Oracle中ROWNUM的使用技巧

Oracle中ROWNUM的使用技巧
作者： fuyuncat

来源： www.HelloDBA.com

ROWNUM是一种伪列，它会根据返回记录生成一个序列化的数字。利用ROWNUM，我们可以生产一些原先难以实现的结果输出，但因为它是伪列的这个特殊性，我们在使用时也需要注意一些事项，不要掉入“陷阱”。下面就介绍一下它的使用技巧及注意事项。

1         特殊结果输出
利用ROWNUM，我们可以做到一些特殊方式的输出。

1.1     Top N结果输出
我们如果希望取输出结果的前面几条数据，通过ROWNUM可以轻松实现：



SQL> select * from t_test4
  2  where rownum <= 5;

USERNAME                          USER_ID CREATED
------------------------------ ---------- ---------
WOW                                    71 26-APR-07
CS2                                    70 15-JAN-07
3                                      69 01-NOV-06
DMP                                    68 12-OCT-06
PROFILER                               67 05-SEP-06


但是，如果你希望对一个排序结果取Top N数据的话，使用ROWNUM存在一些“陷阱”，我们后面部分会介绍这些“陷阱”并且说明如何避免。

1.2     分页查询
利用ROWNUM对结果进行分页，下面返回结果中的第6到第10条记录：

SQL> select * from
  2  (
  3  select a.*, rownum as rn from css_bl_view a
  4  where capture_phone_num = '(1) 925-4604800'
  5  ) b
  6  where b.rn between 6 and 10;

6 rows selected.


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=2770 Card=2183 Bytes =7166789)
   1    0   VIEW (Cost=2770 Card=2183 Bytes=7166789)
   2    1     COUNT
   3    2       TABLE ACCESS (FULL) OF 'CSS_BL_VIEW' (Cost=2770 Card=2183 Bytes=1305434)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
      29346  consistent gets
      29190  physical reads
          0  redo size
       7328  bytes sent via SQL*Net to client
        234  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          5  rows processed


另外一种实现方式：



SQL> select * from css_bl_view a
  2  where capture_phone_num = '(1) 925-4604800'
  3  and rownum <= 10
  4  minus
  5  select * from css_bl_view a
  6  where capture_phone_num = '(1) 925-4604800'
  7  and rownum <= 5
  8  ;


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=5920 Card=10 Bytes=8970)
   1    0   MINUS
   2    1     SORT (UNIQUE) (Cost=2960 Card=10 Bytes=5980)
   3    2       COUNT (STOPKEY)
   4    3         TABLE ACCESS (FULL) OF 'CSS_BL_VIEW' (Cost=2770 Card=2183 Bytes=1305434)
   5    1     SORT (UNIQUE) (Cost=2960 Card=5 Bytes=2990)
   6    5       COUNT (STOPKEY)
   7    6         TABLE ACCESS (FULL) OF 'CSS_BL_VIEW' (Cost=2770 Card=2183 Bytes=1305434)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
         62  consistent gets
         50  physical reads
          0  redo size
       7232  bytes sent via SQL*Net to client
        234  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          2  sorts (memory)
          0  sorts (disk)
          5  rows processed


第三种实现方式：



SQL> select * from
  2  (
  3  select a.*, rownum as rn from css_bl_view a
  4  where capture_phone_num = '(1) 925-4604800'
  5  and rownum <= 10
  6  ) b
  7  where b.rn > 5;


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=2770 Card=10 Bytes=32830)
   1    0   VIEW (Cost=2770 Card=10 Bytes=32830)
   2    1     COUNT (STOPKEY)
   3    2       TABLE ACCESS (FULL) OF 'CSS_BL_VIEW' (Cost=2770 Card=2183 Bytes=1305434)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
         35  consistent gets
         30  physical reads
          0  redo size
       7271  bytes sent via SQL*Net to client
        234  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          5  rows processed


这里特地将三种实现方式的查询计划及统计数据打印出来，大家可以比较一下3中方式的性能。

1.3     利用ROWNUM做分组子排序
对于以下表T_TEST4的内容：



OWNER                                   NAME
------------------------------------------------------
STRMADMIN                               STREAMS_QUEUE
APARKMAN                                JOB_QUEUE
SYS                                     AQ$_AQ_SRVNTFN_TABLE_E
SYS                                     AQ$_KUPC$DATAPUMP_QUETAB_E
APARKMAN                                AQ$_JMS_TEXT_E
STRMADMIN                               AQ$_STREAMS_QUEUE_TABLE_E
SYS                                     AQ$_SCHEDULER$_EVENT_QTAB_E
…


