Re:May I ask some questions about Flink ListView ？

[Xuyang]

Hi, Apollo.

The processing of agg is generated by codegen. You can observe the processing 
log of the generated agg class by setting the following parameters:

env.java.opts.taskmanager: "-Dorg.codehaus.janino.source_debugging.enable=true 
-Dorg.codehaus.janino.source_debugging.dir=/flink/log/"




--

    Best！
    Xuyang




At 2024-09-21 14:21:14, "Apollo Elon" <apolloe...@gmail.com> wrote:

When I tried to calculate the median of the age field in a table with 120 
million rows, I implemented a custom UDAF. However, there was a significant 
performance difference between the two different types of  accumulator. The 
ListView stated that it would enable the state backend when encountering large 
amounts of data. How can I observe this process and which classes are 
responsible for these functions?
```java
public static class State implements Serializable {

public int scale = 2;

@DataTypeHint(value = "ARRAY<DOUBLE>")
public ArrayList<Double> numbers;

public State() {}
}

@Override
public State createAccumulator() {
    State state = new State();
    state.numbers = new ArrayList<>();
return state;
}


```java
public static class State implements Serializable {

public int scale = 2;

public ListView<Double> numbers;

public State() {}
}

@Override
public State createAccumulator() {
    State state = new State();
    state.numbers = new ListView<>();
return state;
}