I don't know whether this is important because we are talking about the
Java API, but for people familiar with Scala or F#, for example, the
methods minBy/maxBy make perfectly sense. These methods are part of the
standard collection API of the respective language and select the element
with the minimum/maximum selected key. Of course, Flink's minBy/maxBy
return a DataSet with the element instead of the element itself.

On Fri, Nov 28, 2014 at 6:59 PM, Ufuk Celebi <[email protected]> wrote:

> @Stephan: Yap. :)
>
> I agree with Viktor's suggestions. In order to not break existing programs
> I would keep the deprecation annotation as you have in your branch. We
> could then remove them in 0.9.0 or so. For minBy/maxBy I would rename it as
> you suggested, but keep the old methods and deprecate them as well.
>
> Maybe someone can comment on the Javadocs changes here:
> https://github.com/apache/incubator-flink/pull/244
>
> I hope that it makes the API clearer for now.
>
> On Fri, Nov 28, 2014 at 6:23 PM, Viktor Rosenfeld <
> [email protected]> wrote:
>
>> Hi,
>>
>> I would drop min/max/sum. I've marked them deprecated in my pull request
>> because I did not change their current implementation based on the old
>> aggregate.
>>
>> I think that it makes more sense to have convenience functions like
>> countByKey which Sebastian suggested that do encapsulate more complex
>> behavior than an aggregation on a single field.
>>
>> I would also rename minBy/maxBy to selectMin/selectMax or something
>> similar
>> (selectMinBy?, selectMinByField?) to make their behavior more explicit.
>>
>> Best,
>> Viktor
>>
>>
>>
>> --
>> View this message in context:
>> http://apache-flink-incubator-user-mailing-list-archive.2336050.n4.nabble.com/Wrong-and-non-consistent-behavior-of-max-tp484p499.html
>> Sent from the Apache Flink (Incubator) User Mailing List archive. mailing
>> list archive at Nabble.com.
>>
>
>

Reply via email to