If Erlang is so great, why hasn't it taken off? | Hacker News Reader

If Erlang is so great, why hasn't it taken off?

11 pointsehsanul5 comments

I just read "What makes Erlang so great?", posted here on HN ( http://www.clickcaster.com/item/view/what-makes-erlang-so-great ). The question it makes me ask is, if it really is that great, why don't I see more of it? Sure I see people commenting on how wonderful it is and whatnot, but other than that, where are the applications?

It seems similar to how Lisp is wonderful, as is Smalltalk, and Haskell. But none of those "took off", for various contested reasons. What's the reason Erlang hasn't (yet)? Or am I just wrong that it hasn't taken off in the first place?

View on HN

I guess you have to define what "taken off" means. I'd agree that Erlang hasn't taken off in the way (say) jQuery has, but the use-cases and audiences are pretty different.

There are many applications written in Erlang (ejabberd, RabbitMQ, CouchDB, SimpleDB) that are the fundamental component of larger apps - I'd say this is a "taking off" of sorts.

Aside from telecoms, where it's been running systems with over 99.99% uptime for decades now, Erlang is also used in production by 37signals and GitHub. Amazon SimpleDB contains Erlang components. Powerset, acquired by Microsoft to become one of the technologies behind Bing, built their load balancing application server (Fuzed) and distributed datastore (Dynomite) in Erlang.

So it depends on what you mean by "taken off". Plenty of large and small companies have been using it successfully for years. In the last two years the number has started growing rapidly.

I've written a little bit of erlang, so I can offer a few reasons why it isn't used widely.

Erlang does not have a string data type. All it has are lists of ints smaller than 256.

    1> [98, 114, 111, 107, 101, 110] == "broken".
    true

If statements are just wrong.

    if
    X == 1 ->
        io:format("you picked one\n");
    X == 2 ->
        io:format("you picked two\n")
    end

If X == 3 an exception is raised. You also can't call functions in if statements. This won't compile:

    if
    p(X) ->
        io:format("Predicate is true for X!\n");
    true ->
        ok
    end

The syntax is unfamiliar (e.g., {} is tuple) and also fugly. You've got 3 separate line terminators, namely ; , . all of which are used at different times.

Records are also broken. In most languages, x.field addresses the field of a record. In erlang, x#ClassName.field is necessary, since records are actually just a macro which compiles down to tuples.

I actually rather like erlang. But make no mistake: it's a domain specific language for message passing. It's not a general purpose programming language, even if it is Turing complete.

I think you are missing some of how Erlang views the world re:if statements. Erlang uses pattern matching rather than assignment, so the statements in the if are Guards.

From Pragmatic Programming in Erlang:

At least one of the guards in the if expression must evaluate to true; otherwise, an exception will be raised.

Often the ﬁnal guard in an if expression is the atom true, which guarantees that the last for m in the expression will be evaluated if all other guards have failed.

Perhaps my "just wrong" was an overstatement. I guess my real objection is that they should use a different keyword than "if". Like in ocaml:

    match x with
        [] -> 0 |
        _ -> 7

Same semantics, but it doesn't look like a C/Python if statement.

They could also use if, but require an else clause (like Haskell). Like it or not, syntax does matter to language adoption.

Reply on news.ycombinator.com