Postgres 9.2.4 "Double Precision" Precision

On 09/13/2013 11:32 AM, NWRFC Portland wrote:

I recently upgraded from postgres 8.2.6 to 9.2.4 . For the most part I
am enjoying the upgrade. I have found one behavior that I can not
explain.....

Below is sample contents of a table. "VALUE" in column 7 is defined as
double precision (table definition is the same in 8.2.6 as 9.2.4).
The change over from 8.2.6 to 9.2.4 happened after 2013-09-10 16:00:00.

In postgres 8.2.6 a value inserted , 6.31, is represented in the
database as 6.31 (7th column,7th row)
In postgres 9.2.4 a value inserted , 6.32, is represented in the
database as 6.32000017166138 (7th column, 6th row)

VALUE
SLMO3 | HG | 0 | RG | Z | 2013-09-10 18:15:00 |
6.32000017166138 | Z | 1879048191 | 0 | MSGPRODID
| 2013-09-10 20:02:00 | 2013-09-10 20:03:15
SLMO3 | HG | 0 | RP | Z | 2013-09-10 18:15:00 |
6.32999992370605 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 19:15:00 | 2013-09-10 19:15:34
SLMO3 | HG | 0 | RP | Z | 2013-09-10 18:00:00 |
6.32999992370605 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 17:45:00 |
6.32999992370605 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 17:30:00 |
6.32000017166138 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 17:15:00 |
6.32000017166138 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 16:00:00 |
6.31 | Z | 1879048191 | 0 | KPQRRR6PQR |
2013-09-10 16:18:23 | 2013-09-10 16:18:23
SLMO3 | HG | 0 | RG | Z | 2013-09-10 16:00:00 |
6.28 | Z | 1879048191 | 0 | KWOHRRSPTR |
2013-09-10 16:16:32 | 2013-09-10 16:17:33
SLMO3 | HG | 0 | RG | Z | 2013-09-10 15:45:00 |
6.28 | Z | 1879048191 | 0 | KWOHRRSPTR |
2013-09-10 16:16:32 | 2013-09-10 16:17:33
SLMO3 | HG | 0 | RP | Z | 2013-09-10 15:45:00 |
6.31 | Z | 1879048191 | 0 | KPQRRR6PQR |
2013-09-10 16:18:23 | 2013-09-10 16:18:23
SLMO3 | HG | 0 | RP | Z | 2013-09-10 15:30:00 |
6.31 | Z | 1879048191 | 0 | KPQRRR6PQR |
2013-09-10 16:18:23 | 2013-09-10 16:18:23
SLMO3 | HG | 0 | RG | Z | 2013-09-10 15:30:00 |
6.27 | Z | 1879048191 | 0 | KWOHRRSPTR |
2013-09-10 16:16:32 | 2013-09-10 16:17:33

As the values effect arithmetic calculations, I wish the database to
represent values as they are given. For instance, 6.345 would be
represented as 6.345..... 6.3 represented as 6.3.
Does anyone have any suggestions?

Are the two databases on the same machine?

If you want defined precision you will need to use numeric, see docs
below for more information.

http://www.postgresql.org/docs/9.2/interactive/datatype-numeric.html

8.1.3. Floating-Point Types

The data types real and double precision are inexact, variable-precision
numeric types. In practice, these types are usually implementations of
IEEE Standard 754 for Binary Floating-Point Arithmetic (single and
double precision, respectively), to the extent that the underlying
processor, operating system, and compiler support it.

Inexact means that some values cannot be converted exactly to the
internal format and are stored as approximations, so that storing and
retrieving a value might show slight discrepancies. Managing these
errors and how they propagate through calculations is the subject of an
entire branch of mathematics and computer science and will not be
discussed here, except for the following points:

If you require exact storage and calculations (such as for monetary
amounts), use the numeric type instead.

If you want to do complicated calculations with these types for anything
important, especially if you rely on certain behavior in boundary cases
(infinity, underflow), you should evaluate the implementation carefully.

Comparing two floating-point values for equality might not always work
as expected.

On most platforms, the real type has a range of at least 1E-37 to 1E+37
with a precision of at least 6 decimal digits. The double precision type
typically has a range of around 1E-307 to 1E+308 with a precision of at
least 15 digits. Values that are too large or too small will cause an
error. Rounding might take place if the precision of an input number is
too high. Numbers too close to zero that are not representable as
distinct from zero will cause an underflow error.

