Dear R-devel, When I was playing around with different kind of indices when subsetting I noticed some unexpected behaviours when using non-integer numeric indices, especially near zero. From the docs: “Numeric values are coerced to integer as by as.integer<http://127.0.0.1:14277/help/library/base/help/as.integer> (and hence truncated towards zero).” But some behaviour differs from that, and the behaviour also differs between [ and [[ : c(1,2)[as.integer(.5)] --> numeric(0) # As expected c(1,2)[.5] --> numeric(0) # As expected c(1,2)[[as.integer(.5)]] --> Error in c(1, 2)[[as.integer(0.5)]] : attempt to select less than one element in integerOneIndex # Also as expected c(1,2)[[.5]] --> [1] 1 # Not so expected c(1,2)[[1.5]] --> [1] 1 # As expected, but this also means somevector[[n]] and somevector[[n+1]] give back the same element for 0<n<1 c(1,2)[[as.integer(-1.5)]] --> [1] 2 # As would be expected, though negative subscript for [[ is of course sketchy c(1,2)[[-1.5]] --> [1] 1 # But coerced to -2 is the last thing I’d expect c(1,2)[as.integer(-.5)] --> numeric(0) # As expected c(1,2)[-.5] --> [1] 2 # Coerced to -1?, this also means that length(union(c(1,2)[-n], c(1,2)[n])) != 2 for -1<n<1 c(1,2)[-1.5] --> [1] 2 # Again as expected, but same problem as before: indexing with n and n+1 can give same element back.
I suspect most of this behaviour is due to the case of special treatment of zero, where first 0-indices are dropped, and only then the casting to integer is done, and when that returns zero some unforeseen behaviour occurs. Along with using negative indices when extracting with [[, which in any case only succeeds with length-2 vectors (We need a length-one index resulting in the return of a single element). For the last case I’d think we’d do best in throwing an error whenever negative indices are used with [[, but for other cases I think we need to change the underlying code, or at the very least update documentation. Any thoughts? Best regards, Emil Bode Data-analyst +31 6 43 83 89 33 emil.b...@dans.knaw.nl<mailto:emil.b...@dans.knaw.nl> DANS: Netherlands Institute for Permanent Access to Digital Research Resources Anna van Saksenlaan 51 | 2593 HW Den Haag | +31 70 349 44 50 | i...@dans.knaw.nl<mailto:i...@dans.kn> | dans.knaw.nl<applewebdata://71F677F0-6872-45F3-A6C4-4972BF87185B/www.dans.knaw.nl> DANS is an institute of the Dutch Academy KNAW<http://knaw.nl/nl> and funding organisation NWO<http://www.nwo.nl/>. [[alternative HTML version deleted]] ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
