Hi All...
On 3/8/07, David Blevins <[EMAIL PROTECTED]> wrote:
On Mar 6, 2007, at 2:29 PM, Mohammad Nour El-Din wrote:
> Hi All...
>
> I've just finished adding the capability to create annotated test
> clients,
> some new classes are added, some others are modified, the target
> was to give
> the capability to developers to create annotated test clients using
> the same
> approach of the un-annotated test clients and without any extra
> effort.
> Please take a look at https://issues.apache.org/jira/browse/
> OPENEJB-519 and
> have a nice time creating annotated test clients :). The patch is
> waiting
> for review.
>
Excellent, Mohammad!
Is there anyone(s) who'd want to want to take a look at this? (i'm a
bit overcommitted myself atm)
Or really, Mohammad you're a committer now. Maybe you just want to
give an overview of what you added and how it works?
Thanks David for introducing me :D. Well, I want to start by what was my aim
while I was making these changes. My aim was to be able to program annotated
test clients, as much as possible, in the same way we currently develop unit
tests.
It all starts from the new TestClient from which
Stateless\StatefulTestClient are extended. This new TestClient contains all
the common features required by all test clients. And it contains the two
new methods processFieldInjections and processSetterInjections. which are
used by the new tests clients
AnnotatedSetter\FieldInjectionStateless\StatefulTestClient and
AnnotatedSetter\FieldInjectionStateless\StatefulLocalTestClient classes.
These two classes take the burden of the new test classes, so they don't
have to call neither of processSetter\FieldInjections. This idea I derived
from the first proposal of Dain regarding annotated test clients. Each of
these new clients extend from BasicStateless\StatefulTestClient or
BasicStateless\StatefulLocalTestClient. You can see that these later test
clients have annotated fields for the EJB homes. By this new hierarchy
structure these fields can work for both un\annotated test clients. So, what
to do to develop a new annotated unit test. All you have to do is to make a
new unit test which extends from either of
AnnotatedSetter\FieldInjectionStateless\StatefulTestClient or
AnnotatedSetter\FieldInjectionStateless\StatefulLocalTestClient according to
the unit test you develop. And if you use a member field\method from a super
class, you don't have to add an annotated version of it into the new unit
test. All you have to do is to annotate the field\method located in the
super class. Sure you don't have to do this if you want a completely new
annotated version of these fields\methods. By this way we have the minor
code changes and minor unit tests development paradigm changes. I hope I
could explain this well. Have a nice time with annotated unit tests :).
-David
>
>
> On 2/13/07, Mohammad Nour El-Din <[EMAIL PROTECTED] > wrote:
>
>> Hi All...
>>
>> I started to work on a rough implementation related to how to
>> inject EJB
>> refs into test clients. I was suggesting to do the injection
>> within test
>> suites, but I found that it is a bad idea as the test suites are
>> initialized
>> before the container, so we will not be able to initialize the EJB
>> refs. I
>> am working on another idea in my mind, and will keep you updated
>> with info
>> when I get to a proper solution.
>>
>>
>>
>> On 2/6/07, Mohammad Nour El-Din <[EMAIL PROTECTED] > wrote:
>>
>> >
>> >
>> > On 2/6/07, David Blevins < [EMAIL PROTECTED] > wrote:
>> > >
>> > >
>> > > On Feb 4, 2007, at 2:59 AM, Mohammad Nour El-Din wrote:
>> > >
>> > > > Hi All...
