Re: [DISCUSS] Extending Frequent Items Sketch to allow returning of all retained items.

[leerho <le...@gmail.com>]

I agree with Jon.

1) If this change is for research purposes, this could be done locally and
not affect all users of the library.

2) It is not clear why we would want to expose items to users that the
sketch considers noise.

Unless there are strong arguments to the contrary, I propose that we reject
this PR.

Lee.

On Tue, Feb 16, 2021 at 2:05 PM Jon Malkin <jo...@gmail.com> wrote:

> I am very hesitant to have the sketch return everything. Being realistic
> about it, I don't think people read and understand the documentation well
> enough for it to be a good idea. "Of course I want everything" And then
> they blame us for returning non-heavy hitters and decide the library is
> broken. RTFM is useful advice and all, but you need to deal with the users
> you have, not idealized users.
>
> I may be misinterpreting the discussion, but the comment on the PR
> suggests it's proposed basically for research purposes. That seems like the
> sort of change one makes locally rather than making a permanent part of the
> library?
>
>   jon
>
>
> On Tue, Feb 16, 2021 at 11:42 AM leerho <le...@gmail.com> wrote:
>
>> Folks,
>>
>> Please read https://github.com/apache/datasketches-java/pull/350 which
>> addresses the issue.
>>
>> Extending the Java Frequent Items ItemsSketch<T> to be able to return all
>> retained items has implications on whether the same change should be made
>> to the non-generic LongsSketch as well as whether the C++ versions should
>> also provide the same capability.
>>
>> We would appreciate comments from the community on this issue.
>>
>> Cheers,
>> Lee.
>>
>>
>>
>>