Paul W Bennett


Net::IPAddress::Util - Version-agnostic representation of an IP address


Version 4.004


  use Net::IPAddress::Util qw( IP );

  my $ipv4  = IP('');
  my $ipv46 = IP('::ffff:0:');
  my $ipv6  = IP('fe80::1234:5678:90ab');

  print "$ipv4\n";  #
  print "$ipv46\n"; #
  print "$ipv6\n";  # fe80::1234:5678:90ab

  print $ipv4->normal_form()  . "\n"; # 0000000000000000ffff0000c0a80001
  print $ipv46->normal_form() . "\n"; # 0000000000000000ffff0000c0a80001
  print $ipv6->normal_form()  . "\n"; # fe8000000000000000001234567890ab

  for (my $ip = IP(''); $ip <= IP(''); $ip++) {
    # do something with $ip


The goal of the Net::IPAddress::Util modules is to make IP addresses easy to deal with, regardless of whether they're IPv4 or IPv6, and regardless of the source (and destination) of the data being manipulated. The module Net::IPAddress::Util is for working with individual addresses, Net::IPAddress::Util::Range is for working with individual ranges of addresses, and Net::IPAddress::Util::Collection is for working with collections of addresses and/or ranges.



Set to a true value to make errors confess(). Set to a false value to make errors cluck(). Defaults to false.


Set to a true value to make new() assume that bare 32-bit (or smaller) numbers are supposed to represent IPv4 addresses, and promote them accordingly (i.e. to do implicitly what n32_to_ipv4() does). Set to a false value to make new() treat all bare numbers as 128-bit numbers representing IPv6 addresses. Defaults to false.


Set to a true value to make new() accept its argument in the "broken" format used by module versions prior to v4.000, and automatically "repair" them to the format used by v4.x of this module. Defaults to false.


Set to a true value to make any "repairs" undertaken by new() issue a warning using the Carp module. Specifically, if set to the number 2 (or higher), then new() will cluck() whenever it performs a repair, and if set to any other true value, then new() will carp() whenever it performs a repair. If set to a false value, then repairs (if any) will occur silently. Defaults to 1.



Transform an IP address to and from an array of 128 bits, MSB-first.


Given two bit arrays (as provided by explode_ip), return the truncated bit array of the prefix bits those two arrays have in common.


Given two bit arrays (as provided by explode_ip), return a truncated bit array of ones of the same length as the shared common_prefix of the two arrays.


Take a truncated bit array, and right-pad it with zeroes to the appropriate length.


Returns a bitmask that can be ANDed against an IP to pull out only the IPv4-relevant bits, that is the N32 portion with the 0xffff appended to its front.


Returns a bitmask that can be ORed onto an N32 to make it a proper "IPv4 stored as IPv6" N128.


Given an array of objects, sorts them in ascending order, faster than Perl's built-in sort command.

For those who understand the math, a radix sort is O(N) instead of O(N log N) (the speed of Perl's builtin sort()), but it does discard duplicates, so ymmv. There are also (rare) corner cases in which radix_sort() can chew up so much RAM that it causes paging / swapping, which will slow down the process dramatically.



These functions are exportable to provide a functionally-identical API to that provided by Net::IPAddress. They will cause warnings to be issued if they are called, to help you in your transition to Net::IPAddress::Util, if indeed that's what you're doing -- and I can't readily imagine any other reason you'd want to export them from here (as opposed to from Net::IPAddress) unless that's indeed what you're doing.



Exports IP() and n32_to_ipv4(), both useful for creating objects based on arbitrary external data.


Exports the functions for low-level "bit-twiddling" of addresses. You very probably don't need these unless you're writing your own equivalent of the Net::IPAddress::Util::Range or Net::IPAddress::Util::Collection modules.


Exports radix_sort(). You only need this if you're dealing with large arrays of Net::IPAddress::Util objects, and runtime is of critical concern.


Exports the Compatibility API functions listed above.


Exports all exportable functions.



Create a new Net::IPAddress::Util object, based on a well-formed IPv4 or IPv6 address string (e.g. '' or 'fe80::1234:5678:90ab'), or based on what is known by this module as the "normal form", a 32-digit hex number (without the leading '0x').

There are a number of acceptable arguments to new(), though it does always take a single argument.

