Content-Type: multipart/alternative;
 boundary="------------KH07GEyuWZsuVuq0tBuzU48j"
Message-ID: <0251f755-00bb-41c7-8f76-14a3604f6d18@tantorlabs.com>
Date: Tue, 3 Sep 2024 16:09:50 +0300
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Content-Language: en-US
To: pgsql-admin@lists.postgresql.org
From: Ilia Evdokimov <ilya.evdokimov@tantorlabs.com>
Subject: Duplicate Extended Statistics
Archived-At: 
 <https://www.postgresql.org/message-id/0251f755-00bb-41c7-8f76-14a3604f6d18%40tantorlabs.com>
Precedence: bulk

This is a multi-part message in MIME format.
--------------KH07GEyuWZsuVuq0tBuzU48j
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit

Hello everyone,

I have a question regarding extended statistics in PostgreSQL. Why is it 
possible to create duplicate extended statistics? To make it clearer, 
here’s an example:

CREATE TABLE t(a int, b int);
INSERT INTO t(a, b) VALUES (...);
CREATE STATISTICS ON a, b FROM t;
ANALYZE t;
....
CREATE STATISTICS ON a, b FROM t;
ANALYZE t;

After executing these queries, the following issues might arise:

 1. ANALYZE will take longer to run because, for example, MCV extended
    statistics would need to be gathered twice.
 2. Duplicate information will be stored.
 3. The planner might take longer to find the relevant statistics since
    it has to search through them in a loop.

Or do duplicate extended statistics practically never occur in practice?

Thanks in advance for your response.

-- 
Regards,
Ilia Evdokimov,
Tantor Labs LCC.

--------------KH07GEyuWZsuVuq0tBuzU48j
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit

<!DOCTYPE html>
<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Hello everyone,</p>
    <p>I have a question regarding extended statistics in PostgreSQL.
      Why is it possible to create duplicate extended statistics? To
      make it clearer, here’s an example:</p>
    <p><span class="hljs-keyword">CREATE</span> <span
        class="hljs-keyword">TABLE</span> t(a <span class="hljs-type">int</span>,
      b <span class="hljs-type">int</span>);<br>
      <span class="hljs-keyword">INSERT</span> <span
        class="hljs-keyword">INTO</span> t(a, b) <span
        class="hljs-keyword">VALUES</span> (...);<br>
      <span class="hljs-keyword">CREATE</span> STATISTICS <span
        class="hljs-keyword">ON</span> a, b <span class="hljs-keyword">FROM</span>
      t;<br>
      ANALYZE t;<br>
      ....
      <br>
      <span class="hljs-keyword">CREATE</span> STATISTICS <span
        class="hljs-keyword">ON</span> a, b <span class="hljs-keyword">FROM</span>
      t;<br>
      ANALYZE t;</p>
    <p>After executing these queries, the following issues might arise:</p>
    <ol>
      <li>ANALYZE will take longer to run because, for example, MCV
        extended statistics would need to be gathered twice.</li>
      <li>Duplicate information will be stored.</li>
      <li>The planner might take longer to find the relevant statistics
        since it has to search through them in a loop.</li>
    </ol>
    <p>Or do duplicate extended statistics practically never occur in
      practice?</p>
    <p>Thanks in advance for your response.</p>
    <p></p>
    <p></p>
    <pre class="moz-signature" cols="72">-- 
Regards,
Ilia Evdokimov,
Tantor Labs LCC.</pre>
  </body>
</html>

--------------KH07GEyuWZsuVuq0tBuzU48j--