Problem with array definition

Help
1

Dear friends,

Please, I would like to request your help with the following problem: I have to define a large array containing zeros which don’t fit in a single row due to I’m exceeding the maximun numbers of character per row. For example:

a = [0, 0, 0, 0, 0, 0, …, 0, &
0, 0, 0, 0, … 0, 0]

However, the second row cannot start with a zero given that, when compiling, I get this error
" Syntax error in array constructor at (1). Zero is not a valid statement label at (1)"

How can I overcome this problem?

Many thanks for the help.

2

How about:

integer, parameter :: n = 1000000
real :: a(n)
a = 0
2 Likes
3

Although @ivanpribec solution is THE solution, note that the label problem could probably be avoided by putting an & at the start of the second line:

a = [0, 0, 0, 0, 0, 0, …, 0, &
   & 0, 0, 0, 0, … 0, 0]

Updated: in fact, no. See the post by @themos downward: Problem with array definition - #9 by themos
In free-form format, there is no error and no need to add an &. And in fixed-form, the trailing & is not accepted.

2 Likes
4

and if a is allocatable you can do

allocate(a(n), source=0)

2 Likes
5

@ivanpribec, @vmagnin, @rwmsu, many thanks for your help.

2 Likes
6

How about using a data statement

integer, parameter :: n = 20
real, dimension(n) :: a; data a/n*0.0/

print *,a

produces 20 values of 0.0

If this is fixed format file, then you need to add a continuation character in column 6

!     |<-- column 7
      a = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 
     &0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
7

By the large number of 0-elements in the examples above, on one hand, and some prior exposure to Python’s scripy library on the other I noticed Fortran seems equally to work with sparse matrices (or, perhaps more suitable for the parlance of Fortran, a sparse arrays). One of the entries found is this one.

8

@johnalx ,

Welcome to the forum.

Re: the “data statement,” note the use of DATA also imparts the SAVE attribute to the variable, essentially making it a static variable and adversely impacting thread-safety. Coders may not want that attribute in modern applications.

An option to consider if implied-do is of interest is

a = [( xx, integer :: i = 1, n )] !<-- xx can be a constant such as zero, or a valid expression involving `i`
3 Likes
9

Hello,
I am guessing you are using the gfortran compiler and you wrote something like this and saved it in a file foo.f.

c234567
      dimension a(6)
      a=[0,0,0, &
 0    ,0,0]
      end

What went wrong here is that you used free-format Fortran but the conventions of gfortran treat the contents of foo.f as fixed-form Fortran. In fixed-form the trailing ampersand (&) is an error (which is the Error: Syntax error in array constructor at (1)) and that first zero on line 3 is interpreted as a statement label and not part of a statement (which is the Zero is not a valid statement label at (1))

If you write free-format Fortran you must either put it in a file with a filename extension that the compiler expects (most compilers will accept .f90 for this) or override the compiler’s defaults with a command line option (in gfortran’s case, -ffree-form)

2 Likes
10

I did not know that about the data statement. Thank you this caveat.

11

It is also possible to initialize the array with a constant:

module mymod
  implicit none
  integer, parameter :: n = 20
  real :: a(n) = 0                      ! implicit save
end module
program main
use mymod
print *, a
end program

Edit: for any new Fortran programmers, the following might be more illustrating of what implicit save means

module mymod
  implicit none
  real :: a(5) = 0    ! implicit save
end module
module usemymod
  use mymod, only: a
  implicit none
  private
  public :: modify_a
contains
  subroutine modify_a()
    a(1) = 42.
  end subroutine
end module
program main
  use usemymod, only: modify_a
  call modify_a
  block
    use mymod, only: a
    print *, a                        ! a(1) equals 42.
  end block
end program