How do I limit the number of rows returned by an Oracle query after ordering?

You can use a subquery for this like

select *
from  
( select * 
  from emp 
  order by sal desc ) 
where ROWNUM <= 5;

Have also a look at the topic On ROWNUM and limiting results at Oracle/AskTom for more information.

Update: To limit the result with both lower and upper bounds things get a bit more bloated with

select * from 
( select a.*, ROWNUM rnum from 
  ( <your_query_goes_here, with order by> ) a 
  where ROWNUM <= :MAX_ROW_TO_FETCH )
where rnum  >= :MIN_ROW_TO_FETCH;

(Copied from specified AskTom-article)

Update 2: Starting with Oracle 12c (12.1) there is a syntax available to limit rows or start at offsets.

SELECT * 
FROM   sometable
ORDER BY name
OFFSET 20 ROWS FETCH NEXT 10 ROWS ONLY;

See this answer for more examples. Thanks to Krumia for the hint.

---

Starting from Oracle 12c R1 (12.1), there is a row limiting clause. It does not use familiar LIMIT syntax, but it can do the job better with more options. You can find the full syntax here. (Also read more on how this works internally in Oracle in this answer).

To answer the original question, here's the query:

SELECT * 
FROM   sometable
ORDER BY name
OFFSET 20 ROWS FETCH NEXT 10 ROWS ONLY;

(For earlier Oracle versions, please refer to other answers in this question)

Examples:

Following examples were quoted from linked page, in the hope of preventing link rot.

Setup

CREATE TABLE rownum_order_test (
  val  NUMBER
);

INSERT ALL
  INTO rownum_order_test
SELECT level
FROM   dual
CONNECT BY level <= 10;

COMMIT;

What's in the table?

SELECT val
FROM   rownum_order_test
ORDER BY val;

       VAL
----------
         1
         1
         2
         2
         3
         3
         4
         4
         5
         5
         6
         6
         7
         7
         8
         8
         9
         9
        10
        10

20 rows selected.

Get first N rows

SELECT val
FROM   rownum_order_test
ORDER BY val DESC
FETCH FIRST 5 ROWS ONLY;

       VAL
----------
        10
        10
         9
         9
         8

5 rows selected.

Get first N rows, if Nth row has ties, get all the tied rows

SELECT val
FROM   rownum_order_test
ORDER BY val DESC
FETCH FIRST 5 ROWS WITH TIES;

       VAL
----------
        10
        10
         9
         9
         8
         8

6 rows selected.

Top x% of rows

SELECT val
FROM   rownum_order_test
ORDER BY val
FETCH FIRST 20 PERCENT ROWS ONLY;

       VAL
----------
         1
         1
         2
         2

4 rows selected.

Using an offset, very useful for pagination

SELECT val
FROM   rownum_order_test
ORDER BY val
OFFSET 4 ROWS FETCH NEXT 4 ROWS ONLY;

       VAL
----------
         3
         3
         4
         4

4 rows selected.

You can combine offset with percentages

SELECT val
FROM   rownum_order_test
ORDER BY val
OFFSET 4 ROWS FETCH NEXT 20 PERCENT ROWS ONLY;

       VAL
----------
         3
         3
         4
         4

4 rows selected.

---

I did some performance testing for the following approaches:

Asktom

select * from (
  select a.*, ROWNUM rnum from (
    <select statement with order by clause>
  ) a where rownum <= MAX_ROW
) where rnum >= MIN_ROW

Analytical

select * from (
  <select statement with order by clause>
) where myrow between MIN_ROW and MAX_ROW

Short Alternative

select * from (
  select statement, rownum as RN with order by clause
) where a.rn >= MIN_ROW and a.rn <= MAX_ROW

Results

Table had 10 million records, sort was on an unindexed datetime row:

Explain plan showed same value for all three selects (323168)

But the winner is AskTom (with analytic following close behind)

Selecting first 10 rows took:

AskTom: 28-30 seconds

Analytical: 33-37 seconds

Short alternative: 110-140 seconds

Selecting rows between 100,000 and 100,010:

AskTom: 60 seconds

Analytical: 100 seconds

Selecting rows between 9,000,000 and 9,000,010:

AskTom: 130 seconds

Analytical: 150 seconds

---

An analytic solution with only one nested query:

SELECT * FROM
(
   SELECT t.*, Row_Number() OVER (ORDER BY name) MyRow FROM sometable t
) 
WHERE MyRow BETWEEN 10 AND 20;

Rank() could be substituted for Row_Number() but might return more records than you are expecting if there are duplicate values for name.

