Jian Yun

@omidb I have received this suggestion. We will evaluate and schedule. thank you so much for your suggestion.

Hello, I'm sorry it took so long to get back to you. When this pulsar-flink connector implements the codec of the data, the codec capability is put into Pulsar to...

@subbu107 The Producer in the sink is sent asynchronously. It is sent in snapshotState. There is an asynchronous queue in the Producer. In the case of back pressure, it is...

I am sorry that the documentation for this part is not good enough. You can refer to the following code to solve the problem. ```java PulsarDeserializationSchema pulsarDeserialization = PulsarDeserializationSchema.valueOnly(JsonDeser.of(LiveRawData.class)); FlinkPulsarSource...

@syhily We can update the documentation to explain how to use json, but `JsonDeser` is something that needs to be deprecated.

@nlu90 In upsert-pulsar, the `key.fields` and `key.fields-prefix` options will no longer be used, you need to use the `PRIMARY KEY` statement ```sql CREATE TABLE some_table ( `order_id` STRING, `transaction_time` STRING,...

Another issue, it seems that `PulsarDynamicTableFactory` was created, suggest to recheck and test it. I was unable to reproduce the problem.

In key-shard mode consumed as a subscriber, the range is persisted by creating a subscriber `test[0,1000]`. The original progress is restored by reading this subscription at startup time. I will...

The expected value is changed to `+[2,1]` and that works. The reason is that Flink version upgrade added `RowKind` field for Row to mark this data as inserted, deleted, etc....

In the test, often use `Row` or `RowData` `toString()` method to compare, `Row.toString()` method to add this `rowKind` field occurred in flink version 1.13