In some cases, MySQL can use an index to satisfy an
ORDER BY
clause without doing any extra
sorting.
The index can also be used even if the ORDER
BY
does not match the index exactly, as long as all of
the unused portions of the index and all the extra
ORDER BY
columns are constants in the
WHERE
clause. The following queries use the
index to resolve the ORDER BY
part:
SELECT * FROM t1 ORDER BYkey_part1
,key_part2
,... ; SELECT * FROM t1 WHEREkey_part1
=constant
ORDER BYkey_part2
; SELECT * FROM t1 ORDER BYkey_part1
DESC,key_part2
DESC; SELECT * FROM t1 WHEREkey_part1
=1 ORDER BYkey_part1
DESC,key_part2
DESC;
In some cases, MySQL cannot use indexes to
resolve the ORDER BY
, although it still uses
indexes to find the rows that match the WHERE
clause. These cases include the following:
You use ORDER BY
on different keys:
SELECT * FROM t1 ORDER BYkey1
,key2
;
You use ORDER BY
on nonconsecutive parts
of a key:
SELECT * FROM t1 WHEREkey2
=constant
ORDER BYkey_part2
;
You mix ASC
and DESC
:
SELECT * FROM t1 ORDER BYkey_part1
DESC,key_part2
ASC;
The key used to fetch the rows is not the same as the one
used in the ORDER BY
:
SELECT * FROM t1 WHEREkey2
=constant
ORDER BYkey1
;
You use ORDER BY
with an expression that
includes terms other than the key column name:
SELECT * FROM t1 ORDER BY ABS(key
); SELECT * FROM t1 ORDER BY -key
;
You are joining many tables, and the columns in the
ORDER BY
are not all from the first
nonconstant table that is used to retrieve rows. (This is
the first table in the
EXPLAIN
output that does not
have a const
join type.)
You have different ORDER BY
and
GROUP BY
expressions.
You index only a prefix of a column named in the
ORDER BY
clause. In this case, the index
cannot be used to fully resolve the sort order. For example,
if you have a CHAR(20)
column, but index only the first 10 bytes, the index cannot
distinguish values past the 10th byte and a
filesort
will be needed.
The type of table index used does not store rows in order.
For example, this is true for a HASH
index in a MEMORY
table.
Availability of an index for sorting may be affected by the use
of column aliases. Suppose that the column
t1.a
is indexed. In this statement, the name
of the column in the select list is a
. It
refers to t1.a
, so for the reference to
a
in the ORDER BY
, the
index can be used:
SELECT a FROM t1 ORDER BY a;
In this statement, the name of the column in the select list is
also a
, but it is the alias name. It refers
to ABS(a)
, so for the reference to
a
in the ORDER BY
, the
index cannot be used:
SELECT ABS(a) AS a FROM t1 ORDER BY a;
In the following statement, the ORDER BY
refers to a name that is not the name of a column in the select
list. But there is a column in t1
named
a
, so the ORDER BY
uses
that, and the index can be used. (The resulting sort order may
be completely different from the order for
ABS(a)
, of course.)
SELECT ABS(a) AS b FROM t1 ORDER BY a;
By default, MySQL sorts all GROUP BY
queries as if you
specified col1
,
col2
, ...ORDER BY
in the query as
well. If you include an col1
,
col2
, ...ORDER BY
clause
explicitly that contains the same column list, MySQL optimizes
it away without any speed penalty, although the sorting still
occurs. If a query includes GROUP BY
but you
want to avoid the overhead of sorting the result, you can
suppress sorting by specifying ORDER BY NULL
.
For example:
INSERT INTO foo SELECT a, COUNT(*) FROM bar GROUP BY a ORDER BY NULL;
With EXPLAIN SELECT ... ORDER BY
, you can
check whether MySQL can use indexes to resolve the query. It
cannot if you see Using filesort
in the
Extra
column. See
Section 7.2.1, “Optimizing Queries with EXPLAIN
”.
MySQL has two filesort
algorithms for sorting
and retrieving results. The original method uses only the
ORDER BY
columns. The modified method uses
not just the ORDER BY
columns, but all the
columns used in the query.
