I am joining several tables. From the joined table I need to select a record with the minimum value in one column. The where clause contains some additional conditions. How can this be achieved without having to list the whole join twice in the select and in the where clause to identify the minimum?
I mean - from the result of the join, I need to select one record that fullfills some conditions and that also includes a minimum in a specific column. It is in Teradata but I am asking about the general principle.
I have something like this. It works, but is ugly as the join is included twice.
SELECT TABLE1.X, TABLE2.Y, TABLE3.Z
FROM TABLE1
INNER JOIN TABLE2
ON TABLE1.A = TABLE2.B
INNER JOIN TABLE3
ON TABLE2.C=TABLE3.D
WHERE TABLE3.M =
(SELECT MIN(TABLE3.M)
FROM TABLE1
INNER JOIN TABLE2
ON TABLE1.A = TABLE2.B
INNER JOIN TABLE3
ON TABLE2.C=TABLE3.D
WHERE TABLE1.K=123 AND TABLE2.L=456
)
Thanks, R.
In a comment you say you only need one row as your output.
In which case, use ORDER BY and LIMIT 1
SELECT TABLE1.X, TABLE2.Y, TABLE3.Z
FROM TABLE1
INNER JOIN TABLE2
ON TABLE1.A = TABLE2.B
INNER JOIN TABLE3
ON TABLE2.C=TABLE3.D
WHERE TABLE1.K=123 AND TABLE2.L=456
ORDER BY TABLE3.M
LIMIT 1
Edit: (To use min() to fulfil unstated requirements...)
SELECT
X, Y, Z
FROM
(
SELECT
TABLE1.X,
TABLE2.Y,
TABLE3.Z,
TABLE3.M,
MIN(TABLE3.M) OVER () AS MIN_M
FROM
TABLE1
INNER JOIN
TABLE2
ON TABLE1.A = TABLE2.B
INNER JOIN
TABLE3
ON TABLE2.C = TABLE3.D
WHERE
TABLE1.K = 123
AND TABLE2.L = 456
)
AS FILTERED
WHERE
MIN_M = M
Even if I was going to use window functions for this, I'd use ROW_NUMBER() OR RANK() rather than using MIN(). Without a clear reason WHY you feel this MUST use it, yet still be DRY, efficient and maintainable, this constraint appears not only pointless, but misguided.
Use min window function as follows:
Select x, y, z from
(SELECT TABLE1.X, TABLE2.Y, TABLE3.Z,
Min(TABLE3.M) over () as mn,
TABLE3.M
FROM TABLE1
INNER JOIN TABLE2
ON TABLE1.A = TABLE2.B
INNER JOIN TABLE3
ON TABLE2.C=TABLE3.D
Where TABLE1.K=123 AND TABLE2.L=456 ) t
Where m = mn
If I am following correctly, you can use qualify:
SELECT TABLE1.X, TABLE2.Y, TABLE3.Z
FROM TABLE1 INNER JOIN
TABLE2
ON TABLE1.A = TABLE2.B INNER JOIN
TABLE3
ON TABLE2.C = TABLE3.D
QUALIFY TABLE3.M = MIN(CASE WHEN TABLE1.K = 123 AND TABLE2.L = 456 THEN TABLE3.M END) OVER ();
I have two tables, t1 and t2, with identical columns(id, desc) and data. But one of the columns, desc, might have different data for the same primary key, id.
I want to select all those rows from these two tables such that t1.desc != t2.desc
select a.id, b.desc
FROM (SELECT * FROM t1 AS a
UNION ALL
SELECT * FROM t2 AS b)
WHERE a.desc != b.desc
For example, if t1 has (1,'aaa') and (2,'bbb') and t2 has(1,'aaa') and (2,'bbb1') then the new table should have (2,'bbb') and (2,'bbb1')
However, this does not seem to work. Please let me know where I am going wrong and what is the right way to do it right.
Union is not going to compare the data.You need Join here
SELECT *
FROM t1 AS a
inner join t2 AS b
on a.id =b.id
and a.desc != b.desc
UNION ALL dumps all rows of the second part of the query after the rows produced by the first part of the query. You cannot compare a's fields to b's, because they belong to different rows.
