Thanks for confirming and giving details on the rationale (... and I'll updated R.utils to use format() instead).
Regarding as.character(x)[j] === as.character(x[j]): I agree with this - is that property of as.character()/subsetting explicitly stated/documented somewhere? I wonder if this is a property we should all strive for for other types of objects? /Henrik On Thu, Sep 23, 2021 at 12:46 AM Martin Maechler <maech...@stat.math.ethz.ch> wrote: > > >>>>> Henrik Bengtsson > >>>>> on Wed, 22 Sep 2021 20:48:05 -0700 writes: > > > The update in rev 80946 > > > (https://github.com/wch/r-source/commit/d970867722e14811e8ba6b0ba8e0f478ff482f5e) > > caused as.character() on hexmode objects to no longer pads with zeros. > > Yes -- very much on purpose; by me, after discussing a related issue > within R-core which showed "how wrong" the previous (current R) > behavior of the as.character() method is for > hexmode and octmode objects : > > If you look at the whole rev 80946 , you also read NEWS > > * as.character(<obj>) for "hexmode" or "octmode" objects now > fulfills the important basic rule > > as.character(x)[j] === as.character(x[j]) > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > > rather than just calling format(). > > The format() generic (notably for "atomic-alike" objects) should indeed > return a character vector where each string has the same "width", > however, the result of as.character(x) --- at least for all > "atomic-alike" / "vector-alike" objects -- > for a single x[j] should not be influenced by other elements in x. > > > > > > Before: > > >> x <- structure(as.integer(c(0,8,16,24,32)), class="hexmode") > >> x > > [1] "00" "08" "10" "18" "20" > >> as.character(x) > > [1] "00" "08" "10" "18" "20" > > > After: > > >> x <- structure(as.integer(c(0,8,16,24,32)), class="hexmode") > >> x > > [1] "00" "08" "10" "18" "20" > >> as.character(x) > > [1] "0" "8" "10" "18" "20" > > > Was that intended? > > Yes! > You have to explore your example a bit to notice how "illogical" > the behavior before was: > > > as.character(as.hexmode(0:15)) > [1] "0" "1" "2" "3" "4" "5" "6" "7" "8" "9" "a" "b" "c" "d" "e" "f" > > as.character(as.hexmode(0:16)) > [1] "00" "01" "02" "03" "04" "05" "06" "07" "08" "09" "0a" "0b" "0c" "0d" > "0e" > [16] "0f" "10" > > > as.character(as.hexmode(16^(0:2))) > [1] "001" "010" "100" > > as.character(as.hexmode(16^(0:3))) > [1] "0001" "0010" "0100" "1000" > > as.character(as.hexmode(16^(0:4))) > [1] "00001" "00010" "00100" "01000" "10000" > > all breaking the rule in the NEWS and given above. > > If you want format() you should use format(), > but as.character() should never have used format() .. > > Martin > > > /Henrik > > > PS. This breaks R.utils::intToHex() > > [https://cran.r-project.org/web/checks/check_results_R.utils.html] > ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
