Hashes and hash references

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

I was reading in Effective Perl Programming about creating hash references such as this:

$student = {}; 
$student->{last} = 'Smith';
[download]

I believe he is saying that this is more efficient to be passed to a subroutine, and I agree with that.
What puzzles me though is that you can, and should, always pass a reference to a hash, or array, to a subroutine anyway:

&sub(\%hash, \@array);
[download]

I have also heard that it is faster to create and use a hash reference in general than using a "regular" hash such as:

my %student = {};
$student{'last'} = 'Smith';
[download]

This might be a silly question but is it more efficient to declare and uses hashes that way (reference) and if it is more efficient then why do we have the possibility to declare and use them the "regular" way?
Is it one of those TIMTOWTDI things?

Confused Monk

Comment on Hashes and hash references Select or Download Code

Replies are listed 'Best First'.
Re: Hashes and hash references by pg (Canon) on Sep 04, 2005 at 17:39 UTC
Depends on what we are talking: Assignment: (reference appears to be consistantly slower than regular) `use Benchmark qw (timethese); timethese(10, {"regular" => \&regular, "reference" => \&reference}); sub regular { my %hash; for my $i (1 .. 100000) { $hash{$i} = $i; } } sub reference { my $hash; for my $i (1 .. 100000) { $hash->{$i} = $i; } }` [download] `Benchmark: timing 10 iterations of reference, regular... reference: 8 wallclock secs ( 7.76 usr + 0.16 sys = 7.92 CPU) @ 1 +.26/s (n=1 0) regular: 8 wallclock secs ( 7.20 usr + 0.14 sys = 7.34 CPU) @ 1 +.36/s (n=1 0) Benchmark: timing 10 iterations of reference, regular... reference: 9 wallclock secs ( 7.67 usr + 0.17 sys = 7.84 CPU) @ 1 +.27/s (n=1 0) regular: 8 wallclock secs ( 7.23 usr + 0.08 sys = 7.31 CPU) @ 1 +.37/s (n=1 0)` [download] Assignment + Passing as ref: (reference appears to be faster than regular) `use Benchmark qw (timethese); timethese(10, {"regular" => \&regular, "reference" => \&reference}); sub regular { my %hash; for my $i (1 .. 100000) { $hash{$i} = $i; } helper(\%hash); } sub reference { my $hash; for my $i (1 .. 100000) { $hash->{$i} = $i; } helper($hash); } sub helper { my $hash = shift; }` [download] `Benchmark: timing 10 iterations of reference, regular... reference: 8 wallclock secs ( 7.72 usr + 0.14 sys = 7.86 CPU) @ 1 +.27/s (n=1 0) regular: 9 wallclock secs ( 7.91 usr + 0.08 sys = 7.98 CPU) @ 1 +.25/s (n=1 0) Benchmark: timing 10 iterations of reference, regular... reference: 8 wallclock secs ( 7.70 usr + 0.11 sys = 7.81 CPU) @ 1 +.28/s (n=1 0) regular: 9 wallclock secs ( 7.95 usr + 0.06 sys = 8.02 CPU) @ 1 +.25/s (n=1 0)` [download]	[reply] [d/l] [select]
Re^2: Hashes and hash references by GrandFather (Saint) on Sep 04, 2005 at 19:09 UTC
I like cmpthese results rather than timethese so I've reworked your benchmark a little. The parameter passing tests were doing a lot of non-parameter passing work so I made the variables global (using our variables) and used the constructed hashes from the first two tests as the test variables for the passing tests. `Rate reference regular reference 8.00/s -- -8% regular 8.73/s 9% -- Rate regularPass referencePass regularPass 7.19/s -- -33% referencePass 10.7/s 49% --` [download] Read more... The test code (982 Bytes) Perl is Huffman encoded by design.	[reply] [d/l] [select]
Re: Hashes and hash references by dave_the_m (Monsignor) on Sep 04, 2005 at 17:17 UTC
This might be a silly question but is it more efficient to declare and uses hashes that way (reference) No, a direct hash will be marginally faster. PS this is wrong: `my %student = {};` [download] it should be `my %student = ();` [download] or even just `my %student;` [download] as the hash is created empty already Dave.	[reply] [d/l] [select]
Re^2: Hashes and hash references by Anonymous Monk on Sep 04, 2005 at 21:11 UTC
oops!... You are right. See I got confused with the reference construct ;-) Confused Monk	[reply]