stackoverflow needs tweaking

stackoverflow is a wonderful site. It's the first place I go when I have a question related to programming.

Unfortunately, I think it needs a few tweaks. My sample may be small, but lately (the last few months) I've been seeing too many cases where there was a cry of "Duplicate! Duplicate! Duplicate!" or even closing the question. That wouldn't be a bad thing if Question A was, 'What does "Error: Variable not defined" mean?" and Question B was, 'I got this error "Error: Variable not defined". What does it mean?"

The problem is that in many cases those voting to close are trolling rather than looking out for the site. In some cases it is probably just trolling by the immature, with too much time on their hands and too little common sense, but in others it is regular users having fun with power.

Here is one good example. The OP lists a couple of things he's been doing to debug his programs and asks for other suggestions. If that's not a perfect question for stackoverflow, nothing is.

Yet it was closed on the grounds that "We expect answers to generally involve facts, references, or specific expertise; this question will likely solicit opinion, debate, arguments, polling, or extended discussion." Good chance that the trolls voting to close have never written a line of R code.

Here is another example. It was reopened, but as user Dan Burton wrote in his comment, "I checked the profiles of the 5 people that closed this, I'm rather appaled that none of them have made a single contribution to the scala tag."

By the standard that seems to be in place right now, even the question I started with, where the user gives an error message, should be closed. There are multiple reasons you might get a certain error message. That leads to opinions rather than facts, because you'd have to make a judgement about the most likely reasons for a certain error to occur in a given language. It might also be interpreted as a question not about a programming language, but rather what is wrong with the programmer's training that would make him arrive at that error message. (If you think I'm stretching things, dig around on stackoverflow and look at the closed questions.)

My goal, though, is not to talk about these specific cases. Maybe they are terrible examples - whatever. My point is that the same thing happens again and again, and it is creating a hostile environment, particularly for new programmers. It used to be that I could recommend stackoverflow for someone learning a language, but I've stopped. It is very discouraging to ask what you think is a reasonable question and then be subjected to a public flogging. (Even when the question is not closed, the answers are more harsh than a couple of years ago.)

Traditional trolling is bad enough, but at least you can delete or ignore those comments and get on with your business. This new-fangled trolling puts your own posts on the line, and the "closed" message is an official "you don't belong here" message, sent by stackoverflow.com, that the whole world can see. That's kind of hard to ignore, particularly if you used your real name.

Maybe one solution would be banning the close trolls for a week if they vote to close two questions in a 30-day period that are then reopened. Cries of duplication are not as easy to monitor, but they're more in line with traditional trolling, so I don't know that it is as much of a concern. Whatever the solution is, something has to be done to make the site friendly again.

Edit: Just came across this gem. Somebody named Michael Petrotta explained the reason for his vote to close: "@Jay: I was one of the people who voted to close. I thought you might be interested in my reasons. I don't program in C++, and I don't feel much, one way or the other, for the language or its adherents. I read your "question", and saw an attack on a project, with a lot of nasty name-calling. I read the answers so far, and saw a bunch of defensive, content-light replies. I didn't see any value in the question or its answers to date, and I voted to close."

So you've got someone who admits he knows nothing about C++, doesn't care about C++, and he's voting to close the discussion? And he's allowed to continue trolling?

Use Counterclockwise or Clooj if you want to start with Clojure

It appears that I didn't do a good job of writing my previous post about Emacs and Clojure. I was trying to give a flavor of what someone new to Clojure would think when encountering "Emacs + Clojure is the best development environment". Seriously, getting it set up sucks. Really bad. I speak from experience, having tried to get someone set up using Clojure and Emacs recently.

What you should do is use Counterclockwise. It's wonderful. It's easy to set up. It has lots of features. It's a joy to program in Clojure when using Counterclockwise. That's what my friend is now using.

The other alternative is Clooj. That is truly the easiest way to get going. Just download and run. The disadvantage is that Swing apps don't look that great on Linux. I've gone through all the guides but the fonts still aren't what I want. To my knowledge it doesn't work with dual monitors. But if you want to get started quickly, you cannot possibly do better than Clooj.

Note: I'm not saying you shouldn't use anything else, just that these are awesome for beginners, and others may or may not be.

Is this a good introduction to Clojure?

So I wanted to help someone get started with Clojure today. For those who haven't done much with Java before, the incredible overhead associated with doing the most trivial tasks leads to a bad impression.

To make matters worse, I was trying to set up Emacs for use with Clojure. I'm not an Emacs noob. I used ESS as my main development environment several years ago, but was not impressed, and moved on.

[Edit: Used bold font to emphasize that I'm writing from the perspective of someone who decided he'd try Clojure using Emacs. My personal recommendation is to use Counterclockwise or Clooj.]

This is what someone new to Clojure would experience.

