Pharyngula

Posted by: Dayv | May 16, 2006 12:39 PM

Is there a more clear version of this chart available?  This one looks like it's been resized without much respect for the type.

Posted by: Bronze Dog | May 16, 2006 1:14 PM

Think I might try making my own version tonight or something.

Posted by: JoeB | May 16, 2006 1:36 PM

I feel sort of gypped regarding my FORTRAN experience. I'm young enough that there were no punch cards associated with the experience, so all I got was the experience of a silly language with too many words.

That being said, flowcharting sucks. It is terrible, and I hate it.

Posted by: quork | May 16, 2006 2:26 PM

I don't see any connection to FORTRAN, which is quite good at FORmula TRANslation. Flowcharting outlines the process of an algorithm, it does not rely on the specific syntax of any particular programming language. You could use flowcharting with newer languages like C++ or Perl, and I know some people who should.

Posted by: PaulC | May 16, 2006 2:49 PM

The one good thing about Fortran is that it has complex numbers as an intrinsic type. You don't need them for a lot of applications, but when you do (e.g. for FFT) it's a pain to have to reinvent the wheel. In C++ you can at least create overloaded operators for them, but it's just weird that they get second-class status. Real numbers aren't even algebraically closed for crying out loud, and complex numbers, far from being esoteric, give you a natural implementation of planar transformations such as scaling, rotation, and translation.

Though my background is CS and I have a big preference for Java these days, I can understand why scientific programmers have held onto Fortran so tenaciously. It's the only programming language that seems to have been designed by someone who studied any continuous math.

Another thing I'd like to see is an intrinstic "int mod n" type that would work with +, -, *, and / but restrict values to the range 0 through n-1. Again, you don't need it that often, but when you do it beats pulling your hair out over the bogus not-really-a-mod-operator % that can return negative numbers.

Posted by: ekzept | May 16, 2006 6:50 PM

... looks like something that was done with an ancient dot-matrix printer

HEY! i still own, use, and love my dot-matrix printer. even bought a new one when the old one finally died. i also have an ink jet, but the dot-matrix is for random stuff and code. its cartridges are cheap. of course, it's an Epson.

Posted by: Rich | May 16, 2006 7:00 PM

Boo!

I go to all that effort and you all g33k out on Fortran and mock my crappy formatting. Listen. I'm building on the work of the Fig Newton of Information Theory, so I can clearly sell a few books based on it and possibly get DaveScott to do a blog for me if I'm lucky.

Just don't ask me to testify in court based in this!

*sulks*

Posted by: PaulC | May 16, 2006 7:05 PM

ekzept:

yeah, but if you do all your calculations in COMPLEX (whether it's in FORTRAN or not), roundoff and similar creep gives purely real numbers gradually larger imaginary parts, even if these parts begin at zero. hence the continuing need for REAL.

I think reals are nice to have around, for a variety of reasons, but it's not obvious that they should be the default.

Posted by: ulg | May 16, 2006 8:35 PM

Folks, this is a science blog. There is nothing scientific about tiresome old discussions of the relative (de)merits of various programming languages or various algorithm diagraming conventions.

Posted by: PaulC | May 16, 2006 8:57 PM

Oops, hybrid came out when I swear typed chimera. My bad.

Would it be fair to say that FORTRAN is more suited to analysis, while Lisp is more suited to algebra, Paul?

Depends on what you mean by analysis and by algebra. If you mean numerical analysis vs. symbolic algebra you might have a point (Macsyma was originally written in Lisp, wasn't it?)

But it's a false dichotomy. S-expressions are good. Complex numbers are good. Both ought to be available at your fingertips in the same programming language.

Posted by: ekzept | May 16, 2006 9:51 PM

Macsyma was originally written in Lisp, wasn't it?

yes, in Maclisp, same thing i wrote the program for my Master's in. see the origins of LISP-based computer algebra.

Folks, this is a science blog. There is nothing scientific about tiresome old discussions of the relative (de)merits of various programming languages or various algorithm diagraming conventions.

putting Wolfram's ideas aside for a moment (or forever AFAIC), most science is being heavily influenced by the availability of large scale and inexpensive computation, if not by ideas and models stolen from discussions and analyses of computation. where does science end and computation begin? what is the difference between mathematics and computation? there is one.

