Regexp_substr

Question

My earlier post was answered by Frank Kulash, thanks Frank.
However i have few questions
SELECT  owner_id
,    owner_name        AS old_owner_name
,    REGEXP_SUBSTR ( owner_name
              , '^\(([^)]+)'
              , 1
              , 1
              , NULL
              , 1
              )        AS legacy_id
,    REGEXP_SUBSTR ( owner_name
              , '-([^(]+)'
              , 1
              , 1
              , NULL
              , 1
              )        AS DEPT_ID
,    REGEXP_SUBSTR ( owner_name
              , '\(([^)]+)\) * (\(OLD\))?$'
              , 1
              , 1
              , NULL
              , 1
              )        AS NEW_OWNER_NAME
FROM    stg_wms_setup_owner_unpiv
ORDER BY  owner_id
;The problem i'm encountering is with the 'OWNER_NAME'. The original field in the table has a record '(1104) - (History) FORT HARRISON' under 'OWNER_NAME', in the above query to get the NEW~_OWNER_NAME i have
REGEXP_SUBSTR ( owner_name
              , '\(([^)]+)\) * (\(OLD\))?$'
              , 1
              , 1
              , NULL
              , 1
              )        AS NEW_OWNER_NAMEBut the above sql takes care when there is '(OLD)' at the end of the record, but i want to get rid of '(History)' as well, please need help , i'm thinking if i change the back reference will that help?
Thanks

Answer 1

Hi,
897837 wrote:
... Frank just like using a negative value in SUBSTR can we do a similar thing in back references? How can i copy the resultant dataset as a table and paste , because once i posted the dataset the fields are again getting closer and difficult for the users to read.Is this one question or two completely unrealted questions?
just like using a negative value in SUBSTR can we do a similar thing in back references? Unfortunately, no. I don't believe there's any good way to tell REGEXP_SUBSTR to return the last matrching pattern, or the next-to-last, or the N-th from the end. Often you can work around that by specifying specify what comes after the part you really want, and anchoring the expression to the end of the string ($). Do you really need to find the N-th-to last expression in this case? You've never explained exactly how you get the new columns from the original owner_name.
How can i copy the resultant dataset as a table and paste , because once i posted the dataset the fields are again getting closer and difficult for the users to read.Use \ tags, just like you did around the CREATE TABLE statement, and again around the INSERT statements.  (By the way, it looks like an editor or something replaced the single-quotes with some kind of fancy leaning quotes.  I had to edit that out manually.)
The word "code" in the \ tags means that the text is to be displayed in the style of code; that is, with a fixed-width font and no whitespace compressed. It has nothing to do with the content. You can use \ tags to format query results, or poetry, or graphics, or any other kind of text.
The query you posted looks pretty close to what you need.  If 'History' is case-insensitive, then use REGEXP_REPLACE instead of REPLACE to get rid of it.  If you're using REGEXP_REPLACE, it's easy to remove any spaces that come before or after '(History)' at the same time.SELECT owner_id
,      REGEXP_SUBSTR ( owner_name
     , '^\(([^)]+)'
     , 1
     , 1
     , NULL
, 1
     ) AS legacy_id
, REGEXP_SUBSTR ( owner_name
     , '\(([^)]+)\) * (\(OLD\))?$'
     , 1
     , 1
     , NULL
, 1
     ) AS dept_id
,      owner_name     AS old_owner_name
, REGEXP_SUBSTR ( REGEXP_REPLACE ( owner_name
                    , ' *\(History\ *'
                    , NULL
                    , 1
                    , 1
                    , 'i'     -- Case-insensitive
     , '-([^(]+)'
     , 1
     , 1
     , NULL
, 1
     ) AS new_owner_name
FROM t
ORDER BY owner_id
This seems to be right except for new_owner_name.  The expression above says there must be a hyphen before new_owner_name, but in your desired results (if I understand them correctly) you way you want new_owner_name to include the only hyphen in one case:2424 PS065413 FW - HM (PS065413) (OLD) FW - HM
If you can explain the rules about how to find new_owner_name, I can help you find a regular expression to do it.  Apparantly, the rule is not "everything from the hyphen to the next left parenthesis, or the end of the string", but I don't know what the rule is.
Are there other places where this is not doing what you need?  Point out those places, and explain how you get the correct results in those places.
Sorry, I'm not at an Oracle 11 database today.  I tested the query above as well as I could using Oracle 10.2.