12 Aug
2011
12 Aug
'11
7:25 a.m.
On 8/12/2011 2:50 PM, Eric Niebler wrote:
On 8/11/2011 10:26 PM, Joel de Guzman wrote:
I reverted to the original implementation but used advance<begin, size-1> instead of prior. pop_back_iterator is an unwieldy hack and while it might be fixable, I am not sure if its CT cost outweigh the CT cost of simply using advance<begin, size-1>.
Yowsa! That instantiates O(N) templates for what should be an O(1) operation. I say leave it as prior until we can think of something better.
Yeah, you are right. Reverted for now. Alas, the tests will fail, but anyway, I think I know a good strategy to make it work. Regards, -- Joel de Guzman http://www.boostpro.com http://boost-spirit.com