“Readable” is mostly a matter of “fits the grooves already worn in my brain”; to me, the Elixir there nonstandard-macro-based-syntax-and-all, is vastly more readable than the Clojure.
I'll admit to being able to read Lisp syntax without choking on the parens, but I don't even use Clojure (or any other dialect that uses special syntax for the different collection types) and I think it's way more readable and malleable (you can literally do whatever you want that makes sense to this data and it'll work, meaning you're free to write whatever middleware you want that takes and produces the expected data).
I've used Elixir since 2015 so I have no problem actually reading the syntax at all; I just think it's factually much worse than the Clojure example because it's basically just a bunch of macros you can't do much with and lots of implicit behaviors.
This also applies to Elixir's greater position in the BEAM ecosystem. I think the idea was that Elixir was going to be a boon to Erlang and other languages in there over time but I think it's demonstrably pretty useless as a BEAM citizen except for some of the patches they've submitted; most big Elixir libraries can't even be used outside of Elixir because they're full of macros and don't even provide interfaces with just functions.
Not sure whose idea it was but I would say this is a very inaccurate take. Of course, we have taken much more from Erlang than given (that's expected from hosted languages), but I personally sit on more than 100 PRs to Erlang/OTP. I made binary matching several times faster by using SIMD. I improved the compiler performance (including Erlang programs) by (conservatively) 10-20%. I contributed compiler optimizations to reduce memory allocation. There is a now faster sets implementation which I contributed. Erlang/OTP 26 ships with faster map lookups for atom keys based on a pull request I submitted. The new logger system in Erlang takes ideas from Elixir logger. The utf-8 string handling in Erlang is based wholesale on Elixir’s (and extended for lists). And this is in no way a comprehensive listing [1].
And this is not counting everyone else in the Elixir community who contributed and those who are now actively working on Erlang tooling and were on-boarded to the ecosystem via Elixir. The Erlang ecosystem now has a package manager [2], used extensively by both languages, rebar3 and Mix, all implemented and powered by Elixir.
And then there are efforts like the Erlang Ecosystem Foundation, which is made of developers and sponsored by companies working with both languages, and it delivers important projects around observability, documentation, and more. For example, you can find several Erlang projects now using ExDoc for documentation, such as Recon [3].
That’s my take. What are your demonstrations that we are pretty useless for the BEAM ecosystem?
1: https://github.com/erlang/otp/pulls?q=is%3Apr+author%3Ajosev... 2: https://hex.pm/ 3: https://hexdocs.pm/recon
It is. That they have not replied to any of your responses shows they don't have much argument.
I mean using "factually" in this subjective opinion is just ridiculous.
> I just think it's factually much worse than the Clojure example
My hyperloglog erlang lib is one of the most advanced in any language (behind what Omar Ertl does for Java) and i come wholesale from elixir.
I have my own code, as an elixir person, in OTP, for the formatting of float to string. The maybe construct was only possible because elixir with showed the way.
I could probably keep going.
There are lots of silent fans of your teams' work. Please keep it up, and please feel proud of your collective outputs!
I was just reading the documentation, where it says "remember that macros are not your API", and smirking. If you write macros and things depend on them, such that if you change the macros, things will break, macros are your API. Syntax is API.