array reference madness

Anonymous Monk has asked for the wisdom of the Perl Monks concerning the following question:

I'm doing an exercise from "Intermediate Perl" -- anyway, in the below code, the YAML Dump comes out correct but the foreach (marked HERE) yields just a single array reference (ARRAY), as if tmp only had one element, although it is the exact same array.

#!/usr/bin/perl -w
use strict;

use YAML::XS;

my %all;

open(INPUT,"<coconet.dat") || die "Can't open coconet.dat";
while (<INPUT>) {
    if ($_ =~ /^#/) { next }
    my ($src, $dest, $bytes) = split;
    $all{$src}{$dest}+=$bytes;    
}
close(INPUT);

my @reflist = sort { ${$$b}[2] <=> ${$$a}[2] } map \&buildlist($_), ke
+ys %all;   

foreach my $ref (@reflist) { 
    print "${$$ref}[0] ${$$ref}[2]\n";
    my @tmp=@{$$ref}[3];
#HERE    foreach my $line (@tmp) { print "\t$line\n" }
    print Dump @tmp;
}

sub buildlist {
    my $site = $_;
    my @tmp;
    $tmp[0]=$site;
    $tmp[1]=\%{$all{$site}};
    foreach (keys %{$tmp[1]}) { $tmp[2]+=${$tmp[1]}{$_} }
    @{$tmp[3]} = sort { ${$tmp[1]}{$b} <=> ${$tmp[1]}{$a} } keys %{$tm
+p[1]};
    return \@tmp;   # 0=site name, 1=hash ref, 2=total bytes, 3=sorted
+ dest refs
}
[download]

Why would there be this difference and what's wrong with my foreach??

Comment on array reference madness Download Code

Replies are listed 'Best First'.
Re: array reference madness by almut (Canon) on Mar 14, 2009 at 21:35 UTC
Your array `@tmp` holds a single array reference, not a list of values — you probably meant `my @tmp = @{${$$ref}[3]};` [download] which can also be written as `my @tmp = @{$$ref->[3]};` [download] BTW, I don't know what the exercise is, but I can't help thinking you have one level of referencing too much... :)	[reply] [d/l] [select]
Re^2: array reference madness by Porculus (Hermit) on Mar 14, 2009 at 21:44 UTC
Indeed; the extra referencing is being added by the `\&buildlist($_)`, which is an unnecessary complication. There's no need to build a reference here at all, since buildlist() already explicitly returns one, and it's arguably confusing because `\&` is normally used to take a reference to a subroutine, not to a return value. Removing the `\&` there would enable all the `${$$...}`-type constructions to be simplified, so for example the assignment to `@tmp` would then become just `@{$ref->[3]}`. This is probably a Good Thing.	[reply] [d/l] [select]
Re^3: array reference madness by Anonymous Monk on Mar 14, 2009 at 22:42 UTC
"Extra-referencing?!!" I was thinking the \& and the \@ matched, wrt the "type" of the return value. So you've given me something to think about -- I'll try and apply this. Thanks much! ps. the exercise is 5-2. I like the book, it's well structured ;)	[reply]
Re^4: array reference madness by Anonymous Monk on Mar 15, 2009 at 07:03 UTC
Re^2: array reference madness by halfcountplus (Hermit) on Mar 15, 2009 at 02:08 UTC
For some reason I must have thought that \& was the correct way to return a reference...anyway it worked out neatly: #!/usr/bin/perl -w use strict; my %all; open(INPUT,"<coconet.dat") \|\| die "Can't open coconet.dat"; while (<INPUT>) { if ($_ =~ /^#/) { next } my ($src, $dest, $bytes) = split; # hash of hashes holds the from->to transmission TOTALS; we do not + keep all the data $all{$src}{$dest}+=$bytes; } close(INPUT); # simple use of map to create array of references to hashes: my @refli +st = map { \%{$all{$_}} } keys %all; # so...create an array of references to arrays (from map callback) of +scalars and REFS TO existing hashes and new anonymous arrays! my @reflist = sort { $$b[2] <=> $$a[2] } map buildlist($_), keys %all; + # sort by the total of TOTALS foreach my $ref (@reflist) { print "$$ref[0] $$ref[2]\n"; # the next two are the same: foreach my $line (@{$ref->[3]}) { print "\t$line ${$ref->[1]}{$li +ne}\n" } # foreach my $line (@{$$ref[3]}) { print "\t$line ${$$ref[1]}{$lin +e}\n" } } sub buildlist { my $site = $_; my @tmp; $tmp[0]=$site; $tmp[1]=\%{$all{$site}}; foreach (keys %{$tmp[1]}) { $tmp[2]+=${$tmp[1]}{$_} } # make element 3 a reference to an anonymous array of scalars (the + dest site names, ranked) @{$tmp[3]} = sort { ${$tmp[1]}{$b} <=> ${$tmp[1]}{$a} } keys %{$tm +p[1]}; return \@tmp; # 0=site name, 1=hash ref, 2=total bytes for all t +ransmissions from site, 3=ranked dest names ref } __DATA__ 2000 lines, eg: gilligan.crew.hut lovey.howell.hut 4721 thurston.howell.hut lovey.howell.hut 4046 professor.hut ginger.girl.hut 5768 gilligan.crew.hut laser3.copyroom.hut 9352 gilligan.crew.hut maryann.girl.hut 1180 fileserver.copyroom.hut thurston.howell.hut 2548 skipper.crew.hut gilligan.crew.hut 1259 [download] This is more complex, but no longer than the solution in the book -- but the solution in the book does not create a permanent data structure either (goal was to just print ranked totals).	[reply] [d/l]
Re^3: array reference madness by halfcountplus (Hermit) on Mar 15, 2009 at 02:34 UTC
`# the next THREE are the same: # foreach my $line (@{$ref->[3]}) { print "\t$line ${$ref->[1]}{$line} +\n" } # foreach my $line (@{$$ref[3]}) { print "\t$line ${$$ref[1]}{$line}\n +" } foreach my $line (@{$$ref[3]}) { print "\t$line $all{$$ref[0]}{$line}\ +n" } }` [download] Notes ;)	[reply] [d/l]