I like the way that kennethk did this quite a bit.

From reading your post, I'm not sure that the points about "my" and file handles is completely clear. You clearly see that a re-open on the same file handle closes the previous file.

In the first foreach loop() for reading, my $handle is already closed before he gets to the point of opening that handle for write! This is a good thing. In your code that is not true. To do the same thing, you would need an explicit close() after the reading loop.

I would have gone further and used a name like "my $infile" in the first loop and a different name, like "my $outfile" for the output file.

Whether or not to close a file when the program is going to end shortly (very short time - some milliseconds) is a minor quibble.

One main reason to do this is that file handles are limited resources and closing a file handle frees resources - there will even be a limit to the maximum number of files that the entire system can have open at one time - albeit large as that limit is. However short of an OS fault (crash), leaving a file handle open for a very short time causes no harm. If the program is going to run for a significant time after you are finished with a file, I would close it. Making these judgements is part of the art of programming.

The rule that I use that the level of the program that opens a file is responsible for closing it. Open it in a loop, close it in that loop. Open it in a sub, close in that sub.

Early in my career, I saw one program run for 5 days and then fail because an open for the output file failed! So handling these situations can definitely have impacts!

I am an absolute fan of the use of "my" to scope variables to the smallest area possible. kennethk's code just has %data and the last use of $my handle at the file level scope. You have lots, like $key is "re-used".

A big advantage of using "my" is that you can "cut-n-paste" code into new programs without having to worry about "spooky unforeseen action at a distance".