Subject: Re: [boost] [flat_map] Any interest in a flat_map variant that is much faster on insertion/erase at the price of a slower lookup?
From: Giacomo Drago (giacomo_at_[hidden])
Date: 2015-03-13 09:50:40
On 2015-03-13 13:37, Ion Gaztañaga wrote:
> El 13/03/2015 a las 10:11, Giacomo Drago escribió:
>> Insertions are one by one, and they are random with respect to the order
>> of the keys. Same applies for lookups and erasures. No cheating.
>
> I didn't want to suggest any cheating, just trying to understand the
> scenario. If insertions are by range, then flat_map can be optimized. If
> insertions are one by one, it's more difficult, unless there are two
> phases, first insertion (without lookups between insertions) and after
> that, only searches. That scenario could be also optimized.
They are one by one, and can interleave in any possible way without
affecting performance. I didn't imply you were suggesting any cheating,
but I know that benchmarks can be crafted to make things look better
than they are. And you can do this without even knowing.
Best
GIacomo