[Raiford@labware.com: Re: float8 column size defined as 15 instead of 53?]

----- Forwarded message from Raiford@labware.com -----

Date: Tue, 19 Jan 2016 17:12:33 -0700
From: Raiford@labware.com
To: pgsql-odbc-owner@postgresql.org
Subject: Re: float8 column size defined as 15 instead of 53?
Message-ID: <OFB569ED94.0C1AE84E-ON85257F40.00006986-87257F40.0000B402@labware.com>

So I am finding conflicting information about this value.  The
documentation for Column Size (
https://msdn.microsoft.com/en-us/library/ms711786%28v=vs.85%29.aspx) shows
the value being the decimal precision.  I know from experience that I can
get binary precision from SQL Server and from Oracle, so maybe there is
something else going on that I don't understand.

I did look at the code for the pgsql-odbc driver and in
pgtype_attr_column_size() the driver is answering 15 for float8 in the
older driver that I am using.  In the current driver it is answering 17.
Now I'm even more confused!  Surely there can only be 15 digits of decimal
precision in a IEEE double floating point?  Is Postgres somehow able to
include 17 digits precision if requesting the value as text (9 for single
precision floats)?  In the current driver these values are from
PG_REAL_DIGITS and PG_DOUBLE_DIGITS.  In the older driver they were simply
hard coded.

Can anyone shed some light on this stuff?  Clearly there is something I'm
missing.

Jon



From:   Jon Raiford/Employee/LW-US
To:     pgsql-odbc-owner@postgresql.org
Date:   01/19/2016 12:27 PM
Subject:        float8 column size defined as 15 instead of 53?


I'm seeing that the columns defined as float8 in Postgres are being
described as SQL_FLOAT(15) instead of SQL_FLOAT(53).  From what I can
tell, according to the ODBC spec, the column size should be reported as
binary precision, but the Postgres ODBC driver is reporting the decimal
precision.

https://msdn.microsoft.com/en-us/library/ms710150%28v=vs.85%29.aspx

"SQL_FLOAT  FLOAT(p)  Signed, approximate, numeric value with a binary
precision of at least p. (The maximum precision is driver-defined.)[5]"

Unfortunately I'm not running on a current ODBC driver and it would be
very inconvenient for me to load it at this moment.  Could someone with
the latest driver verify whether or not SQLDescribeCol(),
SQLGetTypeInfo(), or SQLColAttribute() still report this with the decimal
precision?

I am happy to add a kludge to my code to treat numbers <= 15 as decimal
precision, but at some point it would be nice to see this be fixed if it
hasn't already.

I'm currently running the PostgreSQL Unicode 9.01.01.00 driver from
30-Dec-2011.  I know it is quite old, so I'd be happy to hear that this is
already resolved.  I'd also be happy to hear that I've misunderstood the
spec and would love clarification.  For what its worth, I do see that both
SQL Server and Oracle are reporting binary precision.

Thank you,
Jon

www.labware.com
Results Count






www.labware.com
Results Count




----- End forwarded message -----