What you are probably trying to do is locating records of t1 with ids matching these of t2, but different description. This can be achieved by a JOIN:
SELECT a.id, b.desc
FROM t1 AS a
JOIN t2 AS b ON a.id = b.id
WHERE a.desc != b.desc
This way records of t1 with IDs matching records of t2 would end up on the same row of joined data, allowing you to do the comparison of descriptions for inequality.
I want both the rows to be selected is the descriptions are not equal
You can use UNION ALL between two sets of rows obtained through join, with tables switching places, like this:
SELECT a.id, b.desc -- t1 is a, t2 is b
FROM t1 AS a
JOIN t2 AS b ON a.id = b.id
WHERE a.desc != b.desc
UNION ALL
SELECT a.id, b.desc -- t1 is b, t2 is a
FROM t2 AS a
JOIN t1 AS b ON a.id = b.id
WHERE a.desc != b.desc
The UNION operator is used to combine the result-set of two or more SELECT statements.
Notice that each SELECT statement within the UNION must have the same number of columns. The columns must also have similar data types.
So, if it has same number of columns and same datatype, then use Union otherwise join only Can be used.
SELECT *
FROM t1 AS a
inner join t2 AS b
on a.id =b.id
and a.desc != b.desc
My problem here is that i'm modifying an existing query and i cannot use count(*) in the query.
I have to use inner join subqueries.
What i need to "transform" into my inner join is like this (this works):
SELECT count(distinct t1.id)
FROM table1 t1
WHERE t1.column1 = 'value1' AND
t2.column2 = 'value2' AND
EXISTS(select 1 from table2 t2 where t2.id = t1.id)
My global query looks like this:
SELECT [many many column]
FROM table2 t2
INNER JOIN [...]
LEFT OUTER JOIN [...]
--[I NEED MY COUNT HERE, see below for example]
WHERE [some conditions are true]
ORDER BY [some column]
What i found to help me is something like this:
SELECT [many many column], myJoin.Count
FROM table2 t2
INNER JOIN (
SELECT tt2.id, count(distinct tt2.id) as Count
FROM table2 tt2
WHERE EXISTS (SELECT 1 FROM table1 tt1 where tt1.id = tt2.id)
GROUP BY tt2.id) myJoin
on t2.id = myJoin.id;
See what i'm trying to acheive? I need to count the ids, joining 2 tables, but i can't have a count in my main query, i can't possibly copy-paste all the "group by" condition that would go with it...
I'm on sql server.
If i find the answer i will come back and post it.
Thanks for any advice/tricks about this.
How about the following:
SELECT table2.*, TopQ.MyCount
FROM (
SELECT t2.id, myJoin.MyCount
FROM table2 t2
INNER JOIN (
SELECT tt2.id, count(distinct tt2.id) as MyCount
FROM table2 tt2
WHERE EXISTS
(SELECT 1 FROM table1 tt1 where tt1.id = tt2.id)
GROUP BY tt2.id) AS myJoin
on t2.id = myJoin.id
)AS TopQ
INNER JOIN table2 ON TopQ.id = table2.id
I came across this:
select count(distinct t1.id) over (partition by t1.aColumn) as myCount,
[many many column]
from table2 t2
inner join table1 t1 on [someConditions] = value1 and
[someConditions] = value2 and
t2.id = t1.id;
I get the same results as my first select i posted in my question, and without adding a "group by" anywhere and a lot of inner join that im not that familliar with. I'm gonna stick with this solution.
Thanks!
I have one table that I need to bump against multiple tables with left outer joins excluding the right(s). Is there a best practice for this? Union all the other tables first? Something else?
Here's the first thought that comes to my mind to handle this, but I want to know if there is a better more efficient way.
select
master_table.*
from
master_table
left outer join
(
select customer_id from table_1
union
select customer_id from table_2
union
select customer_id from table_3
union
select customer_id from table_4
) bump_table
on
master_table.customer_id = bump_table.customer_id
where
bump_table.customer_id is null
I should think a NOT EXISTS would be better. It certainly better communicates the intent of the query.
select * from master_table m
where not exists( select 1 from table_1 where m.customer_id=table_1.customer_id)
and not exists( select 1 from table_2 where m.customer_id=table_2.customer_id)
and not exists( select 1 from table_3 where m.customer_id=table_3.customer_id)
and not exists( select 1 from table_4 where m.customer_id=table_4.customer_id)
The basic form is surely faster - similar to the NOT EXISTS that #dbenham already supplied.
