Hi Felix,
You can change the listening port of jobmanager web frontend by setting
`jobmanager.web.port` in configuration (conf/flink-conf.yml).
I attached a link of documentation [1] about this.
Regards,
Chiwan Park
[1]
https://ci.apache.org/projects/flink/flink-docs-release-0.9/setup/config.h
Hi Chiwan,
I'm using Flink 0.9. 1
Cheers,
Giacomo
I created a JIRA issue [1]. After FLINK-2637 [2][3] is resolved, I’ll
submit a patch to solve this.
Currently, there is no way to use derived class with CSV input.
Thank you for reporting.
[1] https://issues.apache.org/jira/browse/FLINK-2690
[2]
I'm using Flink from Zeppelin in local mode and Zeppelin is using 8080
Hi Fabian,
the local file problem would however not exist, if I just copy both halves
to all nodes, right?
Lets say I have a file `1st` and a file `2nd`, which I copy to all nodes.
Now with your approach from above, I do:
// helper broadcast datasets to know on which half to operate
val data1stH
I created a JIRA issue [1]. After FLINK-2637 [2][3] is resolved, I’ll submit a
patch to solve this.
Currently, there is no way to use derived class with CSV input.
Thank you for reporting.
[1] https://issues.apache.org/jira/browse/FLINK-2690
[2] https://issues.apache.org/jira/browse/FLINK-2637
[
It seems like a bug of CsvInputFormat. I succeed in reproducing in my local
machine.
I will create a JIRA issue for this and submit a patch to fix it.
Which version of Flink are used?
Regards,
Chiwan Park
> On Sep 17, 2015, at 12:20 AM, Giacomo Licari wrote:
>
> Yes I did.
>
> if anyone has
Yes I did.
if anyone has a bypass solution, let us know.
Regards,
Giacomo Licari
On Wed, Sep 16, 2015 at 5:15 PM, Chiwan Park wrote:
> Hi Giacomo,
>
> Did you create constructors without arguments in both base class and
> derived class?
> If you do, it seems like a bug.
>
> Regards,
> Chiwan P
Hi Giacomo,
Did you create constructors without arguments in both base class and derived
class?
If you do, it seems like a bug.
Regards,
Chiwan Park
> On Sep 17, 2015, at 12:04 AM, Giacomo Licari wrote:
>
> Hi Chiwan,
> I followed instructions in documentation.
> I have a simple base class wi
Hi Chiwan,
I followed instructions in documentation.
I have a simple base class with some properties (all public).
Then I extend that class with a new public property (tweet in my case), I
provide also getter and setter for that property.
Now when I execute:
DataSet ds_twitter = env.readCsvFile("f
Hi Giacomo,
You should set your field as public. If you are set your field as private or
protected, the class must provide getter and setter to be treated as POJO.
Maybe the documentation in homepage [1] would be helpful.
Regards,
Chiwan Park
[1]
https://ci.apache.org/projects/flink/flink-doc
I run it only implementing java.io.Serializable without disabling the
closure cleaner.
Another question I have is about POJO classes.
I would also create a base POJO class with some common proprerties, and
then extend it in new classes. These classes are used to convert a CSV into
a dataset of POJ
Could you also try the other variant (disabeling the closure cleaner)? I
would be curious if this behavior is expected Java Serialization behavior,
or whether our pre-processing code is causing it.
Greetings,
Stephan
On Wed, Sep 16, 2015 at 3:38 PM, Giacomo Licari
wrote:
> Thank you Martin and
Thank you Martin and Stephan for your help.
I tried directly to implement java.io.Serializable in Base class and it
worked perfectly!
Now I can develop more flexible and maintainable code. Thank you a lot guys.
Greetings,
Giacomo
On Wed, Sep 16, 2015 at 1:46 PM, Stephan Ewen wrote:
> Hi!
>
> I
Hi,
unfortunately the expression syntax is not documented right now. I'm
writing documentation right now, it should be available in a few hours.
For reference, you can look at expressionDsl.scala or
ExpressionParser.scala to see the expressions that are actually implemented.
Cheers,
Aljoscha
On
Cool!
Always happy to help :-)
2015-09-16 14:41 GMT+02:00 Pieter Hameete :
> Fantastic Fabian, that was it :-)! I'm glad it wasn't a more severe/tricky
> programming error though I already spent quite some time wondering about
> this one.
>
> Have a nice day!
>
> - Pieter
>
>
>
> 2015-09-16 14:2
Fantastic Fabian, that was it :-)! I'm glad it wasn't a more severe/tricky
programming error though I already spent quite some time wondering about
this one.
Have a nice day!
- Pieter
2015-09-16 14:27 GMT+02:00 Fabian Hueske :
> Sorry, I was thinking too complicated. Forget about the methods
Sorry, I was thinking too complicated. Forget about the methods I mentioned.
If you are implementing WritableComparable types, you need to override the
hashcode() method.
Flink treats WritableComparable types just like Hadoop [1].
DawnData does not implement hashcode() which causes inconsistent ha
Hi,
I havent been able to find the problem yet, and I dont know exactly how to
check the methods you suggested to check earlier (extractKeys,
getFlatComparators, duplicate) for the Scala API. Do you have some pointers
for me on how I can check these myself?
In my earlier mail I stated that maps,
Hi!
Interesting case. We use plain Java Serialization to distribute UDFs, and
perform additional "cleaning" of scopes, which may be causing the issue.
Can you try the following to see if any of those resolves the problem?
1) On the environment, disable the closure cleaner (in the execution
confi
Where can I find a description of the expressions that can be used with
the Table operations, such as select() and filter()?
I've seen the examples use code like
> .select("word.count as count, word").filter("count = 2");
and
> .select("month as m, cnt.max as max")
The Table docs
(https://ci.a
Hi Giacomo,
I ran into the same issue. Seems to be coupled to the serialization
mechanism of UDFs. I solved it by letting the base class implement the
UDF interface (e.g. FlatMapFunction) and in addition make it generic
(which should be necessary in your example).
public [abstract] class Cul
Hi guys,
I'm trying to create a base class which is inherited from classes
implementing FlatMap method on specific POJO types.
It seems inheritance doesn't work, I can access this.PropertyName or
super.PropertyName from flatMap method but values are always null.
Here the derived class, using Rain
Cheers Till and Fabian for your fast replies, it's much appreciated!
I figured something should be wrong with my data type. I have no doubt the
CoGroup works just fine :-) Its pointers what to investigate about my
datatype what I am looking for. Initially I had problems with serialization
causing
This sound like a problem with your custom type and its (presumably) custom
serializers and comparators.
I assume it is not an issue of partitioning or sorting because Reduce is
working fine, as you reported.
CoGroup does also partition and sort data, but compares the elements of two
sorted stream
Hi Pieter,
your code doesn't look suspicious at the first glance. Would it be possible
for you to post a complete example with data (also possible to include it
in the code) to reproduce your problem?
Cheers,
Till
On Wed, Sep 16, 2015 at 10:31 AM, Pieter Hameete wrote:
> Dear fellow Flinkers,
Dear fellow Flinkers,
I am implementing queries from the XMark (
http://www.ins.cwi.nl/projects/xmark/) benchmark on Flink using a custom
nested data type. Reading the XML data generated by the XMark generator
into my custom nested datatype works perfectly, and the queries that I have
implemented
26 matches
Mail list logo
