Converting a many-to-many relationship to one-to-many in PostgreSQL

问题

I have a many-to-many between foo and bar modeled as a table foo_bar with foo_id and bar_id.

I'd now like to model this as a one-to-many (which my data allows).

I've added a foo_id column to bar but now I want to migrate my data. So, I want to

UPDATE bar SET foo_id = f where id = b;

where each f and b pair are coming from

SELECT foo_id AS f, bar_id AS b FROM foo_bar;

Is it possible to do this in SQL (and specifically PostgreSQL 9.0)?

I know how to do sub-SELECTs in UPDATEs when there's only one value, but stumped how to do it in this case.

回答1:

UPDATE bar b
SET    foo_id = fb.foo_id
FROM   foo_bar fb
WHERE  fb.bar_id = b.bar_id;

If you should have multiple rows for one bar (which you shouldn't, according to your description) the one row will be updated multiple times and the result is arbitrary.

This form of the query generally performs better than a correlated subquery.

Note that the primary key of bar should really be named bar_id - I use that name in the query.

回答2:

You can still join tables in UPDATE statements, try

UPDATE  bar a
SET     foo_id = c.foo_id
FROM    (
            SELECT foo_id, bar_id
            FROM foo_bar
        ) c
WHERE   a.id = c.bar_id

or simply as

UPDATE  bar a
SET     foo_id = c.foo_id
FROM    foo_bar c
WHERE   a.id = c.bar_id

回答3:

If you really have a 1-many relationship, then it doesn't matter which value of foo you take for a given bar -- there is only one or they are all the same.

You can do the following:

update bar
    set foo_id = (select max(foo_id) from foo_bar where foo_bar.bar_id = bar.id)

The subquery limits the results to a single value.

来源：https://stackoverflow.com/questions/12151761/converting-a-many-to-many-relationship-to-one-to-many-in-postgresql