Path: csiph.com!news.mixmin.net!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Erland Sommarskog <esquel@sommarskog.se>
Newsgroups: comp.databases.ms-sqlserver
Subject: Re: Calculating the percentile
Date: Thu, 14 Mar 2024 14:55:22 +0100
Organization: Erland Sommarskog
Lines: 30
Message-ID: <XnsB13597CDC7B36Yazorman@127.0.0.1>
References: <20240311144248.432ce7f1a182654f52237410@g{oogle}mail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 8bit
Injection-Info: dont-email.me; posting-host="435fa701f050f778012518ef71edda4a"; logging-data="1727835"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX18wsHJPfKp8joh6tv2d5vBk"
User-Agent: Xnews/2006.08.24 Mime-proxy/2.1.c.0 (Win32)
Cancel-Lock: sha1:4uPSu4aCDZm4ug52bmLwV9EaqJg=
Xref: csiph.com comp.databases.ms-sqlserver:2241

Anton Shepelev (anton.txt@g{oogle}mail.com) writes:
> I am trying to solve a trivial problem, and my solution
> turns out unexpectedly cumbersome and likeky suboptimal.
> Given a table storing values (val) over dimension (dim), I
> want to query the total, mean and median of the value over
> the dimension.  My solution is:
> 
>   SELECT
>      dim         ,
>      SUM(val) tot,
>      AVG(val) avg,
>      med
>   FROM
>   (  SELECT
>         dim,
>         val,
>         PERCENTILE_CONT( 0.5 )
>         WITHIN GROUP (ORDER BY val)
>         OVER (PARTITION BY dim) AS med
>      FROM #test
>   ) w_med
>   GROUP BY dim, med
> 

If you are on SQL 2022, you can use the new function
APPROXIMATIVE_PERCENTILE_CONT which is a regular aggregate function.
But as the name says, it is approximative. It's intended for large
data sets where exact calculation would take a lot of time. I don't
know how well it works on smaller data sets.