The optimizer selects which filesort
algorithm to use. It normally uses the modified algorithm except
when BLOB
or
TEXT
columns are involved, in
which case it uses the original algorithm.
The original filesort
algorithm works as
follows:
Read all rows according to key or by table scanning. Rows
that do not match the WHERE
clause are
skipped.
For each row, store a pair of values in a buffer (the sort
key and the row pointer). The size of the buffer is the
value of the
sort_buffer_size
system
variable.
When the buffer gets full, run a qsort (quicksort) on it and store the result in a temporary file. Save a pointer to the sorted block. (If all pairs fit into the sort buffer, no temporary file is created.)
Repeat the preceding steps until all rows have been read.
Do a multi-merge of up to MERGEBUFF
(7)
regions to one block in another temporary file. Repeat until
all blocks from the first file are in the second file.
Repeat the following until there are fewer than
MERGEBUFF2
(15) blocks left.
On the last multi-merge, only the pointer to the row (the last part of the sort key) is written to a result file.
Read the rows in sorted order by using the row pointers in
the result file. To optimize this, we read in a big block of
row pointers, sort them, and use them to read the rows in
sorted order into a row buffer. The size of the buffer is
the value of the
read_rnd_buffer_size
system
variable. The code for this step is in the
sql/records.cc
source file.
One problem with this approach is that it reads rows twice: One
time when evaluating the WHERE
clause, and
again after sorting the pair values. And even if the rows were
accessed successively the first time (for example, if a table
scan is done), the second time they are accessed randomly. (The
sort keys are ordered, but the row positions are not.)
The modified filesort
algorithm incorporates
an optimization such that it records not only the sort key value
and row position, but also the columns required for the query.
This avoids reading the rows twice. The modified
filesort
algorithm works like this:
Read the rows that match the WHERE
clause.
For each row, record a tuple of values consisting of the sort key value and row position, and also the columns required for the query.
Sort the tuples by sort key value
Retrieve the rows in sorted order, but read the required columns directly from the sorted tuples rather than by accessing the table a second time.
Using the modified filesort
algorithm, the
tuples are longer than the pairs used in the original method,
and fewer of them fit in the sort buffer (the size of which is
given by sort_buffer_size
). As
a result, it is possible for the extra I/O to make the modified
approach slower, not faster. To avoid a slowdown, the
optimization is used only if the total size of the extra columns
in the sort tuple does not exceed the value of the
max_length_for_sort_data
system
variable. (A symptom of setting the value of this variable too
high is that you should see high disk activity and low CPU
activity.)
For slow queries for which filesort
is not
used, you might try lowering
max_length_for_sort_data
to a
value that is appropriate to trigger a
filesort
.
If you want to increase ORDER BY
speed, check
whether you can get MySQL to use indexes rather than an extra
sorting phase. If this is not possible, you can try the
following strategies:
Increase the size of the
sort_buffer_size
variable.
Increase the size of the
read_rnd_buffer_size
variable.
Use less RAM per row by declaring columns only as large as
they need to be to hold the values stored in them. For
example, CHAR(16)
is better than
CHAR(200)
if values never exceed 16
characters.
Change tmpdir
to point to a
dedicated file system with large amounts of free space.
Also, this option accepts several paths that are used in
round-robin fashion, so you can use this feature to spread
the load across several directories. Paths should be
separated by colon characters
(“:
”) on Unix and semicolon
characters (“;
”) on Windows,
NetWare, and OS/2. The paths should be for directories in
file systems that are located on different
physical disks, not different
partitions on the same disk.
User Comments
To be certain that this optimization is done you often need to force the correct index to be used with FORCE INDEX.
