Release: 0.7.3

Date: 14th Sep 2009

Changes since 0.7.3

Added add, radd, sub, rsub operators to the IPAddress class.

Added support for validation and iteration of simple nmap style IPv4 ranges (raised in  Issue 46 ).

Removed some unused constants from fallback socket module.

Specific bug fixes addressed in this release

• FIXED  Issue 44  - int/long type error
• FIXED  Issue 46  - Question about IPv4 ranges
• FIXED  Issue 47  - IPNetwork cannot be evaluated as a boolean when it has a large size

Release: 0.7.2

Date: 20th Aug 2009

Changes since 0.7.1

Fixed a boundary problem with the iter_iprange() generator function and all associated calls to it throughout the codebase, including unit test coverage and adjustments.

Replaced regular expressions in cidr_merge() with pre-compiled equivalents for a small speed boost.

Adjustments to README raised by John Eckersberg.

Specific bug fixes addressed in this release

• FIXED  Issue 43  - IPNetwork('0.0.0.0/0') not usable in for loop

Release: 0.7.1

Date: 14th Aug 2009

Changes since 0.7

Renamed the netaddr shell script from nash to plain netaddr. This is to avoid a potentially nasty clash with an important Linux tool with the same name. Thanks to John Eckersberg for spotting this one early!

Updated IANA and IEEE data files with latest versions.

Specific bug fixes addressed in this release

•  Issue 42  - Bug in cidr_merge() function when passed the CIDRs 0.0.0.0/0 and/or ::/0

Release: 0.7

Date: 11th Aug 2009

Changes since 0.6.x

Please Note - This release represents a major overhaul of netaddr. It breaks backward compatibility with previous releases. See the API documentation for full details of what is available.

Some highlights of what has changed :-

Internal module hierarchy has been completely overhauled and redesigned. This fixes up a lot of inconsistencies and problems with interdependent imports. All public classes, objects, functions and constants are still published via the main netaddr module namespace as in previous releases.

No more AT_* and ST_* 'constants'.

The Addr base class is gone. This removes the link between EUI and IP functionality so the library is can now easily be split into distinct units without many interdependencies between layer 2 and layer 3 functionality.

The use of custom descriptor classes has been completely discontinued.

Strategy classes and singleton objects have been replaced with a group of strategy modules in their own netaddr.strategy namespace. Each IP or EUI address object now holds a reference to a module rather than a singleton object.

Many operations that were previously static class methods are now presented as functions in the relevant modules. See the API documentation for details. The IP and CIDR classes have been replaced with two new classes called IPAddress and IPNetwork respectively. This name change is important as the IP part of netaddr has been completed redesigned. The notion of an individual IP address and an IP network or subnet has been made more obvious. IPAddress objects are now true scalars and do not evaluate in a list or tuple context. They also do not support any notion of a netmask or CIDR prefix; this is the primary function of an IPNetwork object.

Abritrary IP ranges and are still supported but a lot of their functionality has also been exposed via handy functions.

IP globbing routines (previous known as Wildcards) have been moved into their own submodule.

Added a new IPSet class which fully emulates a mutable Python set. This replaces a lot of half-baked experimental classes found in 0.5.x and 0.6.x such as IPRangeSet and CIDRGroup. See documentation for details.

All methods and properties that previously used or supported the 'fmt' formatting property no longer do so. In all cases, objects are now returned to correctly support pass through calls without side effects. It is up to the user to extract data in the right format from the objects IPAddress objects returned as required.

Unit tests have been completed re-written to support docstring style tests bundled into test suites. These are handy as they double up as documentation being combined with wiki syntax. Implemented code coverage checking using coverage 3.x.

nash - a nascent shell like tool for the netaddr library (requires IPython).

Support for RFC 1924 added ;-)

Specific bug fixes addressed in this release

•  Issue 13  - Searching for a match in a list of CIDR objects
•  Issue 26  - Refactor out use of isinstance()
•  Issue 28  - Add support for network block operations
•  Issue 34  - Addition issue?

Release: 0.6.4

Date: 11th Aug 2009

Major changes since 0.6.3

A small bug fix release for 0.6.x. For new code see release 0.7.

