The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.
use LWP::UserAgent;

package LWP::UserAgent::ExponentialBackoff;
$VERSION = '0.04';
@ISA     = ("LWP::UserAgent");
my @FAILCODES = qw(408 500 502 503 504);

sub new {
	my ( $class, %cnf ) = @_;
	my $before_request = delete $cnf{before_request};
	my $after_request = delete $cnf{after_request};
	my $sum        = delete $cnf{sum};
	my $retryCount = delete $cnf{retryCount};
	$retryCount = 3 unless defined $retryCount;
	my $minBackoff = delete $cnf{minBackoff};
	$minBackoff = 3 unless defined $minBackoff;
	my $maxBackoff = delete $cnf{maxBackoff};
	$maxBackoff = 90 unless defined $maxBackoff;
	my $deltaBackoff = delete $cnf{deltaBackoff};
	$deltaBackoff = 3 unless defined $deltaBackoff;
	my $tolerance = delete $cnf{tolerance};
	$tolerance = .20 unless defined $tolerance;
	my $failCodes = delete $cnf{failCodes};
	$failCodes = { map { $_ => $_ } @FAILCODES } unless defined $failCodes;
	my $self = $class->SUPER::new(@_);
	$self = bless {
		%{$self},
		sum          => $sum,
		retryCount   => $retryCount,
		minBackoff   => $minBackoff,
		maxBackoff   => $maxBackoff,
		tolerance    => $tolerance,
		deltaBackoff => $deltaBackoff,
		deltaLow     => $deltaLow,
		deltaHigh    => $deltaHigh,
		failCodes    => $failCodes
	}, $class;
	$self->deltas($tolerance);
	$self->sum($sum) unless !defined $sum;
	$self->before_request($before_request) unless !defined $before_request;
	$self->after_request($after_request) unless !defined $after_request;
	return $self;
}

sub simple_request {
	my ( $self, @args ) = @_;
	my $failCodes         = $self->{failCodes};
	my $currentRetryCount = 0;
	my $before_c          = $self->before_request;
	my $after_c           = $self->after_request;
	my $retryInterval     = 0;
	
	do {
		$before_c and $before_c->( $self, \@args, $retryInterval);
		sleep $retryInterval;
		$response = $self->SUPER::simple_request(@args);
		$after_c and $after_c->( $self, \@args, $retryInterval, $response);
		$code = $response->code();
		$currentRetryCount++;
	}while ( ( $retryInterval = $self->retry($currentRetryCount-1) )
		&& ${$failCodes}{$code} );
		
	return $response;
}

sub retry {
	my ( $self, $currentRetryCount ) = @_;
	my $retryCount   = $self->{retryCount};
	my $minBackoff   = $self->{minBackoff};
	my $maxBackoff   = $self->{maxBackoff};
	my $deltaLow     = $self->{deltaLow};
	my $deltaHigh    = $self->{deltaHigh};
	my $deltaBackoff = $self->{deltaBackoff};

	if ( $currentRetryCount < $retryCount ) {

		#Calculate Exponential backoff with tolerance (deltaLow & deltaHigh)
		my $r = $deltaBackoff;
		if ( $deltaHigh - $deltaLow != 0 ) {
			$r = rand( $deltaHigh - $deltaLow ) + $deltaLow;
		}
		$increment = ( 2**$currentRetryCount - 1 ) * $r + $minBackoff;
		$retryInterval = $increment <= $maxBackoff ? $increment : $maxBackoff;
	}
	else {
		$retryInterval = 0;
	}
	return $retryInterval;
}

sub retryCount            { shift->_elem( 'retryCount',            @_ ); }
sub minBackoff            { shift->_elem( 'minBackoff',            @_ ); }
sub maxBackoff            { shift->_elem( 'maxBackoff',            @_ ); }
sub failCodes             { shift->_elem( 'failCodes',             @_ ) }
sub before_request { shift->_elem( 'before_request', @_ ) }
sub after_request  { shift->_elem( 'after_request',  @_ ) }

sub tolerance {
	my ( $self, $tolerance ) = @_;
	$self->{tolerance} = $tolerance;
	$self->deltas($tolerance);
}

sub deltas {
	my ( $self, $tolerance ) = @_;
	if ( $tolerance == 0 ) {
		$self->{deltaLow}  = 0;
		$self->{deltaHigh} = 0;
	}
	else {
		$self->{deltaLow}  = $self->{deltaBackoff} * ( 1 - $tolerance );
		$self->{deltaHigh} = $self->{deltaBackoff} * ( 1 + $tolerance );
	}
}

sub deltaBackoff { shift->_elem( 'deltaBackoff', @_ ); }

sub addFailCodes {
	my ( $self, $code ) = @_;
	$self->{failCodes}->{$code} = $code;
}

sub delFailCodes {
	my ( $self, $code ) = @_;
	delete $self->{failCodes}->{$code};
}

