The earth heart of the code is the creation of the needed datastructure with

push @{$r{join (' 'x8,@F[0..3]) }}, [@F[4,5]];'
[download]

Using Data::Dump dd function as first thing in the END block you'll see the datastructure:

(
  "I        33        C        C",
  [[0.5, 2], [1, 2]],
  "I        21        B        A",
  [[1, 6], [1, 6]],
  "I        40        D        D",
  [[1, 2], [1, 5]],
  "I        56        A        E",
  [[1, 2]],
  "I        9        A        B",
  [[0.25, 6], ["0.30", 8]],
)
[download]

When all files are processed the END block comes in play: for each key of the %r hash we use map to process all arrays contained as values of the key: every first value is added to $x (these are coming from all $F[4] values! ) and every second value is added to $y (coming from all $F[5] values) Vars $x and $y are declared with my so they are resetted for every key of the %r hash processed.

Now that all is ready and while we are still processing the key of the %r hash we print the key, a tab, $x divided by how many values we used ( scalar @{$r{$k}} ie: the scalar value of the array contained in the $r{$k} ) or the average you asked for. Then the total value of $y and stop.

L*

Thank you very much for the explanation! That makes a lot more sense now.