16-element ARRAYREF

Creates an IPv6 object from 16 unsigned octets in network (big-endian) order.

4-element ARRAYREF

Creates an IPv6 object from 4 unsigned 32-bit network-order integers, supplied in network order.

An existing Net::IPAddress::Util object (equivalently, call as an object method)

Creates a non-destructive clone of the object.

A well-formed IPv4 or IPv6 string (including SIIT "IPv4 in IPv6" notation)

Examples are, ::ffff:0:, 1:2::3:4. Note that for IPv6 flavor strings, the scope ID (if any) is silently discarded. Note also that this behavior is subject to change. If you feel strongly, go to CPAN RT and file a ticket.

An unsigned 32-bit integer Perl value

Iff the $PROMOTE_N32 package variable is set, creates an IPv4 object.

    Actually, since all objects of this class are underlyingly IPv6, creates an "IPv4 in IPv6" representation of the IPv4 address. This is a very minor technical point, but I don't want the reader going away with incorrect assumptions about the way this module works.

An unsigned 128-bit integer Perl value (or a string holding a decimal representation of one, or a Math::BigInt object containing one)

Creates an IPv6 object, treating the number as network-order.

A 32-character hex string (case insensitive)

Creates an IPv6 object. NB this may be especially useful when you're using the output of the normal_form method (e.g. for round-tripping to a database).

A non-encoded sequence of 16 bytes in Perl string form

Creates an IPv6 object. Be especially sure that use of this argument form is performed correctly. You MUST, for instance, utf8::downgrade and decode your string before providing it. No effort is made to check or ensure anything about Unicode flagging or semantics. This is probably a bug, and is likely to be fixed in some future version (unless you can find a case for it being a security bug, in which case go directly to CPAN RT, please, and I'll fix it ASAP).


The exportable function IP() is a shortcut for Net::IPAddress::Util->new().

  my $xyzzy = Net::IPAddress::Util->new($foo);
  my $plugh = IP($foo); # Exactly the same thing, but with less typing


The exportable function n32_to_ipv4() converts an IPv4 address in "N32" format (i.e. a network-order 32-bit number) into an Net::IPAddress::Util object representing the same IPv4 address.


This module overloads a number of operators (cmp, <=>, &, |, ~, +, -, <<, >>) in hopefully obvious ways. One perhaps non-obvious overload is that cmp performs apparently "numeric" order comparison (the same as <=>) instead of strict string comparison. To understand why, picture it as comparing the normal_form of the addresses stringwise (rather than the as_str form).



Returns true if this object represents an IPv4 address.


Returns the dotted-quad representation of this object, or an error if it is not an IPv4 address, for instance ''.


Returns the "N32" representation of this object (that is, a 32-bit number in network order) if this object represents an IPv4 address, or an error if it does not.


Returns the "N128" representation of this object (that is, a 128-bit number in network order).

You may supply one optional argument. If this argument is true, the return value will be a Math::BigInt object (allowing quickish and easy math involving two such return values), otherwise (if it is false (the default)), then the N128 number will be returned as a bare string. If your platform can handle math with unsigned 128-bit integers, or if you will not be doing math on the results, then I strongly recommend the latter (default / false) option for performance reasons. In the true-argument case, you're advised to stringify the Math::BigInt math results as soon as is practical for performance reasons -- Math::BigInt is not "CPU free".


Returns the canonical IPv6 string representation of this object, for instance 'fe80::1234:5678:90ab' or '::ffff:0:'.


Returns the IPv6 string representation of this object, without compressing extraneous zeroes, for instance 'fe80:0000:0000:0000:0000:1234:5678:90ab'.


Returns the value of this object as a zero-padded 32-digit hex string, without the leading '0x', suitable (for instance) for storage in a database, or for other purposes where easy, fast sorting is desirable, for instance 'fe8000000000000000001234567890ab'.





If this object is an IPv4 address, it stringifies to the result of ipv4, else it stringifies to the result of ipv6.



Either confess()es or cluck()s the passed string based on the value of $Net::IPAddress::Util::DIE_ON_ERROR, and if possible returns undef.


What is the correct thing to do when new is given a flat 16-character string with its Unicode flag set?


May be redistributed and/or modified under terms of the Artistic License v2.0.


PWBENNETT -- paul(dot)w(dot)bennett(at)