如何从R调用Fortran程序 [英] How to call a Fortran program from R

问题描述

我对Fortran完全陌生，并且精通R. 我被传授了一个庞大的Fortran程序，其中包含约30个子例程，约15个函数和许多其他代码行. 有人告诉我我需要从R调用Fortran程序.我一直在网上搜索方法 创建R和Fortran之间的桥梁几乎没有成功. 我可以从命令行成功执行Fortran exe文件并创建所需的输出. 该fortran文件称为"FortFish.f"

I am totally new to Fortran and well versed with R. I was handed down a huge Fortran program with about 30 subroutines and about 15 functions and many other lines of code. I was told that I needed to call the Fortran program from R. I have been searching online for ways to create this bridge between R and Fortran with very little success. I can successfully execute the Fortran exe file from the command line and create the desired outputs. The fortran file is called "FortFish.f"

一个问题:

从R中，我应该调用Fortran程序还是必须分别调用Fortran函数和子例程?

From R, Do I call the Fortran program or do I have to call Fortran functions and subroutines separately?

在R中，我是否这样调用整个Fortran程序?:R CMD SHLIB FortFish.f，然后使用: dyn.load("FortFish.so")

From R, Do I call the entire Fortran program like this?: R CMD SHLIB FortFish.f and then use: dyn.load("FortFish.so")

如果我无法一次运行整个Fortran程序，我将根据要求发布几个小型的fortran函数和子例程. 有没有人使用R和Fortran可以共享的运行示例?

If I cannot run the entire Fortran program at once, I will post a couple of small fortran functions and subroutines upon request. Does anyone have a running example of using R and Fortran that can share?

我的Fortran代码非常大，否则，我将其发布在这里. 谢谢.

My Fortran code is extremely large, otherwise, I would post it here. Thank you.

推荐答案

我看到三种可能性:

1. 您可以单独编译Fortran程序，并使用R函数system()对其进行调用.您将必须通过该程序可以读取的格式通过文件传递数据.

1. You compile the Fortran program separately, and call it with R function system(). You will have to pass data through files, in a format that this program can read.

使用dyn.load()编译从R加载的DLL，然后使用.Fortran()调用Fortran函数.您可以轻松地传递数字数据(标量，向量或数组)，但是字符串数据更难处理.然后数组被复制.

You compile a DLL that you load from R with dyn.load(), then you call a Fortran function with .Fortran(). You can easily pass numeric data (scalars, vectors or arrays), but string data is more difficult to handle. And arrays are copied.

这种调用DLL函数的机制被认为过于简单，现在首选.Call()，但是要使用.Call()，则必须编写C包装程序.

This mechanism to call a DLL function is considered too simplistic and now .Call() is prefered, but to use .Call() you would have to write C wrappers.

我将举一个第二种可能性的例子. 考虑一下Fortran中的一个子例程，该子例程通过Horner算法评估多项式:

I will give an example of the second possibility. Consider a subroutine in Fortran that evaluates a polynomial by Horner's algorithm:

subroutine horner(n, a, x, y)
    implicit none
    integer :: n, i
    double precision :: a(n), x, y

    y = a(n)
    do i = n - 1, 1, -1
        y = y * x + a(i)
    end do
end subroutine

从命令行使用以下命令进行编译:

Compile from the command line with:

R CMD SHLIB horner.f90

要从R调用它:

dyn.load("horner.dll")

horner <- function(a, x) {
  .Fortran("horner", as.integer(length(a)), a, x, y=0)$y
}

horner(c(-2, 0, 1), 1.414)

如果您希望Fortran子例程将某些内容打印到RStudio控制台，则需要这样做(至少在Windows上如此):

If you want your Fortran subroutines to print something to RStudio console, you need to do (at least on Windows):

Sys.unsetenv("GFORTRAN_STDOUT_UNIT")
Sys.unsetenv("GFORTRAN_STDERR_UNIT")

这确实是一个简单的例子，一个更复杂的程序将需要更多的工作，但是您明白了.

This is really a trivial example, and a more complex program will require more work, but you get the idea.

如果您的Fortran程序是独立的(它具有一个程序"单元，应该被编译为从命令行调用的可执行文件)，并且如果您不熟悉Fortran，那么我建议您坚持使用第一个选择，这会简单得多.这就是打包的季节性所做的事情:调用人口普查 X13AS .可执行文件位于

If your Fortran program is standalone (it has a 'program' unit and is supposed to be compiled to an executable called from the command line), and if you are new to Fortran, I would suggest to stick with the first choice, which will be much simpler. That's what the seasonal packaged does: call the executable of Census' X13AS from within R. The executable is in the x13binary package.

这篇关于如何从R调用Fortran程序的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！