I need to cache some data to disk between PHP requests so I decided to compare the various methods available.

Using 100000 element array as source data

Serialize save: 2.958 ms (1.5M)
Serialize read: 5.447 ms

JSON save: 1.880 ms (574.96K)
JSON read: 6.876 ms

PHP save: 8.684 ms (1.7M)
PHP read: 26.863 ms

Memcache set: 5.651 ms
Memcache get: 2.465 ms

IGBinary save: 1.377 ms (720.08K)
IGBinary read: 2.245 ms

MsgPack save: 1.389 ms (359.92K)
MsgPack read: 2.930 ms

I was surprised to see IGBinary and MsgPack so much faster than native JSON. JSON is easy and super portable, but not the fastest.

I need a simple disk based object cache and Cache::File was overkill. I wrote my own dependency free (only core modules) version:

cache($key);                     # Get
cache($key, $val);               # Set expires is 1 hour (default)
cache($key, $val, time() + 900); # Set expires in 15 minutes
cache($key, undef)               # Delete

I purposely wrote it small so it can be copy/pasted in to other scripts simply. I wrote a more robust implementation with some basic tests as well. When an entry is fetched that is expired it will be removed from disk. Abandoned cache entries will persist on disk until cache_clean() is called.

sub cache {
    use JSON::PP; use Tie::File; use File::Path; use Digest::SHA qw(sha256_hex);

    my ($key, $val, $expire, $ret, @data) = @_;

    my $hash = sha256_hex($key || "");
    my $dir  = "/dev/shm/perl-cache/" . substr($hash, 0, 3);
    my $file = "$dir/$hash.json";
    mkpath($dir);

    tie @data, 'Tie::File', $file or die("Unable to write $file"); # to r/w file

    if (@_ > 1) { # Set
        $data[0] = encode_json({ expires => int($expire || 3600), data => $val, key => $key });
    } elsif ($key && -r $file) { # Get
        eval { $ret = decode_json($data[0]); };
        if ($ret->{expires} && $ret->{expires} > time()) {
            $ret = $ret->{data};
        } else {
            unlink($file);
            $ret = undef;
        }
    }

    return $ret;
}