Re: [boost] Accumulators review

From: Eric Niebler (eric_at_[hidden])
Date: 2007-01-30 13:54:21

John Maddock wrote:
> In the tutorial it says:
> 
> " Even the extractors can accept named parameters. In a bit, we'll see a 
> situation where that is useful."
> 
> But I can't find any further mention of extractors with named parameters, I 
> ask because I believe it may be indeed be useful - for example a student's t 
> test could defer requiring the acceptable risk level until the extractor is 
> actually called - so the same test could be conducted multiple times with 
> different risk levels for example just by calling the extractor multiple 
> times.

Whoops. I had intended to document how to use an external accumulator 
for the weight statistics, which can then be passed in as a parameter at 
extraction time. (Essentially, dependencies within an accumulator set 
can be resolved by promising to pass in the required accumulator at 
extraction time.) But apparently I never got around to it. I'll fix that.

> One other thing I've missed is the ability to "reset" an accumulator so it 
> can be used more than once.  Is this possible?  Potentially it might save 
> quite a bit of construction cost in some cases?

Would you like to be able to reset an accumulator, or the whole set? 
Nothing stops you from writing an accumulator with a reset() member 
function. You can use accumulator_set::extract() to fetch the 
accumulator corresponding to a certain feature and then call reset() on it.

It might be nice to have an accumulator_set::reset() member function 
that resets all the accumulators in order. But what should happen to 
accumulators that don't define a reset() function? Destroy and 
reconstruct them in-place? That could be trouble. Skipping them would be 
trouble, too.

-- 
Eric Niebler
Boost Consulting
www.boost-consulting.com