FR: in colClasses of fread, allow column names not present in actual table

[aushev]

I often need to read multiple tables that share some (but not all) columns which would be passed to colClasses. It would be then convenient to make one vector with all those column names, i.e. instead of

chrcols1 <- c('colA', 'colB')
dt1 <- fread(file1, colClasses=list(character=chrcols1))
chrcols2 <- c('colA', 'colC')
dt2 <- fread(file2, colClasses=list(character=chrcols2))
# etc

use

chrcolsAll <- c('colA', 'colB', 'colC')
dt1 <- fread(file1, colClasses=list(character=chrcolsAll))
dt2 <- fread(file2, colClasses=list(character=chrcolsAll))

Right now this is not allowed ( "Column name '%s' in colClasses[[%d]] not found" ) but instead of throwing an error we could just exclude those column names from further processing.

[gdmcdonald]

This would be a great feature!

[MichaelChirico]

In current dev, this is a warning (DTWARN), whereas in the linked commit it was an error (STOP).

Is a warning too strict here? The idea is to relegate it to verbose output only? Or can we close this?

[gdmcdonald]

Ahhhhh yes the current dev solves all my problems! Thankyou