I was quite surprised by how amicable and positive this blog post was; I expected him to have lots of objections. It's admirable that an old hand like Joe Armstrong can be so open and accepting of something that essentially sets out to improve on his lifework.
In fact, if you look at his paper on the development of Erlang [1] while at Ericsson, which is coincidentally a juicy, fascinating read, Armstrong is being very forthcoming about the struggle to develop the language, in particular developing a good syntax, and also very forthcoming about the involved parties' lack of experience with language design; in particular, the Prolog-like syntax seems less of a planned decision than a side-effect of the first implementations being written in Prolog itself, and following the path of least resistance. Also, it worked.
At any rate; it seems clear to me that Armstrong values the principles behind Erlang's design (concurrency, immutability, fault tolerance and so on) more than the actual syntax. I wonder, at this point, how amenable he would be to larger syntax changes to Erlang itself, or if he is happy with the fact that Elixir is a completely separate project.
(There is something weird about the effusive tone and surreal humour in the blog post, though. Is this his new "Erlang evangelist voice", or has he always been writing in this way?)
"Today there is an unhealthy concentration on language and
efficiency and NOT on how things fit together and
protocols - teach protocols and not languages.
And teach ALGORITHMS."
I started working with Erlang for work, on and off, for the past year. I had been working with Ruby for a lot longer. I too, liked the Erlang design, but the syntax is a wreck. It's also true, use something long enough and you get used to it. I've noticed long-time Erlang users tend to have this mystical ability to see the underlying elegance expressed by the Erlang syntax while ignoring the syntactic awkwardness.
Elixir brings a lot of good stuff onto the table. I hope the Elixir community incorporates the changes Joe Armstrong suggests.
One in particular: f.() makes perfect sense in the Ruby world because everything is an object (we're sending the '()' message to the 'f' object); but things are built out of functions in the Erlang world, and f() makes more sense there. There are no objects or messages to send them, so 'f.()' gets parsed as an extra dot at the end of the name. I think Armstrong's other suggestions follow in the same vein.
The point isn't to try to make Elixir have a Ruby-like syntax so much as, having distilled the syntax design principles developed in the Ruby community, apply it in Elixir.
"I've noticed long-time Erlang users tend to have this mystical ability to see the underlying elegance expressed by the Erlang syntax while ignoring the syntactic awkwardness."
It's just practice. It no longer trips me up, but I still find it klunky and unpleasant after nigh on 6 years of medium-duty usage.
lists:map(fun (X) -> X + 1 end, [1, 2])
vs.
map (+1) [1, 2]
Augh. Yes, I'm cheating perhaps a bit with the (+1), but Erlang syntax has enough other stuff in the way for it to still be fair in the end. For instance, the example Joe gave, where you end up with this long sequence of
"Ant turd tokens", or .,; are a big pain in the neck when refactoring stuff. For instance
case Something of
{ok, Val} ->
do_something();
_Error ->
barf()
end,
The ; and the , are easy to forget about. Indeed, the , might be changed to a . if you eliminate the next expression, and those things can be easy to miss when fiddling with code.
I picked that as a simple example. Yes, I suppose I should have used a fold, or any of the other "everything else" functions that don't have syntax support. My point still stands.
Comparing Haskell and Erlang syntax isn't even fair. Haskell threw homoiconicity and homogeneity way out the window in exchange for some very nice, mathematical notation.
In fact, if you look at his paper on the development of Erlang [1] while at Ericsson, which is coincidentally a juicy, fascinating read, Armstrong is being very forthcoming about the struggle to develop the language, in particular developing a good syntax, and also very forthcoming about the involved parties' lack of experience with language design; in particular, the Prolog-like syntax seems less of a planned decision than a side-effect of the first implementations being written in Prolog itself, and following the path of least resistance. Also, it worked.
At any rate; it seems clear to me that Armstrong values the principles behind Erlang's design (concurrency, immutability, fault tolerance and so on) more than the actual syntax. I wonder, at this point, how amenable he would be to larger syntax changes to Erlang itself, or if he is happy with the fact that Elixir is a completely separate project.
(There is something weird about the effusive tone and surreal humour in the blog post, though. Is this his new "Erlang evangelist voice", or has he always been writing in this way?)
[1] https://docs.google.com/viewer?url=http%3A%2F%2Fwebcem01.cem...