How to insert multiple rows from a flow - oracle

Question

I have to insert multiple row into a table from a file structured like this:
BANAC2C100017701007_X75 _CA 4X2 CT MLCR DR SX EP 160 E4
where 4x2, MLCR, 160 E4 have to be inserted into the same column for the same code BANAC2C100017701007. As example, the table should be structured like this:
After to split the elements from the file, how can I put them into the table? Any suggestion?

Answer 1

It can be done with sqlldr. I have made some assumptions, but if the data is one row per line as you describe above, with the same number of elements a line, a properly constructed control file with multiple "into table" statements can write different parts of one row of data as multiple rows to the same table.
The control file:
LOAD DATA
infile "file.dat"
TRUNCATE
INTO TABLE data_table
(entirerow BOUNDFILLER char(4000)
,code expression "regexp_substr(:entirerow, '(.*?)(_)', 1, 1, NULL, 1)"
,desc expression "regexp_substr(:entirerow, '(.*?)( +)', 1, 3, NULL, 1)"
)
INTO TABLE data_table
(entirerow BOUNDFILLER position(1) char(4000)
,code expression "regexp_substr(:entirerow, '(.*?)(_)', 1, 1, NULL, 1)"
,desc expression "regexp_substr(:entirerow, '(.*?)( +)', 1, 5, NULL, 1)"
)
INTO TABLE data_table
(entirerow BOUNDFILLER position(1) char(4000)
,code expression "regexp_substr(:entirerow, '(.*?)(_)', 1, 1, NULL, 1)"
,desc expression "regexp_substr(:entirerow, '(.*?)( +)', 1, 9, NULL, 1) || ' ' ||
regexp_substr(:entirerow, '(.*?)( +|$)', 1, 10, NULL, 1)"
)
A couple of things to note:
Since there are no delimiters, and none are specified, the entire row will be read into the first field "entirerow". Since it is not a column in the table, and it is defined as BOUNDFILLER, it is "remembered" for use later.
The next field "code" is found in the control file. No data field exists to match it with, but sqlldr finds it matches a column in the table and sees it is an expression so it applies the expression with the intention of putting the result into the column. The expression uses REGEXP_SUBSTR against the remembered BOUNDFILLER to cut out the parts we need. For code, get the characters up to but not including the first underscore. For desc, get the 3rd set of characters that are followed by one or more spaces (but not the spaces).
For the second logical row, we need to re-position the logical pointer back to the beginning of the row read in so sqlldr can re-process. Otherwise the logical pointer is at the end of the data and nothing will be returned. This is done with the "position" parameter seen in the "entirerow" definition of the 2nd and 3rd "into table" statements. The last "into table" follows the previous paradigm of just getting the 9th and 10th fields and concatenating them together. I chose to do this rather than come up with another regex to do it as it keeps consistency with the other fields, plus if you want to change it in the future it will be easier to follow.
As you can see it works and is reusable:
SQL> select code, desc
from data_table;
CODE DESC
------------------------- -------------
BANAC2C100017701007 4X2
BANAC2C100017701007 MLCR
BANAC2C100017701007 160 E4
Possible caveat: each row is being scanned 3 times, and the regexp calls are expensive so depending on the amount of data you need to load this may not be a feasible solution for your situation.