Posted by: archgoon | May 17, 2006 1:39 AM

PaulC wrote:

Another thing I'd like to see is an intrinstic "int mod n" type that would work with +, -, *, and / but restrict values to the range 0 through n-1. Again, you don't need it that often, but when you do it beats pulling your hair out over the bogus not-really-a-mod-operator % that can return negative numbers.

Actually, Python implements this (as well as including a complex number package). This is actually somewhat of a problem. I was working on a project and I prototyped it in python, and later moved it over to C. Since I was wrapping values around in an array, I was getting these seg faults, and I didn't know why.

Posted by: Roman Werpachowski | May 17, 2006 4:36 AM

You know.. The sad thing is, I wish I had some experience with Fortran.. It might help me figure out why the heck this:

Doesn't work when I try to convert it to VB (Heh, its the one I have a compiler for). Of course the other question is why the heck they are still using a Fortran version of it anyway anymore, when there are millions of hand held GPS units around that can do it using something more sane, like C++ or even Java.

1. there are free Fortran compilers available
2. people use Fortran becasue there are loads and loads of great numerics code written in it, available for free
3. Fortran will be surely more efficient than Java.

Posted by: Keith Douglas | May 17, 2006 9:55 AM

I escaped learning FORTRAN, though ML did screw with my head, so didn't manage to escape all oddball languages.

What I want is a hypercomputer and a suitable hyperprogramming language ...

Posted by: Torbjörn Larsson | May 17, 2006 11:21 AM

"Part of the input will be routed back to the input, precisely as Rich has hypothesed above.

It is of course nothing remarkable in that applying the new improved explanatory filter on the theory of ID itself yields the theory of ID."

Ouch! A little hasty there. I mean't to write: "Part of the output will be routed back to the input, precisely as Rich has hypothesed above.

This result is of course nothing remarkable, applying the new improved explanatory filter on the theory of ID itself should also yield the theory of ID."

Posted by: arensb | May 17, 2006 1:46 PM

The one good thing about Fortran is that it has complex numbers as an intrinsic type.

No, the one good thing about Fortran was the ability to redefine the number 2. That, and computed GOTOs.

Two good things! The two good things about Fortran are redefining integers, computed GOTOs, and self-modifying code.

Three! Three good things...

Posted by: PaulC | May 17, 2006 4:09 PM

arensb

Meh. So does C, these days, to say nothing of Python and a slew of other languages I'm forgetting.

As far as I know, Java does not. I haven't paid attention to ANSI C in years, but it didn't have a built in complex type back when I could have used one.

I may even be 10 years out of date on my whole assessment, but the inability to simply write infix +,-,*,/ expressions with complex variables has been an annoyance in the past.

I personally have not written Fortran code since college in the late 80s and don't have a lot of use for it. There was, however, a long span of history when it was understandable why such an archaic language was hanging on in numerical applications. If that is no longer the case, so much the better.

Posted by: ulg | May 18, 2006 2:58 PM

ekzept, RavenT, my earlier comment 'Folks, this is a science blog ...' was intended to be humorous on the grounds that (a) unscientific discussions are nonetheless important for communities like science blogs (as PaulC immediately noted), and (b) I am not a scientist myself! (But nobody challenged me on the latter element, so I don't get play a second round.) It's also sour grapes resulting from many bad experiences in other forums involving discussions on the relative merits of various programming languages. Such discussions usually start out with several people making sincere efforts to help a novice choose between a bewildering variety of available tools. Then someone thoughtlessly insults the enormous amount of effort it takes to master a useful programming language that happens to have some attribute they don't care for. Soon after, otherwise respectable students, computer scientists (including people whose peer-reviewed work was positively discussed in classes I took), and professional software developers all ended up acting like foul-mouthed schoolboys with web browsers to grind. I'm pleasantly surprised to find that my assumptions about how the discussion would evolve were wrong.

I'd try to post some kind of positive contribution, but I'm far too distracted by encounters with numerous HOPL and PPIG papers, weather forecasting and climate modeling software written in fortran, genome sequencing software written in perl, symbolic manipulation software written in lisp, condom-buffing software written in Dylan, web servers written C, web-browsers written in C++, etc, etc, etc.