fread bug with colClasses

[arunsrinivasan]

require(data.table)
# Loading required package: data.table
# data.table 1.11.5 IN DEVELOPMENT built 2018-06-02 00:09:06 UTC; travis  Latest news: http://r-datatable.com
str <- "x1,x2,x3,x4,x5\n1,2,1.5,T,cc\n3,4,2.5,F,ff"

# Read correctly
fread(str)
#    x1 x2  x3 x4 x5
# 1:  1  2 1.5  T cc
# 2:  3  4 2.5  F ff

 # col x4 is wrong
fread(str, colClasses=c("integer", "numeric", "numeric", "integer", "character"))
#    x1 x2  x3 x4 x5
# 1:  1  2 1.5 cc cc
# 2:  3  4 2.5 ff ff

# col x4 is wrong here as well
fread(str, colClasses=c("integer", "numeric", "numeric", "logical", "character"))
#    x1 x2  x3 x4 x5
# 1:  1  2 1.5 cc cc
# 2:  3  4 2.5 ff ff

# correct if colClasses for x4 is 'character' type
fread(str, colClasses=c("integer", "numeric", "numeric", "character", "character"))
#    x1 x2  x3 x4 x5
# 1:  1  2 1.5  T cc
# 2:  3  4 2.5  F ff

[renkun-ken]

Looks a little bit similar to #2863 but with numeric and integers.

[mattdowle]

Yep, that's a bad one. It's in 1.11.4 and probably 1.11.0+.

[hstahl]

The "NULL" value in colClasses no longer drops the column:

fread(str, colClasses=c("integer", "NULL", "numeric", "character", "character"))
#    x1 x2  x3 x4 x5
# 1:  1  2 1.5  T cc
# 2:  3  4 2.5  F ff

[MichaelChirico]

@hstahl can you file this as a separate issue?

TBH I had no idea fread ever worked like this -- drop argument is designed for this purpose. Not sure how important consistency with read.csv is for this...

[jesskfullwood]

I'm running 1.11.2 and just came across the same bug as @hstahl

I have often found it's easier just to specify a list of c("character", "integer", "NULL", "integer" ...) rather than using drop, FWIW.