Thanks in advance,
Joanne

--
Adrian Klaver
adrian.klaver@gmail.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Arian,

It is a single database . Postgres was upgraded from 8.2.6 to 9.2.4... the
database contents was not altered in upgrade, that is a 8.2.6 dump was not
created then uploaded to 9.2.4.

Perhaps handling of double precision, changed from 8.2.6 to 9.2.4 ?

Joanne

On Fri, Sep 13, 2013 at 11:40 AM, Adrian Klaver <adrian.klaver@gmail.com>wrote:

On 09/13/2013 11:32 AM, NWRFC Portland wrote:

I recently upgraded from postgres 8.2.6 to 9.2.4 . For the most part I
am enjoying the upgrade. I have found one behavior that I can not
explain.....

Below is sample contents of a table. "VALUE" in column 7 is defined as
double precision (table definition is the same in 8.2.6 as 9.2.4).
The change over from 8.2.6 to 9.2.4 happened after 2013-09-10 16:00:00.

In postgres 8.2.6 a value inserted , 6.31, is represented in the
database as 6.31 (7th column,7th row)
In postgres 9.2.4 a value inserted , 6.32, is represented in the
database as 6.32000017166138 (7th column, 6th row)

VALUE
SLMO3 | HG | 0 | RG | Z | 2013-09-10 18:15:00 |
6.32000017166138 | Z | 1879048191 | 0 | MSGPRODID
| 2013-09-10 20:02:00 | 2013-09-10 20:03:15
SLMO3 | HG | 0 | RP | Z | 2013-09-10 18:15:00 |
6.32999992370605 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 19:15:00 | 2013-09-10 19:15:34
SLMO3 | HG | 0 | RP | Z | 2013-09-10 18:00:00 |
6.32999992370605 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 17:45:00 |
6.32999992370605 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 17:30:00 |
6.32000017166138 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 17:15:00 |
6.32000017166138 | Z | 1879048191 | 0 | KPQRRR6PQR
| 2013-09-10 18:18:00 | 2013-09-10 18:18:12
SLMO3 | HG | 0 | RP | Z | 2013-09-10 16:00:00 |
6.31 | Z | 1879048191 | 0 | KPQRRR6PQR |
2013-09-10 16:18:23 | 2013-09-10 16:18:23
SLMO3 | HG | 0 | RG | Z | 2013-09-10 16:00:00 |
6.28 | Z | 1879048191 | 0 | KWOHRRSPTR |
2013-09-10 16:16:32 | 2013-09-10 16:17:33
SLMO3 | HG | 0 | RG | Z | 2013-09-10 15:45:00 |
6.28 | Z | 1879048191 | 0 | KWOHRRSPTR |
2013-09-10 16:16:32 | 2013-09-10 16:17:33
SLMO3 | HG | 0 | RP | Z | 2013-09-10 15:45:00 |
6.31 | Z | 1879048191 | 0 | KPQRRR6PQR |
2013-09-10 16:18:23 | 2013-09-10 16:18:23
SLMO3 | HG | 0 | RP | Z | 2013-09-10 15:30:00 |
6.31 | Z | 1879048191 | 0 | KPQRRR6PQR |
2013-09-10 16:18:23 | 2013-09-10 16:18:23
SLMO3 | HG | 0 | RG | Z | 2013-09-10 15:30:00 |
6.27 | Z | 1879048191 | 0 | KWOHRRSPTR |
2013-09-10 16:16:32 | 2013-09-10 16:17:33

As the values effect arithmetic calculations, I wish the database to
represent values as they are given. For instance, 6.345 would be
represented as 6.345..... 6.3 represented as 6.3.
Does anyone have any suggestions?

Are the two databases on the same machine?

If you want defined precision you will need to use numeric, see docs below
for more information.

http://www.postgresql.org/**docs/9.2/interactive/datatype-**numeric.html&lt;http://www.postgresql.org/docs/9.2/interactive/datatype-numeric.html&gt;

8.1.3. Floating-Point Types

The data types real and double precision are inexact, variable-precision
numeric types. In practice, these types are usually implementations of IEEE
Standard 754 for Binary Floating-Point Arithmetic (single and double
precision, respectively), to the extent that the underlying processor,
operating system, and compiler support it.

Inexact means that some values cannot be converted exactly to the internal
format and are stored as approximations, so that storing and retrieving a
value might show slight discrepancies. Managing these errors and how they
propagate through calculations is the subject of an entire branch of
mathematics and computer science and will not be discussed here, except for
the following points:

If you require exact storage and calculations (such as for monetary
amounts), use the numeric type instead.

