NA values with stringsAsFactors = true

[JamesOwers]

If I run fread specifying na.strings is this not respected by stringsAsFactors? I see in the code that as.factor is called without the exclude= argument.

I'm asking because I get factors with NA as one of the levels. This means DT[is.na(factorcol),] returns no rows which is not what I was expecting. The workaround it to use DT[is.na(levels(factorcol)),]

[arunsrinivasan]

I don't fully understand this. There's no reproducible example or version number. But I think its duplicate of #1408 and news item 14 under bug fixes for v1.9.7. We don't use as.factor() but our own as_factor().

Please try with devel version before posting.

# v1.9.7 (latest commit)
require(data.table)
fread('a,b\na,2\nb,3\nc,4\nNA,5\n', stringsAsFactors=TRUE, na.strings=c("NA", "c"))[is.na(a)]
#     a b
# 1: NA 4
# 2: NA 5

If I seem to have misunderstood, please explain with a reproducible example. Thanks.

[JamesOwers]

Exactly what I needed. Many thanks. Your mind reading skills are on point. Apologies for poor description 😞