in reply to Re^4: Speed up file write taking weeks
in thread Speed up file write taking weeks

Well I can't comment about improving the SQL since that's not my expertise, but starting from flat files and using perl, I suggest:

Sort both input files by key (either when exporting them from the DB or as a post-export step. Then get perl to read the first line from each of the files and extract its key. If the keys don't match, read another line from the file whose key sorts earlier. Once you have identical keys from both files, read in lines from both files while they still have the same key. At this point you will have extracted two lists of data values. From your original example, for the key 'D' they will be (x, y) and (m,n,o). De-duplicate each list individually (assuming that such duplicates are possible) using a pair of hashes. Then calculate the cartesian product of those two lists, and use a third hash to de-duplicate. Then output to a third file a series of lines containing the same key and the various data pairs.

Then go back to step reading new lines from each file.

Dave.