Hi,
I guess this is more of a Java Problem than a Flink Problem. If you want it
quick and dirty you could implement a class such as:
public class Value {
private boolean isLongSet = false;
private long longValue = 0L;
private boolean isIntegerSet = false;
private int intValue = 0;
public Value(final long value) {
setLong(value);
}
public void setLong(final long value) |
longValue = value;
isLongSet = true;
}
public long getLong() {
if(isLongSet) {
return longValue
}
}
// Add same methods for int
// to satisfy POJO requirements you will also need to add a no-argument
constructor as well as getters and setters for the boolean flags
}
I guess a cleaner solution would be possible using a custom Kryo serializer as
explained here:
https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/custom_serializers.html
Regards
Klemens
> Am 20.04.2021 um 10:34 schrieb Miguel Araújo <miguelaraujo...@gmail.com>:
>
> Hi everyone,
>
> I have a ProcessFunction which needs to store different number types for
> different keys, e.g., some keys need to store an integer while others need to
> store a double.
>
> I tried to use java.lang.Number as the type for the ValueState, but I got the
> expected "No fields were detected for class java.lang.Number so it cannot be
> used as a POJO type and must be processed as GenericType."
>
> I have the feeling that this is not the right approach, but the exact type to
> be stored is only known at runtime which makes things a bit trickier. Is
> there a way to register these classes correctly, or Is it preferable to use
> different ValueState's for different types?
>
> Thanks,
> Miguel
