File::CacheDir - Perl module to aid in keeping track and cleaning up files, quickly and without a cron $Id:,v 1.22 2006/05/17 00:05:34 earl Exp $


CacheDir attempts to keep files around for some ttl, while quickly, automatically and cronlessly cleaning up files that are too old.


The possible named arguments (which can be named or sent in a hash ref, see below for an example) are,

base_dir - the base directory default is '/tmp/cache_dir'

cache_stats - whether or not to try and cache -d checks default is 1

carry_forward - whether or not to move forward the file when time periods get crossed. For example, if your ttl is 3600, and you move from the 278711 to the 278712 hour, if carry forward is set, it will refresh a cookie (if set_cookie is true) and move the file to the new location, and set $self->{carried_forward} = 1 default is 1

cleanup_suffix - in order to avoid having more than one process attempt cleanup, a touch file, that looks like this "$cleanup_dir$self->{cleanup_suffix}" is created and cleaned up cleanup_fork - fork on cleanup default is 1 cleanup_frequency - percentage of time to attempt cleanup cleanup_length - seconds to allow for cleanup, that is, how old a touch file can be before a new cleanup process will start

content_typed - whether or not you have printed a Content-type header default is 0

cookie_brick_over - brick over an old cookie default is 0 cookie_name - the name of your cookie default is 'cache_dir'

cookie_path - the path for your cookie default is '/'

filename - what you want the file to be named (not including the directory), like "storebuilder" . time . $$ I would suggest using a script specific word (like the name of the cgi), time and $$ (which is the pid number) in the filename, just so files are easy to track and the filenames are pretty unique default is time . $$

periods_to_keep - how many old periods you would like to keep

set_cookie - whether or not to set a cookie default is 0

ttl - how long you want the file to stick around can be given in seconds (3600) or like "1 hour" or "1 day" or even "1 week" default is '1 day'


Since CacheDir fits in so nicely with cookies, I use a few CGI methods to automatically set cookies, retrieve the cookies, and use the cookies when applicable. The cookie methods make it near trivial to handle session information. Taking the advice of Rob Brown <>, I use, though it increases load time and nearly doubles out of the box memory required.

The cookie that gets set is the full path of the file with your base_dir swapped out. This makes it nice for users to not know full path to your files. The filename that gets returned from a cache_dir call, however is the full path.


Most of the time, the defaults will suffice, but using your own object methods, you can override most everything CacheDir does. To show which methods are used, I walk through the code with a simple example.

my $cache_dir = File::CacheDir->new({ base_dir => '/tmp/example', ttl => '2 hours', filename => 'example.' . time . ".$$", });

An object gets created, with the hash passed getting blessed in.

my $filename = $cache_dir->cache_dir;

The ttl gets converted to seconds, here 7200. The

$ttl_dir = $base_dir . $ttl;

In our example, $ttl_dir = "/tmp/example/7200";

$self->ttl_mkpath - if the ttl directory does not exist, it gets made

Next, the number of ttl units since epoch, here it is something like 137738. This is

$self->{int_time} = int(time/$self->{ttl});

Now, the full directory can be formed

$self->{full_dir} = $ttl_dir . $self->{int_time};

If $self->{full_dir} exists, $self->{full_dir} . $self->{filename} gets returned. Otherwise, I look through the $ttl_dir, and for each directory that is too old (more than $self->{periods_to_keep}) I run

$self->cleanup - just deletes the old directory, but this is where a backup could take place, or whatever you like.

Finally, I

$self->sub_mkdir - makes the new directory, $self->{full_dir}

and return the $filename


  #!/usr/bin/perl -w

  use strict;
  use File::CacheDir qw(cache_dir);

  my $filename = cache_dir({
    base_dir => '/tmp',
    ttl      => '2 hours',
    filename => 'example.' . time . ".$$",

  `touch $filename`;


Thanks to Rob Brown for discussing general concepts, helping me think through things, offering suggestions and doing the most recent code review. The idea for carry_forward was pretty well all Rob. I didn't see a need, but Rob convinced me of one. Since Rob first introduced the idea to me, I have seen CacheDir break three different programmers' code. With carry_forward, no problems. Finally, Rob changed my non-CGI cookie stuff to use CGI, thus avoiding many a flame war. Rob also recently wrote a taint clean version of rmtree. He also wrote an original version of strong_fork, recently adopted here, and got my logic right on fork'ing and exit'ing.

Thanks to Paul Seamons for listening to my ideas, offerings suggestions, using CacheDir and giving feedback. Using the namespace File::CacheDir, the case of CacheDir and cache_dir are all from Paul. Paul helped me cut down strong_fork to what we actually need here. Finally, thanks to Paul for the idea of this THANKS section.

Thanks to Wes Cerny for using CacheDir, and giving feedback. Also, thanks to Wes for a last minute code review. Wes had me change the existence check on $self->{carry_forward_filename} to a plain file check based on his experience with CacheDir.

Thanks to Allen Bettilyon for discovering some problems with the cleanup scheme. Allen had the ideas of using touch files and cleanup_frequency to avoid concurrent clean ups. He also convinced me to use perhaps_cleanup to allow for backward compatibility with stuff that might be using cleanup.