You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@flink.apache.org by Lorenzo Affetti <lo...@gmail.com> on 2016/01/29 10:49:54 UTC

Issue on watermark meaning

Hi everybody,
I want to signal that I think there is a mismatch between what is the
meaning of emitting a watermark between the code and the documentation:

from Flink docs
<https://ci.apache.org/projects/flink/flink-docs-master/apis/streaming/index.html#working-with-time>:
A watermark with a certain timestamp denotes the knowledge that no event
with timestamp* lower than *the timestamp of the watermark will ever arrive.


from Flink code (Output.java)
<https://raw.githubusercontent.com/apache/flink/master/flink-streaming-java/src/main/java/org/apache/flink/streaming/api/operators/Output.java>
:
A watermark specifies that no element with a timestamp *older or equal* to
the watermark timestamp will be emitted in the future.

I think that the documentation should be fixed to match "lower or equal to"
because it is what the system really does, I think.

Thank you
-- 
Lorenzo Affetti

Re: Issue on watermark meaning

Posted by Till Rohrmann <tr...@apache.org>.
Hi Lorenzo,

you're right that we should stick to the same terminology between the
online documentation and the code, otherwise it's confusing. In this case,
though, a lower numeric timestamp is equivalent to an older event. The
older an element is, the lower is its timestamp.

However, there is a another problem with the formulation. In the first
sentence it's said no event with a lower (=older) timestamp whereas in the
second sentence it's additional said that no event with the same timestamp
as the watermark will be emitted. This is not consistent. We'll fix that.

Cheers,
Till

On Fri, Jan 29, 2016 at 10:49 AM, Lorenzo Affetti <lorenzo.affetti@gmail.com
> wrote:

> Hi everybody,
> I want to signal that I think there is a mismatch between what is the
> meaning of emitting a watermark between the code and the documentation:
>
> from Flink docs
> <https://ci.apache.org/projects/flink/flink-docs-master/apis/streaming/index.html#working-with-time>:
> A watermark with a certain timestamp denotes the knowledge that no event
> with timestamp* lower than *the timestamp of the watermark will ever
> arrive.
>
>
> from Flink code (Output.java)
> <https://raw.githubusercontent.com/apache/flink/master/flink-streaming-java/src/main/java/org/apache/flink/streaming/api/operators/Output.java>
> :
> A watermark specifies that no element with a timestamp *older or equal* to
> the watermark timestamp will be emitted in the future.
>
> I think that the documentation should be fixed to match "lower or equal
> to" because it is what the system really does, I think.
>
> Thank you
> --
> Lorenzo Affetti
>