> My question is why linear search of seq is faster than constant search of > tables?
Sorry to pile on four days later, but the right question would be "why is my benchmark code producing crazy results?" since it's much more likely there's something wrong with your quick benchmark than with Nim's fairly-mature standard library. I made the same mistake a few weeks ago when I posted here saying there seemed to be a bad performance bottleneck in the standard lib ... but it turned out that the real problem was that the build config I was using did not produce an optimized build at all. Oops.
