Question

可以通过以下方式创建数据：

create table rows (
  client_tracking_id text,
  external_id text,
  transaction_date timestamp,
  transaction_type text,
  transaction_amount numeric(18,2)
);

INSERT INTO rows 
VALUES ('281101002240', '281101002240','2019-08-14 18:03:04.000000'::timestamp,'Purchase',2000), 
('RC000610808201930230', '281101002240','2019-08-19 07:56:45.000000'::timestamp,'AuthorizationRequest',0),
('RC000610808201930230', '281101002240','2019-08-19 07:56:46.000000'::timestamp,'AuthorizationRequest',-2000),
('RC000610808201930230', '281101002240','2019-08-19 07:56:46.000000'::timestamp,'Purchase',-2000);

此查询的目标是检索第一次购买的（正）日期时间，但返回client_tracking_id列的RC值。

尝试1：

SELECT max(client_tracking_id) AS client_tracking_id,
       min(transaction_date)   AS min
from rows
GROUP BY external_id;

尝试2：

SELECT max(client_tracking_id) AS client_tracking_id,
       min(transaction_date)   AS min
from rows
where (transaction_type = 'Purchase'::text) AND (transaction_amount > (0)::numeric)
GROUP BY external_id;

在这种情况下尝试尝试正确返回[RC00061080820193023,'2019-08-14 18:03:04.000000']集的情况，但是在其他情况下并非总是如此。因此，额外的要求是将日期时间仅作为尝试两次满足where条件的交易返回。

在SQL Fiddle上可用：http://www.sqlfiddle.com/#!17/642fd/6

Answer 1

首先过滤，您只需要“购买”>0。然后按日期排序并采用最早的日期。

SQL FIDDLE

    fileInfo = fileInfo.Name.Replace("__", "_")

    Dim a() As String = fileInfo.Name.Split(New Char() {"_", "-"})

输出

select * 
from rows
WHERE transaction_amount > 0 
  AND transaction_type = 'Purchase'
ORDER BY transaction_date 
LIMIT 1
;

Answer 2

我已经开发出满足我所有要求的以下内容：

select distinct
(
  select max(client_tracking_id) as id
  from rows r1
  where r1.external_id = r.external_id
) as client_tracking_id,
(
select min(transaction_date) as date
from rows r2
WHERE transaction_type = 'Purchase' AND transaction_amount > 0
and r2.external_id = r.external_id
) as date

from rows r

单分组，多选择。需要子查询吗？

2 个答案: