Brainfunk. With Regex.

[oiskuu]

Greetings

my teacher assigned us a brainfuck interpreter^[1]. Here's what I've arrived at.

#! /usr/bin/perl

($dpos, $data, $/) = (0, "");
sub data :lvalue { vec($data, $dpos, 8) }
<> =~ m{
  (?: [>] (?{++$dpos}) | [+] (?{++(data)}) | [.] (?{print chr data}) |
      [<] (?{--$dpos}) | [-] (?{--(data)}) | [,] (?{data= ord getc}) |
      [[] (?(?{data})(?0)|(?1)) []] | [^][<>.,+-]
  )*+\z | (?(?{data})(?0)|(?1)) | \z((?: [^][] | [[] (?1) []])*) }x;

BEGIN { *ARGV = *DATA unless @ARGV }
__DATA__
+++++[>+++++++++<-],[[>--.++>+<<-]>+.->[<.>-]<<,]

For timing and testing I have used 392quine.b. This, and some other bf programs found here.

I'd say this regex based interpreter is fairly legible and uncomplicated (*cough*), though couple of snags remain.
First and foremost, the unfortunate "recursion" limit is encountered when long (but non-recursing) code is run:

  $ perl -w bf.pl <(perl bf.pl < bf.pl)
  Complex regular subexpression recursion limit (32766) exceeded at bf
+.pl line 5, <> chunk 1.
  ...

It should output whatever it was fed (bf.pl itself), but goes apeshit midway through. The workaround is to use (?:(?:__)*)*; however, I find this just too disgraceful. In this thread, ikegami hints at possible upcoming fix. Well, that was four years ago, and it sure isn't fixed in v5.18.1, let alone v5.12.3. I wonder if this is solved in more recent perls?

Another annoyance is with the data sub. It can be written in several ways

  sub data :lvalue { vec($data,$dpos,8) } # v5.16.1--OK; v5.12.3--Fail
  sub data :lvalue { vec($data,$dpos,8) //= 0 } # OK
  my @data;
  sub data :lvalue { $data[$dpos] &= 255 } # OK

The first one appears more expressive, though the array variant is no slower.

Finally, there may well be some important optimizations I've missed or somesuch. Suggestions are welcome, of course.