Github user mxm commented on the pull request:
https://github.com/apache/flink/pull/931#issuecomment-126218526
I think we can merge this later on.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not
Github user asfgit closed the pull request at:
https://github.com/apache/flink/pull/931
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is
Github user mxm commented on the pull request:
https://github.com/apache/flink/pull/931#issuecomment-125979125
Thanks for the pull request @zentol!
+1 for removing the dill library. As far as I can see, we handle all the
serialization ourselves now. We only used the Dill
Github user zentol commented on the pull request:
https://github.com/apache/flink/pull/931#issuecomment-125983186
Thanks for the review @mxm .
I've addressed the cosmetic issue you mentioned, and added a small fix for
a separate issue as well (error reporting was partially
Github user mxm commented on a diff in the pull request:
https://github.com/apache/flink/pull/931#discussion_r35763245
--- Diff:
flink-staging/flink-language-binding/flink-python/src/main/java/org/apache/flink/languagebinding/api/java/python/streaming/PythonStreamer.java
---
@@
GitHub user zentol opened a pull request:
https://github.com/apache/flink/pull/931
[FLINK-1927][py] Operator distribution rework
Python operators are no longer serialized and shipped across the
cluster. Instead the plan file is executed on each node, followed by
usage of the
Github user zentol closed the pull request at:
https://github.com/apache/flink/pull/638
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is
Github user zentol commented on the pull request:
https://github.com/apache/flink/pull/638#issuecomment-103813351
Implementing this in a clean way has become trickier that i initially
expected, as such I'll postpone it and close this PR for now.
---
If your project is set up for it,
Github user mxm commented on the pull request:
https://github.com/apache/flink/pull/638#issuecomment-97353102
Wow! Great to see that we get rid of the only python-side dependency. It
was a bit unclear under which terms we could ship the library anyways. Have you
done any measurements
Github user zentol commented on the pull request:
https://github.com/apache/flink/pull/638#issuecomment-97389037
I didn't check performance, it shouldn't have any noticeable effect on it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on
Github user zentol commented on the pull request:
https://github.com/apache/flink/pull/638#issuecomment-97417603
@aljoscha that variable must be declared somewhere within the plan file.
during the plan rebuild this would be executed as well, so i don't think this
is a problem. in
Github user aljoscha commented on the pull request:
https://github.com/apache/flink/pull/638#issuecomment-97412959
But doesn't this mean that the lambdas now must be stateless, i.e. if a
user refers to some variable outside the lambda this will not be serialised
with the closure
Github user zentol commented on the pull request:
https://github.com/apache/flink/pull/638#issuecomment-97418740
oh snap i just noticed a big flaw... well let's put this PR on hold for a
bit.
I'm simply re executing the plan file on each node, but forgot to deal with
GitHub user zentol opened a pull request:
https://github.com/apache/flink/pull/638
[FLINK-1927] [py] Operator distribution rework
Python operators are no longer serialized and instead rebuilt on each node.
This also means that the dill library is no longer necessary.
You can merge
14 matches
Mail list logo