Surprising behaviour of Fortran (90/95)

Do, 08/01/2013 - 14:28 — Draketo

1 Introduction

I recently started really learning Fortran (as opposed to just dabbling with existing code until it did what I wanted it to).

Here I document the surprises I found along the way.

If you want a quick start into Fortran, I’d suggest to begin with the tutorial Writing a commandline tool in Fortran and then to come back here to get the corner cases right.

As reference: I come from Python, C++ and Lisp, and I actually started to like Fortran while learning it. So the horror-stories I heard while studying were mostly proven wrong. I uploaded the complete code as base60.f90.

1. Introduction
2. Testing Skelleton
- 2.1. Helpers
3. Surprises

2 Testing Skelleton

This is a code sample for calculating a base60 value from an integer.

The surprises are taken out of the program and marked with double angle brackets («surprise»). They are documented in the chapter Surprises.

base60-program: Base60 encoder, program part. For reference see Tantek Çelik's Newbase60 In Python And Java and NewBase60.

program base60
  ! first step: Base60 encode. 
  ! reference: http://faruk.akgul.org/blog/tantek-celiks-newbase60-in-python-and-java/
  ! 5000 should be 1PL
  implicit none
  <<declare-function-type-program>>
  <<function-test-calls>>
end program base60

base60-function: Base60 encoder, function definition (in the same file)

<<declare-function-type-function>>
  implicit none
  !!! preparation
  <<unchanged-argument>>
  <<parameter>>
  ! work variables
  integer :: n = 0
  integer :: remainder = 0
  ! result
  <<variable-declare-init>>
  ! actual algorithm
  if (number == 0) then
     <<return>>
  end if
  ! calculate the base60 string
  <<variable-reset>>
  n = number ! the input argument: that should be safe to use.
  ! catch number = 0
  do while(n > 0)
     remainder = mod(n, 60)
     n = n/60
     <<indizes-start-at-1>>
     ! write(*,*) number, remainder, n
  end do
<<return-end>>

2.1 Helpers

function-test-calls: Function test calls to see the results.

write(*,*) 0, trim(numtosxg(0))
write(*,*) 100000, trim(numtosxg(100000))
write(*,*) 1, trim(numtosxg(1))
write(*,*) 2, trim(numtosxg(2))
write(*,*) 60, trim(numtosxg(60))
write(*,*) 59, trim(numtosxg(59))

3 Surprises

3.1 I have to declare the return type of a function in the main program and in the function

declare-function-type-program: Declare the return type of the function in the program

! I have to declare the return type of the function in the main program, too.
character(len=1000) :: numtosxg

declare-function-type-function: Declare the return type of the function in the function definition again

character(len=1000) function numtosxg( number )

Alternatively to declaring the function in its header, I can also declare its return type in the declaration block inside the function body:

Alternative: Declare the return type of a function inside the function

function numtosxg (number)
  character(len=1000) :: numtosxg
end function numtosxg

3.2 Variables in Functions accumulate over several function calls

This even happens, when I initialize the variable when I declare it:

variable-declare-init: Declare and initialize the variable.

character(len=1000) :: res = ""

Due to that I have to begin the algorithm with resetting the required variable.

variable-reset: reset the variable at each run of the algorithm.

res = " " ! I have to explicitely set res to " ", otherwise it
          ! accumulates the prior results!

This provides a hint that initialization in a declaration inside a function is purely compile-time.

Accumulation example. Prints 1 2 3.

program accumulate
  implicit none
  integer :: acc
  write(*,*) acc(), acc(), acc() ! prints 1 2 3
end program accumulate

integer function acc()
  implicit none
  integer :: ac = 0
  ac = ac + 1
  acc = ac
end function acc

Safer version (best practices thanks to Alexey)

program accumulate
  implicit none
  integer :: acc
  write(*,*) acc(), acc(), acc() ! prints 1 1 1
end program accumulate

integer function acc()
  implicit none
  integer :: ac
  ac = 0
  ac = ac + 1
  acc = ac
end function acc

3.3 parameter vs. intent(in)

Defining a variable as parameter gives a constant, not an unchanged function argument:

parameter: Constants are marked with the keyword parameter

! constants: marked as parameter: not function parameters, but
! algorithm parameters!
character(len=61), parameter :: base60chars = "0123456789"&
     //"ABCDEFGHJKLMNPQRSTUVWXYZ_abcdefghijkmnopqrstuvwxyz"

