[v3] Document implementation-defined bucket counts

Jonathan Wakely jwakely.gcc@gmail.com
Tue Jan 8 20:29:00 GMT 2013


On 8 January 2013 20:14, François Dumont wrote:
>     I hadn't notice that default bucket count for range constructors was 0.
> That's surprising as the equivalent on internal _Hashtable type is 10 like
> for the other constructors. Why not generalizing to 10 ? The advantages
> would be:
> - doc simplification
> - constructors consistency
>
>     0 is also surprising for a container that has such an important memory
> overhead that creating it to store 1 or 2 elements looks like using a tank
> to kill a fly.

Yes, good question and good points. I don't know the answers, I was
just documenting what we have, as required for any
implementation-defined properties.



More information about the Libstdc++ mailing list