https://issues.dlang.org/show_bug.cgi?id=15401
--- Comment #4 from Andrei Alexandrescu <and...@erdani.com> --- (In reply to Infiltrator from comment #3) > (In reply to Andrei Alexandrescu from comment #2) > > (In reply to Infiltrator from comment #1) > > > Once issue 15421 is fixed, this is a simple matter of > > > topN(l, r); > > > sort(l); > > > > > > Which brings us to the question of: should partialSort(Range, index) be > > > changed to call partialSort(r[0..n], r[n..$]) to reduce duplication or is > > > there a large performance difference in the two topNs? > > > > Affirmative. > > Sorry, do you mean affirmative to how to implement partialSort(Range, Range); > or affirmative to changing partialSort(Range, index) to call the (Range, > Range) version; > or affirmative to there being too large of a performance difference between > the two versions of topN? My bad, I missed the "or". I don't think there's a loss of efficiency if partialSort with index calls partialSort with the two subranges. --