for many years now many IDEs offer to run code-snippets thru perltidy in a launched sub-process.

But this comes with noticeable delay, because most of the time is lost for startup of perltidy, while Perl::Tidy offers a server mode.

A faster tidying would allow formatting on-the-fly on key-triggers, like when typing return or closing a sub.

Proof

The following code is formatting itself 1,10 and 100 times thru perltidy, and an average run takes less than 0,07 secs

Challenge

• Write an IDE solution which starts a constant Perltidy server-process in the background and sends code-snippets back and forth.

Possible technologies

• comint-mode in emacs
• Language server protocol plugin for various IDEs (reference implementation Visual Studio Code)
• ... whatever your favourit IDE

Extra Points

• implement a GUI to try out perltidy configs on the fly, be it in Tk or inside an IDE.

demo code

use v5.12;
use warnings;
use Perl::Tidy;
use Time::HiRes qw/time/;

seek DATA,0,0;
my $code = join "",<DATA>;
#say $code;

my $show;


my $rc = <<'__CFG__';
--indent-columns=4
--maximum-line-length=80
--variable-maximum-line-length
--whitespace-cycle=0
__CFG__


time_it();


sub time_it {
    for my $times (1,10,100) {
        my $start=time;
        run_it() for 1..$times;
        my $end =time;
        warn "$times took: ", ($end-$start);
    }

}


sub run_it {
    my $clean;
    my $stderr;

    my $error = Perl::Tidy::perltidy
      (
       source      => \$code,
       destination => \$clean,
       stderr      => \$stderr,
       perltidyrc  => \$rc,
      );

    return unless $show;
    say $code;
    say '--------';
    if ($error) {
        say 'ERROR';
        say $stderr;
    } else {
        say $clean;
    }

}

__DATA__
[download]

1 took: 0.0720160007476807 at c:/tmp/pm/my_tidy.pl line 29, <DATA> lin
+e 59.
10 took: 0.643707036972046 at c:/tmp/pm/my_tidy.pl line 29, <DATA> lin
+e 59.
100 took: 6.37499022483826 at c:/tmp/pm/my_tidy.pl line 29, <DATA> lin
+e 59.
[download]