Re: Syntactic sugar for tail call optimizations

I experimented with Sub::Call::Recur:

#!/usr/bin/perl

use strict;
use warnings;
use Sub::Call::Recur qw(:all);
use Memoize;
memoize('sum');

sub sum { 
    my ( $n, $sum ) = @_;
    if ( $n == 0 ) {
        return $sum;
    }
    else {
        recur ( $n - 1, $sum + 1);
    }
}

foreach my $sum ( \&sum ) { 
  print $sum->(0, 0),    "\n",
        $sum->(0, 1),    "\n",
        $sum->(1, 0),    "\n", 
        $sum->(1, 1),    "\n",
        $sum->(2, 2),    "\n",
        $sum->(10, 1),   "\n",
        $sum->(1000, 1), "\n";
}
[download]

It uses the Clojure - special_forms 'recur' to replace 'redo'.

Comment on Re: Syntactic sugar for tail call optimizations Download Code

Replies are listed 'Best First'.
Re^2: Syntactic sugar for tail call optimizations by LanX (Saint) on May 27, 2011 at 15:24 UTC
Thank you for pointing me to Sub::Call::Recur, very interesting! especially: `It can be thought of as the redo operator, but for subroutines instead of loops.` :) Actually I once read about `recur` in a Clojure book and wondered about the advantages to `goto &sub` Clearly - and as you demonstrated - there are two: 1. One can use the "usual" call interface `f(PARAS)` instead of the cumbersome `@_=(PARAS);goto &f` 2. Memoize can't intercept and cache calls when goto-like constructions are used. OTOH Sub::Call::Recur relies on XS code... so I think I'd rather prefer to chose between the need of memoization for repeated runs or for speed and stack-sanity of one time runs. I think if I need both I could still realize my own memoization hash. Cheers Rolf	[reply] [d/l] [select]

Replies are listed 'Best First'.

Re^2: Syntactic sugar for tail call optimizations
by LanX (Saint) on May 27, 2011 at 15:24 UTC

Sub::Call::Recur

especially:

It can be thought of as the redo operator, but for subroutines instead of loops.

Actually I once read about recur in a Clojure book and wondered about the advantages to goto &sub

Clearly - and as you demonstrated - there are two:

1. One can use the "usual" call interface f(PARAS) instead of the cumbersome @_=(PARAS);goto &f

2. Memoize can't intercept and cache calls when goto-like constructions are used.

OTOH Sub::Call::Recur relies on XS code... so I think I'd rather prefer to chose between the need of memoization for repeated runs or for speed and stack-sanity of one time runs.

I think if I need both I could still realize my own memoization hash.

Cheers Rolf

[reply]
[d/l]
[select]