Difference between data stream window function and cep within

[Joshua Fan]

Hi All,

I'd like to know the difference between data stream window function and cep within, I googled this issue but found no useful information.

Below the cep within, is there a tumbling window or sliding window or just a process function?

Your explanation will be truly appreciated.

Yours sincerely

Joshua

[Dian Fu]

Hi Joshua,

There is no tumbling/sliding window underlying the cep within implementation. 

The difference between datastream window and cep within is that:

1) Regarding to datastream window, the window is unified for all the elements (You can think that the window already exists before the input elements come). For example, for sliding window: (window size: 60s, slide size: 10s), then the windows will be [0s, 60s], [10s, 70s], [20s, 80s], etc. When the input elements come, they are put into the windows they belong to.

2) Regarding to cep within, it defines the maximum time interval for an event sequence to match the pattern. So a unified window is not suitable for this requirement. Regarding to the underlying implementation, for each matching/partial-matching sequence, the time interval between the first element and the last element of the sequence will be checked against the within interval. You can refer to [1] for details.

Regards,

Dian

[1] https://github.com/apache/flink/blob/459fd929399ad6c80535255eefa278564ec33683/flink-libraries/flink-cep/src/main/java/org/apache/flink/cep/nfa/NFA.java#L251

在 2019年9月17日，下午7:47，Joshua Fan <[hidden email]> 写道：

Hi All,

I'd like to know the difference between data stream window function and cep within, I googled this issue but found no useful information.

Below the cep within, is there a tumbling window or sliding window or just a process function?

Your explanation will be truly appreciated.

Yours sincerely

Joshua

[Joshua Fan]

Hi Dian

Thank you for your explanation.

After have a look at the source code, the cep within just executes by a time interval according to each state.

Thank you.

Yours sincerely

Joshua

On Wed, Sep 18, 2019 at 9:41 AM Dian Fu <[hidden email]> wrote:

Hi Joshua,

There is no tumbling/sliding window underlying the cep within implementation. 

The difference between datastream window and cep within is that:

1) Regarding to datastream window, the window is unified for all the elements (You can think that the window already exists before the input elements come). For example, for sliding window: (window size: 60s, slide size: 10s), then the windows will be [0s, 60s], [10s, 70s], [20s, 80s], etc. When the input elements come, they are put into the windows they belong to.

2) Regarding to cep within, it defines the maximum time interval for an event sequence to match the pattern. So a unified window is not suitable for this requirement. Regarding to the underlying implementation, for each matching/partial-matching sequence, the time interval between the first element and the last element of the sequence will be checked against the within interval. You can refer to [1] for details.

Regards,

Dian

[1] https://github.com/apache/flink/blob/459fd929399ad6c80535255eefa278564ec33683/flink-libraries/flink-cep/src/main/java/org/apache/flink/cep/nfa/NFA.java#L251

在 2019年9月17日，下午7:47，Joshua Fan <[hidden email]> 写道：

Hi All,

I'd like to know the difference between data stream window function and cep within, I googled this issue but found no useful information.

Below the cep within, is there a tumbling window or sliding window or just a process function?

Your explanation will be truly appreciated.

Yours sincerely

Joshua