Small bugs with the C++ implementation

[pierreroudier]

There seem to have a few remaining issues related to the C++ part of the package, that was kindly contributed by @kdaust (see #15).

Technically, it seems the problem is actually not as much in the C++ code itself, but in the R code that precedes it.

First, this for loop (now commented out):

clhs/R/clhs-data.frame.R

Lines 79 to 82 in 8d45408

	# for(i in 1:ncol(data_factor)){
	# data_factor[,i] <- as.numeric(as.factor(data_factor[,i]))
	# }
	data_factor <- apply(data_factor, 2, function(x) as.numeric(as.factor(x)))

This converts automatically any character data to NAs.
I think it can be fixed by using as.numeric(as.factor(data_factor[,i)). I've implemented a version of this using apply.

Second issue, probably minor, but this line here looks like it's a hardcoded "5":

clhs/R/clhs-data.frame.R

Lines 89 to 92 in 8d45408

	data <- as.matrix(x)
	factIdx <- 1:5 # why five??
	ncont <- ncol(data)
	data_continuous <- x

I'm guessing it should be a 0?

[kdaust]

Hi @pierreroudier - thanks for catching this! The fix for the factor conversion looks good. I think I was assuming that any character columns were already factorised, but I agree that's not a safe assumption.

Regarding the hardcoded 5, I think its actually fine; its not in the block of code dealing with factors. The C++ function just needs some value for the vector, so I'm using 1:5 as a placeholder basically, even though it's not used at all in the computation.