如何比较具有多个联合和联接的表？

Question

I'm trying to find rows which are present only in one table. Big table name (consists all rows) is kwf_uploads , small tables present in many names. My goal is find rows which exists in kwf_uploads , but not exists in other (this example are primary and foreigh keys in tables without relations).

What I did in sql:

select id from
kwf_uploads
left join

(SELECT picture_id as id
FROM documents where picture_id is not null
UNION 
SELECT file_id as id
FROM books where file_id is not null
UNION 
SELECT picture_id as id
FROM employee where picture_id is not null
UNION 
SELECT file_id as id
FROM flightFiles where file_id is not null
UNION 
SELECT picture_id as id
FROM tasks where picture_id is not null
UNION 
SELECT picture_id as id
FROM trainingContentQuestions where picture_id is not null
UNION 
SELECT picture_id as id
FROM trainingQuestions where picture_id is not null) foo  

ON kwf_uploads.id = foo.id

Find it on: SQL: cascade UNION and JOIN

But it fails with error: SQL (1052): Column "id" in field list is ambiguous .

I don't want to use join with every table, because I'm not good in joins and sql becomes very large and unreadable. I also tried not exists contrustion without any results. I think we can find a better solution =)

Result from select id from kwf_uploads query (9690 rows):

Result from union queries (6096 rows):

In result I want to see 3594 (9690 minus 6096) rows.

Answer 1

UNION (when used by itself, without ALL) is an "expensive" operation.

There is an alternative, NOT EXISTS. This construct is a "semi join" and I suspect may be less expensive than a union approach.

SELECT
      id
FROM kwf_uploads AS u
WHERE NOT EXISTS     (SELECT NULL FROM documents WHERE u.id = picture_id )
      AND NOT EXISTS (SELECT NULL FROM books WHERE u.id = file_id )
      AND NOT EXISTS (SELECT NULL FROM employee WHERE u.id = picture_id )
      AND NOT EXISTS (SELECT NULL FROM flightFiles WHERE u.id = file_id )
      AND NOT EXISTS (SELECT NULL FROM tasks WHERE u.id = picture_id )
      AND NOT EXISTS (SELECT NULL FROM trainingContentQuestions WHERE u.id = picture_id )
      AND NOT EXISTS (SELECT NULL FROM trainingQuestions WHERE u.id = picture_id )

Answer 2

To fix your current error

Prefix Id column with proper resultset alias. In this case kwf_uploads.id . Because Id exists in both resultsets.

To get desired output

Add Where foo.id is null . This will fetch those ids which failed in join, those are the ids you want.

  select kwf_uploads.id from
        kwf_uploads
        left join

        (SELECT picture_id as id
        FROM documents where picture_id is not null
        UNION 
        SELECT file_id as id
        FROM books where file_id is not null
        UNION 
        SELECT picture_id as id
        FROM employee where picture_id is not null
        UNION 
        SELECT file_id as id
        FROM flightFiles where file_id is not null
        UNION 
        SELECT picture_id as id
        FROM tasks where picture_id is not null
        UNION 
        SELECT picture_id as id
        FROM trainingContentQuestions where picture_id is not null
        UNION 
        SELECT picture_id as id
        FROM trainingQuestions where picture_id is not null) foo  

        ON kwf_uploads.id = foo.id
        where foo.id is null

Answer 3

Beside DarkKnight answer:

You dont need include where on each union select . Worst case you get one NULL from all the UNION but that wont appear on the LEFT JOIN anyway

select kwf_uploads.id 
from kwf_uploads
left join
    (
    SELECT picture_id as id FROM documents
    UNION 
    SELECT file_id as id FROM books
    UNION 
    SELECT picture_id as id FROM employee
    UNION 
    SELECT file_id as id FROM flightFiles
    UNION 
    SELECT picture_id as id FROM tasks
    UNION 
    SELECT picture_id as id FROM trainingContentQuestions 
    UNION 
    SELECT picture_id as id FROM trainingQuestions
    ) foo      
ON kwf_uploads.id = foo.id
where foo.id is null