---

SQL Standard

Since version 12c Oracle supports the SQL:2008 Standard, which provides the following syntax to limit the SQL result set:

SELECT
    title
FROM
    post
ORDER BY
    id DESC
FETCH FIRST 50 ROWS ONLY

Oracle 11g and older versions

Prior to version 12c, to fetch the Top-N records, you had to use a derived table and the ROWNUM pseudocolumn:

SELECT *
FROM (
    SELECT
        title
    FROM
        post
    ORDER BY
        id DESC
)
WHERE ROWNUM <= 50

---

On Oracle 12c (see row limiting clause in SQL reference):

SELECT * 
FROM sometable
ORDER BY name
OFFSET 20 ROWS FETCH NEXT 10 ROWS ONLY;

---

Pagination queries with ordering are really tricky in Oracle.

Oracle provides a ROWNUM pseudocolumn that returns a number indicating the order in which the database selects the row from a table or set of joined views.

ROWNUM is a pseudocolumn that gets many people into trouble. A ROWNUM value is not permanently assigned to a row (this is a common misunderstanding). It may be confusing when a ROWNUM value is actually assigned. A ROWNUM value is assigned to a row after it passes filter predicates of the query but before query aggregation or sorting.

What is more, a ROWNUM value is incremented only after it is assigned.

This is why the followin query returns no rows:

 select * 
 from (select *
       from some_table
       order by some_column)
 where ROWNUM <= 4 and ROWNUM > 1; 

The first row of the query result does not pass ROWNUM > 1 predicate, so ROWNUM does not increment to 2. For this reason, no ROWNUM value gets greater than 1, consequently, the query returns no rows.

Correctly defined query should look like this:

select *
from (select *, ROWNUM rnum
      from (select *
            from skijump_results
            order by points)
      where ROWNUM <= 4)
where rnum > 1; 

Find out more about pagination queries in my articles on Vertabelo blog:

Oracle ROWNUM Explained

Top-N and pagination queries

---

As an extension of accepted answer Oracle internally uses ROW_NUMBER/RANK functions. OFFSET FETCH syntax is a syntax sugar.

It could be observed by using DBMS_UTILITY.EXPAND_SQL_TEXT procedure:

Preparing sample:

CREATE TABLE rownum_order_test (
  val  NUMBER
);

INSERT ALL
  INTO rownum_order_test
SELECT level
FROM   dual
CONNECT BY level <= 10;
COMMIT;

Query:

SELECT val
FROM   rownum_order_test
ORDER BY val DESC
FETCH FIRST 5 ROWS ONLY;

is regular:

SELECT "A1"."VAL" "VAL" 
FROM  (SELECT "A2"."VAL" "VAL","A2"."VAL" "rowlimit_$_0",
               ROW_NUMBER() OVER ( ORDER BY "A2"."VAL" DESC ) "rowlimit_$$_rownumber" 
      FROM "ROWNUM_ORDER_TEST" "A2") "A1" 
WHERE "A1"."rowlimit_$$_rownumber"<=5 ORDER BY "A1"."rowlimit_$_0" DESC;

db<>fiddle demo

Fetching expanded SQL text:

declare
  x VARCHAR2(1000);
begin
 dbms_utility.expand_sql_text(
        input_sql_text => '
          SELECT val
          FROM   rownum_order_test
          ORDER BY val DESC
          FETCH FIRST 5 ROWS ONLY',
        output_sql_text => x);

  dbms_output.put_line(x);
end;
/

WITH TIES is expanded as RANK:

declare
  x VARCHAR2(1000);
begin
 dbms_utility.expand_sql_text(
        input_sql_text => '
          SELECT val
          FROM   rownum_order_test
          ORDER BY val DESC
          FETCH FIRST 5 ROWS WITH TIES',
        output_sql_text => x);

