Golang IP <-> int conversion

ipint.go

func ip2int(ip net.IP) uint32 {
	if len(ip) == 16 {
		return binary.BigEndian.Uint32(ip[12:16])
	}
	return binary.BigEndian.Uint32(ip)
}

func int2ip(nn uint32) net.IP {
	ip := make(net.IP, 4)
	binary.BigEndian.PutUint32(ip, nn)
	return ip
}

Yeah, i got it. I put it here for future adventurers of GO

func ipv6ToInt(IPv6Addr net.IP) *big.Int {
    IPv6Int := big.NewInt(0)
    IPv6Int.SetBytes(IPv6Addr)
    return IPv6Int
} 

func IntToIpv6(intipv6 *big.Int) net.IP {
    ip := intipv6.Bytes()
    var a net.IP = ip
    return a
}

Regards

Hi @srPuebla, are you using these 2 functions, especially for IntToIpv6 ? I think there is a bug for it, you can try the case:

1. convert ::1 to bigInt by ipv6ToInt
2. reverse the conversion by using IntToIpv6
   After above steps, you may find the problem. Besides, there should be a new solution to handle it.

@zhlsunshine The reason is the net.IP accepts an array of byte. When converting back from BigInt, it needs to fill the prefix with 0s up to the length of the IP or it can't recognize the IP. See this example: https://pkg.go.dev/net#IP
praserx has a solution for this: https://github.com/praserx/ipconv/blob/master/ipconv.go
Or just write a loop to determine the IP type (v4 or v6), then fill out the prefix slice.

@ammario Can you explain why you check len(ip) == 16? Wouldn't it return only a quarter of the ipv6 IP?

@ammario Can you explain why you check len(ip) == 16? Wouldn't it return only a quarter of the ipv6 IP?

I honestly can't remember why I did that. The behavior doesn't seem useful, it should probably panic instead. I updated the gist.

@ammario Can you explain why you check len(ip) == 16? Wouldn't it return only a quarter of the ipv6 IP?

I honestly can't remember why I did that. The behavior doesn't seem useful, it should probably panic instead. I updated the gist.

I'm surprised no one has mentioned it yet, but this latest change actually breaks the code. The length of both ipv4 and ipv6 addresses in net.IP representation is 16 bytes due the data structure being designed to accommodate both types of addresses. As proof of this point, see the following unit test:

func TestIPRepresentation(t *testing.T) {
	ipv4Str := "192.168.1.1"
	ipv6Str := "2001:0db8:85a3:0000:0000:8a2e:0370:7334"

	ipv4 := net.ParseIP(ipv4Str)
	ipv6 := net.ParseIP(ipv6Str)

	t.Logf("IPv4: %s, Length: %d bytes", ipv4, len(ipv4))
	t.Logf("IPv6: %s, Length: %d bytes", ipv6, len(ipv6))

	if len(ipv4) != 16 {
		t.Errorf("IPv4 length is not 16 bytes")
	}

	if len(ipv6) != 16 {
		t.Errorf("IPv6 length is not 16 bytes")
	}
}

This test results in the following output:

=== RUN   TestIPRepresentation
    ./networking_test.go:72: IPv4: 192.168.1.1, Length: 16 bytes
    ./networking_test.go:73: IPv6: 2001:db8:85a3::8a2e:370:7334, Length: 16 bytes
--- PASS: TestIPRepresentation (0.00s)
PASS
ok      gitlab.eitccorp.com/elicsar/elx/elx-manager/pkg/eitccommon      0.231s

I've come up with this alternative implementation that depends on net.IP's To4() method, which returns nil for anything that's not a valid v4 IP.

func IPv4toInt(ipv4 net.IP) (uint32, error) {
	ipv4Bytes := ipv4.To4()
	if ipv4Bytes == nil {
		return 0, errors.New("not a valid IPv4 address")
	}
	return binary.BigEndian.Uint32(ipv4Bytes), nil
}

Note that my implementation does return error, so, if you need a function that only returns a single value, you can always change to a penic like the original function, but I personally like returning an error better.

@wilrodriguez I didn't realize how popular this gist was and recklessly updated it. I reverted it back per your message. Thank you!

Really helpful!