In my pet project I'm casting a lot of strings to named enum
members.
enum Animal { Gorilla, Shark, Alien, Rambo, Dolphin }
auto foo = "Dolphin";
auto fooAsEnum = foo.to!Animal;
While micro-optimizing because it's fun, I see that it's much
faster (by some factor of >3.5x) to do such casts as associative
array lookups rather than by using std.conv.to as above. As in,
populate an Animal[string] array with all enum members indexed by
the strings of their names, allowing you to get the Animal you
want via animalAA[foo] or (foo in animalAA).
In comparison, what code is generated from the foo.to!Animal
cast? A big final switch? A long if-else-if-else chain?
http://dpaste.dzfl.pl/7e700a1053c0
(Can the compiler not generate such code instead?)
