[
https://issues.apache.org/jira/browse/FLINK-6740?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
sunjincheng updated FLINK-6740:
-------------------------------
Description:
When we define UDTF as follows:
{code}
class TableFuncPojo extends TableFunction[TPojo] {
def eval(age: Int, name:String): Unit = {
collect(new TPojo(age.toLong,name))
}
def eval(age: Date, name:String): Unit = {
collect(new
TPojo(org.apache.calcite.runtime.SqlFunctions.toInt(age).toLong,name))
}
}
{code}
TableAPI:
{code}
val table = stream.toTable(tEnv,
'long2, 'int, 'double, 'float, 'bigdec, 'ts, 'date,'pojo, 'string,
'long.rowtime)
val windowedTable = table
.join(udtf('date, 'string) as 'pojo2).select('pojo2)
{code}
We will get the error as following:
{code}
org.apache.flink.table.api.ValidationException: Found multiple 'eval' methods
which match the signature.
at
org.apache.flink.table.functions.utils.UserDefinedFunctionUtils$.getUserDefinedMethod(UserDefinedFunctionUtils.scala:180)
at
org.apache.flink.table.plan.logical.LogicalTableFunctionCall.validate(operators.scala:700)
at org.apache.flink.table.api.Table.join(table.scala:539)
at org.apache.flink.table.api.Table.join(table.scala:328)
at
org.apache.flink.table.runtime.datastream.DataStreamAggregateITCase.test1(DataStreamAggregateITCase.scala:84)
{code}
The reason is in `parameterTypeEquals` method, logical as follows:
{code}
candidate == classOf[Date] && (expected == classOf[Int] || expected ==
classOf[JInt])
{code}
I think we can modify the logical of `parameterTypeEquals` method.
What do you think? Welcome anybody feedback...
was:
When we define UDTF as follows:
{code}
class TableFuncPojo extends TableFunction[TPojo] {
def eval(age: Int, name:String): Unit = {
collect(new TPojo(age.toLong,name))
}
def eval(age: Date, name:String): Unit = {
collect(new
TPojo(org.apache.calcite.runtime.SqlFunctions.toInt(age).toLong,name))
}
}
{code}
TableAPI:
{code}
val table = stream.toTable(tEnv,
'long2, 'int, 'double, 'float, 'bigdec, 'ts, 'date,'pojo, 'string,
'long.rowtime)
val windowedTable = table
.join(udtf('date, 'string) as 'pojo2).select('pojo2)
{code}
We will get the error as following:
{code}
org.apache.flink.table.api.ValidationException: Found multiple 'eval' methods
which match the signature.
at
org.apache.flink.table.functions.utils.UserDefinedFunctionUtils$.getUserDefinedMethod(UserDefinedFunctionUtils.scala:180)
at
org.apache.flink.table.plan.logical.LogicalTableFunctionCall.validate(operators.scala:700)
at org.apache.flink.table.api.Table.join(table.scala:539)
at org.apache.flink.table.api.Table.join(table.scala:328)
at
org.apache.flink.table.runtime.datastream.DataStreamAggregateITCase.test1(DataStreamAggregateITCase.scala:84)
{code}
The reason is in {{ parameterTypeEquals }} method, logical as follows:
{code}
candidate == classOf[Date] && (expected == classOf[Int] || expected ==
classOf[JInt])
{code}
But sometimes user really need two eval
So, We should modify the logical of {{ parameterTypeEquals }} method.
What do you think? Welcome anybody feedback...
> Fix "parameterTypeEquals" method error.
> ---------------------------------------
>
> Key: FLINK-6740
> URL: https://issues.apache.org/jira/browse/FLINK-6740
> Project: Flink
> Issue Type: Sub-task
> Components: Table API & SQL
> Affects Versions: 1.3.0
> Reporter: sunjincheng
> Assignee: sunjincheng
>
> When we define UDTF as follows:
> {code}
> class TableFuncPojo extends TableFunction[TPojo] {
> def eval(age: Int, name:String): Unit = {
> collect(new TPojo(age.toLong,name))
> }
> def eval(age: Date, name:String): Unit = {
> collect(new
> TPojo(org.apache.calcite.runtime.SqlFunctions.toInt(age).toLong,name))
> }
> }
> {code}
> TableAPI:
> {code}
> val table = stream.toTable(tEnv,
> 'long2, 'int, 'double, 'float, 'bigdec, 'ts, 'date,'pojo, 'string,
> 'long.rowtime)
> val windowedTable = table
> .join(udtf('date, 'string) as 'pojo2).select('pojo2)
> {code}
> We will get the error as following:
> {code}
> org.apache.flink.table.api.ValidationException: Found multiple 'eval' methods
> which match the signature.
> at
> org.apache.flink.table.functions.utils.UserDefinedFunctionUtils$.getUserDefinedMethod(UserDefinedFunctionUtils.scala:180)
> at
> org.apache.flink.table.plan.logical.LogicalTableFunctionCall.validate(operators.scala:700)
> at org.apache.flink.table.api.Table.join(table.scala:539)
> at org.apache.flink.table.api.Table.join(table.scala:328)
> at
> org.apache.flink.table.runtime.datastream.DataStreamAggregateITCase.test1(DataStreamAggregateITCase.scala:84)
> {code}
> The reason is in `parameterTypeEquals` method, logical as follows:
> {code}
> candidate == classOf[Date] && (expected == classOf[Int] || expected ==
> classOf[JInt])
> {code}
> I think we can modify the logical of `parameterTypeEquals` method.
> What do you think? Welcome anybody feedback...
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
