Allow setting iout?d to 0 for no output.

[soaringxmc]

This PR disables output when iout[0-3]d, icheck, or isave are set to 0.

Closes #121, #127.

[p-costa]

Excellent, thanks Mao @soaringxmc!

This sounds good, I just want to do some small checks because evaluating mod(istep,0) may break a run, as it is undefined. I'll get back to you soon.

[soaringxmc]

Hi, Pedro!
The implementation should be safe, because

When iout0d=0:
First part iout0d > 0 evaluates to FALSE
Due to Fortran's short-circuit evaluation of .and., the second part mod(istep,iout0d) is never evaluated
Therefore, the potential division by zero is avoided

My tests show this does not cause problems.

[p-costa]

This program does breaks for me:

  integer :: a,i
  a = 0
  i = 10
  if (a > 0.and.mod(i,a) == 0) then
    print *, '1'
  else
    print *, '2'
  end if
end

when compiled with gfortran -fcheck=all test.f90 && ./a.out. A solution without resorting to nested ifs could be evaluating mod(istep,max(1,out1d)).

[soaringxmc]

You are right! I also think mod(istep,max(1,out1d)) is a good solution.

[p-costa]

Sounds good, do you want to update your PR with this change? We should also allow for masking the post-processing of the initial conditions when the simulation starts: https://github.com/CaNS-World/CaNS/blob/main/src/main.f90#L328,L330.

[soaringxmc]

Okay. I will update my PR.

[p-costa]

Thanks! I see now that I can also push to your fork, so if you'd like, I can also make the changes.

[soaringxmc]

Hi, Pedro! I have updated the PR.

[p-costa]

Thanks Mao! I pushed some minor changes. I think this PR now also nicely addresses #121 more elegantly.