[ psqlodbc-Bugs-1010731 ] Incorrect info for varchar(255), text columns

Поиск
Список
Период
Сортировка
От
Тема [ psqlodbc-Bugs-1010731 ] Incorrect info for varchar(255), text columns
Дата
Msg-id 20091118231651.C58C81072088@pgfoundry.org
обсуждение исходный текст
Список pgsql-odbc
Bugs item #1010731, was opened at 2009-11-18 20:55
You can respond by visiting:
http://pgfoundry.org/tracker/?func=detail&atid=538&aid=1010731&group_id=1000125

Category: None
Group: None
Status: Open
Resolution: None
Priority: 3
Submitted By: Farid Zidan (faridz)
Assigned to: Nobody (None)
Summary: Incorrect info for varchar(255), text columns

Initial Comment:
PostgreSQL ANSI ODBC driver 8.04.01 returns incorrect info for
SQLColAttribute SQL_DESC_TYPE and/or SQL_DESC_OCTET_LENGTH

Example,
create table test_varchar(col1 varchar(20), col2 varchar(255), col3 text);

I get the following information from SQLColAttribute:
column  SQL_DESC_TYPE    SQL_DESC_OCTET_LENGTH
col1    SQL_VARCHAR    20
col2    SQL_LONGVARCHAR    255
col3    SQL_LONGVARCHAR    8190

col1 info is correct.
col2 SQL_DESC_TYPE  should be SQL_VARCHAR not SQL_LONGVARCHAR.
col3 SQL_DESC_OCTET_LENGTH should be something large such as 2147483647 (2**31 -1)

Attached is screenshot of the Firebird vs PostgreSQL ODBC driver results for the test table above (MS SQL Server ODBC
driverreturns the same info as the Firebird ODBC driver) 

----------------------------------------------------------------------

Comment By: Richard Broersma (rabroersma)
Date: 2009-11-18 23:16

Message:
one last comment:

notice that the maximum field size in PostgreSQL is 1GB.

http://www.postgresql.org/about/

Also the number of possible characters is less that this since there is some meta data associated with VARCHAR and TEXT
columns.

http://www.postgresql.org/docs/8.4/interactive/datatype-character.html

I'm not sure how this relates to the correct value for SQL_DESC_OCTET_LENGTH.

Regarding the choice of the default setting for these data types, I can only guess as to why the ODBC maintainers made
thechoices that they did.  I believe that your view point is technical correct from a standards point of view. 

But if the vast majority of PostgreSQL ODBC users today are MS-Access users, I doubt they would have adopted PostgreSQL
fortheir database back-end if it didn't automatic work for their use. So for them it is a nice convenience.  

On the other hand, PostgreSQL is probably far more recognized in the MS community today that it was back is the PG
version6 days.  So they might be will to take the time necessarily to read the manual for appropriate adjustments.
Perhapsthe time is ripe to change the defaults to values actually supported by PostgreSQL. 

----------------------------------------------------------------------

Comment By: Farid Zidan (faridz)
Date: 2009-11-18 22:44

Message:
It may very well be so. A workaround for MS Access (doc also mentions Borland). Note that MS SQL Server own driver does
notdo  
anything of the kind. It reports the datatypes and size as they are at the database for the test table.

I do not actively use MS Access, but for people that do and require this workaround, then it is good to have it. My
issueis that  
everybody is getting the workaround without asking for it or requiring it and in the process the driver is masking the
real 
datatype/size of the database table columns by default!

The PostgreSQL ODBC driver by default should report the datatype information accurately as those at the server, so by
defaultthe  
limit on varchar should be the same as the server limit and the limit on TEXT datatype should be the same as the server
limit.It  
is up to he client application to determine what buffer type/size to use for a very large varchar column (inline or
pointerto  
allocated memory).

That's how all other DBMS vendor ODBC drivers behave by default, MS SQL Server, Oracle, DB2, Informix, Firebird, Sybase
ASE, 
SQLAnywhere, Ingres, just to name few that I have tested.

For my test table I only get the incorrect info when using the PostreqSQL driver out-of-the box. I don't have to alter
theoptions  
of any of mentioned DBMS ODBC drivers to get the expected results.

Best regards

----------------------------------------------------------------------

Comment By: Richard Broersma (rabroersma)
Date: 2009-11-18 22:22

Message:
> I suspect the limit on the long varchar size was done as
> I workaround for something, in that case only people that
> need the workaround should get it and everybody who does
> not need it or want it by default.

It I were to make a guess, I would say that these setting were defaulted so that there would be an exact mapping to the
MS-Accessdata types: text{255} and memo{8190}. 

I would expect that there is special magic to make the data type mappings work nicely with MS SQL Server using ODBC
linkedtables.  I not sure how nicely Firebird ODBC driver would work when mapping data type to MS-Access. 

For example, try linking a table with a TEXT PRIMARY KEY (perfectly valid in PostgreSQL). While MS-Access will not
allowits mapped memo data type to be assigned as a primary key. 

----------------------------------------------------------------------

Comment By: Farid Zidan (faridz)
Date: 2009-11-18 22:08

Message:
Thanks. That is very related. I updated the driver page 1 options Max Varchar to 2000
and Max LongVarChar to 2147483647 and now I get identical results to other ODBC
drivers.

I understand the need to set a limit to SQL_VARCHAR buffers, the default value 254 is
very small a more reasonable value is 2000 or 8190.

However, setting a limit on SQL_LONGVARCHAR by default to 8190 seems to be very odd.
No other driver does that. If I define a column at the Server as TEXT then I want to
get all its data and not the first 8190 characters (by default).

I suspect the limit on the long varchar size was done as I workaround for something,
in that case only people that need the workaround should get it and everybody who does
not need it or want it by default.

This issue is similar to http://pgfoundry.org/tracker/?
func=detail&aid=1010516&group_id=1000125&atid=538
where the driver does LF <-> LF/CR by default.

I believe both are the wrong default behavior by the driver and by default the max on
SQL_LONGVARCHAR column should be the same as the server supported max for the datatype
TEXT (2**31 - 1) so user will not have to change anything to use the driver.

OK, I need to update this issue title to say:
Driver Max LongVarChar value should be the same as Server max for TEXT datatype by
default, and update driver Max Varchar option to something bigger such as 2000 (like
Oracle) or 8000 (like MS SQL Server), otherwise every user who wants to use the driver
to handle TEXT columns bigger than 8190 will have to fiddle with the driver options
which should be not required nor necessary if the driver used the server max size for
the TEXT column like you expect to do

----------------------------------------------------------------------

Comment By: Richard Broersma (rabroersma)
Date: 2009-11-18 21:21

Message:
I have one thought that might be unrelated to your problem:

Notice the ODBC driver setting for Max LongVarChar.


http://cvs.pgfoundry.org/cgi-bin/cvsweb.cgi/~checkout~/psqlodbc/psqlodbc/docs/config.html?rev=1.10&content-type=text/html

I wonder if the ODBC driver is masking the actual size that a TEXT column could be?

----------------------------------------------------------------------

You can respond by visiting:
http://pgfoundry.org/tracker/?func=detail&atid=538&aid=1010731&group_id=1000125

В списке pgsql-odbc по дате отправления:

Предыдущее
От:
Дата:
Сообщение: [ psqlodbc-Bugs-1010731 ] Incorrect info for varchar(255), text columns
Следующее
От:
Дата:
Сообщение: [ psqlodbc-Bugs-1010731 ] Incorrect info for varchar(255), text columns