> On Dec 31, 2015, at 12:20, Brent Royal-Gordon via swift-evolution
> <[email protected]> wrote:
>
> 2. It forces a naïve implementation, which may not be the best idea. In the
> Perl world, for instance, we would usually use a Schwartzian transform to
> implement this, particularly if the key might be expensive to compute:
>
> array.map { ($0, sortKey($0)) }.sort { $0.1 < $1.1 }.map { $0.0 }
+1 and -1 to this. Computing the sort key N times instead of 2*(# of
comparisons) can be a big win sometimes. On the other hand, allocating memory
for the sort keys might be a net loss, especially if the collection being
sorted is large. I guess that means it's better to be made explicit, but it
would be nice™ if it were more convenient than it is now. Brent's way does a
lot of copies; I tried to avoid that but quickly ran into trouble…
var keys = array.map { $0.key }
array.sort { ??? }
…because the current 'sort' takes a comparator, which just takes values, not
indices.
Jordan_______________________________________________
swift-evolution mailing list
[email protected]
https://lists.swift.org/mailman/listinfo/swift-evolution