SELECT m.*
FROM master_table m
LEFT JOIN table_1 t1 ON t1.customer_id = m.customer_id
LEFT JOIN table_2 t2 ON t2.customer_id = m.customer_id
LEFT JOIN table_3 t3 ON t3.customer_id = m.customer_id
LEFT JOIN table_4 t4 ON t4.customer_id = m.customer_id
WHERE t1.customer_id IS NULL
AND t2.customer_id IS NULL
AND t3.customer_id IS NULL
AND t4.customer_id IS NULL;
I want to do a full outer join in MySQL. Is this possible? Is a full outer join supported by MySQL?
You don't have full joins in MySQL, but you can sure emulate them.
For a code sample transcribed from this Stack Overflow question you have:
With two tables t1, t2:
SELECT * FROM t1
LEFT JOIN t2 ON t1.id = t2.id
UNION
SELECT * FROM t1
RIGHT JOIN t2 ON t1.id = t2.id
The query above works for special cases where a full outer join operation would not produce any duplicate rows. The query above depends on the UNION set operator to remove duplicate rows introduced by the query pattern. We can avoid introducing duplicate rows by using an anti-join pattern for the second query, and then use a UNION ALL set operator to combine the two sets. In the more general case, where a full outer join would return duplicate rows, we can do this:
SELECT * FROM t1
LEFT JOIN t2 ON t1.id = t2.id
UNION ALL
SELECT * FROM t1
RIGHT JOIN t2 ON t1.id = t2.id
WHERE t1.id IS NULL
The answer that Pablo Santa Cruz gave is correct; however, in case anybody stumbled on this page and wants more clarification, here is a detailed breakdown.
Example Tables
Suppose we have the following tables:
-- t1
id name
1 Tim
2 Marta
-- t2
id name
1 Tim
3 Katarina
Inner Joins
An inner join, like this:
SELECT *
FROM `t1`
INNER JOIN `t2` ON `t1`.`id` = `t2`.`id`;
Would get us only records that appear in both tables, like this:
1 Tim 1 Tim
Inner joins don't have a direction (like left or right) because they are explicitly bidirectional - we require a match on both sides.
Outer Joins
Outer joins, on the other hand, are for finding records that may not have a match in the other table. As such, you have to specify which side of the join is allowed to have a missing record.
LEFT JOIN and RIGHT JOIN are shorthand for LEFT OUTER JOIN and RIGHT OUTER JOIN; I will use their full names below to reinforce the concept of outer joins vs inner joins.
Left Outer Join
A left outer join, like this:
SELECT *
FROM `t1`
LEFT OUTER JOIN `t2` ON `t1`.`id` = `t2`.`id`;
...would get us all the records from the left table regardless of whether or not they have a match in the right table, like this:
1 Tim 1 Tim
2 Marta NULL NULL
Right Outer Join
A right outer join, like this:
SELECT *
FROM `t1`
RIGHT OUTER JOIN `t2` ON `t1`.`id` = `t2`.`id`;
...would get us all the records from the right table regardless of whether or not they have a match in the left table, like this:
1 Tim 1 Tim
NULL NULL 3 Katarina
Full Outer Join
A full outer join would give us all records from both tables, whether or not they have a match in the other table, with NULLs on both sides where there is no match. The result would look like this:
1 Tim 1 Tim
2 Marta NULL NULL
NULL NULL 3 Katarina
However, as Pablo Santa Cruz pointed out, MySQL doesn't support this. We can emulate it by doing a UNION of a left join and a right join, like this:
SELECT *
FROM `t1`
LEFT OUTER JOIN `t2` ON `t1`.`id` = `t2`.`id`
UNION
SELECT *
FROM `t1`
RIGHT OUTER JOIN `t2` ON `t1`.`id` = `t2`.`id`;
You can think of a UNION as meaning "run both of these queries, then stack the results on top of each other"; some of the rows will come from the first query and some from the second.
