Adding demo showing how to use CDI + JPA + JTA (@Transactional, trans…

[gunnarmorling]

…actional observer methods) in unittests

Hey @gsmet, could you take a look at this one? The other day I wanted to run CDI and Hibernate ORM in unit tests and it was quite some effort. So I thought I'd share my findings in a blog post, for which this will be the accompanying code.

[gunnarmorling]

@manovotn, @mkouba, I'm using Weld JUnit in this example; could you take a look and let me know whether I'm doing something stupid?

Also is there no simpler way than this (specifically, hooking up Narayana and a JNDI server) for using @Transactional and transactional observer methods in JUnit tests that don't run in WildFly proper (via Arquillian)?

I'd like to blog about this, hoping it's useful to others, too. But at the same time I'm wondering whether these bits should be added to Weld JUnit eventually, making the experience much simpler out of the box. WDYT?

[gunnarmorling]

Also filed weld/weld-testing#65 to see whether this all could be baked right into Weld JUnit, giving a smoother experience out-of-the-box.

[manovotn]

@manovotn, @mkouba, I'm using Weld JUnit in this example; could you take a look and let me know whether I'm doing something stupid?

Scanned the code, weld-junit bits are looking good!

I'd like to blog about this, hoping it's useful to others, too. But at the same time I'm wondering whether these bits should be added to Weld JUnit eventually, making the experience much simpler out of the box. WDYT?

Sounds like a good idea to me, let's discuss that in the weld-junit issue you created.

[manovotn]

In order to avoid type cast here, you can simply use Weld API instead of CDI itself.
Just doing WeldInitiator.from(new Weld())... should do the trick.

Also note that by doing this (e.g. by using pure new Weld()), you are running with enabled discovery (your CP wll be scanned for beans) whereas commonly a synthetic archive is used to reduce bootstrap time. Not sure you intended that, so just pointing it out.

[gunnarmorling]

Ha, didn't know it'd be as simple as calling new Weld() :-)

you are running with enabled discovery...

Yes, I did this intentionally. Having to specify all the classes explicitly is one of the things I find a bit cumbersome when using Arquillian. Bootstrap time was perfectly fine when I ran the test. But it's definitely good to keep this in mind. Thanks!

[gsmet]

Looks like an interesting addition to me.

[gunnarmorling]

Cool, thanks for reviewing. I'll merge it then and write something for in.relation.to within the next few days.

[gunnarmorling]

Merged. Thanks, everone.