Re: [pystatsmodels] Pandas as RRD (Round Robin Database) Replacment

[Wes McKinney]

On Mon, Aug 6, 2012 at 11:07 PM, Casey <casey....@gmail.com> wrote:
> I'm interesting in converting a python application using RRD
> [http://oss.oetiker.ch/rrdtool/] over to pandas. RRD provides a rolling
> time-series database where time periods are increased as you go further back
> in time. For example, storing a 60s sample rate for 24 hours, and then
> converting the samples to a 1h sample rate for 30 days. The other
> requirement would be to perform range-based queries on this data, and
> ideally returning the highest resolution of data available to the query. Can
> anyone comment how suited pandas is for these requirements.
>
> Casey

hi Casey,

Well, I don't know exactly the API requirements that you need, but
there's no reason you can't store mixed-frequency time series
together. If you have a time series that is all of a particular
frequency, that frequency can be easily inferred
(ts.index.inferred_freq).

Anyway, the time series representation for pandas is very lightweight
and efficient, so no reason you shouldn't give it a shot. I'll be
interested to see how it goes.

- Wes