CODEWITHSUNDEEP
r
John Doucette
John Doucette

Reputation: 4540

Why does Rscript fail to run FSelector scripts that R can handle fine?

This is a personal record, because the problem is difficult to Google, and I imagine I'll run into it again in a couple of years.

I have a script like: 

library(FSelector)
table <- read.csv("somefile", header=0);
table <- table[,colSums(is.na(table))<nrow(table)]
imputed_table <- apply(table, 2, function(x){x <- replace(x, is.na(x), mean(x, na.rm=TRUE))});
nms <- colnames(table)
model <- information.gain(as.formula(paste(nms[length(nms)],"~.")), table)

When run with:

R --no-save < IG.R

this works fine, and prints the model.

When run with:

Rscript ./IG.R

this crashes with the error:

Error in .jarray(x) : could not find function "getClass"
Calls: information.gain ... read_model_frame_into_Weka -> read_data_into_Weka -> .jcall -> .jarray -> .Call
Execution halted

Why does this happen?

Upvotes: 0

Views: 1132

Answers (1)

John Doucette
John Doucette

Reputation: 4540

This happens because Rscript does not load the rJava library by default, which is required by FSelector, but apparently is not loaded when FSelector is loaded. In contrast, the standard "R" command does load rJava by default.

To resolve this, add:

library(rJava)

to the script before the information.gain call.

Upvotes: 1

Related Questions