Elixir Streams(blog.drewolson.org)
blog.drewolson.org
Elixir Streams
http://blog.drewolson.org/elixir-streams/
3 comments
When people say Elixir is just syntactic sugar on top of erlang I point them to Stream and Enum. It's a great example of how polymorphism via protocols is an enabler for powerful designs.
Elixir is so much more than syntactic sugar... but you can't deny that it's also very, very sweet.
I think it's the best language out there-- the power of erlang and OTP and the fun of, well, elixir. It's more fun than Ruby and Python.
I think it's the best language out there-- the power of erlang and OTP and the fun of, well, elixir. It's more fun than Ruby and Python.
That's the best way to describe it. The concurrency model and metaprogramming bits mean that effortless to make cool stuff with it. And it turns out that's way more fun that wrestling with the jvm or other high friction tools.
There's a lot of neat stuff in Elixir, and the community is great too. I honestly it may be the first functional language that becomes widely used. Hopefully.
There's a lot of neat stuff in Elixir, and the community is great too. I honestly it may be the first functional language that becomes widely used. Hopefully.
Interesting to see this as a novel thing in another ecosystem. Rust actually uses "stream" manipulation as a default way of dealing with things that might otherwise be expressed as full data structures. For example, the equivalent to one of the code snippets in the article would be:
BufReader::new(File::open("myfile.txt").unwrap())
.lines()
.enumerate()
.map(|(i, line)| format!("{}: {}", i, line.unwrap()))
.take(1)
.next().unwrap()
(formatted for non-monospace font readability). Note the .unwrap() is where error handling should normally happen.
BufReader::new(File::open("myfile.txt").unwrap())
.lines()
.enumerate()
.map(|(i, line)| format!("{}: {}", i, line.unwrap()))
.take(1)
.next().unwrap()
(formatted for non-monospace font readability). Note the .unwrap() is where error handling should normally happen.
The problem with "stream/generator as default" is the boilerplate you constantly have to add.
In Python 3, for example, map() returns a generator. So you can no longer transform a list into another list with `new = map(func, lst)`. You have to do `new = list(map(func, lst))`. One of the most subtle gotchas when upgrading from 2 to 3 is having to wrap list() around functions and methods like map(), range(), dict.items(), etc.
Discussion I've seen about Rust on less moderated websites often seems to make fun of the .unwrap() spam seen in a lot of programs, to the point of the mockery even dominating some discussions.
Streams and lazy evaluation are really nice, but sometimes you just want some nice procedural handling.
In Python 3, for example, map() returns a generator. So you can no longer transform a list into another list with `new = map(func, lst)`. You have to do `new = list(map(func, lst))`. One of the most subtle gotchas when upgrading from 2 to 3 is having to wrap list() around functions and methods like map(), range(), dict.items(), etc.
Discussion I've seen about Rust on less moderated websites often seems to make fun of the .unwrap() spam seen in a lot of programs, to the point of the mockery even dominating some discussions.
Streams and lazy evaluation are really nice, but sometimes you just want some nice procedural handling.
As another comment mentioned, .unwrap() is really useful for testing things out, but any legitimate (read: with intent to be used for some practical purpose) library or program is going to handle most errors. One edge case may be calling unwrap() in a thread that should die if the unwrapped operation doesn't succeed.
When using Rust for more than toy programs (which, assuming I am thinking of the same "less moderated" websites that you are, is unlikely to be the experience of the commenters), unwrap() isn't particularly common outside of main(). You generally want to handle the error cases. It's used a lot in examples because they have to function as one-offs.
All the boilerplate you need to add is
lmap = lambda f, l: list(map(f, l))
One has to wonder why this is not included in the functools or some other module.
Dunno, to promote generators? I'd say people often convert to list for no good reason, so adding friction might discourage that.
Then again, map() and friends are not exactly the recommended way, list comprehensions and generator expressions are usually preferred by the BDFL.
Then again, map() and friends are not exactly the recommended way, list comprehensions and generator expressions are usually preferred by the BDFL.
I wonder what the Erlang equivalent would look like (passing around data that are funs)?
I'm mostly an Erlang guy, but I do like that about Elixir: they really seem to be taking the Erlang base, and building on it in some interesting ways.
That's the macros. Phoenix's routing is an awesome example of the power of Elixir macros. It leverages the BEAMs pattern matching capabilities to do the routing, but the syntax is very clear and concise.
Minor nitpick. BEAM doesn't have pattern matching built-in. Core Erlang does. Core Erlang compiles to BEAM bytecode with functions using regular if-statements. So basically: Erlang/Elixir -> Core Erlang -> BEAM. (Actually, I think there is one more step there).
Minor nitpick to the nitpick, since I dug down in Elixir's compiler before. Elixir spits out a normal Erlang AST, not a Core Erlang AST, which is a bit different. There are a half dozen levels between the Erlang AST and BEAM as well. There's several instances of a talk by Robert Virding about implementing languages on the BEAM where he talks about which advantages different levels of the compiler to hook into.
The closest thing I can think of outside of using funs or macros to define the usual `delay` and `force` operators, and that currently exists, would be something like the QLC interface, which allows you to turn whatever you want into a type that can be used in list comprehensions and functional queries around it: http://www.erlang.org/doc/man/qlc.html