Add a common check for `is_foo?` function names that mix up two separate naming patterns

[angelikatyborska]

I saw this in one of the community solutions:

defp is_allowed_char?(char), do: (char >= 97 and char <= 122) or char == 95

The correct name is either allowed_char? or is_allowed_char, not both. Source: https://hexdocs.pm/elixir/naming-conventions.html#is_-prefix-is_foo

We should add a common check that looks for functions and macros named is_foo? and points to the explanation that it should be either foo? or is_foo (but we don't need to try to detect which it should be).

Edit based on the discussion below:

We should add checks that follow this table:

definition	is_foo?	is_foo	foo?
def/depf	🚫	🚫	✅
defguard/defguardp	🚫	✅	🚫
defmacro/defmacrop	🚫	✅	✅

[jiegillet]

is_foo is reserved for guards clauses though right?
That's what mentor extraordinaire @chriseyre2000 taught me anyway :)

It would be simple enough to check the distinctions and only allow
def/depf foo?
defguard/defguardp is_foo

[angelikatyborska]

It's not that straight-forward. Yes, for custom guards we could completely ban foo?, and for functions we could completely ban is_foo but the gray area are macros. Macros can be used in guards, but only if they consist of expressions that can be used in guards. For example:

defmodule Foo do
  import List, only: [wrap: 1]

  defmacro is_foo(x) do
    quote do
      unquote(x) == :foo
    end
  end
  defmacro foo?(x) do
    quote do
      # List.wrap
      wrap(unquote(x)) == [:foo]
    end
  end
end

defmodule Bar do
  require Foo
  import Foo
  
  # works
  def seven_for_foo(x) when is_foo(x), do: 7

  # won't work
  # "cannot invoke remote function List.wrap/1 inside guards"
  def seven_for_foo(x) when foo?(x), do: 7

  def seven_for_foo(_), do: nil
end

[jiegillet]

Macros, man...
How about this then?

definition	is_foo?	is_foo	foo?
def/depf	🚫	🚫	✅
defguard/defguardp	🚫	✅	🚫
defmacro/defmacrop	🚫	✅	✅

And we can have a common message for all explaining the convention and mention that for macros it depends on the use case.

[angelikatyborska]

How about this then?
(table)

That looks correct to me.