Parsing CSV to Oracle columns

Question

Hi Gurus,
Need a help from you. I have a table(say, INSERT_CSV) in which a column(say, CSVFILE) with CLOB datatype is present. That column consists of a file in CSV format stored in as a CLOB.
Another table(Say, RGDEMO) has three columns (Employee_name, employee_id, employer). These data are present in the CSV format. Need to parse the csv from the first table and store the same in the second. There aremore than 20K records in there.
(DB is 9i)
Your help is much appreciated
Thanks!
Edited by: 796057 on Sep 20, 2010 2:14 PM

Answer 1

796057 wrote:
Actually I'm compelled to use CSV files stored in a db. We can not use SQL*Loader and/or External Tables also. And from the ASP UI, CSV file is getting stored in the Database.You have two basic choices.
Spool the CLOB to an external file (using the DBMS_LOB and UTL_FILE packages) and then load it as an external table.
The other choice is to use a custom token parser (changes a string into tokens) and then insert these tokens as column values into the destination table. I posted an example of such a token parser, written as a PL/SQL function, in {message:id=2416747}.
In pseudo code terms, you need to use the following approach to parse and insert the data:
  // get the glob
  select my_clob column into csvClob from my_table where col1 = primaryKeyValue;
  // read a line from the clob (each line will be terminated by a LF or CRLF set of characters)
  loop
     line := GetNextLineFromClob( csvClob ); // custom function that find the next line in a clob
     exit when line is null;  // no more lines to process
     tokens := Tokenise( line );  // convert the line into tokens (string array)
     insert into new_tab( col1, col2 ) values( tokens(1), tokens(2) ); // insert tokens into table
  end loop;
..The only real complexities here are the Tokenise and GetNextLineFromClob functions. You need to apply standard programming design and logic to deal with these - both does string processing. And this is one of the most common programming problems around. So it is not that difficult to design and code - and fairly easy to test too (do not need to create complex test data as input is string data).