Scroll to navigation

CHI::Memoize(3pm) User Contributed Perl Documentation CHI::Memoize(3pm)


CHI::Memoize - Make functions faster with memoization, via CHI


version 0.07


    use CHI::Memoize qw(:all);
    # Straight memoization in memory
    # Memoize to a file or to memcached
    memoize( 'func', driver => 'File', root_dir => '/path/to/cache' );
    memoize( 'func', driver => 'Memcached', servers => [""] );
    # Expire after one hour
    memoize('func', expires_in => '1h');
    # Memoize based on the second and third argument to func
    memoize('func', key => sub { $_[1], $_[2] });


"`Memoizing' a function makes it faster by trading space for time. It does this by caching the return values of the function in a table. If you call the function again with the same arguments, "memoize" jumps in and gives you the value out of the table, instead of letting the function compute the value all over again." -- quoted from the original Memoize

For a bit of history and motivation, see

"CHI::Memoize" provides the same facility as Memoize, but backed by CHI. This means, among other things, that you can

  • specify expiration times (expires_in) and conditions (expire_if)
  • memoize to different backends, e.g. File, Memcached, DBI, or to multilevel caches
  • handle arbitrarily complex function arguments (via CHI key serialization)


All of these are importable; only "memoize" is imported by default. "use Memoize qw(:all)" will import them all as well as the "NO_MEMOIZE" constant.

Creates a new function wrapped around $func that caches results based on passed arguments.

$func can be a function name (with or without a package prefix) or an anonymous function. In the former case, the name is rebound to the new function. In either case a code ref to the new wrapper function is returned.

    # Memoize a named function
    # Memoize an anonymous function
    $anon = memoize($anon);

By default, the cache key is formed from combining the full function name, the calling context ("L" or "S"), and all the function arguments with canonical JSON (sorted hash keys). e.g. these calls will be memoized together:

    memoized_function({a => 5, b => 6, c => { d => 7, e => 8 }});
    memoized_function({b => 6, c => { e => 8, d => 7 }, a => 5});

because the two hashes being passed are canonically the same. But these will be memoized separately because of context:

     my $scalar = memoized_function(5);
     my @list = memoized_function(5);

By default, the cache namespace is formed from the full function name or the stringified code reference. This allows you to introspect and clear the memoized results for a particular function.

"memoize" throws an error if $func is already memoized.

See OPTIONS below for what can go in the options hash.

Returns a CHI::Memoize::Info object if $func has been memoized, or undef if it has not been memoized.

    # The CHI cache where memoize results are stored
    my $cache = memoized($func)->cache;
    # Code references to the original function and to the new wrapped function
    my $orig = memoized($func)->orig;
    my $wrapped = memoized($func)->wrapped;
Removes the wrapper around $func, restoring it to its original unmemoized state. Also clears the memoize cache if possible (not supported by all drivers, particularly memcached). Throws an error if $func has not been memoized.



The following options can be passed to "memoize".

Specifies a code reference that takes arguments passed to the function and returns a cache key. The key may be returned as a list, list reference or hash reference; it will automatically be serialized to JSON in canonical mode (sorted hash keys).

For example, this uses the second and third argument to the function as a key:

    memoize('func', key => sub { @_[1..2] });

and this is useful for functions that accept a list of key/value pairs:

    # Ignore order of key/value pairs
    memoize('func', key => sub { %@_ });

Regardless of what key you specify, it will automatically be prefixed with the full function name and the calling context ("L" or "S").

If the coderef returns "CHI::Memoize::NO_MEMOIZE" (or "NO_MEMOIZE" if you import it), this call won't be memoized. This is useful if you have a cache of limited size or if you know certain arguments will yield nondeterministic results. e.g.

    memoize('func', key => sub { $is_worth_caching ? @_ : NO_MEMOIZE });
You can pass any of CHI's set options (e.g. expires_in, expires_variance) or get options (e.g. expire_if, busy_lock). e.g.

    # Expire after one hour
    memoize('func', expires_in => '1h');
    # Expire when a particular condition occurs
    memoize('func', expire_if => sub { ... });
Any remaining options will be passed to the CHI constructor to generate the cache:

    # Store in file instead of memory
    memoize( 'func', driver => 'File', root_dir => '/path/to/cache' );
    # Store in memcached instead of memory
    memoize('func', driver => 'Memcached', servers => [""]);

Unless specified, the namespace is generated from the full name of the function being memoized.

You can also specify an existing cache object:

    # Store in memcached instead of memory
    my $cache = CHI->new(driver => 'Memcached', servers => [""]);
    memoize('func', cache => $cache);


By default "CHI", and thus "CHI::Memoize", returns a deep clone of the stored value even when caching in memory. e.g. in this code

    # func returns a list reference
    my $ref1 = func();
    my $ref2 = func();

$ref1 and $ref2 will be references to two completely different lists which have the same contained values. More specifically, the value is serialized by Storable on "set" and deserialized (hence cloned) on "get".

The advantage here is that it is safe to modify a reference returned from a memoized function; your modifications won't affect the cached value.

    my $ref1 = func();
    push(@$ref1, 3, 4, 5);
    my $ref2 = func();
    # $ref2 does not have 3, 4, 5

The disadvantage is that it takes extra time to serialize and deserialize the value, and that some values like code references may be more difficult to store. And cloning may not be what you want at all, e.g. if you are returning objects.

Alternatively you can use CHI::Driver::RawMemory, which will store raw references the way "Memoize" does. Now, however, any modifications to the contents of a returned reference will affect the cached value.

    memoize('func', driver => 'RawMemory');
    my $ref1 = func();
    push(@$ref1, 3, 4, 5);
    my $ref2 = func();
    # $ref1 eq $ref2
    # $ref2 has 3, 4, 5


The caveats of Memoize apply here as well. To summarize:

  • Do not memoize a function whose behavior depends on program state other than its own arguments, unless you explicitly capture that state in your computed key.
  • Do not memoize a function with side effects, as the side effects won't happen on a cache hit.
  • Do not memoize a very simple function, as the costs of caching will outweigh the costs of the function itself.


Memoizing a function will affect its call stack and its prototype.


A number of modules address a subset of the problems addressed by this module, including:

  • Memoize::Expire - pluggable expiration of memoized values
  • Memoize::ExpireLRU - provides LRU expiration for Memoize
  • Memoize::Memcached - use a memcached cache to memoize functions


Questions and feedback are welcome, and should be directed to the perl-cache mailing list:

Bugs and feature requests will be tracked at RT:

The latest source code can be browsed and fetched at:
    git clone git://


CHI, Memoize


Jonathan Swartz <>


This software is copyright (c) 2011 by Jonathan Swartz.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.

2018-10-03 perl v5.26.2