Oracle 分析函式詳解

1、自動彙總函式rollup,cube,
2、rank 函式, rank,dense_rank,row_number
3、lag,lead函式
4、sum,avg,的移動增加,移動平均數
5、ratio_to_report報表處理函式
6、first,last取基數的分析函式

基礎資料

06:34:23 SQL> select * from t;

BILL_MONTH AREA_CODE NET_TYPE LOCAL_FARE
--------------- ---------- ---------- --------------
200405 5761 G 7393344.04
200405 5761 J 5667089.85
200405 5762 G 6315075.96
200405 5762 J 6328716.15
200405 5763 G 8861742.59
200405 5763 J 7788036.32
200405 5764 G 6028670.45
200405 5764 J 6459121.49
200405 5765 G 13156065.77
200405 5765 J 11901671.70
200406 5761 G 7614587.96
200406 5761 J 5704343.05
200406 5762 G 6556992.60
200406 5762 J 6238068.05
200406 5763 G 9130055.46
200406 5763 J 7990460.25
200406 5764 G 6387706.01
200406 5764 J 6907481.66
200406 5765 G 13562968.81
200406 5765 J 12495492.50
200407 5761 G 7987050.65
200407 5761 J 5723215.28
200407 5762 G 6833096.68
200407 5762 J 6391201.44
200407 5763 G 9410815.91
200407 5763 J 8076677.41
200407 5764 G 6456433.23
200407 5764 J 6987660.53
200407 5765 G 14000101.20
200407 5765 J 12301780.20
200408 5761 G 8085170.84
200408 5761 J 6050611.37
200408 5762 G 6854584.22
200408 5762 J 6521884.50
200408 5763 G 9468707.65
200408 5763 J 8460049.43
200408 5764 G 6587559.23

BILL_MONTH AREA_CODE NET_TYPE LOCAL_FARE
--------------- ---------- ---------- --------------
200408 5764 J 7342135.86
200408 5765 G 14450586.63
200408 5765 J 12680052.38
40 rows selected.
Elapsed: 00:00:00.00

1. 使用rollup函式的介紹下面是直接使用普通sql語句求出各地區的彙總資料的例子
06:41:36 SQL> set autot on
06:43:36 SQL> select area_code,sum(local_fare) local_fare
06:43:50 2 from t
06:43:51 3 group by area_code
06:43:57 4 union all
06:44:00 5 select '合計' area_code,sum(local_fare) local_fare
06:44:06 6 from t
06:44:08 7 /

AREA_CODE LOCAL_FARE
---------- --------------
5761 54225413.04
5762 52039619.60
5763 69186545.02
5764 53156768.46
5765 104548719.19合計 333157065.31
6 rows selected.
Elapsed: 00:00:00.03
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=7 Card=1310 Bytes=
24884)
1 0 UNION-ALL
2 1 SORT (GROUP BY) (Cost=5 Card=1309 Bytes=24871)
3 2 TABLE ACCESS (FULL) OF 'T' (Cost=2 Card=1309 Bytes=248
71)
4 1 SORT (AGGREGATE)
5 4 TABLE ACCESS (FULL) OF 'T' (Cost=2 Card=1309 Bytes=170
17)
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
6 consistent gets
0 physical reads
0 redo size
561 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
6 rows processed下面是使用分析函式rollup得出的彙總資料的例子
06:44:09 SQL> select nvl(area_code,'合計') area_code,sum(local_fare) local_fare
06:45:26 2 from t
06:45:30 3 group by rollup(nvl(area_code,'合計'))
06:45:50 4 /
AREA_CODE LOCAL_FARE
---------- --------------
5761 54225413.04
5762 52039619.60
5763 69186545.02
5764 53156768.46
5765 104548719.19
333157065.31
6 rows selected.
Elapsed: 00:00:00.00
Execution Plan
----------------------------------------------------------
0 SELECT STATEMENT Optimizer=ALL_ROWS (Cost=5 Card=1309 Bytes=
24871)
1 0 SORT (GROUP BY ROLLUP) (Cost=5 Card=1309 Bytes=24871)
2 1 TABLE ACCESS (FULL) OF 'T' (Cost=2 Card=1309 Bytes=24871
)
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
4 consistent gets
0 physical reads
0 redo size
557 bytes sent via SQL*Net to client
503 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
6 rows processed從上面的例子我們不難看出使用rollup函式,系統的sql語句更加簡單,耗用的資源更少,從6個consistent gets降到4個consistent gets,如果基表很大的話,結果就可想而知了.

