Unless you are concerned that a particular e-mail address will appear on more than one line (and you would then put all the files for that address into one outgoing message), why not just process the file one line at a time and avoid storing it in memory at all?

Perhaps you can elaborate on what you are going to do with these data. For example, if you want to make sure you only send one message to each address, and that no file is attached more than once to any message, read the whole file into a hash of hashes, which will ensure unique mail addresses and unique file names. This is the 'most correct' solution to me, but it would also probably consume the most memory (depending on how much redundancy there is in the data). As you said, TIMTOWTDI!

--
I'd like to be able to assign to an luser