Translate "Add delete/1, mocked/0, and mocked_process/1"

[paulo-ferraz-oliveira]

Translated⁽¹⁾ from 2Latinos/nuntius#28...

Notes on this implementation

• 🟧 Credo has A LOT of non-default stuff
  • even though we may end up with a shorter rule list we wanna go full strict, at this moment, to test its/our limits
• 🟩 macros ftw 🎉
  • even if small, in this context, so easily replaced by a function, I wanted to test this out, so went with it (don't hate me yet!)
• 🟩 defdelegate for expressiveness/semantics
  • I'd used it in the past, in the context of behaviour implementations, and here it just feels like syntactic sugar (I'm not sure it's much more than that, though)
• 🟧 . after function names
  • odd, when coming from the Erlang world, but easy to adapt to
• 🟧 using & to pass a function reference
  • odd, when coming from the Erlang world, I think I'd preferred fn for consistency, but I get where the language's coming from
• 🟩 if used
  • not common, for me, in Erlang, but easy to use in Elixir, since the syntax is much cleaner (if/else, instead of if/true)
• 🟩 string interpolation
  • I think something's being cooked up for Erlang, but "#{__MODULE__}.mocked_process" is a ftw
• 🟩 @doc false
  • using this instead of @hidden or @private (since @doc is already what you use for regular doc.) is cleaner and makes a lot of sense
• 🟧 jumping to Erlang via :
  • it probably wouldn't make sense to replicate all the Erlang public API in Elixir, but e.g. proc_lib and timer surely could do with some wrappers (this is just aesthetics for me, really 😄)
• 🟧 defp/def
  • I quite like that Erlang modules' API is exposed at the top (and you don't have to mess with indentation when making something private/public), but there's (dis)advantages to both approaches
  • I miss rebar3 xref 😢
• 🟩 tests' on_exit(
  • this is so much more ergonomic than init_per_testcase/end_per_testcase, so it's a win
• 🟧 variable rebinding (^)
  • I don't love this (yet?): coming from an Erlang world and being used to "pinned by default": I shot myself in the foot twice while writing the tests (look at where ^ is used and you can see how easy it is to forget it and have the tests pass)
  • I wished https://www.erlang.org/eeps/eep-0055 would've gotten traction, implemented as a trial, and potentially faded out if the community at large disliked it...
• 🟧 compile-time validations
  • there's some compile-time validations (like checking if a function exists) which are odd when coming from the Erlang world. While the "real world" problem is to potentially not load functions after they're used (and I guess that's what rebar3 xref is also used for) I am torn between having better compile-time validation and forcing me to use a compiler option to not have it (choices, choices, ...)

---

⁽¹⁾ an Erlang > Elixir translation, as presented here, is not a copy-paste exercise but rather a re-thinking of the implementation in a different language using that language's constructs as best as possible. The goal is to 1. make the API similar, 2. make it work, 3. make it idiomatic.

[paulo-ferraz-oliveira]

Thanks. Merged!