如果我们希望结果按照OWNER进行分组后，再对每组中成员进行编号，结果类似如下：



OWNER                                   NO NAME
------------------------------------------------------
APARKMAN                                1 JOB_QUEUE
                                        2 AQ$_JMS_TEXT_E
STRMADMIN                               1 STREAMS_QUEUE
                                        2 AQ$_STREAMS_QUEUE_TABLE_E
SYS                                     1 AQ$_AQ_SRVNTFN_TABLE_E
                                        2 AQ$_KUPC$DATAPUMP_QUETAB_E
                                        3 AQ$_SCHEDULER$_EVENT_QTAB_E
…


在没有ROWNUM时要实现这样的功能会很复杂，但通过ROWNUM我们可以轻松实现：



SQL> SELECT DECODE(ROWNUM-min_sno,0,a.owner,NULL) owner,DECODE(ROWNUM-min_sno,0,1,rownum+1-min_sno) sno, a.name
  2  FROM (SELECT *
  3        FROM t_test8
  4        ORDER BY owner, name ) a,
  5       (SELECT owner, MIN(rownum) min_sno
  6        FROM( SELECT *
  7              FROM t_test8
  8              ORDER BY owner, name)
  9        GROUP BY owner) b
10  WHERE a.owner=b.owner;

OWNER                                 SNO NAME
------------------------------ ---------- ------------------------------
APARKMAN                                1 JOB_QUEUE
                                        2 AQ$_JMS_TEXT_E
STRMADMIN                               1 STREAMS_QUEUE
                                        2 AQ$_STREAMS_QUEUE_TABLE_E
SYS                                     1 AQ$_AQ_SRVNTFN_TABLE_E
                                        2 AQ$_KUPC$DATAPUMP_QUETAB_E
                                        3 AQ$_SCHEDULER$_EVENT_QTAB_E
                                        4 AQ$_SCHEDULER$_JOBQTAB_E
                                        5 AQ$_STREAMS_QUEUE_TABLE_E
                                        6 AQ$_SYS$SERVICE_METRICS_TAB_E
                                        7 AQ$_AQ_EVENT_TABLE_E
                                        8 AQ$_AQ$_MEM_MC_E
                                        9 AQ$_ALERT_QT_E
                                       10 ALERT_QUE
                                       11 AQ_EVENT_TABLE_Q
                                       12 SYS$SERVICE_METRICS
                                       13 STREAMS_QUEUE
                                       14 SRVQUEUE
                                       15 SCHEDULER$_JOBQ
                                       16 SCHEDULER$_EVENT_QUEUE
                                       17 AQ_SRVNTFN_TABLE_Q
SYSMAN                                  1 AQ$_MGMT_NOTIFY_QTABLE_E
                                        2 MGMT_NOTIFY_Q
SYSTEM                                  1 DEF$_AQERROR
                                        2 DEF$_AQCALL
                                        3 AQ$_DEF$_AQERROR_E
                                        4 AQ$_DEF$_AQCALL_E
WMSYS                                   1 AQ$_WM$EVENT_QUEUE_TABLE_E
                                        2 WM$EVENT_QUEUE

29 rows selected.
2         性能
我们很多程序员在确认某个表中是否有相应数据时，喜欢加上ROWNUM=1，其思路就是只要存在一条数据就说明有相应数据，查询就可以直接返回了，这样就能提高性能了。但是在10G之前，使用ROWNUM=1是不能达到预期的性能效果的，而是需要通过<2或<=1作为过滤条件才能达到预期效果，看以下查询计划：



SQL> select * from t_test1
  2  where object_id <100
  3  and rownum = 1;


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=37 Card=1 Bytes=86)
   1    0   COUNT (STOPKEY)
   2    1     TABLE ACCESS (BY INDEX ROWID) OF 'T_TEST1' (Cost=37 Card=89 Bytes=7654)
   3    2       INDEX (RANGE SCAN) OF 'T_TEST1_PK' (UNIQUE) (Cost=2 Card=89)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
         62  consistent gets
          0  physical reads
          0  redo size
        654  bytes sent via SQL*Net to client
        234  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