1. 使用cube函式的介紹

為了介紹cube函式我們再來看看另外一個使用rollup的例子
06:53:00 SQL> select area_code,bill_month,sum(local_fare) local_fare
06:53:37 2 from t
06:53:38 3 group by rollup(area_code,bill_month)
06:53:49 4 /
AREA_CODE BILL_MONTH LOCAL_FARE
---------- --------------- --------------
5761 200405 13060433.89
5761 200406 13318931.01
5761 200407 13710265.93
5761 200408 14135782.21
5761 54225413.04
5762 200405 12643792.11
5762 200406 12795060.65
5762 200407 13224298.12
5762 200408 13376468.72
5762 52039619.60
5763 200405 16649778.91
5763 200406 17120515.71
5763 200407 17487493.32
5763 200408 17928757.08
5763 69186545.02
5764 200405 12487791.94
5764 200406 13295187.67
5764 200407 13444093.76
5764 200408 13929695.09
5764 53156768.46
5765 200405 25057737.47
5765 200406 26058461.31
5765 200407 26301881.40
5765 200408 27130639.01
5765 104548719.19
333157065.31
26 rows selected.
Elapsed: 00:00:00.00

系統只是根據rollup的第一個引數area_code對結果集的資料做了彙總處理,而沒有對bill_month做彙總分析處理,cube函式就是為了這個而設計的.下面,讓我們看看使用cube函式的結果

06:58:02 SQL> select area_code,bill_month,sum(local_fare) local_fare
06:58:30 2 from t
06:58:32 3 group by cube(area_code,bill_month)
06:58:42 4 order by area_code,bill_month nulls last
06:58:57 5 /
AREA_CODE BILL_MONTH LOCAL_FARE
---------- --------------- --------------
5761 200405 13060.43
5761 200406 13318.93
5761 200407 13710.27
5761 200408 14135.78
5761 54225.41
5762 200405 12643.79
5762 200406 12795.06
5762 200407 13224.30
5762 200408 13376.47
5762 52039.62
5763 200405 16649.78
5763 200406 17120.52
5763 200407 17487.49
5763 200408 17928.76
5763 69186.54
5764 200405 12487.79
5764 200406 13295.19
5764 200407 13444.09
5764 200408 13929.69
5764 53156.77
5765 200405 25057.74
5765 200406 26058.46
5765 200407 26301.88
5765 200408 27130.64
5765 104548.72 200405 79899.53
200406 82588.15
200407 84168.03
200408 86501.34
333157.05

30 rows selected.
Elapsed: 00:00:00.01

可以看到,在cube函式的輸出結果比使用rollup多出了幾行統計資料.這就是cube函式根據bill_month做的彙總統計結果

2、 rollup 和 cube函式的再深入

從上面的結果中我們很容易發現,每個統計資料所對應的行都會出現null,我們如何來區分到底是根據那個欄位做的彙總呢,這時候,oracle的grouping函式就粉墨登場了.如果當前的彙總記錄是利用該欄位得出的,grouping函式就會返回1,否則返回0
1 select decode(grouping(area_code),1,'all area',to_char(area_code)) area_code,
2 decode(grouping(bill_month),1,'all month',bill_month) bill_month,
3 sum(local_fare) local_fare
4 from t
5 group by cube(area_code,bill_month)
6* order by area_code,bill_month nulls last
07:07:29 SQL> /
AREA_CODE BILL_MONTH LOCAL_FARE
---------- --------------- --------------
5761 200405 13060.43
5761 200406 13318.93
5761 200407 13710.27
5761 200408 14135.78
5761 all month 54225.41
5762 200405 12643.79
5762 200406 12795.06
5762 200407 13224.30
5762 200408 13376.47
5762 all month 52039.62
5763 200405 16649.78
5763 200406 17120.52
5763 200407 17487.49

5763 200408 17928.76
5763 all month 69186.54
5764 200405 12487.79
5764 200406 13295.19
5764 200407 13444.09
5764 200408 13929.69
5764 all month 53156.77
5765 200405 25057.74
5765 200406 26058.46
5765 200407 26301.88
5765 200408 27130.64
5765 all month 104548.72all area 200405 79899.53
all area 200406 82588.15
all area 200407 84168.03
all area 200408 86501.34
all area all month 333157.05
30 rows selected.
Elapsed: 00:00:00.01

可以看到,所有的空值現在都根據grouping函式做出了很好的區分,這樣利用rollup,cube和grouping函式,我們做資料統計的時候就可以輕鬆很多了.

2. rank函式的介紹介紹完rollup和cube函式的使用,下面我們來看看rank系列函式的使用方法.問題2.我想查出這幾個月份中各個地區的總話費的排名.

