Getting parts of a string (using regexp_instr, REVERSE, ...)

Question

Hi,
I'm trying to alter our company's existing procedure, so that it can extract the Nr of the house, and the bus ('mailbox') from the field "pio_address.street_nr_bus".
Now, if the input is something simple like "Testlane 15 A", there's no problem ofcourse, but since the data comes from what people entered in a form (on our site for example), it can get really messy and the input might be like:
[streetpart1] [streetpart2] [streetpart 3] [NR] [NRpart2] [mailbox] ==> ex: "Test 2 Lane 15 B A"
So in that example, I'd like to get "15 B A" (being a combination of house-nr and mailbox-nr)
So what I've come op with for now (would love to simplify it though) is the following code; but the procedure takes 30 minutes to complete now (~7000 records), which seems rather long, especially since it used to last like 5-10 minutes.
Any ideas/tips?
code:
SELECT REGEXP_REPLACE(trim(REVERSE(pio_address.street_nr_bus)),' *',' ') INTO pio_address.street_nr_bus
FROM DUAL;
SELECT REGEXP_INSTR(pio_address.street_nr_bus, '[[:digit:]]') INTO v_posnr FROM DUAL;
SELECT REGEXP_INSTR(pio_address.street_nr_bus, '[[:alpha:]]', v_posnr) INTO v_posstreet FROM DUAL;
IF v_posstreet = 0 THEN
SELECT REVERSE(pio_address.street_nr_bus) INTO v_nrbus FROM DUAL;
ELSE
SELECT trim(REVERSE(substr(pio_address.street_nr_bus,1, v_posstraat-1))) INTO v_nrbus FROM DUAL;
END IF;
*****UPDATE*****
it appears that this part of the code isn't what's slowing the procedure down, but if someone still knows a "shorter" way to accomplish what I'm doing with this code, tips are always welcome
Edited by: user11236545 on Mar 27, 2012 2:47 AM
Edited by: user11236545 on Mar 27, 2012 2:49 AM

Answer 1

thanks, normally that would have been a great solution!
...buuuut I've just got some more examples of what my input might be, and instead of just 'Test 2 Lane 15 B A'
it could also be something like 'Test 2 Lane 15B BUS 2'
so, since the last character can also be a number ...
this is what I've got atm:
SELECT REVERSE(REGEXP_REPLACE(UPPER(pio_adres),'POSTBUS|BUS',''))
INTO pio_adres FROM DUAL;
SELECT REGEXP_REPLACE(trim(pio_adres),' *',' ') INTO pio_adres
FROM DUAL;
SELECT regexp_instr(pio_adres, '[[:alpha:]]{3,50}')
INTO v_streetpos FROM dual;
Now, like before, when using the old source; running the procedure takes 5minutes;
and running the new source (with this piece of code added), it takes like 35 minutes...
I'm testing it on about 7000 records, so yeah, the procedure suddenly has to do 3 extra selects 7000 times, but 30 minutes extra is waaaay too long. It doesn't seem normal to me, does it?