SQL> select * from t_test1
  2  where object_id <100
  3  and rownum <= 1;


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=37 Card=1 Bytes=86)
   1    0   COUNT (STOPKEY)
   2    1     TABLE ACCESS (BY INDEX ROWID) OF 'T_TEST1' (Cost=37 Card=89 Bytes=7654)
   3    2       INDEX (RANGE SCAN) OF 'T_TEST1_PK' (UNIQUE) (Cost=2 Card=89)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          3  consistent gets
          0  physical reads
          0  redo size
        654  bytes sent via SQL*Net to client
        234  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

SQL> /


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=37 Card=1 Bytes=86)
   1    0   COUNT (STOPKEY)
   2    1     TABLE ACCESS (BY INDEX ROWID) OF 'T_TEST1' (Cost=37 Card=89 Bytes=7654)
   3    2       INDEX (RANGE SCAN) OF 'T_TEST1_PK' (UNIQUE) (Cost=2 Card=89)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          3  consistent gets
          0  physical reads
          0  redo size
        654  bytes sent via SQL*Net to client
        234  bytes received via SQL*Net from client
          4  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed


10G以后，这个问题就被修正了：



SQL> select * from t_test1
  2  where rownum = 1;


Execution Plan
----------------------------------------------------------
Plan hash value: 536364188

------------------------------------------------------------------------------
| Id  | Operation          | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |         |     1 |    86 |     2   (0)| 00:00:01 |
|*  1 |  COUNT STOPKEY     |         |       |       |            |          |
|   2 |   TABLE ACCESS FULL| T_TEST1 |     1 |    86 |     2   (0)| 00:00:01 |
------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(ROWNUM=1)


Statistics
----------------------------------------------------------
          1  recursive calls
          0  db block gets
          4  consistent gets
          1  physical reads
          0  redo size
       1201  bytes sent via SQL*Net to client
        385  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

SQL> select * from t_test1
  2  where rownum <= 1;


Execution Plan
----------------------------------------------------------
Plan hash value: 536364188

------------------------------------------------------------------------------
| Id  | Operation          | Name    | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |         |     1 |    86 |     2   (0)| 00:00:01 |
|*  1 |  COUNT STOPKEY     |         |       |       |            |          |
|   2 |   TABLE ACCESS FULL| T_TEST1 |     1 |    86 |     2   (0)| 00:00:01 |
------------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   1 - filter(ROWNUM<=1)


Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          4  consistent gets
          0  physical reads
          0  redo size
       1201  bytes sent via SQL*Net to client
        385  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed


3         ROWNUM的使用“陷阱”
由于ROWNUM是一个伪列，只有有结果记录时，ROWNUM才有相应数据，因此对它的使用不能向普通列那样使用，否则就会陷入一些“陷阱”当中。

3.1     对ROWNUM进行>、>=、=操作
不能对ROWNUM使用>（大于1的数值）、>=（大于或等于1的数值）、=（大于或等于1的数值），否则无结果



SQL> select count(*) from css_bl_view a where rownum>0;

  COUNT(*)
----------
361928


SQL> select count(*) from css_bl_view a
  2  where rownum > 1;

  COUNT(*)
----------
         0


这是因为：

1、ROWNUM是伪列，必须要要有返回结果后，每条返回记录就会对应产生一个ROWNUM数值；

2、返回结果记录的ROWNUM是从1开始排序的，因此第一条始终是1;



这样，当查询到第一条记录时，该记录的ROWNUM为1，但条件要求ROWNUM>1，因此不符合，继续查询下一条；因为前面没有符合要求的记录，因此下一条记录过来后，其ROWNUM还是为1，如此循环，就不会产生结果。上述查询可以通过子查询来替代：



SQL> select count(*)
  2  from
  3  (select BL_REF_CDE, rownum rn from css_bl_view)
  4  where rn > 1;

  COUNT(*)
----------
    361927


我们可以通过以下方式来实现对ROWNUM的>、=的查询：

查询ROWNUM=5的数据：



SQL> select object_id,object_name
  2  from (select object_id,object_name, rownum as rn from t_test1)
  3  where rn = 5;

OBJECT_ID OBJECT_NAME
---------- ------------------------------
        29 C_COBJ#


查询ROWNUM > 25的数据：



SQL> select * from t_test4
  2  minus
  3  select * from t_test4
  4  where rownum <= 25;

