Re: [boost] [tokenmap] updated perfect hash container in sandbox

Thorsten Ottosen wrote:

If there is no reason that performance cannot be compared to boost::unordered_map, then please also perform this comparison.

will do..

Consider also to reorder your template arguments to make the interface as close to normal maps as possible.

The intention was that the token-type, being limited to an integral type, would default to an int which would work for vast majority of cases. So the user could simply write: tokenmap<my_class> fast_map; instead of: tokenmap<int, my_class> fast_map; The second may give an impression that the key can be anything other, whereas in reality it's quite limiting. I don't feel strong about either method though. -sl