Hacker News new | comments | show | ask | jobs | submit login
Perl is no more accurate than randomly-generated language (neverworkintheory.org)
131 points by j_baker 2099 days ago | hide | past | web | 53 comments | favorite



I’ll point out what a dozen others are probably typing right now: what works for beginners is not necessarily what’s best for experts.

An analogy: I don’t know how to read sheet music with any proficiency. It would be easier for me if note duration were shown by horizontal length instead of the system of tails and open/filled circles. However, this wouldn’t be as good for experts, who care more about fingering than about being able to see the melody in the most literal way.

I don’t happen to like perl’s syntax, but this study by itself isn’t proof that it’s objectively bad for the people who use it heavily.


"what works for beginners is not necessarily what’s best for experts.... this study by itself isn’t proof that it’s objectively bad for the people who use it heavily."

More than that. Three languages were chosen. One was based on what previous studies had shown works, and one sounds like it was basically a ROT13'ed perl (only with symbols), and from what some were saying on Reddit it was an overcomplicated dialect of perl (though IMHO that would change little). Then, in a study in which they pitted Perl against a language designed to do well in studies, the language designed to do well in studies prevailed.

Even though I do question whether there may be a wee whiff of tautology about this whole exercise, this is interesting because it shows that studies may be able to distinguish between the features of two languages w.r.t. how easily they can be taught and learned. This is valuable stuff, millions upon millions of dollars to society's worth of valuable stuff. What it says about Perl is simply that it was not designed to do well in such studies, quelle surprise, and in particular it doesn't say anything at all about Perl relative to any other real language. $YOUR_FAVORITE_LANGUAGE may well have fared worse, and $YOUR_MOST_HATED_LANGUAGE may well have fared much better.

It is a true shame that the underlying valuable message of the study is being missed by so many people in a tearing hurry to go "hurr hurr, perl t3h suckz, hurr hurr".


Then, in a study in which they pitted Perl against a language designed to do well in studies, the language designed to do well in studies prevailed.

No, it's not surprising that perl performed poorer than Quorum. The surprise is that perl did as poorly as Randomo, which should not be sensible given that the operators and keywords are gibberish.

I generally agree with what you're saying, but the point of the study is not that the well-designed language did better. Who wouldn't expect that? But there is no tautological question when you realize that the result they are focusing on is the near-even performance with Randomo, not the failure to exceed or match Quorum.

It's pretty clear that they aren't engaging in perl bashing, but trying to provide data that backs up their experiential findings as teachers.


From what I've understood, Randomo and Quorum's syntaxes share the same structure. Only keywords and operators changed. Therefore, Randomo has a much better structure than what we could expect from a truly randomly generated language.

So, while it still looks pretty bad for Perl, I'd say it's not that bad.


It doesn't look bad for Perl at all, because the entire setup of this study is flawed to the point of making the conclusion a non sequitur.


It always amazes me when people can dismiss academic papers -- the result of (at least) several weeks of work -- with a glib comment uttered anonymously, and no support given.

Have you read the study? If not, you can find it here: http://www.cs.siue.edu/~astefik/papers/StefikPlateau2011.pdf


I have read the study which is why i made that comment. In fact, i've even contacted the author to gain more details about how they selected code samples to serve as learning aids and how all the other code samples looked like.

Additionally, i am not anonymous. Google my name and you will find all about me, including, if you look for longer than a minute, my full name and address.

Lastly, there was no need to give support in that comment by repeating what others and myself have said copiously all over the comments here. This was my main post towards that. Feel free to argue it: http://news.ycombinator.com/item?id=3153249


What it says about Perl is simply that it was not designed to do well in such studies...

I disagree. If it says anything, it's that if you give bad code to people who don't know anything about programming, they'll have trouble writing good code.


If you give a noob the task of writing a program in a language the noob has never seen before it makes sense the noob would write code that looks like random gibberish to a Perl programmer. Therefore, it's not surprising the random gibberish code would would resemble "a language design by chance" more than it would look like Perl.


The point he was making wasn't about the output, but about the fact that the learning material given to the test subjects was the worst perl imaginable.


There were 18 frigging participants to the study, for chrissake! You can't get any meaningful statistical data from such a ridiculously small sample!

The methodology may be sound, they may be good students and deserve a good grade, but there are absolutely zero conclusion to draw from this paper except for the fact that it may be valuable to make a larger study.


These are teachers teaching software. They like the study because it somewhat confirm their experience (switching to python makes learning significantly faster).

It is not The Final Proof (against what? perl?), but it is nice to find one way, among maybe other, to give some ground to empirical teaching experience.

I don't know much of perl, but I tried bash, and know PHP well, and really, really, the language used does matter, it will shape your models and logic, and change the way you handle, debug, write, read, maintain code.