>> > > >
>> > > > On 2/4/07, David Blevins < [EMAIL PROTECTED]> wrote:
>> > > >>
>> > > >> On Feb 3, 2007, at 5:50 PM, Dain Sundstrom wrote:
>> > > >>
>> > > >> > On Feb 3, 2007, at 5:00 PM, Jacek Laskowski wrote:
>> > > >> >
>> > > >> >> On 2/4/07, Dain Sundstrom < [EMAIL PROTECTED]> wrote:
>> > > >> >>> I was thinking it would be nice if client code could
>> call use
>> > > the
>> > > >> >>> OpenEJB injection facilities. Say you have some unit
>> > > tests. You
>> > > >> >>> could in the setUp method, lookup a bunch of stuff in
>> JNDI and
>> > > >> write
>> > > >> >>> it into fields, but it would be much nicer to simple
>> annotate
>> > > >> your
>> > > >> >>> fields and have OpenEJB inject directly. I'm thinking of
>> > > >> something
>> > > >> >>> like this:
>> > > >> >>>
>> > > >> >>> public class SomeTest extends TestCase {
>> > > >> >>> @Resource
>> > > >> >>> private SomeEjb someEjb;
>> > > >> >>>
>> > > >> >>> public void setUp() {
>> > > >> >>> OpenEJB.inject(this);
>> > > >> >>> }
>> > > >> >>> }
>> > > >> >>>
>> > > >> >>> I can think of a bunch of ways we could implement this,
>> but
>> > > >> what do
>> > > >> >>> you think of the idea generally?
>> > > >
>> > > >
>> > > > I think it would be nice to let other *ENTITY* initialize
>> the test
>> > > > case, I
>> > > > think of the test suite itself. For example the the local test
>> > > > suite could
>> > > > be initialized with the properties needed to get the local
>> initial
>> > > > context,
>> > > > and the same for the remote test suite to get the remote
>> initial
>> > > > context and
>> > > > we can override the addTest method of the test suite so it
>> > > > initializes the
>> > > > added tests and then preforms the normal test case addition
>> to the
>> > > > suite, by
>> > > > this way we will have minor code changes, thoughts ???
>> > >
>> > > I wouldn't want to mess with our own tests suite too much as
>> we try
>> > > to keep it strictly ejb-spec compliant.
>> > >
>> > > But definitely for test code in user-land the approach you
>> describe
>> > > could be a pretty slick test framework. I say go for it.
>> > >
>> > > I think the InitialContext approach I mention would be a good
>> > > alternate for those who don't want to have any compile-time
>> deps in
>> > > their test code on OpenEJB classes. It'd also work for any
>> client
>> > > (doesn't have to be a test case).
>> > >
>> > > -David
>> >
>> >
>> > AFAIU you, you mean that we should initialize the test suite
>> with the
>> > initial context, or at least the info it needs to create one, if
>> this what
>> > you mean I totally agree with, if not it would be nice to
>> explain more for a
>> > not talented man like me :).
>> >
>> > >
>> > > >>>
>> > > >> >> Yeah, that would be of help.
>> > > >> >>
>> > > >> >> There's one caveat with it. When I first read the email
>> of yours
>> > > I
>> > > >> >> couldn't see the point because of the name which would
>> perform
>> > > the
>> > > >> >> injection. That's what the method would do, wouldn't it?
>> So why
>> > > it
>> > > >> >> isn't called performInjection or doInjection or alike?
>> Other
>> > > than
>> > > >> >> that
>> > > >> >> the idea's excellent.
>> > > >> >
>> > > >> > I don't really care what it is called. I just found the
>> phrase
>> > > >> > "inject this" funny in english.
>> > > >>
>> > > >> Check out this thread. http://www.nabble.com/%40EJB-
>> injection-for-
>> > > >> TestCases-%28would-be-cool%29-tf2889639s2756.html#a8072860
>> > > >>
>> > > >> One of the ideas we came up with there was people could pass a
>> > > >> reference to their TestCase instance along with the other
>> > > >> InitialContext params.
>> > > >>
>> > > >> -David
>> > > >>
>> > > >>
>> > > >
>> > > >
>> > > > --
>> > > > Thanks
>> > > > - Mohammad Nour
>> > >
>> > >
>> >
>> >
>> > --
>> > Thanks
>> > - Mohammad Nour
>>
>>
>>
>>
>> --
>> Thanks
>> - Mohammad Nour
>
>
>
>
> --
> Thanks
> - Mohammad Nour
--
Thanks
- Mohammad Nour
