db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrew McRae (JIRA)" <j...@apache.org>
Subject [jira] Commented: (DERBY-4251) Make Derby EmbedResultSet support getRow() for FORWARD_ONLY result sets
Date Sat, 04 Jul 2009 12:55:47 GMT

    [ https://issues.apache.org/jira/browse/DERBY-4251?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12727232#action_12727232
] 

Andrew McRae commented on DERBY-4251:
-------------------------------------

The most relevant fact is that the vast majority of Java database code already running in
the real world was written according to JDBC spec 3.0 or earlier versions. What I would like
you to think about and ask yourself is "What is the most relevant JDBC specification for Derby?"

Consider that Derby still (thankfully) runs on Java 1.4 - thus allowing it to work in the
widest possible range of Java runtimes already out there. The current JDBC api version at
the time JRE 1.4 was released (2003) until Java 6 came out (2008) was JDBC 3.0 as defined
in the  JDBC 3.0 spec (December 2001) and the Java 1.4 API Javadoc.

Here's what the spec says about compliance:
--8<------8<----
JDBC 3.0 API Compliance
A driver that is compliant with the JDBC 3.0 API must do the following:
*  Comply with the JDBC 2.0 API requirements
*  Include the following required interfaces:
     java.sql.ParameterMetaData
     java.sql.Savepoint
...
--8<------8<----

Here's the Java 1.4 API:
http://java.sun.com/j2se/1.4.2/docs/api/java/sql/ResultSet.html#getRow()
--8<------8<----
getRow

public int getRow()
           throws SQLException

    Retrieves the current row number. The first row is number 1, the second number 2, and
so on.

    Returns:
        the current row number; 0 if there is no current row 
    Throws:
        SQLException - if a database access error occurs
    Since:
        1.2
--8<------8<----

This is the behaviour required by over 5 years of software developed throughout the IT industry.
There is no point in claiming to support later versions of the JDBC specification if Derby
does not support the JDBC 2.0 API.

Here is what the JDBC 4.0 spec says about the goals of the 4.0 specification:
--8<------8<----
8. Maintain backward compatibility with existing applications and drivers
     Existing JDBC technology-enabled drivers ( JDBC drivers) and the applications
     that use them must continue to work in an implementation of the Java virtual
     machine that supports the JDBC 4.0 API. Applications that use only features
     defined in earlier releases of the JDBC API will not require changes to continue
     running. It should be straightforward for existing applications to migrate to
     JDBC 4.0 technology.
--8<------8<----

The latest JDBC documentation has added a new comment saying getRow() is optional for the
special case of forward-only result sets, but clearly by the above reasoning this is self-contradictory
to the goals of JDBC 4.0 as stated by Sun, and so should be ignored. The getRow() method has
never been optional and by the chain of logic going back to JDBC 2 it cannot become optional.

Let me explain this in a different way in case it has not become clear.

Specification compliance is not a matter of being cool, it's a matter of being practically
useful due to fully functional interoperability. The more recent the spec the software supports
then the more features it will support and the cooler it will be, but not at the expense of
breaking compatibility with the older specifications because that would make it useless to
most existing software. (That would also be very uncool.)

This issue is not a request for a feature enhancement as you have now incorrectly recategorised
it.

This is a bug, it is a nonconformance to the JDBC 2.0 API, and Derby must comply with that
API for three reasons:
1) to allow all existing Java software to use Derby (assuming you do actually want people
to use Derby), and
2) to merit compliance with BOTH the JDBC 3.0 and JDBC 4.0 specifications, and
3) to be cool.

This makes sense to me. What does everyone else think?

> Make Derby EmbedResultSet support  getRow() for FORWARD_ONLY result sets
> ------------------------------------------------------------------------
>
>                 Key: DERBY-4251
>                 URL: https://issues.apache.org/jira/browse/DERBY-4251
>             Project: Derby
>          Issue Type: Improvement
>          Components: JDBC
>    Affects Versions: 10.4.2.0, 10.5.1.1
>         Environment: Windows XP, and Java 6 or Java 1.4.2
>            Reporter: Andrew McRae
>   Original Estimate: 2h
>  Remaining Estimate: 2h
>
> According to source of the Derby version I first found it in, and the version just released
today, here is the offending code:
> https://svn.apache.org/repos/asf/db/derby/code/tags/10.5.1.1/java/engine/org/apache/derby/impl/jdbc/EmbedResultSet.java
> {quote}
> 	public int getRow() throws SQLException {
> 		// getRow() is only allowed on scroll cursors
> 		checkScrollCursor("getRow()");
> 		/*
> 		 * * We probably needn't bother getting the text of * the underlying
> 		 * statement but it is better to be * consistent and we aren't
> 		 * particularly worried * about performance of getRow().
> 		 */
> 		return theResults.getRowNumber();
> 	}
> {quote}
> The comment and the check for scrollability is incorrect. The Javadoc comment for this
getRow is a duplicate of the Sun JDK API for ResultSet.getRow(), and if the author had just
written it according to their own javadoc there would be no problem. 
> See spec here:
>  http://java.sun.com/j2se/1.3/docs/api/java/sql/ResultSet.html#getRow()
> The Sun Javadoc does not spell out any exceptions to the required behaviour. getRow should
always work regardless of whether the result set is scrollable because the issue of scrollability
is whether the client can *reposition* the cursor to change the current row of the result
set. Simply *reading* the current row number is different and unrelated to changing the current
row number. 
> Therefore getRow() should work for FORWARD_ONLY ResultSets, but Derby currently fails
on this.
> Quite aside from the spec nonconformance, it's a bit annoying that my application has
worked fine on Oracle, MS SQL, the JDBC-LDAP bridge driver, and even MS Access MDB files via
the Sun JDBC-ODBC bridge driver, but then the Derby engine falls over just trying to run a
simple Select statement on a non-updatable forward-only result set - the simplest kind of
all!
> I haven't looked at what " theResults.getRowNumber() " does, but I hope the fix for this
bug is as simple as removing the call to checkScrollCursor; 

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message