I’m curious if part of the motivation for this is in fact lowering the barrier of entry for developers coming from other languages, specifically Python, as part of a long term strategy of moving Elixir into ML and AI arenas. This would be consistent with recent focus on launching/growing Nx. That’s not an unreasonable motivation if so.
stevensonmt
jeregrine
Phoenix Core Team
I want to voice my support for this whole heartedly and enthusiastically. This is great and should proceed.
If there are concerns about the similar to module constants then maybe ::local_accumulator = would suffice, though would require new AST I believe.
We could also make a special syntax for it similar to how clojure does atom’s though they aren’t a one to one match for this, they do mark the values as distinct from normal values.
I also do not think we should let FUD about education hold us back either. This change is relatively small in the grand scheme of things and will MOSTLY make things like Advent of Code and other learning problems easier to accomplish, not harder. We should assume people learning Elixir are capable enough to understand what a scope is.
14
hst337
Since you’re supporting this feature, could you please explain what benefits it brings to the language? I mean, like readability of code, shorter solutions or lower learning curve?
1
jeregrine
Phoenix Core Team
I agree with the reasoning provided, by José. The problem statement and proposed solution.
2
hst337
Yeah, I understand that 
I just couldn’t find the exact problem in the original post. Original post just compares Elixir’s and Python solutions and shows how we can have Elixir’s solution to look like Python’s one. And I am asking why do we want Elixir’s solution to look like Python’s one?
3
hst337
I have a few examples above where scope of the feature is extremely unclear and is confusing for me as a professional Elixir and Python developer.
0
jeregrine
Phoenix Core Team
I do not know what to say here. I think the annoyances it solves are greater than the annoyances you pointed out. And that José makes a compelling case for his proposed solution.
3
benwilson512
Author of Craft GraphQL APIs in Elixir with Absinthe
I think as far as the details of the proposal goes, this is likely the best one yet.
Personally I am on the fence about whether it’s actually a good idea though. I think it sort of trades messier code for a messier mental model. The code with @@ is certainly more clean in a sense, but it’s hard to beat the “cleanliness” of looking at the single line
And knowing that that encapsulates the entire preceding function. When functions are representable by their return value you can just |> debug that one line and watch the whole thing unfold. You can extract it and know that nothing has changed.
I am sure that with some compiler warnings and so forth the experience with @@ can be good. I have just come to really love the “return value has everything” idea because while there may be more code, you can focus on less of it at a time.
21