  dbms_output.put_line(x);
end;
/

SELECT "A1"."VAL" "VAL" 
FROM  (SELECT "A2"."VAL" "VAL","A2"."VAL" "rowlimit_$_0",
              RANK() OVER ( ORDER BY "A2"."VAL" DESC ) "rowlimit_$$_rank" 
       FROM "ROWNUM_ORDER_TEST" "A2") "A1" 
WHERE "A1"."rowlimit_$$_rank"<=5 ORDER BY "A1"."rowlimit_$_0" DESC

and offset:

declare
  x VARCHAR2(1000);
begin
 dbms_utility.expand_sql_text(
        input_sql_text => '
          SELECT val
FROM   rownum_order_test
ORDER BY val
OFFSET 4 ROWS FETCH NEXT 4 ROWS ONLY',
        output_sql_text => x);

  dbms_output.put_line(x);
end;
/


SELECT "A1"."VAL" "VAL" 
FROM  (SELECT "A2"."VAL" "VAL","A2"."VAL" "rowlimit_$_0",
             ROW_NUMBER() OVER ( ORDER BY "A2"."VAL") "rowlimit_$$_rownumber" 
       FROM "ROWNUM_ORDER_TEST" "A2") "A1" 
       WHERE "A1"."rowlimit_$$_rownumber"<=CASE  WHEN (4>=0) THEN FLOOR(TO_NUMBER(4)) 
             ELSE 0 END +4 AND "A1"."rowlimit_$$_rownumber">4 
ORDER BY "A1"."rowlimit_$_0"

---

Less SELECT statements. Also, less performance consuming. Credits to: anibal@upf.br

SELECT *
    FROM   (SELECT t.*,
                   rownum AS rn
            FROM   shhospede t) a
    WHERE  a.rn >= in_first
    AND    a.rn <= in_first;

---

With 21c version, you can simple apply a limit as follows:

select * from course where ROWNUM <=10;

---

I'v started preparing for Oracle 1z0-047 exam, validated against 12c While prepping for it i came across a 12c enhancement known as 'FETCH FIRST' It enables you to fetch rows /limit rows as per your convenience. Several options are available with it

- FETCH FIRST n ROWS ONLY
 - OFFSET n ROWS FETCH NEXT N1 ROWS ONLY // leave the n rows and display next N1 rows
 - n % rows via FETCH FIRST N PERCENT ROWS ONLY

Example:

Select * from XYZ a
order by a.pqr
FETCH FIRST 10 ROWS ONLY

---

For each row returned by a query, the ROWNUM pseudocolumn returns a number indicating the order in which Oracle selects the row from a table or set of joined rows. The first row selected has a ROWNUM of 1, the second has 2, and so on.

  SELECT * FROM sometable1 so
    WHERE so.id IN (
    SELECT so2.id from sometable2 so2
    WHERE ROWNUM <=5
    )
    AND ORDER BY so.somefield AND ROWNUM <= 100 

I have implemented this in oracle server 11.2.0.1.0

---

select * FROM (SELECT 
   ROW_NUMBER() OVER (ORDER BY sal desc),* AS ROWID, 
 FROM EMP ) EMP  where ROWID=5

greater then values find out

select * FROM (SELECT 
       ROW_NUMBER() OVER (ORDER BY sal desc),* AS ROWID, 
     FROM EMP ) EMP  where ROWID>5

less then values find out

select * FROM (SELECT 
       ROW_NUMBER() OVER (ORDER BY sal desc),* AS ROWID, 
     FROM EMP ) EMP  where ROWID=5

---

(untested) something like this may do the job

WITH
base AS
(
    select *                   -- get the table
    from sometable
    order by name              -- in the desired order
),
twenty AS
(
    select *                   -- get the first 30 rows
    from base
    where rownum < 30
    order by name              -- in the desired order
)
select *                       -- then get rows 21 .. 30
from twenty
where rownum > 20
order by name                  -- in the desired order

There is also the analytic function rank, that you can use to order by.

---