Somewhat related to #4060 in appearance but crashes on all versions of R. Different mechanism, similar appearance. Found this while working on amending #4272. I'll submit a fix soon (hopefully). Should I put it in with #4272 or submit a separate PR? These are all related in that they are triggered by passing zero length icols/xcols to Cbmerge.
I'm not a hundred percent on what this should do, but I'd suggest to have it behave the same way for all values of roll.
DT = data.table(A="a", key="A")
DT[data.table(), roll = 'nearest'] # Will segfault
# Proposed return Value == DT[0]:
# Empty data.table (0 rows and 1 cols): A
Somewhat related to #4060 in appearance but crashes on all versions of R. Different mechanism, similar appearance. Found this while working on amending #4272. I'll submit a fix soon (hopefully). Should I put it in with #4272 or submit a separate PR? These are all related in that they are triggered by passing zero length
icols/xcolstoCbmerge.I'm not a hundred percent on what this should do, but I'd suggest to have it behave the same way for all values of
roll.