Mailing-List: contact user-help@flink.apache.org; run by ezmlm
Precedence: bulk
MIME-Version: 1.0
From: Vishal Santoshi <vishal.santoshi@gmail.com>
Date: Sun, 17 Dec 2017 01:52:20 -0500
Message-ID: <CAMq=OU7hhGXuEO4FKFFaXYbGwj+qOu7Dr56p49dDqHKtY6nW9g@mail.gmail.com>
Subject: A question about Triggers
To: user <user@flink.apache.org>
Content-Type: multipart/alternative; boundary="001a11c1829edb8087056083ac3e"
archived-at: Sun, 17 Dec 2017 06:52:35 -0000

--001a11c1829edb8087056083ac3e
Content-Type: text/plain; charset="UTF-8"

I want to augment a POJO in  Trigger's onElement method, specifically
supply the POJO with the watermark from the TriggerContext. The sequence of
execution is this sequence

1. call to add() in the accumulator for the window  and save the POJO
reference in the Accumulator.
2. call to onElement on Tigger
3. set watermark to the POJO

The next add() method should have the last reference and any mutation done
in step 3.

That works in a local test case, using LocalFlinkMiniCluster, as in I have
access to the mutation by the onElement() in the POJO in the subsequent
add(),  but not on a distributed cluster. The specific question I had is
whether  add() on a supplied accumulator on a window and onElement() method
of the trigger on that window are inline executions, on the same thread or
is there any serialization/deserialization IPC that causes these divergence
( local versus distributed )

Regards.

--001a11c1829edb8087056083ac3e
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">I want to augment a POJO in=C2=A0 Trigger&#39;s onElement =
method, specifically supply the POJO with the watermark from the TriggerCon=
text. The sequence of execution is this sequence<div><br></div><div>1. call=
 to add() in the accumulator for the window=C2=A0 and save the POJO=C2=A0 r=
eference in the Accumulator.</div><div>2. call to onElement on Tigger=C2=A0=
</div><div>3. set watermark to the POJO</div><div><br></div><div>The next a=
dd() method should have the last reference and any mutation done in step 3.=
</div><div><br></div><div>That works in a local test case, using LocalFlink=
MiniCluster, as in I have access to the mutation by the onElement() in the =
POJO in the subsequent add(),=C2=A0 but not on a distributed cluster. The s=
pecific question I had is whether=C2=A0 add() on a supplied accumulator on =
a window and onElement() method of the trigger on that window are inline ex=
ecutions, on the same thread or is there any serialization/deserialization =
IPC that causes these divergence ( local versus distributed )=C2=A0<br></di=
v><div><br></div><div>Regards.</div></div>

--001a11c1829edb8087056083ac3e--