A simple Fortran module for passing Fortran output to a Python script for later using Matplotlib to make figure

Fortran is weak in making figure, while Matplotlib is strong in making figure. It is not necessary to reinvent a Fortran wheel of making figure, instead, a simple way to remedy the weakness of Fortran is building a simple bridge from Fortran to Matplotlib.

The module fdata2pyplot is defined in src/fdata2pyplot.f90. Three public subroutines fdata2pyplot_pass_data, fdata2pyplot_add_others and fdata2pyplot_plt are defined in it.

program test
 use iso_fortran_env, only: sp => real32, dp => real64
 use fdata2pyplot
 implicit none
 integer :: i, j
 real(sp) :: x(-500:500), y(-500:500)
 real(dp) :: z(-500:500, -500:500)
 x = [(real(i, sp), i=-500, 500)] / 500.0_sp * 3.0_sp
 y = [(real(j, sp), j=-500, 500)] / 500.0_sp * 3.0_sp
 do i = -500, 500
 do j = -500, 500
 z(i, j) = exp(-x(i)**2.0_sp/2.0_sp) * exp(-y(j)**2.0_sp/2.0_sp)
 end do
 end do
 call fdata2pyplot_pass_data(x, "X")
 call fdata2pyplot_pass_data(y, "Y")
 call fdata2pyplot_pass_data(z, "Z", "Gaussian")
 call fdata2pyplot_add_others("plt.contourf(X, Y, Z)")
 call fdata2pyplot_add_others("plt.axis('square')")
 call fdata2pyplot_add_others("plt.show()")
 call fdata2pyplot_plt()
end program test

After running this program:

1. Fortran arrays x, y, z will be saved to X.txt, Y.txt and Gaussian.txt respectively;

2. A Python script named plt.py will be generated, and it contains:

import numpy as np
import matplotlib.pyplot as plt
X = np.loadtxt('X.txt')
Y = np.loadtxt('Y.txt')
Z = np.loadtxt('Gaussian.txt')
plt.contourf(X, Y, Z)
plt.axis('square')
plt.show()

3. The Python script plt.py will be run for making figure.

1. fdata2pyplot_pass_data has three dummy arguments: fortran_arr, py_arr_name, and optional txt_name. For convenience, You can only associate fortran_arr and py_arr_name with effective arguments, and make py_arr_name be the name of fortran_arr. For example, you can run:

 call fdata2pyplot_pass_data(x, "x")
 call fdata2pyplot_pass_data(y, "y")
 call fdata2pyplot_pass_data(z, "z")

2. In the above example, I called fdata2pyplot_add_others several times, which seems to be cumbersome. However, since different operating systems use different line-break characters, I have to do it like this to make the example be safely cross-platform. If you are sure that you will not meet with this trouble, you can call fdata2pyplot_add_others one time for adding all "others". For example, if you will only run the program on a Unix-like platform, you can run:

 call fdata2pyplot_add_others("plt.contourf(X, Y, Z)"//new_line('A')// &
 "plt.axis('square')"//new_line('A')// &
 "plt.show()")

3. Usually we don't know how to make the clearest figure. You need not call fdata2pyplot_add_others and fdata2pyplot_plt. You can only call fdata2pyplot_pass_data, and then modify and run plt.py by hand to try different methods of making figure.

• Numpy
• Matplotlib