db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Army <qoz...@sbcglobal.net>
Subject Re: [jira] Commented: (DERBY-334) Add initial SQL support for XML datatype/functions in Derby, based on SQL/XML specs.
Date Sat, 04 Jun 2005 00:20:42 GMT
Daniel John Debrunner (JIRA) wrote:
>     [ http://issues.apache.org/jira/browse/DERBY-334?page=comments#action_12312598 ]

> 
> Daniel John Debrunner commented on DERBY-334:
> ---------------------------------------------
> 
> From a quick look at the initial patch it seems the XML data type 
> spills over into the SQL layer, through use of classes in the new
> package  impl.sql.xml.

[ snip ]

> It may be that  you just created a sql.xml package instead of 
> puting the classes in the iapi.types package, but I did see a 
> reference to the DataDictionary  in one of the classes in 
> impl.sql.xml.

At first I had indeed put the XML implementation classes (the ones in sql.xml) 
into the iapi.types package--and that worked just fine (I just imported the 
DataDictionary into the iapi.types package, which didn't show any problems).

When I started the code "re-org" though, I decided to move the classes to 
impl.sql.xml because 1) the package name seemed to make more sense to me 
("sql.xml" => SQL/XML, and "impl" => these are the implementation classes, not 
the interface/type classes); and 2) I don't know how many more XML type 
implementations we're going to have in the future, but it seemed like it would 
be cleaner to have all of them isolated into their own package, instead of 
cluttering the iapi.types package with who-knows-how-many XML impl files.

That said, moving impl.sql.xml back to iapi.types is certainly doable, but I'm 
not sure if that would handle the issue you bring up--by using DataDictionary in 
the XML impl class, we're still crossing the two layers...So maybe that's the 
more fundamental issue...

> I'm not sure if we need the double indirection, XMLDataValue->XML->implementation.

I've gone over that question a couple of times, too.  I've re-organized the XML 
type classes several times since I first began, so maybe this double indirection 
is a result of an old class organization.  I'll try to look at it again to see 
if I can explain why I did it...

 > Once the new patch is posted I will look at the data type closely,

As for the next patch: I'm done making code changes to address Satheesh's 
comments (aside from aesthetics), but I still need to run "derbyall".  That 
means I won't be able to post the new patch until 8 hours after I start the 
test, which means sometime in the middle of the night or (more likely) tomorrow 
morning...(sorry, but posting the patch without running derbyall is a bit too 
risky, since my changes weren't just to the new XML files).

Army


Mime
View raw message