Dynamic sql with pivot/unpivot - oracle11g

Question

I want to flip all rows to columns. So if have following 5 columns
with 3 data rows
ACCT_ID NAME PHONE MOBILE ALTERNATIVE_NAME
01 JOE BROWN 0456-9992-6666 07767828432 ZOE BROWN
02 GILL SHARP 0456-9992-6666 07763928432 BILL SHARP
03 ZAC LOWE 0236-9992-5644 07663925672 LUKE LOWE
I want the result set to look as follows. So have 3 columns and 5 rows.
COL headings (COL1 - COL3) are only added to make clearer, i don't need
columns headings
COL1 COL2 COL3
01 02 03
JOE BROWN GILL SHARP ZAC LOWE
0456-9992-6666 0456-9992-6666 0236-9992-5644
07767828432 07763928432 07663925672
ZOE BROWN BILL SHARP LUKE LOWE
The following SQL will work if know the acct_id's. But i wont know the acct_id, so need to write some dynamic sql to select the acct_id from table REPORTER.TEMP_PSR_REGION and then do the pivot/unpivot.
CREATE TABLE REPORTER.TEMP_PSR_REGION
(
ACCT_ID VARCHAR(50) NOT NULL,
NAME VARCHAR(50) NOT NULL,
PHONE VARCHAR(50) NOT NULL,
MOBILE VARCHAR(50) NOT NULL,
ALTERNATIVE_CONTACT VARCHAR(50) NOT NULL
)
INSERT INTO TEMP_PSR_REGION
(ACCT_ID,
NAME,
PHONE,
MOBILE,
ALTERNATIVE_CONTACT)
VALUES ('01',
'JOE BROWN',
'0456-9992-6666',
'07767828432',
'ZOE BROWN')
INSERT INTO TEMP_PSR_REGION
(ACCT_ID,
NAME,
PHONE,
MOBILE,
ALTERNATIVE_CONTACT)
VALUES ('02',
'GILL SHARP',
'0456-9992-6666',
'07763928432',
'BILL SHARP')
INSERT INTO TEMP_PSR_REGION
(ACCT_ID,
NAME,
PHONE,
MOBILE,
ALTERNATIVE_CONTACT)
VALUES ('03',
'ZAC LOWE',
'0236-9992-5644',
'07663925672',
'LUKE LOWE')
--- PIVOT/UNPIVOT Example
select col1, col2, col3
from (
select t.*, t.acct_id as col_id
from TEMP_PSR_REGION t
)
unpivot
(
value FOR heading in (acct_id, name, phone, mobile, ALTERNATIVE_CONTACT)
)
pivot
(
max(value) for col_id in ('01' as col1, '02' as col2, '03' as col3)
)
order by case heading
when 'ACCT_ID' then 1
when 'NAME' then 2
when 'PHONE' then 3
when 'MOBILE' then 4
when 'ALTERNATIVE_CONTACT' then 5
end
What would the dynamic sql look like to do this. I have not written dynamic sql before?

Answer 1

You can get the pivot's IN clause with a query:
select listagg('''' || acct_id || ''' as col' || rownum, ', ') within group (order by acct_id)
from TEMP_PSR_REGION;
LISTAGG(''''||ACCT_ID||'''ASCOL'||ROWNUM,',')WITHINGROUP(ORDERBYACCT_ID)
------------------------------------------------------------------------
'01' as col1, '02' as col2, '03' as col3
You can then use that as part of a query to generate the dynamic SQL statement, and then open that statement as a ref cursor. In this example I'm using a SQL*Plus (or SQL Developer or SQLcl) client refcursor variable to simplify how to declare and sisplay it:
var rc refcursor
declare
l_sql varchar2(32767);
begin
select
q'[
select *
from (
select t.*, t.acct_id as col_id
from TEMP_PSR_REGION t
)
unpivot
(
value FOR heading in (acct_id, name, phone, mobile, ALTERNATIVE_CONTACT)
)
pivot
(
max(value) for col_id in (]'
|| listagg('''' || acct_id || ''' as col' || rownum, ', ') within group (order by acct_id)
|| q'[)
)
order by case heading
when 'ACCT_ID' then 1
when 'NAME' then 2
when 'PHONE' then 3
when 'MOBILE' then 4
when 'ALTERNATIVE_CONTACT' then 5
end]'
into l_sql
from TEMP_PSR_REGION;
dbms_output.put_line(l_sql); -- for debugging only
open :rc for l_sql;
end;
/
Most of that is your original query. Essentially it puts your query into a string variable, but changes
select col1, col2, col3
from (
...
to
q'[
select *
from (
...]'
though you could generate that select-list with another listagg() if you prefer; and more importantly:
...
pivot
(
max(value) for col_id in ('01' as col1, '02' as col2, '03' as col3)
)
...
to:
q'[...
pivot
(
max(value) for col_id in (]'
|| listagg('''' || acct_id || ''' as col' || rownum, ', ') within group (order by acct_id)
|| q'[)
)
...]'
When that block is executed the dbms_output debug shows the generated statement, which is the same as your original once listagg() has been evaluated:
select *
from (
select t.*, t.acct_id as col_id
from TEMP_PSR_REGION t
)
unpivot
(
value FOR heading in (acct_id, name, phone, mobile, ALTERNATIVE_CONTACT)
)
pivot
(
max(value) for col_id in ('01' as col1, '02' as col2, '03' as col3)
)
order by case heading
when 'ACCT_ID' then 1
when 'NAME' then 2
when 'PHONE' then 3
when 'MOBILE' then 4
when 'ALTERNATIVE_CONTACT' then 5
end
PL/SQL procedure successfully completed.
And you can display the opened ref cursor:
print rc
COL1 COL2 COL3
-------------------------------------------------- -------------------------------------------------- --------------------------------------------------
01 02 03
JOE BROWN GILL SHARP ZAC LOWE
0456-9992-6666 0456-9992-6666 0236-9992-5644
07767828432 07763928432 07663925672
ZOE BROWN BILL SHARP LUKE LOWE
Using rownum is slightly dodgy as the headings might be out of order (as there is no ordering), but it seems you don't care about those anyway. If it does turn out to matter you can use a subquery to assign a numeric index to each row in the base table and then use that value instead of rownum.
You could potentially put this in a function that returns a ref cursor, but it depends how you're going to call it and use the results. The var and print are just a demonstration and probably not what you would really use.
You might also want to consider letting a reporting layer, if you have one, deal with the pivoting.