Another problem is that the optimization is lost when you have an OR on a prefix of the index:
SELECT * FROM t1 FORCE INDEX (key1_key2_key3)
WHERE key1=1 OR key1=2 ORDER BY key2,key3 LIMIT 5, 5;
This can be worked around by using UNION:
(SELECT * FROM t1 FORCE INDEX (key1_key2_key3)
WHERE key1=1 ORDER BY key2,key3 LIMIT 10)
UNION
(SELECT * FROM t1 FORCE INDEX (key1_key2_key3)
WHERE key1=2 ORDER BY key2,key3 LIMIT 10)
ORDER BY key2,key3 LIMIT 5, 5
This rewrite can make the query go ~20000 times faster on a table with ~2M rows.
===========================================================
Posted by Ravenous Bugblatter Beast on August 15 2006
If you are selecting a wide result set and using ORDER BY RAND() with a LIMIT, you can often speed things up by changing a query of the form:
SELECT id, col1, col2, ... , colN FROM tab WHERE conditions ORDER BY RAND() LIMIT m
To a query of the form:
SELECT id, col1, col2, ... , colN FROM tab WHERE id IN (SELECT id FROM tab WHERE conditions ORDER BY RAND() LIMIT m)
Although the second query has to perform an additional select, it only has to sort a result set containing the single id column, rather than the full result set you are returning from the query.
===========================================================
it's true, but if we have a big database with 1000ths of table rows and we must to join them ...... so oooops ;)
Lately I saw the following link http://jan.kneschke.de/projects/mysql/order-by-rand/ .....you can read here how to get more speed from your executing query in some cases. First read the explanation and see bottom of the page:
============================================================
Performance
Now let's see what happends to our performance. We have 3 different queries for solving our problems.
* Q1. ORDER BY RAND()
* Q2. RAND() * MAX(ID)
* Q3. RAND() * MAX(ID) + ORDER BY ID
Q1 is expected to cost N * log2(N), Q2 and Q3 are nearly constant.
The get real values we filled the table with N rows ( one thousand to one million) and executed each query 1000 times.
----------------------------------------------------------
Rows ||100 ||1.000 ||10.000 ||100.000 ||1.000.000
----------------------------------------------------------
Q1||0:00.718s||0:02.092s||0:18.684s||2:59.081s||58:20.000s
Q2||0:00.519s||0:00.607s||0:00.614s||0:00.628s||0:00.637s
Q3||0:00.570s||0:00.607s||0:00.614s|0:00.628s ||0:00.637s
----------------------------------------------------------
As you can see the plain ORDER BY RAND() is already behind the optimized query at only 100 rows in the table.
============================================================
All this is much faster than an ORDER BY RAND(), the problem is, it does not do the same.
Imagine you have a table with two columns (id, name) of thousands of names of people.
If you want to produce a result of only one record, then this is by far the fastest way to do it. You select a random ID from the table, and query it directly.
The problem is, if you want to retrieve two random records from the table.
ORDER BY RAND() will give you, for example, ID 390 and ID 4957 in a result, but the sugested way, will calculate a random index (ex. 390) and return two records based on that index, ID 390 and ID 391...
This means, that the order of the result is not random, only the first "fetched" ID. If you query for more than two records, it will give you ID 390, 391, 392, 393...
In other words, this does not produce an actual random selected result on more than one record.
When performing an ORDER BY col1 ASC - where col1 is an ENUM data type - the ORDER BY clause will not order by the natural language order, but the order of the values in the definition of the ENUM field.
To order by an ENUM field make sure the ENUM values are in alphabetical order or reverse-alphabetical order and then perform the ORDER BY clause.
If you need to "hard" sort a table, i.e. you want the table to be sorted in some predefined order, so that your query doesn't have to do it, use the following syntax:
alter table `yourtablename` order by `yourfieldname` desc;
In this case, when you execute:
select * from `yourtablename` limit 1;
you get the record for which yourfieldname is the largest.
* Q2. RAND() * MAX(ID)
* Q3. RAND() * MAX(ID) + ORDER BY ID
You should NOT use these examples, because it could return invalid data:
With that you receive the highest stored index. Then a random number between 0 and MAX(id) is returned. 0 is a unwanted result. Imagine ID=3 is returned and this row has been deleted. Then this is a unwanted result, too.
If you need only 1 row in case of quotes or similar things then you ever should use the LIMIT 1 at the end of the query.
Add your own comment.