Thank you for the tips. Unfortunately, the guides don't follow the same case since I'm not using SQL DDL and I don't convert from DataStream to Table, I register the DataStream as a table in the SQL API.
I have a custom DataStream source, which I have as DataStream<Row>. The time characteristic is normalized time. I call inputStream .assignTimestampsAndWatermarks(watermarkStrategy). Then, I call tableEnv.registerDatastream(). In this process, I don't see where I could set the proctime as the time attribute. Also, I'm currently using TimeIndicatorTypeInfo.ROWTIME_INDICATOR as the type, so I'm unsure how to cast it to TIMESTAMP(3). On Mon, Aug 2, 2021 at 7:18 PM JING ZHANG <beyond1...@gmail.com> wrote: > Hi Pranav, > Yes, The root cause is the `timecol` is not a time attribute column. > If you use processing time as time attribute, please refer [1] for more > information. > If you use. event time as time attribute, please refer[2] for more > information. And only if choose event time, `assignTimestampsAndWatermarks` > is needed. > If the problem still exists after you update the program based on the demo > in the document, please let me know and provide your program. > > [1] > https://ci.apache.org/projects/flink/flink-docs-release-1.13/docs/dev/table/concepts/time_attributes/#processing-time > [2] > https://ci.apache.org/projects/flink/flink-docs-release-1.13/docs/dev/table/concepts/time_attributes/#event-time > > Best, > JING ZHANG > > > Pranav Patil <pranav.pa...@salesforce.com> 于2021年8月3日周二 上午8:51写道： > >> Hi, >> >> I'm upgrading a repository from Flink 1.11 to Flink 1.13. I have Flink >> SQL command that used to do tumbling windows using the following in the >> GROUP BY clause: >> >> SELECT ... FROM ... GROUP BY TUMBLE(proctime, INTERVAL '1' MINUTE) >> >> However now, it gives me the error: >> >> org.apache.flink.table.api.TableException: Window aggregate can only be >> defined over a time attribute column, but TIMESTAMP(9) encountered. >> >> I'm not sure why this isn't a time attribute column anymore. Thinking it's a >> syntax change, I try: >> >> Table(TUMBLE(Table inputStream,DESCRIPTOR(proctime),INTERVAL '1' MINUTE)) >> >> This too doesn't work, with the error: >> >> org.apache.flink.table.api.ValidationException: The window function >> TUMBLE(TABLE table_name, DESCRIPTOR(timecol), datetime interval) requires >> the timecol is a time attribute type, but is TIMESTAMP(9). >> >> So I'm realizing the problem is actually that it isn't a time attribute >> column. However, I'm confused because this wasn't a problem in previous >> versions. In the DataStream API file source, assignTimestampsAndWatermarks >> is called on the source stream, which I believed to be enough. I then call >> registerDatastream to access it from Flink SQL. Are there additional steps >> that must be taken in Flink 1.13? >> >> Thank you. >> >>
