Re: [boost] Time Series "review"

From: Lewis Hyatt (lhyatt_at_[hidden])
Date: 2007-08-08 10:08:18

> I wouldn't mind it, either. :-) If you're so motivated, You could polish 
> the rolling average implementation I sent around not long ago.

I could take a look at it in a week or two. There are a lot of other 
things that would be useful, like spline interpolation, FFT, etc. The 
only downside is that most of those algorithms could be implemented in 
two seconds using other open source software as the back end (clapack or 
fftw3 or gsl or whatever), so it seems like a waste to re-do that just 
to get them into boost.

> I don't like the idea of commit() called automatically, and I also don't 
> like the idea of needing to keep a reference count to know when to call it.

Fair enough. In any case, it's a pretty straightforward thin wrapper 
around ordered_inserter that I might just use for myself anyway.

> That's true, the word "invert" has a couple of meanings here. Perhaps 
> "reciprocal()"?

That's probably the best option.

> 
>> * Do you think the library should be accepted as a Boost library?
>>
>> Given that I didn't have time to look into this much more, I'm not sure 
>> how seriously my opinion can be taken, but I do vote yes.
> 
> Thanks for taking the time.
> 
> 

I did think about this library some more since I wrote this, and I'm 
getting excited about using it in my next project.

-Lewis