K as a Prototyping Language

http://www.cs.nyu.edu/courses/fall02/G22.3033-007/kintro.html

by cubix • 18 years ago

42 25 18 years ago

cs.nyu.edu

tdoggette• 18 years ago

/ Right to left precedence:

3*2+5 / yields 21

Why should I learn this language that I've never heard of if it can't even do math in any sensible way? Not left-to-right, not order of operations, not lisp-like prefix. Who thought that was a good idea?

cousin_it2• 18 years ago

I'm lucky to know a little K. The language basically takes this one idea: instead of processing objects one by one with lambdas and the like, use bulk high-level operations on whole datastructures at once. And really runs with it. For example, it has an SQL DBMS written without loops, using only bulk vector operations.

prakash2• 18 years ago

do you work for a bank? where did you learn K?

halo• 18 years ago

My own personal opinion is that whatever language you're using (or, indeed, in mathematics) using redundant brackets is generally the best idea anyway.

Removes any ambiguity, easier to read, doesn't need you to remember precedence rules and there's negligible difference in execution speed.

eru2• 18 years ago

Try using redundant brackets in Lisp.

ajross3• 18 years ago

Isn't python (no brackets at all) an existence proof that you're wrong, though? The eye learns rapidly to ignore brackets. What it reads is indentation. And most editors will try to enforce indentation levels even without brackets present.

Or maybe you're just talking about parentheses in precedence grammars? I guess I have no objections there, but still: the fact that "multiplication has higher precedence than addition" is taught in grade schools and is used everywhere, in every language I know that supports precedence at all. What conceivable purpose could one have to buck this trend? All you gain is maybe a few lines of complexity in the parser (I mean, it's not like these things are hard to write) at the cost of a lifetime of newbie syntax errors.

stcredzero2• 18 years ago

Smalltalk also has different precedence rules. Assignment is highest, followed by unary ops, binary ops, then keyword messages. In some cases, this results in expressions evaluating left to right.

This is often the biggest complaint from outsiders. But it's easily solved by including parenthesis now and then. In 10 years of doing Smalltalk, I've never encountered a situation where this is ever an issue. I've never encountered a but caused by math expression precedence. Newb Smalltalkers might make the mistake, but they generally correct it in the next second. So much for the first impression of programmers who are unfamiliar with the language.

In 10 years of Smalltalk, I've also never encountered a situation where Smalltalk execution was slow -- it was always poorly designed (low-latency favoring in a high-latency environment) database IO. If anything, I've always been impressed with how fast the commercial VMs are.

My point: the "popular notions" about programming languages are often as misleading as urban legends.

cubix2• 18 years ago

Who thought that was a good idea?

Well, Ken Iverson did, and he won a Turing for his trouble.

kcl• 18 years ago

>Why should I learn this language that I've never heard of if it can't even do math in any sensible way? Not left-to-right, not order of operations, not lisp-like prefix. Who thought that was a good idea?

Of the three choices you mention only one makes sense as an alternative: left-to-right instead of right-to-left.

Order of operations, as it turns out, has nothing at all to do with real computation (in the same way that pi really should have been defined as 2pi). Most educated people, after a moment of reflection, should agree with this statement. Getting rid of order of operations makes the parsing process simpler, both for the compiler and for your head. If order of operations wasn't so ingrained in your head there would be no reason at all to use it. K makes a lot of design decisions like these: doing what's right at the expense of what's common.

I'm not sure where order of operations came from, probably as a nicety for mathematicians dealing with complicated polynomials, but when computers came about it must've been clear that it was the wrong way to do things.

Here's a code snippet that makes order of operations look just as silly:

  a*b+c%d%e+f*g^h*i*j*k-l / ugly with order of operations, hidden levels of complexity

Lisp prefix form, from a mathematical point of view, certainly makes more sense than order of operations. The reason it isn't used in K is, for one, K is inline, and two, Lisp syntax demands a lot more written code than K, and this goes against the philosophy of K. K wants to create the shortest programs in the shortest amount of time.

