On Tuesday 15 Feb 2005 8:03 pm, Antonios Christofides wrote:
> I'm also considering a table, of course, where each timeseries record
> will be one row. I have experimented only a little with that (id
> integer, date timestamp, value double precision, flags text, primary
> key(id, date)). It appears to be fast alright, a problem being that it
> appears to consume much disk space (about 4 times more than an array
> of text, which is about 4 times more than a single huge text).
Perhaps you could attempt to store a fix small number of records per row, say
4-6? Or may be a smaller fixed size array, That should make the row overhead
less intrusive...
As far as replacing the time series is concerned, you can just drop the table
which should be pretty fast..
Just a thought..
Shridhar