NA as a factor level malformed by rbindlist

[sindribaldur]

data.table(V1 = factor(as.character(c(NA, 1:100, NA)), exclude = NULL))
# Error in as.character.factor(x) : malformed factor
data.table(V1 = factor(as.character(c(NA, 1:3, NA)), exclude = NULL))
#      V1
# 1: <NA>
# 2:    1
# 3:    2
# 4:    3
# 5: <NA>

Is this the expected behaviour?

Another example here: https://stackoverflow.com/q/58103098/4552295

I'm getting this result with data.table_1.12.2 and R version 3.6.1.

[jangorecki]

Same in devel. Thanks for reporting. Happens when printing DT having > 100 rows because of head&tail print format.

root cause of the issue is rbindlist in this line

data.table/R/print.data.table.R

Line 61 in 5ab1884

toprint = rbindlist(list(head(x, topn), tail(x, topn)), use.names=FALSE) # no need to match names because head and tail of same x, and #3306

producing malformed factor, so the issue is actually in rbindlist. fyi @mattdowle
minimal example

x = data.table(V1 = factor(as.character(c(NA, 1:3, NA)), exclude = NULL))
rbindlist(list(x), use.names=FALSE)$V1
#Error in as.character.factor(x) : malformed factor