Postgresql：LAG直到找到具有特定值的行，并返回该值

Question

I have a dataset containing emails from customers.

One of the columns is Type. In Type exists a value 'Duplicate Case' which flags that a customer has just sent us a barrage of emails about the same topic. We only reply to the original and close all other cases as duplicates. But what I want to do, is get the type of the original email.

I want to be able to create the column Original Type:

There generally aren't more than 5 duplicates of a cases per customers.I would like to add logic which only returns results where duplicate cases are 24 hours older than the original

I have this awful piece of code:

CASE
 WHEN type = 'Duplicate Case'
 AND LAG(type,4) OVER (PARTITION BY c.client_code ORDER BY case_number ASC) = 'Duplicate Case'
  THEN LAG(type,5) OVER (PARTITION BY c.client_code ORDER BY case_number ASC)

 WHEN type = 'Duplicate Case'
 AND LAG(type,3) OVER (PARTITION BY c.client_code ORDER BY case_number ASC) = 'Duplicate Case'
  THEN LAG(type,4) OVER (PARTITION BY c.client_code ORDER BY case_number ASC)

 WHEN type = 'Duplicate Case'
 AND LAG(type,2) OVER (PARTITION BY c.client_code ORDER BY case_number ASC) = 'Duplicate Case'
  THEN LAG(type,3) OVER (PARTITION BY c.client_code ORDER BY case_number ASC)

WHEN LAG(type) OVER (PARTITION BY c.client_code ORDER BY case_number ASC) = 'Duplicate Case'
AND type = 'Duplicate Case'
  THEN LAG(type,2) OVER (PARTITION BY c.client_code ORDER BY case_number ASC)

WHEN type = 'Duplicate Case'
  THEN LAG(type) OVER (PARTITION BY c.client_code ORDER BY case_number ASC)
    END AS original_type

And this gives me sort of what I want:

But how can I add a time logic? I want to put in Type only from the first non-duplicate if this duplicate was created less than 24 hours before the original type?

Answer 1

So what you need to do here is create a temporary table with a blank field, then update this blank field later on with some criteria:

/* Main table */

DROP TABLE IF EXISTS cases;
CREATE TEMPORARY TABLE cases AS
SELECT
created_date,
type,
CASE WHEN type = 'Duplicate Case'
  THEN CAST('Unknown' AS VARCHAR(40))
  ELSE type END AS original_type,
 CAST('' AS VARCHAR(40)) AS original_case_number,
client_code,
case_number
FROM case
 ORDER BY c.client_code DESC, case_number DESC;

/* Append previous case data */

UPDATE cases
SET original_case_number = prev_case_number
FROM
(
    SELECT
      a.case_number,
      MAX(b.case_number) AS prev_case_number
      FROM cases a
     LEFT JOIN cases b ON a.client_code = b.client_code
     AND b.created_date BETWEEN a.created_date - INTERVAL '48 hour' AND 
     a.created_date + INTERVAL '1 second'
     AND b.type <> 'Duplicate Case'
     AND a.type = 'Duplicate Case'
     GROUP BY 1
 ) prev
  WHERE cases.case_number = prev.case_number;

UPDATE cases
SET original_type = b.type
FROM (SELECT *
      FROM cases) b
  WHERE cases.original_case_number = b.case_number;

SELECT * FROM cases