From: | Andrew Gierth <andrew(at)tao11(dot)riddles(dot)org(dot)uk> |
---|---|
To: | Jeff Anton <antonpgsql(at)hesiod(dot)org> |
Cc: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Zero-padding and zero-masking fixes for to_char(float) |
Date: | 2015-03-24 16:47:32 |
Message-ID: | 87r3se7403.fsf@news-spur.riddles.org.uk |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
>>>>> "Jeff" == Jeff Anton <antonpgsql(at)hesiod(dot)org> writes:
Jeff> Postgresql seems to be using the first interpretation and
Jeff> reporting fewer digits. I've noticed this with pg_dump. That a
Jeff> dump and restore of floating point values does not produce the
Jeff> same floating point values. To me, that is inexcusable. Using
Jeff> the -Fc format, real values are preserved. I have a large
Jeff> database of security prices. I want accuracy above all.
-Fc doesn't do anything different with floats than text dumps do.
pg_dump (in all modes) uses the extra_float_digits setting to try and
get a value that restores exactly. There have been issues with this in
the past (the limit of extra_float_digits had to be raised from 2 to 3,
iirc), but if it's happening now then that should probably be considered
a bug.
--
Andrew (irc:RhodiumToad)
From | Date | Subject | |
---|---|---|---|
Next Message | Andres Freund | 2015-03-24 18:25:25 | Re: INT64_MIN and _MAX |
Previous Message | Jeff Anton | 2015-03-24 16:26:10 | Re: Zero-padding and zero-masking fixes for to_char(float) |