agora inbox for pgsql-sql@postgresql.org  
help / color / mirror / Atom feed
From: MS (direkt) <martin.stoecker@stb-datenservice.de>
To: pgsql-sql@postgresql.org
Subject: Re: Problems with PARTITION BY with count() in window-func
Date: Wed, 8 Nov 2017 15:51:05 +0100
Message-ID: <41eca9c2-e777-9084-8665-d9d6fda582a1@stb-datenservice.de> (raw)
In-Reply-To: <VisenaEmail.47.ec56caa56a402a62.15f9bdc4735@tc7-visena>
References: <VisenaEmail.47.ec56caa56a402a62.15f9bdc4735@tc7-visena>
List-Unsubscribe: <mailto:majordomo@postgresql.org?body=unsub%20pgsql-sql>

a solution without distinct:

select * from (
select
       date_trunc('month', log.start_date::TIMESTAMP WITHOUT TIME ZONE) 
AS month,
     log.activity_id,
     count(log.entity_id) over (partition by date_trunc('month', 
log.start_date::TIMESTAMP WITHOUT TIME ZONE), activity_id),
     count(log.entity_id) over (partition by date_trunc('month', 
log.start_date::TIMESTAMP WITHOUT TIME ZONE))
FROM
     log_entry log
order by 1,2) as x
group by 1,2,3,4
order by 1,2;

If one think of introducing new column i prefer using grouping sets and 
rollup.
If you do so, an enclosing query will help to filter just the subtotals 
you need.

Regards, Martin



Am 08.11.2017 um 14:50 schrieb Andreas Joseph Krogh:
> På onsdag 08. november 2017 kl. 11:17:39, skrev Andreas Joseph Krogh 
> <andreas@visena.com <mailto:andreas@visena.com>>:
>
>     Hi all.
>     I'm trying to count() all log-entries per activity per month in a
>     separate column using count() over(partition by ...) but get an
>     error I don't understand.
>     [snip]
>
>     select date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)AS month ,log.activity_id
>          ,count(log.entity_id)AS num_logs
>          ,count(log.entity_id)OVER total_for_month_windowAS total_for_month
>     FROM log_entrylog WHERE 1 =1 AND date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)BETWEEN '2017-01-01'::DATEAND '2017-05-01'::DATE
>     GROUP BY month,log.activity_id
>     WINDOW total_for_month_windowAS (PARTITION BY date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE))
>
>     ORDER BY date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)ASC ;
>
>     Note the missing log.entity_id in GROUP BY.
>     but this gives the error:
>     [42803] column "log.entity_id" must appear in the GROUP BY clause
>     or be used in an aggregate function
>     Details
>
> This query gives the desired results:
> SELECT DISTINCT date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)AS month ,log.activity_id
>      ,count(log.entity_id)OVER(partition by date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE),log.activity_id)AS num_logs_per_activity
>      ,count(log.entity_id)OVER (partition by date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE))AS total_for_month
> FROM log_entrylog ORDER BY date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)ASC,log.activity_idASC ;
> month 	activity_id 	num_logs_per_activity 	total_for_month
> 2017-01-01 00:00:00.000000 	1 	2 	4
> 2017-01-01 00:00:00.000000 	2 	2 	4
> 2017-02-01 00:00:00.000000 	1 	5 	6
> 2017-02-01 00:00:00.000000 	2 	1 	6
>
> But I'd like a solution without the DISTINCT, if one exists?
> If I introduce a new column, logged_for, and want to list number of 
> logged entries per person per activity I can use this:
> SELECT DISTINCT date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)AS month ,log.logged_for
>      ,log.activity_id
>      ,count(log.entity_id)OVER(partition by date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE),log.logged_for,log.activity_id)AS num_logs_per_person_for_activity
>      ,count(log.entity_id)OVER (partition by date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE),log.logged_for)AS total_for_person_for_month
> FROM log_entrylog ORDER BY date_trunc('month',log.start_date::TIMESTAMP WITHOUT TIME ZONE)ASC ;
> Which gives:
> month 	logged_for 	activity_id 	num_logs_per_person_for_activity 
> total_for_person_for_month
> 2017-01-01 00:00:00.000000 	5 	1 	2 	4
> 2017-01-01 00:00:00.000000 	5 	2 	2 	4
> 2017-01-01 00:00:00.000000 	6 	1 	2 	4
> 2017-01-01 00:00:00.000000 	6 	2 	2 	4
> 2017-02-01 00:00:00.000000 	5 	1 	5 	6
> 2017-02-01 00:00:00.000000 	5 	2 	1 	6
> 2017-02-01 00:00:00.000000 	6 	1 	5 	6
> 2017-02-01 00:00:00.000000 	6 	2 	1 	6
>
> Is this the recommended way to do this, I mean - having "group by" in 
> the "partition by" belonging to the OVER()-clause of the count-aggregates?
> Thanks.
> -- 
> *Andreas Joseph Krogh*
> CTO / Partner - Visena AS
> Mobile: +47 909 56 963
> andreas@visena.com <mailto:andreas@visena.com>
> www.visena.com <https://www.visena.com;
> <https://www.visena.com;

-- 
Widdersdorfer Str. 415, 50933 Köln; Tel. +49 / 221 / 9544 010
HRB Köln HRB 75439, Geschäftsführer: S. Böhland, S. Rosenbauer

Attachments:

  [image/png] ihkhhplfhmpfbhho.png (1.9K, ../41eca9c2-e777-9084-8665-d9d6fda582a1@stb-datenservice.de/3-ihkhhplfhmpfbhho.png)
  download | view image

view thread (3+ messages)

Message-ID: <41eca9c2-e777-9084-8665-d9d6fda582a1@stb-datenservice.de>
Permalink:  ../41eca9c2-e777-9084-8665-d9d6fda582a1@stb-datenservice.de/
Also on:    postgresql.org/message-id/41eca9c2-e777-9084-8665-d9d6fda582a1@stb-datenservice.de

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-sql@postgresql.org
  Cc: martin.stoecker@stb-datenservice.de
  Subject: Re: Problems with PARTITION BY with count() in window-func
  In-Reply-To: <41eca9c2-e777-9084-8665-d9d6fda582a1@stb-datenservice.de>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox