views:

1257

answers:

2

I'm writing SQL (for Oracle) like:

INSERT INTO Schema1.tableA SELECT * FROM Schema2.tableA;

where Schema1.tableA and Schema2.tableA have the same columns. However, it seems like this is unsafe, since the order of the columns coming back in the SELECT is undefined. What I should be doing is:

INSERT INTO Schema1.tableA (col1, col2, ... colN) 
SELECT (col1, col2, ... colN) FROM Schema2.tableA;

I'm doing this for lots of tables using some scripts, so what I'd like to do is write something like:

INSERT INTO Schema1.tableA (foo(Schema1.tableA)) 
SELECT (foo(Schema1.tableA)) FROM Schema2.tableA;

Where foo is some nifty magic that extracts the column names from table one and packages them in the appropriate syntax. Thoughts?

A: 

You may need to construct the insert statements dynamically using USER_TAB_COLUMNS and execute them using EXECUTE IMMEDIATE.

EddieAwad
+3  A: 

This PL/SQL should do it:

declare
    l_cols long;
    l_sql  long;
begin
    for r in (select column_name from all_tab_columns
              where  table_name = 'TABLEA'
              and    owner = 'SCHEMA1'
             )
    loop
       l_cols := l_cols || ',' || r.column_name;
    end loop;

    -- Remove leading comma
    l_cols := substr(l_cols, 2);

    l_sql := 'insert into schema1.tableA (' || l_cols || ') select ' 
             || l_cols || ' from schema2.tableA';

    execute immediate l_sql;

end;
/
Tony Andrews