Some people seem to adhere to the relativist cliché: all languages are good for something, in their field of usage. It is a matter of personal taste (implying it doesn't matter at all).

I don't agree. All languages are not equal. The breadth or their application field is often not comparable (eg compare bash, which is useful for some unix scripting tasks, with python, which can do web, and math, and stats, and windows grey-hacking, and multiplatform desktop clients, and image processing, and Unix scripting, etc.)

Having worked with (and fought against) PHP for many years and recently switched away, I can tell, it is not a matter of taste, some languages are simply worse, which implies that some are better. There is no "equal rights" to enforce in this domain.

[edit: clarify]


I first learned programming writing BBC BASIC; perl is now my weapon of choice but I'm not sure I would have liked it nearly as much had I learned it first.

Interestingly (perhaps) I repeatedly hear from people with language/humanities background that they tried both python and perl and found perl much, much easier to start to get things done with - the more natural-language-ish aesthetic of perl being a strong contributor to that (I'm a pure maths grad but my parents are both Eng. Lit. grads, so I guess I sort of count in the "humanities background" category).

On the other hand, people with non-humanities backgrounds almost always found python easier to start to get things done with, even if they later found perl to be their preference.

Above comments based purely on anecdata; take them for what you will.


In the case of sheet music it's more about adapting yourself to the standard, so that when you write down music and some random guy in Russia writes down music you can both understand what you wrote without difficulty. (Unless you're that classical asshole who does C sharp sharp).

Most jazz musicians I have talked to and taken clinics from really don't like sheet music though, as it is an ineffective and inconcise way to represent a tune's structure and melody, hence why they prefer charts. Charts really are worlds ahead of sheet music in my opinion.


Double-sharps are actually a great example of syntax for experts being inappropriate for novices. The double-sharp on a note indicates that that particular note is part of an upward motion. Writing the same note as the equivalent natural would instead indicate downward motion (assuming a sharp key). Knowing the intended motion of a note can influence phrasing and help a reader predict the next note of the melody without looking ahead.

Of course when you start out reading notes, double-sharps are confusing as heck.


Haha, yeah I shouldn't of been so harsh with my mentioning of double-sharps.

I do still think that in the end people are just conforming to the standard to make life much much more easier, and that it isn't so much that certain things are better for the expert then the novice, but that experts are more used to the standards than a novice is.

I feel it's more of a matter of personal style than a programming language x being more efficient when you become an expert when compared to language y. Some people really fly with Perl, whereas other people are more of a Python kind of person.


Wait a minute: the amount of information in a typical Jazz chart is vastly inferior compared to the information typically found in a score for a classical piece. This is because in Jazz, much more is left to the discretion of the musicians. So, it is quite difficult to correctly compare the two, since they don't have the same uses.

About the double sharp, it often has to do with tonality. Classical music is typically more tonal than jazz. The sharp is needed when you temporarily change tonality, and some tonalities just need double sharps then and there –Gsharp dur for instance. If you wrote the natural note instead, it would not convey the intended tonality.


That's exactly my point; programming languages aren't about x is better then y, but they are about personal style (and what they are built for). Other people really take off with x language but not y, and vice-versa.

I was also unnecessarily harsh towards double-sharps as well, when I was learning piano I would sight read a lot of easier pop/rock sheet music, and there were a few pieces I encountered with an unnecessary double-sharp that would catch me off guard because it was uncommon for the style.


Can you provide an example of what you mean re. the difference between sheet music and a jazz chart? I have played classical and jazz winds for years and consider them almost synonymous. (You probably won't find "changes" in the kind of sheet music other than that musicians call a "chart.")

I don't feel that any written music can or could adequately convey the feel, shape, or nature of the music--you've got to work through it with the group, or hear someone who already understands the piece.


Most jazz charts I have used just have the chord structure and the melody all on one page, whereas most classical charts I have seen (and the very very few I have used) have a lot of very intense information, and I understand that they have that much information because that is the style of classical music, but I personally very much prefer concise jazz charts.


I don't think anyone (the paper's authors included) would claim that this is objective proof that Perl is bad. I personally share your concerns. But wouldn't you at least agree that it's an interesting concept to try to test?


Oh, I certainly don’t begrudge them running the experiment. I just wanted to put a check on the excessively anti-Perl interpretations that I thought were suggested by the title.

A more interesting test to me would be something like seeing how fast experts in a number of languages could extend or find a bug in code by other experts. I think there’s been a little research on this, but I don’t know enough to link to the best of it.

As I said, I dislike Perl’s syntax. But I think that, on the available evidence, it makes sense to see that an an opinion on the level of disliking cilantro, not on the level of disliking geocentric cosmology.


