P: n/a

(asked last week on .questions, no response)
Can anyone explain why this happens? (under 7.4.1)
select '20040527 09:00:00.50000104' :: timestamp(0) ;
timestamp

20040527 09:00:01
select '20040527 09:00:00.50000004' :: timestamp(0) ;
timestamp

20040527 09:00:00
That is, why doesn't the second operation result in the same timestamp
as the first? Is it a floatingpoint representation issue, or are the
mathematical rules of rounding not being followed correctly (as I
understand them, anyway)?

Jeff Boes vox 269.226.9550 ext 24
Database Engineer fax 269.349.9076
Nexcerpt, Inc. http://www.nexcerpt.com
...Nexcerpt... Extend your Expertise  
Share this Question
P: n/a

On Thu, 3 Jun 2004, Jeff Boes wrote: (asked last week on .questions, no response)
Can anyone explain why this happens? (under 7.4.1)
select '20040527 09:00:00.50000104' :: timestamp(0) ;
timestamp  20040527 09:00:01
select '20040527 09:00:00.50000004' :: timestamp(0) ;
timestamp  20040527 09:00:00
That is, why doesn't the second operation result in the same timestamp as the first? Is it a floatingpoint representation issue, or are the mathematical rules of rounding not being followed correctly (as I understand them, anyway)?
My first guess would be that your system probably implements its default
rounding as nearest even for .5 results, what does 9:00:01.5 give you?
(end of broadcast)
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribenomail command to ma*******@postgresql.org so that your
message can get through to the mailing list cleanly  
P: n/a

Stephan Szabo <ss****@megazone.bigpanda.com> writes: That is, why doesn't the second operation result in the same timestamp as the first? Is it a floatingpoint representation issue, or are the mathematical rules of rounding not being followed correctly (as I understand them, anyway)? My first guess would be that your system probably implements its default rounding as nearest even for .5 results, what does 9:00:01.5 give you?
Fwiw, the floating point timestamp representation is secondsbased. So 0.5s
should be exactly representable. (Though 0.500001 wouldn't, but that shouldn't
matter.)
On my machine it seems to always round away from 0, but this comment from
timestamp.c seems relevant. It would imply my build was build with integer
timestamps and yours was built with floating point timestamps:
/*
* Note: this roundtonearest code is not completely consistent
* about rounding values that are exactly halfway between integral
* values. On most platforms, rint() will implement
* roundtonearesteven, but the integer code always rounds up
* (away from zero). Is it worth trying to be consistent?
*/
And this is from the glibc Info page:
IEEE 754 defines four possible rounding modes:
Round to nearest. This is the default mode. It should be used unless there is a specific need for one of the others. In this mode results are rounded to the nearest representable value. If the result is midway between two representable values, the even representable is chosen. "Even" here means the lowestorder bit is zero. This rounding mode prevents statistical bias and guarantees numeric stability: roundoff errors in a lengthy calculation will remain smaller than half of `FLT_EPSILON'.
[And the other rounding directions are useless;
this is the default and the only one that matters.]

greg
(end of broadcast)
TIP 1: subscribe and unsubscribe commands go to ma*******@postgresql.org  
P: n/a

On Thu, 3 Jun 2004, Jeff Boes wrote: (asked last week on .questions, no response)
Can anyone explain why this happens? (under 7.4.1)
select '20040527 09:00:00.50000104' :: timestamp(0) ;
timestamp  20040527 09:00:01
select '20040527 09:00:00.50000004' :: timestamp(0) ;
timestamp  20040527 09:00:00
That is, why doesn't the second operation result in the same timestamp as the first? Is it a floatingpoint representation issue, or are the mathematical rules of rounding not being followed correctly (as I understand them, anyway)?
My first guess would be that your system probably implements its default
rounding as nearest even for .5 results, what does 9:00:01.5 give you?
(end of broadcast)
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribenomail command to ma*******@postgresql.org so that your
message can get through to the mailing list cleanly  
P: n/a