Specific bug fixes addressed in this release

 Issue 40  - Building RPM wth "python setup.py bdist_rpm" fails, multiple errors

Release: 0.6.3

Date: 23rd Jun 2009

Major changes since 0.6.2

• Fixed line endings in a number of new files created under Windows.
• Tweaked the ordering of values in tuple passed into the hash() function in the hash method of the IP and IPRange classes to make it the same as the values used for comparisons implemented in the eq method (Python best practice).
• Added a number of unit tests to improve code coverage.
• Refreshed IEEE Registry files.

Specific bug fixes addressed in this release

•  Issue 33  - CIDR subtraction is broken for out-of-range CIDR objects
•  Issue 35  - install error (on Python interpreters where socket.has_ipv6 is False)
•  Issue 36  - netaddr.CIDR fails to parse default route CIDR
•  Issue 37  - Bug in bitwise AND operator for IP addresses
•  Issue 38  - Feature request: Addr.nonzero
•  Issue 39  - CIDR.abbrev_to_verbose() not applying implicit classful netmask rules consistently

Release: 0.6.2

Date: 13th Apr 2009

Major changes since 0.6.1

Refreshed IEEE and IANA data files with latest revisions from their respective URLs.

• IANA IPv4 Address Space Registry (last updated 2009-03-11)
• Internet Multicast Addresses (last updated 2009-03-17)
• IEEE OUI and IAB files (last updated 2009-04-13)

Added get_latest_files() functions to both the netaddr.eui and netaddr.ip modules to assist in automating release builds.

Specific bug fixes addressed in this release

•  Issue 32  - Addr.ne returns wrong answer

Release: 0.6.1

Date: 7th Apr 2009

Major changes since 0.6

Added a COPYRIGHT file with details and attribution for all 3rd party files bundled with netaddr.

Minimum Python version required is now 2.4.x changed from 2.3.x.

• Python 2.3 compatibility code in many sections of code have been removed.
• the @property and @staticmethod decorators are now used throughout the code along with the reversed() and sorted() builtin iterators.
• A specific version check has also been added that will raise RuntimeError exceptions if you run netaddr on a Python interpreter version < 2.4.x.

Integer addresses passed to the IP() and EUI() constructors no longer require a mandatory second address type (AT_*) argument in most cases. This is now only really required to disambiguate between IPv4/IPv6 addresses with the same numerical value. The same behaviour applies to EUI-48/EUI-64 identifiers. A small speed boost is achieved if the 2nd address type argument is explicitly provided. IPv6 addresses returned by EUI.ipv6_link_local() now always have a subnet prefix of /64.

Default sort order of aggregate classes (IPRange, CIDR and Wildcard) has been changed (again). They now sort initially by first address and then by network block size from largest to smallest which feels more natural.

Fixed a bug in the CIDR.abbrev_to_verbose() static method where IPv4 addresses with 4 octets (i.e. non-partial addresses) were being assigned subnet prefixes using abbreviated rules. All complete IPv4 addresses should always get a /32 prefix where it is not explicitly provided.

Abbreviated address expansion in the CIDR constructor is now optional and can be controlled by a new expand_abbrev boolean argument.

Added the new CIDR.summarize() static method which transforms lists of IP addresses and CIDRs into their most compact forms. Great for trimming down large ad hoc address lists!

Added the previous() and next() methods to the CIDR classes which return the CIDR subnets either side of a given CIDR that are of the same size. For the CIDR 192.0.2.0/24, previous will return 192.0.1.0/24 and next will return 192.0.3.0/24. Also accepts and optional step size (default is 1).

Added the supernet() method to the CIDR class which returns a generator of all the subnets that contain the current CIDR found by decrementing the prefixlen value for each step until it reaches zero.

Changed the way the fallback code works when the socket module is missing important constants and functions.

Removed the uppercase options from the Strategy constructors and internals as this behaviour can be easily replicated using the word_fmt option instead and requires less code (word_fmt='%X').

