PL/SQL Speeding up long function - regex

Question

I am looking to speed up the following PL/SQL function. Right now it has run for over 2 hours with no sign of completing. We aborted that one and attempting it again with a EXIT WHEN of 20 and it still shows no signs of actually completing.
We are running these through SQLDeveloper 17.3, and each of the (4) tables has about 15k rows.
The goal is to grab all of the SSN's in our database and change the first character to an illegal char and the last 2 characters to a random A-Z combination. We then have to update that SSN in every table that uses it (4).
declare
v_random varchar2(2);
v_origin_ssn varchar2(100);
v_working_start varchar2(100);
v_working_middle varchar2(100);
v_new_ssn varchar2(100);
begin
for o in (
select distinct ssn --loop all rows in tbl_customer
from program_one.tbl_customer
)
loop
if regexp_like(o.ssn, '^[A-Za-z9].*[A-Z]$') then continue; --if this is already scrambled, skip
else
select dbms_random.string('U', 2) --create random 2 cap letters
into v_random
from dual;
v_origin_ssn := o.ssn; --set origin ssn with the existing ssn
if regexp_like(o.ssn, '^[A-Za-z]') --if first char is already alpha, leave it alone, otherwise 9
then v_working_start := substr(o.ssn, 1, 1);
else v_working_start := 9;
end if;
v_working_middle := substr(o.ssn, 2, 6); --set middle ssn with the unchanged numbers
v_new_ssn := v_working_start||v_working_middle||v_random; --create new sanitized ssn
update program_one.tbl_customer --update if exists in tbl_customer
set ssn = v_new_ssn
where ssn = v_origin_ssn;
commit;
update program_one.tbl_mhc_backup --update if exists ssn tbl_mhc_backup
set ssn = v_new_ssn
where ssn = v_origin_ssn;
commit;
update program_two.tbl_waiver --update if exists ssn tbl_waiver
set ssn = v_new_ssn
where ssn = v_origin_ssn;
commit;
update program_two.tbl_pers --update if exists in tbl_pers
set ssan = v_new_ssn
where ssan = v_origin_ssn;
commit;
end if;
--dbms_output.put_line(v_origin_ssn||' : '||v_new_ssn); --output test string to verify working correctly
end loop;
end;

Answer 1

I'd do it without a function in plain SQL:
Create a table with old and new ssn:
CREATE TABLE tmp_ssn AS
SELECT ssn, '9'||substr(ssn,2,6)||dbms_random.string('U',2) as new_ssn
FROM (SELECT distinct ssn FROM program_one.tbl_customer);
CREATE UNIQUE INDEX ui_tmp_ssn ON tmp_ssn(ssn, new_ssn);
EXEC DBMS_STATS.GATHER_TABLE_STATS(null,'tmp_ssn');
... and then update the tables one by one:
MERGE INTO program_one.tbl_customer z USING tmp_ssn q ON (z.ssn=q.ssn)
WHEN MATCHED THEN UPDATE z.ssn = q.new_ssn;
COMMIT;
MERGE INTO program_one.tbl_mhc_backup z USING tmp_ssn q ON (z.ssn=q.ssn)
WHEN MATCHED THEN UPDATE z.ssn = q.new_ssn;
COMMIT;
etc
If that is still to slow, I'd do
RENAME tbl_customer to tbl_customer_old;
CREATE TABLE tbl_customer as
SELECT s.new_ssn as ssn, t.col1, t.col2, ... , t.coln
FROM tbl_customer_old t JOIN tmp_ssn s USING(ssn);
DROP TABLE tbl_customer_old;