Hi!
It seems that what you want is for *each* row compute the sum of the five
previous rows (including the current row). This is the use case of over
aggregation, not sliding window. I don't know if table API supports this,
but you can see [1] for over aggregation in Flink SQL.
But if you insist
Hi. Thank you for the clarification.
I updated my code as below and got the desired result.
result = table.window(Slide.over(
row_interval(WINDOW_SIZE)).every(row_interval(WINDOW_SLIDE)).on(col('proctime')).alias("w"))
\
.group_by(col('w')) \
.select(call(read_raw_data, co
Hi!
You're not only grouping by the over window but also grouping by the value,
thus only the records with the same value will be in the same group. I
guess this is no intended.
Long Nguyễn 于2021年11月2日周二 上午3:05写道：
> I have set up a program that takes bits 0 and 1 from a Kafka topic and
> then u
I have set up a program that takes bits 0 and 1 from a Kafka topic and then
uses Flink to create a sliding count window of size 5. In that window, I'd
like to output 1 if there are 3 or more of the bit 1, otherwise, output 0.
Currently, I follow the way of calculating the sum of bits in the window.
