Exported file from R using `haven` cannot be opened by SAS

When exporting data from R using haven::write_sas(), the resulting sas7bdat file is not recognized (i.e. cannot be loaded) by SAS EG/9.4. Although there are several other packages such as foreign that provide alternative approaches, I was hoping to find a relatively automated way to push a dataset from my R session directly into SAS.

When using haven, the file is made but cannot be opened by SAS EG nor 9.4:

# Load package
library(haven)

# Save data
write_sas(mtcars, "mtcars.sas7bdat")

---

Using foreign as alternative to haven:

library(foreign)

write.foreign(df = mtcars,
              datafile = 'mtcars.txt',
              codefile = 'mtcars.sas',
              dataname = 'libraryname.tablename', # Destination in SAS to save the data
              package = 'SAS')

Running the SAS code output from foreign is successful.

* Written by R;
*  write.foreign(df = mtcars, datafile = "mtcars.txt", codefile = "mtcars.sas",  ;

DATA  libraryname.tablename ;
INFILE  "mtcars.txt" 
     DSD 
     LRECL= 43 ;
INPUT
 mpg
 cyl
 disp
 hp
 drat
 wt
 qsec
 vs
 am
 gear
 carb
;
RUN;

---

However, neither of these methods help with automatically pushing the data directly from R into a SAS library, which would be preferable.

There is a lengthy discussion on GitHub describing some of the challenges when exporting data from R for use in SAS via haven. In addition to providing a solution on how to automate data transfer from R to SAS, I hope this can serve as an answer to some related questions.

If one wants to use tools designed by SAS for interoperability with R, RSWAT on GitHub is likely a more robust option. However, this will assume that you have access to SAS Cloud Analytics Services configured for this purpose.

If you are working with a SAS 9.4 on your machine and perhaps also connect to SAS servers (i.e. using rsubmit; commands), it should be relatively straightforward to pass a data-set directly from R into a SAS library. There are three steps:

1. Format dataset for SAS; although foreign will do a lot of the formatting changes, I prefer converting factors back to characters and having NA replaced with "". This I find ensures that no special formatting is needed by colleagues to open the final table in SAS.

# Example data
data <- data.frame(ID = c(123, NA, 125),
                        disease = factor(c('syphilis', 'gonorrhea', NA)),
                        AdmitDate = as.Date(c("2014-04-05", NA, "2016-02-03")),
                        DOB = as.Date(c("1990-01-01", NA, NA)))

# Function defined for converting factors and blanks
convert_format_r2sas <- function(data){
  data <- data %>%
    dplyr::mutate_if(is.factor, as.character) %>%
    dplyr::mutate_if(is.character, tidyr::replace_na, replace = "")
  return(data)
}

# Convert some formatting
data <- convert_format_r2sas(data)

2. Use foreign to export the data and associated code

library(foreign)

# Ensure the data and code files are saved in an easily accessible location (ideally in or downstream of your R project directory)
write.foreign(df = data ,
              datafile = 'data.txt',
              codefile = 'data.sas',
              dataname = 'libraryname.tablename', # Destination in SAS to save the data
              package = 'SAS')

3. Pass code to local SAS installation using custom function. You may need to adjust the location of the SAS.exe as well as the configuration file. This will work both passing a list of SAS files, or SAS code written directly in R as a character vector.

# Define function for passing the code to SAS and upload data (may require tweaking the local SAS installation location and configuration file)
pass_code_to_sas <- function(sas_file_list = NULL, inputstring = NULL,
         sas_path = "C:/LocationTo/SASHome/SASFoundation/9.4/sas.exe",
         configFile = "C:/LocationTo/SASHome/SASFoundation/9.4/SASV9.CFG") {

  # If provided list of scripts, check they are all valid
  if(!is.null(sas_file_list)){
    if(any(purrr::map_lgl(sas_file_list, file.exists)) == FALSE | is.list(sas_file_list) == F){
      stop("You entered an invalid file location or did not provide the locations as a list of characters")
    }
  }

  sink(file.path(R.home(), "temp_codePass.sas"))

  if(!is.null(sas_file_list)){
    for(i in 1:length(sas_file_list)){
      cat(readLines(sas_file_list[[i]]), sep = "\n")
    }
  }

  cat(inputstring)
  sink()

  # Output message to view what code was sent...
  message(paste0("The above info was passed to SAS: ",
                 if(!is.null(sas_file_list)){for(i in 1:length(sas_file_list)){cat(readLines(sas_file_list[[i]]), sep = "\n")}},
                 print(inputstring)))

  # Run SAS
  system2(sas_path,
          args = paste0(
            "\"", file.path(R.home(), "temp_codePass.sas"), "\"",
            if(!is.null(configFile)) { paste0(" -config \"", configFile, "\"")}
          )
  )

  # Delete the SAS file
  file.remove(file.path(R.home(), "temp_codePass.sas"))
}

# Pass data to SAS
pass_code_to_sas(sas_file_list = 'path2codefile/data.sas')