问题

I need your help to modify a query to accomplish the requirement.

According to the GATHER_TIME, I want to show the VALUE of the NAMEs. The query runs every 1 minute, and it should get 2 rows all the time. In case, there is no data at that time, it should get the past( 1 minute earlier ) data for the Name.

Here, specific requirement is that I should not consider seconds value in Timestamp value(GATHER_TIME).

I tried to make a query like below. It only gets the latest data.

Can you help me out?

Query

select NAME,sum(VALUE)
FROM TestTable
WHERE substring(to_char(gather_time,'YYYY-MM-DD HH24:MI:SS'),1,16) IN ( SELECT MAX(substring(to_char(gather_time,'YYYY-MM-DD HH24:MI:SS'),1,16)) FROM TestTable )
GROUP BY NAME

Table Data

NAME    COL1   COL2    GATHER_TIME         VALUE
------------------------------------------------
first   prince PQ1     2015-12-29 13:10:33 11
first   prince PQ2     2015-12-29 13:10:33 14
first   prince PQ3     2015-12-29 13:10:33 18
first   prince PQ4     2015-12-29 13:10:33 19
second  prince TT1     2015-12-29 13:10:59 20
second  prince TT2     2015-12-29 13:10:59 29
second  prince TT3     2015-12-29 13:10:59 43
first   prince PQ1     2015-12-29 13:11:37 71
first   prince PQ2     2015-12-29 13:11:37 74
first   prince PQ3     2015-12-29 13:11:37 78
first   prince PQ4     2015-12-29 13:11:37 79

Required result

@ Query time: 2015-12-29 13:10:59
first       62
second      92

@ Query time: 2015-12-29 13:11:59
first       302
second      92

Added:

Sorry. I should have put more detail test data. Updated it.

If two columns are used in GROUP BY:: ( Need to be verified )

SELECT DISTINCT ON (name,col1,gather_time) name,col1,SUM(value) 
FROM test_table
GROUP BY name,col1,gather_time
ORDER BY gather_time DESC
LIMIT 2;

回答1:

-- DDL
CREATE TABLE test_table(
  name TEXT,
  col2 TEXT,
  gather_time TIMESTAMP,
  value INTEGER
);

--sample data
INSERT INTO test_table  VALUES
  ('first','PQ1','2015-12-29 13:10:33'::TIMESTAMP, 11),
  ('first','PQ2','2015-12-29 13:10:33'::TIMESTAMP, 14),
  ('first','PQ3','2015-12-29 13:10:33'::TIMESTAMP, 18),
  ('first','PQ4','2015-12-29 13:10:33'::TIMESTAMP, 19),
  ('second','TT1','2015-12-29 13:10:59'::TIMESTAMP, 20),
  ('second','TT2','2015-12-29 13:10:59'::TIMESTAMP, 29),
  ('second','TT3','2015-12-29 13:10:59'::TIMESTAMP, 43),
  ('first','PQ1','2015-12-29 13:11:37'::TIMESTAMP, 71),
  ('first','PQ2','2015-12-29 13:11:37'::TIMESTAMP, 74),
  ('first','PQ3','2015-12-29 13:11:37'::TIMESTAMP, 78),
  ('first','PQ4','2015-12-29 13:11:37'::TIMESTAMP, 79);

--query to run every minute
SELECT name,SUM(value) FROM test_table
  WHERE gather_time <= now()
  GROUP BY name,gather_time
  ORDER BY gather_time DESC
  LIMIT 2;

OR using function:

CREATE OR REPLACE FUNCTION get_2latest_agg_rows(qtime timestamp) 
  RETURNS TABLE (
    order_name TEXT,
    sum_value BIGINT
 ) AS $$
BEGIN
  RETURN QUERY
  SELECT name as order_name,SUM(value) as agg_result FROM test_table
  WHERE gather_time <= qtime::TIMESTAMP
  GROUP BY name,gather_time
  ORDER BY gather_time DESC
  LIMIT 2;
END;
$$ LANGUAGE plpgsql;

--sample run
SELECT * FROM get_2latest_agg_rows('2015-12-29 13:11:59'::TIMESTAMP);

回答2:

What if you will sort by gather_time and use limit 2 to get the latest 2 rows? Will that fit the requirments?

select NAME,sum(VALUE) FROM TestTable ORDER BY gather_time DESC LIMIT 2;

来源：https://stackoverflow.com/questions/34575271/how-to-get-the-latest-2-rows-postgresql

标签

sql

postgresql

greatest-n-per-group

How to get the latest 2 rows ( PostgreSQL )

问题