OK, after reading some more of the spec (you can get it here ->
http://tinyurl.com/kgz7w), here are some more thougts on this...
First, IMO, in an almost unrelated thought: annotations are a neat
idea, but it seems like we are going through another classic IT
industry over-correction: "XML sucks, let's use annotations for
everything!!"
This JDBC4 spec seems to follow that pattern. If (in addition to the
annotation support) it allowed you to externalize the SQL in a related
XML file (like make the MyQueries class look in it's package for
MyQueries.xml), it would be cleaner. It could also be used to support
multiple SQL dialects if done right.
The DataSet interface does currently extend java.util.Collection<T>,
java.lang.Iterable<T>, and java.util.List<T> - that makes things
simpler.
So if it could externalize the SQL, handle dynamic SQL, allowed for
$runtimeSubstitution$, and caching...that would not leave much more.
Larry
On 8/5/06, Ben Munat <[EMAIL PROTECTED]> wrote:
I'm not really sure why I don't like annotations... just seems to add more
noise for my
overworked brain to sift through, I suppose. I actually love @Override -- has
saved me
some frustration a few times already -- and there are some other annotations
that are
useful and not too instrusive.
However, I'm with Gwyn on this... The last thing I want to put in my java code
is this:
<select id="getEmployeesByCriteria" parameterClass="map"
resultClass="foo.bar.something.domain.EmployeeApprovalBean">
Holy mother of all mapped statements!
</select>
Ok, that's a monster search query with a lot of dynamic stuff, so maybe it's
not a
candidate for an annotation (the result gets put in a bean though, so it could).
But I think Larry's point is very valid. Books and tutorials just love to whip
out
stultified examples for stuff like this... "select * from foo where blah blah".
But life
is almost never that simple. Well, at least not in the world of consulting for
State
government.
I also think that avoiding recompilation is never the reason for externalizing
things from
code (configs, sql, etc.). That's often talked about and it can be nice to be
able to make
a change to a config file and just reload the app (though in practice, the app
is always
deployed as a .war so we always recompile and redeploy the whole thing anyway),
but for me
the most important reason is just keeping the code clean and as narrowly
focused as
possible. That's the IOC ideal in my mind.
b
Clinton Begin wrote:
> I've already started some preliminary design work for iBATIS 3, and it
> does include both annotation and XML support.
>
> I have to get it on the wiki...
>
> But anyway, I agree about the one-liner thing...but that could be easily
> fixed with very simple multiline support in the compiler. In both the
> C# version and the Ruby version you can do multiline string, for
> example, in C#:
>
> @"SELECT
> FIRST_NAME
> LAST_NAME
> BIRTH_DATE
> FROM PERSON
> WHERE LAST_NAME LIKE @lastName
> AND DECEASED = false"
>
> Super nice. So annotations can be great, I actually prefer them to
> XML, as I rarely need to change my SQL without recompiling and running
> my tests anyway. But I can understand others have different situations
> --supporting both is fairly straight forward. Luckily, Sun made
> annotations such just as bad as XML, so it's an easier translation in Java.
>
> The C# Attribute support will be harder to keep consistent with the XML,
> because there are so many improvements one can make using ordinal
> parameters and simple things like supporting multiple attributes of the
> same name clean up the code significantly.
>
> Cheers,
> Clinton
>
>
> On 8/5/06, *Larry Meadors* <[EMAIL PROTECTED]
> <mailto:[EMAIL PROTECTED]>> wrote:
>
> Heh, my bet is that Sun delivers enough to make iBATIS easier to
> write, but still manages to overcomplexificate it enough to make
> people still want to use something else.
>
> It is cool to see things like DataSet<Employee> coming from the API,
> but unless DataSet extends List, I do not want that anywhere but in my
> DAO layer, just like I do not want a ResultSet anywhere but in my DAO
> layer. If it does, great! That is a huge advancement.
>
> While stuff like @Select("SELECT * FROM Employee WHERE status = 'A'")
> looks very attractive, and simple crud accounts for probably 35-50% of
> the statements I use, only about 2% of the SQL I write beyond CRUD
> fits into the 1-liner category.
>
> The rest of the SQL statements are those 80-100 line select statements
> with dynamic elements...and as much as I like the StringBuilder
> class...I still prefer <dynamic> elements to java code. ;-)
>
> So...top question on my list...can/should we use this for iBATIS3? ;-)
>
> Larry
>
> On 8/4/06, Clinton Begin <[EMAIL PROTECTED]
> <mailto:[EMAIL PROTECTED]>> wrote:
> > Hi all,
> >
> > I just picked this up from InfoQ. What do you think?
> >
> >
> > =====================
> > Blogger Patrick Lightbody noted "I could see JDBC 4.0 replacing
> my need for
> > iBatis" after reading another blog post on JDBC 4.0
> functionality. JDBC 4.0
> > includes basic O/R mapper functionality that eases the repetitive
> task of
> > copying ResultSet data into objects. It does not handle
> inheritance or table
> > relations however.
> > =====================
> >
> >
> > http://www.infoq.com/news/Easier-Database-Development-JDBC
> >
> > Patrick's a good guy and a respected developer. I appreciate his
> opinion.
> > Obviously, this is only useful for people running JDK 5 or
> possibly only JDK
> > 6. And of course, Annotation support would be an almost
> brainless effort to
> > implement for iBATIS. Your thoughts?
> >
> > Clinton
> >
>
>
