kylin-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From 张 佑铖 <edouard...@hotmail.com>
Subject 答复: Calculate count distinct with 2 segments not correct
Date Thu, 11 Jan 2018 12:01:48 GMT
Hi shaofeng,


Thank you for your reply.

We use the current version 2.2.0.

Actually I did not know the usage of Advanced Dictionaries before,

after I add global dictionary for the column, it seems to be correct.

So the global dictionary is for the situation like us?

If I do not use  global dictionary, every segment will associate with a dictionary for a count
distinct separately?

And I also find out after I add the global dictionary the rate of expansion does not increase



________________________________
发件人: ShaoFeng Shi <shaofengshi@apache.org>
发送时间: 2018年1月11日 16:58
收件人: dev
主题: Re: Calculate count distinct with 2 segments not correct

Hi yousheng,

Whatever the time range you queried, the result of count distinct
should/need be correct. And the segment merge shouldn't change the query
result.

What's your Kylin version? Did you use the GlobalDictionary for the column
that doing distinct count?

2018-01-11 15:57 GMT+08:00 张 佑铖 <edouardZYC@hotmail.com>:

> Hi Kylin team,
>
>
>
> In my cube, partition by date, auto merge by 7 days.
>
> And I use bitmap for count distinct.
>
> The case is I have 2 segments which is 20171229-20180105 and
> 20180105-20180112, I found that if the time of query in both 2 segments,
> the result of count distinct is not correct,
>
> after I merge the 2 segments and finally the result seems correct.
>
> So I have the question,  if I want to calculate a count distinct with
> whenever start and whenever end, I should always merge all the segments?
>
>
> Thank you
>



--
Best regards,

Shaofeng Shi 史少锋
Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message