On 03 Feb 2006 22:30:46 +0100, Knut Anders Hatlen <[EMAIL PROTECTED]> wrote:
Hi,
I have a question about a test case for DERBY-44 (which appears to
have been fixed by Myrna).
In jdbcapi/parameterMetaDataJdbc30.java, I found this test case:
// variation, and also test out empty string in the escape (jira 44).
System.out.println("variation 1, testing jira 44");
ps = con.prepareStatement("select * from sys.systables where tablename like ? escape ?");
ps.setString (1, "SYS%");
ps.setString (2, "");
paramMetaData = ps.getParameterMetaData();
System.out.println ("parameters count for prepared statement is " + paramMetaData.getParameterCount());
dumpParameterMetaData(paramMetaData);
ps.execute();
I wonder what the purpose of this test is. Is it to test that the
escape keyword is accepted by the SQL parser/compiler, or is it to
test that specifiying an empty escape string shouldn't cause
ps.execute() to fail?
It was not at all clear, but after much digging through old tests and repros ;-)
it appears this test is trying to make sure ERROR 22019 is returned. At one point, passing on the empty string in a piece of code like this would generate an StringIndexOutOfBoundsException instead of ERROR 22019 on the
ps.execute().
The reason why I'm asking is that this case failed when I was testing
a patch for DERBY-822. In DERBY-822, we want the client driver to
prefetch data in the execute call. Since having an empty string as
escape character is not allowed by the SQL spec, this test will fail
if we do prefetching in execute(). However, the test code does not
attempt to fetch the results from the query, so it doesn't fail if we
don't prefetch data.
To me, this test case looks a bit odd (testing that executing invalid
SQL doesn't fail), but I thought I'd ask before changing it, in case
there's a reason for doing it like this.
Thx for checking, I hope my reply makes sense. It was attempting to check it didn't fail in an ugly way. I'm wondering how it's failing for you, you didn't say.
Myrna
