Fortran 77 making an array but changing the increment value

[1ytrewq]

Hi
I have an array going from 0 to n (where n has been previously defined) but I want the array to be in steps of 5 rather than increasing each value by 1

eg 0,5,10,15,20,25...n


how can I do this?

Thanks

 

[xts]

What makes you use F-77 ? As it name says it is obsolete by 34 years.

But if you really must do that in F77 - either don't care about memory (you have modern computer with GBs of it) and just use only 1 of 5, or be smart F77 prgrammer and write something like
REAL A(number of elements)
A(N/5+1) = whatever

 

[1ytrewq]

I'm starting a piece of working with an ancient supervisor and he only knows fortran 77 so I've had to start learning it too. Seems a completely ridiculous language from what I've learned so far

but thanks for your help

 

[JeffKoch]

Fortran is still commonly used for scientific applications, I use it regularly myself - in F77 flavor. Some co-workers use Basic. Newfangled languages aren't relevant to simple scientific number-crunching applications, the odds of screwing up your results far outweighs any advantages in efficiency, interfacing, sexiness, whatever.

For every element i in your original array [0,1,2,...,n], you can access it from your preferred array [0,5,10,...,5n] element j through i = (j+1)/5.

 

[xts]

(j+1)/5.

?

Either :
J/5 - if you number arrays starting from 0,
or
J/5+1 - if you are 11 years more obsolete and use F-66 (Fortran-4...) convention of numbering arrays from 1

I am not much younger (probably) than your suprevisor, but I switched to do number crunching in C++ over 20 years ago...

 

[JeffKoch]

Seems clear enough. (0+1)/5 = 0, (5+1)/5 = 5 , (10+1)/5 = 2, etc., in standard integer math, which lops off anything after a decimal point. So given j on [0,5,10,...], i on [0,1,2,..] follows the transformation i = (j+1)/5.

C++ in particular is a poor language choice, it's far too easy to make a mistake that won't be caught by a compiler or even crash your program, it will just give you the wrong answer.

?
Either :
J/5 - if you number arrays starting from 0,
or
J/5+1 - if you are 11 years more obsolete and use F-66 (Fortran-4...) convention of numbering arrays from 1

I am not much younger (probably) than your suprevisor, but I switched to do number crunching in C++ over 20 years ago...

 

[DuncanM]

The general format for a DO loop in FORTRAN is the following:

DO 10 I=1,N,INC
          . . .
          DO SOMETHING
           . . .
   10     CONTINUE

The control variable (I in this case) is assigned an initial value of 1.
N is the limiting value, so the program will break out of the loop after executing for I = N (basically, if I > N).
INC is the increment value. If INC is not given in a program, it defaults to 1. If you give it the value 5, I will increase by INC each time through the loop.

Hope this helps.

 

[xts]

Seems clear enough. (0+1)/5 = 0, (5+1)/5 = 5 , (10+1)/5 = 2,

(0)/5 = 0, (5)/5 = 5 (should be 1, I guess, mistype..) , (10)/5 = 2,
(0+3)/5 = 0, (5+3)/5 = 1, (10+3)/5 = 2,
(0+4)/5 = 0, (5+4)/5 = 1, (10+4)/5 = 2...
Why have you chosen +1 rather than +3?

C++ in particular is a poor language choice, it's far too easy to make a mistake that won't be caught by a compiler or even crash your program, it will just give you the wrong answer.

I am far from being an advocate of C++, it is awful language, encouraging you to make programming errors, but don't say that Fortran is safer!
We both are not quiche eaters!
(Real Programmers do not use Pascal, 1983: http://www.pbm.com/~lindahl/real.programmers.html)

ADDED>
Yes, Duncan, DO-loop is my favourite semantics for loop. Anyway, I am not advocating for

for (i=1; i<=n; i+=inc) {...}

- that is equally awful ;)

 

[JeffKoch]

Why +1 ?

Just good, safe habit - you're guaranteed to add some numbers after the decimal that will be lopped off, so you're always on the same side of the integer that you want.

We both are not quiche eaters!

And I even learned Pascal in college, though I've never actually used it except for that one class in like 1984.

 

[xts]

Well, Pascal is even more obsolete than F-77... But in 1984 it was much more elegant than F77... Actually - it is much more elegant than any modern language, but, of course, something a bit more geeky and efficient had to won...

Anyway: answer to OP is

REAL A(1:N)
DO 10 I=1, 5*N, 5
   do something to A(I/5+1)
10 CONTINUE

ADDED>
I just read the "Real Programmers..." again after 20 years... I forgot Fortran had had a very useful tool: Assigned-Go-To... I really miss that feature in C++ - if I only had it now it would really help me to avoid C++ programming pitfalls!

 

[rcgldr]

I forgot Fortran had had a very useful tool: Assigned-Go-To... I really miss that feature in C++.

C / C++ version of this is switch - case, you can even create a multiple entry point loop inside the switch, using one case for each entry point, called Duff's device:

http://en.wikipedia.org/wiki/Duff's_device

C / C++ also has pointers to functions, which is similar in concept to assigned goto's, using functions instead of code fragments. C++ extended this concept with overriding of member functions in structures or classes.

Getting back to the OP, if the array starts with 0, then normally it's indexes are described as going from 0 to (n-1), where n is the number of elements in the array. Previous posts already explained how to handle arrays that start at zero or start at one.