Elixir's biggest gotchas?

For me, it’s the functional programming. Elixir is not the first functional programming I learn. But it’s the first functional programming I learn by doing, not just by reading the books. Another things are like pattern matching, pipe operator, recursive call, etc.

erlang charlists. Newbie exercises are bound to create a list of integers and iex will print it out in a surprising way. It’s a fundamental ambiguity that’s completely understandable, but it introduces itself to newcomers in a very surprising way.

For me the trickiest thing has been new patterns for managing state. The BEAM ecosystem has extraordinarily powerful tools for this, but learning new abstractions for it like the proper use of genservers and supervision trees has been a road with some missteps for me.

Yep–that’s one of those questions that keeps coming up again and again.

Allowing rebinding of values leads people new to Elixir to think we allow mutation.

iex> x = 1

iex> x = 2

Perfectly legal and people think it means that someone is mutating x–which they’re not. But this seems to be one of those things that people inevitably see and say “Hey waitaminute–I thought Elixir doesn’t allow mutation? WTF?”

I don’t know that I’d call this a “gotcha” but it does seem to confuse new folks on a fairly regular basis.

For me so far syntax for piping into anonymous function was a bit surprising:
|> (&(&1)).()

I’d say most gotchas are related to OTP and state management—it’s really a whole different beast from the language itself.

Language-level gotchas and common pitfalls include:

• = is not assignment
• Using if instead of function pattern matching
• Lists and keyword lists—a lot of people think they’re the same as their Ruby or JavaScript arrays

Erlang charlists can be confusing but it is not restricted to just erlang charlists. You have exactly the same issue with elixir strings:

iex(1)> <<195,165,195,164,195,182>>
"åäö"
iex(2)> 

It’s a result of the erlang system only having a fixed set of datatypes and not being able to define your own datatypes, and there is no character or string base type. This means choosing an existing type to use as strings: erlang chose lists which was common in the functional world, while elixir has chosen binaries.

Robert

I think Elixir may actually make it slightly more confusing because a lot of folks coming from other language expect 'foo' == "foo". That’s why one of the proposals I want to send after v1.3 is out is to deprecate the use of 'foo' in the long-term (and I really mean long-term) in favor of the sigil ~c"foo". It is going to be more explicit and I find using sigils for char lists in Elixir at this point more idiomatic.

I like that. The experienced developers needing to interact with erlang libs will be comfortable with sigils, and the newcomers won’t have to learn it until later. Brilliant!