Go to the main Clojure page instructions.

"Install clojure-mode."

Don't know what that means, but I'll click the link.

"If you use package.el, you can install with M-x package-install clojure-mode. Otherwise you can do a manual install by downloading clojure-mode.el and placing it in the ~/.emacs.d/ directory, creating it if it doesn't exist. Then add this to the file ~/.emacs.d/init.el:"

Well, not being a total Emacs noob, I tried M-x package-install clojure mode.  I got [No Match] in return.

What is clojure-mode.el? Where do I download it? Google found it for me. I downloaded it to the ~/.emacs.d directory. I added the lines to init.el (though I had to create that file, something the instructions didn't mention).

The vast majority of new Clojure users would have given up by now. Users should not have to figure things out for themselves - that is the purpose of documentation.

So let's move on. There's a discussion of how to set up package.el. Okay, that doesn't apply to me.

Then there's a discussion of clojure-test-mode. It says "This source repository also includes clojure-test-mode.el, which provides support for running Clojure tests (using the clojure.test framework) via SLIME and seeing feedback in the test buffer about which tests failed or errored. The installation instructions above should work for clojure-test-mode as well."

So do I need it? Is it something that most users install? I'm going to skip it. I can do tests when I get to that point.

Then there's something about paredit, which I hate, but it's recommended for all users.

Following that, there's a section titled "Basic REPL", which says, "
Use M-x run-lisp to open a simple REPL subprocess using Leiningen. Once that has opened, you can use C-c C-r to evaluate the region or C-c C-l to load the whole file.
If you don't use Leiningen, you can set inferior-lisp-program to a different REPL command"

The only way you'd have any idea what that means is if you knew beforehand. That may have set a record for the least informative paragraph ever written. I get tired when I read something like that. What is meant by "different REPL command"? Heck with it, I'm returning to the main page. {Note: this is explained on the main page, but the reader isn't told that.}


I created a new clojure file in Emacs (ending with .clj). There's no automatic indentation! How on earth do you write Lisp code without automatic indentation?

Eventually I got it sort of working, though I never got Clojure 1.3.0 to work. This is not a pretty introduction to Clojure. {I'm not saying it's my introduction to Clojure. It is the introduction to Clojure for anyone who reads Emacs + Clojure is the best alternative.} There's a lot of room for improvement.

I recommend immediately eliminating all references to Emacs when talking about Clojure, with an exception for "Don't waste your time with Emacs if you want to learn Clojure."

The book the world needs

The world needs "Introduction to Functional Programming Using Clojure".

This is needed for at least two reasons. First, the existing books on Clojure require an understanding of basic functional programming. All of the currently available books do a good job of explaining the language. For the average Java or Python programmer, though, it would be tough to learn Clojure if the only resource were one of those books (or all of them, for that matter).

Second, "introductions" to functional programming are usually too academic. SICP, for as much as I love that book, is a book about computer science, not software development. Scheme will never, ever be a programming language embraced by industry. That doesn't mean there's anything wrong with Scheme, but it does mean that if you use Scheme, the perception will be that functional programming is not something you'd do with a "real" language, so I guess that is something wrong with Scheme.

I've been reading about Standard ML recently, and while I have learned a lot, I can't imagine anyone taking ML (SML or OCaml) seriously. I mean, in the sense that the perception would be any better than the perception of Scheme. Moreover, I've tried to read about Haskell, but it's hard to go more than about five minutes without falling asleep. There's just too much religion about pure functional programming. I can see why it appeals to mathematicians, and I can see the point of Steve Yegge's post on Haskell.

Even worse, there's discussion of type systems, which is an advanced topic. It's my belief that dynamic typing is better for beginners. Don't get me wrong: on technical grounds, the ML and Haskell crowds might be right. I'm not sufficiently informed to take sides in the debate. It's just a matter of perception, and ML and Haskell are non-starters IMO.

Clojure's not-so-secret weapon is the JVM. It can run any Java libraries, and thus it has instant credibility. Unlike Scala, it has dynamic typing. It has the coolness factor, too, because of all the concurrency-related stuff that makes it appear to be cutting edge. The language has an outstanding community with members that have - get this - people skills. The creator gives interesting talks. He is happy to incorporate academic ideas, yet comes at everything from a practical perspective, based on years of real world software development.

Perception matters. Clojure may be what those of us who believe in functional programming have been waiting for, in terms of acceptance. The fans of Scheme, Common Lisp, Haskell, ML, and all other functional languages should be happy about this. Clojure is the first functional programming language that doesn't have to worry about the perception problem. Once a functional programming language, as opposed to functional features in non-functional languages, gains acceptance, developers will ask "Which functional language should I use?" Then the door is open for all the other languages.

So back to the book. I promise I'll preorder it when it becomes available. You can even use my title.