I've played music for years and read sheet music well, but the first time I saw Singstar it was a huge wtf moment.


Perl was designed around two slogans: "the UNIX swiss-army chainsaw" and "there's more than one way to do it" (for any value of "it").

I first met Perl after spending some years working in the UNIX command line environment, and it made intuitive sense ... to someone who primarily used awk, sed, and bash (and able to get stuff done in C). I was up to speed and productive in it within a couple of days. (This was in the days of Perl 4 -- the OO stuff in Perl 5 came years later.) At that level, Perl was basically every UNIX text-handling tool rolled into one handy bundle oriented around script execution rather than interactive shell work, with a bunch of POSIX extras and things like the socket() API thrown in to take it beyond the toy language level.

The point is, Perl isn't a beginner's language. Rather, it's an elaboration of a pre-existing toolset that assumes you're already familiar with the UNIX way before you get started. Pitching it against a teaching language isn't even an apples/oranges comparison: it's like comparing a circus merry-go-round to a helicopter (they're simply not intended for the same purpose or the same user demographic).


"Perl is, in intent, a cleaned up and summarized version of that wonderful semi-natural language known as 'Unix'."

Larry Wall, http://groups.google.com/group/comp.lang.misc/msg/1ebc89bbc0...


The language was not designed by chance. Some of the syntax was designed by chance. “Some of the syntax” is a very tiny part of a language.

While this is interesting, presenting it as a “randomly-generated language” is extremely misleading.


Fair point. They can't, for example, randomly generate language semantics (how the hell would that work? :)

All in all, though, this paper goes much further in trying to derive language usability than any I've seen. And it does seem disturbing that randomly selected characters (even for a subset of the language) should fare as well as Perl's hand-picked syntax.


This study is entirely useless because they invalidated themselves from the start by providing the worst possible Perl examples they could. I quote the salient parts:

"participants were given a code sample worksheet for the particular language group they were in. The general idea of the experiment is to give novice users code samples similarly to if a participant was learning to program from home on the Internet.

This code shows one of the code samples provided to participants."

    $x = &z(1, 100, 3);

    sub z{
        $a = $ [0];
        $b = $ [1];
        $c = $ [2];
        $d = 0.0;
        $e = 0.0;
        for ($i = $a; $i <= $b; $i++){
            if ($i % $c == 0) {
                $d = $d + 1;
            }
            else {
                $e = $e + 1;
            }
        }
        if ($d > $e) {
            $d;
        }
        else {
            $e;
        }
    }
This ridiculous thing was supposed to help their test subjects learn Perl.

If all their perl samples looked like that it should not be a surprise to anyone that their test subjects were entirely confounded. What is missing in that paper is a description of how they actually sourced their code samples, as well as a list of all code samples provided.


Misleading title.

s/no more accurate/no easier for beginnners to learn/


Yes, upon reading the obvious-flamebait title I asked myself, "What does it mean for a language to be accurate?", and "How could something randomly generated possibly be more accurate than something designed?". I did not expect my questions to be answered by the article, and in that I was not disappointed.


I didn't intend the title to be flame bait. I only wanted to get to the gist of the matter as tersely as possible, which isn't easy. What would you have titled this?

EDIT: And why the downvote? I would actually be interested to hear a better title.


"Perl's syntax is as intuitive to beginners as that of another language whose tokens and keywords were randomly-generated" is verbose, but accurate. The structure of Randomo was entirely based on Quorum, with certain tokens and keywords randomized. The study really only speaks to the "intuitiveness" of the most superficial parts of those languages' syntax, which is something that Quorum happens to focus on:

    action z(integer a, integer b,
    integer c) returns number
      number d = 0.0
      number e = 0.0
      integer i = a

      repeat b - a times
        if i mod c = 0 then
          d = d + 1
        end
      else then
        e = e + 1
      end

      i = i + 1
    end
edit: I realize the title I gave isn't short enough to post on HN, but I wanted to reply just to make a point about the study, not your title.


Actually Quorum looks much more verbose than Perl. That may be the main distinctive feature helping absolute newcomers to grasp the code. In that vein, Java probably would fare way better than most terser languages, too. This would be something to test for (on a significant number of guinea pigs preferably).


What an awful title.

What a silly thing to test in isolation; does the syntax of a programming language present a greater barrier to a novice than the semantics of programming?

A better study would have controlled for the ability of the eighteen participants to describe their algorithms in pseudocode.


It's not silly if it's significant.


If you can't isolate that sole characteristic, your findings aren't empirically significant.


