How to outer-join two tables (main and many-to-one sub-table) to get only ONE item from second table?

问题

I have two tables that are something like this:

Main table: id (int), title (varchar), etc. Sub-table: main_table_id (foreign key into main table), tag (varchar), etc.

There can be zero or more subtable rows for a given row in the main table.

I want to do a query that will return every row of the main table, with the columns of the main table, and a the columns from only a single row (doesn't matter which) of the sub-table, if there are any, otherwise NULL in those columns.

Of course if I just do a basic LEFT OUTER JOIN then of course I get the main table repeated multiple times, one for each match in the sub-table.

I'm sure I have seen this done before using a LEFT OUTER JOIN and some sort of trickery that forces only one row to be selected from the sub-table, not all of them -- maybe picking out the minimum or maximum OID. However, more than an hour of googling has not yielded any solutions.

Does anybody have this trick in their toolbelt?

回答1:

I like Charlie's answer best, but i'm not sure if Postges has a TOP/LIMIT function, so here's an alternate solution that doesn't need one (but does assume sub_table has a primary key called "id"):

SELECT * 
FROM main_table m LEFT OUTER JOIN sub_table s 
ON s.main_table_id = m.id
WHERE s.id IS NULL OR s.id IN ( 
  SELECT MAX(id) FROM sub_table GROUP BY main_table_id
)

回答2:

If you're using SQL Server, you can use the TOP clause. If it is something other than SQL Server, you'll have to see if that database offers something equivalent (many of them do). Something like this ...

Select * from Main m
left outer join 
(select top 1 * from subtable s
  where s.main_table_id = m.id) q
on q.main_table_id = m.id;

Note: That is to show you the general idea. I didn't have a chance to run it, so there might be a couple of changes necessary, but the concept is there.

回答3:

If you're just checking of there is something in the second table that goes with the items in the first table you can use an outer join with a group by clause and an aggregate:

select t1.[name], count(t1.[name]) as t2count
   from table1 t1
   left outer join table2 t2 on t1.fk = t2.pk
group by t1.[name]

Then, anything that has 0 for t2count will be those that have nothing in table2

edit: actually, I don't remember if t2count will have null or 0 in it...it should be one of those values.

回答4:

Funny method in Postgres (not schooll at all) :

SELECT * 
 FROM main_table m left join (
   select (max(s::character varying)::subtable).* from subtable s group by main_table_id 
 ) s ON s.main_table_id = m.id

Explanation :
You select the complete record as one field :

select s from subtable s

You cast this as character varying to be able to use max aggregat (or min ...)

 select max(s::character varying) from subtable s group by main_table_id

You obtain only one line by main_table_id containing a varchar representing your record, so we have to cast again in "subtable" and explode it -> (myvarchar::subtable).*

select (max(s::character varying)::subtable).* from subtable s group by main_table_id

You obtain your table subtable well formed with just one line by main_table_id

来源：https://stackoverflow.com/questions/667459/how-to-outer-join-two-tables-main-and-many-to-one-sub-table-to-get-only-one-it

标签

sql

postgresql

join

outer-join