Inverse column selection

[Moohan]

Is it possible (and if not could it be implemented) to inversely select colums?

If I have a file with columns v1:v5 I could use columns = c("v1", "v2", "v3") I would like to be able to do something like columns = c(-"v4", -"v5") i.e. read all except the specified columns.

[MarcusKlik]

Hi @Moohan, thanks for your feature request!

Yes, so something similar to:

# write some sample data with 3 columns
tmp_file <- tempfile(fileext = ".fst")
fst::write_fst(data.frame(X = 1:26, Y = LETTERS, Z = sample(1:100, 26)), tmp_file)

# get reference to file
ft <- fst::fst(tmp_file)

# determine reverse selection of col_sel
col_sel <- c("X", "Z")
rev_sel <- setdiff(colnames(ft), col_sel)

# read all columns but 'X' and 'Z'
fst::read_fst(tmp_file, rev_sel)
#>    Y
#> 1  A
#> 2  B
#> 3  C
#> 4  D
#> 5  E
#> ...```

I know that in dplyr that can be done with

select (mydata, -c(mpg, cyl, wt))

The easiest way to implement that behavior in fst would be to allow for an argument reverse_columns that can be set to TRUE to reverse the column selection. Is that a feature that you would regularly use?