NetAddress: tools for IP and Mac addresses

Pretty useful, thanks for that.

I would only change from_string/1 to parse/1 and parse!/1 instead to match the Integer and Float API more closely.

Parse returns a tuple that contains an extra binary useful for unit matching, no?

Yes, but in general I would assume IP.from_string/1 (or IP.parse/1) to return error tuple instead of throwing on error. You can pick any naming, but I would expect to have “failure-safe” way to parse data.

I think in elixir, .to_x and .from_x are designed to output naked values (String.to_integer/1 raises on a non-integer string), and be in pipelines, but your point about wanting something that outputs an error tuple on conversion is taken under advisement! function that returns an error tuple instead of raising on conversion · Issue #1 · ityonemo/net_address · GitHub

While you are generally correct I believe that having a bang variant like parse! signals even more clearly that the function might raise an exception which also implies that parse would return an ok/error tuple.

As another anecdotal evidence data point, I would also naively believe that from_string would not raise and would return an ok/error tuple as well (although this expectation might be severely misguided – I really don’t know).

The standard library’s Map.from_struct/1 does not return an error tuple. But doing a more comprehensive look, I see that the datetime/time/naivedatetime modules have from_* functions which do emit error tuples. I guess network addresses seem more like datetimes than structs, so I’ll go ahead and make the from_ functions emit tuples and have the bang equivalents throw.

updated in 0.2.0 to have bang functions and tuple functions for from_* functions, and as a bonus I implemented random and IP.SockAddr module.

I like how you used the m modifier on sigil_i. Nice touch.

updated in 0.3.0 to have ranges and subnets work in each other’s Kernel.in/2 implementations; also implements is_in function to detect membership in ranges and subnets, in guards.