Specific bug fixes addressed in this release

•  Issue 23  - Improve IPv6 IPv4 mapped/compatible address formatting
•  Issue 24  - bug in CIDR.subnet() when using the fmt argument
•  Issue 29  - CIDR.subnet method's count argument isn't working as documented
•  Issue 30  - not compatible with Python 2.3
•  Issue 31  - byte order in documentation confusing or wrong

Release: 0.6

Date: 22nd Jan 2009

Major changes since 0.5.x

Namespace changes

3 new sub namespaces have been added :-

netaddr.eui

Currently contains IEEE OUI and IAB classes and lookup code.

netaddr.ip

Currently contains IANA IPv4, IPv6 and IPv4 multicast lookup code.

netaddr.core

Currently contains only a couple of classes that are shared between code in netaddr.eui and netaddr.ip.

Please Note This change is part of a two stage internal restructuring of netaddr. In future releases, layer-2 MAC/EUI functionality will be separated from and layer-3 IP, CIDR and Wildcard functionality. All shared code will be moved to netaddr.core. When the migration is complete (expected in 0.7) the netaddr.address and netaddr.strategy namespaces will be removed. Please endeavour to access everything you need via the top-level netaddr namespace from this release onwards rather than diving down into the sub module. See the netaddr.__all__ variable for details of constants, objects, classes and functions exposed and intended for the public interface.

Addition of IEEE and IANA informational lookups

• the IP() and EUI() classes now have an additional .info() method through which contextual information about your addresses can be accessed. This data is published by IANA and the IEEE respectively and sourced directly from text files bundled with netaddr that are available for download publicly online. Details on these files and where they can be found are available in the doc strings of the relevant parsing classes. Subsequent netaddr releases will endeavour to keep up-to-date with the latest revisions of these files.

• the EUI() class has been updated with the addition of the OUI() and IAB() classes. They provide object based access to returned via the EUI.info() method. Please see API docs included with netaddr for details.

• added new NotRegisteredError exception that is raised when an EUI doesn't match any currently registration entries in the IEEE registry files.

Addr() class removed from the public interface

• This class is only ever meant to be used internally and its usage may soon be deprecated in favour converting it into an abstract base class in future releases.

Deletion of AddrRange() class

• replaced with the more specific IPRange() class. AddrRange() wasn't very useful in practice. Too much time has been spent explaining its theoretical merits over its actual practicality for every day use.

Addition of new IPRange() class

• the new base class for CIDR() and Wildcard().

• a killer feature of this new class are the new methods .iprange(), .cidrs() and .wildcard() which allow you to use and switch between all 3 formats easily. IPRange('x', 'y').cidrs() is particularly useful returning all the intervening CIDRs between 2 arbitrary IP addresses.

• IPRange() is a great place to expose several new methods available to sub classes. They are issupernet(), issubnet(), adjacent() and overlaps().

• previous method called data_flavour() has been renamed (again) to a more suitable format().

IP() class updates

• is_netmask() and is_hostmask() methods have been optimised and are now both approximately 4 times faster than previously!

• added wildcard() and iprange() methods that return pre-initialised objects of those classes based on the current netmask / subnet prefix.

• copy constructor methods ipv4() and ipv6() now preserve the value of the prefixlen property now also support IPv6 options for returning IPv4-mapped or IPv4-compatible IPv6 addresses.

• added new methods is_loopback(), is_private(), is_link_local(), is_ipv4_mapped() and is_ipv4_compat() which are all self explanatory.

• added a bin() method which provides an IP address in the same format as the standard Python bin() builtin type ('0bxxx') that is usually only available in Python 2.6.x and higher.

• added a packed() method which provides an IP address in packed binary string format, suitable for passing directly to Python socket calls.

nrange() generator function updates

• by default this now returns IP() objects instead of Addr() objects.

CIDR() class updates

• the strict_bitmask option in the CIDR class constructor has been had a name change and is now just strict (less typing).

• support for Cisco ACL-style (hostmask) prefixes. Also available to the IP() class. They are converted to their netmask equivalents before being applied to the base address.

• added a new subnet() generator method that returns iterators to subnet CIDRs found within the current CIDR object's boundaries e.g. a /24 CIDR can provide address with subnet prefixes between a /25 and /32.

• added a new span() method which takes a list of IP, IPRange, CIDR and/or Wildcards returning a single CIDR that 'spans' the lowest and highest boundary addresses. An important property of this class is that only a single CIDR is returned and that it (potentially) overlaps the start and end addresses. The most important aspect of this method is that it identifies the left-most set of bits that are common to all supplied addresses. It is the plumbing that makes a lot of other features function correctly.