為了將rank,dense_rank,row_number函式的差別顯示出來,我們對已有的基礎資料做一些修改,將5763的資料改成與5761的資料相同.
1 update t t1 set local_fare = (
2 select local_fare from t t2
3 where t1.bill_month = t2.bill_month
4 and t1.net_type = t2.net_type
5 and t2.area_code = '5761'
6* ) where area_code = '5763'
07:19:18 SQL> /

8 rows updated.

Elapsed: 00:00:00.01我們先使用rank函式來計算各個地區的話費排名.
07:34:19 SQL> select area_code,sum(local_fare) local_fare,
07:35:25 2 rank() over (order by sum(local_fare) desc) fare_rank
07:35:44 3 from t
07:35:45 4 group by area_codee
07:35:50 5
07:35:52 SQL> select area_code,sum(local_fare) local_fare,
07:36:02 2 rank() over (order by sum(local_fare) desc) fare_rank
07:36:20 3 from t
07:36:21 4 group by area_code
07:36:25 5 /

AREA_CODE LOCAL_FARE FARE_RANK
---------- -------------- ----------
5765 104548.72 1
5761 54225.41 2
5763 54225.41 2
5764 53156.77 4
5762 52039.62 5

Elapsed: 00:00:00.01

我們可以看到紅色標註的地方出現了,跳位,排名3沒有出現下面我們再看看dense_rank查詢的結果.

07:36:26 SQL> select area_code,sum(local_fare) local_fare,
07:39:16 2 dense_rank() over (order by sum(local_fare) desc ) fare_rank
07:39:39 3 from t
07:39:42 4 group by area_code
07:39:46 5 /

AREA_CODE LOCAL_FARE FARE_RANK
---------- -------------- ----------
5765 104548.72 1
5761 54225.41 2
5763 54225.41 2
5764 53156.77 3 這是這裡出現了第三名
5762 52039.62 4

Elapsed: 00:00:00.00

在這個例子中,出現了一個第三名,這就是rank和dense_rank的差別,
rank如果出現兩個相同的資料,那麼後面的資料就會直接跳過這個排名,而dense_rank則不會,差別更大的是,row_number哪怕是兩個資料完全相同,排名也會不一樣,這個特性在我們想找出對應沒個條件的唯一記錄的時候又很大用處

1 select area_code,sum(local_fare) local_fare,
2 row_number() over (order by sum(local_fare) desc ) fare_rank
3 from t
4* group by area_code
07:44:50 SQL> /

AREA_CODE LOCAL_FARE FARE_RANK
---------- -------------- ----------
5765 104548.72 1
5761 54225.41 2
5763 54225.41 3
5764 53156.77 4
5762 52039.62 5

在row_nubmer函式中,我們發現,哪怕sum(local_fare)完全相同,我們還是得到了不一樣排名,我們可以利用這個特性剔除資料庫中的重複記錄.這個帖子中的幾個例子是為了說明這三個函式的基本用法的. 下個帖子我們將詳細介紹他們的一些用法.

2. rank函式的介紹a. 取出資料庫中最後入網的n個使用者
select user_id,tele_num,user_name,user_status,create_date
from (
select user_id,tele_num,user_name,user_status,create_date,
rank() over (order by create_date desc) add_rank
from user_info
)
where add_rank <= :n;

b.根據object_name刪除資料庫中的重複記錄
create table t1 as select obj#,name from sys.obj$;再insert into t1 select * from t1 數次.
delete from t1 where rowid in (
select row_id from (
select rowid row_id,row_number() over (partition by obj# order by rowid ) rn
from t1 ) where rn <> 1
);

c. 取出各地區的話費收入在各個月份排名.
SQL> select bill_month,area_code,sum(local_fare) local_fare,
2 rank() over (partition by bill_month order by sum(local_fare) desc) area_rank
3 from t
4 group by bill_month,area_code
5 /

BILL_MONTH AREA_CODE LOCAL_FARE AREA_RANK
--------------- --------------- -------------- ----------
200405 5765 25057.74 1
200405 5761 13060.43 2
200405 5763 13060.43 2
200405 5762 12643.79 4
200405 5764 12487.79 5
200406 5765 26058.46 1
200406 5761 13318.93 2
200406 5763 13318.93 2
200406 5764 13295.19 4
200406 5762 12795.06 5
200407 5765 26301.88 1
200407 5761 13710.27 2
200407 5763 13710.27 2
200407 5764 13444.09 4
200407 5762 13224.30 5
200408 5765 27130.64 1
200408 5761 14135.78 2
200408 5763 14135.78 2
200408 5764 13929.69 4
200408 5762 13376.47 5

20 rows selected.