[RFC] optional fast iterators for set/map in libstdc++

Benjamin Kosnik bkoz@redhat.com
Tue May 12 03:45:00 GMT 2009


> It seems that to be really useful, different container classes should
> be created.  In this way, users could more easily use them based on
> their performance and memory constraints.  A compile flag or switch
> just cannot give the flexibility required.

Agreed. This is an interesting idea, and seems promising, but this is
the wrong form. You'd think this would be a policy. I'll try to work up
an example of this.

And the idea of an ABI-different container with a compile flag and no
linkage difference is not the kind of thing that should be encouraged,
ever. 

-benjamin



More information about the Libstdc++ mailing list