ORA-00918: column ambiguously defined in SELECT *

Question

Getting ORA-00918: column ambiguously defined: running this SQL:

SELECT *
FROM
  (SELECT DISTINCT(coaches.id),
    people.*,
    users.*,
    coaches.*
  FROM "COACHES"
  INNER JOIN people ON people.id = coaches.person_id
  INNER JOIN users ON coaches.person_id = users.person_id
  LEFT OUTER JOIN organizations_users ON organizations_users.user_id = users.id
) WHERE rownum <= 25

Any suggestions please?

Solution 1

A query's projection can only have one instance of a given name. As your WHERE clause shows, you have several tables with a column called ID. Because you are selecting * your projection will have several columns called ID. Or it would have were it not for the compiler hurling ORA-00918.

The solution is quite simple: you will have to expand the projection to explicitly select named columns. Then you can either leave out the duplicate columns, retaining just (say) COACHES.ID or use column aliases: coaches.id as COACHES_ID.

Perhaps that strikes you as a lot of typing, but it is the only way. If it is any comfort, SELECT * is regarded as bad practice in production code: explicitly named columns are much safer.

Solution 2

You have multiple columns named the same thing in your inner query, so the error is raised in the outer query. If you get rid of the outer query, it should run, although still be confusing:

SELECT DISTINCT
    coaches.id,
    people.*,
    users.*,
    coaches.*
FROM "COACHES"
    INNER JOIN people ON people.id = coaches.person_id
    INNER JOIN users ON coaches.person_id = users.person_id
    LEFT OUTER JOIN organizations_users ON organizations_users.user_id = users.id
WHERE
    rownum <= 25

It would be much better (for readability and performance both) to specify exactly what fields you need from each of the tables instead of selecting them all anyways. Then if you really need two fields called the same thing from different tables, use column aliases to differentiate between them.

Solution 3

You can also see this error when selecting for a union where corresponding columns can be null.

select * from (select D.dept_no, D.nullable_comment
                  from dept D
       union
               select R.dept_no, NULL
                 from redundant_dept R
)

This apparently confuses the parser, a solution is to assign a column alias to the always null column.

select * from (select D.dept_no, D.comment
                  from dept D
       union
               select R.dept_no, NULL "nullable_comment"
                 from redundant_dept R
)

The alias does not have to be the same as the corresponding column, but the column heading in the result is driven by the first query from among the union members, so it's probably a good practice.

Comments

[APC]

+1 This solves the problem because the outer SELECT is unnecessary as the code stands. If the outer query was required - say because the inner query had had an ORDER BY - then it would be a different matter.

[user43685]

The code is generated by an ORM library so we have no control over the outer SELECT generation. Just removing outer select helps but unfortunately it is not an option.

[dburges]

Anytime you use select * when you have a join, you have duplicated columns of data and that wastes resources.

[Admin]

As it’s currently written, your answer is unclear. Please edit to add additional details that will help others understand how this addresses the question asked. You can find more information on how to write good answers in the help center.

[D.Zotov]

How your answer is related to the question?