I was thinking in the lines of using this as an alternative clustering /
distributed computing engine.
--
View this message in context:
http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/Apache-Ignite-tp5115p5199.html
Sent from the Apache Flink Mailing List archive. mailing list arc
Zoltán Zvara created FLINK-1902:
---
Summary: Web interface reports false (the default)
jobmanager.rpc.port on YARN
Key: FLINK-1902
URL: https://issues.apache.org/jira/browse/FLINK-1902
Project: Flink
So, the first thing is a "feature" of the Java API that removes
duplicate fields in keys, so an equi-join on (0,0) with (0,1) would
throw an error because one 0 is removed from the first key.
The second thing is a feature of the Table API where the error message
is hinting at the problem:
Could no
Aljoscha Krettek created FLINK-1903:
---
Summary: Joins where one side uses a field more than once don't
work
Key: FLINK-1903
URL: https://issues.apache.org/jira/browse/FLINK-1903
Project: Flink
Why not doing two separate joins, union the results and doing a distinct
operation on the combined key?
On Fri, Apr 17, 2015 at 9:42 AM, Aljoscha Krettek
wrote:
> So, the first thing is a "feature" of the Java API that removes
> duplicate fields in keys, so an equi-join on (0,0) with (0,1) would
Aljoscha Krettek created FLINK-1904:
---
Summary: Add cross() to Table API
Key: FLINK-1904
URL: https://issues.apache.org/jira/browse/FLINK-1904
Project: Flink
Issue Type: Improvement
I am also against the manual cross method. Isn't it the idea of the table
API to hide the actual implementation from the user?
Best regards,
Felix
Am 17.04.2015 10:09 vorm. schrieb "Till Rohrmann" :
> Why not doing two separate joins, union the results and doing a distinct
> operation on the comb
Hi Gabor,
the vertex-centric iteration implements the Pregel model, according to
which a vertex is "active" during a superstep only when there are messages
for this vertex. So, the behavior you are seeing is intended, yes :-)
The logic behind this is that a vertex value gets updated by performing
Yes, that is the idea, but I think in this case the user must be
protected from an operation that can get ridiculously expensive.
On Fri, Apr 17, 2015 at 10:20 AM, Felix Neutatz wrote:
> I am also against the manual cross method. Isn't it the idea of the table
> API to hide the actual implementat
Robert Metzger created FLINK-1905:
-
Summary: Add tests for parallelism > 1 for Kafka sources
Key: FLINK-1905
URL: https://issues.apache.org/jira/browse/FLINK-1905
Project: Flink
Issue Type: S
Hey,
Thank you both for your help :)
I was missing this point in the Pregel model. I thought it would be
expected to receive empty message iterators.
Yes, I know it is not the best use-case for vertex-centric iteration as
there are well-defined steps, and also know the Flink examples. I only
want
I agree with Aljoscha.
Let's give a good error message and offer a cross operator.
2015-04-17 4:52 GMT-05:00 Aljoscha Krettek :
> Yes, that is the idea, but I think in this case the user must be
> protected from an operation that can get ridiculously expensive.
>
> On Fri, Apr 17, 2015 at 10:20 A
Fabian Hueske created FLINK-1906:
Summary: Add tip to work around plain Tuple return type of project
operator
Key: FLINK-1906
URL: https://issues.apache.org/jira/browse/FLINK-1906
Project: Flink
I also agree with Aljoscha.
It is widely considered a big mistake in SQL that cross products are
implicit rather than explicit (because they are so expensive)
Let's not make the same mistake here to put the theoretical algebra over
the practical user experience and program safety.
On Fri, Apr 17
Okay, it seems the consensus forms around not breaking the API.
When it comes to the *OperatorBase - should we rename them or simply get
rid of them (remove the common API). If we want to remove them, a
precondition is to remove the Record API, and for that, we should migrate
the Record-API-based
Nikolaas Steenbergen created FLINK-1907:
---
Summary: Scala Interactive Shell
Key: FLINK-1907
URL: https://issues.apache.org/jira/browse/FLINK-1907
Project: Flink
Issue Type: New Feature
16 matches
Mail list logo