Stephan Szabo <ss****@megazone.bigpanda.com> writes: That is, why doesn't the second operation result in the same timestamp as the first? Is it a floatingpoint representation issue, or are the mathematical rules of rounding not being followed correctly (as I understand them, anyway)? My first guess would be that your system probably implements its default rounding as nearest even for .5 results, what does 9:00:01.5 give you?
Fwiw, the floating point timestamp representation is secondsbased. So 0.5s
should be exactly representable. (Though 0.500001 wouldn't, but that shouldn't
matter.)
On my machine it seems to always round away from 0, but this comment from
timestamp.c seems relevant. It would imply my build was build with integer
timestamps and yours was built with floating point timestamps:
/*
* Note: this roundtonearest code is not completely consistent
* about rounding values that are exactly halfway between integral
* values. On most platforms, rint() will implement
* roundtonearesteven, but the integer code always rounds up
* (away from zero). Is it worth trying to be consistent?
*/
And this is from the glibc Info page:
IEEE 754 defines four possible rounding modes:
Round to nearest. This is the default mode. It should be used unless there is a specific need for one of the others. In this mode results are rounded to the nearest representable value. If the result is midway between two representable values, the even representable is chosen. "Even" here means the lowestorder bit is zero. This rounding mode prevents statistical bias and guarantees numeric stability: roundoff errors in a lengthy calculation will remain smaller than half of `FLT_EPSILON'.
[And the other rounding directions are useless;
this is the default and the only one that matters.]

greg
(end of broadcast)
TIP 1: subscribe and unsubscribe commands go to ma*******@postgresql.org  
P: n/a

Stephan Szabo wrote: On Thu, 3 Jun 2004, Jeff Boes wrote: (asked last week on .questions, no response)
Can anyone explain why this happens? (under 7.4.1)
select '20040527 09:00:00.50000104' :: timestamp(0) ;
timestamp  20040527 09:00:01
select '20040527 09:00:00.50000004' :: timestamp(0) ;
timestamp  20040527 09:00:00
That is, why doesn't the second operation result in the same timestamp as the first? Is it a floatingpoint representation issue, or are the mathematical rules of rounding not being followed correctly (as I understand them, anyway)?
My first guess would be that your system probably implements its default rounding as nearest even for .5 results, what does 9:00:01.5 give you?
20040527 09:00:02, so I guess that would confirm it.

Jeff Boes vox 269.226.9550 ext 24
Database Engineer fax 269.349.9076
Nexcerpt, Inc. http://www.nexcerpt.com
...Nexcerpt... Extend your Expertise
(end of broadcast)
TIP 6: Have you searched our list archives? http://archives.postgresql.org  
P: n/a

Stephan Szabo wrote: On Thu, 3 Jun 2004, Jeff Boes wrote: (asked last week on .questions, no response)
Can anyone explain why this happens? (under 7.4.1)
select '20040527 09:00:00.50000104' :: timestamp(0) ;
timestamp  20040527 09:00:01
select '20040527 09:00:00.50000004' :: timestamp(0) ;
timestamp  20040527 09:00:00
That is, why doesn't the second operation result in the same timestamp as the first? Is it a floatingpoint representation issue, or are the mathematical rules of rounding not being followed correctly (as I understand them, anyway)?
My first guess would be that your system probably implements its default rounding as nearest even for .5 results, what does 9:00:01.5 give you?
20040527 09:00:02, so I guess that would confirm it.

Jeff Boes vox 269.226.9550 ext 24
Database Engineer fax 269.349.9076
Nexcerpt, Inc. http://www.nexcerpt.com
...Nexcerpt... Extend your Expertise
(end of broadcast)
TIP 6: Have you searched our list archives? http://archives.postgresql.org   This discussion thread is closed Replies have been disabled for this discussion.   Question stats  viewed: 4184
 replies: 6
 date asked: Nov 23 '05
