Personally I do:

defp halt_if_not_signed_in(conn, nil), do: :whatever
defp halt_if_not_signed_in(conn, signed_in_account) do
  ...
end

Unless you just want nil to error out, in which case I’d optimally do a precise match:

defp halt_if_not_signed_in(conn, signed_in_account) when is_binary(signed_in_account)

Or so.

I’m personally a fan of val != nil because nil is just an atom, so I don’t even know why it has a specialized function like is_nil/1 anyway, especially since the empty list [] is the nil value on the system, those confused me for a bit when I started Elixir (nil should have been [] as the empty list is named nil standardly, not :nil, blah).

agreed. I find using pattern matching to handle this is very readable.

I’ve helped introduce two Java engineers to Elixir and they both found this a really useful language feature, especially as they discovered some of the powerful variations of pattern matching.

As for other uses of not is_nil(val) vs val != nil I’d say just pick one and be consistent within your project. Personally not is_nil(val) is not my favorite.