EXPLAIN EXTENDED

How to create fast database queries

INNER JOIN vs. CROSS APPLY

with 7 comments

From Stack Overflow:

Can anyone give me a good example of when CROSS APPLY makes a difference in those cases where INNER JOIN will work as well?

This is of course SQL Server.

A quick reminder on the terms.

INNER JOIN is the most used construct in SQL: it joins two tables together, selecting only those row combinations for which a JOIN condition is true.

This query:

SELECT  *
FROM    table1
JOIN    table2
ON      table2.b = table1.a

reads:

For each row from table1, select all rows from table2 where the value of field b is equal to that of field a

Note that this condition can be rewritten as this:

SELECT  *
FROM    table1, table2
WHERE   table2.b = table1.a

, in which case it reads as following:

Make a set of all possible combinations of rows from table1 and table2 and of this set select only those rows where the value of field b is equal to that of field a

These conditions are worded differently, but they yield the same result and database systems are aware of that. Usually both these queries are optimized to use the same execution plan.

The former syntax is called ANSI syntax, and it is generally considered more readable and is recommended to use.

However, it didn't get into Oracle until recently, that's why there are many hardcore Oracle developers that are just used to the latter syntax.

Actually, it's a matter of taste.

To use JOINs (with whatever syntax), both sets you are joining must be self-sufficient, i. e. the sets should not depend on each other. You can query both sets without ever knowing the contents on another set.

But for some tasks the sets are not self-sufficient. For instance, let's consider the following query:

We table table1 and table2. table1 has a column called rowcount.

For each row from table1 we need to select first rowcount rows from table2, ordered by table2.id

We cannot formulate a join condition here. The join condition, should it exists, would involve the row number, which is not present in table2, and there is no way to calculate a row number only from the values of columns of any given row in table2.

That's where the CROSS APPLY can be used.

CROSS APPLY is a Microsoft's extension to SQL, which was originally intended to be used with table-valued functions (TVF's).

The query above would look like this:

SELECT  *
FROM    table1
CROSS APPLY
        (
        SELECT  TOP (table1.rowcount) *
        FROM    table2
        ORDER BY
                 id
        ) t2

For each from table1, select first table1.rowcount rows from table2 ordered by id

The sets here are not self-sufficient: the query uses values from table1 to define the second set, not to JOIN with it.

The exact contents of t2 are not known until the corresponding row from table1 is selected.

I previously said that there is no way to join these two sets, which is true as long as we consider the sets as is. However, we can change the second set a little so that we get an additional computed field we can later join on.

The first option to do that is just count all preceding rows in a subquery:

SELECT  *
FROM    table1 t1
JOIN    (
        SELECT  t2o.*,
                (
                SELECT  COUNT(*)
                FROM    table2 t2i
                WHERE   t2i.id <= t2o.id
                ) AS rn
        FROM    table2 t2o
        ) t2
ON      t2.rn <= t1.rowcount

The second option is to use a window function, also available in SQL Server since version 2005:

SELECT  *
FROM    table1 t1
JOIN    (
        SELECT  t2o.*, ROW_NUMBER() OVER (ORDER BY id) AS rn
        FROM    table2 t2o
        ) t2
ON      t2.rn <= t1.rowcount

This functions returns the ordinal number a row would have be the ORDER BY condition used in the function applied to the whole query.

This is essentially the same result as the subquery used in the previous query.

Now, let's create the sample tables and check all these solutions for efficiency:

SET NOCOUNT ON
GO
DROP TABLE [20090716_cross].table1
DROP TABLE [20090716_cross].table2
DROP SCHEMA [20090716_cross]
GO
CREATE SCHEMA [20090716_cross]
CREATE TABLE table1
        (
        id INT NOT NULL PRIMARY KEY,
        row_count INT NOT NULL
        )
CREATE TABLE table2
        (
        id INT NOT NULL PRIMARY KEY,
        value VARCHAR(20) NOT NULL
        )
GO
BEGIN TRANSACTION
DECLARE @cnt INT
SET @cnt = 1
WHILE @cnt <= 100000
BEGIN
        INSERT
        INTO    [20090716_cross].table2 (id, value)
        VALUES  (@cnt, 'Value ' + CAST(@cnt AS VARCHAR))
        SET @cnt = @cnt + 1
END
INSERT
INTO    [20090716_cross].table1 (id, row_count)
SELECT  TOP 5
        id, id % 2 + 1
FROM    [20090716_cross].table2
ORDER BY
        id
COMMIT
GO

table2 contains 100,000 rows with sequential ids.

table1 contains the following:

id row_count
1 2
2 1
3 2
4 1
5 2

Now let's run the first query (with COUNT):

SELECT  *
FROM    [20090716_cross].table1 t1
JOIN    (
        SELECT  t2o.*,
                (
                SELECT  COUNT(*)
                FROM    [20090716_cross].table2 t2i
                WHERE   t2i.id <= t2o.id
                ) AS rn
        FROM    [20090716_cross].table2 t2o
        ) t2
ON      t2.rn <= t1.row_count
ORDER BY
        t1.id, t2.id
id row_count id value rn
1 2 1 Value 1 1
1 2 2 Value 2 2
2 1 1 Value 1 1
3 2 1 Value 1 1
3 2 2 Value 2 2
4 1 1 Value 1 1
5 2 1 Value 1 1
5 2 2 Value 2 2
8 rows fetched in 0.0000s (498.4063s)
Table 'table1'. Scan count 2, logical reads 200002, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 
Table 'Worktable'. Scan count 100000, logical reads 8389920, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 
Table 'Worktable'. Scan count 0, logical reads 0, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 
Table 'table2'. Scan count 4, logical reads 1077, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 

SQL Server Execution Times:
   CPU time = 947655 ms,  elapsed time = 498385 ms. 

