Re: [jira] Created: (DERBY-447) getBoolean() throws data conversion exception for DECIMAL type in J2ME/CDC/Foundation

[Lance J. Andersen]

The DatabaseMetaData method supportsConvert is used to validate what the scalar escape function CONVERT can be used for based on the driver/database combo. The javadoc for the supportsConvert(arg1, arg2) should clarified a bit more. As far as the name of the SQLTYPE for the Convert scalar function, drivers were doing this in an inconsistant manner.  adding the SQL_ prefix brings it in line with ODBC.  Drivers that do not support the SQL_ prefix (some drivers support with and without the prefix), will be required to support the prefix and to continue to optional support without the prefix.  Is there an additional question here that i might have missed? -lance Dag H. Wanvik wrote: Hi, (Lance, see question to you below) "DJD" == Daniel John Debrunner <[EMAIL PROTECTED]> wrote: DJD> DJD> Dag H. Wanvik wrote: DJD> > DJD> There's some inconsistency here, the ResultSet getter methods are DJD> > DJD> converting from SQL data types to Java data types. This supportsConvert DJD> > DJD> method takes two arguments, both of which describe SQL data types. DJD> > DJD> > Yes, you are right. Something for Lance to clarify :) DJD> > But it would seem the no-args supportsConvert() should return false DJD> > until we add support for CONVERT. DJD> DJD> And just to be clear, this is the JDBC CONVERT function from section C.5 DJD> of JDBC 3.0, correct? Yes. DJD> DJD> I think there is some issue with CONVERT in JDBC 4.0, where the format DJD> of the second argument is changing, from INTEGER to SQL_INTEGER or DJD> something like that? Hmm.. I did a short check.. The DataBaseMetaData.supportsConvert has int (from java.sql.Types aka JDBC types/SQLtypes) for both arguments (both in JDBC 3.0 and early draft 4.0), which makes sense since the API speaks of describing the support/behavior of CONVERT, and this conversion all happens in the SQL domain. [On a side note: actually the Javadoc for supportsConvert speaks of (unqualified) CONVERT, without being explicit on this being the C.5 escape JDBC CONVERT, not the SQL character set CONVERT, but I guess that's kind of obvious..?] Also, in both the JDBC 3.0 and 4.0 spec, the escape function CONVERT has SQLtype as the second argument, "value" being given as the first. The PreparedStatement "setter"/ ResultSet "getter" functions mappings, though, convert from and to Java types, given by the tables in the B appendix. I think the problem is that section 15.2.3.1 in the JDBC 4.0 spec (14.2.3.1 in 3.0) erroneously states that DataBaseMetaData.supportsConvert says ANYTHING about the mapping for getter methods. Interestingly, the corresponding section on conversion for *setter* methods makes no mention of DataBaseMetaData.supportsConvert (section 13.2.2.1 in 4.0 as well as in 3.0). Nor do the IN, OUT, INOUT descriptions. 15.2.3.1 Data Type Conversions The recommended ResultSet getter method for each JDBC type is shown in TABLE B-6 on page B-210. This table also shows all of the possible conversions that a JDBC driver may support. The method DataBaseMetaData.supportsConvert(int fromType, int toType) returns ************************************ true if the driver supports the given conversion. Lance, do you agree this is wrong? If it is correct, please explain what I am missing here..! Notwhithstanding this, the Derby implementation of no-args DataBaseMetaData.supportsConvert() returns true, indicating we support C.5 CONVERT, which we do not AFAICS, and I suggest we make a JIRA for that. Dag