For example, something that is perfectly natural in K

  2+7%3*2+5 / k, 9 characters - 1 level deep

looks ridiculous in Lisp

  (+ 2 (% 7 (* 3 (+ 2 5))))
  ;Lisp, 25 characters - 8 spaces, 8 parentheses, 4 levels deep
  ;Two parentheses and one level of nesting for each operation

for the reason that Lisp is not as rabid about brevity as K is. (And how could it be. K is a much newer language.)

This leaves left-to-right evaluation. One argument against left-to-right evaluation is the following code:

f g

where f is any function. For instance

   g:3
   f:{1+x}
 
   f f g 
  5

and so on. But this presumes that you accept two axioms (1. you should read left-to-right as in English 2. functions should precede operands). I am currently of the opinion that left-to-right evaluation is better given a left-to-right read order

  g f f / 5, fictional example

because I've yet to discover sound reasoning why mathematicians believe the function should precede the operand. (The answer may be that there isn't a reason, and mathematicians did it that way only because they never cared enough to pay close attention to syntax. Or perhaps it's because written mathematics "executes" in English, and not in an interpreter.)

The language designer has defended right-to-left evaluation with the following statements. "Why right-to-left?"

  so that it reads left to right.
  note k executes left of right, e.g. syracuse operation: 1+3*
  the basic math notation is:
  f g h x
  executes right to left but reads
  f of g of h of x

I remain unconvinced, especially since there are benefits to left-to-right execution (presuming left-to-right read order) including on-the-fly processing. It just seems more natural to me, mathematically, though I admit it would be very unnatural for me to write in such a way.

K is a beautiful language with a lot of nice features. Anything truly innovative is going to look new and different, and in K you get the statement

  3*2+5 / yields 21

There are a lot of unusual looking things in K. There are also lots of very powerful ones. But these won't reveal themselves until you spend a time looking at the language. So my suggestion is if you are intrigued by the powerful stuff, keep looking, and don't be turned off by the strangeness of the simple stuff. In some cases you may find that your opinion changes on how things should be done. There is a lot of wisdom hidden in K.

42flicks• 18 years ago

Very interesting, thanks for posting the link.

I'm also interested in what industries this language perfoms well.

cubix2• 18 years ago

From what I gather, it's used mainly by quantitative analysts in finance (so called "quants"), but I'm sure it could be applied to a wide variety of other fields.

dats• 18 years ago

Last year someone benchmarked a simplified raytracer in a few different languages--primarily to show off OCaml. Later, someone wrote the raytracer in 7 lines of K. http://www.nsl.com/k/ray/ray.k. The original comparison: http://www.ffconsultancy.com/languages/ray_tracer/index.html

ajross3• 18 years ago

Nifty reference, thanks. You absolutely have to read the description in the original page though. The K looks pretty much line line noise; I've seen raw sendmail scripts that look cleaner. As terse as the language might be, a little whitespace and naming conventions might still be a good idea for us newbies...

gruseom2• 18 years ago

a little whitespace and naming conventions might still be a good idea for us newbies

The culture around K is fanatical about concision. This goes back to APL, though the K people (following the lead of Arthur Whitney, the designer of K and Q) take it further than anybody else. To the unfamiliar, K looks like a bad joke (check out the 40 lines of code at http://www.nsl.com/k/s/s.k, which implement the spreadsheet at http://www.nsl.com/papers/spreadsheet.htm), but these people know what they're doing. Whitney's product includes the language interpreter, a massively scalable RDBMS, a web server, and God knows what else, and the last time I downloaded it, the total size was 380 KB. My experience with that stuff was that the more I looked at it, the more in awe I became. It's a shame it's not more widely available.

ahold• 18 years ago

Is it "K" or "Q" language?

Kaizyn• 18 years ago

It's K, a successor to the J programming language.

deepster• 18 years ago

Soon to be replaced by the L language.

gruseom2• 18 years ago

Not so much a successor as a sibling rival. I've always assumed the name was a bit of (typically cryptic) humor.

dangoldin• 18 years ago

If anyone is interested in some examples, take a look at projecteuler.net - it's a series of mathematical/programming puzzles.

After you solve a particular problem you gain access to a forum where people detail their solution. A good amount of the solutions tend to be written in these APL languages (J, K) and may serve as a nice way to learn.

K as a Prototyping Language

Related Stories