Super Programming Languages

Jonathan Edwards, in The Myth of the Super Programming Language makes the extraordinary claim that language choice makes little difference.

There is no reason to believe that language choice makes a big difference. People have been doing studies for decades, and there are no reproducible strongly positive results. The burden of proof is on the poster, who is making a surprising claim.

To be fair, I believe he actually intended to single out Haskell and Lisp and state that it is impossible to get an order of magnitude increase in effectiveness. I’m not sure he is right there either, but I’ll dispute the actual claim he made.

Proof By Obviousness

First of all, there are studies that show that different languages required a different amount of code to implement a certain amount of functionality. Take a look at the language / function point here. Smalltalk (28 lines/fp) is more than 5 times better than C (148 lines/fp) in this metric. And comments about APL not-withstanding, code that you don’t write takes the least time.

Second of all, it is hardly a surprising claim. In the leap from Assembler to C, I’m obviously going to save programming time by not having to write code to push my parameters on to the stack every time I want a function.

Moving from C to Java I don’t have to write [so much] memory management code. Java to Perl, I don’t have to specify my data structures ahead of time.

Each language change, there is code that I just don’t have to write any more¹. How can that not make a difference?

I’m not smart enough to get Haskell. If I had to guess, I’d say that Jonathan isn’t either, and is mildly upset that folks can out-perform him by choosing these languages.

Perl is the most super programming language I know.

And just to demonstrate the amount of code I need to specify a simple data structure that I get for free in perl, here is some C++.

struct MyData {
    vector<string> list;
    map<string, string> dict;
};

MyData data;
if (data.dict.find("a") == data.dict.end()) {
    data.dict.insert(make_pair("a", "b"));
}
data.list.push_back("text");
// data.list.push_back(1.5);

The perl is clearly terser. Is it less understandable (assuming you know both languages)? No way – it’s clearer.

And I know I’ve been kinda mean with the heterogenous list.

my %data;
$data{dict}->{a} = 'b' unless exists $data{a};
push @{$data{list}}, 'text';
push @{$data{list}}, 1.5; #   

Choice Of Language Isn’t The Most Important Factor

As long as you don’t choose something crazy, like Intercal, or Assembler.

The important thing is that there are libraries that solve the majority of your problem. As many people have pointed out, a lot of development these days is gluing together libraries. For me, as much as 95% of the systems I write are third-party libraries.

And my favourite example of something I wrote that would have been much longer in either Perl or C++ is my Emacs Database Mode. Emacs Lisp isn’t a great language, although it is good enough, but for certain problems it has some amazing libraries.

---

1. Of course, there are trade-offs.

About these ads