It should be noted that a UNION in MySQL will eliminate exact duplicates: Tim would appear in both of the queries here, but the result of the UNION only lists him once. My database guru colleague feels that this behavior should not be relied upon. So to be more explicit about it, we could add a WHERE clause to the second query:
SELECT *
FROM `t1`
LEFT OUTER JOIN `t2` ON `t1`.`id` = `t2`.`id`
UNION
SELECT *
FROM `t1`
RIGHT OUTER JOIN `t2` ON `t1`.`id` = `t2`.`id`
WHERE `t1`.`id` IS NULL;
On the other hand, if you wanted to see duplicates for some reason, you could use UNION ALL.
Using a union query will remove duplicates, and this is different than the behavior of full outer join that never removes any duplicates:
[Table: t1] [Table: t2]
value value
----------- -------
1 1
2 2
4 2
4 5
This is the expected result of a full outer join:
value | value
------+-------
1 | 1
2 | 2
2 | 2
Null | 5
4 | Null
4 | Null
This is the result of using left and right join with union:
value | value
------+-------
Null | 5
1 | 1
2 | 2
4 | Null
SQL Fiddle
My suggested query is:
select
t1.value, t2.value
from t1
left outer join t2
on t1.value = t2.value
union all -- Using `union all` instead of `union`
select
t1.value, t2.value
from t2
left outer join t1
on t1.value = t2.value
where
t1.value IS NULL
The result of the above query that is as the same as the expected result:
value | value
------+-------
1 | 1
2 | 2
2 | 2
4 | NULL
4 | NULL
NULL | 5
SQL Fiddle
#Steve Chambers: [From comments, with many thanks!]
Note: This may be the best solution, both for efficiency and for generating the same results as a FULL OUTER JOIN. This blog post also explains it well - to quote from Method 2: "This handles duplicate rows correctly and doesn’t include anything it shouldn’t. It’s necessary to use UNION ALL instead of plain UNION, which would eliminate the duplicates I want to keep. This may be significantly more efficient on large result sets, since there’s no need to sort and remove duplicates."
I decided to add another solution that comes from full outer join visualization and math. It is not better than the above, but it is more readable:
Full outer join means (t1 ∪ t2): all in t1 or in t2
(t1 ∪ t2) = (t1 ∩ t2) + t1_only + t2_only: all in both t1 and t2 plus all in t1 that aren't in t2 and plus all in t2 that aren't in t1:
-- (t1 ∩ t2): all in both t1 and t2
select t1.value, t2.value
from t1 join t2 on t1.value = t2.value
union all -- And plus
-- all in t1 that not exists in t2
select t1.value, null
from t1
where not exists( select 1 from t2 where t2.value = t1.value)
union all -- and plus
-- all in t2 that not exists in t1
select null, t2.value
from t2
where not exists( select 1 from t1 where t2.value = t1.value)
SQL Fiddle
None of the previous answers are actually correct, because they do not follow the semantics when there are duplicated values.
For a query such as (from this duplicate):
SELECT * FROM t1 FULL OUTER JOIN t2 ON t1.Name = t2.Name;
The correct equivalent is:
SELECT t1.*, t2.*
FROM (SELECT name FROM t1 UNION -- This is intentionally UNION to remove duplicates
SELECT name FROM t2
) n LEFT JOIN
t1
ON t1.name = n.name LEFT JOIN
t2
ON t2.name = n.name;
If you need this to work with NULL values (which may also be necessary), then use the NULL-safe comparison operator, <=> rather than =.
