I have the following piece of code in my subroutine:
character x*256 ,y*80
common /foo/ x ,y(999)
Well, I did not actually write this. So I don't understant the dimensions here. Is y an 999 element wide array of 80 character long strings?
If so, how can I define this properly in Fortran 90, without the common block?
I will first say that the code you have is "proper" Fortran 90, but I agree with wanting to move away from common blocks.
There is, essentially, nothing specific to the character nature of the declaration. Whenever
<type> A
common /foo/ A(<size>)
is used there are two parts to the declaration of A, as well as the common association: the type and the dimension. Ignoring the association, declaration of the dimension in the common statement is allowed and the above is like
<type> A
dimension A(<size>)
This is in turn the same as
<type>, dimension(<size>) :: A
Coming to the specific example, the type is a character of length 80. Your non-common declaration would simply be
character(len=80), dimension(999) :: y
Indeed, then, y is a rank-1 array of size 999 of length-80 characters. y(10) is a scalar length-80 character (the 10th element of the array y).
x(10) isn't correct syntax, as the (10) is array indexing, and x is a scalar. For substrings a different indexing is required. x(10:10) is the 10th character of the character variable x; y(10)(10:10) is the 10th character of the 10th element of the character array y.