在 Oracle/Postgresql 查询中需要帮助

Question

这是一个要求，希望你能帮助我。

要求：

我需要获取发送者和接收者之间的最后一次对话，反之亦然。

详细来说，我在表格中有以下数据。

    sen | rec |            day             |  msg  
    -----+-----+----------------------------+-------
    A   | B   | 2014-12-05 11:51:38.213723 | HI
    B   | A   | 2014-12-05 11:51:54.486329 | HELLO
    A   | B   | 2014-12-05 11:52:10.289694 | HwRU
    B   | A   | 2014-12-05 11:52:21.613091 | FINE
    A   | C   | 2014-12-05 11:52:32.608677 | HI
    C   | A   | 2014-12-05 11:53:00.695742 | HELLO
    C   | D   | 2014-12-05 11:53:10.058141 | HELLO
    E   | D   | 2014-12-05 11:53:17.473893 | HELLO
    D   | E   | 2014-12-05 11:53:53.675839 | HI

注意：对话不分发送者和接收者。

所以，A 到 B 之间的最后一次对话是

    B   | A   | 2014-12-05 11:52:21.613091 | FINE

A 和 C 之间的最后一次对话是

    C   | A   | 2014-12-05 11:53:00.695742 | HELLO

这样，无论发送者和接收者，我都需要两个人之间的最后一次对话。

请帮帮我。

Answer 1

要在不考虑排序的情况下评估(sen,rec) 组合，您可以使用least(sen,rec), greatest(sen,rec) 这可用于对数据进行分组，然后使用窗口函数使用最新的行：

select sen, rec, day, msg
from (
  select sen, rec, day, msg,
         row_number() over (partition by least(sen,rec), greatest(sen,rec) order by day desc) as rn
  from conversations
) t
where rn = 1;

Answer 2

CREATE TABLE sendrecv(
    sen CHAR(1) NOT NULL
    , rec CHAR(1) NOT NULL
    , zday TIMESTAMP NOT NULL
    , msg varchar
      , PRIMARY KEY(sen,rec,zday)  -- this will cause an index to be created
      , UNIQUE(rec, sen,zday)      -- this will create the "swapped" index
        );

    INSERT INTO sendrecv(sen , rec , zday, msg) values
    ('A', 'B', '2014-12-05 11:51:38.213723', 'HI' )
    ,('B', 'A', '2014-12-05 11:51:54.486329', 'HELLO' )
    ,('A', 'B', '2014-12-05 11:52:10.289694', 'HwRU' )
    ,('B', 'A', '2014-12-05 11:52:21.613091', 'FINE' )
    ,('A', 'C', '2014-12-05 11:52:32.608677', 'HI' )
    ,('C', 'A', '2014-12-05 11:53:00.695742', 'HELLO' )
    ,('C', 'D', '2014-12-05 11:53:10.058141', 'HELLO' )
    ,('E', 'D', '2014-12-05 11:53:17.473893', 'HELLO' )
    ,('D', 'E', '2014-12-05 11:53:53.675839', 'HI' )
        ;

---

SELECT * FROM sendrecv sr
WHERE NOT EXISTS (
        SELECT * FROM sendrecv nx
        WHERE nx.sen = sr.sen AND nx.rec = sr.rec
        AND nx.zday > sr.zday
        )
AND NOT EXISTS (
        SELECT * FROM sendrecv nx
        WHERE nx.sen = sr.rec AND nx.rec = sr.sen
        AND nx.zday > sr.zday
        )
        ;

---

CREATE TABLE
INSERT 0 9
 sen | rec |            zday            |  msg  
-----+-----+----------------------------+-------
 B   | A   | 2014-12-05 11:52:21.613091 | FINE
 C   | A   | 2014-12-05 11:53:00.695742 | HELLO
 C   | D   | 2014-12-05 11:53:10.058141 | HELLO
 D   | E   | 2014-12-05 11:53:53.675839 | HI
(4 rows)

Answer 3

如果你想找到特定两个人之间的最后一次对话（例如'A'和'B'），解决方案很简单：

SELECT *
FROM conversations
WHERE 'A' IN (sen, rec) AND 'B' IN (sen, rec)
ORDER BY "day" DESC
LIMIT 1;

为了找到任何两方之间的最后一次对话，之前发布的两个解决方案都可以使用，但我再给你一个建议：

SELECT DISTINCT ON (LEAST(sen, rec), GREATEST(sen, rec)) *
FROM conversations
ORDER BY LEAST(sen, rec), GREATEST(sen, rec), "day" DESC;

以上语句是为 PostgreSQL 编写的。