Re: time + date_part oddness? - Mailing list pgsql-hackers

From Thomas Lockhart
Subject Re: time + date_part oddness?
Date
Msg-id 3A556E80.F18AE49B@fourpalms.org
Whole thread Raw
In response to time + date_part oddness?  (Tatsuo Ishii <t-ishii@sra.co.jp>)
Responses Re: Re: time + date_part oddness?  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
> Any thought?

Hmm. It is not consistantly implemented :(

timestamp_part() uses just the fractional part of the seconds to
calculate milliseconds. But interval_part() uses both fractional and
integer parts to figure the result, and that is the routine which ends
up being used for the time value.

I'd assume that it should be using fractional seconds only, just like
timestamp_part() does. Any reason not to change it for 7.1?

btw, what should 'microseconds' return? It suffers from the problems
mentioned already, plus leaves the "milliseconds" part in the result.
That should probably only return the pieces which are less than a
millisecond...
                     - Thomas


pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Well, we seem to be proof against cache-inval problems now
Next
From: Thomas Lockhart
Date:
Subject: Re: Missing ColLabel tokens