# Given sum and deltaBackoff, compute retryCount and maxBackoff such that total retryIntervals will always be equal to or "slightly" greater than sum.
sub sum {
	my ( $self, $sum ) = @_;
	$self->{retryCount} = log2( $sum / $self->{deltaBackoff} + 1 );

# maxBackoff should be at least as big as the largest retry interval which is never bigger than the sum, so just make it equal the sum
	$self->{maxBackoff} = $sum;
	$self->{sum} = $sum;
}

sub log2 {
	my $n = shift;
	return log($n) / log(2);
}
1;
__END__

=head1 NAME

LWP::UserAgent::ExponentialBackoff - LWP::UserAgent extension that retries errors with exponential backoff

=head1 SYNOPSIS

  my $before_request = sub {
    print "# /Trying ", $_[1][0]->uri, " in ", $_[3], " seconds\n";
    ++$before_count;
    $sum+=$_[3];
  };

  my @failCodes    = qw(500 503);
  my %failCodesMap = map { $_ => $_ } @failCodes;

  %options = (
  	  before_request => $before_request,
  	  after_request  => $after_request,
  	  tolerance      => .20,
	  retryCount     => 5,
	  minBackoff     => 3,
	  maxBackoff     => 120,
	  deltaBackoff   => 3,
	  failCodes      => \%failCodesMap
  );

  my $ua = LWP::UserAgent::ExponentialBackoff->new(%options);
  my $request   = HTTP::Request->new( 'GET', $uri );
  my $response  = $ua->request($request);

=head1 DESCRIPTION

LWP::UserAgent::ExponentialBackoff is a LWP::UserAgent extention.
It retries requests on error using an exponential backoff algorthim.

=head1 CONSTRUCTOR METHODS

The following constructor methods are available:

=over 4

=item $ua = LWP::UserAgent::ExponentialBackoff->new( %options )

This method constructs a new C<LWP::UserAgent::ExponentialBackoff> object and returns it.
Key/value pair arguments may be provided to set up the initial state.

   KEY                     DEFAULT
   -----------             --------------------
   sum                     undef
   retryCount              3
   minBackoff              3
   maxBackoff              90
   tolerance               .20
   deltaBackoff            3
   failCodes               { map { $_ => $_ } qw(408 500 502 503 504) }	
   

See L<LWP::UserAgent> for additional key/value pair arguments that may be provided.

=head1 METHODS

This module inherits all of L<LWP::UserAgent>'s methods,
and adds the following.

=over

=item $browser->before_request()

=item $browser->before_request( \&some_routine );

=item $browser->after_request()

=item $browser->after_request( \&some_routine );

These read (first two) or set (second two) callbacks that are
called before the actual HTTP/FTP/etc request is made.  By default,
these are set to undef, meaning nothing special is called.  If you
want to alter try requests, or inspect responses before any retrying
is considered, you can set up these callbacks.

The arguments passed to these routines are:

=over

=item 0: the current $browser object

=item 1: an arrayref of the arguments we pass to LWP::UserAgent::simple_request
(the first of which is the request object)

=item 2: the number of seconds we sleep on this retry.

=item (3): And, only for after_request, the response we
just got.

=back

=head1 IMPLEMENTATION

This class works by overriding LWP::UserAgent's C<simple_request> method 
with an exponential backoff algortihm.


=head1 SEE ALSO

L<LWP>, L<LWP::UserAgent>, L<LWP::UserAgent::Determined>

=head1 COPYRIGHT AND DISCLAIMER

Copyright 2011, Michael Marrotte C<marrotte@gmail.com>, all rights
reserved.  This program is free software; you can redistribute it
and/or modify it under the same terms as Perl itself.

This program is distributed in the hope that it will be useful,
but without any warranty; without even the implied warranty of
merchantability or fitness for a particular purpose.

=head1 AUTHOR

Michael Marrotte C<marrotte@gmail.com>

This library is free software; you can redistribute it and/or modify
it under the same terms as Perl itself.

=cut