As per documentation, it should understand the same arguments as IPv*Network.

Unfortunately it does not recognize netmask in string form. Hence the following code will fail:

    ipaddress.ip_interface(('192.168.1.10', '255.255.255.0'))

while the following will work:

ipaddress.ip_network(('192.168.1.10', '255.255.255.0'), strict=False)

This should be easy enough to fix, at least in IPv4Interface.__init__. It needs to copy some of IPv4Network.__init__, dealing with address[1] and calling _make_netmask(). Currently, it just calls int(address[1]).

I haven't looked at IPv6Interface.

Tests are also needed, of course.

The document here says: https://docs.python.org/3/library/ipaddress.html#interface-objects

""IPv4Interface is a subclass of IPv4Address""

trying with:

>>> ipaddress.IPv4Address(('192.168.128.0', '255.255.255.0'))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.6/ipaddress.py", line 1284, in __init__
    self._ip = self._ip_int_from_string(addr_str)
  File "/usr/lib/python3.6/ipaddress.py", line 1118, in _ip_int_from_string
    raise AddressValueError("Expected 4 octets in %r" % ip_str)
ipaddress.AddressValueError: Expected 4 octets in "('192.168.128.0', '255.255.255.0')"

So the behavior of IPv4Interface seem to be correct?

While an IPv4Interface may be a subclass of an IPv4Address, it definitely has more information attached to it: the netmask of the network it's on. So an interface (like a network) needs to allow additional parameters to specify the netmask.

It should be true that IPv4Interface is like IPv4Network, but it will allow arbitrary host bits. A IPv4Network in strict mode will have all zeros for the host bits, while a IPv4Network in non-strict mode will allow ones in the host bits (see Ilya's original example).

If you look at IPv4Interface.__init__, it actually does just that: creates an IPv4Network with strict=False, then extracts the network parts. I'm not exactly sure why it also has the int(address[1]) code there, too, since IPvv4Network deals with the address[1] part. I would think extracting _prefixlen from the network (as it does later in __init__ for the non-tuple case) would be good enough.

Eric: I made the patch, reference to which IPv*Network dealing with tuple. Should I also add the unittest for it?

Also, can you help me code review this, thanks.

Thanks! Yes, we'll need tests.

I'm out of town most of the weekend, but I'll look at this as soon as I can.

Add unittest. Since IPv6 do not support prefix netmask ('ffff:ff00::'), it have only test like this:

>> a = ipaddress.ip_interface(('dead:beaf::', '32'))
>> b = ipaddress.ip_interface('dead:beaf::/32')
>> str(a) == str(b)

Can this be included into the next bugfix release?

Do you need any help with the change?

@JelleZijlstra The issue seems to be resolved in this PR.

Yes, thanks!