This query, as was expected, is very unoptimal. It runs for more than 500 seconds.

Here's the query plan:

SELECT
  Sort
    Compute Scalar
      Parallelism (Gather Streams)
        Inner Join (Nested Loops)
          Inner Join (Nested Loops)
            Clustered Index Scan ([20090716_cross].[table2])
            Compute Scalar
              Stream Aggregate
                Eager Spool
                  Clustered Index Scan ([20090716_cross].[table2])
          Clustered Index Scan ([20090716_cross].[table1])

For each row selected from table2, it counts all previous rows again an again, never recording the intermediate result. The complexity of such an algorithm is O(n^2), that's why it takes so long.

Let's run he second query, which uses ROW_NUMBER():

SELECT  *
FROM    [20090716_cross].table1 t1
JOIN    (
        SELECT  t2o.*, ROW_NUMBER() OVER (ORDER BY id) AS rn
        FROM    [20090716_cross].table2 t2o
        ) t2
ON      t2.rn <= t1.row_count
ORDER BY
        t1.id, t2.id

id row_count id value rn
1 2 1 Value 1 1
1 2 2 Value 2 2
2 1 1 Value 1 1
3 2 1 Value 1 1
3 2 2 Value 2 2
4 1 1 Value 1 1
5 2 1 Value 1 1
5 2 2 Value 2 2
8 rows fetched in 0.0006s (0.5781s)
Table 'Worktable'. Scan count 1, logical reads 214093, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 
Table 'table2'. Scan count 1, logical reads 522, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 
Table 'table1'. Scan count 1, logical reads 2, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 

SQL Server Execution Times:
   CPU time = 578 ms,  elapsed time = 579 ms. 

This is much faster, only 0.5 ms.

Let's look into the query plan:

SELECT
  Inner Join (Nested Loops)
    Clustered Index Scan ([20090716_cross].[table1])
  Lazy Spool
    Sequence Project (Compute Scalar)
      Compute Scalar
        Segment
          Clustered Index Scan ([20090716_cross].[table2])

This is much better, since this query plan keeps the intermediate results while calculating the ROW_NUMBER.

However, it still calculates ROW_NUMBERs for all 100,000 of rows in table2, then puts them into a temporary index over rn created by Lazy Spool, and uses this index in a nested loop to range the rns for each row from table1.

Calculating and indexing all ROW_NUMBERs is quite expensive, that's why we see 214,093 logical reads in the query statistics.

Finally, let's try a CROSS APPLY:

SELECT  *
FROM    [20090716_cross].table1 t1
CROSS APPLY
        (
        SELECT  TOP (t1.row_count) *
        FROM    [20090716_cross].table2
        ORDER BY
                id
        ) t2
ORDER BY
        t1.id, t2.id


id row_count id value
1 2 1 Value 1
1 2 2 Value 2
2 1 1 Value 1
3 2 1 Value 1
3 2 2 Value 2
4 1 1 Value 1
5 2 1 Value 1
5 2 2 Value 2
8 rows fetched in 0.0004s (0.0008s)
Table 'table2'. Scan count 5, logical reads 10, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 
Table 'table1'. Scan count 1, logical reads 2, physical reads 0, read-ahead reads 0, lob logical reads 0, lob physical reads 0, lob read-ahead reads 0. 

SQL Server Execution Times:
   CPU time = 0 ms,  elapsed time = 1 ms. 

This query is instant, as it should be.

The plan is quite simple:

SELECT
  Inner Join (Nested Loops)
    Clustered Index Scan ([20090716_cross].[table1])
    Top
      Clustered Index Scan ([20090716_cross].[table2])

For each row from table1, it just takes first row_count rows from table2. So simple and so fast.

Summary:

While most queries which employ CROSS APPLY can be rewritten using an INNER JOIN, CROSS APPLY can yield better execution plan and better performance, since it can limit the set being joined yet before the join occurs.

Written by Quassnoi

July 16th, 2009 at 11:00 pm

Posted in SQL Server

7 Responses to 'INNER JOIN vs. CROSS APPLY'

Subscribe to comments with RSS or TrackBack to 'INNER JOIN vs. CROSS APPLY'.

  1. Extremely helpful, thank-you.

    Dave

    1 Feb 13 at 01:36

  2. In the first two alternative queries you post, you have:
    t1.rowcount = t2.rn

    I think you meant >=, otherwise the result is totally different from the output query

    mas

    7 Mar 13 at 01:27

  3. @mas: thanks, corrected.

    Quassnoi

    7 Mar 13 at 14:02

  4. Thanks very much for this explanation of INNER JOIN vs. CROSS APPLY. I finally get it — the key idea (for me) being “The sets here are not self-sufficient: the query uses values from table1 to define the second set, not to JOIN with it.”

    Sam

    12 May 13 at 22:58

  5. “This is much faster, only 0.5 ms.”

    Did you mean to write “0.5 s”? As shown in the screenshots.

    theghostofc

    13 Aug 13 at 12:11

  6. You mister have just optimized my query from 1200 ms to 10 ms. THANK YOU!

    I was using rowNumber over Name column but joining result over id and that was killing MSSQL. If rowNumber was also over Id everything was fine.

    Milan

    5 Sep 13 at 13:09

  7. Thanks for the detailed explanation, this CROSS APPLY is becoming an interview question now but still difficult to find a good explanation, even worse when comparing with other execution plans like our beloved JOIN. THanks Again.

    Pap Perlada

    10 Jun 14 at 00:15

Leave a Reply