USERNAME                          USER_ID CREATED
------------------------------ ---------- ---------
DIP                                    19 21-NOV-05
OUTLN                                  11 21-NOV-05
PUBLIC                              99999 18-JUL-07
SYS                                     0 21-NOV-05
SYSMAN                                 32 21-NOV-05
SYSTEM                                  5 21-NOV-05

6 rows selected.
3.2     ROWNUM和Order BY
要注意的是：在使用ROWNUM时，只有当Order By的字段是主键时，查询结果才会先排序再计算ROWNUM，下面OBJECT_ID是表T_TEST1的主键字段：



SQL> select object_id,object_name from t_test1
  2  where rownum <= 5
  3  order by object_id;

OBJECT_ID OBJECT_NAME
---------- ------------------------------
         2 C_OBJ#
         3 I_OBJ#
         4 TAB$
         5 CLU$
         6 C_TS#


但是，对非主键字段OBJECT_NAME进行排序时，结果就混乱了：



SQL> select object_id,object_name from t_test1
  2  where rownum <= 5
  3  order by object_name;

OBJECT_ID OBJECT_NAME
---------- ------------------------------
        28 CON$
        29 C_COBJ#
        20 ICOL$
        44 I_USER1
        15 UNDO$

SQL> select count(*) from t_test1
  2  where object_name < 'CON$';

  COUNT(*)
----------
     21645


出现这种混乱的原因是：Oracle先按物理存储位置（rowid）顺序取出满足rownum条件的记录，即物理位置上的前5条数据，然后在对这些数据按照Order By的字段进行排序，而不是我们所期望的先排序、再取特定记录数。



如果需要对非主键字段排序再去前n条数据，我们可以以以下方式实现：



SQL> select object_id,object_name
  2  from (select object_id,object_name from t_test1
  3        order by object_name)
  4  where rownum <= 5;

OBJECT_ID OBJECT_NAME
---------- ------------------------------
     35489 /1000e8d1_LinkedHashMapValueIt
     35490 /1000e8d1_LinkedHashMapValueIt
     21801 /1005bd30_LnkdConstant
     21802 /1005bd30_LnkdConstant
     17205 /10076b23_OraCustomDatumClosur


3.3     排序分页
当对存在重复值的字段排序后再分页输出，我们很容易会陷入到另外一个“陷阱”。



请看以下例子，我们希望对T_TEST1的OWNER字段排序后，以每页输出10个结果的方式分页输出：



SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name from t_test1 order by owner) a
  4  where rownum <= 10)
  5  where rn >= 1;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWADAPTER
AFWOWNER                       AFWADAPTERCONFIGURATION
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX1
AFWOWNER                       AFWADAPTERFQN_PK
AFWOWNER                       AFWADAPTERCONFIGURATION_PK
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX2
AFWOWNER                       AFWSERVERCODE_PK
AFWOWNER                       AFWSERVER
AFWOWNER                       AFWADAPTERLOOKUP_IDX1
AFWOWNER                       AFWADAPTERLOOKUP

10 rows selected.

SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name from t_test1 order by owner) a
  4  where rownum <= 20)
  5  where rn >= 11;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWTOKENSTATUSCODE_PK
AFWOWNER                       AFWTOKENSTATUS
AFWOWNER                       AFWTOKENADMIN_IDX1
AFWOWNER                       AFWTOKENADMINCODE_PK
AFWOWNER                       AFWTOKENADMIN
AFWOWNER                       AFWTOKEN
AFWOWNER                       AFWSERVERCONFIGURATION_PK
AFWOWNER                       AFWSERVERCONFIGURATION
AFWOWNER                       AFWSERVER
AFWOWNER                       AFWADAPTERLOOKUP

10 rows selected.


仔细比较结果，你会发现“AFWSERVER”、“AFWADAPTERLOOKUP”在两次分页结果中都出现了。但是OBJECT_NAME在每个OWNER中的值是唯一的，说明这个输出结果是错误的，我们又陷入了一个“陷阱”。这是怎么回事呢，请先看下上述语句的查询计划：



SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name from t_test1 order by owner) a
  4  where rownum <= 20)
  5  where rn >= 11;

10 rows selected.


