I agree. My post was mainly intended to demonstrate the use of %n as you'd requested. It was not a recommendation for production-grade code.

I would probably just generate the potentially long string with sprintf("%.2f %d/%d/%d  %s", @$record); then use printf to truncate and print. You have a number of options here; I've shown a selection below — the first pair indicates a gotcha which you should avoid; the remaining three pairs are all potential candidates (depends on the final output you want).

$ perl -e '
    my $x = "12345";
    my $y = "1234567890";

    printf q{%-*2$s|}."\n", $x, 9;
    printf q{%-*2$s|}."\n", $y, 9;

    printf q{%-.*2$s|}."\n", $x, 9;
    printf q{%-.*2$s|}."\n", $y, 9;

    printf q{%-*2$.*2$s|}."\n", $x, 9;
    printf q{%-*2$.*2$s|}."\n", $y, 9;

    printf q{%-*2$.*3$s|}."\n", $x, 10, 9;
    printf q{%-*2$.*3$s|}."\n", $y, 10, 9;
'
12345    |
1234567890|
12345|
123456789|
12345    |
123456789|
12345     |
123456789 |
[download]

The (possibly odd-looking) 'q{...}."\n"' is needed to avoid a 'Global symbol "$s" requires explicit package name ...' error. You could escape the '$' signs (e.g. "%-*2\$.*3\$s|\n") but, in my opinion, that makes the already somewhat cryptic format even less readable.

I hadn't previously used '%n', so that was an interesting learning exercise for me. Thanks for the question.