An argument the function is not allowed to change is defined via intent(in):

unchanged-argument: constant function arguments are marked with intent(in)

! input: ensure that this is purely used as input.
! intent is only useful for function arguments.
integer, intent(in) :: number

3.4 To return values from functions, assign the value to the function itself

This feels surprisingly obvious, but it was surprising to me nontheless.

return: Values can be returned from functions at any time by assigning the function the return value and issuing return to end the function call.

numtosxg = "0"
return

The return statement is only needed when returning within a function. At the end of the function it is implied.

return-end: At the end of the function, return is not needed.

  numtosxg = res
end function numtosxg

3.5 Fortran array indizes start at 1 - and are inclusive

For an algorithm like the example base60, where 0 is identified by the first character of a string, this requires adding 1 to the index.

indizes-start-at-1: Fortran array-indexing starts at 1.

! note that fortran indizes start at 1, not at 0.
res = base60chars(remainder+1:remainder+1)//trim(res)

Also note that the indizes are inclusive. The following actually gets the single letter at index n+1:

Getting a single letter from a character array

base60chars(n+1:n+1)

In python on the other hand, the second argument of the array is exclusive, so to get the same result you would use [n:n+1]:

Python-example for getting a single letter from a list

pythonarray[n:n+1]

3.6 I have to trim strings when concatenating

It is necessary to get rid of trailing blanks (whitespace) from the last char to the end of the declared memory space, otherwise there will be huge gaps in combined strings - or you will get missing characters.

Example for string concatenaniong with and without trim.

program test
  character(len=5) :: res
  write(*,*) res ! undefined. In the last run it gave me null-bytes, but
                 ! that is not guaranteed.
  res = "0"
  write(*,*) res ! 0
  res = trim(res)//"a"
  write(*,*) res ! 0a
  res = res//"a"
  write(*,*) res ! 0a: trailing characters are silently removed.
  ! who else expected to see 0aa?
  write(res, '(a, "a")') trim(res) ! without trim, this gives an error!
                                   ! *happy*
  write(*,*) res
end program test

Hint from Alexey: use trim(adjustl(…)) to get rid of whitespace on the left and the right side of the string. Trim only removes trailing blanks.

Author: Arne Babenhauserheide

Emacs 24.3.1 (Org mode 8.0.2)

Anhang	Größe
surprises.org	8.42 KB
accumulate.f90	226 Bytes
accumulate-not.f90	231 Bytes
base60-surprises.f90	1.6 KB
trim.f90	501 Bytes
surprises.pdf	206.83 KB
surprises.html	22.47 KB
base60.f90	2.79 KB

Druckversion
Login to post comments

Use Node:

⚙ Babcom is trying to load the comments ⚙

This textbox will disappear when the comments have been loaded.

If the box below shows an error-page, you need to install Freenet with the Sone-Plugin or set the node-path to your freenet node and click the Reload Comments button (or return).

If you see something like Invalid key: java.net.MalformedURLException: There is no @ in that URI! (Sone/search.html), you need to setup Sone and the Web of Trust

If you had Javascript enabled, you would see comments for this page instead of the Sone page of the sites author.

Note: To make a comment which isn’t a reply visible to others here, include a link to this site somewhere in the text of your comment. It will then show up here. To ensure that I get notified of your comment, also include my Sone-ID.

Link to this site and my Sone ID: sone://6~ZDYdvAgMoUfG6M5Kwi7SQqyS-gTcyFeaNN1Pf3FvY

This spam-resistant comment-field is made with babcom.

Surprising behaviour of Fortran (90/95)

1 Introduction

Table of Contents

2 Testing Skelleton

2.1 Helpers

3 Surprises

3.1 I have to declare the return type of a function in the main program and in the function

3.2 Variables in Functions accumulate over several function calls

3.3 parameter vs. intent(in)

3.4 To return values from functions, assign the value to the function itself

3.5 Fortran array indizes start at 1 - and are inclusive

3.6 I have to trim strings when concatenating

Beliebte Inhalte

Heute:

Zuletzt angezeigt:

Draketo neu: Beiträge

Ein Würfel System

sn.1w6.org news

Sep. 12th decides the fate of the internet in the EU!