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=205 Card=20 Bytes=94
          0)

   1    0   VIEW (Cost=205 Card=20 Bytes=940)
   2    1     COUNT (STOPKEY)
   3    2       VIEW (Cost=205 Card=30670 Bytes=1042780)
   4    3         SORT (ORDER BY STOPKEY) (Cost=205 Card=30670 Bytes=858760)
   5    4           TABLE ACCESS (FULL) OF 'T_TEST1' (Cost=42 Card=30670 Bytes=858760)


看下这个“陷阱”是怎么形成的。从查询计划上，我们可以注意到，对于子查询

select a.*, rownum as rn from
     (select owner, object_name from t_test1 order by owner) a
where rownum <= 20
优化器采用了“SORT (ORDER BY STOPKEY)”。



“SORT (ORDER BY STOPKEY)”不需要对所有数据进行排序，而是只要找出结果集中的按特定顺序的最前N条记录，一旦找出了这N条记录，就无需再对剩下的数据进行排序，而直接返回结果。这种算法我们可以视为是“快速排序”算法的变种。快速排序算法的基本思想是：先将数据分2组集合，保证第一集合中的每个数据都大于第二个集合中每个数据，然后再按这个原则对每个集合进行递归分组，直到集合的单位最小。在进行“SORT (ORDER BY STOPKEY)”时，首先找出N条数据（这些数据并没有做排序）放在第一组，保证第一组的数据都大于第二组的数据，然后只对第一组数据进行递归。

可以看到，基于这样的算法基础上，如果N的数值不同，数据的分组也不同（如N=20时，第一次分组比例为12:8，然后继续递归；当N=10时，第一次分组比例为3:7 … …），这样，在数据的排序字段值都相等时，输出结果的顺序就会因为N值不同而不同。



知道原因后，我们可以通过以下几种方法来避免这个“陷阱”。

1、让查询计划避免“SORT (ORDER BY STOPKEY)”，采用“SORT (ORDER BY)”，使数据排序不受ROWNUM的影响。但这样会使所有数据都做排序：



SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name, rowid from t_test1 order by owner) a)
  4  where rn <= 10
  5  and rn >= 1;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWADAPTER
AFWOWNER                       AFWADAPTERCONFIGURATION
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX2
AFWOWNER                       AFWADAPTERCONFIGURATION_PK
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX1
AFWOWNER                       AFWADAPTERFQN_PK
AFWOWNER                       AFWADAPTERLOOKUP_IDX1
AFWOWNER                       AFWSERVERCODE_PK
AFWOWNER                       AFWSERVERCONFIGURATION_IDX1
AFWOWNER                       AFWTOKENTYPECODE_PK

10 rows selected.

SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name, rowid from t_test1 order by owner) a)
  4  where rn <= 20
  5  and rn >= 11;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWTOKENTYPE
AFWOWNER                       AFWTOKENSTATUSCODE_PK
AFWOWNER                       AFWTOKENSTATUS
AFWOWNER                       AFWTOKENADMIN_IDX1
AFWOWNER                       AFWTOKENADMINCODE_PK
AFWOWNER                       AFWTOKENADMIN
AFWOWNER                       AFWTOKEN
AFWOWNER                       AFWSERVERCONFIGURATION_PK
AFWOWNER                       AFWTOKEN_PK
AFWOWNER                       AFWTOKEN_IDX6

10 rows selected.

SQL> set autot trace
SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name, rowid from t_test1 order by owner) a)
  4  where rn <= 20
  5  and rn >= 11;

10 rows selected.


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=237 Card=30670 Bytes=1441490)
   1    0   VIEW (Cost=237 Card=30670 Bytes=1441490)
   2    1     COUNT
   3    2       VIEW (Cost=237 Card=30670 Bytes=1042780)
   4    3         SORT (ORDER BY) (Cost=237 Card=30670 Bytes=1073450)
   5    4           TABLE ACCESS (FULL) OF 'T_TEST1' (Cost=42 Card=30670 Bytes=1073450)


2、在排序时，加上一个或多个字段（如主键字段、ROWID），使排序结果具有唯一性：



SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name, rowid from t_test1 order by owner, object_id) a
  4  where rownum <= 10)
  5  where rn >= 1;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWADAPTER
AFWOWNER                       AFWADAPTERFQN_PK
AFWOWNER                       AFWADAPTERCONFIGURATION
AFWOWNER                       AFWADAPTERCONFIGURATION_PK
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX1
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX2
AFWOWNER                       AFWADAPTERLOOKUP
AFWOWNER                       AFWADAPTERLOOKUP_IDX1
AFWOWNER                       AFWSERVER
AFWOWNER                       AFWSERVERCODE_PK

10 rows selected.

SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name, rowid from t_test1 order by owner, object_id) a
  4  where rownum <= 20)
  5  where rn >= 11;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWSERVERCONFIGURATION
AFWOWNER                       AFWSERVERCONFIGURATION_PK
AFWOWNER                       AFWSERVERCONFIGURATION_IDX1
AFWOWNER                       AFWTOKEN
AFWOWNER                       AFWTOKEN_PK
AFWOWNER                       AFWTOKEN_IDX1
AFWOWNER                       AFWTOKEN_IDX2
AFWOWNER                       AFWTOKEN_IDX3
AFWOWNER                       AFWTOKEN_IDX4
AFWOWNER                       AFWTOKEN_IDX5

10 rows selected.

SQL> set autot trace
SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select owner, object_name, rowid from t_test1 order by owner, object_id) a
  4  where rownum <= 20)
  5  where rn >= 11;

10 rows selected.


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=253 Card=20 Bytes=940)
   1    0   VIEW (Cost=253 Card=20 Bytes=940)
   2    1     COUNT (STOPKEY)
   3    2       VIEW (Cost=253 Card=30670 Bytes=1042780)
   4    3         SORT (ORDER BY STOPKEY) (Cost=253 Card=30670 Bytes=1196130)
   5    4           TABLE ACCESS (FULL) OF 'T_TEST1' (Cost=42 Card=306
          70 Bytes=1196130)


3、对排序字段建立索引，并强制使用索引。这样就能利用索引已经建立好的排序结果：

SQL> create index t_test1_idx1 on t_test1(owner);

Index created.

SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select /*+index(t T_TEST1_IDX1)*/owner, object_name from t_test1 t order by owner) a
  4  where rownum <= 10)
  5  where rn >= 1
  6  ;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWADAPTER
AFWOWNER                       AFWADAPTERCONFIGURATION
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX1
AFWOWNER                       AFWADAPTERCONFIGURATION_IDX2
AFWOWNER                       AFWADAPTERCONFIGURATION_PK
AFWOWNER                       AFWADAPTERFQN_PK
AFWOWNER                       AFWADAPTERLOOKUP
AFWOWNER                       AFWADAPTERLOOKUP_IDX1
AFWOWNER                       AFWSERVER
AFWOWNER                       AFWSERVERCODE_PK

10 rows selected.

SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select /*+index(t T_TEST1_IDX1)*/owner, object_name from t_test1 t order by owner) a
  4  where rownum <= 20)
  5  where rn >= 11;

OWNER                          OBJECT_NAME
------------------------------ ------------------------------
AFWOWNER                       AFWSERVERCONFIGURATION
AFWOWNER                       AFWSERVERCONFIGURATION_IDX1
AFWOWNER                       AFWSERVERCONFIGURATION_PK
AFWOWNER                       AFWTOKEN
AFWOWNER                       AFWTOKENADMIN
AFWOWNER                       AFWTOKENADMINCODE_PK
AFWOWNER                       AFWTOKENADMIN_IDX1
AFWOWNER                       AFWTOKENSTATUS
AFWOWNER                       AFWTOKENSTATUSCODE_PK
AFWOWNER                       AFWTOKENTYPE

10 rows selected.

SQL> set autot trace
SQL> select owner, object_name from
  2  (select a.*, rownum as rn from
  3  (select /*+index(t T_TEST1_IDX1)*/owner, object_name from t_test1 t order by owner) a
  4  where rownum <= 20)
  5  where rn >= 11;

10 rows selected.


Execution Plan
----------------------------------------------------------
   0      SELECT STATEMENT Optimizer=CHOOSE (Cost=414 Card=20 Bytes=940)
   1    0   VIEW (Cost=414 Card=20 Bytes=940)
   2    1     COUNT (STOPKEY)
   3    2       VIEW (Cost=414 Card=30670 Bytes=1042780)
   4    3         TABLE ACCESS (BY INDEX ROWID) OF 'T_TEST1' (Cost=414Card=30670 Bytes=858760)
   5    4           INDEX (FULL SCAN) OF 'T_TEST1_IDX1' (NON-UNIQUE) (
          Cost=26 Card=30670)



以上就是ROWNUM的使用技巧及其注意事项，希望编程成员正确使用ROWNUM，也希望DBA遇到相关问题能迅速定位。