The provided perl is a little dated:

  $x = &z(1, 100, 3);
  
  sub z{
    $a = $_[0];
    $b = $_[1];
    $c = $_[2];
    $d = 0.0;
    $e = 0.0;
  
    for ($i = $a; $i <= $b; $i++){
      if($i % $c == 0) {
        $d = $d + 1;
      }
      else {
        $e = $e + 1;
      }
    }

    if($d > $e) {
      $d;
    }
    else {
      $e;
    }
  }
Fixed?

  $x = z(1, 100, 3);
  
  sub z{
    my( $a, $b, $c ) = @_;
    my $d = 0;
    my $e = 0;
  
    for ($a..$b){
      if($_ % $c == 0) {
        $d++;
      }
      else {
        $e++;
      }
    }
  
    if($d > $e) {
      return $d;
    }
    else {
      return $e;
    }
  }


This also clearly reminds me of one of the uses of perl. It's nice to use for prototyping something that can be implemented in ANSI C when necessary.


"While they spend their time explaining how Perl can't be used for productive work, we do commit even more code to CPAN and make our sh$t just work by the end of the day."


The whole experiment is based on learning from other code samples (see http://www.cs.siue.edu/~astefik/papers/StefikPlateau2011.pdf). That's what they measured -- how easy is for the beginner to learn from the code samples.

If you don't explain the logic behind the special characters, of course it's easier to learn what the code is about from the notation which has less special characters and less semantic behind the special characters, like their better faring language:

  action z(integer a, integer b,
  integer c) returns number
  number d = 0.0
  number e = 0.0
  integer i = a
  repeat b - a times
  if i mod c = 0 then
I'd say the students were talented if they managed to make reasonable results in any language, from my perspective, their "better" language is still way too hard for absolute beginners.


You're absolutely right.

I blame the title for turning this into Perl sucks vs. no it doesn't discussion. Not going into the problems with their study (N=18, poor Perl samples,...), all they prove is that a novice will much better understand pseudo code designed to be easy to understand than a real programming language.

I'm a Perl hacker and I would be the last to claim that it has an easy to understand syntax. Perl is full of shortcuts and magic that is hard for beginners, but it makes my life as a programmer much more easier. If they compare their pseudo language with Perl, Ruby, and Python, Perl would be at the bottom of the list.


This shouldn't be surprising. Even experts get surprised by what perl's syntax actually does [1].

[1] http://blog.plover.com//prs/accidental-syntax.html


N = 18


There seems to be a number of "studies" with a ridiculously low number of participants (here, 18), from which "scientists" (actually, students) try to draw conclusion from quasi-random, statistically insignificant results.

Thousands of people learned programming using perl. This is much more significant than this ridiculous study (randomo language? quorum? give me a break would you?).

Fortunately, at least this poor-piss stuff won't make the headlines, like too many such ridiculous studies run on 30 students do (remember lately "women wearing make-up appear more competent", augh).


I have spent years upon years programming in perl, and I love it dearly, but I laughed pretty hard at this. No offense to Larry Wall, but if you've spent any time writing perl code, surely you've had at least one moment where you wondered if someone hadn't gotten one over on you. I don't knw about relative ease of learning, but perl can be obscure. Of course, some think that adds to the fun. ;)


Suddenly I'd like to know how other languages fare. I have a feeling they might not be much better.


Of the current suite of top 20 languages at TIOBE (which for all its faults is as good as any other top-20 list), the ones that I know of that have heavy pedagogical backgrounds (that is, designed to teach students, not merely used to teach students because they were already popular) would be Python, Visual Basic, and Pascal, with Python's background probably being the most recent with its ABC predecessor, which some Googling suggests dates to the late 1980s, at which point the other two were already heading down the "industrial strength" path. Those would be the most interesting to test.

(If you're in a hurry to mention Scheme, that's #27. And on the list of "pedagogical languages" it would still be beaten by Logo at #23, which, frankly, is a sign we probably shouldn't trust the list at all below about #20....)


You may be in luck. The paper says:

"Given this interesting first result, we plan to test a number of additional languages using the same procedures."

Hopefully they'll use larger sample sizes too. This one involved 18 people across 3 groups.


If they use the same procedure the result will be completely useless.


In other words: When a beginner writes code, the programming language does not matter.


This title is misleading to the point of being nonsensical. Attributing the inaccuracy to Perl itself is just plain weird. It's the people programming in Perl who failed to be accurate in their code — it's not like they wrote correct code and Perl (or Randomo for that matter) moved things around on them. But that doesn't get as much of a guffaw out of Perl-haters, I suppose.


Unless I missed something, I don't think being easy for novices to use was one of Larry's design goals for Perl. Nevertheless, I found Perl pretty easy to learn, and it's the only language I ever "stuck with" long enough to build complicated programs.


Shall we compare new users of TextEdit to new users of vi?




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: