Re: Problems with semicolon trying to create a trigger function via jdbc

Hmm. I'm curious as to what your web app is doing differently. If
you're getting the connection directly from the DriverManager, I don't
believe it would be proxied (although perhaps I'm wrong here--maybe
some other driver is registering as a proxy for the jdbc:postgresql
subprotocol?). Can you try to find out the actual type of the
Connection you get? I think if it is a PGConnection, this implies an
issue with the driver, and if that is the case, it'd be nice to figure
it out and fix it. Are there differences in your JUnit environment and
webapp environment? Jdbc driver versions? JRE versions? Anything else
that's different?
---
Maciek Sakrejda | Software Engineer | Truviso

1065 E. Hillsdale Blvd., Suite 230
Foster City, CA 94404
(650) 242-3500 Main
(650) 242-3501 F
msakrejda(at)truviso(dot)com
www.truviso.com

On Thu, Mar 25, 2010 at 2:43 PM, Collin Peters
<cpeters(at)intouchtechnology(dot)com> wrote:
> Thanks for the replies Tom and Maciek,
>
> I wasn't able to get the DriverManager.getConnection() approach to
> work (still worked in the unit test, but not the running webapp), but
> reverting to the old 'non-dollar' quoting approach worked.
>
> Thanks
>
> Collin
>
>
> On Thu, Mar 25, 2010 at 2:02 PM, Maciek Sakrejda <msakrejda(at)truviso(dot)com> wrote:
>> The parseQuery() method in QueryExecutorImpl breaks up a query if
>> you're executing more than one statement in a single JDBC query.
>> However, it seems to take quotes (and comments) into account. I tried
>> your simple test case (replacing DriverManager.getConnection() for
>> ds.getConnection()), and it works fine. I have a feeling that the
>> DataSource is wrapping the Connection (and Statement) in proxies that
>> also try to break up the individual queries, but that do not take
>> dollar-quotes into account.
>> ---
>> Maciek Sakrejda | Software Engineer | Truviso
>>
>> 1065 E. Hillsdale Blvd., Suite 230
>> Foster City, CA 94404
>> (650) 242-3500 Main
>> (650) 242-3501 F
>> msakrejda(at)truviso(dot)com
>> www.truviso.com
>>
>>
>>
>> On Thu, Mar 25, 2010 at 1:19 PM, Collin Peters
>> <cpeters(at)intouchtechnology(dot)com> wrote:
>>> Hi all,
>>>
>>> I have some framework code that needs to dynamically generate a
>>> function.  There seems to be a problem where the SQL gets truncated at
>>> the first semicolon encountered in the function.  I have tried this
>>> with a very simple function and duplicated it.
>>>
>>> The test trigger function is as follows:
>>> CREATE OR REPLACE FUNCTION test()
>>>   RETURNS trigger AS
>>> $BODY$
>>> DECLARE
>>>     foo integer;
>>> BEGIN
>>>     foo = 4;
>>>     RAISE NOTICE 'Foo: %', foo;
>>> END;
>>> $BODY$
>>> LANGUAGE 'plpgsql' VOLATILE
>>> COST 100;
>>> ALTER FUNCTION test() OWNER TO mcrtdbms;
>>>
>>> The simple test code is:
>>> String sql = "CREATE OR REPLACE FUNCTION
>>> history.history_insert_trigger()     RETURNS trigger AS   $BODY$
>>> DECLARE   foo integer;   BEGIN    foo = 4; RAISE NOTICE 'Foo: %', foo;
>>>      END;   $BODY$     LANGUAGE 'plpgsql' VOLATILE; ";
>>>         DataSource ds = getDataSource();
>>>         try
>>>         {
>>>             Connection conn = ds.getConnection();
>>>             conn.setAutoCommit(true);
>>>             Statement st = conn.createStatement();
>>>             st.executeUpdate(sql);
>>>             st.close();
>>>             conn.close();
>>>         }
>>>         catch (SQLException e)
>>>         {
>>>             e.printStackTrace();
>>>         }
>>>
>>> When I try to run this via my webapp I get the error:
>>> org.postgresql.util.PSQLException: ERROR: unterminated dollar-quoted
>>> string at or near "$BODY$   DECLARE   foo integer"
>>>     at org.postgresql.core.v3.QueryExecutorImpl.receiveErrorResponse(QueryExecutorImpl.java:1591)
>>>     at org.postgresql.core.v3.QueryExecutorImpl.processResults(QueryExecutorImpl.java:1340)
>>>     at org.postgresql.core.v3.QueryExecutorImpl.execute(QueryExecutorImpl.java:192)
>>> <snip>
>>>
>>> So it seems to be truncating the SQL at the first semicolon it
>>> encounters which, of course, borks the whole thing.  What is even
>>> stranger in my quest to get this working is that the above code
>>> actually WORKS when I run it through a JUnit test!!!  I have made no
>>> progress in trying to figure out what is different between the unit
>>> test and the running webapp.  At first I though it was my ORM so I
>>> tried with the straight JDBC code used above and so eliminated that.
>>> Now I am trying to determine if the Postgres JDBC driver is at fault.
>>>
>>> Here is the rest of the details
>>>  * JDBC version 8.3-605 JDBC 3
>>>  * Postgres 8.3
>>>  * JUnit 4
>>>  * Application is built w/ Spring (but this manual query shouldn't be
>>> affected by taht)
>>>  * When the above query is run through the webapp, it is initially
>>> triggered by a Quartz (scheduling api) trigger (which runs when the
>>> webapp starts)
>>>
>>> Any help at all would be appreciated!!  I am about to give up and
>>> write a function which will do the job of creating the trigger for
>>> me!!
>>>
>>> Regards,
>>> Collin
>>>
>>> --
>>> Sent via pgsql-jdbc mailing list (pgsql-jdbc(at)postgresql(dot)org)
>>> To make changes to your subscription:
>>> http://www.postgresql.org/mailpref/pgsql-jdbc
>>>
>>
>