Question

I suspect this is a bug in library(foreign) but I am not sure.

When I export a dataframe containing a boolean vector with write.foreign, the resulting SASDATA file causes an error on SAS import.

Example:

df = data.frame(id = c(1,2,3), boolean = c(TRUE,FALSE,TRUE))
library(foreign)
write.foreign(df, "test.sasdata", "test.sas",
    "SAS", dataname="WORK.TEST", validvarname="V7")

Contents of test.sasdata:

1,TRUE
2,FALSE
3,TRUE

Output from SAS when I run “test.sas”:

NOTE: Invalid data for boolean in line 1 3-6
RULE: ----+----1--...
1     1,TRUE 6
id=1 boolean=. _ERROR_=1 _N_=1
... (errors for remaining rows)

The workaround is to do:

df$boolean <- as.numeric(df$boolean)

before export. Or, more generically:

logicals <- sapply(df, is.logical)
if (any(logicals))
    df[logicals] <- lapply(df[logicals], as.numeric)

Am I doing something wrong with write.foreign? Or is this a bug?

EDIT: Revised code above to use sapply/lapply as writeForeignSAS does. I have submitted a patch to writeForeignSAS which incorporates the above.

Thanks to DWin for his assistance.

Answer 1

Given that ‘logical’ is not listed among the classes that are claimed to be properly exported to SAS, you could hardly call it a bug.

?write.foreign