Thank you all for your help. Looks like if I remove ^@ from my select string
then whether I use a prepared statement with bind variable or not (statement
contains control characters in string constants), the statement executes
without exceptions.
Whether I use a bind variable as Barry suggested, I still get the same
exception as long as the string contains ^@. The driver does not reject that
character and instead throws the exception and closes the database
connection.
Given this, I guess, the application will have to filter out ^@ from data
and not let the sql data contain it?
As for ^@ being character code 0, that is correct I think. I was not aware
that character code 0 or nul gets displayed as ^@. In java when I tried '^@'
== 0x0000, I got back true so I assume they are the same.
In the postgres 7.2 documentation, I found the following mentioned in
section 1.1.2.1:
"The character with the code zero cannot be in a string constant.".
I did not find any escape sequence documented for this character. If you are
aware of one, I would like to know.
Thanks again.
Vrinda
-----Original Message-----
From: Barry Lind
To: Tom Lane
Cc: Mushran, Vrinda; 'pgsql-jdbc@postgresql.org'
Sent: 1/17/2003 12:49 PM
Subject: Re: [JDBC] Control characters in sql statements close db connection
Tom Lane wrote:
> Barry Lind <blind@xythos.com> writes:
>+
>>Yes I would say this is expected behavior. If you use the regular
>>Statement object it is your responsibility to properly quote and
escape
>>the data. However if you use a PreparedStatement then the driver
>>handles it for you.
>
>
>>So you would issue the following via a PreparedStatement:
>>SELECT * FROM NEB_IPSNMPDEVICES WHERE NEB_IPSNMPDevices.PHY_ADDRESS =
?
>>and then do a stmt.setString(1,"'^@`^]:u'");
>
>
> But there is no escape for a \0, except in the limited context of
> BYTEA values. It'd probably be better for the driver to reject 'em.
>
> Does JDBC understand about BYTEA fields? If so, it could generate the
> correct escape when filling a parameter for a BYTEA value, and fail
> otherwise.
>
Yes the driver understands bytea fields and deals with them
appropriately. But there are not currently any checks that prevent
passing \0 in other areas. The jdbc API uses different methods for
binary data versus textual data and thus there are obvious places to add
such checks.
--Barry