If you want to do complicated calculations with these types for anything
important, especially if you rely on certain behavior in boundary cases
(infinity, underflow), you should evaluate the implementation carefully.

Comparing two floating-point values for equality might not always work as
expected.

On most platforms, the real type has a range of at least 1E-37 to 1E+37
with a precision of at least 6 decimal digits. The double precision type
typically has a range of around 1E-307 to 1E+308 with a precision of at
least 15 digits. Values that are too large or too small will cause an
error. Rounding might take place if the precision of an input number is too
high. Numbers too close to zero that are not representable as distinct from
zero will cause an underflow error.

Thanks in advance,
Joanne

--
Adrian Klaver
adrian.klaver@gmail.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/**mailpref/pgsql-general&lt;http://www.postgresql.org/mailpref/pgsql-general&gt;

--

Joanne R. Salerno
Sr Hydrologist

503.326.7291
NOAA/Northwest River Forecast Center
Portland, OR

On 09/13/2013 12:36 PM, Joanne Salerno - NOAA Federal wrote:

Arian,

It is a single database . Postgres was upgraded from 8.2.6 to 9.2.4...
the database contents was not altered in upgrade, that is a 8.2.6 dump
was not created then uploaded to 9.2.4.

So you used pg_upgrade to move the data?

If not you can't move from one major version to another without a
dump/restore.

Perhaps handling of double precision, changed from 8.2.6 to 9.2.4 ?

Joanne

--
Adrian Klaver
adrian.klaver@gmail.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

On 09/13/2013 12:36 PM, Joanne Salerno - NOAA Federal wrote:

Arian,

It is a single database . Postgres was upgraded from 8.2.6 to 9.2.4...
the database contents was not altered in upgrade, that is a 8.2.6 dump
was not created then uploaded to 9.2.4.

Perhaps handling of double precision, changed from 8.2.6 to 9.2.4 ?

To follow up, I don't think that is the case. By way of example, granted
on a 9.3 instance:

create table float_test (id int, f_fld double precision);
insert into float_test values (1, 6.31);
insert into float_test values (2, 6.32);
select * from float_test ;

id | f_fld
----+-------
1 | 6.31
2 | 6.32
(2 rows)

A couple of questions:

How is the data entered into the database?

Just wondering if there was any client side changes along with the
database change?

Joanne

--
Adrian Klaver
adrian.klaver@gmail.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Adrian Klaver <adrian.klaver@gmail.com> wrote:

create table float_test (id int, f_fld double precision);
insert into  float_test values (1, 6.31);
insert into  float_test values (2, 6.32);
select * from float_test ;

   id | f_fld
----+-------
   1 |  6.31
   2 |  6.32
(2 rows)

If, instead of those inserts I use these:

insert into  float_test values (1, '6.31');
insert into  float_test values (1, '6.32');
insert into  float_test values (1, '6.32'::double precision);
insert into  float_test values (1, '6.32'::real);

I get:

 id |      f_fld      
----+------------------
  1 |             6.31
  1 |             6.32
  1 |             6.32
  1 | 6.32000017166138
(4 rows)

Apparently the value is being treated as a real value somewhere.

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

On 09/14/2013 08:51 AM, Kevin Grittner wrote:

Adrian Klaver <adrian.klaver@gmail.com> wrote:

create table float_test (id int, f_fld double precision);
insert into float_test values (1, 6.31);
insert into float_test values (2, 6.32);
select * from float_test ;

id | f_fld
----+-------
1 | 6.31
2 | 6.32
(2 rows)

If, instead of those inserts I use these:

insert into float_test values (1, '6.31');
insert into float_test values (1, '6.32');
insert into float_test values (1, '6.32'::double precision);
insert into float_test values (1, '6.32'::real);

I get:

id | f_fld
----+------------------
1 | 6.31
1 | 6.32
1 | 6.32
1 | 6.32000017166138
(4 rows)

Apparently the value is being treated as a real value somewhere.

Interesting, more grist for the mill.

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

--
Adrian Klaver
adrian.klaver@gmail.com

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Adrian, Kevin,

Thank you for the clues. It turns out a java process was added (between
the data source and database) at same time of postgres upgrade. It was the
java process that incorrectly handled the double precision data.

Joanne

On Sat, Sep 14, 2013 at 9:57 AM, Adrian Klaver <adrian.klaver@gmail.com>wrote: