I've created a pivot table with data from multiple tables (using JOINS). How can I add another column to the table which adds up each column from each row?
Example:
Category | A | B | C |
ABC 1 1 1
A 1 0 0
B 0 1 0
C 0 0 1
Category | A | B | C | TOTAL
ABC 1 1 1 3
A 1 0 0 1
B 0 1 0 1
C 0 0 1 1
SCOTT#research 15-APR-15> select * from testing ;
CATEG A B C
----- ---------- ---------- ----------
ABC 1 1 1
A 1 0 0
B 0 1 0
C 0 0 1
SCOTT#research 15-APR-15> select category,a,b,c, sum(a+b+c) as "total" from testing group by category,a,b,c order by category;
CATEG A B C total
----- ---------- ---------- ---------- ----------
A 1 0 0 1
ABC 1 1 1 3
B 0 1 0 1
C 0 0 1 1
In case you want to add a column, then can add one use a procedure to update the values using this,
alter table testing add total int;
use this procedure to update the values
create or replace procedure add_Test
is
sqlis varchar2(10);
total1 int;
begin
for i in (select * from testing) loop
select sum(a+b+c) into total1 from testing where category=i.category;
update testing set total=total1 where category=i.category;
end loop;
commit;
end;
exec add_test;
SCOTT#research 15-APR-15> select * from testing;
CATEG A B C TOTAL
----- ---------- ---------- ---------- ----------
ABC 1 1 1 3
A 1 0 0 1
B 0 1 0 1
C 0 0 1 1
Related
I'd like to make "SORTKEY" like the below. It's not the same observations for each one.
Basically, each one is 3 obs but if flg=1 then "SORTKEY" includes that observation.
In this example, it means SORTKEY = 2 is 4 obs, SORTKEY ^=2 is 3 obs.
Is there the way to make the SORTKEY manually?. If you have a good idea, please give me some advice.
I want the following dataset, using the "test" dataset.
/*
SORTKEY NO FLG
1 1 0
1 2 0
1 3 0
2 4 0
2 5 0
2 6 0
2 7 1
3 8 0
3 9 0
3 10 0
*/
data test;
input no flg;
cards;
1 0
2 0
3 0
4 0
5 0
6 0
7 1
8 0
9 0
10 0
;
run;
Use a sequence counter to track the 3-rows-per-sortkey requirement.
Example:
data want;
set have;
retain sortkey 1;
seq+1;
if seq > 3 and flag ne 1 then do;
seq = 1;
sortkey+1;
end;
run;
I have the following table - with a lot of rows -
ID A_1 B_1 A_2 B_2 A_3 B_3
-- ---- --- --- ---- --- ---
1 0 0 0 0 0 0
2 1 0 0 0 0 0
I need to get the following output table -
the rows will be ID, A_1, B_1 and so on.
ID A B
--- -- --
1 0 0
1 0 0
1 0 0
2 1 0
2 0 0
2 0 0
I tried with union, unpivot - I get only one row for each ID instead three.
How can I do this?
You have to use Union all to include the duplicate:
Demo
Result:
SELECT ID, A_1 A, B_1 B FROM TABLE1
UNION ALL
SELECT ID, A_2, B_2 FROM TABLE1
UNION ALL
SELECT ID, A_3, B_3 FROM TABLE1 ORDER BY ID;
select *
from t
unpivot (
(A,B)
for z in (
(A_1, B_1),
(A_2 , B_2),
(A_3, B_3)
)
);
Full test case with the results:
with t (ID, A_1, B_1, A_2 , B_2 , A_3, B_3) as (
select 1, 0 , 0, 0, 0, 0, 0 from dual union all
select 2, 1 , 0, 0, 0, 0, 0 from dual
)
select *
from t
unpivot (
(A,B)
for z in (
(A_1, B_1),
(A_2 , B_2),
(A_3, B_3)
)
);
Results:
ID Z A B
---------- ------- ---------- ----------
1 A_1_B_1 0 0
1 A_2_B_2 0 0
1 A_3_B_3 0 0
2 A_1_B_1 1 0
2 A_2_B_2 0 0
2 A_3_B_3 0 0
I would like to get a null value when i SUM UP and divide multiple values in event any of the values that i am summing up has a null. in the example below i would like the return value to be be a null if any of the values i am summing up have a null or zero.
(((CAST (NVL(XYY.SCR,NULL)AS NUMBER) - 57.81114) / 24.79211) + ((CAST(NVL(WPM_SCR,NULL)AS NUMBER) - 40.7836082505127) / 17.5946375921401) + ((CAST (NVL(SLOT3,NULL) AS NUMBER) - 50.204190919674 ) / 25.5100093808846) ) / 3 BASE
A simplified example:
anything + null or / null will be null anyway, so you don't have to do anything about it
for + 0 or / 0, use CASE (see lines #7 and #11)
SQL> with test (a, b) as
2 (select 6, 3 from dual union all
3 select 5, 0 from dual union all
4 select 2, null from dual
5 )
6 select a, b,
7 case when b = 0 then null
8 else a/b
9 end result_div,
10 --
11 case when a = 0 or b = 0 then null
12 else a + b
13 end result_sum
14 from test;
A B RESULT_DIV RESULT_SUM
---------- ---------- ---------- ----------
6 3 2 9
5 0
2
SQL>
I'm using SAS Base/ SAS Enterprise Guide and I'm stuck.
I need to create the Association Rules matrix to calculate the support, lift, confidence and other association metrics.
My table is similar to: (one line per client with all the products associated to)
data test;
input item1 item2 item3 item4 ;
datalines ;
1 0 1 0
1 1 1 0
1 0 1 0
1 0 1 1;
As you know for the market basket analysis I want something like that:
item1 item2 item3 item4
item1 4 1 4 1
item2 1 1 1 0
item3 4 1 4 1
item4 1 0 4 4
Any idea?
Thank you for your help.
I think your example WANT is incorrect.
data test;
input item1-item4;
datalines;
1 0 1 0
1 1 1 0
1 0 1 0
1 0 1 1
;;;;
proc print;
run;
proc corr noprint sscp out=sscp(drop=int: where=(_type_ eq 'SSCP' and _NAME_ ne: 'Int'));
var item:;
run;
proc print;
run;
create or replace procedure prcdr_Clustering is
v_sampleCount number;
v_sampleFlag number;
v_matchPercent number;
v_SpendAmount Number(18, 2);
cursor cur_PDCSample is
SELECT *
FROM TBL_BIL
WHERE UDF_CHK = 'N';
rec_Pdcsample TBL_BIL%rowtype;
BEGIN
OPEN cur_PDCSample;
LOOP
FETCH cur_PDCSample
into rec_Pdcsample;
EXIT WHEN cur_PDCSample%NOTFOUND;
SELECT COUNT(*)
INTO v_sampleCount
FROM TBL_BIL
WHERE UDF_TOKENIZED = rec_Pdcsample.UDF_TOKENIZED;
IF v_sampleCount <> 0 THEN
UPDATE TBL_BIL
SET UDF_CHK = 'Y'
WHERE UDF_TOKENIZED = rec_Pdcsample.UDF_TOKENIZED;
IF v_sampleCount > 1 THEN
v_sampleFlag := 1;
ELSE
IF v_sampleCount = 1 THEN
v_sampleFlag := 2;
ELSE
v_sampleFlag := 0;
END IF;
END IF;
UPDATE TBL_BIL
SET UDF_SAMPLECOUNT = v_sampleCount, UDF_SAMPLEFLAG = v_sampleFlag
WHERE uniqueid = rec_Pdcsample.uniqueid;
UPDATE TBL_BIL
SET UDF_PID = rec_Pdcsample.uniqueid
WHERE UDF_TOKENIZED = rec_Pdcsample.UDF_TOKENIZED;
UPDATE TBL_BIL
SET UDF_PIDSPEND = v_SpendAmount
WHERE uniqueid = rec_Pdcsample.uniqueid;
UPDATE TBL_BIL
SET UDF_MATCHPERCENT = 1
WHERE uniqueid <> rec_Pdcsample.uniqueid
AND UDF_TOKENIZED = rec_Pdcsample.UDF_TOKENIZED;
END IF;
IF cur_PDCSample%ISOPEN THEN
CLOSE cur_PDCSample;
END IF;
OPEN cur_PDCSample;
END LOOP;
IF cur_PDCSample%ISOPEN THEN
CLOSE cur_PDCSample;
END IF;
end PrcdrClustering;
It takes me days to execute, my table has 225,846 rows of data.
The structure of my table is :-
UNIQUEID NUMBER Notnull primary key
VENDORNAME VARCHAR2(200)
SHORTTEXT VARCHAR2(500)
SPENDAMT NUMBER(18,2)
UDF_TOKENIZED VARCHAR2(999)
UDF_PID NUMBER(10)
UDF_SAMPLEFLAG NUMBER(4)
UDF_SAMPLECOUNT NUMBER(4)
UDF_MATCHPERCENT NUMBER(4)
UDF_TOKENCNT NUMBER(4)
UDF_PIDSPEND NUMBER(18,2)
UDF_CHK VARCHAR2(1)
Where to start? I've a number points to make.
