3

With Postgres 9.4, I'm doing the following query quite often:

SELECT DISTINCT ON(recipient) * FROM messages
LEFT JOIN identities ON messages.recipient = identities.name
WHERE timestamp BETWEEN timeA AND timeB
ORDER BY recipient, timestamp DESC;

So I decided to create a view:

CREATE VIEW myView AS SELECT DISTINCT ON(recipient) * FROM messages
LEFT JOIN identities ON messages.recipient = identities.name
ORDER BY recipient, timestamp DESC;

I just realized if I query my view like SELECT * FROM myView WHERE timestamp BETWEEN timeA AND timeB I get a significantly worse performance.

Doing EXPLAIN ANALYZE on both queries, I found out the reason is that the database in the second case brings up all the records, does the left join and then applies the WHERE clause. In other words, the WHERE clause is not pushed down into the view's query. I also tried to remove the ORDER BY from the view, but still the database performs the LEFT JOIN on full data rather on the filtered set.

What is the reason of this behavior? Is there a way I can get a comparable performance when using view?

  • But is there a way to force the logic of the first query? – NumberFour Aug 8 '15 at 16:45
  • I've edited my answer with a link. You also have a new answer with code just for that. – ypercubeᵀᴹ Aug 8 '15 at 22:25
6

You can create a function like this;

CREATE OR REPLACE FUNCTION public.get_messages_by_timestamp (
  time_a timestamp,
  time_b timestamp
)
RETURNS TABLE (
  recipient varchar,
  "timestamp" timestamp
) AS
$$
BEGIN
  RETURN QUERY
    SELECT DISTINCT ON (m.recipient) 
        m.recipient,
        m."timestamp"
      FROM messages m
      LEFT JOIN identities i ON m.recipient = i.name
      WHERE 
        m."timestamp" BETWEEN time_a AND time_b
      ORDER BY 
        m.recipient,
        m."timestamp" DESC;
END;
$$
LANGUAGE 'plpgsql';

Then you can use the function like a table

  SELECT *
         FROM get_messages_by_timestamp('2015-01-01', '2015-01-02')
12

This query:

SELECT DISTINCT ON(recipient) * FROM messages
LEFT JOIN identities ON messages.recipient = identities.name
WHERE messages.timestamp BETWEEN timeA AND timeB
ORDER BY recipient, timestamp DESC;

says:

For all messages between timeA and timeB, find the recipients and for every recipient, find one message (the latest in between timeA and timeB).


This query (which you would get if you use the view):

SELECT *
FROM 
  ( SELECT DISTINCT ON(recipient) * FROM messages
    LEFT JOIN identities ON messages.recipient = identities.name
    ORDER BY recipient, timestamp DESC
  ) AS myView
WHERE timestamp BETWEEN timeA AND timeB;

says:

For all messages, find the recipients and for every recipient, find one message, (the latest in all times) and then, show only those messages that are between timeA and timeB.


As a result, the first query will show messages that are between time A and B which the second query will not show (because there may be one or more messages for the same recipient, later than time B).

So, the queries are logically different and the condition cannot (and shouldn't) be pushed down for your view.

If you want to have parameters passed to your view, look the two answers (by @a_horse_with_no_name and @Erwin Brandstetter) in this question for how to use a set returning function: Pass In “WHERE” parameters to PostgreSQL View?

  • Many thanks for a clarification! In the end, I used solution described by Sahap below. – NumberFour Aug 8 '15 at 23:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.