Tom Brinkman wrote:

...

...

Not 100% sure I understand the question. Is "foo::high" supposed to be the value or the offset in the sparse series?

The value. If my data is already in a vector, it seems non-optimal to populate another vector in the "time_series" containers. I wonder if there is a way to act the original vector.

Ah, I see the confusion. Ultimately, the time series will need to "own" the data, so it can't just refer to your vector, right? There needs to be at least one copy. With the ordered inserter, there is *exactly* one copy. It builds the internal representation of the series (often in a vector of some sort), and then when you say ".commit()", that internal representation is swapped with the one in the time series.

...

...

This certainly has some overhead, and I could consider adding a reserve()

Ok.

This point is still valid. If you know ahead of time how many "runs" are going to be in your series, you can save some reallocations.

I'm going to spend the next few weeks studying this library and your statistics library, as they go together. What areas of these libraries remain unfinished and need some attention? I would like to help. I have a backround in econometrics and I would like to create some real-world examples.

Great! If you really wanted to contribute to Time_series in a big way, you could look into implementing a rolling averages algorithm. -- Eric Niebler Boost Consulting www.boost-consulting.com The Astoria Seminar ==> http://www.astoriaseminar.com