• although IPv6 doesn't support the concept of a broadcast address, after some pondering I've decide to add network() and broadcast() methods to the CIDR class that return the start and end address of a block. It is an interface quirk that users expect so it has been added to aid ease of use.

• the methods .network(), .broadcast(), .hostmask() and .netmask() have been wrapped in property() builtin calls to make them appear as read-only properties.

• added an iterator that returns all valid address within a given CIDR called iter_host_addrs(). This excludes network and broadcast addresses for IPv4 CIDRs and the unspecified address for IPv6 CIDRs.

Support for many more MAC and IPv4 string address representation added

• Improvements to both EUI and IP classes. They now accept many more valid address formats than previously. Thanks for all the bugs tickets raised.

repr() method behaviour change

• Using repr() now assume that you import the relevant names from netaddr before you execute them. They no longer specify the originating namespace of objects which is a bit unnecessary and a lot to read on-screen.They will also be moving around within the namespace shortly anyway so its best not to think of them as being anywhere other than directly below netaddr itself.

'klass' property renamed to 'fmt' (format)

• now referred to as the format callable property and or argument depending on context. An unfortunately but necessary change. klass was a bad initial name choice as it most often doesn't even reference a class object also supporting references to Python types, builtin functions and user defined callables.

Complete re-work and consolidation of unit tests

• now over 110 tests covering all aspects of the API and library functionality.

• Moved all tests into a single file. Lots of additional tests have been added along with interface checks to ensure netaddr's always presents a predictable set of properties and methods across releases.

Nascent support for Python eggs and setuptools

• Help is need to test this as it is not something I use personally.

Significant bug fixes

• fixed the IPv6 string address compression algorithm so that it is now compliant with the socket modules calls inet_ntop() and inet_pton().

EXPERIMENTAL FEATURES

added bitwise operators to the IP class

• does what it says on the tin. Does not effect that value of the IP object itself but rather, returns a new IP after the operation has been applied.

IPRangeSet() class added

• Experimental. May stay, may go away or change significantly. DO NOT RELY ON IT AT PRESENT.
• the intention with this class is to allows you to create collections of unique IP(), IPRange(), CIDR() and Wildcard() objects. It provides iteration over IPs in the collection as well as several membership based operations such as any_match(), all_matches(), min_match() and max_match().

• lots more work to do here. Please raise bugs and feature requests against this as you find them. Improvements to this are coming in 0.7.

Release: 0.5.2

Date: 29th Sep 2008

Bug fixes for 0.5 - release 2

• Fixed Issue 15 in bug tracker. Bad validation and conversion of IPv4 mapped IPv6 address values in IPv6Strategy class. Covered with unit test cases.

• Updated PrefixLenDescriptor() class so that modifications to the property CIDR.prefixlen also update CIDR.first and CIDR.last keeping them in sync. Covered by unit test cases.

• IP.hostname() method returns None when DNS lookup fails.

Release: 0.5.1

Date: 23rd Sep 2008

Bug fixes for 0.5 - release 1

• CIDR class constructor was throwing a TypeError for valid unicode string addresses which worked in previous releases. Fixed and covered with a unit test case.

• The methods CIDR.netmask() and CIDR.hostmask() contained code errors that were causing them to fail. Problem fixed and covered with unit test case.

Release: 0.5

Date: 19th Sep 2008

Major changes since 0.4.x

General changes

• access to all important object attributes in all netaddr classes now takes place via custom Python descriptor protocol classes. This has greatly simplified internal class logic and made external attributes changes much safer and less error prone. It has also made aggregate classes such as CIDR and Wildcard effectively read-write rather than read-only which they have been up until this release.

• ammended the way sort order is calculated for Addr and AddrRange (sub)class instances so that the address type is taken into account as well as as the numerical value of the address or address range. The ascending sort order is IPv4, IPv6, EUI-48 and EUI-64. Sequences of AddrRange (sub)class instances now sort correctly!

• comparisons between instances of Addr and AddrRange (sub)classes now return False, rather than raising an AttributeError.

