Using Regex_substr in Oracle to select string up to the last occurence of a space within \n characters length - sql

Question

We have an issue where a column in our Oracle database has a longer character length than a field in another system.
Therefore I am trying to use case statements along with substr in order to split strings that are more than 40 characters in length. My case statements so far do what I want them to do in the fact that it leaves the first 40 characters of a string in column_a and then puts the remainder of the string in column_b.
However, the problem that I have is that by just using substr, the strings are being split midway through words.
So I was wondering if anybody knew of a couple of regular expressions that I could use with regex_substr that will -
select a string UP TO the last space within 40 characters - for
column_a
select a string AFTER the last space within 40 characters - for
column_b
These are the case statements that I have so far with substr:
CASE WHEN Length(column_a) > 10 THEN SubStr(column_a, 0, 40) END AS column_a,
CASE WHEN Length(column_a) > 40 THEN SubStr(addressnum, 41) END AS column_b
I am not familiar with regular expressions at all and so any help would be very much appreciated!

Answer 1

I've solved with instr/substr:
select substr(column_a,1,instr(substr(column_a,1,40), ' ', -1 )) column1,
substr(column_a,instr(substr(column_a,1,40), ' ', -1 )+1, 40) column2
from table1

Answer 2

A very similar problem was posted today on OTN. https://community.oracle.com/message/13928697#13928697
I posted a general solution, which will cover the problem proposed here as well. It may come in handy if there are similar needs in the future.
For the problem posted here on SO, the row_lengths table will have only one row, with r_id = 1 and r_len = 40. For demo purposes I am showing below an input_strings different from what I used on OTN.
Setup:
create table input_strings (str_id number, txt varchar2(500));
insert into input_strings values (1,
'One Hundred Sixty-Nine Thousand Eight Hundred Seventy-Four Dollars And Nine Cents');
insert into input_strings values (2, null);
insert into input_strings values (3, 'Mathguy rules');
create table row_lengths (r_id number, r_len number);
insert into row_lengths values (1, 40);
commit;
select * from input_strings;
STR_ID TXT
------- ---------------------------------------------------------------------------------
1 One Hundred Sixty-Nine Thousand Eight Hundred Seventy-Four Dollars And Nine Cents
2
3 Mathguy rules
3 rows selected
select * from row_lengths;
R_ID R_LEN
------- ----------
1 40
1 row selected.
Query and output: (NOTE: I include token length to verify that the first token is no more than 40 characters. OP did not answer if the SECOND token can be more than 40 characters; if it can't, one can add rows to the row_lengths table, perhaps with r_len = 40 for every row.)
with
r ( r_id, r_len ) as (
select r_id , r_len from row_lengths union all
select max(r_id) + 1, 4000 from row_lengths union all
select max(r_id) + 2, null from row_lengths
),
b (str_id, str, r_id, token, prev_pos, new_pos) as (
select str_id, txt || ' ', -1, null, null, 0
from input_strings
union all
select b.str_id, b.str, b.r_id + 1,
substr(str, prev_pos + 1, new_pos - prev_pos - 1),
b.new_pos,
new_pos + instr(substr(b.str, b.new_pos + 1, r.r_len + 1) , ' ', -1)
from b join r
on b.r_id + 2 = r.r_id
)
select str_id, r_id, token, nvl(length(token), 0) as len
from b
where r_id > 0
order by str_id, r_id;
STR_ID R_ID TOKEN LEN
------- ------- ------------------------------------------------ -------
1 1 One Hundred Sixty-Nine Thousand Eight 37
1 2 Hundred Seventy-Four Dollars And Nine Cents 43
2 1 0
2 2 0
3 1 Mathguy rules 13
3 2 0
6 rows selected.