SQL查询选择值的唯一组合，两列顺序被忽略

Question

I am trying to build a SQL query where unique rows are selected based on specific criteria, but I am at a loss on how to construct it.

The rows have a varchar Conversation ID, varchar sender, and varchar recipient.

I need the results to be all distinct combinations of the Conversation ID and sender and recipient, but repeats of senders in the recipient column or vice versa are treated as treated as being the same.

For example:

ID      ,Sender      ,Recip
-------------------------
Convo1  ,PersonA     ,PersonB
Convo1  ,PersonB     ,PersonA
Convo1  ,PersonC     ,PersonA
Convo1  ,PersonC     ,PersonA
Convo1  ,PersonA     ,PersonC
Convo1  ,PersonC     ,PersonA
Convo2  ,PersonB     ,PersonD
Convo2  ,PersonB     ,PersonA

A query that would return:

Convo1, PersonA, PersonB
Convo1, PersonC, PersonA
Convo2, PersonB, PersonD
Convo2, PersonB, PersonA

Answer 1

In standard SQL, you would do this using the case statement:

select id, (case when sender < recip then sender else recip end) as person1,
       (case when sender < recip then recip else sender end) as person2
from conversations
group by id,
         (case when sender < recip then sender else recip end),
         (case when sender < recip then recip else sender end);

Many databases support the least() and greatest() functions which simplify the code:

select id, least(sender, recip) as person1, greatest(sender, recip) as person2
from conversations
group by id, least(sender, recip) as person1, greatest(sender, recip);

Answer 2

One option is to order the conversation in some way. Since your sender and recipient are varchar s, alphabetically would seem the be the most straightforward:

SELECT DISTINCT
    ID, 
    CASE WHEN Sender <= Recip
        THEN Sender
        ELSE Recip
        AS Person1,
    CASE WHEN Sender <= Recip
        THEN Recip
        ELSE Sender
        AS Person2
FROM Conversations