MySQL does not have FULL-OUTER-JOIN syntax. You have to emulate it by doing both LEFT JOIN and RIGHT JOIN as follows:
SELECT * FROM t1
LEFT JOIN t2 ON t1.id = t2.id
UNION
SELECT * FROM t1
RIGHT JOIN t2 ON t1.id = t2.id
But MySQL also does not have a RIGHT JOIN syntax. According to MySQL's outer join simplification, the right join is converted to the equivalent left join by switching the t1 and t2 in the FROM and ON clause in the query. Thus, the MySQL query optimizer translates the original query into the following -
SELECT * FROM t1
LEFT JOIN t2 ON t1.id = t2.id
UNION
SELECT * FROM t2
LEFT JOIN t1 ON t2.id = t1.id
Now, there is no harm in writing the original query as is, but say if you have predicates like the WHERE clause, which is a before-join predicate or an AND predicate on the ON clause, which is a during-join predicate, then you might want to take a look at the devil; which is in details.
The MySQL query optimizer routinely checks the predicates if they are null-rejected.
Now, if you have done the RIGHT JOIN, but with WHERE predicate on the column from t1, then you might be at a risk of running into a null-rejected scenario.
For example, the query
SELECT * FROM t1
LEFT JOIN t2 ON t1.id = t2.id
WHERE t1.col1 = 'someValue'
UNION
SELECT * FROM t1
RIGHT JOIN t2 ON t1.id = t2.id
WHERE t1.col1 = 'someValue'
gets translated to the following by the query optimizer:
SELECT * FROM t1
LEFT JOIN t2 ON t1.id = t2.id
WHERE t1.col1 = 'someValue'
UNION
SELECT * FROM t2
LEFT JOIN t1 ON t2.id = t1.id
WHERE t1.col1 = 'someValue'
So the order of tables has changed, but the predicate is still applied to t1, but t1 is now in the 'ON' clause. If t1.col1 is defined as NOT NULL
column, then this query will be null-rejected.
Any outer-join (left, right, full) that is null-rejected is converted to an inner-join by MySQL.
Thus the results you might be expecting might be completely different from what the MySQL is returning. You might think its a bug with MySQL's RIGHT JOIN, but that’s not right. Its just how the MySQL query optimizer works. So the developer in charge has to pay attention to these nuances when he/she is constructing the query.
I modified shA.t's query for more clarity:
-- t1 left join t2
SELECT t1.value, t2.value
FROM t1 LEFT JOIN t2 ON t1.value = t2.value
UNION ALL -- include duplicates
-- t1 right exclude join t2 (records found only in t2)
SELECT t1.value, t2.value
FROM t1 RIGHT JOIN t2 ON t1.value = t2.value
WHERE t1.value IS NULL
In SQLite you should do this:
SELECT *
FROM leftTable lt
LEFT JOIN rightTable rt ON lt.id = rt.lrid
UNION
SELECT lt.*, rl.* -- To match column set
FROM rightTable rt
LEFT JOIN leftTable lt ON lt.id = rt.lrid
You can do the following:
(SELECT
*
FROM
table1 t1
LEFT JOIN
table2 t2 ON t1.id = t2.id
WHERE
t2.id IS NULL)
UNION ALL
(SELECT
*
FROM
table1 t1
RIGHT JOIN
table2 t2 ON t1.id = t2.id
WHERE
t1.id IS NULL);
You can just convert a full outer join, e.g.
SELECT fields
FROM firsttable
FULL OUTER JOIN secondtable ON joincondition
into:
SELECT fields
FROM firsttable
LEFT JOIN secondtable ON joincondition
UNION ALL
SELECT fields (replacing any fields from firsttable with NULL)
FROM secondtable
WHERE NOT EXISTS (SELECT 1 FROM firsttable WHERE joincondition)
Or if you have at least one column, say foo, in firsttable that is NOT NULL, you can do:
SELECT fields
FROM firsttable
LEFT JOIN secondtable ON joincondition
UNION ALL
SELECT fields
FROM firsttable
RIGHT JOIN secondtable ON joincondition
WHERE firsttable.foo IS NULL
SELECT
a.name,
b.title
FROM
author AS a
LEFT JOIN
book AS b
ON a.id = b.author_id
UNION
SELECT
a.name,
b.title
FROM
author AS a
RIGHT JOIN
book AS b
ON a.id = b.author_id
I fix the response, and works include all rows (based on the response of Pavle Lekic):
(
SELECT a.* FROM tablea a
LEFT JOIN tableb b ON a.`key` = b.key
WHERE b.`key` is null
)
UNION ALL
(
SELECT a.* FROM tablea a
LEFT JOIN tableb b ON a.`key` = b.key
where a.`key` = b.`key`
)
UNION ALL
(
SELECT b.* FROM tablea a
right JOIN tableb b ON b.`key` = a.key
WHERE a.`key` is null
);
Use:
SELECT * FROM t1 FULL OUTER JOIN t2 ON t1.id = t2.id;
It can be recreated as follows:
SELECT t1.*, t2.*
FROM (SELECT * FROM t1 UNION SELECT name FROM t2) tmp
LEFT JOIN t1 ON t1.id = tmp.id
LEFT JOIN t2 ON t2.id = tmp.id;
Using a UNION or UNION ALL answer does not cover the edge case where the base tables have duplicated entries.