• obeying the Law of Demeter, the address type of Addr and AddrRange (sub)class instances can be accessed using the property directly :-

obj.addr_type # 0.5 onwards

rather than having to go via the strategy object :-

obj.strategy.addr_type # 0.4 and earlier

• renamed the AT_DESCR lookup dictionary to AT_NAMES. Removed invalid and duplicated imports from all modules.

Addr class changes

• removed the setvalue() method from the Addr class and replaced all uses of __setattr__() replaced by custom descriptors throughout.

IP class changes

• removed the ambiguity with masklen and prefixlen attributes in the IP class. prefixlen now denotes the number of bits that define the netmask for an IP address. The new method netmask_bits() returns the number of non-zero bits in an IP object if the is_netmask() method returns True. A prefixlen value other than /32 for an address where is_netmask() returns True is invalid and will raise a ValueError exception.

• removed the family() method from the IP class. It duplicates information now provided by the newly exposed addr_type property.

• IP class has several new methods. is_multicast() and is_unicast() quickly tell you what category of IP address you have and while ipv4() and ipv6() act as IPv4

  <->

  IPv6 conversions or copy constructors depending on context.

• reverse DNS lookup entries now contain a trailing, top-level period (.) character appended to them.

AddrRange class changes

• the AddrRange class no longer stores instances of Addr (sub)classes for the first and last address in the range. The instance variables self.start_addr and self.stop_addr have been renamed to self.first and self.last and the methods obj.first() and obj.last() have been removed.

Instead, self.first and self.last contain integer values and a reference to a strategy object is stored. Doing this is a lot more useful and cleaner for implementing internal logic.

CIDR class changes

• the strictness of the CIDR class constructor in relation to non-zero bits once the prefix bitmask has been applied can be disabled use the optional argument strict_bitmask=False. It is True (strictness enabled) by default.

• fixed a bug in abbreviated CIDR conversion. Subnet prefix for multicast address 224.0.0.0 is now /4 instead of /8.

• the CIDR class now supports subtraction between two CIDR objects, returning a list of the remainder. Please note that the bigger of the two CIDR objects must be on the left hand side of the the expression, otherwise an empty list is return. Sorry, you are not allowed to create negative CIDRs ;-)

• the function abbrev_to_cidr() has been renamed to and turned into the static method CIDR.abbrev_to_verbose(). No major changes to the logic have been made.

Wildcard class changes

• the Wildcard class now defines a static method Wildcard.is_valid() that allows you to perform validity tests on wildcard strings without fully instantiation a Wildcard object.

Release: 0.4

Date: 7th Aug 2008

Major changes since 0.3.x

• all general Exception raising has been replaced with more specific exceptions such as TypeError and ValueError and with the addition of two custom exception classes, AddrFormatError and AddrConversionError.

• The IP class now accepts a subnet prefix. It is NOT strict about non-zero bits to the right of implied subnet mask, unlike the CIDR class (see below).

• The CIDR class is now completely strict about non-zero bits to the right of the implied subnet netmask and raises a ValueError if they exist, with a handy hint as to the correct CIDR to be used based on the supplied subnet prefix.

• The CIDR class now also supports abbreviated CIDR ranges and uses older classful network address rules to decided on a subnet prefix if one is not explicitly provided. Supported forms now include 10, 10/8 and 192.168/16. Currently only supports these options for IPv4 CIDR address ranges.

• __repr__() methods have been defined for all classes in the netaddr module producing executable Python statements that can be used to re-create the state of any object.

• CIDR and Wildcard classes now have methods that support conversions between these two aggregate types :-

• CIDR() -> Wildcard()
• Wildcard() -> CIDR()

Housekeeping Changes

• massive docstring review and tidy up with the inclusion of epydoc specific syntax to spruce up auto-generated API documentation.

• Thorough review of code using pylint.

• netaddr module now has the special __version__ variable defined which is also referenced by setup.py.

• Some minor changes to setup.py and MANIFEST.in.

• constants and custom Exception classes have been moved to __init__.py from strategy.py.

• A from netaddr import * friendly __all__ has been defined for the netaddr namespace which should remove the need to delve too much into the address and strategy submodules.

• fixed a number of line-ending issues in several files.