqlLambdaScope sets lambda parameters to ANY. But ANY does not have fields.
> 3. LambdaOperandTypeChecker sets the actual types. I think this is not the
> right place to set actual types [2]. (or it is ?)
> [1]
> [https://github.com/apache/calcite/commit/44bf468d867a6623e905c98ec8e3c64daa2711b7]
> [2]
> [https://github.com/apache/calcite/blob/7ecd358b2e4678e8c74e6689a235912361aa6a8a/core/src/main/java/org/apache/calcite/sql/type/OperandTypes.java#L1815-L1817]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