You're doing bulk updates; this implies that bulk collect ... forall would be far more efficient.
You're doing multiple updates of the same table, which doubles the amount of DML.
As you've already selected from the table, re-entering it to do another count is pretty pointless, use an analytic function to get the result you need.
Indentation, indentation, indentation. Makes your code much easier to read.
You can use elsif to reduce the amount of statements to be evaluated ( very, very minor win )
If the uniqueid is unique you can use rowid to update the table.
You're updating udf_pidspend to null, whether this is intentional or not there's no need to do a separate update for it.
You can do a lot more in the cursor, but there's obviously no need to select everything, which'll decrease the amount of data you need to read from the disks.
You may need a couple of commits in there; though this means you can't rollback if it fails midway.
I hope tbl_bil is indexed on uniqueid
As GolzeTrol noted you're opening the cursor multiple times. There's no need for this.
As general rules:
If you're going to select / update or delete from a table do it once if possible and as few times as possible if not.
If you're doing bulk operations use bulk collect.
Never write select *
Use rowid where possible it avoids all index problems.
This will only work in 11G, I answered this question recently where I provided my own way of dealing with this implementation restriction in versions prior to 11G and linked to Ollie's, Tom Kyte's and Sathya's
I'm not entirely certain what you're trying to do here so please forgive me if the logic is a little off.
create or replace procedure prcdr_Clustering is
cursor c_pdcsample is
select rowid as rid
, count(*) over ( partition by udf_tokenized ) as samplecount
, udf_chk
, max(uniqueid) over ( partition by udf_tokenized ) as udf_pid
from tbl_bil
where udf_chk = 'N';
type t__pdcsample is table of c_pdcsample%rowtype index by binary_integer;
t_pdcsample t__pdcsample;
begin
open c_pdcsample;
loop
fetch c_pdcsample bulk collect into t_pdcsample limit 1000;
exit when t_pdcsample.count = 0;
if t_pdcsample.samplecount <> 0 then
t_pdcsample.udf_chk := 'y';
if t_pdcsample.samplecount > 1 then
t_pdcsample.samplecount := 1;
elsif t_pdcsample.samplecount = 1 then
t_pdcsample.samplecount := 2;
else
t_pdcsample.samplecount := 0;
end if;
end if;
forall i in t_pdcsample.first .. t_pdcsample.last
update tbl_bil
set udfsamplecount = t_pdcsample.samplecount
, udf_sampleflag = t_pdcsample.sampleflag
, udf_pidspend = null
, udf_pid = t_pdcsample.udf_pid
where rowid = t_pdcsample(i).rowid
;
for i in t_pdcsample.first .. t_pdcsample.last loop
update tbl_bil TBL_BIL
set udfmatchpercent = 1
where uniqueid <> t_pdcsample.uniqueid
and udf_tokenized = t_pdcsample.udf_tokenized;
end loop;
commit ;
end loop;
close c_pdcsample;
end PrcdrClustering;
/
Lastly calling all tables tbl_... is a little bit unnecessary.
Here is a variant using a single SQL statement. I'm not 100% certain that the logic is exactly the same, but for my test set, it is. Also the current procedure is non deterministic when you have more than one record with udf_chk = 'N' and the same udf_tokenized ...
This is the refactored procedure
SQL> create procedure prcdr_clustering_refactored
2 is
3 begin
4 merge into tbl_bil t
5 using ( select tb1.uniqueid
6 , count(*) over (partition by tb1.udf_tokenized) cnt
7 , max(decode(udf_chk,'N',uniqueid)) over (partition by tb1.udf_tokenized order by tb1.udf_chk) pid
8 from tbl_bil tb1
9 where udf_chk = 'N'
10 or exists
11 ( select 'dummy'
12 from tbl_bil tb2
13 where tb2.udf_tokenized = tb1.udf_tokenized
14 )
15 ) q
16 on ( t.uniqueid = q.uniqueid )
17 when matched then
18 update
19 set t.udf_samplecount = decode(t.udf_chk,'N',q.cnt,t.udf_samplecount)
20 , t.udf_sampleflag = decode(t.udf_chk,'N',decode(q.cnt,1,2,1),t.udf_sampleflag)
21 , t.udf_pid = q.pid
22 , t.udf_pidspend = decode(t.udf_chk,'N',null,t.udf_pidspend)
23 , t.udf_matchpercent = decode(t.udf_chk,'N',t.udf_matchpercent,1)
24 , t.udf_chk = 'Y'
25 ;
26 end;
27 /
Procedure created.
And here is a test:
SQL> select *
2 from tbl_bil
3 order by uniqueid
4 /
UNIQUEID VENDORNAME SHORTTEXT SPENDAMT UDF_TOKENI UDF_PID UDF_SAMPLEFLAG UDF_SAMPLECOUNT UDF_MATCHPERCENT UDF_TOKENCNT UDF_PIDSPEND U
-------- ---------- ---------- -------- ---------- ------- -------------- --------------- ---------------- ------------ ------------ -
1 a a 1 bl 0 0 0 0 0 0 N
2 a a 1 bla 0 0 0 0 0 0 N
3 a a 1 bla 0 0 0 0 0 0 Y
4 a a 1 bla 0 0 0 0 0 0 Y
5 a a 1 bla 0 0 0 0 0 0 Y
6 a a 1 blah 0 0 0 0 0 0 N
7 a a 1 blah 0 0 0 0 0 0 Y
8 a a 1 blah 0 0 0 0 0 0 Y
9 a a 1 blah 0 0 0 0 0 0 Y
10 a a 1 blah 0 0 0 0 0 0 Y
11 a a 1 blah 0 0 0 0 0 0 Y
11 rows selected.
SQL> exec prcdr_clustering
PL/SQL procedure successfully completed.
SQL> select *
2 from tbl_bil
3 order by uniqueid
4 /
UNIQUEID VENDORNAME SHORTTEXT SPENDAMT UDF_TOKENI UDF_PID UDF_SAMPLEFLAG UDF_SAMPLECOUNT UDF_MATCHPERCENT UDF_TOKENCNT UDF_PIDSPEND U
-------- ---------- ---------- -------- ---------- ------- -------------- --------------- ---------------- ------------ ------------ -
1 a a 1 bl 1 2 1 0 0 Y
2 a a 1 bla 2 1 4 0 0 Y
3 a a 1 bla 2 0 0 1 0 0 Y
4 a a 1 bla 2 0 0 1 0 0 Y
5 a a 1 bla 2 0 0 1 0 0 Y
6 a a 1 blah 6 1 6 0 0 Y
7 a a 1 blah 6 0 0 1 0 0 Y
8 a a 1 blah 6 0 0 1 0 0 Y
9 a a 1 blah 6 0 0 1 0 0 Y
10 a a 1 blah 6 0 0 1 0 0 Y
11 a a 1 blah 6 0 0 1 0 0 Y
11 rows selected.
SQL> rollback
2 /
Rollback complete.
SQL> exec prcdr_clustering_refactored
PL/SQL procedure successfully completed.
SQL> select *
2 from tbl_bil
3 order by uniqueid
4 /
UNIQUEID VENDORNAME SHORTTEXT SPENDAMT UDF_TOKENI UDF_PID UDF_SAMPLEFLAG UDF_SAMPLECOUNT UDF_MATCHPERCENT UDF_TOKENCNT UDF_PIDSPEND U
-------- ---------- ---------- -------- ---------- ------- -------------- --------------- ---------------- ------------ ------------ -
1 a a 1 bl 1 2 1 0 0 Y
2 a a 1 bla 2 1 4 0 0 Y
3 a a 1 bla 2 0 0 1 0 0 Y
4 a a 1 bla 2 0 0 1 0 0 Y
5 a a 1 bla 2 0 0 1 0 0 Y
6 a a 1 blah 6 1 6 0 0 Y
7 a a 1 blah 6 0 0 1 0 0 Y
8 a a 1 blah 6 0 0 1 0 0 Y
9 a a 1 blah 6 0 0 1 0 0 Y
10 a a 1 blah 6 0 0 1 0 0 Y
11 a a 1 blah 6 0 0 1 0 0 Y
11 rows selected.
Regards,
Rob.
I don't know why, but you open the cur_PDCSample, which select (I suspect) thousands of records. And then, in a loop, you close the cursor and reopen it, each time processing only the first record that is returned.
If you open the cursor once, process each record and then close it, your procedure will probably go a lot faster.
Actually, since you do not always update TBL_BIL.UDF_CHK to 'Y', it seems to me that your current procedure may run infinitely.