On Mon, 7 May 2007, Tom Lane wrote:
No, that's a truly awful way to do it. The correct way is to use number times interval multiplication, eg date_issued + term * '1 year'::interval; This reduces to not much more than a floating-point multiply, whereas the other way involves string-forming and string-parsing. Plus you can easily use whatever multiplier you like, eg '7 days' if weeks strike your fancy.
Thank you, Tom. This makes sense to me and I did not pick up on this in my readings.
It might be that converting those columns to interval is the best answer, depending on what other processing needs to be done with them. But if Rich wants to leave them as numbers, the above is the best way to convert them to intervals on-the-fly.
No, we'll use whatever data type makes extracting rows the easiest and most efficient. I don't see 'interval' as a data type in the docs. Is it a single-quoted string? We can do converstions between the UI and storage (in both directions), so the type in the DDL can be whatever's best. Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc. | Accelerator(TM) <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