Explanation:
There is an edge case that a UNION or UNION ALL cannot cover. We cannot test this on MySQL as it doesn't support full outer joins, but we can illustrate this on a database that does support it:
WITH cte_t1 AS
(
SELECT 1 AS id1
UNION ALL SELECT 2
UNION ALL SELECT 5
UNION ALL SELECT 6
UNION ALL SELECT 6
),
cte_t2 AS
(
SELECT 3 AS id2
UNION ALL SELECT 4
UNION ALL SELECT 5
UNION ALL SELECT 6
UNION ALL SELECT 6
)
SELECT * FROM cte_t1 t1 FULL OUTER JOIN cte_t2 t2 ON t1.id1 = t2.id2;
This gives us this answer:
id1 id2
1 NULL
2 NULL
NULL 3
NULL 4
5 5
6 6
6 6
6 6
6 6
The UNION solution:
SELECT * FROM cte_t1 t1 LEFT OUTER JOIN cte_t2 t2 ON t1.id1 = t2.id2
UNION
SELECT * FROM cte_t1 t1 RIGHT OUTER JOIN cte_t2 t2 ON t1.id1 = t2.id2
Gives an incorrect answer:
id1 id2
NULL 3
NULL 4
1 NULL
2 NULL
5 5
6 6
The UNION ALL solution:
SELECT * FROM cte_t1 t1 LEFT OUTER join cte_t2 t2 ON t1.id1 = t2.id2
UNION ALL
SELECT * FROM cte_t1 t1 RIGHT OUTER JOIN cte_t2 t2 ON t1.id1 = t2.id2
Is also incorrect.
id1 id2
1 NULL
2 NULL
5 5
6 6
6 6
6 6
6 6
NULL 3
NULL 4
5 5
6 6
6 6
6 6
6 6
Whereas this query:
SELECT t1.*, t2.*
FROM (SELECT * FROM t1 UNION SELECT name FROM t2) tmp
LEFT JOIN t1 ON t1.id = tmp.id
LEFT JOIN t2 ON t2.id = tmp.id;
Gives the following:
id1 id2
1 NULL
2 NULL
NULL 3
NULL 4
5 5
6 6
6 6
6 6
6 6
The order is different, but otherwise matches the correct answer.
Use a cross join solution:
SELECT t1.*, t2.*
FROM table1 t1
INNER JOIN table2 t2
ON 1=1;
It is also possible, but you have to mention the same field names in select.
SELECT t1.name, t2.name FROM t1
LEFT JOIN t2 ON t1.id = t2.id
UNION
SELECT t1.name, t2.name FROM t2
LEFT JOIN t1 ON t1.id = t2.id
The SQL standard says full join on is inner join on rows union all unmatched left table rows extended by nulls union all right table rows extended by nulls. Ie inner join on rows union all rows in left join on but not inner join on union all rows in right join on but not inner join on.
Ie left join on rows union all right join on rows not in inner join on. Or if you know your inner join on result can't have null in a particular right table column then "right join on rows not in inner join on" are rows in right join on with the on condition extended by and that column is null.
Ie similarly right join on union all appropriate left join on rows.
From What is the difference between “INNER JOIN” and “OUTER JOIN”?:
(SQL Standard 2006 SQL/Foundation 7.7 Syntax Rules 1, General Rules 1 b, 3 c & d, 5 b.)