[R] Fortran Code to R Code

[Duncan Murdoch]

On 11/28/2007 12:54 PM, Dave Roberts wrote:
> Bryan,
> 
>      The previous responses will point you in the right direction.  I 
> have found, however, that it takes a while to get used to the 
> requirements, and there are many possible sources of error.

This is true, and the advice below is all good.  One addition:  the 
inline package can help when you're starting.  Ask it to write a simple 
Fortran wrapper function for you, and you'll see what it should look like.

Duncan Murdoch

> 
> 1) You have to convert the main program to a subroutine.  Any arrays 
> declared in the main program or any nested subroutine have to created in 
> R before the subroutine is called, and passed to the subroutine.  On the 
> other hand, this provides the advantage of allocatable array size in 
> FORTRAN 77.
> 
> 2) You have to make sure that you are very careful about storage.  I 
> have reluctantly decided that attempting minimize storage by using 
> integer*2 and real*4 is not worth it, and I now use just integer and 
> double precision to minimize problems with R.
> 
> 3) Consider casting every variable passed in the .Fortran() call.  Many 
> of them would be correct by default possibly, but it is easier to simply 
> cast them all, e.g. as.integer() for all integers, as.double() for real 
> numbers.  These get the storage right, and strip of any attributes 
> besides he actual values.  Fortunately, R and FORTRAN agree about 
> storage order in arrays and you don't have to micro-manage that.
> 
> 4) When you get it wrong, the probability of a seg fault and hard crash 
> is very high.  So, every time you modify the R function that calls the 
> .Fortran you have written, do a save.image() before you try the code. 
> Otherwise you lose all your changes.  If you have the luxury of working 
> in linux (and presumably other *nixes), you can use "write(6,*) 
> whatever" debug statements in your code and the output goes to your R 
> session.  In Windows, you have to use the specific debug routines 
> described in the R manuals.
> 
> 5) Managing the returned values has to handled in the .Fortran() call by 
> giving the argument a name.  Then, that name can be used as the 
> component name of the returned object.  Just below is an example from 
> labdsv that converts any dissimilarity or distance matrix to the nearest 
> euclidean distance object.  The PACKAGE= argument is used when the 
> routine is pat of a package; it's not necessary for ad hoc functions.
> 
> euclidify <- function (x,upper=FALSE,diag=FALSE)
> {
>      x <- as.dist(x)
>      tmp <- .Fortran("euclid",
>          x=as.matrix(x),
>          as.integer(attr(x,"Size")),
>          PACKAGE='labdsv')
>      tmp2 <- as.dist(tmp$x)
>      attr(tmp2, "Labels") <- dimnames(x)[[1]]
>      attr(tmp2, "Diag") <- diag
>      attr(tmp2, "Upper") <- upper
>      attr(tmp2, "method") <- paste("euclidify", attr(x, "method"))
>      attr(tmp2, "call") <- match.call()
>      tmp2
> }
> 
> Dave Roberts
> 
> 
> Bryan Klingaman wrote:
>> How do you convert Fortran Code to R Code to use and execute in R?  Or how do you take Fortran code and make it run in R?  So what I'm getting at is, I have some code in fortran and I want to be able to run that same code for R.  Please email me back and let me know how to do that, thanks.
>>    
>>   Bryan
>>   Email: BKling9er at yahoo.com
>> 
>>        
>> ---------------------------------
>> 
>> 	[[alternative HTML version deleted]]
>> 
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>> 
>>
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.