PLSQL: How to create a cursor using a variable as table_name

Question

I'm trying the following in PLSQL:

• I have a table who contain a list of table names, something like: Table_1, Table_2, Table_3, etc...

• I rewrite that table on a daily basis, so the list of tables is dynamic, one day you might get 3 tables, tomorrow 7 tables and so on.

• Based on that list of tables, i want to export the content of those tables using the UTL_FILE.

So far, i tried to fetch the table names into a variable and then loop over the variable, but it doesn't work.

I have the following codes:

Declare
    var1 SYS_REFCURSOR;
    var2 varchar2(20);
    var3 varchar2(20);    
    ARCHIVE UTL_FILE.FILE_TYPE;
    Cursor cur2 IS SELECT TABLE_NAME FROM LIST_OF_TABLES; 

BEGIN   

ARCHIVO:=UTL_FILE.FOPEN('test_path','test.txt','W');

        for i in cur2 loop

        var2:= i.table_name;

        OPEN var1 for 'SELECT SKUID, CMRPRICE FROM '||VAR2;
        loop

       FOR C IN MICURSOR LOOP
  UTL_FILE.PUT_LINE(ARCHIVE,(''||C.SKUID||''||','||''||  C.CMRPRICE||''));
      END LOOP;
      UTL_FILE.FCLOSE(ARCHIVE);
        close var1;
        end loop;

END;

i expect to obtain the same amount of files as tables on the table list

Thanks in advance

Answer 1

i expect to obtain the same amount of files as tables on the table list

In which case you need to open a different file for each table.

Also, you need to be consistent with variable naming ( archivo | archive , cur2 | micursor ). Generally it is better to give variables meaningful names which reflect their usage. It is too easy to confuse var1 and var2 .

I think you want to have the files' contents wrapped in quotes. I have used double-quotes, which is the standard for CSV format.

This code defines a PL/SQL record type, tgt_rec , with a projection which matches the projection of the query string. We open a dynamic ref cursor then in a loop fetch records into that record variable until the cursor is exhausted.

declare
    file_handle utl_file.file_type;

    rc sys_refcursor;          
    cursor cur_tables is 
        select table_name from list_of_tables; 
    type tgt_rec is record ( skuid number, cmrprice number); 
    l_rec tgt_rec;

begin   

    << tables >>
    for i in cur_tables loop

        file_handle := utl_file.fopen('test_path', i.table_name||'.csv','w');

        open rc for 'select skuid, cmrprice from '||i.table_name;

        << table_rows >>
        loop

           fetch rc into l_rec;
           exit when rc%not found;

            utl_file.put_line(file_handle, '"'||l_rec.skuid||'","'||  l_rec.cmrprice||'"');

        end loop table_rows;        

        utl_file.fclose(file_handle);

        close rc;

    end loop tables;

end;
/

You may not like the cosmetic changes I have made to your program (such as .csv for file extensions) and obviously you are free to revert them. After all, you are the Josiah Spode of your executable code.

---

"Do you know if I can print the result of two scripts using the same code structure?"

It depends. You can convert this anonymous block into a procedure and pass the query string - 'select skuid, cmrprice from ' - as a parameter. That would allow you to vary the executed query. However, you still need to fetch the result set into something, and that something must match the structure of the projection: same number of columns and same datatype. So that limits how flexible you can be.

Fortunately PL/SQL is a proper programming language, with lots of capabilities (although not Java-style Reflection). So you have the option of coding some seriously modular suite of programs, with sub-routines for generic things such as file handling and subroutines for query-specific data wrangling.