fread suggests fill for dropped column

[tdhock]

Hi all!
Today I was using fread, on some data with one more column name than there is data, and I observed the following warning, which I think is a false positive. I would have expected no warning, because I told fread to drop the last column. Is this a bug?

> fread("x,y,z\n1,2",drop=3)
       x     y
   <int> <int>
1:     1     2
Warning message:
In fread("x,y,z\n1,2", drop = 3) :
  Detected 3 column names but the data has 2 columns. Filling rows automatically. Set fill=TRUE explicitly to avoid this warning.
> fread("x,y,z\n1,2",colClasses=list(NULL=3))
       x     y
   <int> <int>
1:     1     2
Warning message:
In fread("x,y,z\n1,2", colClasses = list(`NULL` = 3)) :
  Detected 3 column names but the data has 2 columns. Filling rows automatically. Set fill=TRUE explicitly to avoid this warning.

[joshhwuu]

As far as I can tell, this is because the check for the warning happens before drop is applied internally:

fread("x,y,z\n1,2",drop=3,verbose=TRUE)

...
[07] Detect column types, good nrow estimate and whether first row is column names
  Number of sampling jump points = 1 because (3 bytes from row 1 to eof) / (2 * 3 jump0size) == 0
Types in 1st data row match types in 2nd data row but previous row has 3 fields. Taking previous row as column names.
Warning message in fread("x,y,z\n1,2", drop = 3, verbose = TRUE):
“Detected 3 column names but the data has 2 columns. Filling rows automatically. Set fill=TRUE explicitly to avoid this warning.”
  All rows were sampled since file is small so we know nrow=0 exactly
[08] Assign column names
[09] Apply user overrides on column types # where drop happens
...

I agree that this behavior is unintuitive, but it may be hard to fix without refactoring the internals of fread?